Population Structure , Genetic Diversity , and Gene Introgression of Two Closely Related Walnuts ( Juglans regia and J . sigillata ) in Southwestern China Revealed by EST-SSR Markers

The common walnut (Juglans regia L.) and iron walnut (J. sigillata Dode) are well-known economically important species cultivated for their edible nuts, high-quality wood, and medicinal properties and display a sympatric distribution in southwestern China. However, detailed research on the genetic diversity and introgression of these two closely related walnut species, especially in southwestern China, are lacking. In this study, we analyzed a total of 506 individuals from 28 populations of J. regia and J. sigillata using 25 EST-SSR markers to determine if their gene introgression was related to sympatric distribution. In addition, we compared the genetic diversity estimates between them. Our results indicated that all J. regia populations possess slightly higher genetic diversity than J. sigillata populations. The Geostatistical IDW technique (HO, PPL, NA and PrA) revealed that northern Yunnan and Guizhou provinces had high genetic diversity for J. regia while the northwestern Yunnan province had high genetic diversity for J. sigillata. AMOVA analysis revealed that significant genetic variation was mainly distributed within population as 73% in J. regia and 76% in J. sigillata. The genetic differentiation (FST) was 0.307 between the two walnut species (p < 0.0001), which was higher than FST values within populations (J. regia FST = 0.265 and J. sigillata FST = 0.236). However, the STRUCTURE analysis of the J. regia and J. sigillata populations revealed two genetic clusters in which gene introgression exists, therefore, the boundary of separation between these two walnut species is not clear. Moreover, these results were validated by NJ and UPGMA analysis with additional conformation from the PCoA. Based on the SSR data, our results indicate that J. sigillata is an ecotype of J. regia. Taken together, these results reveal novel information on population genetics and provide specific geographical regions containing high genetic diversity of the Juglans species sampled, which will assist in future conservation management.


Introduction
Walnuts trees have been valued since ancient times for their edible nuts and high-quality timber [1,2].Walnuts belong to the family Juglandaceae (genus Juglans) which includes more than 20 species, with the most well-known being J. regia L. (common walnut) [1][2][3][4][5].The Juglans have a diploid genome with a

Sampling and DNA Extraction
Leaf samples from a total of 506 individuals from 28 populations of J. regia and J. sigillata were collected within the main range of their sympatric distribution in Yunnan, Guizhou, and Sichuan province of southwestern China (15 populations from J. regia (n = 190) and 13 populations from J. sigillata (n = 316), details in Table S1).Most leaf samples were collected from primary forests in rural locations.On average, the distance between collections for each population was greater than 500 m while the distance of individuals between a pair of populations within the same species was greater than 200 meters.Samples of J. regia and J. sigillata were identified based on their morphology including leaflet number (5-9 in J. regia and 9-11 in J. sigillata), flower, nut, and shell (J.regia has a wrinkled fruit surface; J. sigillata has deep pits and seal-like depressions) [20].Fresh leaves were first dried with silica gel prior to high-quality genomic DNA extraction using a modified CTAB method [21,22] and storage at −20 • C at the Evolutionary Botany Lab, Northwest University, Xi'an, China.DNA samples deriving from J. regia are designated by an "r" suffix and J. sigillata by an "s" suffix.

Genetic Diversity Analysis
Genetic diversity per locus and between populations was evaluated through the following descriptive summary statistics: number of alleles (N A ), observed (H O ) and expected (H E ) heterozygosity, Percentage of polymorphic loci (PPL), Private alleles (PrA), Shannon index (I), and Fixation Index (F) using the program GenAlEx6.5 [28].The presence of null alleles was monitored using MICRO-CHECKER 2.2.3 [29].Arlequin 3.5 [30] was used to test the Hardy-Weinberg equilibrium (HWE), to calculate F ST for identification of potential outlier loci under selection, and for determination of Analysis of molecular variance (AMOVA) for investigation into genetic variations among and within populations for each species and species pair using 10,000 permutations.Prior to AMOVA analysis, populations TC-r, DL-r, HD-r, and WXT-r were removed due to low sample size (3-6 individuals per population).The linkage disequilibrium (LD) for all loci was tested using FSTAT [31].GenAlEx6.5 was used to conduct a Mantel test [32] and analyze the correlation between geographical distance and genetic distance.

