Following climate cooling at the end of the Tertiary, arctic-alpine plants attained most of their extant species diversity. Because East Asia was not heavily glaciated, the importance of this region as a location for the long-term persistence of these species and their subsequent endemism during the Pleistocene was proposed in early discussions of phytogeography. However, this hypothesis remains to be fully tested.
Here, we address this hypothesis by elucidating the phylogenetic history of Phyllodoce (Ericaceae).
A phylogenetic tree based on multiple nuclear loci revealed that Phyllodocenipponica was not derived from widespread species such as the arctic-alpine Phyllodocecaerulea, but rather represented an independent lineage sister to the clade of widespread relatives. Molecular dating indicated a mid-Pleistocene divergence of P. nipponica.
These findings exclude the hypothesis that P. nipponica was derived from an arctic-alpine species that extended its range southwards during recent glacial periods. Instead, our results support the hypothesis that P. nipponica is an ancestral species which persisted in the Japanese archipelago during the mid- and late Pleistocene. Our findings demonstrate support for the early proposal and shed light on the importance of the Japanese archipelago for the evolution and persistence of arctic-alpine species.
Arctic-alpine plants occupy enormous areas in the Arctic as well as high mountains in temperate regions of the Northern Hemisphere (Hultén & Fries, 1986). Following the development of a cooler climate at the end of the Tertiary, arctic-alpine plants attained most of their extant species diversity as well as widespread distributions (Matthews & Ovenden, 1990; Murray, 1995). Given that the Pleistocene was characterized by repeated cycles of glacial and interglacial periods, the evolutionary history of arctic-alpine plants has been largely influenced by climate oscillations (Abbott & Comes, 2004). Together with fossil evidence (Birks, 2008), molecular analyses have shed much light on the biogeography of the arctic-alpine flora (Abbott & Brochmann, 2003; Brochmann & Brysting, 2008). Because of the lack of extensive glaciations in East Asia during the Pleistocene (Fig. 1; Frenzel, 1968; Frenzel et al., 1992), this region is considered to be one of the most important areas for the long-term persistence of populations of arctic-alpine plants (Hultén, 1937). In contrast to their biogeographical history in the Arctic (Abbott & Comes, 2004; Alsos et al., 2005, 2007; Ronikier et al., 2012), the origins of arctic-alpine plants in East Asia have been less well elucidated (but see, Ikeda et al., 2012; DeChaine et al., 2013).
In East Asia, the distributions of arctic-alpine plants extend as far south as c. 35 N in the Japanese archipelago, where one-third of the taxa assigned to the arctic-alpine element, more than 100 species, are endemic (Shimizu, 1982, 1983). However, Japanese alpine communities are mostly dominated by species that are shared with the Arctic (e.g. Empetrum nigrum and Vaccinium vitis-idaea), suggesting that recent migration has resulted in the occurrence of arctic-alpine plants in this region. A previous study determined that a Japanese endemic, Cardamine nipponica, diverged from its arctic relative, Cardamine bellidifolia, during the penultimate interglacial period (c. 208 000 yr before present; Ikeda et al., 2012). This indicates that some Japanese endemic alpine plants evolved from arctic-alpine relatives that migrated southwards in the late Pleistocene. Furthermore, a recent phylogenetic study of Saxifraga sect. Trachyphyllum revealed that two species found in the Japanese archipelago (Saxifraganishdae and Saxifragarebunshirensis) were derived from species that occur in Beringia (DeChaine et al., 2013). Although these two species formed a unique lineage, most species in sect. Trachyphyllum harboured closely related chloroplast DNA (cpDNA) haplotypes, corroborating the recent divergence in the Japanese archipelago of alpine plants from widespread species.
