For each species, we tried to sample at least one accession from its typical geographical range for complete plastome sequencing ( Table 1 , Figure 1 ): P. sikokianus, P. kurilensis, P. selskianus, and P. litoralis. In some species (P. aizoon, P. middendorffianus, P. ellacombeanus, and P. kamtschaticus), there was more than one accession based on wild or cultivated origins. The plant materials from the Botanical Garden Institute (Vladivostok, Russia) represent transplanted ones from nature into cultivation with accurate species identification. In the case of the Korean endemics, P. zokuriensis and P. latiovalifolium, we sampled one accession from the type locality, Mt. Sokri, and Geumdae-bong Peak, Mt. Taebaek, respectively, in central Korea. Unfortunately, a newly described Korean endemic, P. daeamensis, was not included in this study owing to its recent description and lack of plant materials. Thus, we sequenced the complete plastome of 15 accessions, representing 11 taxa. For sequencing of nrDNA ITS sequences, we sampled a total of 89 accessions representing 10 taxa ( Supplementary Table 1 ). All these accessions were of wild origin, with accurate species identification by Seung-Chul Kim, Masayuki Maki, Takuro Ito, and Marina Koldaeva.

Total genomic DNA was isolated using the DNeasy Plant Mini Kit (Qiagen, Carlsbad, CA, USA), following the manufacturer’s protocol. For nrDNA ITS sequencing, we followed the same protocols of polymerase chain reaction (PCR) amplification and subsequent sequencing as described by Seo et al. (2020). The nrDNA ITS is a multi-copy nuclear gene, and different ribotypes may exist within a single individual, complicating inferences about species relationships. To minimize the potential complications of ribotypes in species relationship inference, we used direct sequencing of PCR products rather than cloning of ribotypes in Phedimus species (Gontcharova et al., 2006; Chao, 2020; Seo et al., 2020). In addition, we only included accessions with clean sequencing results. The pairwise sequence divergence was calculated based on the Kimura 2-parameter method (Kimura, 1980) using MEGA11 (Tamura et al., 2021). According to earlier studies, we sequenced the entire plastome (e.g., Kim et al., 2019; Cho et al., 2020; Yang et al., 2021; Yun and Kim, 2022). An Illumina paired-end (PE) genomic library was constructed and sequenced using the Illumina HiSeq platform (Illumina, Inc., San Diego, CA, USA) at Macrogen Corporation (Seoul, Korea). Sequence reads of the plastomes were assembled using the de novo genomic assembler Velvet 1.2.10 (Zerbino and Birney, 2008) or NOVOPlasty version 4.3.1 (Dierckxsens et al., 2017). Annotation was performed using Geneious R10 (Biomatters, Auckland, New Zealand) and ARAGORN v1.2.36 (Laslett and Canback, 2004). The pairwise sequence divergence was also calculated based on the Kimura 2-parameter method (Kimura, 1980) using MEGA11 (Tamura et al., 2021). The annotated plastome sequences were deposited in GenBank under accession numbers ( Table 1 ). The annotated GenBank (NCBI, Bethesda, MD, USA) format sequence file was used to draw a circular plastid genome map using the OGDRAW software v1.2 (CHLOROBOX) (Lohse et al., 2007). We performed DnaSP v6.10 (Rozas et al., 2017) sliding window analysis, with a step size and window length of 200 bp and 800 bp, respectively, to determine mutation hotspots (that is, the most divergent regions of the plastome). To determine genes under positive selection, a site-specific model was developed using EasyCodeML (Gao et al., 2019) with CodeML algorithms (Yang, 1997). Seven codon substitution models (M0, M1a, M2a, M3, Mt, M8, and M8a) were constructed and compared to detect positively selected sites using the likelihood ratio test (LRT).

For the ITS phylogeny, we included 106 accessions, representing two species (three accessions) of subg. Phedimus (S. obtusifolius was not available) and 15 species (103 accessions) of subg. Aizoon, and based on previous studies, the genus Rhodiola was used as an outgroup (Mayuzumi and Ohba, 2004; Gontcharova et al., 2006; Chao, 2020; Messerschmid et al., 2020) ( Supplementary Table 1 ). Of the 106 accessions, newly sequenced ITS sequences (a total of 89 accessions) included P. latiovalifolium (16 accessions), P. zokuriensis (four accessions), P. takesimensis (14 accessions), P. middendorffianus (six accessions), P. kurilensis (one accession), P. ellacombeanus (nine accessions from Korea and Japan), P. litoralis (two accessions), P. aizoon (18 accessions), P. selskianus (one accession), and P. kamtschaticus (18 accessions). All the sequences were edited and assembled using Sequencher v4.2.2 (Gene Codes, Ann Arbor, MI, USA) and Geneious R10 (Biomatters, Auckland, New Zealand). The ITS sequences for the following species (P. stellatus, P. spurius, P. odontophyllus, P. yangshanicus, P. hybridus, P. sichotensis, and P. sikokianus) were also obtained from GenBank. We included these species (totaling 17 accessions) because some represent members of subg. Phedimus (P. stellatus, P. spurius, P. odontophyllus, and P. yangshanicus), and other species (P. hybridus, P. selskianus, P. sichotensis, and P. sikokianus) have distinct species diagnostic features, minimizing potential misidentifications and providing overall species relationships in subg. Aizoon. The sequences were aligned using Clustal X v1.83 (Thompson, 1997) with a final manual adjustment using MacClade (Maddison and Maddison, 2002). Maximum likelihood (ML) analysis was conducted using IQ-TREE v1.4.2 (Nguyen et al., 2015), with 1,000 replicate bootstrap (BS) analyses, based on the best-fit model of TIM3e + G4 selected by ModelFinder (Kalyaanamoorthy et al., 2017). For plastome phylogeny, the complete plastome sequences were aligned using MAFFT v7 (Katoh and Standley, 2013), and an ML phylogenetic tree was constructed using IQ-TREE with 1,000 bootstrap replicates (Nguyen et al., 2015). The best-fit evolutionary model for the complete plastome sequences, TVM + F + I + G4, was selected based on ModelFinder (Kalyaanamoorthy et al., 2017), implemented in IQ-TREE v1.4.2. A representative species of Rhodiola was used as the outgroup in ITS analysis (Messerschmid et al., 2020). Given the lack of representative plastomes of subg. Phedimus, we included Rhodiola species as part of the ingroup and Umbilicus as the outgroup.