Genetic Structure Analysis
An analysis to detect genetic structure was performed using STRUCTURE (Pritchard Lab, Stanford University, Stanford, CA, USA, version 2.3.4) with a burn-in of 200,000 Markov Chain Monte Carlo (MCMC) iterations, a duration of 500,000 iterations, ten replicates per run for K (from 2 to 10 clusters), and the admixture model [33,34].STRUCTURE was run based on 15 neutral loci, 10 selected loci, and all 25 EST-SSR loci (Figures S1 and S2) using four datasets: (1) all samples from J. regia and J. sigillata (Table S1), (2) seven pairs of the main populations displaying sympatric distribution of J. regia and J. sigillata, (3) all fifteen populations of J. regia, and (4) all thirteen populations of J. sigillata.The program STRUCTURE HARVESTER was used to calculate the optimal value of K using the deltaK criterion [35].The inferred clusters were drawn as colored box-plots using the program DISTRUCT [36].Principal coordinate analysis (PCoA) was performed with GenAlEx 6.5 [28].In order to detect signatures of bottlenecks in studied populations, the program Bottleneck 1.2 was used [37] and excess heterozygosity in each population was assessed by applying the two-tailed Wilcoxon's sign-rank test [38].P values associated with bottleneck tests were calculated by performing 10,000 permutations under both the two-phase model (TPM) and stepwise mutation model (SMM).

Genetic Barrier Analysis
The presence of genetic barriers, employed to highlight geographical features corresponding to pronounced genetic discontinuity, was investigated using Monmonier's maximum difference algorithm as implemented in BARRIER 2.2 [39].

Landscape Genetics
The Inverse Distance Weighted (IDW) [8,40] interpolation function implemented in the Geographic Information System (GIS) software ArcGIS 9.3 (ESRI, Redlands, CA, USA) was used to display the geographic patterns of observed heterozygosity (H O ), percentage of polymorphic loci (PPL), the number of alleles (N A ), and private alleles (PrA) of the 28 populations sampled and to derive maps of genetic diversity.The IDW algorithm used a linear weighted combination of a set of sample points to estimate the value of a target variable in a new position.IDW assumed that points close to each other are more relevant than those that are more distant and is weighted more closely to the predicted position than the farther distances [40].

2.7.
Inter-Specific Gene Flow MIGRATE 3.6.4[41] was used to estimate the historical gene flow parameter (M) and the frequency of migration events through coalescent history between J. regia and J. sigillata for the EST-SSRs.The mode and 95% highest posterior density were then estimated after checking for data convergence.

Phylogenetic Relationship Analysis
The genetic relationship among populations was explored by generating neighbor-joining (NJ) trees [42] and the unweighted pair-group method with arithmetic means (UPGMA) tree using the software POPTREE2 [43] based on Nei genetic distance [44] and visualized with the software Fig Tree 1.4.2[45].

