Chloroplast DNA insights into the phylogenetic position and anagenetic speciation of Phedimus takesimensis (Crassulaceae) on Ulleung and Dokdo Islands, Korea

Phedimus takesimensis (Ulleungdo flat-leaved stonecrop) is endemic to Ulleung and Dokdo Islands off the east coast of the Korean Peninsula. It was suggested that P. takesimensis originated via anagenetic speciation from the continental progenitor species P. kamtchaticus or P. aizoon. However, little is known of the phylogenetic relationships and population genetic structure among species of Phedimus in the Korean Peninsula and Ulleung/Dokdo Islands. We inferred the phylogenetic relationships among congeneric species in Korea based on nuclear ribosomal DNA internal transcribed spacer and chloroplast noncoding regions. We also sampled extensively for P. takesimensis on Ulleung Island and the continental species, P. kamtschaticus and P. aizoon, to assess the genetic consequences of anagenetic speciation. We found (1) the monophyly of P. takesimensis, (2) no apparent reduction in genetic diversity in anagenetically derived P. takesimensis compared to the continental progenitor species, (3) apparent population genetic structuring of P. takesimensis, and (4) two separate colonization events for the origin of the Dokdo Island population. This study contributes to our understanding of the genetic consequences of anagenetic speciation on Ulleung Island.


Introduction
The unique and diverse flora and fauna of the oceanic archipelagos are excellent model systems to investigate patterns and processes of evolution; thus, oceanic islands are often considered as natural laboratories [1][2][3][4][5]. In particular, numerous insular endemic plant lineages with spectacular morphological, anatomical, and ecological diversity in the Pacific and Atlantic Oceans provide ample opportunities for naturalists to uncover their fascinating natural histories on islands (e.g., the silversword alliance of the Hawaiian Islands [6,7]; Echium of the Canary Islands [8]; Dendroseris and Robinsonia of the Juan Fernandez Islands [9][10][11] woody Sonchus alliance of the Canary Islands [12]). Considering the fact that approximately 25% of the extant vascular plant species occur on islands, of which a large portion are highly vulnerable to natural and anthropogenic threats, it is critical and urgent for plant systematists to discover hidden diversity and generate base line information for the conservation of critical endemic lineages [13][14][15]. Detailed systematic and ecological studies combined with the use of high throughput genomic sequence data and a diverse array of analytical tools are essential to elucidate the complex evolutionary history and facilitate the conservation of numerous endemic plant lineages that are shaped by different evolutionary processes (e.g., cladogenesis, anagenesis, and reticulation) [16]. Ulleung Island, an oceanic volcanic island in the East Sea, is located approximately 137 km east of the Korean Peninsula and has never been connected to the adjacent continental land mass. It is estimated to be approximately 1.8 million years old (myr) with a total area of 73 km 2 and a highest peak of 984 m above sea level [17]. Albeit small, Ulleung Island is home to approximately 500 native vascular plant species, approximately 36 of which are endemic species primarily driven via anagenetic speciation. During anagenetic speciation, an initial founder lineage simply transforms genetically and morphologically through time without further specific differentiation [18,19]. The diversity estimates based on the endemic plant species per 100 km 2 and the number of plant species per km 2 are much higher than those of the Hawaiian and Canary Islands [20,21]. Given the important role of anagenetic speciation (also known as phyletic or gradual speciation; see also anacladogenesis, [22]) for explaining the origin of the majority of plant endemics, Ulleung Island has been recognized as an ideal place to investigate this particular mode of speciation among other oceanic archipelagos.
