Population Structure of and Conservation Strategies for Wild Pyrus ussuriensis Maxim. in China

Pyrus ussriensis Maxim. is native to the northern part of China, but whose habitats are currently being destroyed by environmental changes and human deforestation. An investigation of population structure and genetic diversity of wild Ussurian pear is a priority in order to acquire fundamental knowledge for conservation. A total of 153 individuals of wild Ussurian pear from the main habitats, Heilongjiang, Jilin, and Inner Mongolia in China, possessed low genetic diversity as a result of habitat fragmentation. The genetic diversity of the populations in Inner Mongolia and north east of Heilongjiang was especially low and there was the possibility of inbreeding. Wild Ussurian pears were divided into 5 groups based on the Bayesian clustering method using 20 nuclear SSRs (nSSRs) and 5 groups by haplotype distributions using 16 chloroplast SSRs (cpSSRs), and the populations in Inner Mongolia and north east of Heilongjiang represented unique genotypes. AMOVA indicated there was a 20.05% variation in nSSRs and a 44.40% variation in cpSSRs among populations. These values are relatively high when compared to those of other tree species. Haplotype E, positioned in the center of the cpSSR analysis network and showed the largest number of connections with other haplotypes, represented the most important haplotype. Inner Mongolia and the north east of Heilongjiang are two areas that need urgent conservation because of their genetic vulnerability and peculiarity. We determined 4 conservation units based on the clustering by nSSRs and cpSSRs, and geographic factor. This information is helpful in deciding the conservation strategies for wild Ussurian pear in China.


Introduction
Investigating the genetic population structure, which includes genetic diversity, population differentiation and degree of gene flow [1], is very important in revealing the origin and the diversity of Chinese wild Ussurian pear using hypervariable regions of chloroplast DNA. But, to date, there has been no large-scale investigation of Chinese wild Ussurian pear using nuclear DNA.
Genetic analysis of population structure requires suitable markers. Microsatellite (Simple Sequence Repeat, SSR) has been used widely in construction of genetic maps, parentage and population genetic structure analysis [22]. Because SSR marker is PCR-based, highly reproducible, polymorphic, generally co-dominant and abundant throughout the eukaryotic genome, it has become a popular genetic marker in many species [23]. Recently many SSR primers were also developed for Pyrus [24,25,26,27,28] and Malus [29,30,31], and Yamamoto et al. [31] reported that SSR primers based on apple can also be transferred for Pyrus. Chloroplast DNA marker is a useful tool for evolutionary studies and population structure analysis because of the non-recombinant and uniparentally inherited nature of organelle genomes [32,33]. Detecting useful polymorphisms at population levels is however often difficult because of the low level of substitutions in the chloroplast genome. The discovery of polymorphic mononucleotide repeats (cpSSR) in the chloroplast genome has revolutionized plant evolutionary biology and population genetics [34,35]. The location of cpSSRs in Pyrus was reported in the entire sequence data of P. pyrifolia cpDNA [36]. The investigation of population genetic structure using these two kind of markers, can help to define conservation priorities for Pyrus [37].
The aim of this study is (1) to reveal genetic structure of wild Ussurian pear in China, such as genetic diversity, gene flow, and genetic divergence. This study is an initial step in determining suitable conservation units, and considering possible conservation strategies for the wild Ussurian pear. The second aim of this study is (2) to reveal the phylogenetic relationships between P. ussuriensis var. aromatica in Japan and P. ussuriensis Maxim., and get an insight into the origin of P. ussuriensis.

Collection sites and DNA extraction
Almost no investigations of wild Ussurian pears in China have been conducted until now, so their habitats were unclear. After explorations based on the old documentary records and interviews, 13 populations of Ussurian pears growing wildly were recognized. A total of 153 wild Ussurian pear (Pyrus ussuriensis Maxim.) individuals were collected from their main habitat, Inner Mongolia, Heilongjiang, and Jilin in China. We collected 13 populations comprised as follows: 91 individuals from 6 populations in Inner Mongolia, 70 individuals from 6 populations in Heilongjiang, and 12 individuals from a single population in Jilin (Fig 1). All of the samples were used for nSSRs analysis, and 145 individuals (86 individuals from 6 populations in Inner Mongolia, 49 individuals from 6 populations in Heilongjiang, and 10 individuals from a single population in Jilin) were used for cpSSRs analysis. Sampled locations were recorded as global positioning system (GPS) coordinates ( Table 1). Classification of Pyrus based on those of Yu (1979) and Ohwi (1965). To avoid multiple sampling of closely related individuals such as half sib or full sib., a minimum distance of 10 m between individuals was maintained.