Comparisons of Genetic Diversity and Differentiation between J. regia and J. sigillata
The null allele test indicates a lower frequency of null alleles at 19 loci when compared to the threshold frequency (* = 0.15) across all the J. regia and J. sigillata populations studied while the probability of null alleles was slightly significant for the remaining six loci.No evidence was found for the presence of linkage disequilibrium (LD).The Hardy-Weinberg equilibrium (HWE) test revealed highly significant deviations for many EST-SSR loci.These deviations may result from a deficiency of heterozygosity among the J. regia and J. sigillata populations.Only two loci (JR6439 and JR3147) showed no significant deviations which suggests that almost all populations may be affected by factors of interference such as introgression, mutation, and selection by migration (Table 1).
Genetic diversity parameters for each population based on allelic frequencies are summarized in Table 2.The results indicate that all J. regia populations possess slightly higher genetic diversity than the J. sigillata populations.The mean value of alleles (N A ) is 2.261 in J. regia and 2.098 in J. sigillata.The number of effective alleles (N E ) per population varies from 1.053 (LM-r) and 2.139 (QZ-r) in J. regia, while in J. sigillata, the values vary between 1.269 and 1.925.For J. regia, the lowest observed heterozygosity (H O ) and expected heterozygosity (H E ) are found in Liming, Yunnan population; the highest in DL-r (0.467) and QZ-s (0.419).Both values of heterozygosity in the majority of J. regia populations are somewhat higher than in J. sigillata.Shannon's information index (I) ranges from 0.111 to 0.735, with an average of 0.518 in J. regia populations, while the mean number of I is 0.449 in J. sigillata (Table 2).Moreover, the Percentage of Polymorphic Loci (PPL) in J. regia (mean = 0.328) is higher than in J. sigillata (mean = 0.273), with the majority of J. regia populations having higher PPL values.Taken together, these values support higher genetic diversity among J. regia populations.The fixation index (F) determined for most populations showed significant deviations from zero, indicating a high level of inbreeding within individuals of each population.The Wilcoxon test was used to detect possible bottlenecks in each population, and signatures of significant recent population bottlenecks were detected in LJ-r and GZ-r populations of J. regia, and in the TZ-s population of J. sigillata (Table 3).
The within-population inbreeding coefficient (F IS ) varies from −0.481 to 0.387, with a mean of 0.043 (Table 1).The inbreeding coefficient determined for the total population (F IT ) per locus ranges from −0.012 to 0.991 with an average of 0.482.Meanwhile, the genetic differentiation (F ST ) ranges from 0.206 to 0.991, with a mean of 0.479, indicating limited gene flow and high differentiation between populations.As shown in Table 1, F IS is significantly negative for many loci while significantly positive for some others.F IS values that are significantly greater than zero indicate a deficiency of heterozygosity at this locus, probably as a consequence of the allelic dropout, including the presence of null alleles [46].The AMOVA analysis revealed the percentage of variation found between J. regia and J. sigillata is 7.311 % (p < 0.0001), demonstrating that the differences between the two species is significant.Therefore, analyzing the genetic structure of the J. regia population and the J. sigillata population independently is necessary.Within the population of each species, the proportion of variance is notable based on AMOVA (J.regia 73 %, p < 0.0001; J. sigillata 76 %, p < 0.0001).These data indicate that higher genetic variation of J. regia and J. sigillata is mainly distributed within populations (Table 3).

Spatial Genetic Structure of Populations and Divergence
The possibility exists that EST-SSRs may be located within functional genes and that the EST-SSR marker selected for study are not necessarily neutral.Inclusion of non-neutral loci may bias the population genetic diversity and genetic structure analysis.Prior to conducting genetic structure analysis, Arlequin3.5 software was used to identify possible selected loci for all 25 EST-SSRs.Ten selected loci were identified: JM5969, JM61666, JR4964, JR4616, JR3773, JR1165, JM6638, JH6160, JR3147, and JH2096 (Figure S3).Finally, 15 polymorphic and neutral EST-SSR markers were selected to assess population genetic structure among the 28 walnut populations studied herein.
The genetic structure of the 15 J. regia and 13 J. sigillata populations was inferred using STRUCTURE (Figure 1).