Of the nearly 40 currently recognized endemic plant lineages on Ulleung Island that presumably have their origins in anagenetic speciation, Phedimus takesimensis (Nakai) 't Hart (= Sedum takesimense Nakai) represents an endemic species in the family Crassulaceae of the core eudicots clade. The genus Phedimus, with around 20 species primarily in Asia and Europe, represents one lineage of non-monophyletic subfamily Sedoideae [23]. Traditionally, species of Phedimus have been treated as members of Sedum, but a recent phylogenetic study strongly supported the monophyly of Phedimus and its segregation from the genus Sedum [24]. Phedimus takesimensis occurs exclusively on the seaside and sunny slopes of Ulleung and Dokdo Islands in Korea. It is a robust perennial herb with overwintering shoots on the lower part of stem, erect or decumbent with a height of 20-50 cm [21]. Phedimus takesimensis is morphologically similar to P. aizoon (L.) 't Hart, but the former species can be distinguished from the latter primarily based on overwintering shoots on the lower part of the stems and much shorter internodes, giving the appearance of tightly clustered leaves ( [25]; S-C. Kim, personal observation). In addition, P. aizoon does not branch much and has yellow anthers, while P. takesimensis is much more branched, like P. kamtschaticus, with reddish yellow anthers [26]. Phedimus kamtschaticus, which occurs commonly in northeastern Asia, is morphologically similar to P. aizoon, with intermediate morphological forms between the two species, but it can be distinguished based on the presence of more branched stems. Given the morphological similarities among these taxa and significant variation within each taxon, there is substantial confusion and numerous taxonomic classifications abound. For example, in addition to the newly described species Sedum zokuriense Nakai [= Phedimus zokuriensis (Nakai) 't Hart] from Mt. Seokrisan [27], S. takesimense was recognized as infraspecific taxon of S. kamtschaticum, i.e., S. kamtschaticum var. takesimense and S. kamtschaticum var. zokuriensis [28]. However, a distinct species status was proposed based on floristic and taxonomic studies [26,29]. Several infraspecific taxa (up to seven varieties) within S. aizoon were recognized (see the various treatments described in [26]), while Lee [30] recognized only one taxon. In addition, a morphometric study of the subgenus Aizoon in Korea suggested that although they formed distinct clusters within S. kamtschaticum in a broad sense, the three species S. takesimense, S. zokuriense, and S. ellacombianum [= Phedimus ellacombianus (Praeger) 't Hart] could be recognized as infraspecific taxa of S. kamtschaticum [31]. Interspecific gene flows and hybridizations among the species of the subgenus Aizoon in Korea (i.e., S. latiovalifolium, S. kamtschaticum, and S. aizoon) were also documented [32]. Furthermore, extensive polyploidy and aneuploidy in members of the subgenus Aizoon were demonstrated, partly explaining a certain degree of morphological variation [33][34][35].
Considering the origin of endemic Phedimus takesimensis on Ulleung and Dokdo Islands in the East Sea, few continental species could be proposed as likely candidate progenitor species. Based on a morphological analysis, the two species S. kamtschaticum [= P. kamtschaticus] and S. ellacombianum [= P. ellacombianus] were clustered with S. takesimense, while S. aizoon [= P. aizoon] formed a distinct and separate cluster from the S. kamtschaticum group [31]. While P. kamtschaticus occurs widely in the Korean Peninsula and neighboring regions (eastern China, Russia, and Japan), P. ellacombianus is restricted to the southern part of the Korean Peninsula [26]. It was shown, based on very limited sampling, that P. aizoon is closely related to P. takesimensis genetically based on RAPDs (randomly amplified polymorphic DNAs) [36]. The first assessment of the genetic diversity of P. takesimensis and its relationship to a continental progenitor was based on one noncoding chloroplast (cp) region (trnL-F) [37]. Based on a total of 32 individuals (30 individuals of P. takesimensis and two individuals of P. kamtschaticus), two chlorotypes were found; "type01" was found exclusively on Ulleung Island (15 individuals), while "type02" was found in 17 individuals from the Ulleung and Dokdo Islands. In addition, two individuals of P. kamtschaticus-as a continental progenitor-were similar to the chlorotype "type02" in that they had a 6-bp deletion but were distinguished by a single point mutation [37]. Recently, a Phedimus cultivar identification study, including several purported wild progenitor species, demonstrated unresolved species relationships within the genus [38]. For example, it was shown that the highly polymorphic P. takesimensis was unresolved with P. middendorffianus and P. kamtschaticus based on nuclear ribosomal DNA (nrDNA) internal transcribed spacer (ITS) sequences, while one cp noncoding region (psbA-trnH) sequence suggested an unresolved relationship with P. aizoon, P. kamtschaticus, and P. middendorffianus. It seems likely that this lack of clear species relationships could be attributed to various factors, such as polyploidization, aneuploidization, interspecific gene flow, concerted evolution of nrDNA, incomplete lineage sorting of ancient polymorphisms, lack of sufficient cpDNA variation, and any combination of these. While precise species relationships within the genus remain to be determined, it is reasonable to consider that P. kamtschaticus is one of plausible continental progenitor species of P. takesimensis on Ulleung Island based on morphological characteristics.
