•The Quaternary climatic changes resulted in range shifts of species, providing chances for hybridization. However, the genetic signatures of such ancient introgression have rarely been reported. To investigate such signatures, we performed a phylogeographical study on the perennial plant Veratrum album ssp. oxysepalum, which may have hybridized long ago with another congeneric species, V. stamineum.
•Sequence variations in chloroplast DNA (cpDNA) were examined in 43 populations in Japan and adjacent areas. Phylogenetic analyses of different cpDNA haplotypes were conducted on the basis of cpDNA and nuclear ribosomal internal transcribed spacer (nrITS) variations.
•In the Japanese archipelago, two major groups of haplotypes were detected, one of which was distributed in a disjunct pattern. The major haplotype, occupying the central part of the species’ distribution, formed a monophyletic group with V. stamineum in phylogenetic trees on the basis of cpDNA variation, although the two species did not form a monophyletic group in phylogenetic trees on the basis of nrITS variation.
•Historical hybridization between V. album ssp. oxysepalum and V. stamineum in refugia during the Quaternary climatic oscillations, and the resulting chloroplast capture of V. stamineum by V. album ssp. oxysepalum, are most probably responsible for the disjunct distribution of cpDNA in V. album ssp. oxysepalum.
Over the last decade, molecular phylogeographical appro-aches have generally supported the idea that climate oscillations during the Quaternary period influenced the current distributions of plant species (Taberlet et al., 1998; Hewitt, 2000). Many phylogeographical studies, in particular of European tree species, have revealed the existence of refugia in the last glacial era, and postglacial expansion routes to northern temperate zones from these refugia (Petit et al., 2003b; Grive et al., 2006; Magri et al., 2006; Svenning et al., 2008).
These phylogeographical studies have indicated genetic consequences, such as the paucity of genetic variation in northern areas to which the species have spread in recent times, and the existence of many rare haplotypes in putative refugium regions of plant species experiencing expansion and contraction of their ranges. Range shifts resulting from expansions and contractions of plant species’ distributions caused by climatic oscillations would have provided the opportunity for contact between species that occurred allopatrically in the past, and would have been expected to result in hybridization between them in suture zones (Heuertz et al., 2006; Dixon et al., 2007; Nettel et al., 2008; Nordström & Hedrén, 2008; Prentice et al., 2008). Patterns of chloroplast DNA (cpDNA) variation in plant species can sometimes provide a genetic signature of the occurrence of past introgressive hybridization (Rieseberg & Soltis, 1991; Rieseberg & Welch, 2002).
One conspicuous example of introgressive hybridization revealed by cpDNA variation concerns the European deciduous oaks, Quercus robur and Quercus petraea, which share the same cpDNA haplotypes in many stands over a wide range, suggesting that these species hybridized with each other during their northward expansion into Europe after the last ice age (Kremer et al., 2002; Petit et al., 2003a). When genetic markers specific to one species are found in another species with a remote distribution, gene transfer events between the two species are assumed to have occurred in a sympatric or parapatric phase in their history (Alvarez & Wendel, 2006; Besnard et al., 2007). Furthermore, we can infer phylogeographical patterns, such as range contractions or expansions, from evidence of ancient introgression having occurred between two species. For example, Dixon et al. (2007) used evidence of chloroplast capture to show the past presence of Androsace in areas in which it has subsequently become extinct.
The Japanese archipelago was connected to the Eurasian continent by four landbridges as a result of the lower sea level during the last glacial era: Sakhalin, the Kuriles, the Korean peninsula and the Ryukyu Islands. The flora of Japan contains many plant species common to the Eurasian continent, and these species appear to have migrated from the continent into Japan via the landbridges (Hotta, 1974). Thus, to investigate the phylogeography of these species with widespread distributions between the Japanese archipelago and the Eurasian continent, we must include the intraspecific populations located peripherally to the Japanese archipelago. However, such studies of plant species have rarely been performed.
