Sophora koreensis Nakai (Fabaceae), endemic to the Korean Peninsula, is a deciduous broad-leaved small shrub (less than 1 m tall) narrowly distributed on lower mountain hillsides (altitude ca. 150–650 m above sea level). The species reproduces both sexually and vegetatively via rhizomes. Until recently it was included in its own genus, Echinosophora Nakai (Lee, 1996; Lee et al., 2004). In North Korea (the Democratic People’s Republic of Korea, hereafter the “DRPK”), about 11 localities are known in one city and four counties (Ra et al., 2005; Chang et al., 2016; Fig. 1), whereas in South Korea (the Republic of Korea, hereafter the “ROK”), up to 16 localities have been found to date in one city and four counties, all in Kangwon Province (Cheon et al., 2009, and the present study; Fig. 1). In 1992 the ROK government designated two of the localities (HAN, Hanjeon Rural village, and YIM, Imdang Rural village; Fig. 1) as National Monument of ROK No. 372, in order to preserve and protect those areas (Yim, 1993).

In ROK, the number of shoots in each population is on the order of tens to hundreds (Kim et al., 2016; J. M. Chung, pers. comm.). Owing to the species’ limited distribution and endemic nature, S. koreensis has been assessed by several authors and authorities in ROK following the criteria of the IUCN (International Union for Conservation of Nature) Red List Categories (IUCN, 2001). However, the survey results were inconsistent: “Vulnerable” [VU B2ab (Chang et al., 2005); VU B2ab(iii, iv) (MOE, 2012)] or “Endangered” [EN B1ab(iii)+2ab(iii) (Chang et al., 2016; Kim et al., 2016)]. Sophora koreensis had been included in ROK’s official listing of special protection of endangered wild plants (the 2nd grade) until 2012 (MOE, 2012), when it was delisted without providing any scientific justification (Kim et al., 2016). For further information on ecology and conservation status, we suggest referring Kim et al. (2016).

Given its threatened status and the lack of current conservation measures (Kim et al., 2016), it is necessary to implement appropriate conservation and management strategies to ensure the long-term maintenance of genetic diversity in S. koreensis. It has been generally agreed that demographic and genetic information of endangered plant species is essential for formulating comprehensive conservation strategies (Falk & Holsinger, 1991; Hamrick & Godt, 1996). Although several short-term ecological and demographic surveys have been conducted for S. koreensis in ROK (Cheon et al., 2009; Oh et al., 2009), levels of genetic diversity at regional (e.g. ROK) or species levels are currently not well understood. For the latter case, it might be nearly impossible because of practical unavailability of samples from DPRK. The only genetic study available to date, apart from the isolation and characterization of 13 microsatellite loci for the species (Lee et al., 2012), is a fine-scale genetic structure (FSGS) survey using ISSR in two populations (that were nearly the same as HAN-2 and YIM-1; Fig. 1) in Yanggu County; a strong genet-based FSGS was found (most clones’ aggregation at less than 6-m interplant distance), presumably due to limited seed and pollen dispersal (Chung et al., 2006), together with a moderate degree of genetic differentiation between the two populations (Φ _ST = 0.184).

In this study, we expand the previous study of Chung et al. (2006) to investigate levels and distribution of genetic variation at the landscape level (in a range of ca. 6 km) in a part of Yanggu County (Fig. 1). With this spatial scale, we are especially interested in getting insights into the degree of genetic differentiation between adjacent populations within each locality (e.g. Williams, 1994; Williams & Guries, 1994; Chung et al., 2002). In general, marginal populations of plant species exhibit significantly reduced within-population genetic variability and increased among-population differentiation relative to geographically central populations (reviewed in Eckert et al., 2008). This might be the consequence of lower gene flow and higher random genetic drift in small peripheral populations (Schnabel & Hamrick, 1990; Gibson & Hamrick, 1991). Given the fact that populations in ROK are within the southernmost limit of the species distribution, we would expect low levels of within-population genetic variation and a moderate degree of genetic differentiation. With the genetic data obtained we are providing suggestions for in situ and ex situ conservation strategies.