Divergence times based on complete plastome sequences were estimated using the Bayesian method (Drummond et al., 2006) using BEAST version 1.10.4 (Suchard et al., 2018). The XML file for analysis was prepared using the Bayesian evolutionary analysis utility (BEAUTi). Owing to the lack of reliable fossils of Phedimus and related genera of Crassulaceae, we considered two secondary calibration points based on ITS phylogeny: an estimated Rhodiola crown mean age of 7.17 Myr and a standard deviation of 4.87, giving a range of 3.09–12.03 Myr, and the Phedimus and Rhodiola clade stem mean age of 39.43 Myr, and a standard deviation of 14.31, giving a range of 24.91–53.74 Myr (Messerschmid et al., 2020). We used the Yule process speciation prior, a lognormal relaxed clock model, and the GTR-γ substitution model, and then the ucld.mean parameter was specified to be uniform with 0.333 as the initial value, 0.00 as the lower limit, and 1 as the upper limit (Drummond et al., 2006). Posterior distributions for each parameter were estimated using an MCMC run for 400 million generations with a sampling frequency of 50,000 generations. The posterior distribution of all statistics was checked using Tracer version 1.5 (Rambaut and Drummond, 2009) to assess convergence and confirm whether the effective sample sizes (ESS) for all parameters were larger than 200 (Drummond et al., 2012). In addition, we used TreeAnnotator version 1.5 (http://beast.bio.ed.ac.uk/TreeAnnotator) to produce a maximum credibility tree of mean divergence time and 95% highest posterior density (HPD) intervals with a posterior probability (PP) limit (0.5) after removing the first 25% of trees as burn-in (Drummond et al., 2012).

The complete plastome length of the subg. Aizoon ranged from 151,557 bp (P. middendorffianus; Primorsky Krai, Russia) to 151,869 bp (P. ellacombeanus; Hokkaido, Japan) ( Table 1 ; Figure 2 ). The large single copy (LSC) region, small single copy (SSC) region, and two inverted repeat (IR) regions ranged from 82,890 bp (P. middendorffianus; Primorsky Krai, Russia) to 83,179 bp (P. ellacombeanus; Hokkaido, Japan), 16,677 bp (P. aizoon; Geumdae-bong Peak, Korea) to 16,718 bp (P. sikokianus; Tokushima Pref., Japan), and 25,976 bp (P. aizoon; Primorsky Krai, Russia) to 25,988 bp (P. aizoon var. floribundus; Iwate Pref. Japan), respectively ( Table 1 ). All 15 accessions of the subg. Aizoon contains 135 genes, including 88 protein-coding, eight ribosomal RNA, and 37 transfer RNA genes. The overall guanine-cytosine (GC) content ranged from 37.7% to 37.8%, and 20 duplicate genes were found in the IR regions. Sliding window analysis using the DnaSP program identified several highly variable genic and intergenic (“–”) regions in 15 plastomes of subg. Aizoon: atpF–atpH (Pi = 0.01223), ycf1 (Pi = 0.01015), trnH–psbA (Pi = 0.0076), rbcL–accD (Pi = 0.00747), ndhF–rpl32 (Pi = 0.00622), rpoB–trnC (Pi = 0.00601), psbZ–trnG (Pi = 0.00533), trnW–trnP (Pi = 0.0052), and rps8–rpl14 (Pi = 0.00489) ( Figure 3 ). The average nucleotide diversity value (Pi) over the entire plastome was 0.00188. Among the conserved genes, we identified seven genes with positively selected sites ( Table 2 ). These genes included the c-type cytochrome synthesis gene (ccsA), chloroplast envelope membrane protein (cemA), maturase K gene (matK), NADH dehydrogenase subunit gene (ndhC), photosystem II protein gene (psbJ), cytoplasmic ribosomal protein L22 gene (rpl22), and RNA polymerase C2 gene (rpoC2).

