The speciation history of northern‐ and southern‐sourced Eranthis (Ranunculaceae) species on the Korean peninsula and surrounding areas

Abstract The temporal and spatial origins and evolution of the genus Eranthis have not been previously studied. We investigated the speciation and establishment histories of four Eranthis species: Eranthis byunsanensis, E. pungdoensis, E. stellata, and E. pinnatifida. The sampling localities were Korea, Japan, Jilin in China, and the area near Vladivostok in Primorskiy, Russia. We used 12 chloroplast microsatellite loci (n = 935 individuals) and two chloroplast noncoding regions (rpl16 intron, petL‐psbE intergenic spacer; n = 33 individuals). The genetic diversity, genetic structure, phylogenetic relationships of the four species were analyzed, and their ancestral areas were reconstructed. The high genetic diversity of the Jeju island population of E. byunsanensis and Russian populations of E. stellata indicated these species’ northward and southward dispersal, respectively. The genetic structure analyses suggest that the populations in these four species have limited geographical structure, except for the Chinese E. stellata population (SCP). The phylogenetic analyses suggest that E. byunsanensis and E. pinnatifida are sister species and that Chinese SCP may not belong to E. stellata. The ancestral area reconstruction revealed that the most recent common ancestor of the four species existed in the current Chinese habitat of E. stellata. This study shows that E. byunsanensis and E. pinnatifida originated from a southern Eranthis species and speciated into their current forms near Jeju island and near western regions of Japan, respectively, during the Miocene. E. stellata may have dispersed southward on and near the Korean peninsula, though its specific origin remains unclear. Interestingly, the Chinese E. stellata population SCP suggests that the Chinese population might be most ancient among all the four Eranthis species. E. pungdoensis may have allopatrically speciated from E. byunsanensis during the Holocene. The Korean peninsula and the surrounding areas can be considered interesting regions which provide the opportunity to observe both northern‐ and southern‐sourced Eranthis species.