However, considering the limited glaciation in East Asia during the Pleistocene and early proposals by Hultén (1937) based on phytogeography, it is feasible that at least some endemics in East Asia, especially in the Japanese archipelago, were not recently derived from arctic-alpine plants but rather persisted in isolation for a prolonged period during the Pleistocene. Phylogeographical studies of arctic-alpine plants have detected a unique genetic structure in populations from the high mountains of central Japan (Ikeda et al., 2008, 2009), demonstrating that isolated populations could have persisted in this region throughout several cycles of cold and warm periods. Given that altitude shifts could allow long-term population persistence in mountain regions (Hewitt, 2004), plants that originated from a higher latitude may have experienced complicated Pleistocene population dynamics in the Japanese archipelago. Thus, ancestral species may have persisted exclusively and then may have experienced subsequent isolation leading to the origin of endemic species. Accordingly, Hultén's (1937) hypothesis of East Asia as an important area for long-term isolation as well as endemism of arctic-alpine plants is plausible, although it has never been tested empirically.
Here we examined this hypothesis by investigating the phylogenetic history of a Japanese endemic alpine plant, Phyllodoce nipponica Makino (Ericaceae). Phyllodoce consists of about seven species of the Arctic and alpine areas in the circum-Pacific (Mabberley, 2008). Phyllodoce nipponica is exclusively distributed in the Japanese archipelago, especially in central Japan (Fig. 1a). Two widespread species, Phyllodoce aleutica and Phyllodoce caerulea, also grow in the archipelago; the former is found from central Japan to Alaska, while the latter has a distribution encompassing northern Japan, central Asia, and the Arctic (Fig. 1b; Hultén & Fries, 1986; Yamazaki, 1993). In North America, two widespread species, Phyllodoce empetriformis and Phyllodoce glanduliflora, are distributed from eastern Alaska to southern Canada, with the former encompassing northern California (Fig. 1b; Wallace, 1993). Their hybrid Phyllodoce × intermedia occurs wherever the two species are sympatric. In addition to these widespread species, Phyllodoce breweri occurs exclusively in interior mountain ranges of California (Fig. 1a; Wallace, 1993). Furthermore, a recent nuclear/plastid phylogeny of the Tribe Phyllodoceae revealed that a lineage of the genus Kalmiopsis, including two species (Kalmiopsis fragrans and Kalmiopsisleachiana) that are endemic to Oregon, USA (Fig. 1a), is closely related to Phyllodoce (Gillespie & Kron, 2013); that is, Phyllodoce and Kalmiopsis formed a well-supported monophyletic group, and were not reciprocally monophyletic. In particular, the phylogenetic study of Gillespie & Kron (2013) revealed that, based on cpDNA, these North American endemic species of Kalmiopsis represent basal lineages of the Phyllodoce–Kalmiopsis clade.
Given that the geographical distribution of P. nipponica is adjacent to that of P. caerulea and overlaps that of P. aleutica in Japan, this Japanese endemic may have diverged at the southernmost range margin of these widespread species following their recent southwards migration. However, P. aleutica and P. caerulea share an urn-shaped corolla with P. glanduliflora, whereas P. nipponica shares a bell-shaped corolla with P. empetriformis. This morphological similarity between allopatric P. nipponica and P. empetriformis suggests an alternative scenario, namely that P. nipponica may represent an ancient relict linage. In the phylogenetic study of Gillespie & Kron (2013), which included the Phyllodoce species mentioned above, the clade of P. nipponica and P. caerulea was highly supported by both cpDNA and nuclear DNA (nDNA) phylogeny. This suggests that P. nipponica diverged from widespread P. caerulea. However, the phylogenetic relationship of Phyllodoce species showed some discordance between cpDNA and nDNA sequence variation; that is, the monophyly of widespread species (P. aleutica, P. caerulea, P. empetriformis and P. glanduliflora) and P. nipponica was highly supported in the cpDNA phylogeny, whereas this monophyletic group collapsed in the nDNA phylogeny. This implies that incomplete lineage sorting or introgression may have influenced the estimation of the phylogenetic relationship among these species. In this case, analysing multiple individuals is essential to infer their phylogenetic relationships (Funk & Omland, 2003), whereas the previous study included only one individual per species (Gillespie & Kron, 2013). Accordingly, further investigation including multiple individuals, especially of P. nipponica, is required to confirm the previous phylogenetic position of P. nipponica.