Nuclear and chloroplast SSR markers
Development of cpSSR markers. Terakami et al [36] detected a total of 67 SSR regions (≧10 repeated motifs) by sequencing the entire chloroplast DNA of Pyrus pyrifolia. In this study, the software program, Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www. cgi), was utilized to design primer pairs flanking SSRs. The major parameters for primer design were as follows: primer length, about 20 bp; PCR product size, between 100 to 250 bp;  Table 3). The haplotypes are indicated by circles and the colors correspond with the color of the haplotypes in Fig 1 (a), and small red circles show median vectors. The size of each pie chart is proportional to the frequency of corresponding haplotype. (1)-1 including 2 haplotypes was detected mostly in the north east of Heilongjiang, and (1)-2, (1)-3, and (1)-4 contained the haplotypes found in Jilin, Inner Mongolia, and Mudanzhang City, Heilongjiang, respectively. Most of the median vectors were included in (2). doi:10.1371/journal.pone.0133686.g001 optimum annealing temperature, 60°C; GC content, about 50%. Synthesized primer pairs (Life Technologies Japan Ltd) were tested to detect polymorphisms within 95 individuals of wild Ussurian pear endemic in Japan, and 16 cpSSR primers were selected for the present study (Table 2 and S1 Table).

Statistical data analysis
Genetic diversity in wild Ussurian pears. The genetic diversity as defined by nSSRs in each population was quantified in terms of the number of alleles per locus (Na), effective number of alleles (Ne), allelic richness (Ar), the mean observed heterozygosity (Ho), expected heterozygosity (He), and fixation index (F IS = 1-Ho / He) using POPGENE 32 version 1.3.2 [39], FSTAT version 2.9.3.2 [40], Arlequin version 3.11 [41]. Deviations from Hardy-Weinberg expectations were determined by F IS and their significance (F IS 6 ¼ 0) was examined using 1,000 permutation tests. For cpSSRs, the number of haplotypes per locus (Na), effective number of haplotypes (Ne) and Nei's gene diversity were also calculated using POPGENE 32 version 1.3.2.
Genetic structure. The Bayesian model-based clustering programs STRUCTURE 2.3.4 [42] were employed to detect population structure and assign individuals to groups using the nSSRs under the admixture model and the option of correlated allele frequencies between populations. The cluster number (K) was set to vary from 1 to 20. The model was run as 10 independent simulations for each K and used a burn-in length of 100,000 and a run length of 1,000,000 MCMC iterations. The relationships between the K value and the data likelihood of K were plotted. To know the detail population structure of wild Ussurian pear in China, the run was conducted with only wild Ussurian pear in China. The cluster number (K) was set to vary from 1 to 10. The model was run as 10 independent simulations for each K and used a burn-in length of 100,000 and a run length of 1,000,000 MCMC iterations.
Analysis of molecular variance (AMOVA) was performed to examine the hierarchical genetic structure using the program Arlequin version 3.11 [43] for wild Ussurian pears. Genetic variance was partitioned into two levels using wild Ussurian pears, among populations and within a population. The significance of variance components and the differentiation statistics were tested with 1023 permutations.
To estimate the relative amounts of pollen and seed flow, we calculated the pollen to seed migration ratio (r = m p / m s ) as described by Mitsui et al (2010) using only wild Ussurian pears in China. The equation of m p / m s is based on the prediction that the levels of population differentiation can differ between nuclear and cytoplasmic markers because of their different modes of inheritance and that the extent of population differentiation is related to the relative levels of pollen and seed migration among populations [44,45]. For biparentally inherited nuclear markers, gene flow occurs in both pollen and seeds, whereas the genes for maternally inherited cytoplasmic markers are only dispersed in seeds.
Mantel tests [46] were carried out using Arlequin to test the significance of isolation by distance patterns (IBD), regressing pairwise population F ST / (1-F ST ) values against the geographic distances between the respective pairs of populations [47]. The geographic distances used in this study were calculated using the Lambert-Andoye Method on the Internet; LatLng2Distance (http://lab.uribou.net/ll2dist/). Phylogenetic tree. A dendrogram was constructed using nSSRs data to interpret relationships among populations by a neighbor-joining (NJ) method [48] based on Nei's DA distance [49] by. Populations 1.2.32 [50]. Genetic distance was calculated in POPGENE 32 version.
A median-joining network including potential median vectors was constructed with wild Ussrian pears in China using Network 4.5.1.6 [51] using cpSSRs data. Any individuals with missing data were removed from this analysis (Table 3). We also calculated network including closely related species with wild Ussurian pears in China such as 'cultivated Ussrian pears in China' and 'Iwateyamanashi'.
To determine the morphological features for each population, we performed a discriminant analysis, using the program JMP 8 (SAS Institute Inc., Cary NC, USA). Discriminant analysis was performed using all of the morphological traits except for calyx existence.