Spatial Genetic Structure of Populations and Divergence
The possibility exists that EST-SSRs may be located within functional genes and that the EST-SSR marker selected for study are not necessarily neutral.Inclusion of non-neutral loci may bias the population genetic diversity and genetic structure analysis.Prior to conducting genetic structure analysis, Arlequin3.5 software was used to identify possible selected loci for all 25 EST-SSRs.Ten selected loci were identified: JM5969, JM61666, JR4964, JR4616, JR3773, JR1165, JM6638, JH6160, JR3147, and JH2096 (Figure S3).Finally, 15 polymorphic and neutral EST-SSR markers were selected to assess population genetic structure among the 28 walnut populations studied herein.
The genetic structure of the 15 J. regia and 13 J. sigillata populations was inferred using STRUCTURE (Figure 1).The optimal cluster (K) was identified based on posterior probability (∆K) values.K was tested from two to ten with 10 replicated runs performed.For seven pairs of sympatric distribution of J. regia and J. sigillata, ∆K values computed for all classes indicated a strong signal for K = 2.This implies that the 14 sympatric populations under study are grouped into two clusters (Figure S4).Several populations displayed a clearly intermixed composition, which could be attributed to introgression among QZ-r, TC-r, DL-r, HD-r, SG-r, and WXT-r populations.When all populations were included in the STRUCTURE analysis based on EST-SSR data, the most probable division with the highest ∆K value (∆K = 1683.81)was also detected at K = 2.The results revealed two genetic clusters of J. regia and J. sigillata, which identified gene introgression among different populations; however, evidence for a distinct boundary is not apparent.Eight J. regia populations (BS-r, GY-r, LJ-r, YW-r, YN-r, GZ-r, EM-r, SC-r) appear to be embedded within J. sigillata populations based on SSRs (Figure 1).The estimated population structure inferred for K = 2 is shown in Figure 1d.Results of STRUCTURE analysis using two datasets (based on only J. regia populations or J. sigillata populations) were in accordance with each other; admixture was detected between J. regia and J. sigillata and the same population genetic pattern (K = 2) was observed (Figure 2).
Additional evidence for the genetic distribution in J. regia and J. sigillata was obtained using PCoA analysis (Figure 3c).The results corroborated those obtained from the STRUCTURE (Figure 1b; Figure 1c) analysis and the first three axes in Figure 3c explain 27% of the cumulative variation.The first and second coordinates accounted for nearly 15% and 22% of the molecular variation, respectively (Table S2).The individuals of J. regia and J. sigillata sampled were mostly separated by their EST-SSR data, which indicates that interspecific differentiation is stronger than intraspecific differentiation, however, considerable overlap between the J. sigillata and J. regia samples was still detected, providing additional information on the level of genomic admixture among individuals.
To further illustrate the genetic relationships among the 28 populations studied herein, a phylogenetic analysis was completed using both the Neighbor-joining (NJ) and unweighted pair-group method with arithmetic means (UPGMA) methods for tree construction, both based on Nei's genetic distances (Figure 3a,b).The resulting dendrograms revealed a high level of gene introgression between J. regia and J. sigillata, especially for several J. regia populations (BS-r, GY-r, LJ-r, YW-r, YN-r GZ-r, EM-r, and SC-r) (Figure 3b).These results further support those obtained by STRUCTURE (Figure 1).At the individual level, a distinct boundary is not obvious, while introgression between these two closely related species is, however, apparent as individuals with gene introgression identified by STRUCTURE are roughly similar to PCoA.The resulting phylogenies are congruent and similar to the PCoA and STRUCTURE results, demonstrating that the J. regia and J. sigillata populations do cluster into two groups according to their presumed walnut species; however, the results also indicate gene introgression or hybridization between the different populations based on STRUCTURE analysis, NJ/UPGMA analysis, and PCoA analysis (Figure 1, Figure 3 and Figure S5).

Gene Introgression between J. regia and J. sigillata
The consequences of gene introgression from sympatric populations are strongly dependent on the extent of gene flow.In the Bayesian analysis of population structure, the results showed that J. regia individuals appear to be admixed with J. sigillata (Figure 1c,d).Moreover, the PCoA, NJ and UPGMA analyses confirmed the results obtained by STRUCTURE (Figure 3).The Migrate-n analysis produced θ and M values greater than zero.The θ-value and the size of the immigration rate (M) revealed a highly asymmetric historical gene flow across the two species.Gene flow occurred predominantly from J. regia into J. sigillata (16.29 vs. 8.63) (Table 4, Table S4).Meanwhile, the gene flow values (Nm) per locus ranged from 0.002 to 0.962 (average, 0.435) (Table 1).Thus, this current study provides direct evidence for historical interspecific gene flow between J. regia and J. sigillata in southwestern China.

Gene Introgression between J. regia and J. sigillata
The consequences of gene introgression from sympatric populations are strongly dependent on the extent of gene flow.In the Bayesian analysis of population structure, the results showed that J. regia individuals appear to be admixed with J. sigillata (Figure 1c,d).Moreover, the PCoA, NJ and UPGMA analyses confirmed the results obtained by STRUCTURE (Figure 3).The Migrate-n analysis produced θ and M values greater than zero.The θ-value and the size of the immigration rate (M) revealed a highly asymmetric historical gene flow across the two species.Gene flow occurred predominantly from J. regia into J. sigillata (16.29 vs. 8.63) (Table 4, Table S4).Meanwhile, the gene flow values (Nm) per locus ranged from 0.002 to 0.962 (average, 0.435) (Table 1).Thus, this current study provides direct evidence for historical interspecific gene flow between J. regia and J. sigillata in southwestern China.The mode of the posterior distribution is shown in bold and the values in square brackets give the 95% credibility interval; θ, 4 Neµ; →, source populations; M, mutation-scaled immigration rate; m, immigration rate; µ, mutation rate.