In this study, we aimed to assess whether the occurrence of endemic arctic-alpine plants in East Asia represents not only recent southwards migration but also the persistence of ancient relicts in East Asia during the Pleistocene (Hultén, 1937). We first revealed the phylogenetic position of P. nipponica in the Ericaceae using cpDNA markers (rbcL and matK). We then addressed the phylogenetic history of P. nipponica using sequences of multiple nuclear loci. Together with four widespread species (P. aleutica, P. caerulea, P. empetriformis and P. glanduliflora), we included K. fragrans in the phylogenetic analyses as a representative of a basal taxon of the Phyllodoce–Kalmiopsis clade. Although we did not include P. breweri or K. leachiana in the analysis, our taxon sampling was sufficient to address whether P. nipponica represents an independent lineage from the widespread species or was derived from them, especially from P. caerulea as indicated in the previous study (Gillespie & Kron, 2013). Given that concatenating sequences of multiple loci could result in a less accurate inference of species relationships as a consequence of genealogical discordance across loci (Carstens & Knowles, 2007; Edwards et al., 2007; Kubatko & Degnan, 2007; Degnan & Rosenberg, 2009), we incorporated a multispecies coalescent model to infer species relationships (Heled & Drummond, 2010). Furthermore, after evaluating migration history between pairs of species, we discuss future perspectives regarding the evolutionary history of Phyllodoce species.
Materials and Methods
DNA samples, PCR and sequencing
Previously extracted DNA was used for 12 individuals of Phyllodoce nipponica Makino (Ikeda, 2012). In addition to P. nipponica, dried leaf material of Phyllodoce aleutica (Spreng.) A. Heller, Phyllodoce caerulea (L.) Bab., Phyllodoce empetriformis and (Sm.) D.Don Phyllodoce glanduliflora (Hook.) Coville was collected from wild populations around the Pacific, in which nine to 13 individuals from each species were analysed. DNA was freshly extracted from dried leaf material using a DNeasy Plant Mini Kit (Qiagen, Hilden, Germany). In addition to these widespread Phyllodoce species, leaf tissues of Kalmiopsis fragrans Meinke & Kaye were collected from three cultivated individuals that had originated from wild individuals. Epigaea asiatica Maxim. was used as the outgroup and its DNA was extracted from dried leaf tissues. Details of each sample are given in Supporting Information Table S1.
Sequences of matK and rbcL were determined for two to four randomly selected individuals of each Phyllodoce species as well as Epigaea asiatica, Loiseleuria procumbens (L.) Desv., Therorhodion camtschaticum (Pall.) Small and Cassiope lycopodioides (Pall.) D.Don. PCR was conducted with previously reported primers of rbcL (1F and 1367R) (Gillespie & Kron, 2010) and the reverse primer of matK (trnK-R2) (Johnson & Soltis, 1994; Gillespie & Kron, 2010), while a newly designed primer was used as the forward primer of matK (matK-F520) instead of trnK-582F (Table S2). PCR and sequencing procedures were conducted as previously reported using fast reaction polymerase (SapphireAmp® Fast PCR Master Mix; TaKaRa Bio Inc., Shiga, Japan) (Ikeda, 2012), but with the application of a longer extension time (20 s).
For nuclear loci, we first attempted to directly sequence 16 intron-flanking expressed sequence tag (EST) markers of Rhododendron (Wei et al., 2005) as well as the coding region of several phytochrome genes (Ikeda & Setoguchi, 2010) that were successfully sequenced previously in P. nipponica (Ikeda, 2012). According to the unambiguity of electropherograms, 12 loci (B1, B2, B4, B7, C11 C12, C16, C18, C20, PHYA, (PHYTOCHROME A), PHYB and PHYE) were selected for study. PCR products were directly sequenced from both directions using an ABI 3130 Avant Genetic Analyzer (POP-7 polymer and 50-cm capillary; Applied Biosystems, Foster City, CA, USA). When ambiguous electropherograms were obtained, TA cloning was conducted using a Mighty TA-cloning Reagent Set for PrimeSTAR (TaKaRa Bio). High-fidelity enzyme (PrimeSTAR® Max DNA Polymerase; TaKaRa Bio) was used for PCR for TA-cloning. The amplified products were purified by gel extraction, and then ligated into a pMD20-T vector followed by dA attachment using a Mighty TA-cloning Reagent Set for PrimeSTAR (TaKaRa Bio). After the blue-white selection of successful insertions, we sequenced several inserts using M-13 primers and determined at least two sequences that were responsible for production of ambiguous electropherograms.