Chloroplast haplotypes characterized by 16 cpSSR markers
A total of 105 chloroplast haplotypes were identified amongst the samples such as wild individuals and cultivars using 16 cpSSRs. These cpSSRs harbored enough genetic diversity to represent the genetic variation of the Pyrus species. Twenty five haplotypes were represented in wild Ussurian pears in China (Fig 1A, Table 3). Wild Ussurian pear in China were divided into 5 groups based on the haplotype distribution; (1) Inner Mongolia; IMQS, IMTHL, IMPJG, IMRSL, IMLMD, and IMSLG, (2) The north of Heilongjiang; HLSWX, HLYCS2, HLYCS3 and HLFYX, (3) HLYCS1: This population had unique haplotype and didn't have the haplotype E which is representative haplotype in Heilongjiang, (4) HLMTZ and (5) JLGZL. Four haplotypes, A, B, C, and D, were found among the 6 Inner Mongolian populations. Seventy six percent of the 73 individuals from Inner Mongolia could be assigned to haplotype A, which was considered to be the major haplotype amongst 6 Inner Mongolian populations. Thirteen haplotypes, E, F, G, H, I, J, K, L, M, V, W, X, and Y were detected in the 6 populations from Heilongjiang, and 9 haplotypes, E, N, O, P, Q, R, S, T, and U were identified in the population from Jilin.
Haplotype E was detected in both the Heilongjiang and Jilin populations, and represented 41% of the 51 individuals from these populations. Seventy one percent of the 20 individuals represented by haplotype E were detected in the north east in 2 populations from Heilongjiang (HLYCS3 and HLFYX), therefore haplotype E was the most common haplotype for wild Ussurian pears in the north east of Heilongjiang (Fig 1A, Table 3). The relationships between 25 cpSSR haplotypes detected in wild Ussurian pears in China were analyzed by a median-joining network model (Fig 1B). The haplotype network is an indication of the minimum number of evolutionary events separating each haplotype. The Network indicated that haplotype A is derived from E via T, Q, and R, and the other haplotypes also seemed to be subdivided from E. Some cultivated Ussurian pears in China also derived from haplotype E (S1 Fig). Consequently, haplotype E is considered to be the divergent center of wild Ussurian pears. Nine median vectors (mv) inferred from the network might have arisen from homoplasious mutation based on mononucleotide repeats in cpSSRs (Fig 1B). Possible shortest least complex phylogenetic tree (maximum parsimony tree) was also reconstructed by using the Network software, and the result was same to that with median-joining method.

Genetic structure in Chinese wild Ussurian pears
The results of the AMOVA for nSSR showed that 20.05% of the variation was due to differences among populations, and 79.95% within each population. For cpSSR, 44.40% of the total genetic variation was due to differences among populations, and 55.60% within each population. Differentiation among populations for cpSSRs was more than twice as large as that of nSSRs. The calculated ratio of pollen flow to seed flow were 5.8: 1 in 13 wild populations from China.
In the STRUCTURE analysis, the plot of the average log-likelihood values reached plateau at K = 10 (Figs 2A and 3). At K = 10, wild Ussurian pears were divided into 4 groups. But when only wild Ussurian pears were applied to STRUCTURE analysis, the average log-likelihood values reached plateau at K = 5 (Figs 2B and 4A). K = 5 reflects more reasonable result, because STRUCTURE analysis seek to identify the admixture among the species that when including the related species. So we divided wild Ussurian pears into 5 groups; (1) Inner Mongolia; IMQS, IMTHL, IMPJG, IMRSL, IMLMD, and IMSLG, (2) The north east of Heilongjiang; HLYCS3 and HLFYX, (3) the central part of Heilongjiang; HLSWX, HLYCS2 and HLYCS1, (4) Jilin; JLGZL, (5) the southern part of Heilongjiang; HLMTZ (Fig 3A). HLYCS2 was found to be admixtures of genotypes found in Jilin. Geographical distribution of 5 groups was shown in Fig 3B. Most of the cultivars formed admixtures at K = 10, but QZL1, QZL2, and JAP were made up of relatively uniform gene pool (Fig 4).
As a result of the Mantel test, IBD pattern was detected for both nSSR and cpSSR. Pairwise F ST / (1-F ST ) values were significantly correlated with geographic distance among populations for nSSR (r = 0.687, P = 0.000) and cpSSR (r = 0.810, P = 0.000).