For the first time, we obtained robust and well-resolved whole plastome-based phylogenetic relationships among representative species of subg. Aizoon in East Asia ( Figure 4 ). The plastome tree suggested that P. sikokianus, which is diploid and endemic to southern Japan, diverged early within the subg. Aizoon and further identified three major lineages, clades A–C. The first lineage, clade A (100% bootstrap support, BS), included P. ellacombeanus, sampled as P. aizoon var. floribundus from the type locality of the species in Japan (Hokkaido, Japan), P. kurilensis (Sakhalin, Russia), and P. kamtschaticus (Hokkaido, Japan). All these taxa occur from northern Japan (Hokkaido) to the southern Kuriles (Russia). In this clade, P. kurilensis, endemic to the southern Kuril Islands and considered a synonym of P. sikokianus, is sister to P. kamtschaticus from Hokkaido, Japan (100% BS). After the divergence of clade A, P. aizoon var. floribundus (Iwate, Japan) is sister to the remaining clades of B and C (100% BS). The second lineage, clade B (100% BS), included P. middendorffianus (two accessions from Russia and Korea), P. kamtschaticus (Korea), two accessions of P. ellacombeanus, and P. zokuriensis (Korea). One Korean accession of P. middendorffianus sampled from Gangwon-do Province is sister to P. kamtschaticus, which was also sampled from Gangwon-do Province (moderate support, 74% BS). The central South Korean Peninsula Endemic P. zokuriensis is sister to the P. ellacombeanus sampled from Sochi Island, which is in the southeastern part of the Korean Peninsula (95% BS). Lastly, the third lineage, clade C (100% BS), included two Korean endemic species (P. latiovalifolium and P. takesimensis), P. aizoon (Korea), and species from Russia (P. selskianus, P. litoralis, and P. aizoon). The monophyletic Ulleung Island endemic P. takesimensis shares its most recent common ancestor with Phedimus species from Russia (P. selskianus, P. litoralis, and P. aizoon; 92% BS). The central South Korean peninsula endemic P. latiovalifolium is sister to the clade containing species from Russia and the Ulleung Island endemic P. takesimensis (84% BS). Overall, plastome-based phylogeny of subg. Aizoon suggested that P. takesimensis is monophyletic, while the more widely distributed P. aizoon and P. kamtschaticus appear to be polyphyletic, which requires further confirmation based on multiple accessions. Pairwise sequence divergence based on Kimura 2-parameter distances was shown in Supplementary Table 2 . The average pairwise sequence divergence for all 15 accessions was 0.189%, ranging from 0.008% (between P. kamtschaticus OP344954 and P. kurilensis OP344953) to 0.425% (between P. takesimensis NC026025 and P. sikokianus OP344947). The average pairwise sequence divergence for four accessions of P. takesimensis on Ulleung Island and three accessions of P. ellacombeanus was 0.053% and 0.215%, respectively. The type locality accession of P. ellacombeanus (OP344955) in Japan showed quite divergent pairwise sequence divergence from its conspecific populations sampled in Korea: 0.307% between type locality accession OP344955 and two Korean accessions versus 0.030% between two Korean accessions of P. ellacombeanus.

The ITS ML tree revealed two major lineages within the genus Phedimus ( Figure 5 ; phylogram shown in Supplementary Figure 1 ). Pairwise sequence divergence based on the Kimura 2-parameter distances was shown in Supplementary Table 3 . For all 106 accessions (including two subgenera Phedimus and Aizoon), the average pairwise sequence divergence was 1.654%: 2.967% for species of subg. Phedimus and 1.371% for species of subg. Aizoon. Pairwise sequence divergence between two species of subg. Aizoon (i.e., P. odontophyllus and P. yangshanicus) and the remaining species was 7.064%, while between P. odontophyllus/P. yangshanicus and subg. Phedimus was 1.254%. Accordingly, P. odontophyllus and P. yangshanicus are more closely related to species of subg. Phedimus than to consubgeneric species in subg. Aizoon. The first lineage (99% BS) included two species of subg. Phedimus (P. stellatus and P. spurius) and two species of subg. Aizoon (P. odontophyllus and P. yangshanicus), confirming non-monophyly of subg. Aizoon (Chao, 2020). According to ITS sequence divergence, two species of subg. Aizoon, P. odontophyllus and P. yangshanicus, are more closely related to members of subg. Phedimus than those of subg. Aizoon ( Supplementary Table 3 ). The second lineage (100% BS) included all but two species of subg. Aizoon (100% BS), and monophyletic P. hybridus (98% BS) diverged first within this clade. Although bootstrap support values within the subg. Aizoon restricted us from rigorously inferring species monophyly and interspecific relationships, but some species relationships could be postulated with caution. The ITS tree identified one weakly supported clade 1 (59% BS), which included P. middendorffianus, P. selskianus, P. kurilensis, P. sichotensis, and P. takesimensis. The Ulleung Island endemic P. takesimensis is monophyletic (96% BS) and embedded within the non-monophyletic P. middendorffianus, sampled primarily from northeastern (Jilin) China. One accession of P. middendorffianus sampled from Primorsky Krai, Russia, was sister to the remaining accessions within this clade. Phedimus sichotensis (AM039913 from GenBank), which is often treated as a subspecies of P. middendorffianus (P. middendorffianus subsp. sichotensis), is sister to the clade containing primarily P. middendorffianus and P. takesimensis (73% BS). The southern Kuriles (Russia) endemic P. kurilensis accession sequenced in this study is sister to the clade containing P. middendorffianus and P. takesimensis (80% BS).