Study species

Sophora koreensis is a pioneer species that can potentially invade open, sunny forest gaps (Yim, 1993). The study species is rapidly spreading by the underground rhizomes in disturbed areas (e.g. burnt areas). For example, we identified a 180-cm long rhizome in HAN-2 (Fig. 1) in May 2004, about three years and half after a fire affected the whole locality. Density of individuals in HAN locality was relatively high until 2009 (Cheon et al., 2009), when it started to decrease due to more shade conditions produced by the growth of broad-leaved deciduous trees such as oaks and ashes (J. M. Chung, pers. comm.). In May, 10 to 20 yellow flowers (ca. 3–5 cm long) per inflorescence bloom and are frequently visited by the bumblebee Bombus diversus diversus (Chung et al., 2006). The species is self-compatible under greenhouse conditions (Chung et al., 2006). Fruits (legumes, ca. 7 cm long) ripen from July to September and contain two or three seeds. The drying of the pods makes them shrink and seeds fall around the parent plant (“shattering”; Quick et al., 2017). There are apparently no specialized mechanisms of seed dispersal.

Population sampling

To determine levels of allozyme diversity in S. koreensis, in May 2004 we collected one leaf from 397 shoots belonging to 10 populations, that are distributed among four different localities (Fig. 1 and Table 1): three populations in HAN (ca. 80 × 100 m area) and YIM (ca. 80 × 150 m area) localities, and two populations in Dosa Rural village (DOS; 40 × 50 m area; Fig. 1) and in Deokgoek Rural village (DGR; 50 × 60 m area). Sample sizes ranged from 78 in HAN-2 to 12 in DGR-2 (TABLA1). To prevent sampling of clonal ramets, we collected shoots at ca. 4 m intervals across each population. All sampled leaf material was kept on ice until it could be transported to the laboratory, where it was stored at 4°C until protein extraction.

¹ Abbreviations: N: sample size; NP: number of populations surveyed; %P: percentage of polymorphic loci; AR: mean allelic richness (adjusted for a sample size of 12); A: mean number of alleles per locus; H _o : observed heterozygosity; H _e : H–W expected heterozygosity or Nei’s gene diversity; SE: standard error; F: fixation index; P: level of significance (or P-value); na: not applicable.

² Grand mean (F _IS ) following Weir & Cockerham (1984).

^{3, 4} data from Liu et al. (2006) and Crawford et al. (2001), respectively.

Enzyme electrophoresis

Using a phosphate-polyvinylpyrrolidone buffer (Mitton et al., 1979), we extracted enzymes and conducted horizontal starch-gel electrophoresis using 13% gels. We resolved 14 putative loci from eight enzyme systems using two buffer systems (Table 2). Stain recipes were taken from Soltis et al. (1983), except for diaphorase (Cheliak & Pitel, 1984). We designated putative loci sequentially, with the most anodally migrating isozyme designated as 1, the next 2, and so on. We also designated different alleles per locus sequentially by alphabetical order (a, b).

¹ Buffer systems 6 and MC (morpholine-citrate system, pH 6.1) were from Soltis et al. (1983) and Clayton & Tretiak (1972), respectively.

Data analyses

To estimate genetic diversity and genetic structure, a locus was considered polymorphic if two or more alleles were observed, regardless of their frequencies. The following genetic diversity parameters were estimated using the programs POPGENE (Yeh et al., 1999) and FSTAT (Goudet, 2002): percent polymorphic loci (%P), mean number of alleles per locus (A), allelic richness (AR; El Mousadik & Petit, 1996) corrected by minimum sample size (N = 12 in DGR-2; Table 1); observed heterozygosity (H _o ), and Nei’s (1978) unbiased gene diversity (H _e ).

We calculated mean fixation indices (F) across polymorphic loci for each population, and their levels of significance (or P values) by gene permutation tests (999 replicates) assuming the null hypothesis (F = 0) is true using the program SPAGeDi (Hardy & Vekemans, 2002). We estimated Wright’s (1965) F _IS and F _ST following the method of Weir & Cockerham (1984) to measure the average level (grand mean) of inbreeding within populations and genetic differentiation among them, with the aid of FSTAT. We estimated the 95% confidence intervals (CI) by bootstrapping over loci (999 replicates) around mean of the F-statistics, and considered the observed F-statistics to be significant when 95% CI did not overlap zero. We tested differences between the two population groups (HAN and YIM) for observed statistics, OSx (AR, H _o , H _e , F _IS , and F _ST ). We chose these two localities because they are national monuments; thus, any differences in genetic parameters might be informative for conservation plans. To do this, we used a permutation scheme (999 replicates) by randomly allocating whole samples to the different groups, keeping the number of samples in each group constant and calculating differences between the two groups for randomized statistics, RSx. We then obtained the P value of the test as the proportion of randomized data sets giving a larger RSx than OSx. These calculations were also performed using FSTAT (Goudet, 2002).