Genetic diversity within populations and the Hardy-Weinberg equilibrium
For nSSRs, which harbored sufficient genetic diversity to represent the genetic variation of the Pyrus species (S3 Table), the average values of the indicators of na, ne, Ar, Ho, He, and FIs were 4.96, 2.90, 3.34, 0.40, 0.60 and 0.25 respectively for the wild Ussurian pear in China when the genetic diversity was calculated for 5 groups based on the STRUCTURE analysis (Fig 3A  and 3B), and 5.85, 3.67, 1.71, 0.56, 0.71, and 0.16 respectively for Iwateyamanashi (IWT) ( Table 2). These values except Ar shows that the wild Ussurian pear in China has less genetic diversity than Iwateyamanashi.
In  Table 2) as did HLFYX, which come from the north east of Heilongjiang (na = 2.85, Ne = 1.99, Ar = 1.39, Ho = 0.30, and He = 0.36). The F IS values were positive in all cases, except for the population, HLYCS1, and 3 populations from Inner Mongolia (IMTHL, IMPJG, and IMLMD). Two populations from Heilongjiang (HLYCS3 and HLFYX) showed significant deviations from the Hardy-Weinberg equilibrium. In cpSSRs, wild Ussurian populations from China also showed less genetic diversity than Chinese Ussurian cultivars (QZL) and Iwateyamanashi (IWT). JLGZL and HLYCS2 showed high genetic diversity, and all of the populations in Inner Mongolia had low diversity ( Table 2). When the genetic diversity of wild Ussurian pear was calculated for the population based on the collection site, there was the population which had small sampling size (HLYCS1). This possibly caused minimize the values of Allelic richness, but when removed the population, HLYCS1, there was nearly unchanged in the values (S9 Table).

Genetic relationships between wild Ussurian and cultivated pear
Thirteen populations of wild Ussurian pears from China and Japan, and 6 cultivar groups were separated into 6 clades on an unrooted NJ tree; (1) The 6 populations (IMQS, IMTHL, IMPJG, IMRSL, IMLMD, and IMSLG) originating from Inner Mongolia which were all closely related.

Flower and fruit morphology
In order to demonstrate the relationships between populations based on flower and fruit morphologies, discriminant analysis was carried out using the values of 5 floral and 4 fruit morphological characters (Fig 6I and 6II). The 5 populations from Inner Mongolia (IMQS, IMTHL,  (Fig 6II). HLYCS3 was plotted on left upper side. An Ussurian pear cultivar 'Nanguoli'  was located on the right middle side. The fruit size and peduncle length of 'Nanguoli' were bigger and longer than those of wild Ussurian pears. With regard to wild populations, all of the fruits were much smaller than that of 'Nanguoli'. The fruit morphologies of wild Ussurian pears could be divided into two groups: circle (1) the group plotted on left lower side such as Mongolian populations harboring globular shape and short peduncles, and circle (2) the group located on left upper part such as HLYCS3 harboring ellipse shape and relatively long peduncles.

Low genetic diversity of wild Ussurian pears
Using nSSR markers, we have shown that wild Ussurian pears from China have much lower genetic diversity Iwateyamanashi. The genetic diversity of wild Ussurian pear in China inferred from the values of na, ne, Ho, and He was low compared to the values of wild grapevine from Spain, wild apple (Malus sieversii and M. sylvestris), African fruit tree of Rosaceae (Prunus africana), wild Callery pear (P. calleryana), and the average of many perennial and outcross species [52,53,54,55,56].
The low genetic diversity of Chinese wild Ussurian pears might be the result of restrictive gene flow which is important in maintaining genetic diversity and minimizing genetic drift arising from habitat fragmentation because of deforestation caused by climate change in Inner Mongolia and human development in Heilongjiang. Pairwise F ST were also significantly different among populations with the exception of several populations within Inner Mongolia (S3 Table and S4 Table).
The measurement of within population diversity revealed that the values of F IS were high in most of the wild populations in China (Table 2). This result could be attributed to nonrandom mating promoted by habitat fragmentation and limited gene flow causing inbreeding. S-locus usually can avoid self-pollination and crossing closely-related individuals, but when belonging to small population, inbreeding such as sister brother mating can be occurred.