Veratrum album L. ssp. oxysepalum (Turcz.) Hultén (Melanthiaceae) is widely distributed in the Kuriles, the Kamchatka Peninsula, Sakhalin, Far East Russia, the Korean peninsula, the northern part of China and Japan (Takahashi, 2010). This species is very widespread from the north to the south of Japan (Fig. 1a, a′), and occurs primarily in the understories of cool temperate forests or marshy meadows, especially in the north of Japan. Another species, Veratrum stamineum Maxim., also occurs in Japan. Veratrum stamineum is distinct from V. album ssp. oxysepalum, and many differences in morphological characters exist between them, such as stamen form (upright in the lower part and outwardly spreading in V. album ssp. oxysepalum and V. stamineum, respectively), perianth color (pale green vs white, respectively) and tepal size (6–18 mm long and 4–8 mm wide vs 4–8 mm long and 3–5 mm wide, respectively). Neither species has any specialized mechanism for seed dispersal. Veratrum album ssp. oxysepalum is pollinated by small insects belonging to Coleoptera and Diptera (Kato et al., 2009).
Paleopalynological data (Tsukada, 1983) and a few phylogeographical studies (Fujii et al., 2002; Okaura & Harada, 2002; Takahashi et al., 2005) have indicated that temperate forests in Japan have expanded from refugia. Recolonization by the cool temperate understory herb V. album ssp. oxysepalum probably accompanied these temperate forest movements. By contrast, V. stamineum, an intrinsically alpine herb, probably had a wider distribution in the glacial era and, during the postglacial period, retreated to high-altitude habitats or specific lowland habitats.
Veratrum album ssp. oxysepalum rarely co-occurs with V. stamineum, except in the northern limit of the distribution of V. stamineum, because V. stamineum var. stamineum normally inhabits alpine meadows and V. album ssp. oxysepalum grows in understories at lower altitudes even in the same mountain region. Consequently, the habitats of these two species are separated spatially, although putative hybrid individuals between the two species are sometimes recorded (Takahashi, 2010).
Kato et al. (1996) investigated the phylogeny of Veratrum species in Japan using restriction fragment length polymorphism (RFLP) of cpDNA. They found that V. maackii is sister to the clade composed of V. album ssp. oxysepalum and V. stamineum with very high bootstrap support. They also showed that the populations of V. album ssp. oxysepalum are separated into two clades: one clade (Group II) contains only V. album ssp. oxysepalum populations, whereas the other (Group I) contains eight populations of V. stamineum and seven populations of V. album ssp. oxysepalum.Kato et al. (1996) proposed two hypotheses to explain the results: (1) introgressive hybridization had occurred between V. stamineum and V. album var. oxysepalum; or (2) V. stamineum had diverged recently from V. album ssp. oxysepalum. However, they analyzed a small number of populations of V. album ssp. oxysepalum and only one individual per population. Furthermore, they examined only three individuals from the Eurasian continent (two from the Korean peninsula and one from the Eurasian continent). Thus, more comprehensive analyses are needed to test their hypotheses. If introgressive hybridization between V. album ssp. oxysepalum and V. stamineum occurred in the past and is not ongoing at present, we may be able to determine the phylogeographical patterns of the species by determining the ranges in which the symptoms of ancient introgression appear.
In this study, we focused on the geographical pattern of cpDNA sequence variations in V. album ssp. oxysepalum to infer the phylogeography of the species, and considered its relationship with V. stamineum. Although many technical advances in molecular biology, such as genetic markers [e.g. amplified fragment length polymorphism (AFLP) and simple sequence repeats (SSRs)], have been used in phylogeographical studies, intraspecific cpDNA variation is still very useful for examining population structures and the phylogeography of species, because it can infer historical processes, such as refugial isolation, genetic divergence, founder events and migration routes (Comes & Kadereit, 1998; Stehlik et al., 2002). To address the shortcomings of cpDNA in revealing only maternal lineages, we also used biparentally inherited internal transcribed spacer (ITS) of nuclear ribosomal DNA (nrDNA) variation following a previous study (Wang et al., 2009). We analyzed 43 populations covering almost the whole distribution of V. album ssp. oxysepalum in the Japanese archipelago and some Eurasian continental populations located near Japan, in addition to some populations of V. stamineum. During the course of the present study, we found that the main haplotype group of cpDNA was disjunctively distributed in V. album var. oxysepalum in the Japanese archipelago. On the basis of additional phylogenetic and phylogeographical analyses of other Veratrum species using nuclear ribosomal ITS (nrITS) variation, we discuss the geographical pattern of cpDNA found in V. album ssp. oxysepalum from the viewpoint of phylogeography and ancient introgressive hybridization.