Comparative Landscape Genetics
The genetic diversity of J. regia and J. sigillata were investigated by a landscape genetic overlay approach in southwestern China which offers promise for evaluating the important role of landscape in shaping the genetic diversity of the two species.The values of observed heterozygosity (H O ), percentage of polymorphic Loci (PPL), the number of alleles (N A ) (reported in Table 2), and private alleles (PrA) (summarized in Table S3), are geographically displayed in Figure 4.The geospatial interpolation of these indices allowed the generation of new spatial data representing the genetic diversity of J. regia and J. sigillata populations in southwestern China.The observed heterozygosity (H O ) values of J. regia ranged between 0.124 (LM-r) and 0.467 (DL-r) (Table 2).High values of H O were observed in J. regia populations located in the north of its distribution in Yunnan and Guizhou provinces, while lower values of H O were shown in Sichuan and in the western and southern regions in the Yunnan populations (Figure 4a).In J. sigillata, significant differences among populations were observed for the values of observed heterozygosity (H O ) (Figure 4b) where higher numbers of H O were observed in the Himalayas (which decreased to the west) and some populations from Yunnan province, while lower values were observed in the other regions including some populations from Sichuan and Guizhou provinces.In addition, for J. regia, higher values of Percentage of Polymorphic Loci (PPL) and the number of alleles (N A ) were observed in most of the populations from Guizhou and Sichuan provinces while some of the populations from northwestern Yunnan province had the lowest numbers (Figure 4c,e).Meanwhile, in J. sigillata, the highest values of Percentage of Polymorphic Loci (PPL) and the number of alleles (N A ) are found in the area of the northwestern Yunnan (Himalayas), while regions of low PPL values were found in some populations from Yunnan and Guizhou provinces (Figure 4d,f).High values for private alleles were detected in most of Guizhou province (J.regia) and Yunnan province (J.sigillata) (Figure 4g,h).Taken together, these results reveal regions with high levels of genetic diversity that could be the result of gene flow and be associated with genetic drift.Local populations with a high genetic diversity of private alleles may be more primitive.In fact, the results of Monmonier's maximum difference algorithm implemented in BARRIER identified three statistically significant (0.10 < p < 0.40) genetic barriers in J. regia and J. sigillata (Figures S6 and S7) in accordance with the hot map of genetic diversity.The main genetic boundary was mainly distributed in the area of Guizhou province (J.regia) and Yunnan (J.sigillata) province.genetic differentiation of populations, leading to high genetic diversity.Genetic diversity, however, will also increase if existing gene flow between populations occurs for other reasons.In addition, landscape genetics has provided new opportunities to analyze spatial patterns of neutral or functional genetic diversity in a forest tree species [8].It is essential to evaluate the geographical patterns of genetic diversity and identify the populations within, and determine areas that display high values of genetic diversity and divergence.However, until now, practitioners have seldom taken into account the relevance of genetics research [58] and few studies on landscape genetics have been applied in practical forest management [59].From our point of view, landscape genetics can be considered an easy and self-explaining tool to transfer information about spatial distribution of genetic variation into practice.In this work, the application of novel spatial analysis provides more exhaustive and critical information on the genetic diversity of J. regia and J. sigillata.However, as it was discussed in a recent study, Pollegioni et al. [5] demonstrated the spatial genetic structure of walnuts in Europe using the software IDW function.
Genomic studies of introgression between closely related species are beginning to offer novel insights into the evolutionary consequences of hybridization and genetic structure.Overall, we believe that our work will provide convincing evidence for the introgression of sympatric distributions for J. regia and J. sigillata in southwestern China.Evidence for introgression between species of the genus Juglans is known, and its extent may be due to strong gene flow between closely related species with sympatric distributions.The polymorphic microsatellite loci and the genetic variation index used in this study are of great significance for research in the population genetics of walnut, and our findings, including locations of high genetic diversity, will play a significant role in conservation management of J. regia and J. sigillata.