In addition, the ITS tree also suggested that P. sikokianus (AB08863 from GenBank), which is narrowly endemic to southern Japan (Shikoku), is sister to the members of clade 2 (52% BS weak support; thus, it could be either sister to clade 1 or 2). Excluding P. sikokianus, the remaining clade 2 is strongly supported (99% BS). Within clade 2, the central South Korean Peninsula endemics P. zokuriensis and P. latiovalifolium are monophyletic (84% and 94% BS, respectively), whereas the widely distributed P. aizoon and P. kamtschaticus are not monophyletic. The ITS tree also showed that P. litoralis, endemic to Russia, was deeply embedded within the P. aizoon lineage sampled primarily from Primorsky Krai (Russia) and Heilongjiang (China) (56% BS). Phedimus ellacombeanus sampled from the type locality of Japan and Korea was polyphyletic. One accession sampled from southeastern Korea, Igarigani Beach, Pohang (Gyeongsangbuk-do Province), is closely related to P. kamtschaticus, sampled from various parts of the Korean Peninsula, and P. aizoon, sampled from Gangwon-do Province, Korea. Conversely, two accessions sampled from southeastern Korea, that is, Uje-bong Peak and Geoje Island (Gyeongsannam-do Province), are closely related to P. aizoon and P. kamtschaticus collected from Korea, Russia, and China. Furthermore, two accessions sampled from Hachuja Island in the southern part of Korea are sister to the clade containing all but P. sikokianus and one accession of P. kamtschaticus (Korea) in clade 2. Two accessions of P. ellacombeanus from Sochi Island in the southern part of Korea share their most recent common ancestor with P. latiovalifolium in Gangwon-do Province. Lastly, one accession collected from the type locality of P. ellacombeanus in Hakodate (southern Hokkaido, Japan) is embedded within P. aizoon from northeastern (Jilin) China (61% BS).

Several distinct lineages of highly polyphyletic P. aizoon were found exclusively in clade 2 ( Figure 5 ), including a lineage largely from Gangwon-do Province (Korea)/Jilin (China) and a lineage primarily from Primorsky Krai (Russia), Mongolia, and P. litoralis (56% BS). Several lineages of P. kamtschaticus were also revealed in Korea, including one lineage sampled from various parts of Korea (<50% BS) and the other lineage, which is sister to the clade of P. ellacombeanus (Sochi Island) and P. latiovalifolium. Both P. aizoon and P. kamtschaticus were not only intermixed with each other but also closely related to other congeneric species, such as P. ellacombeanus and P. litoralis.

Based on the complete plastome sequences, we estimated the divergence times of the major lineages within the subg. Aizoon ( Figure 6 ). Without any representative species from subg. Phedimus, the age of subg. Aizoon was estimated to be 27.09 Ma (95% HPD, 13.02–44.34 Ma), suggesting its origin in the late Oligocene. Within the subg. Aizoon, the early divergence of diploid P. sikokianus in southern Japan was immediately followed by the divergence of the remaining lineages (26.79 Ma; 95% HPD, 13.12–43.86 Ma). The age of the estimated divergence of the Ulleung Island endemic P. takesimensis from its continental sister species was estimated to be 2.03 Ma (95% HPD, 1.15–3.70 Ma), suggesting its early colonization soon after the formation of Ulleung Island (ca. 1.8 Ma). The estimated divergence of the narrow endemic P. zokuriensis in Korea from its sister P. ellacombianus was inferred to be very recent at 0.15 Ma (95% HPD, 0.05–0.33 Ma) during the Pleistocene (late Ionian), whereas that of the other narrow Korean endemic P. latiovalifolium was inferred at 10.53 Ma (95% HPD, 4.02–34.35 Ma) during the late Miocene (late Messinian).