| INTRODUC TI ON
The Korean peninsula and surrounding areas, including Japanese archipelago, East Sea, Yellow Sea, East China Sea, and the Chinese and Russian regions near North Korea, have a variety of geographical features and have experienced a number of dynamic orogenic and ecological events. This has led to diverse distributional patterns for a wide range of species in these regions. For example, Korea, which is surrounded on three sides by sea and contains large mountainous regions, exhibits high geographical complexity and provides the combinations of diverse geography and ecology for many taxa on this area. Also, interestingly, these regions can be appropriate for observing many closely related species exhibiting various distributions.
To aid in our understanding of various distributional patterns of the taxa in these regions and investigate their origins and speciation histories based on these patterns, we propose the genus Eranthis (Ranunculaceae) as an intriguing model for evolutionary study. The genus Eranthis is an early flowering herbaceous plant which has a tuberous rhizome, petaloid sepals, and palmately divided leaves. This genus is made up of 8-9 species that range from Europe to East Asia, and exhibits interesting distributions. This genus generally exhibits high levels of endemism and is distributed in both mainlands and islands. Its constituent species seldom cooccur, and the sizes of these species' distributional ranges usually vary much.
A number of morphological, ecological, and genetic studies have been conducted on the genus Eranthis. For example, previous research has investigated the genetic variation of the populations of the Korean endemic Eranthis byunsanensis ( Figure 1) and the environmental characteristics of its habitat (Kim et al., 2012;So, Lee, & Park, 2012). E. byunsanensis and its closely related species Eranthis pungdoensis were compared based on their genetic variation, and taxonomic analysis was conducted for these two species (Lee, Yeau, & Lee, 2012). In addition, the seeds of E. byunsanensis and E. stellata have been compared morphologically (Jung, Shin, & Heo, 2010), the flowering rate and pollen release of E. hyemalis, which inhabits Europe (Rysiak & Żuraw, 2011), have been analyzed, and cytological studies on E. stellata and E. pinnatifida have been conducted (Kurita, 1955;Yuan & Yang, 2006). Despite this past research interest, however, the origin and evolution of this genus have not received much attention.
In this study, we aimed to investigate the speciation as establishment histories of this genus on the Korean peninsula and in nearby regions, thereby adding to the basic knowledges of this genus, and gaining important evolutionary insights into its constituent species.
Eranthis species are morphologically clustered into two groups: one with yellow sepals ranging from Western Europe to Central Asia, and the other with white sepals ranging from Central China to Russia, Korea, and Japan. Here, we focus on four Eranthis species with white sepals: (a) Eranthis stellata, found all across the Korean peninsula, in Jilin and Liaoning in China, in the western land near the Sea of Okhotsk and in Primorskiy Krai, in eastern-most Russia; (b) Eranthis pinnatifida, endemic to Japan and found in the central and southern areas of Honshu; (c) E. byunsanensis, endemic to Korea and found across the Korean peninsula, with its northern limit possibly in North Korea; and (d) E. pungdoensis, endemic to Korea and found only on the small island of Pungdo in the Yellow Sea (Oh & Ji, 2009;Oh et al., 2016;Sun, Kim, & Kim, 1993). We analyzed 12 chloroplast microsatellites and two chloroplast noncoding regions in these species, looking specifically at (a) the genetic diversity of the populations of each species; (b) the genetic structure of each species; (c) phylogenetic reconstructions at both population and species levels; (d) the divergence times for each lineage and species; and (e) ancestral area reconstructions.
In this study, we examined to which direction these four Eranthis species dispersed in the past, and where and when they speciated.
We revealed that the Korean peninsula, as a contact area for "northern" E. stellata and "southern" E. byunsanensis and E. pungdoensis, can offer the opportunity to infer the origins of these species, and that the origin of Japanese E. pinnatifida can also be analyzed in relation to other Eranthis species in Korea. In short, our research helps to clarify the current and historical distributions of these species, to determine their origins and speciation patterns, and at the same time, to reinforce the utility of the Korean peninsula and its neighboring areas as subjects of evolutionary biology.  (Figure 2b,c), and it is inferred that E. pungdoensis diverged from E. byunsanensis not very long ago, they were analyzed together as if they were the same species.
The sampled bracts or leaves were dried with silica gel, and the total genomic DNA was extracted using DNeasy Plant Kit (QIAGEN, Seoul, Korea). The voucher specimens were deposited in the herbarium of Chungbuk National University.

| cpSSR genotyping and cpDNA sequencing
All of the 935 individuals of the four Eranthis species were genotyped at 12 chloroplast microsatellite loci (Ebp01, Ebp40, Ebp27, Ebp31,Ebp25,Ebp12,Ebp10,Ebp06,Ebp38,Ebp11,Ebp28,Ebp32) which were randomly selected from the 24 cpSSR loci isolated by Oh and Oh (2017). The PCR procedure was the same as that of Oh and Oh (2017), and the length of PCR products was measured using the ABI3730xl DNA Analyzer (Applied Biosystems) and GeneMapper v.

(Applied Biosystems).
In addition, one individual from each population of the four focal Eranthis species (giving a total of 33 individuals) and five outgroup species, were sequenced at two chloroplast noncoding regions, rpl16 intron and petL-psbE (Shaw et al., 2005;Shaw, Lickey, Schilling, & Small, 2007

| Genetic diversity
The 12 chloroplast microsatellite markers were used to estimate the genetic diversity of each sampled population. We estimated the number of haplotypes in each population and number of private haplotypes. We also calculated the effective number of haplotypes, haplotype richness, and Nei's index of genetic diversity estimated without bias. These analyses were performed in Haplotype Analysis ver. 1.05 (Eliades & Eliades, 2009).