Materials and Methods
Leaf samples of V. album ssp. oxysepalum were collected from 10 individuals in each of 33 populations in the Japanese archipelago and seven populations in the Korean peninsula. We sampled each individual at intervals of a few meters apart to prevent the collection of the same genets. Three Russian populations of V. album ssp. oxysepalum were also examined. Except for the Russian populations, leaf materials were kept in an ultracold refrigerator (−70°C) in Ziplock plastic bags until DNA extractions were carried out. For the Russian populations, DNA was extracted from dried specimens of 10 individuals from Okhotsk and one individual from the Kamchatka Peninsula population (Fig. 1a, a′, Supporting Information Table S1), and a DNA solution extracted from an individual of a population in Vladivostok was provided by Dr T. Kawahara, AFFRI, Japan. Similarly, leaf samples of V. stamineum (sensu lato) were collected from 10 individuals from each of seven populations in the Japanese archipelago. The samples were kept in an ultracold refrigerator, as above. Among the populations sampled in this study, only two populations of V. album ssp. oxysepalum and V. stamineum (Populations A1 and S44) were located near each other.
DNA extraction and sequencing
Total genomic DNA was isolated from 50–100 mg of leaf tissues based on the method of Maki et al. (2002). PCR amplification was conducted in a total reaction volume of 15 μl containing 10–20 ng of total DNA, each primer at 0.15 μM, 0.1 mM deoxynucleoside triphosphates (dNTPs), 50 mM KCl, 2 mM MgCl2, 10 mM Tris-HCl (pH 8.3) and 0.375 units of Taq DNA polymerase (Amplicon Inc., Irvine, CA, USA). Double-stranded DNA was amplified after incubation at 94°C for 3 min, followed by 30 cycles of incubation at 94°C for 30 s, 55°C for 30 s and 72°C for 30 s, with a final extension at 72°C for 15 min. To examine the geographical distribution of cpDNA in V. album ssp. oxysepalum and V. stamineum, we amplified the two noncoding regions of cpDNA, the trnS(GCU)–trnG(UCC) spacer and the trnL(UAA) intron, using the primers reported by Hamilton (1999) and Taberlet et al. (1991), which had been successfully amplified and shown to exhibit some variations in preliminary experiments. After amplification, the PCR products were purified using a Gene Clean Kit II (MP Biomedicals, Irvine, CA, USA). We sequenced the purified PCR products using a Big-Dye™ Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Carlsbad, CA, USA) and analyzed these on a Genetic Analyzer Model 3100 (Applied Biosystems) according to the manufacturer’s protocol. Sequencing was conducted from both ends using the same primers as in PCR. For all minor variants, we performed the sequencing at least twice.
For phylogenetic analyses, we amplified the following two noncoding regions of cpDNA, the trnL(UAA)–trnF(GAA) spacer and the atpB–rbcL spacer, and nrDNA ITS regions. PCR and sequencing were performed according to the method of Liao et al. (2007). We used different cpDNA regions from the above phylogeographical analysis to incorporate the samples examined in the present study into datasets generated for a previous phylogenetic study of other Veratrum species (Liao et al., 2007). Liao et al. (2007) treated V. album and V. oxysepalum as separate taxa, although the latter seems to correspond to V. album ssp. oxysepalum on the basis of the sampling locality. In this study, we tentatively followed the taxonomic treatment of Liao et al. (2007) for their samples when analyzing the data. Although the genus Veratrum comprises c. 27 species from around the world, in the phylogenetic analyses, we used our data and data for an additional 13 species from Liao et al. (2007), for which both cpDNA and nrITS sequences were available.
In V. album ssp. oxysepalum, we sequenced the above regions in several individuals with major cpDNA haplotypes and at least one individual with the minor cpDNA haplotype (Table S1). In V. stamineum, five individuals per haplotype covering the species’ entire distribution were sequenced (Table S1). For sequencing, we used the same primers as employed for amplification. All sequences have been deposited in the DDBJ/EMBL/GenBank databases (AB514874–AB514928).