Genome size, gene order, and number were highly conserved within the subg. Aizoon ( Table 1 ). In addition, the overall genome size, gene numbers, and GC content were like those of the sister lineage Rhodiola and other closely related genera (Hylotelephium, Kalanchoe, Orostachys, Rosularia, Sedum, Sinocrassula, and Umbilicus) (Kim and Kim, 2020; Zhao et al., 2020; Zhao et al., 2022). This suggests that despite being morphologically and cytologically extremely variable, major lineages of the subfamily Sedoideae have highly conserved plastomes. In the case of the hypervariable regions of plastomes ( Figure 3 ), which can be used as barcoding markers, we identified one ycf1 gene and three intergenic regions (trnH–psbA, ndhF–rpl32, and trnW–trnP), which showed high sequence variation in the two sister lineages Rhodiola and Phedimus (Zhao et al., 2022). These sets of mutation hotspots in the two sister genera could provide useful phylogenetic information for phylogeographic and population genetic studies at the intraspecific level. Rhodiola species, which are mainly adapted to alpine habitats in the Qinghai-Tibet Plateau and the Hengduan Mountains, have been shown to contain three positively selected genes (rpl16, ndhA, and ndhH) and one gene with a faster than average rate of evolution (psaA) (Zhao et al., 2020). The products of these genes may have been involved in the adaptive radiation of Rhodiola to high altitudes, an environment with low CO₂ concentrations and high-intensity light. In the subg. Aizoon of the Phedimus lineage, which was demonstrated to have significant niche divergence from its sister Rhodiola (Zhao et al., 2020), we identified different sets of positively selected genes (ccsA, cemA, matK, ndhC, psbJ, rpl22, and rpoC2) ( Table 2 ). Because Phedimus species occur in various wider habitats (e.g., grassy slopes, shrub thickets, meadows, rocky streamsides, sandy cliffs, mountain steppes, stony and gravelly soils in forests, sandy shores, deciduous fields and forests, subalpine meadows, limestone hills, and rocks), these plastomic adaptations are assumed to contribute to the diversification and range expansion of subg. Aizoon throughout its geographical range. This also suggests that these genes were most likely selected when the common ancestor of the genus Phedimus diverged from its sister lineage, Rhodiola. However, it is yet to be determined whether some species of subg. Phedimus, which occur at high altitudes (e.g., P. stevenianus, P. spurius, and P. obtusifolius), show any evidence of positive selection for the same genes (ndhA, ndhH, rpl16, and psaA) detected in Rhodiola. It is also necessary to investigate what drives radiating diversification and adaptation to diverse ecological niches (e.g., Kapralov et al., 2013; Zhang et al., 2021; Izuno et al., 2022).

One major objective of this study was to determine the origin and evolution of the Phedimus species endemic to Korea. This study, based on extensive sampling in eastern Asia, allowed us to gain insights into the origin of P. takesimensis on Ulleung Island for the first time. A previous study showed the monophyly of P. takesimensis on Ulleung Island without concretely determining the closest continental sister lineage(s) (Seo et al., 2020). The current study further corroborates previous findings that the morphological and genetic variation of P. takesimensis on Ulleung Island were accumulated after a single colonization event of a continental ancestral population on the island. However, the ITS tree strongly suggests that monophyletic P. takesimensis (96% BS) is embedded within paraphyletic P. middendorffianus ( Figure 5 and Supplementary Figure 1 ), whereas the species of Phedimus from Russia (P. aizoon, P. litoralis, and P. selskianus) are the closest continental relatives based on the plastome tree (100% BS; Figure 4 ). All accessions of P. middendorffianus sister to P. takesimensis, on Ulleung Island were from northeastern (Jilin) China. It also commonly occurs from eastern Siberia to the Russian Far East. Therefore, this species is a possible continental progenitor species of P. takesimensis on Ulleung Island, suggesting cautiously its origin from northeastern Asia as a geographical source area, based on ITS phylogeny ( Figure 5 ). Although this geographical region seems likely to be the origin of P. takesimensis, the plastome phylogeny suggests that different sets of species, i.e., P. aizoon, P. litoralis, and P. selskianus, are most likely progenitor species ( Figure 4 ). Of these three species, P. selskianus has a distinct characteristic of densely grayish pubescent leaves. Phedimus litoralis is a glabrous herb with an elongated, creeping, simple rhizome, and strong stems. The last species in this clade was P. aizoon from Primorsky Krai. Therefore, based on plastome phylogeny, we cautiously suggest that P. aizoon-like species from northeastern Asia (Jilin, China, or the Russian Far East) may also be involved in the origin of P. takesimensis. The estimated divergence time of P. takesimensis from its continental progenitor species was estimated to be 2.03 Ma (95% HPD, 1.15–3.70 Ma), suggesting its origin soon after the formation of Ulleung Island (ca. 1.8 Ma) ( Figure 6 ). Of the nearly 40 endemic species on Ulleung Island, P. takesimensis is an example of its geographical origin from the Russian Far East and northeastern China rather than from the Korean Peninsula (Rubus takesimensis, Yang et al., 2019; Campanula takesimana, Cheong et al., 2020; Prunus takesimensis, Cho et al., 2021), southern Korean Peninsula/southern Japan (Rubus takesimensis, Yang et al., 2019), and Japan/Sakhalin, Russia (Scrophularia takesimensis, Gil et al., 2020). The phylogenetic incongruence between maternally inherited plastome sequences and the multicopy nature of nrDNA ITS sequences further complicate the origin of P. takesimensis. Additional studies based on genome-wide SNPs (e.g., Cho et al., 2021) or reduced representation sequencing approaches (e.g., Johnson et al., 2019) would be useful for future phylogenetic studies of subg. Aizoon in East Asia.