In this study, we sampled extensively P. takesimensis on Ulleung and Dokdo Islands and estimated their genetic variation and population structure based on five cpDNA noncoding regions. Then, we compared the genetic diversity and population genetic structure of P. takesimensis with two representative continental progenitor species (P. kamtschaticus and P. aizoon). Lastly, we took a phylogenetic approach to determine the phylogenetic position of P. takesimensis among the congeneric species occurring in Korea. Based on extensive sampling and the combination of population genetics and phylogenetic approaches, we hope to (1) estimate the level of genetic variation between insular endemic and purported continental progenitor species, (2) reveal the population genetic structure of P. takesimensis on Ulleung and Dokdo Islands, (3) determine the relationship between Ulleung and Dokdo Island populations of P. takesimensis, and (4) infer the relationship between insular derived endemic P. takesimensis and continental congeneric species.

Plant materials
For the phylogenetic analysis of the genus Phedimus in Korea, we sampled nine accessions of P. aizoon, 10 accessions of P. kamtschaticus, 15 accessions of P. takesimensis, and two accessions of P. ellacombianus (Fig 1 and Table 1). In the case of P. kamtschaticus, we sampled one accession from a representative locality in the Korean Peninsula: Gangwon-do Province (three locations: Mt. Hambaek, Sokcho-si, and Cheorwon-gun), Gyeonggi-do Province (two locations: Mt. Ungil and Mt. Munsu), Chungcheongnam-do Province (one location: Taean-gun), Gyeongsanbuk-do Province (one location: Pohang-si), Jeollanam-do Province (one location: Mudeungsan National Park), and Jeollabuk-do Province (two locations: Mt. Deogyu and Mt. Jiri). For P. latiovalifolius, which is endemic to Korea, we sampled a total of 10 accessions from Gangwon-do Province (Taebaek-si) and one accession from the same 15 populations sampled for the population genetic study for P. takesimensis (Table 1). No permits were required to collect the species in Korea given their conservation status. For the outgroups, we selected Rhodiola and Sedum sensu stricto, and DNA sequences of the representative species were obtained from GenBank ( Table 1). The population codes for the sampling locations of each species are given in Table 1.
For the population-level study of the continental progenitor and insular derivative species pairs, we sampled a total of two populations for P. aizoon (Table 1). In the case of the continental progenitor species P. kamtschaticus, we sampled a total of 10 populations (eight accessions per population, except for SAM) from Korea. Phedimus kamtschaticus occurs widely in East Asia, including eastern China, Russia, and Japan, but population-level sampling was not possible for those regions. The 10 populations from the Korean Peninsula represent the typical distribution range of P. kamtschaticus, which include the northeastern most population (YON), the southernmost population (MUD), and the westernmost population (AN), enabling us to assess their relationships to P. takesimensis (Fig 2 and Table 1). For the insular derived P. takesimensis species, which occurs exclusively on the seaside and sunny slopes of Ulleung Island and Dokdo Island, we sampled a total of 15 populations and 148 accessions; 14 populations were from Ulleung Island and one was from Dokdo Island, which is located about 87 km east of Ulleung Island. We sampled a total of 10 accessions per population (except for HYE with eight accessions), representing the northern region (HYE, CHU, and CHE), eastern region (SUM, NAE, JEO, DOD, and BON), southern region (DON, SAD, and TON), and southwestern/western region (NAM, GUA, HAK, and TEA) (Fig 2 and Table 1). Voucher specimens for the representative individuals of each population were deposited in the Ha Eun Herbarium (SKK), Sungkyunkwan University (Table 1).