Genetic population structure in wild Ussurian pears in China
Wild Ussurian pears in China were highly structured. Differentiation among collecting sites was higher than other congeneric species (20.02%), referring to previous studies such as P. calleryana: 8.86% for among populations, 91.14% for within a population [61], P. ussuriensis var. aromatica: 7.20% for among populations, 92.80% for within a population (S6 Table). For cpSSRs, genetic variations among populations of wild Ussurian pears (44.40%) from China were twice higher than that of wild Callery pears (22.9%) from China [61]. These results indicated that the wild Ussurian pears in China had high population structure. Actually, wild Ussurian pear in China divided into 5 groups based on the STRUCTURE analysis and chloroplast haplotype distribution (Figs 1A and 4B). The grouping using nSSRs and cpSSRs almost corresponded except north of Heilongjiang. STRUCTURE analysis divided the populations in north of Heilongjiang into 2 groups (Group2; HLYCS3 and HLFYX, Group3; HLYCS1, HLYCS2, and HLSWX). The populations in north of Heilongjiang such as HLYCS1, HLYCS2, and HLSWX were belonging to the same group by STRUCTURE analysis, but in chloroplast haplotype distribution HLYCS1 was unique and didn't have haplotype E which was representative haplotype in Heilongjiang. HLSWX, HLFYX, and HLYCS3 belonging to green colored Group 2 by cpSSRs locate near the river, the first 2 populations are in close to Amur River and the latter is in close vicinity to Tangwang River flowing into Songhua River which is the greatest tributary of Amur River (Fig 1A). The circumstances of these 3 populations are similar, such as woody and damp ground. However, in STRUCTURE analysis (K = 5) most of the individuals from HLSWX were assigned to dark blue colored cluster same as HLYCS1 and HLYCS2 (Fig 3A and 3B). HLSWX, HLYCS1, and HLYCS2 are located on skirt of Xiaoxinganling Mountains, and therefore gene flow by pollen may often occur among these 3 populations. HLYCS3 also close to location of HLYCS1 and HLYCS2, however, HLYCS3 is located on small woods separated from Xiaoxinganling Mountains. Thus the gene flow might not be so frequent between HLYCS3 and HLYCS1 and HLYCS2. By taking into account geographic factor such as mountains, 5 groups revealed by STRUCTURE analysis are suitable for representing current population structure of wild Ussurian pear in China (Fig 3B). A significant correlation between genetic distance and geographic distance was detected by a Mantel test using both nSSRs and cpSSRs. Isolation by distance was expected to be the main factor causing the high level of population divergence in wild Ussurian pears, due to limited gene flow and the effects of genetic drift as a result of habitat fragmentation. Geographic distance influences migration of pollen and seed and restricts the gene flow among populations. The ratio of pollen to seed flow was calculated using the equation of Ennos [45], and the result showed that pollen flow was 5.8 times higher than seed flow. The ratios reported previously for other tree species showed much higher values than that for wild P. ussuriensis from China (Quercus petraea; 196: 1, Pinus contorta; 28: 1, P. radiate; 31: 1, P. attenuat; 44: 1, P. muricata; 24: 1) [45]. This is mainly caused by the activities of frugivorous animals. Small pear fruits are favored by many kinds of animals, and the seeds are carried relatively long distances. A similar result was obtained for wild P. calleryana from China [61]. The ratio of pollen to seed was 0.56: 1, this means that seed flow plays an important role in wild callery pear trees. It is thought that wild Ussurian pear seed could also be moved easily.
'Group 1' included all of the populations from Inner Mongolia (Fig 3B). (2) Bottleneck effect: the wild populations in Inner Mongolia have been conjectured that they were experienced bottleneck from the demographic reduction by environmental changes, so that inbreeding has occurred in small populations leading to uniform genotype and low genetic diversity.
'Group 5' included HLMTZ, from Muling, Mudanjiang City, Heilongjiang Province. Based on STRUCTURE analysis, this population was separated from the other wild populations in Heilongjiang, and positioned near the Ussurian pear cultivar group in the phylogenic tree ( Fig 5). Moreover the flower morphologies of this population were similar to those of the Ussurian pear cultivars (Fig 6). Historically, cultivation of Ussurian pear was prosperous in Mudanjiang City, so introgression of Ussurian pear cultivars into wild Ussurian pear trees might have occurred frequently in HLMTZ over time. This was also known from low bootstrap value in the phylogenetic tree (Fig 5). This result is consistent with the previous reports that P. ussuriensis growing wild in Northern part of Japan was introgressed from Japanese pear cultivars [17,62].