The taxonomic status of P. zokuriensis and its relationship with other congeneric species is contentious. It can be distinguished from congeneric species by its weak and creeping stems. Such phenotypic variation from its congeneric species could be caused by its forest habitats and is well within the range of broadly distributed species, such as P. kamtschaticus or P. aizoon. Because P. kamtschaticus tends to occur on rocky surfaces in shaded forests and sunny forest edges instead of the more open grassland habitats of P. aizoon, it is highly plausible that P. zokuriensis is conspecific to P. kamtschaticus or at least closely related to it. Morphologically, P. zokuriensis is more closely related to P. kamtschaticus than to P. aizoon. In addition, based on extensive morphological analyses, Chung and Kim (1989) argued that P. zokuriensis and P. ellacombeanus are not infraspecific taxa of P. kamtschaticus, and this could be difficult to determine based on the ITS phylogeny owing to weak bootstrap support (<50% BS; Figure 5 ). Plastome phylogeny suggests that P. zokuriensis is embedded within the clade of P. ellacombeanus, which was sampled from the southern Korean Peninsula (100% BS). This plastome connection between accessions of forest habitat from the central Korean Peninsula and coastal habitat from the southern Korean Peninsula, including isolated islands, was unexpected, owing to their geographical distance and morphological differences. Nevertheless, based on a previous extensive morphological investigation (Chung and Kim, 1989) and its monophyly and close maternal relationship with P. ellacombeanus from this study, the species status of P. zokuriensis would be reasonable to maintain until different lines of evidence suggest otherwise.

Phedimus latiovalifolium is also narrowly endemic to Gangwon-do Province in the northern part of South Korea. It was originally described by Lee (1992) but was later hypothesized based on morphological intermediacy to have a hybrid origin between P. aizoon and P. kamtschaticus or between P. aizoon and P. ellacombeanus (Lee, 2000). Yoo and Park (2016), based on morphological and allozyme studies, refuted its hybrid origin and suggested a distinct taxonomic status for P. latiovalifolium. Our current study, for the first time, demonstrated the monophyly of P. latiovalifolium (94% BS, Figure 5 ) and suggested that it shared its most recent common ancestor with broad-leaved maritime P. ellacombeanus sampled from Hachuju and Sochi Island (Jeju and Gyeongsangnam-do Province, respectively) and P. kamtschaticus (71% BS). However, the complete plastome tree is less conclusive but shows its relationship with close relatives, suggesting that it shares the most recent common ancestor with species from the Russian Far East (P. selskianus, P. litoralis, and P. aizoon) and Ulleung Island endemic P. takesimensis in Korea (92% BS, Figure 4 ). Like P. zokuriensis, we suggest that it would be reasonable to maintain the species status of P. latiovalifolium because of its morphological and allozyme distinctions (Yoo and Park, 2016) and the strong monophyly demonstrated in this study. Unlike the recent origin of two other Korean endemic species, P. takesimensis and P. zokuriensis, it was suggested that the split of P. latiovalifolium from its sister lineage might have occurred at 10.53 Ma (95% HPD, 4.02–34.35 Ma) during the late Miocene ( Figure 6 ).

This study allowed us to broadly assess species relationships within Phedimus, thus giving us an opportunity to evaluate species boundaries. Within the genus Phedimus, two major lineages are recognized: the first clade includes two species of subg. Phedimus (P. stellatus and P. spurius) and two species of subg. Aizoon (P. yangshanicus and P. odontophyllus), while the second one includes all but two species of subg. Aizoon ( Figure 5 ). Two species (P. stellatus and P. spurius) in this clade are diploids (P. odontophyllus and P. yangshanicus of subg. Aizoon are unknown) with a simple descending dysploidy series of x = 7 to x = 6 to x = 5 (Hart et al., 1993; Hart and Bleij, 2003). These two species and other members of subg. Phedimus are morphologically distinct and owing to their largely contiguous geographical distribution in Eurasia, a recent origin has been suggested (Hart and Bleij, 2003). Although subg. Aizoon is a very distinct taxon based on morphological and cytological characteristics (x = 8), its component species are much less clear, except for the hirsute P. selskianus. Other taxa are very difficult to separate and can be merged because of the uniformity of their floral characters despite the extreme variation in vegetative characters (Hart and Bleij, 2003). However, Hart and Bleij (2003) followed a less rigorous and much more conservative approach, recognizing approximately 14 species in subg. Aizoon; Fröderström (1931) included all but one species (P. hybridus), either as a subspecies or synonym in P. aizoon. We fully agree that large-scale, comprehensive biosystematics studies of natural populations are required to properly understand morphological and cytological variation (Amano, 1990; Amano and Ohba, 1992; Hart and Bleij, 2003). However, our current study provides some insights into the species boundaries in subg. Aizoon from East Asia. We found that the monophyletic P. hybridus represents the earliest diverged lineage within this subgenus in the ITS phylogeny. Despite the poorly described nature of this species, P. hybridus appears to be a distinct taxon within subg. Aizoon. The ITS phylogeny suggested that P. selskianus, a very distinct hirsute species, is sister to the clade containing P. sichotensis (Russia), P. kamtschaticus (Japan), P. kurilensis (Russia), P. middendorffianus (China), and P. takesimensis (Korea) ( Figure 5 ). According to the plastome phylogeny ( Figure 4 ), it is sister to the clade of P. litoralis–P. aizoon (Russia). Therefore, if P. selskianus can be recognized as a distinct species, other species in this “clade 1,” such as P. sichotensis, P. middendorffianus, and P. takesimensis (Hart and Bleij, 2003), could also maintain cautiously distinct species status until different lines of evidence suggest otherwise. We could also maintain distinct species status for P. sikokianus (Japan), P. zokuriensis (Korea), and P. latiovalifolium (Korea) in “clade 2.” Phedimus sikokianus has been suggested to be a key species for understanding the evolution of the whole subg. Aizoon because of the low diploid chromosome number and the opposite leaves (Amano and Ohba, 1992; Hart and Bleij, 2003). Its phylogenetic position is unresolved in the ITS tree ( Figure 5 ), but in the plastome tree, P. sikokianus first diverged within the subg. Aizoon ( Figure 4 ). The precise phylogenetic position of P. sikokianus requires further study based on multiple samples and a robust phylogenetic framework. We were unable to include any members of subg. Phedimus, but as in the ITS tree, it is still possible that P. sikokianus represents one of the early diverged lineages in subg. Aizoon, as supported by cytological evidence (Amano and Ohba, 1992).