| Population structure
To determine the population structure based on cpSSR loci for the four Eranthis species, STRUCTURE 2.3.4 (Pritchard, Stephens, & Donnelly, 2000) was used. Ten independent runs were performed for each K (K = 1~10), with each run composed of 100,000 burn-in steps, followed by 100,000 MCMC steps. The admixture model and the correlated allele frequencies model were applied. The optimal number of populations (K) was determined using Evanno, Regnaut, and Goudet (2005) with the software STRUCTURE HARVESTER 0.6.94 (Earl & vonHoldt, 2012).
For each species, a spatial analysis of molecular variance (SAMOVA) was performed using the cpSSR data to define groups of populations that are geographically homogeneous and maximally differentiated with each other (SAMOVA 2.0; Dupanloup, Schneider, & Excoffier, 2002). In this analysis, a simulated annealing procedure was used, and the number of initial configuration of groups was set to 100.
The genetic variation among groups of populations, among populations within group, and within populations was estimated using analysis of molecular variance, AMOVA (Excoffier, Smouse, & Quattro, 1992) with Arlequin 3.5 (Excoffier & Lischer, 2010). In this analysis, the groups of populations were defined according to the results of SAMOVA analysis.
The software POPART (Leigh & Bryant, 2015) was used to construct the network of the sequences for the four species, with one sequence from each population, and here, median-joining method (Bandelt, Forster, & Röhl, 1999) was applied. In the analysis process, the columns with gaps or ambiguous characters in the alignment input files were masked, which often led to originally different sequences becoming identical. Because of this, even though all the sequences were different from one another in the input data, there were haplotypes with more than one sample in the result.

| Phylogenetic inferences and divergence time estimation
Bayesian inference and maximum likelihood were applied in the reconstructions of the phylogenetic trees. The two chloroplast noncoding sequences, rpl16 intron and petL-psbE, were aligned with MUSCLE (Edgar, 2004) using the software MEGA7 (Kumar, Stecher, & Tamura, 2016), and manually edited with the same software.
These two regions were concatenated (2,293 bp) for the Bayesian analysis of phylogenetic relationships, which was conducted in BEAST 1.8.0 (Drummond & Rambaut, 2007;Drummond, Suchard, Xie, & Rambaut, 2012). One sequence from each population across the four species (33 sequences in total), were analyzed, with C. simplex (CSD), A. asiatica (ND), E. longistipitata (LC), E. albiflora (HR) used as outgroups. E. hyemalis (HYE), which is European Eranthis species, was excluded from this analysis, because the sequencing for rpl16 intron region in this species was not successful.
In Bayesian inferences, in which tree topology and divergence times were sought, nucleotide substitution model of HKY + G + I and uncorrelated lognormal relaxed clock were applied, and the Yule process was chosen as a tree prior. The length of the MCMC chain was set to 800,000 generations, and parameters were logged every 200 generations. In TreeAnnotator 1.8.0 (Rambaut & Drummond, 2013), burn-in was set to 400 trees and maximum clade credibility tree was generated with mean height as node height. Software FigTree 1.4.3 (Rambaut, 2017) was used to graphically present the phylogenetic trees.
Since we did not have any fossil record which can be used to estimate the divergence times, we obtained two calibration points from a previous phylogeographical study which included the genus Eranthis in its phylogeny (Wang et al., 2016). In that study, the crown group age of the genus Eranthis, when including seven extant species in its phylogenetic analysis, was <47 Ma. Also, in another phylogeny reconstruction in the same paper, the most recent common ancestor (MRCA) for E. stellata, A. asiatica, and C. simplex was determined to be 57.55 Ma with 95% highest posterior density of 56.41-59.21.
Maximum likelihood phylogenetic trees were constructed in RAxML 7.0.3 (Stamatakis, 2006) with GTR model. The rpl16 intron region and petL-psbE region were analyzed separately; for the petL-psbE region, E. hyemalis was included in the outgroup, because the sequencing was successful.