The inference of primitive haplotype of wild Ussurian pear in China
Haplotype E which is the representative haplotype of Heilongjiang, and according to coalescent theory [63], there is a possibility that haplotype E is primitive haplotype detected in this study, because it shows the largest number of connections with other haplotypes including wild and cultivated Ussurian pears (S1 Fig) and positioned in the center of the network. Previous research by Wuyun et al. [11] using hypervariable regions of cpDNA also revealed that a haplotype (Hcp3) found in populations such as HLSWX, HLYCS3, HLFYX, and JLGZL in Heilongjiang and Jilin was at the divergent center of a haplotype network. Therefore, haplotype E is very important to investigate the origin of Ussurian pear.

The inference of domestication of Ussurian pear
Although the flower shape in the population 'HLFYX' represented intermediate flower morphology between wild and its derived cultivars, wild Ussurian pear and cultivated Ussurian pear in China were genetically divergent from each other (Figs 4, 5 and 6I). Another study by Cao et al [21] using M13-tailed SSR markers showed that Ussurian pear cultivars had close relationships with the wild accessions found in the southern areas of north east of China, such as Liaoning and Hebei. All of the wild Ussurian pear trees from Heilongjiang and Jilin were clustered in a group by UPGMA tree, and were separate from the cluster of Ussurian pear cultivars [21]. From molecular data, wild pears in Heilongjiang and Jilin were not considered to be related with domestication of cultivated Ussurian pear. More cultivated Ussurian pears need to be analysed to know the origin of Ussurian cultivars.
The phylogenetic relationship of Ussurian pear in China and Iwateyamanashi In this study, P. ussuriensis Maxim. (wild Ussurian pear in China) and P. ussuriensis var. aromatica (Iwateyamanashi) were genetically divergent from the result of STRUCTURE analysis and phylogenetic tree (Figs 4 and 5). This reason is as follows; Most of the oldest definitive members of the Rosaceae were already present in the Eocene upland floras of the Okanogan Highlands of northeastern Washington State and British Columbia, Canada [64]. The fossils of leaves recognized as Pyrus were also discovered in the stratum of Eocene [65]. Neogene rosaceous occurrences are more widespread with reports known in Asia, particularly Japan, the Arctic, several Gondwanan regions, and northern Africa [64]. At that time the Japanese archipelago have been connected with the Asian continent by land deformation, and P. ussuriensis Maxim. was considered to be already distributed in the region which was the present northern part of Japan and northeast part of Eurasian continent. From the latter Miocene to the Pliocene, the Sea of Japan became to expand and the original form of present island of Japan were developed, consequently both of P. ussuriensis Maxim. in Japan and continent were geographically isolated. There was little gene flow between P. ussuriensis Maxim. (wild Ussurian pear in China) and P. ussuriensis var. aromatica (Iwateyamanashi), then each of these Pyrus followed the evolutionally different way.

Conservation strategy
The detection of population differentiation using cpSSRs haplotypes and STRUCTURE analysis using nSSRs can help to define appropriate conservation units, and provide a good focus for conservation management [66]. Wild Ussurian pear in China can be divided into 5 genetically distinct groups (Fig 3A and 3B), and their flower and fruit morphologies diverged in each population (Fig 6I and 6II). But group 5 was considered to be the population introgressed by cultivars, so 4 conservation units were set up with removing group 5. Group 2 and 3 had high genetic diversity. In Group 2, the genetic diversity of HLYCS2 having rich ecological conditions which included all of the genotypes in Heilongjiang revealed by STRUCTURE analysis was especially high. Conservation priority should be given to populations with a high level of genetic diversity and unique genotypes possessing the risk of extinction in the future. HLYCS2 should be perfectly conserved for in situ level. This study also revealed that the genetic diversity of group 1 was especially low (Table 2). In these populations, inbreeding and the shortage of gene flow were indicated ( Table 2, Figs 3A and 5), thus the risk of extinction in the future was high. Because some individuals in Inner Mongolian populations had unique genotypes, these individuals should be reserved with ex situ conservation.