DNA isolation, amplification, and sequencing
Silica-gel dried leaves collected from the natural populations were used as DNA sources for the DNA extraction using DNeasy Plant mini kit (Qiagen, Carlsbad, California, USA) following the standard protocol. We screened 21 noncoding regions of chloroplast DNA and selected five variable noncoding regions (i.e., atpI-atpH, trnC (GCA) -ycf6, trnL-trnF, 3'rps16-5'trnK (UUU) , and ndhJ-trnF (GAA) ) [39]. We also amplified the nrDNA ITS region using the same primers described previously [40,41]. Polymerase chain reaction (PCR) amplification was conducted using the standard protocol of the Inclone Taq DNA polymerase kit (Inclone Biotech Co., Yongin, Korea) with a final volume of 50 μl consisting of 1 μl purified DNA, 5 μl reaction buffer, 1 μl dNTPs, 1 μl of each universal forward and reverse primer, 0.25 μl Taq polymerase, and 40.75 μl of ddH 2 O. The following PCR conditions were used for the chloroplast DNA regions: an initial denaturation of 95˚C for 2 min, 35 cycles of 95˚C denaturation for 1 min, 50-56˚C annealing for 2 min, 72˚C extension for 2 min, and a final extension at 72˚C for 10 min. The following were the nrDNA ITS region reaction conditions: an initial denaturation of 94˚C for 1 min, 35 cycles of 94˚C denaturation for 1 min, 54˚C annealing for 2 min, 72˚C extension for 2 min, and a final extension at 72˚C for 10 min. The PCR products were visualized using electrophoresis on a 1% agarose gel, purified using an Inclone Gel and PCR purification kit (Inclone Biotech Co., Yongin, Korea), and sequenced using Big Dye Terminator Cycle Sequencing reagents (Applied Biosystems, Forest City, CA, USA) at Geno Tech. Corp. (Daejeon, Korea). For the sequencing reactions, we used the same primers as those used for the PCR.

Phylogenetic analysis, network construction, and population structure
All sequences were edited and assembled using Sequencher 4.2.2 (Gene Codes, Ann Arbor, MI, USA) and Geneious Pro v R.8 (Biomatters, Ltd, New Zealand), and were deposited at Gen-Bank with the accession numbers MK752203-MK752392. The sequences were aligned using Clustal X version 1.83 [42], with a final manual adjustment using MacClade [43]. Gaps were coded as simple binary characters [44] using the program SeqState 1.4.1 [45]. To determine the species relationships among Phedimus in Korea, we conducted a separate phylogenetic analysis for the nrDNA ITS and four chloroplast noncoding regions (trnC-ycf6, atpI-atpH, trnL-trnF, and ndhJ-trnF). The maximum likelihood (ML) analysis was conducted using IQ-TREE v. 1.4.2. [46], with 1,000 replicate bootstrap (BS) analyses, based on the best-fit model for each data set: "TVM+F+G4" for combined cpDNA and "TIM3e+G4" for nrDNA ITS.
For the population level analysis of insular derived P. takesimensis and the purported continental progenitors P. kamtschaticus and P. aizoon in Korea, we concatenated four previously used chloroplast noncoding regions and one more highly variable region (rps16-trnK) to achieve more resolutions. We constructed a haplotype network using TCS version 1.21 [47]. Gaps were treated as missing data and the connection limit excluding homoplastic changes was set to 95%, in accordance with Hart and Sunday [48]. To determine the overall relationships among accessions of the three species, we conducted ML analysis using IQ-TREE v. 1.4.2. [46] with the best-fit model of "K3Pu+F+I" and an unrooted tree was drawn. The pairwise distance based on the Kimura 2-parameter method [49] was also calculated using PAUP � 4.0b10 [50].