| Biogeographical analysis
We performed a statistical dispersal vicariance analysis (S-DIVA) in the software RASP 4.0 (Yu, Harris, Blair, & He, 2015). In this ancestral area reconstruction, the distributional ranges of the four Eranthis

| Genetic diversity
The genetic variation of the 33 populations from E. byunsanensis, E. pungdoensis, E. stellata, and E. pinnatifida was analyzed at 12 chloroplast microsatellites loci (Table 1)  were mostly assigned to cluster 2 (Supporting information Figure   S1 in Appendix S1). For E. pinnatifida, the Hiroshima populations PH5, PH8, PH9 were assigned to cluster 1, the Shiga populations PS2, PS5 assigned to cluster 2, and the Yamaguchi population PY to cluster 3 (Supporting information Figure S1 in Appendix S1).
The clustering patterns for E. pinnatifida were consistent with their geographical distributions (Figure 2). In the case of E. stellata, all the seven Russian populations (SR1, SR2, SR3, SR5, SR7, SR8, and SR9) and the two Chinese populations (SCW, SCP) were assigned to cluster 1, and the others were assigned to cluster 2 (Supporting information Figure S1 in Appendix S1  Figure 4), except for the Chinese population SCP, which was not grouped with other E. stellata populations (Figure 4). In this network, the population SCP behaved as if it was a separate species. The starlike network in E. stellata indicates that this species has experienced population expansion (Figure 4).

| Phylogenetic inferences and divergence time estimation
The Bayesian analysis of phylogenetic relationships including 37 concatenated sequences of rpl16 intron and petL-psbE region showed that the Eranthis species are generally monophyletic, except for E. stellata ( Figure 5), in which the Chinese population SCP was clustered with E. byunsanensis and E. pinnatifida, rather than with other E. stellata populations, and thus, is a paraphyletic group.
E. byunsanensis and E. pinnatifida appear to be sister species, while E. pungdoensis was nested within the E. byunsanensis clade.
Maximum likelihood analyses in the software RAxML, which used rpl16 intron region and petL-psbE intergenic region separately, produced noticeably different tree topology to that from Bayesian inference using concatenated sequences ( Figure 5; Supporting information Figure S2 and S3 in Appendix S1  Figure 5).

| Ancestral area reconstruction
S-DIVA analysis revealed that the divergence between Korean endemic Eranthis species (A)-E. byunsanensis, E. pungdoensis -and Japanese endemic E. pinnatifida (D) was a vicariant event, and that F I G U R E 5 Phylogenetic relationships between the sequences (one sequence from each population) in the four Eranthis species constructed with BEAST. The numbers in brackets show 95% HPD of the divergence times, and two black circles on the nodes denote the calibration points. * mark beside E. stellata for SCP population indicates this population's different genetic characteristics from other E. stellata populations in E. stellata, some divergences between Chinese (B), Korean (A), and Russian (C) populations were also vicariances ( Figure 6). For the four focal Eranthis species, Chinese range (B) was determined as an ancestral range. Overall, seven dispersal events and ten vicariant events are thought to have occurred.