Conclusions
In the current study, we analyzed population genetics, genetic structure, and introgression of J. regia and J. sigillata from southwestern China based on EST-SSR (expressed sequence tag-simple sequence repeat) markers from sympatric walnut populations.The genetic diversity of all J. regia populations was slightly higher than J. sigillata populations.STRUCTURE analysis revealed two genetic clusters of all samples, which showed gene introgression among different walnut populations, but failed to distinguish a boundary between the two species.The generation of phylogenetic trees, along with PCoA analysis, confirmed the results from STRUCTURE.Taken together, our results demonstrate that J. sigillata is an ecotype, landrace, or sub-species of J. regia based on SSR data.These combined results, with an emphasis on regions discovered with high genetic variation of each species, provide significant information on population genetics and structure of J. regia and J. sigillata, which can be used for conservation management in the future.The letter of "s" at the end of population names means J. sigillata and "r" means J. regia.(c) Principal coordinates analysis (PCoA) of 28 populations from J. regia and J. sigillata based on EST-SSR (25 loci); each walnut species is labeled with two different colors, representing 506 individuals that were grouped into two clusters.Figure S6.BARRIER analyses of the main genetic barriers of Juglans regia.Delaunay triangulation and detected barriers with bootstrap values over 1000 replicates using Nei's genetic distances.Blue line represents delaunay triangulation, red line indicates statistically significant genetic boundaries.Figure S7.BARRIER analyses of the main genetic barriers of Juglans sigillata.Delaunay triangulation and detected barriers with bootstrap values over 1000 replicates using Nei's genetic distances.Blue line represents delaunay triangulation, red line indicates statistically significant genetic boundaries.Table S1.Number of samples (N), location, geographic coordinates (latitude and longitude) for 28 Juglans populations surveyed in southwestern China.Table S2.Percentage of variation explained by the first 3 axes using Principal Coordinate analysis for model based approach.Table S3.Description of private alleles by 12 populations.Table S4.Summary of profile likelihood percentiles of all parameters.

Supplementary Materials:
Note: ND-Not determined; NS-Not significant; *** p < 0.0001, N A -Number of alleles, F IS -within-population inbreeding coefficient; F IT -total population inbreeding coefficient; F ST -among-population genetic differentiation coefficient; P HW -Hardy-Weinberg equilibrium; Nm-gene flow.

Figure 1 .
Figure 1.Structure clustering results for 506 individuals from 28 populations of walnut (15 J. regia and 13 J. sigillata) based on variation at 15 neutral microsatellite loci.(a) Map of China showing the region of study.(b) Geographic origin of each population sampled and their color-coded grouping (K = 2 as determined using the deltaK method of Evanno et al. [34]).The location of each pie indicates the sampled locations.The larger pie chart represents the results showing the main sympatric distribution while the smaller pie chart represents the results of all populations.The proportion of colors in each pie chart reflects the proportion of genetic affiliation within each of the two populations as determined by STRUCTURE (averaged over all samples from that location).(c) Histogram of individual assignments for seven pairs of main sympatric distributions of J. regia and J. sigillata.(d) Histogram of individual assignments for all populations.Dark blue vertical lines separate the different populations as indicated by the codes below the histogram.

Figure 1 .
Figure 1.Structure clustering results for 506 individuals from 28 populations of walnut (15 J. regia and 13 J. sigillata) based on variation at 15 neutral microsatellite loci.(a) Map of China showing the region of study.(b) Geographic origin of each population sampled and their color-coded grouping (K = 2 as determined using the deltaK method of Evanno et al. [34]).The location of each pie indicates the sampled locations.The larger pie chart represents the results showing the main sympatric distribution while the smaller pie chart represents the results of all populations.The proportion of colors in each pie chart reflects the proportion of genetic affiliation within each of the two populations as determined by STRUCTURE (averaged over all samples from that location).(c) Histogram of individual assignments for seven pairs of main sympatric distributions of J. regia and J. sigillata.(d) Histogram of individual assignments for all populations.Dark blue vertical lines separate the different populations as indicated by the codes below the histogram.

Figure 2 .
Figure 2. Structure clustering results of 15 J. regia and 13 J. sigillata populations based on variation at 15 neutral microsatellite loci.(a)The population structure based on 15 J. regia populations at K = 2 as determined using the deltaK method of Evanno et al.[34].(b) The population structure based on 13 J. sigillata populations at K = 2 as determined using the deltaK method of Evanno et al.[34].

Figure 2 .
Figure 2. Structure clustering results of 15 J. regia and 13 J. sigillata populations based on variation at 15 neutral microsatellite loci.(a)The population structure based on 15 J. regia populations at K = 2 as determined using the deltaK method of Evanno et al.[34].(b) The population structure based on 13 J. sigillata populations at K = 2 as determined using the deltaK method of Evanno et al.[34].