Genetic variation between taxa, and among and within populations was evaluated by an AMOVA (analysis of molecular variance) using ARLEQUIN version 3.5 [51]. The AMOVA analyses were performed using four categories, (a) all samples from the three species, (b) the insular derivative species P. takesimensis, (c) the continental progenitor species P. aizoon, and (d) the continental progenitor species P. kamtschaticus. The significance levels of the AMOVA were evaluated using 1,023 permutations. A SAMOVA (spatial analysis of molecular variance; [52]) was used to characterize patterns of genetic structure across the species distribution. For the demographic history, mismatch distributions were calculated and tested against sudden demographic expansion [53] using Arlequin 3.5 [51] and DnaSP v. 5 [54]. Unimodal patterns are expected for recent sudden population expansions, while multimodal distributions are suggestive of demographic stability or multiple colonizations [55]. Tajima's D [56] and Fu's F S [57] were calculated to test for evidence of range expansion. To validate
We also assessed the relationships among three species, two purported continental progenitor species (P. aizoon and P. kamtschaticus) and insular derived P. takesimensis. The unrooted ML tree (Fig 4) showed that the three species all formed strongly supported groups; 99% BS for P. takesimensis, 100% for P. aizoon, and 93% for P. kamtschaticus. We also calculated the pairwise distance based on the Kimura 2-parameter method [49] within each species and among species. The average pairwise distance within P. takesimensis was 0.000878 ± 0.000564 (± standard deviation, SD), while that of P. aizoon and P. kamtschaticus was 0.000617 ± 0.0005 and 0.00174 ± 0.000131, respectively. In the case of the pairwise distances among species, between P. takesimensis and P. aizoon it was 0.003231 ± 0.0006, between P. takesimensis and P. kamtschaticus it was 0.003477 ± 0.0006, and between P. aizoon and P. kamtschaticus it was 0.00391 ± 0.0006.

Genetic structure and demographic change
Although only two populations were sampled for P. aizoon, we estimated genetic variation among and within the populations by AMOVA. We conducted the AMOVA using four  categories, (a) all samples from the three species, (b) P. takesimensis, (c) P. aizoon, and (d) P. kamtschaticus. When all samples from the three species were analyzed, 62% of the variation was found among the three taxa, followed by among populations (27.9%) and within populations (9.78%) ( Table 4). In the case of continental progenitor species P. kamtschaticus, most of the variation was found among populations (90.84%), with 9.16% of the variation within populations. In contrast, 77% of the variation was found within populations and approximately 23% among populations in P. aizoon. In the case of the anagenetically derived insular P. takesimensis, around 70% of the variation was found among populations and 30% was found within populations.
Mismatch distributions showed that two species, P. takesimensis and P. kamtschaticus, had multimodal distributions, suggesting constant population sizes, multiple colonization, and/or sustained subdivision for a long period of time (Fig 6). In addition, Tajima's D and Fu's F S , and Ramos-Onisins and Rozas' R 2 failed to detect significant population expansion in the two species (Table 5). In the case of the SAMOVA analysis, we detected seven groups (F CT = 0.79116) as the optimal number of genetic groups (K) based on spatial locations and cpDNA haplotypes ( Table 6). The first four SAMOVA groups (Groups 1-4) included populations of P. takesimensis on Ulleung and Dokdo Islands, whereas the remaining three groups exclusively occurred on the Korean Peninsula (Fig 7). Group 1 contained five populations of P. takesimensis, four (CHU, GUA, NAM, and BON) on Ulleung Island and one (DON) on Dokdo Island Table 4. Summary of the analysis of molecular variance (AMOVA) for genetic differences between three Phedimus species (P. takesimensis, P. aizoon, and P. kamtschaticus).

Taxon
Source of variation Degrees of freedom (d.f.), sum of squares (SS), variance components and the percentage of variation (%) and its associated significance (n = 1,023 permutations) for the analyses are shown.