We conducted an analysis of molecular variance (AMOVA) using the program GenAlEx v6.5 (Peakall & Smouse, 2006) to test for the influence of individuals within populations, populations within localities, and localities on the observed genetic variation. In addition, we constructed an UPGMA (unweighted pair-group method using arithmetic averages) phenogram from Nei et al.’s (1983) genetic distance matrix with branch support produced by 1000 bootstrapping over loci, utilizing Populations v1.2.30 (Langella, 1999) and TreeView v1.6 (Page, 1996). We conducted a linear regression analysis between all pairwise F _ST /(1 – F _ST ) (F _ST was calculated following Weir & Cockerham, 1984) and the corresponding logarithm (ln) of pairwise geographical distances (Rousset, 1997) to find the overall pattern of genetic structure at the regional scale (i.e. isolation-by-distance effects). Using FSTAT, we tested a linear regression model using a Mantel test (by making 999 replicates) under the null hypothesis of no spatial genetic structure (regression slope, b = 0).

Genetic diversity and fixation index

Of the 14 putative loci surveyed, only five (Dia, Fe-1, Fe-2, 6Pgd-1, and Tpi; Table 3) were polymorphic, resulting in low levels of allozyme variation (Table 1). At the population level, the mean percentage of polymorphic loci (%P) was 13.6, the mean number of alleles per locus (A) was 1.14, and the mean genetic diversity (H _e ) was 0.026 (Table 1). Higher levels were found when the samples were pooled (N = 397, %P = 35.7, A = 1.36, and H _e = 0.037; Table 1). YIM-1 exhibited the highest H _e estimate, whereas no variation was found in DOS-1 and DOS-2 (Table 1).

Average F values in four populations (HAN-1, HAN-2, HAN-3, and YIM-3; Table 1) indicated a significant deficiency of heterozygotes, with a grand mean of F _IS = 0.556 (Table 1; 95% CI was from 0.016 to 0.297). Permutation test (999 replicates) revealed no significant differences for AR (P = 0.617), H _o (P = 0.568), H _e (P = 0.734), and F _IS (P = 0.340) between HAN and YIM.

Genetic differentiation among populations

Significant differentiation among populations was found in S. koreensis at the landscape level (mean F _ST = 0.203, 95% CI was from 0.032 to 0.245). Permutation test indicated no significant difference for F _ST (P = 0.813) between HAN and YIM. We unexpectedly found high (F _ST among three populations in HAN = 0.583; F _ST between two populations in DGR = 0.389) or moderate (F _ST among three populations in YIM = 0.233) degree of genetic differentiation among populations within the three localities (no allozyme variation in DOS). Consistent with this, the AMOVA analyses revealed a considerable genetic differentiation among populations within localities (17%; Table 4). This estimate was considerably higher than that of variation among the four localities (5%; Table 4).

¹ Abbreviations: df, degrees of freedom; SS, sum of squares; MS, mean squares; percent variation, percentage of total variance contributed by each component.

Despite the non-statistical support of the UPGMA phenogram, it provides a general idea about the clustering of the studied populations (Fig. 2). Most groupings showed little relationship with populations’ geographic location; a notable exception were the populations DOS-1 and DOS-2, that were placed together because of the same banding patterns (no variation). The correlation between pairwise genetic differentiation estimates and their corresponding natural ln of pairwise geographic distances was very low (r = 0.032, P = 0.842) and indicated that ca. 100% of the variation in genetic differentiation was due to unknown factors other than distance (Fig. 3).