Although Chung and Kim (1989) argued for the distinct taxonomic status of P. ellacombeanus, it is uncertain whether it should be recognized as a distinct taxon in Korea based on the current study. One accession collected from the type locality in Hakodate (Hokkaido, Japan) was embedded within the A. aizoon clade in the ITS phylogeny ( Figure 5 ). However, some accessions of P. ellacombeanus sampled, including those previously reported by Chung and Kim (1989), were positioned in various lineages within “clade 2.” In the plastome phylogeny ( Figure 4 ), the type locality P. ellacombeanus was sister to the clade of P. kamtschaticus–P. kurilensis from Russia, whereas two accessions sampled from Korea were closely related to P. zokuriensis and P. middendorffianus–P. kamtschaticus in Korea. The accessions from Korea occurred on seashores or sunny rock surfaces in forests on islands and tended to have broader spatulate leaves. However, we noticed that the leaf characteristics described in previous reports (Praeger, 1917; Praeger, 1921; Jacobsen, 1960; Clausen, 1975; Evans, 1985; Chung and Kim, 1989) were not observed in mature plants in the current study. As we cultivated P. ellacombeanus sampled from Geoje and Hachuja Islands at Sungkyunkwan University, we observed that young emerging stems tend to have opposite to subopposite very broad leaves with two to four crenate margins, which fits the general description of P. ellacombeanus by Chung and Kim (1989). In addition, two specimens cited by Chung and Kim (1989) as P. ellacombeanus from Geoje Island (SNU, 66782) and Chuja Island (SNU, 66785) are very similar to our collections, especially the young Geoje Island accession. Therefore, this study suggests that P. ellacombeanus, previously reported in Korea, may not truly represent P. ellacombeanus originally described from the type locality in Hakodate, Japan. Given their phylogenetic positions, P. ellacombeanus accessions sampled from Korea could be considered P. kamtschaticus sensu Korea (= P. aizoon var. floribundus sensu Japan). Since P. ellacombeanus was described as a new species based on cultivated materials and was originally collected as P. kamtschaticus by Maximowicz, it is necessary to investigate its taxonomic distinction from P. kamtschaticus based on broad sampling, especially from Hokkaido, northern Japan, and southern Kuriles, Russia.

Phedimus kurilensis was described by Voroshilov (1965) from the island of Kunashir and is considered endemic to southern Kuriles. Voroshilov later considered P. kurilensis to be a subspecies of P. sikokianus based on several characteristics (Ohba, 2002). However, P. sikokianus is endemic to the southern part of Japan, Shikoku, and occurs in the mountains (Ohba, 2002). Therefore, the distribution and habitat may not support the subspecies treatment of P. kurilensis as P. sikokianus. The same species of P. kurilensis is thought to occur in the northern part of the island of Hokkaido in Japan, perhaps under different names (personal observation, Marina Koldaeva). The ITS phylogeny in our study suggests that P. kurilensis is sister to the clade containing P. middendorffianus and P. takesimensis ( Figure 5 ). In addition, the plastome phylogeny suggests that P. kurilensis is sister to P. kamtschaticus, which was sampled from Hokkaido in our study (100% BS). In the Flora of Japan (Ohba, 2002), P. kurilensis is considered a synonym of P. kamtschaticus (Uhl and Moran, 1972). This species is known to occur in Kamchatka, the Kuriles, and Japan. The species description of P. kamtschaticus in the Flora of Japan (2002) is quite different from that of Korea, China, and Russia and from that by Hart and Bleij (2003). Although P. kurilensis and P. kamtschaticus appear to be closely related, it is uncertain whether P. kurilensis in the southern Kuriles and P. kamtschaticus in Hokkaido are conspecific based on species description and chromosome number. Based on the ITS phylogeny ( Figure 5 ), it seems that P. kurilensis is distinct from the rest of the P. kamtschaticus lineages from Korea and China. Given its close relationship with other species currently recognized (P. selskianus, P. sichotensis, P. middendorffianus, and P. takesimensis), P. kurilensis could maintain its distinct species status until other lines of evidence indicate otherwise.