Figure 3
Figure 3 (a) Dendrogram generated by NJ cluster analysis of 506 individuals of J. regia and J. sigillata and (b) UPGMA cluster analysis of 28 Juglans populations (comprising of 15 J. regia and 13 J. sigillata) based on Nei's unbiased genetic distances.The letter "s" at the end of a population name indicates J. sigillata and "r" refers to J. regia.(c) Principal coordinates analysis (PCoA) of 506 individuals of J. regia and J. sigillata based on EST-SSR (25 loci).Each species is labeled with a different color, highlighting the existence of two distinct clusters.

Figure 3 .
Figure 3. (a) Dendrogram generated by NJ cluster analysis of 506 individuals of J. regia and J. sigillata and (b) UPGMA cluster analysis of 28 Juglans populations (comprising of 15 J. regia and 13 J. sigillata) based on Nei's unbiased genetic distances.The letter "s" at the end of a population name indicates J. sigillata and "r" refers to J. regia.(c) Principal coordinates analysis (PCoA) of 506 individuals of J. regia and J. sigillata based on EST-SSR (25 loci).Each species is labeled with a different color, highlighting the existence of two distinct clusters.

Figure 4 .
Figure 4. Genetic diversity maps of 15 Juglans regia and 13 J. sigillata populations (black dots) in China: IDW interpolation of (a) observed heterozygosity (H O ) values of J. regia, (b) observed heterozygosity (H O ) values of J. sigillata, (c) Percentage of Polymorphic Loci (PPL) values of J. regia, (d) Percentage of Polymorphic Loci (PPL) values of J. sigillata, (e) the Number of Alleles (N A ) values of J. regia, (f) the Number of Alleles (N A ) values of J. sigillata, (g) Private alleles (PrA) values of J. regia, and (h) Private alleles (PrA) values of J. sigillata.
Figure S1.The population structure clustering results at K = 2 as determined using the deltaK method of Evanno et al. (2005).(a) Histogram of individual assignments for all 28 populations of J. regia and J. sigillata based on 25 EST-SSRs loci.(b) Histogram of individual assignments for J. regia and J. sigillata including 28 populations, based on 10 selected EST-SSRs loci.(c) Histogram of individual assignments for seven pairs of J. regia and J. sigillata populations exhibiting sympatric distribution based on 25 EST-SSR loci.(d) Histogram of individual assignments for seven pairs of J. regia and J. sigillata population exhibiting sympatric distribution based on 10 selected EST-SSRs loci.Population codes are located below the histogram.

Figure S2 .
The population structure clustering results determined using the deltaK method of Evanno et al. (2005).(a) Histogram of individual assignments for 15 J. regia populations at K = 2 based on 25 EST-SSRS loci.(b) Histogram of individual assignments for 15 J. regia populations at K = 4 based on 10 selected microsatellite loci.(c) Histogram of individual assignments for 13 J. sigillata populations at K = 2 based on 25 EST-SSRS loci.(d) Histogram of individual assignments for 13 J. sigillata populations at K = 2 based on 10 selected microsatellite loci.Black lines separate different populations.The population codes are located below the histogram.
Figure S3.Detection of loci under selection from genome scans based on F ST .Figure S4.(a) Histogram of individual assignments for seven pairs of J. regia and J. sigillata populations displaying sympatric distribution for K = 2. (b,d) Distribution of delta K for K = 2 to 9 to determine the true number of populations (K) as described in Evanno et al. (2005).Mean log likelihood of the data at varying estimates of K. (c) Histogram of individual assignments for all populations for K = 3.The magnitude of ∆K as a function of K suggests the existence of two major clusters as the most likely scenario.Figure S5.(a) Dendrogram generated by UPGMA cluster analysis of 506 individuals of J. regia and J. sigillata and (b) NJ cluster analysis of 28 Juglans populations based on Nei's unbiased genetic distances.

Table 2 .
Genetic diversity of 28 Juglans populations based on 25 EST-SSR loci.A -The number of alleles; N E -Number of Effective Alleles; PPL-Percentage of Polymorphic Loci; H O -Observed Heterozygosity, H E -Expected Heterozygosity, I-Shannon's Information Index; and F-Fixation Index; Two mutation models of microsatellites: TPM-two-phase model; SMM-stepwise mutation model.p values lower than 0.05 are indicated in bold.
Note: N-Sample Size; N

Table 3 .
Hierarchical analyses of molecular variance (AMOVA) of Juglans samples based on 25 EST-SSR loci.

Table 4 .
Historical gene flow as estimated by Migrate 3.6.4among Juglans regia and J. sigillata.