PCR-RFLP of nrITS
To obtain information about the nuclear genomes of the populations examined for cpDNA, we performed PCR-RFLP of nrITS on the same populations and individuals based on the sequence differences of nrITS determined in this study. The PCR conditions were the same as above, and the amplification products (2.5 μl) were digested by 2.5 units of DdeI (Toyobo, Tokyo, Japan) in a total reaction volume of 5 μl at 37°C for 1 h. The digested DNAs were electrophoresed in 1% agarose gel with ethidium bromide and observed under a UV transilluminator. After digestion, four fragments (657, 120, 53 and 23 bp; ribotype 1) and three fragments (777, 53 and 23 bp; ribotype 2) were expected to be generated in the cp-group I of V. album ssp. oxysepalum and V. stamineum, respectively (Fig. S3). Thus, we could discriminate between ribotypes by checking band numbers and sizes.
All sequences were aligned with ClustalW (Thompson et al., 1994), and ambiguously aligned regions, caused by indels, were corrected manually. The cpDNA haplotypes were determined on the basis of these aligned sequences, and a haplotype network based on statistical parsimony was created to evaluate possible relationships between the haplotypes using TCS 1.06 (Clement et al., 2000). All indels were treated as point mutations and evenly weighted with other mutations. Population relationships were determined using neighbor-joining (NJ) trees (Saitou & Nei, 1987), obtained with PHYLIP ver. 3.68 (Felsenstein, 2005). The Euclidean distance was used to estimate divergence among haplotypes, and bootstrap support was estimated with 1000 replicates using PHYLIP ver. 3.68 (Felsenstein, 2005).
To determine the phylogeographical structure of V. album ssp. oxysepalum, spatial analyses of molecular variance (SAMOVAs) were performed to identify groups of populations that are geographically homogeneous and maximally differentiated from each other using SAMOVA 1.0 (Dupanloup et al., 2002). This simulates different partitions of n populations into K groups and retains the partition with the highest FCT value (i.e. the proportion of genetic variation among groups). Assuming that the final configuration was influenced by the initial configuration, 100 initial conditions were used as recommended by Dupanloup et al. (2002). Values of K in the range 2–8 were tested. Although K with the highest FCT represents the best number of groups and the best population configuration, it does not consistently represent a significant configuration. In particular, the final configuration of K with one or more single population groups could not derive the group structure (Heuertz et al., 2004). Thus, the number of groups (K) and the geographical structure were inferred from the configuration with the highest FCT that did not contain any single population group. The simple analysis of molecular variance (AMOVA) was used for the case without a group level (Excoffier et al., 1992).
Based on the method described by Pons & Petit (1995, 1996), measures of genetic diversity and differentiation were estimated using the program PERMUT ver. 1.0 (http://www.pierroton.inra.fr/genetics/labo/Software/PermutCpSSR/index.html). Genetic diversity and differentiation were determined in two ways: either the genetic similarity between haplotypes was taken into account (vs, vt, NST) or only frequencies were used (hs, ht, GST). The two differentiation parameters (NST and GST) were compared by a permutation test using 1000 permutations. All parameters were calculated for the populations in the Japanese archipelago (Populations A1–A33) and separately for the Japanese populations belonging to the two major clusters in the NJ tree.
To infer phylogenetic relationships among individuals with different cpDNA haplotypes in other Veratrum species, we performed maximum parsimony (MP) and maximum likelihood (ML) analyses on the basis of sequence data of cpDNA and nrDNA ITS regions, in accordance with the method of Liao et al. (2007), by incorporating our samples into their dataset.
MP analyses were carried out using PAUP* ver. 4.0b10 (Swofford, 2002) through a heuristic search with tree bisection–reconnection (TBR) branch swapping and the MULTREES option. Multiple islands of equally most parsimonious trees were searched for using the heuristic option with 1000 random sequence additions. To estimate the confidence levels of monophyletic groups, the bootstrap method with 1000 replications was used as above (Felsenstein, 1985).