Discussion
In this study, we provide the first convincing evidence for the monophyly of P. takesimensis on Ulleung and Dokdo Islands, despite its morphological variations within the islands and geographical proximity to continental source areas [26]. Based on the extensive sampling of the islands and using chloroplast and nuclear DNA, there is conclusive evidence that a single colonization event is responsible for the morphological and molecular diversity of P. takesimensis on the islands (Fig 3). In addition, both nuclear (ITS; 95% BS) and chloroplast (four noncoding   [60,61]) and Scrophularia takesimensis (Scrophulariaceae; [62]). However, the single origin of P. takesimensis is in line with other anagenetically derived endemics, such as Dystaenia takesimana (Apiaceae; [63]), Hepatica maxima (Ranunculaceae; [64]), Acer okamotoanum (Sapindaceae; [65]), A. takesimense (Sapindaceae; [66]), Fagus multinervus (Fagaceae; [67]), and Campanula takesimana (Campanulaceae; [68]). Therefore, despite the geographical proximity of Ulleung Island to possible source areas, a single origin for anagenetically originated endemics is the norm, with few exceptions. Although we established the single origin of P. takesimensis on the Ulleung and Dokdo Islands, the closest sister species on the continents and geographical source areas are yet to be determined. While the nuclear ITS phylogeny suggested a sister relationship of monophyletic P. takesimensis to the clade containing P. kamtschaticus, P. ellacombianus, P. latiovalifolius, and P. aizoon. (Fig 3A), the cpDNA phylogeny suggested that P. takesimensis was unresolved with species of P. aizoon and P. latiovalifolius (Fig 3B). This unresolved relationship was also suggested based on nearly equal genetic distances of P. takesimensis to either P. aizoon or P. kamtschaticus (Fig 4) and the haplotype network (Fig 5). Although full picture of Phedimus phylogeny is yet to be revealed, it seems highly likely that the conflicts between nuclear and plastid phylogeny are due to hybridization and introgression among the sampled species in the current study. In addition, aneuploidization and polyploidization, which are known to occur commonly in genus Phedimus and related genera, likely contributed to such phylogenetic incongruences [32][33][34][35]. The important role of hybridization and aneuploidization/polyploidization in the evolution of Phedimus and related genera is yet to be investigated. It is not feasible to identify source areas for reciprocally monophyletic continental progenitors and insular derived species pairs (e.g., P. kamtschaticus/P. aizoon with P. takesimensis, A. pseudosieboldianum with A. takesimense, A. mono with A. okamotoanum, Dystaenia ibukiensis with D. takesimana, and Fagus japonica/F. engleriana with F. multivervis). However, given that the monophyletic insular endemics were deeply embedded within the paraphyletic continental species relationships, it was possible to determine the continental source populations for certain taxa, e.g., the origin of Hepatica maxima from the eastern Korean Peninsula population of H. asiatica [64] and the origin of Campanula takesimana from the southeastern Korean Peninsula population of C. punctata [68]. In the case of multiple origins of Rubus takesimensis, it was shown that the northern mainland Korean Peninsula and southern Korean Peninsula/ Japanese archipelago populations were responsible for its origin on Ulleung Island [61]. As we described in the Introduction, P. kamtschaticus is one of most likely continental progenitor species of anagenetically derived endemic P. takesimensis on Ulleung Island. Therefore, this study also provided an opportunity to assess the genetic consequences of anagenetic speciation [69,70]. Based on the emerging genetic patterns for anagenetically derived species on islands, we expected no significant reduction in genetic diversity compared to continental progenitor species and a lack of strong genetic differentiation for P. takesimensis on the island. Although our sampling between the two species pairs was uneven, it seems there is no apparent reduction in genetic diversity in the insular derived P. takesimensis. For example, P. takesimensis showed much higher gene diversity (0.9346) and nucleotide diversity (0.003290) compared to P. kamtchaticus (gene diversity of 0.7699 and nucleotide diversity of 0.00197)  (Table 3). In addition, unlike the highly monomorphic haplotypes that exist within the population of the continental progenitor P. kamtchaticus, we found highly diverse haplotypes in P. takesimensis on Ulleung Island, with 30 and 7 haplotypes for P. takesimensis and P. kamtchaticus, respectively (Table 3 and Fig 5). These results, however, should be interpreted cautiously given the uneven sampling of progenitor and derivative species pairs. Nevertheless, these results are consistent with previously studied taxa, such as Rubus takesimensis [61], Acer takesimense and A. okamotoanum [65,66], and Dystaenia takesimana [63].