Low levels of within-population variation and inbreeding

As expected, populations of S. koreensis exhibited low levels of genetic variation (%P = 13.6, A = 1.14, H _e = 0.026). These estimates were clearly lower than the reference values for both endemic species and narrowly-distributed ones (%P = 26.3, A = 1.39, H _e = 0.063, N = 100; %P = 30.6, A = 1.45, H _e = 0.105, N = 115; Hamrick & Godt, 1989). The population-level estimates, however, are similar to those found in its congener S. fernandeziana (Phil.) Skottsb. (%P = 19 and H _e = 0.022; Crawford et al., 2001). Sophora fernandeziana is a hummingbird-pollinated species endemic to Robinson Crusoe Island (ca. 48 km², known also as Más a Tierra or Más Atierra until 1966), a part of the Juan Fernández Archipelago, located 667 km west of South America (Chile) at 33° S latitude, in the South Pacific Ocean. Crawford et al. (2001) attributed the low levels of allozyme diversity to random genetic drift operating in the small, scattered populations. Another studied congener is the insect-pollinated S. moorcroftiana (Benth.) Benth. ex Baker from the Qinghai-Tibetan Plateau (QTP) in western China. Liu et al. (2006) examined allozyme diversity from 10 populations in the middle reaches (in a range of ~630 km) of the Yarlung Zangbo (Brahmaputra) River. As one may expect for a species with a wider geographic distribution, moderate levels of within-population genetic variation were found in S. moorcroftiana (%P = 27.5, A = 1.50, H _e = 0.122; Liu et al., 2006).

These three Sophora species would have followed their unique evolutionary trajectories. Past histories (e.g. uplift of the QTP for S. moorcroftiana and island colonization for S. fernandeziana) might have greatly contributed to levels of genetic diversity and genetic structure. The broad-leaved deciduous shrub S. koreensis is one of the elements of the mixed (boreal-temperate) forest on the Korean Peninsula. At the Last Glacial Maximum (LGM, 21,000 years before present, BP) S. koreensis might have occurred within a narrow mixed forest zone in the southern and western parts of the Peninsula (further south compared to its present distribution; see Fig. 2A in Chung et al., 2017). The onset of the Holocene (ca. 12,000 years BP) and the accompanying climatic amelioration would had led to a post-glacial recolonization process for the mixed forests towards the north (i.e. to central and northern Korea) from putative refugial population(s) in southern Korea. Under this scenario, levels of genetic diversity in the current populations of S. koreensis in central and northern Korea would be low due to multiple founder effects (e.g. Kim et al., 2005; reviewed in Chung et al., 2017). Kim et al. (2015) revealed that within-population genetic diversity (H _e ) of the draft bamboo Sasa borealis (Hack.) Makino & Shibata populations in South Korea was inversely proportional to the latitude (r = 0.719, P = 0.009). If this scenario is true for S. koreensis, we will expect that levels of within-population genetic diversity would decrease with increasing latitudes. However, a test of this hypothesis seems to be impossible because of political problems (at present it is not feasible to sample in DPRK by scientists from ROK).

The significant deficit of heterozygotes seen in the four populations (HAN-1, HAN-2, HAN-3, and YIM-3) can be attributed to the occurrence of selfing and/or biparental inbreeding, likely resulting from limited pollen and/or seed dispersal (Chung et al., 2006). Even though we tried to prevent duplicate samples, unintentional inclusion of some samples would lead to over- or underestimation of fixation indices. In addition, as individuals in some populations are patchily distributed (J. M. Chung, pers. comm.), a Wahlund effect might be involved in some of the apparent inbreeding observed. Detailed field studies are needed to evaluate the relative contribution of selfing vs. biparental inbreeding to total inbreeding.

Genetic differentiation among populations

Given the ~6 km study range, the F _ST value (0.203) is notably high. A similar result was found in ISSR-based genetic differentiation between HAN-2 and YIM-1 (Φ _ST = 0.184; Chung et al., 2006). These estimates are only slightly lower than those reported for endemic species [Nei’s (1973) G _ST = 0.248] and narrowly-distributed species (G _ST = 0.242; Hamrick & Godt, 1989). Our estimate of genetic differentiation for S. koreensis is even higher than that S. moorcroftiana populations (F _ST = 0.199; Liu et al., 2006), which is quite unexpected given the substantially larger sampling area of S. moorcroftiana (the distances between each locality range from 30 to 630 km; Liu et al., 2006). The authors attributed the moderate degree of genetic differentiation among 10 populations to the secondary seed movement by water streams (hydrochoric gene flow; note that the species occurs along relatively open river valleys). Furthermore, their findings of a significant correlation between genetic and geographic distance (r = 0.707, P = 0.002) suggest a balance between random genetic drift and gene flow. As insect movement would be limited, secondary seed dispersal via river water current might be possible. In contrast to these studies, Crawford et al. (2001) found high degree of genetic differentiation among five S. fernandeziana populations on Robinson Crusoe Island (ca. 48 km²) (G _ST = 0.351). The authors suggested that high genetic differentiation is due to the fact that its populations are small and have a scattered distribution along the island. Thus, random genetic drift would have operated in these small populations in the remote oceanic island.