ML analyses were conducted using PAUP* ver. 4.0b10 (Swofford, 2002), applying the best-fit molecular evolution model and parameters determined using Modeltest version 3.7 (Posada & Crandall, 1998) with the Akaike information criterion (AIC) (Akaike, 1974). For the ML search, multiple accessions of the same taxon with identical sequences were reduced to one, because the inclusion of identical sequences in the search would require an extremely long calculation time. The best-fit models for cpDNA and ITS regions were K81uf+I and TIM, respectively. Based on the parameter values suggested by Modeltest, a heuristic search using the MP tree as the starting tree was run with PAUP* ver. 4.0b10 (random addition sequence with 10 replicates, TBR branch swapping and MULTREES on; Swofford, 2002). Clade supports were assessed by nonparametric bootstrapping using 1000 replicates with a heuristic search strategy [random sequence addition with 10 replicates, nearest neighbour interchange (NNI) branch swapping].
Distributions of cpDNA haplotypes and nrITS ribotypes in V. album ssp. oxysepalum and V. stamineum
The sequences of two noncoding regions (1350–1377 bp) of cpDNA in a total of 412 individuals from 43 populations revealed 17 haplotypes of V. album ssp. oxysepalum. By contrast, only three haplotypes (haplotypes N, R and S) were found in a total of 70 individuals from seven populations of V. stamineum (Fig. 1b, Table S1). The alignments of regions containing indels are shown in Fig. S1. The TCS haplotype network (Fig. 1c), based on mutational steps, showed a clear grouping pattern with two main groups: cp-group I and cp-group II. These two groups were separated from each other by 12 mutational steps.
cp-Group I consisted of 13 haplotypes differing from adjacent ones by one or two mutation(s); the most distinct haplotypes were separated by five steps. In the Japanese archipelago, six haplotypes (haplotypes A–F) were in cp-group I, and these haplotypes occurred disjunctively in the southwestern part of the main island, Honshu, and most of the northernmost island, Hokkaido (Fig. 1a′, b). This group only contained the cpDNA haplotypes of V. album ssp. oxysepalum. By contrast, cp-group II consisted of six haplotypes distinguished from adjacent ones by one or two mutation(s); the most distinct haplotypes were separated by four steps. Of the six haplotypes, three (haplotypes O, P, Q) and two (haplotypes R, S) were specific to V. album ssp. oxysepalum and V. stamineum, respectively. The remaining one (haplotype N) was widely shared by both species. In V. album ssp. oxysepalum, these haplotypes were distributed from the center of the Japanese main island, Honshu, to the southernmost end of Hokkaido (Fig. 1a′). Most of the populations in the Japanese archipelago were fixed by one haplotype, although some populations comprised more than two haplotypes.
In the Korean peninsula, on Jeju Island (A40) and in Far East Russia, all populations were composed of haplotypes belonging to cp-group I. In the Korean peninsula, all populations consisted primarily of haplotypes unique to the peninsula, which were separated by one to five steps from the haplotypes found in Japan, although a few populations contained haplotypes common to Japan. On Jeju Island (A40), located between the Korean peninsula and the Japanese archipelago, the population was fixed for the haplotype (haplotype A) found predominantly in Hokkaido and the southwestern part of Japan. In Far East Russia (A41, A43), the populations also consisted of unique haplotypes, which were separated by one to four steps from the haplotypes of Japan, whereas the population on the Kamchatka Peninsula (A42) consisted of the same haplotype (haplotype A in Fig. 1c) found in the Japanese archipelago.
The NJ tree (Fig. S2) based on Euclidean distances contained two distinct major clusters (Cluster I and Cluster II) and one population (Population A11), comprising haplotypes of both major clusters, as expected from the TCS network of the cpDNA haplotypes and the distribution of haplotypes in populations (Fig. 1c, Table S1). Clusters I and II were composed of populations consisting of individuals with haplotypes of cp-group I and individuals with haplotypes of cp-group II, respectively. Population A11 contained individuals with haplotypes of both cp-group I and cp-group II.
The results of SAMOVA are summarized in Table 1. The number of groups with the highest FCT that included no single population group was four (K = 4). The first group contained populations with haplotypes of cp-group I in Hokkaido (A1–A6), Kyushu (A32, A33), the Korean peninsula (A36), Jeju Island (A40) and Far East Russia (A41–A43). The second group consisted of two populations on Hokkaido (A7 and A10) and seven populations (A25–A31) in the southwestern part of the Japanese archipelago, and most of the individuals in these populations had haplotype C, which belongs to cp-group I. The third group consisted of five populations (A34, A35, A37–A39) on the Korean peninsula. The last group consisted of 16 populations (A8, A9 and A11–A24), which primarily had haplotypes of cp-group II. Most of the variation (91.6%, P <0.001) was explained by differences among groups. Overall, this configuration was consistent with the population clusters depicted in the NJ tree (Fig. 2).