Phylogenetic analysis based on cpDNA
Sequences of matK and rbcL from species other than Phyllodoce spp., E. asiatica, L. procumbens, T. camtschaticum and C. lycopodioides were downloaded from the National Center for Biotechnology Information (NCBI; Table S3). Enkianthus campanulatus (Ericaceae) was used as the outgroup to root the tree. Because few polymorphisms were detected within each species of Phyllodoce, a randomly selected sequence was used to construct the phylogeny except for the inclusion of two samples of P. nipponica. The underlying model of molecular evolution (GTR+G for both matK and rbcL) was determined and a maximum likelihood (ML) tree was estimated using treefinder (Jobb et al., 2004). The significance of the branches was evaluated based on 1000 bootstrap resamplings. A phylogenetic tree was also estimated using a Bayesian method with the program beast v. 1.7.4 (Drummond et al., 2012). By assuming the Yule speciation process and lognormal relaxed molecular clock model, five independent Markov chain Monte Carlo (MCMC) steps were run for 10 000 000 iterations and sampled every 1000 iterations. After discarding 1 000 000 iterations as burn-in and checking convergence using tracer ver. 1.5 (http://tree.bio.ed.ac.uk/software/tracer/), the five runs were combined into a single trace file using LogCombiner ver. 1.7.4, from which a phylogenetic tree was reconstructed using TreeAnnotator ver. 1.7.4.
Evaluating genetic admixture
Bayesian clustering was conducted based on combinations of alleles from multiple nuclear loci using structure ver. 2.1 (Pritchard et al., 2000; Falush et al., 2003). The probability of assigning individuals into clusters was estimated using an admixture model with the correlated allele frequency using 1000 000 generations following 500 000 burn-in periods. The number of clusters (K) was set to the number of species in the present study (K =6) and 20 runs were repeated. In addition, clustering was also conducted after excluding P. aleutica and P. empetriformis which showed apparent genetic admixture. The symmetric similarity coefficient (SCC; H′) was calculated between all pairs of runs for the same K and the final assignment of each individual was estimated as the mean overall configuration using clumpp (Jakobsson & Rosenberg, 2007).
Phylogenetic analysis based on nuclear DNA
Species trees of five Phyllodoce species as well as K. fragrans and E. asiatica based on sequences of nuclear loci were reconstructed using the ∗BEAST option in beast (Drummond & Rambaut, 2007; Heled & Drummond, 2010). Because introgression as well as genetic admixture mimics genealogies with incomplete lineage sorting and leads to inaccurate phylogenetic inferences, two individuals (Table S1) with apparent genetic admixture were excluded from the estimation of the species tree. To identify the influence of recombination on tree reconstruction, the entire sequence and also the longest sequence block showing no recombination were analysed separately for each locus. A four-gamete test (Hudson & Kaplan, 1985) was applied to explore recombination at each locus using DnaSP ver. 5.0 (Librado & Rozas, 2009). In preliminary analyses, we assumed all combinations of the prior speciation process (the birth-death or the Yule model), population size (constant, linear or constant and linear model) and two clock models (strict clock or lognormal relaxed clock) and MCMC searches were run for 100 000 000 iterations and sampled every 2000 iterations. After discarding 10 000 000 iterations as burn-in and checking convergence effective sample size (ESS > 200 for likelihood), each model was assessed against the others using an approximate Bayes factor (BF) comparison (Kass & Raftery, 1995; Suchard et al., 2001) with tracer ver. 1.5.