| The origin of E. byunsanensis
The  Figure S2 in Appendix S1), which is inconsistent with the abovementioned assumption regarding population age and genetic diversity. This observation of phylogeny can partially support southward dispersal of this species, although more evidence is needed for drawing conclusion.
F I G U R E 6 Ancestral area reconstructions performed with S-DIVA in RASP. The map shows the four ranges (a-d) defined from the sampling strategy in this study. Pie charts at the nodes of the tree show the possible ancestral ranges, for which in this analysis, the maximum number was set to four The northward dispersal of E. byunsanensis enables the establishment of two different hypotheses regarding which species and region it speciated from. E. byunsanensis could have speciated from an Eranthis species in Japan, or one in China. The westward dispersal and following speciation from a Japanese species (assuming that it was the Japanese endemic E. pinnatifda and not an earlier common ancestor) can be refuted by the Bayesian phylogeny, in which these two species are sister taxa, and the crown group of E. pinnatifida (7.03 mya) is slightly younger than that of E. byunsanensis (9.47 mya) ( Figure 5). In contrast, if E. byunsanensis speciated from Chinese Eranthis species, it's possible that this Eranthis species might have been distributed in the east of Sichuan (the habitat of current Chinese E. albiflora), in the western sea of the Korean peninsula, and in East China Sea, since at some points in the past, these regions were all connected by land bridge (Ota, 1998   the TCS haplotype network revealed that Jeju was the likely center of differentiation for East Asian Eranthis species, and this conclusion is consistent with our results regarding the speciation of E. byunsanensis. In SAMOVA, Jeju population was differentiated from the other six populations, possibly due to its high genetic diversity. This result can also indicate that, after the geographical isolation of Jeju, genetic differentiation has accumulated, maybe due to genetic drift to some extent, over the period of less than 10,000 years between the sea level rising during the Holocene and the present day. There is one more possibility regarding the speciation of E. byunsanensis-it may have speciated from E. stellata, which is found in Russia, China, North Korea, and South Korea. The distributional patterns of E. stellata and E. byunsanensis on the Korean peninsula seem to show allopatry, in which E. stellata distributes on mountain ranges with high altitudes, and E. byunsanensis on the regions with relatively low altitudes in South Korea (See Supporting information Figure S4 in Appendix S1). If these distributional patterns indicate allopatric speciation of E. byunsanensis from E. stellata, this speciation might have occurred after a Korean geological event, Tertiary tilted flexure (Shin & Hwang, 2014). In this event, the mountain ranges running in a north-south direction were uplifted on the eastern side of the Korean peninsula, tilting the topography. This theory is concordant with current distributions of the two species. The tilted flexure of the Korean peninsula is inferred to have started about 23 mya in the early Miocene (Shin & Hwang, 2014). However, according to Bayesian phylogenetic analysis, the divergence between E. stellata and E. byunsanensis occurred prior to this time, at about 27.15 ± mya, in the Oligocene ( Figure 5). As a possible explanation, E. stellata and E. byunsanensis might have diverged for some reasons prior to the tilted flexure, and then their distributions might have been affected by environmental divergence following the uplift event, presenting allopatric pattern. However, given that the factors which might have caused these two species' divergence in the Oligocene is unknown, the speciation of E. byunsanensis from E. stellata may be considered unlikely. This is also supported by STRUCTURE analysis (Figure 3), in which the genetic structures for E. stellata and E. byunsanensis are significantly different from each other.

| The origin of E. pinnatifida
For E. pinnatifida, which inhabits the central and southern parts of Honshu, there is high possibility that the populations dispersed northward after speciation. This is evidenced by Bayesian and maximum likelihood phylogenetic analyses, in which the northernmost two populations of Shiga prefecture (PS2 and PS5) were the last to diverge of the six E. pinnatifida populations ( Figure 5; Supporting information Figure S2 and S3 in Appendix S1). This was observed in all the phylogenies constructed in this study. From the analysis of genetic diversity, it seems clear that the three Hiroshima populations (PH5, PH8, and PH9), which have more southern distributions than the Shiga populations, have a longer evolutionary history than the other three populations, weakly supporting this hypothesis as well.
Also, the fact that this species does not inhabit northern Honshu and Hokkaido partially supports this northward dispersal hypothesis. be inferred that the tendency of these two species to disperse northward and the restricted gene flow between these two species during the disconnection of Korea and Japan may have played a role to some extent. When comparing the distributional regions of E. byunsanensis and E. pinnatifida, the latitudinal ranges of these two species are quite similar, which suggests their niche similarity and the close relationships between these two species (their similar genetic structure is also indicated in STRUCTURE analysis ( Figure 3)), though this should be accepted cautiously.
In addition, we investigated the possibility of southward dispersal of E. pinnatifida, even though the absence of this species from northern Honshu and Hokkaido can be interpreted in a number of ways, and cannot be currently explained. According to unpublished data, E. stellata can be found in Sakhalin; because Sakhalin, Hokkaido, and Honshu were connected at some points in the past, (Igarashi & Zharov, 2011;Ohnishi, Uno, Ishibashi, Tamate, & Oi, 2009;Suzuki et al., 2004) the speciation of E. pinnatifida from northern E. stellata in these regions cannot be completely dismissed. Moreover, the petal morphologies of E. stellata and E. pinnatifida are rather similar (Figure 2d,e), supporting to a minor degree the speciation from E. stellata. However, more decisive evidence for this hypothesis is needed for meaningful conclusion. In this study, most of the analyses, including phylogenetic reconstructions, indicate that E. pungdoensis is genetically very similar to E. byunsanensis ( Figure 5; Supporting information Figure S2 and S3 in Appendix S1), and is nested within E. byunsanensis in the phylogenies ( Figure 5; Supporting information Figure S2 in Appendix S1). A previous study also reported a similarity between the two species using cpDNA sequences and ITS regions . This may clearly suggest E. pungdoensis speciated from E. byunsanensis.