Table 1. Results of spatial analyses of molecular variance (AMOVA and SAMOVAs) of Veratrum album ssp. oxysepalum
Source of variation
Percentage of variation (%)
The final configurations with the highest FCT that included no single population group (K = 4) are shown.
The measures of genetic diversity and differentiation are shown in Table 2. The genetic diversity values in Cluster I in the NJ tree were larger than those in Cluster II. NST was significantly larger than GST for all populations of the Japanese archipelago, indicating a phylogeographical structure in the archipelago as a whole, although trends, such as NST > GST, were not found in separate analyses for Clusters I and II.
Table 2. Genetic diversity and differentiation parameters for the population in the Japanese archipelago and within two major clusters of the neighbor-joining tree in the Japanese population of Veratrum album ssp. oxysepalum
Standard errors are shown in parentheses.
*NST is significantly larger than GST (P < 0.05).
Cluster I [A1–A7, A10, A25–A33]
Cluster II [A8, A9, A12–A24]
Figure S3 shows the results of the PCR-RFLP of nrITS. All populations of V. album ssp. oxysepalum had the same ribotype (ribotype 1), whereas those of V. stamineum had another ribotype (ribotype 2), except for one population (S44) that included a heterozygous individual with both ribotypes.
Veratrum species were separated into two major clades in cpDNA and ITS trees using ML (Clade A and Clade B; Fig. 2). Individuals of both V. album ssp. oxysepalum and V. stamineum were included in the same clade (Clade A) in both trees (Fig. 2). The same results were obtained from phylogenetic trees constructed using the MP method.
The cpDNA trees supported the relationships between individuals with cp-group I and cp-group II haplotypes shown in the TCS network. The individuals of cp-group I formed a monophyletic group (Clade A1) with V. lobelianum, V. album (sensuLiao et al., 2007), V. oxysepalum Turcz. (sensuLiao et al., 2007), V. grandiflorum and V. dahuricum. By contrast, individuals of cp-group II formed another monophyletic group (Clade A2) composed of the individuals examined in the present study (Fig. 2a). The latter monophyletic group (Clade A2) contained individuals of both V. album ssp. oxysepalum and V. stamineum.
As in the cpDNA trees, two distinct monophyletic groups (Clade A1 and Clade A2) were also found in ML trees based on the nrDNA ITS sequences (Fig. 2b). In contrast with the cpDNA trees, the individuals with the same cp-group in the TCS network did not form a monophyletic group. Clade A1 included V. album ssp. oxysepalum individuals of both cp-group I and cp-group II, and did not contain individuals of V. stamineum. By contrast, another monophyletic group (Clade A2) contained only individuals of V. stamineum, all of which had haplotypes belonging to cp-group II. The same results were obtained from the phylogenetic trees using the MP method. Thus, phylogenetic patterns were incongruent between the cpDNA trees and ITS trees.
Disjunct distribution of two major haplotype groups in V. album ssp. oxysepalum in the Japanese archipelago
Kato et al. (1996) reported two monophyletic groups among populations of V. album ssp. oxysepalum in an MP tree based on cpDNA restriction site variations. We also found two major haplotype groups of cpDNA (cp-group I and cp-group II) by network analysis of sequence variations in two noncoding regions of cpDNA (Fig. 1c). As Kato et al. (1996) examined only three populations in the northern part of the Japanese archipelago, the geographical pattern of the haplotypes they resolved was not clearly determined. In the present study, we found a clear disjunctive distribution pattern of cpDNA haplotypes of V. album ssp. oxysepalum (Fig. 1a, a′, Table S1) with one major haplotype group occurring in the central part of Honshu and a limited part of Hokkaido (cp-group II); the other major haplotype was distributed in the southwestern part of the Japanese archipelago and most of Hokkaido (cp-group I).
