One-hundred and thirty nine L. chinense accessions, originating from four different countries, were obtained from the National Genebank of the Rural Development Administration (RDA-Genebank), Republic of Korea. The samples were mainly from the Republic of Korea (120) and China (17). A description of the accessions used in this study is shown in Table 1.

A set of 18 highly polymorphic microsatellite markers enriched using a modified biotin–streptavidin capture method as described earlier [19] was used for the present study (Table 1). A three-primer system [21] including a universal M13 oligonucleotide (TGTAAAACGACGGCCAGT) labeled with one of the fluorescent dyes (6-FAM, NED, or HEX) was used, which allows PCR products to be triplexed during electrophoresis. A special forward primer composed by the concatenation of the M13 oligonucleotide and the specific forward primer was used with the normal reverse primer for SSR PCR amplification. Primer sequences and PCR amplification conditions for each set of primers have been described previously [19]. SSR alleles were resolved on an ABI PRISM 3100 DNA sequencer (Applied Biosystems, Foster City, CA, USA) using GENESCAN 3.7 software and were sized precisely using GeneScan 500 ROX (6-carbon-X-rhodamine) molecular size standards (35–500 bp) with GENOTYPER 3.7 software (Applied Biosystems).

The number of alleles, gene diversity (GD), heterozygosity (H), and polymorphism information content (PIC) per locus as well as the genetic distance were calculated with the PowerMarker 3.25 program [22]. The unweighted pair group method with an arithmetic mean (UPGMA) tree from shared allele frequencies was constructed using the MEGA 4.0 program [23], which is embedded in PowerMarker.

The possible population was analyzed using the Structure 2.2 model-based program [20] with a burn-in of 10,000, a run length of 150,000, and a model allowing for an admixture and correlated allele frequencies. Five runs of Structure were performed by setting the number of populations (K) from 1 to 12, and an average likelihood value, L(K), was calculated for each K across all runs. The model choice criterion to detect the most probable value of K was ΔK, which is an ad hoc quantity related to the second order change of the log probability of data with respect to the number of clusters inferred by Structure [24].

The molecular variance for model-based subgroups, F_ST, and the correlation of alleles within subpopulations were calculated using an analysis of molecular variance (AMOVA) approach in the Arlequin 3.11 program [25].

The 18 SSR markers revealed 108 alleles among the 139 L. chinense accessions representing the four countries (Table 1). The SSR loci diversity data are summarized in Table 2. The allelic richness per locus varied widely among the markers, ranging from two (GB-LCM-029; GB-LCM-111; GB-LCM-119; GB-LCM-199) to 17 (GB-LCM-022) alleles (average, six alleles). The frequency of major alleles per locus varied from 0.254 (GB-LCM-167) to 0.959 (GB-LCM-092). The allelic frequency database showed that rare alleles (frequency < 0.05) comprised 63.9% of all alleles, whereas intermediate (frequency of 0.05–0.50) and abundant alleles (frequency > 0.50) comprised 23.1 and 13.0% of all detected alleles, respectively. These results indicated the presence of a relatively large proportion of rare alleles, and most alleles were at a low frequency among the L. chinense accessions studied (Fig. 1). The high frequency of rare alleles (36.3%) among L. chinense accessions (especially among Korean accessions) indicates that they make a greater contribution to the overall genetic diversity of the collection. Hence, it is important to include rare alleles to maximize the genetic variation in the gene bank collections and to utilize them for breeding. The values for heterozygosity ranged from 0.00 at GB-LCM-037 to 1.00 at GB-LCM-025 with an average of 0.439. The average gene diversity and PIC values were 0.3792 and 0.3296, with a range from 0.0793 (GB-LCM-092) to 0.8023 (GB-LCM-167) and from 0.0775 (GB-LCM-092) to 0.7734 (GB-LCM-167), respectively.

Effective conservation and management strategies for L. chinense accessions require a fundamental understanding of their population structure. The model-based clustering method was performed using all 139 accessions and 18 SSR markers [20]. At this level, individual proportions of membership in each group, estimated using the multi-allele data set, suggested the existence of some population structure. Estimated likelihood values for a given K in five independent runs yielded consistent results, but the distribution of L(K) did not show a clear mode for the true K (Fig. 2) due to expected behaviour when factors such as inbreeding and departures from Hardy–Weinberg equilibrium are present [26]. These factors could lead to an overestimation of the number of K populations. Thus, another ad hoc quantity (ΔK) was used to overcome the difficulty of interpreting the real K values [24]. ΔK was developed and tested under different simulation routines in which real population structure was present. ΔK showed a clear peak at the true value of K. In this study, the highest value of ΔK for the 139 accessions was K = 3 (Fig. 3), which was consistent with clustering based on the genetic distance (Fig. 4), so we choose a value of K = 3 for the final analysis. The relatively small value of the alpha parameter (α = 0.099) indicates that most accessions originated from one primary ancestor, with a few admixed individuals [26]. As shown in Fig. 4, most of the accessions were clearly classified into one of the three subpopulations (S1–S3) including 65, 51, and 23 L. chinense accessions, respectively (Table 3). S1 consisted of 65 accessions, originating from three different countries but predominantly from Korea (60) and China (4). S2, with 51 accessions, consisted predominantly of Korean accessions (49), whereas the remaining accessions were from China (2). S3 consisted of 23 accessions, predominantly from China (11) and Korea (11) (Table 1). In addition to the accessions that were clearly assigned to a single population, i.e., greater than 70% of their inferred ancestry was derived from one of the model-based populations, 10 accessions (8.2%) in the sample were categorized with admixed ancestry (Fig. 4).