| The origin of E. stellata and E. pungdoensis
Pungdo, the small western island in Korea, might have been connected to the Korean peninsula for a long time, during the Last Glacial Maximum (LGM; when the sea level was lower and the land bridge existed; Millien-Parra & Jaeger, 1999), and long before LGM, considering its proximity to the Korean peninsula (Ota, 1998 Appendix S1). From these observations, we suggest that the SCP population has the genetic characteristics of an ancient Eranthis species, even though we cannot clearly conclude which Eranthis species this may have been. According to the phylogenies, the SCP population may be genetically rather similar to E. albiflora, considering their close divergence times. S-DIVA analysis also supports this argument. The results of S-DIVA analysis suggest that the ancestral location for the common ancestor of the four Eranthis species may have been in China ( Figure 6).
This provides evidence for the possibility that Chinese E. stellata is the most ancestral population among all these four Eranthis species. Even though the highest genetic diversity was found in Russian E. stellata populations, and the sampling of E. stellata in Russia and North Korea is incomplete, this argument should be seriously considered.
Given the unique genetic characteristics of the SCP population, if these characteristics represent ancestral genetic traits, then we can suggest that it is an extremely stable population that has undergone a very slow genetic evolution. An alternative hypothesis is that the SCP population has genetically evolved rapidly for some reasons, resulting in the high level of differentiation, even though there are no geological or ecological factors which may have driven this. Another interpretation, considering that the bract and flower morphology of the SCP population is the same as the other E. stellata populations, is that the SCP population may be a cryptic species of E. stellata, even though explaining this is challenging-the explanation would have to involve a very rare mutational event.
Again, the genetic characteristics of the SCP population should be interpreted with caution, partly because these results depend on only chloroplast DNA, even though it is meaningful that chloroplast noncoding sequences and chloroplast microsatellites presented the concordant results. Related experiments using nuclear and mitochondrial DNA should be conducted for clear identification of SCP population and definite conclusions on its molecular evolution.

| CON CLUS ION
When combining the possible origins of these Eranthis species described above, it can be concluded that the Korean peninsula and nearby regions harbor northern-sourced (E. stellata) and southern-sourced (E. byunsanensis, E. pungdoensis, E. pinnatifida) species of Eranthis, and these regions are phylogeographically valuable in terms of determining complex origin patterns for closely related species. Also, the observation of unique Chinese SCP population provides the opportunity to infer the ancestral location for these four focal species, thereby adding an important description to their evolutionary history.

ACK N OWLED G M ENTS
The authors thank Dr. Dooil Kim and Dr. Kyong-Sook Chung for helping with the data analysis.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N
Byoung-Un Oh contributed to project design, sample collection and writing manuscript. Ami Oh contributed to project design, sample collection, lab work, data analysis and writing manuscript.

DATA ACCE SS I B I LIT Y
DNA sequences:GenBank accessions: MH660625-MH660701.