Two hypotheses were proposed by Kato et al. (1996) to explain the disjunct distribution of cpDNA haplotypes found in V. album ssp. oxysepalum. The first is that hybridization with V. stamineum resulted in the capture of V. stamineum cpDNA by V. album ssp. oxysepalum. Our phylogenetic analyses of the two focal species and other Veratrum species support this hypothesis, in that incongruent patterns were found between the cpDNA trees and ITS trees constructed (Fig. 2). According to this introgression hypothesis, gene exchange may have occurred by current and/or ancient hybridization between V. album ssp. oxysepalum and V. stamineum. As these two species rarely occur together, except in the northernmost part of Hokkaido, and putative hybrids are found infrequently and are almost sterile (H. Takahashi, unpublished data), current hybridization between these two species seems unlikely to have played a major role in causing the disjunct distribution of cpDNA haplotype groups found in this study. In addition, even if occasional interspecific hybridization currently occurs in some populations of V. album ssp. oxysepalum following long-distance seed dispersal of V. stamineum, this would not explain why all individuals in the central–north range of V. album ssp. oxysepalum have the cp-group II haplotype, even though asymmetrical hybridization is relatively common in plants (Cruzan & Arnold, 1994; Caraway et al., 2001; Wu & Campbell, 2005; Minder et al., 2007; Tagane et al., 2008). Furthermore, in the admixed populations of these two species (Populations A1 and S44), all individuals of V. album ssp. oxysepalum examined possessed the cpDNA haplotype belonging to cp-group I, and there was no evidence of chloroplast capture by current hybridization.
An alternative scenario is that hybridization and chloroplast capture may have occurred once or a few times when these two species came into contact with each other in refugia during the last glacial era. Then, subsequently, V. album ssp. oxysepalum with the cpDNA haplotype(s) of V. stamineum (cp-group II) spread widely in the northern part of Honshu and the southernmost part of Hokkaido. This scenario is partly supported by the observation that one major haplotype (haplotype N) was overwhelmingly dominant in the range occupied by cp-group II, and that the populations consisting mainly of individuals with cp-group II had less genetic diversity than those with cp-group I (Table 2). Coalescent simulation based on the isolation with migration model suggests that the minimum time since splitting of the cpDNA groups almost corresponds to the last interglacial period (see Notes S1).
Another hypothesis proposed by Kato et al. (1996) is that V. stamineum originated from the populations located in the central parts of the Japanese archipelago and rapidly diverged with adaptation to alpine meadow environments. Some plant species’ groups are known to have diverged very rapidly and show wide morphological and ecological variations (Hodges & Arnold, 1994; Richardson et al., 2001; Yamashiro et al., 2004). However, the phylogenetic tree based on ITS sequences is not consistent with this hypothesis, because ITS ribotypes of V. stamineum and V. album ssp. oxysepalum formed different monophyletic groups, each of which was supported by a high bootstrap value.
Incongruence of cpDNA and ITS patterns
Incongruence between nuclear and cytoplasmic DNA phylogenies is common in plants and animals (e.g. Nevado et al., 2009; Acosta & Premoli, 2010), and such patterns may be indicative of past hybridization events (Rieseberg & Soltis, 1991; Fehrer et al., 2007). However, only a few studies have examined the symptoms of ancient introgression throughout a species’ range and have revealed that some populations show such traces (Dixon et al., 2007). In this study, we found traces of ancient introgression only in populations of V. album ssp. oxysepalum from central Japan, and infer from this the existence of glacial refugia in this part of Japan in the past. Thus, evidence of ancient hybridization between the two species indicates historical range contraction and subsequent expansion.
Although a few cpDNA haplotypes were shared between V. album ssp. oxysepalum and V. stamineum, nrITS delimited these two species throughout their geographical distributions. Recent theoretical and empirical studies (Currat et al., 2008; Du et al., 2009; Petit & Excoffier, 2009) have suggested that rates of introgression should be inversely related to the intraspecific gene flow level, and this appears to be the case in the present study, because gene flow via cpDNA occurs only by seeds as a result of its maternal inheritance, whereas nuclear DNA is biparentally inherited and its gene flow occurs by both seeds and pollen.