Phedimus litoralis is endemic to the Ussuriysk floristic region (Hart and Bleij, 2003). We sequenced two additional accessions of P. litoralis from Primorsky Krai, and these accessions were closely related to accessions of P. aizoon, primarily sampled from Primorsky Krai (Russia) and one accession from Heilongjiang (China) ( Figure 5 ). In addition, plastome phylogeny suggested that P. litoralis is sister to P. aizoon sampled from Russia (100% BS). Therefore, as recently suggested by Hart and Bleij (2003), P. litoralis could be conspecific to P. aizoon or at least closely related to it. This P. aizoon lineage, including P. litoralis, appears to be distinct from other lineages of P. aizoon. Phedimus sichotensis, another Russian Far East endemic in southern Primorsky Krai, is almost indistinguishable from many small forms of the variable P. kamtschaticus and was considered a subspecies of P. middendorffianus (Gontcharova, 2000). ITS phylogeny showed that P. sichotensis was sister to the clade containing primarily P. kurilensis–P. middendorffianus–P. takesimensis (73% BS). Therefore, it is highly unlikely that P. sichotensis is related to P. kamtschaticus, but it is closely related to P. middendorffianus, which requires further confirmation based on a broader sampling.

Lastly, the taxonomic recognition of two widely occurring species, P. aizoon and P. kamtschaticus, as separate species or infraspecific levels could be problematic given their polyphyletic nature and close relationships with other more distinct species. Some distinct geographical lineages could be recognized for each species, while two species could be intermixed in certain clades, making the species boundaries of the two species difficult ( Figures 4 , 5 ). This could be due, in part, to different species descriptions depending on the country. In the Flora of Japan (2001), two infraspecific taxa were recognized for A. aizoon (var. aizoon and var. floribundus) based on the number of flowering stems and leaves: flowering stems fascicled and oblanceolate to lanceolate-ovate to ovate leaves with apically serrated leaf margins for var. floribundus versus flowering stems one or two and rhombic-elliptic to elliptic leaves with regularly serrated leaf margins, except near the base for var. aizoon. In Korea, these descriptions match those of P. kamtschaticus in the former (var. floribundus) and P. aizoon in the latter (var. aizoon). The species description of P. kamtschaticus by Hart and Bleij (2003) fits that of P. kamtschaticus in Korea and China and that of P. aizoon var. floribundus in Japan. Therefore, it is possible that plants with taller, one- to two-pronged stems and more uniform characters could be recognized as P. aizoon, whereas plants with smaller and more variable characters could be recognized as P. kamtschaticus or P. aizoon var. floribundus. The species description of P. kamtschaticus (Fisch. et C.A. Mey) in Japan seems quite different from that in Korea, China, Russia, and Hart and Bleij (2003). It is described as having numerous slender, fascicled flowering stems, 5–10 cm tall; small oblanceolate leaves (1–2 cm long and 0.6–1 cm wide) with irregularly serrate leaf margins; and carpels patent in fruits. It is known to occur in Hokkaido in Japan, southern Kuriles, and Kamchatka (Russia); it has a chromosome number of n = 16 (2n = 4x = 32, tetraploid) (Uhl and Moran, 1972). P. aizoon subsp. aizoon (= P. aizoon) is a highly polyploid species comprising an aneuploid series with 37 different chromosome numbers, ranging from 2n = 71 to 2n = 124 (2n = 12x = 96, dodecaploid, most frequent). In contrast, P. aizoon var. floribundus sensu Japan (=P. kamtschaticus sensu Korea, China, and Russia) is tetraploid (2n = 32), hexaploid (2n = 48), and octoploid (2n = 64) (Ohba, 1982; Amano, 1990; Amano and Ohba, 1992). Therefore, it is plausible that P. kamtschaticus sensu Japan (n = 16) is conspecific to P. aizoon var. floribundus sensu Japan. Owing to the existence of different ribotypes, we were unable to generate clean ITS sequences for accessions of P. kamtschaticus sampled from Hokkaido, Japan. However, the complete plastome sequence strongly suggests that P. kamtschaticus from Japan is closely related to P. kurilensis (considered a synonym of P. kamtschaticus sensu Korea, China, and Russia) and P. ellacombeanus (Hokkaido, Japan) and distantly related to any accessions of P. aizoon var. floribundus from Japan and P. aizoon accessions from Korea and Russia ( Figure 5 ). This suggests that P. kamtschaticus from Hokkaido and southern Kuriles could represent a distinct taxon from P. kamtschaticus, as currently recognized in Korea, China, and Japan.

In summary, the following relationships and diversification processes have been proposed based on complete plastome and nrDNA ITS sequences, pending independent confirmation based on genome-wide nuclear data. After the divergence of P. sikokianus and P. hybridus within the subgenus Aizoon during the mid-Oligocene, the P. aizoon and P. kamtschaticus lineages diverged further. It is hypothesized that certain distinct geographical lineages of each species have become narrowly occurring local endemics: P. takesimensis, P. zokuriensis, and P. latiovalifolium in Korea; P. litoralis and P. sichotensis in the Russian Far East. It is also highly plausible that species occurring in northern Japan and the Russian Far East (P. middendorffianus, P. sichotensis, P. selskianus, and P. kurilensis) shared their most recent common ancestor and contributed to the origin of the insular endemic P. takesimensis on Ulleung Island, Korea.