Samples from the three populations (YIM-1, YIM-2, and YIM-3) in YIM were collected in an area of ca. 80 × 150 m and those from three populations (HAN-1, HAN-2, and HAN-3) in HAN were from an area of ca. 80 × 100 m. Even though the three populations are very closely located each other both in YIM and HAN, we found significant differences in allelic frequencies. For example, the allele 6Pgd-1 ^a was present in HAN-1 and YIM-1 with frequencies of 0.015 and 0.210, but it did not occur in the remaining populations (Table 3). Unexpectedly, this allele was not present in YIM-2 and YIM-3. Similarly, the allele Fe-1 ^b was present in HAN-2, YIM-1, and YIM-2 with frequencies of 0.038, 0.306, and 0.021, respectively, but not in the remaining populations of both HAN and YIM localities. These findings strongly suggest limited gene dispersal within approximately 1 ha area, which is supported by the strong FSGS detected in HAN-2 and YIM-1 (Sp, a statistic that represents the strength of FSGS, was 0.072 in HAN-2 and 0.093 in YIM-1; Chung et al., 2006). Consistent with these results, we found little association between genetic and geographic distances (Figs. 2 and 3) and a considerable genetic differentiation among populations within localities (Table 4).

Implications for conservation

Before discussing on genetic aspects, and based on available ecological information, our own field observations, and personal communications, we provide two recommendations for in situ conservation in the study area. First, we suggest performing a selective removal of trees in the natural populations, as S. koreensis prefers sunny places, open forest gaps, or disturbed areas (e.g. burnt areas); when relatively fast-growing shrubs or trees (such as oaks and ashes) invade the populations producing shady conditions, many shoots are likely to disappear (J. M. Chung, pers. comm.). Indeed, removal of trees and have been conducted in the areas of Natural Monuments No. 372 (HAN and YIM) with some success (J. M. Chung, per. comm.). However, special care should be needed because the human-assisted clearance may also allow other native trees that can grow in a similar environment to thrive. Second, for in situ protection it would be necessary to take preventive measures against damage caused by the human-mediated pressure. To do this, it seems to be preferable to install a wooden fence at the boundary of the natural populations in HAN and YIM.

The population genetic structure of S. koreensis is summarized as: (1) low levels of genetic variation within populations, (2) moderate degree of differentiation among populations and considerable genetic variation among populations within localities, (3) little association between genetic and geographic distance, (4) lower genetic variation in smaller populations (e.g. DOS), and (5) a strong FSGS in populations in HAN-2 and YIM-1 (Chung et al., 2006). In addition, as S. koreensis also reproduces vegetatively, effective population sizes will be especially small in highly clonally-formed populations. Thus, random genetic drift should have a more pronounced effect in those populations over generations (e.g. Chung et al., 2004). Given our genetic results, we recommend that all extant populations in ROK should be conserved in situ. Preserving entire localities (i.e. packs of close populations) instead of single populations seems a reasonable policy, given the fact that populations within each locality are separated by a few meters (localities range from 0.2 to 1.2 ha); preserving together packs of close populations also allows to maintain gene flow between populations. To ensure long-term ex situ genetic variability of S. koreensis, and based on the results of the FSGS study of Chung et al. (2006), we suggest to collect seeds from shoots positioned at distances of ca. 10–12 m across entire localities. It is also highly advisable to carry out a genetic study across all known 16 populations in ROK with the microsatellite loci developed by Lee et al. (2012), that have been proven to be highly polymorphic. In parallel to this genetic survey, long-term ecological studies (e.g. minimum viable population size and demographic dynamics, seed dispersal, pollination biology, seed germination ecology) are also necessary to reach a comprehensive management of S. koreen sis.