We also found that, based on the same molecular marker type of chloroplast DNA and the AMOVA results, a significant portion of the genetic variation existed between taxa (62%), suggesting genetic divergence among the three species (Table 4). Furthermore, genetic variations appeared to be partitioned primarily into geographical regions (SAMOVA results ; Fig 7), although around 30% of the variation still existed within populations of P. takesimensis. When we compared these results to other endemics on Ulleung Island, we found 43% of the variation in Rubus takesimensis was within populations [61] and 48% was within populations in Campanula takesimana [68]. Therefore, P. takesimensis showed apparent population genetic structure and differentiation after its divergence from the continental progenitor species. Given the fruit type of P. takesimensis are aggregates of follicles, with the fruit opening along an upward facing suture (i.e., cup-shaped), and with small light-weighted seeds, it is highly conceivable that raindrops are responsible for splashing out and flushing away seeds over some distances [71][72][73]. This type of splash rain dispersal mechanism was documented for Sedum formosanum and S. subtile from the same family (Crassulaceae) in Japan and it was shown experimentally that the maximum seed dispersal distance of S. subtile was 63.7 ± 7.8 cm [73]. Therefore, it is highly likely that splash seed dispersal by raindrops is responsible for the limited seedmediated gene flow of populations of P. takesimensis, which have apparent population genetic structure and differentiation.
In this study, we also demonstrated that the population of P. takesimensis on eastern Dokdo Island (i.e., the eastern islet) most likely originated from two separate populations, one in southwestern Ulleung Island (GUA and NAM) and the other in southeastern Ulleung Island (JEO and DOD) (Fig 2). Dokdo Island is located approximately 90 km away from the east of Ulleung Island, with an estimated age of 4.6-2.5 myr, and is much older than Ulleung Island (approximately 2 myr) [17,74]. One common haplotype found on Dokdo Island, T29, was also found in two southwestern parts of Ulleung Island, while one low frequency haplotype on Dokdo Island, T4, was also found in the southeastern part of Ulleung Island with a relatively high frequency, especially in JEO (nine accessions with the T4 haplotype). One rare haplotype, T30, was found exclusively on Dokdo Island and given its haplotype network relationship ( Fig  5), it is highly likely that this unique haplotype originated after colonization from Ulleung Island, especially from the southwestern haplotype (T29). However, we cannot completely rule out the possibility of sampling error for the lack of this haplotype on Ulleung Island. In the case of C. takesimana, which occurs on both Ulleung and Dokdo (western islet) Islands, we demonstrated that Dokdo Island was the steppingstone island for its origin on Ulleung Island [68].
In conclusion, we provide the first convincing evidence, based on extensive sampling, that P. takesimensis on Ulleung Island is a strongly supported monophyletic group that is closely related to P. aizoon and P. kamtschaticus in the Korean Peninsula. We also found that P. takesimensis on Ulleung Island and P. aizoon on the peninsula harbor diverse haplotypes, while P. kamtschaticus shows highly monomorphic haplotypes. The overall genetic diversity parameters also suggested that insular derived P. takesimensis maintains a high genetic diversity compared to its continental progenitor species. Unlike other anagenetically originated endemics, P. takesimensis showed population genetic structuring and we suggest that limited seed-mediated gene flow via a splash rain dispersal mechanism is responsible for this genetic pattern. Lastly, we demonstrated that the Dokdo Island was colonized by two separate gene pools of P. takesimensis (southwestern and southeastern) from Ulleung Island.