Although the deep divergence between the two cp groups suggests that ancient introgression occurred between the two species, we cannot fully exclude the possibility that incomplete lineage sorting was the cause of shared cpDNA haplotypes between them. Genes and genomes with high dispersal rates should be sorted more rapidly than those with low dispersal rates (Hoelzer, 1997), and a recent study of two pine species showed that cpDNA lineage sorting had been accelerated because of its higher dispersability, as compared with mtDNA (Zhou et al., 2010). In the present study, this was also the case, in that biparentally inherited nrITS showed a species-specific pattern, whereas maternally inherited cpDNA had haplotypes shared between two species.
Biogeographical history of V. album ssp. oxysepalum in Japan and its adjacent area
The intrapopulational variation in cpDNA haplotypes was relatively small in V. album var. oxysepalum, with only 10 of 43 populations examined containing more than one cpDNA haplotype. Only one population included individuals with the cp-group I haplotype as well as individuals with the cp-group II haplotype (Population A11). This population is located near the geographical border of the ranges occupied by cp-group I and cp-group II, and may have been formed by individuals of both cp-groups following range expansion. Although the geographical border of the ranges of cp-group I and cp-group II is rather clear, no distinct geographical barrier exists at either the southern or northern ends. With climate warming after the last glacial era, both cp-type populations are likely to have expanded their distributions from refugia and come into contact with each other at these borders. As no reproductive isolation was found between plants containing the two major cpDNA haplotypes (see Notes S2), the haplotypes belonging to the major groups are likely to coexist in the same populations in the future if unidirectional hybridization does not occur between them.
In this study, we examined seven and three populations of V. album ssp. oxysepalum in the Korean peninsula and Far East Russia, respectively. The populations adjacent to Japan have cpDNA haplotypes belonging to cp-group I, which showed a disjunct distribution in Hokkaido and the southwestern part of the Japanese archipelago. These results indicate that populations of V. album ssp. oxysepalum in Japan migrated from the Eurasian continent across the landbridge(s) in the glacial era(s), although migration in the reverse direction is also possible. How many migration routes existed and how many times V. album ssp. oxysepalum migrated from the Eurasian continent remain uncertain, because the cpDNA haplotypes are infrequently shared between the populations in Japan and those in adjacent areas.
If the disjunct distribution of cpDNA haplotypes found in this study was caused by ancient introgression in refugia during the glacial era as discussed above, V. album ssp. oxysepalum may have migrated to the Japanese archipelago from the Eurasian continent via one or two landbridge(s) (e.g. via the Korean peninsula and/or the Sakhalin landbridges) and expanded throughout the archipelago before the last glaciation. This would mean that more than three refugia would have existed during the last glaciation, and introgression and resultant cpDNA captures would have occurred in one or more refugia in the north–central part of the Japanese archipelago (Fig. 3).
As the different haplotypes occur mostly between Japan and the adjacent areas, and the same cpDNA haplotypes are shared between disjunct areas of the Japanese archipelago, geographical isolation among the populations in Japan and the adjacent areas was most probably established after migration of V. album ssp. oxysepalum into Japan, although this could not have occurred a long time ago because the haplotypes differ by only one or two steps between the Japanese and the continental populations (Fig. 1c). That is, these cpDNA differences probably arose during the Quaternary climatic oscillations.
Few comparable phylogeographical studies have been carried out on forest plant species distributed in cool temperate regions of the Japanese archipelago and adjacent areas. Okaura et al. (2007) examined the geographical distribution of cpDNA haplotypes in populations of four oak species in Japan and the Eurasian continent. They found that the same haplotypes were shared among plants in Japan, South Korea and inland China, and suggested that the haplotypes had diversified before the last glacial period. More data are needed to clarify the role of the landbridges during the ice ages as migration routes to Japan for temperate forest plant species.
We are very grateful to all of the following for sampling materials, comments and technical advice: I. Dohzono, Y. Horii, H. Ikeda, T. Itagaki, K-S. Jeong, T. Kawahara, T. Kobayashi, M. Maruoka, S. I. Morinaga, H. Takahashi, K. Nishida, Y. Valentin, T. Yamashiro and H. Yamaji. This study was partly supported by a Grant-in-Aid from the Japan Society for the Promotion of Science to M.M.