Across all combinations of models, either the birth-death or the Yule model of speciation, assuming constant population size prior and a lognormal relaxed clock, was better than the remaining models (2logeBF > 10) for analyses that included entire sequences and also the longest sequence block showing no recombination per locus. By contrast, the likelihoods of the birth-death and Yule models were similar (2logeBF < 6). Because the Yule model is simpler than the birth-death model, we used it for the speciation process. Under the final model (the Yule speciation, constant population size prior, and lognormal relaxed clock), four independent MCMC searches were run under the same conditions as in the preliminary analysis. After checking convergence using tracer ver. 1.5., all runs were combined into a single file using LogCombiner ver. 1.7, from which a phylogenetic tree was reconstructed using TreeAnnotator ver. 1.7.4.
Speciation history incorporating genetic admixture among five Phyllodoce species
The isolation with migration (IM) model was used to infer the approximate timing of the divergence of P. nipponica and to examine genetic admixture among widespread species whose phylogenetic relationships were less resolved. The IM model incorporates parameters of the population sizes of extant species as well as ancestral species (θ), their divergence times (t), and migration rates between pairs of species including ancestral species (m) expressed in terms of a mutation rate (μ). Probability distributions of these demographic parameters were estimated using the program IMa2 (Hey, 2010). Because the phylogenetic relationships among the widespread species (P. caerulea, P. empetriformis, P. aleutica and P. glanduliflora) were unclear, we assumed all possible pairs of species. To estimate the divergence time of P. nipponica (Analysis 1), P. caerulea, P. empetriformis, P. aleutica and P. glanduliflora were all treated separately as sister species of P. nipponica. To evaluate migration history (Analysis 2), parameter estimations were conducted for all possible pairs of the four widespread species. In all cases, the demographic parameters were estimated from 10 000 000 MCMC steps following a 100 000-generation burn-in period with eight to 15 heated chains under the geometric increment model. Three independent runs were conducted, from which 100 000 genealogies were saved from each run. Prior probability densities were optimized based on several preliminary runs using a wide range of densities (m =0.1 (the mean of the exponential prior distribution); q =1.5–10; t =1.0–4.0). Using all of the saved genealogies, the marginal posterior distribution, maximum likelihood estimates (MLEs) and 95% highest probability densities (HPDs) of each parameter were obtained. The estimated divergence time was scaled using the geometric mean of substitution rates per locus based on 5.3–7.8 × 10−9 substitutions per site per year (Ossowski et al., 2010). Although this range was based on a report of the experimental estimation of substitution rates in Arabidopsis thaliana, it was similar to an earlier estimation based on the divergence between monocots and dicots (5.8–8.1 × 10−9 substitutions per site per year; Wolfe et al., 1987). Thus, we used this range to infer the approximate timing of divergence of P. nipponica.
Phylogenetic position of Phyllodoce nipponica based on cpDNA
The phylogenetic tree based on concatenated sequences of rbcL (alignment length = 1247 bp) and matK (alignment length = 1024 bp) indicates that the genus Phyllodoce formed a highly supported clade that included K. fragrans both in the ML (Fig. 2; bootstrap = 98) and in the Bayesian tree (topology not shown; posterior probability (pp) = 1.00). Most branches within this clade were not well supported (Fig. 2; bootstrap < 95 or pp < 0.95). The two samples of P. nipponica analysed in this study formed their own clade (Fig. 2; bootstrap/pp = 100/1.00), while the published sequence for P. nipponica formed a supported clade with P. caerulea (Fig. 2; bootstrap/pp = 99/1.00).
Genetic structure among species
Twenty replicate runs of the Bayesian clustering that assigned individuals to six clusters (K =6) revealed a consistent result (H′ = 1.00; Fig. 3a). One individual of each of P. aleutica and P. empetriformis were placed in two different clusters with nearly equal probabilities: these two individuals were collected from an area where both P. aleutica and P. caerulea and both P. empetriformis and P. glanduliflora occurred, respectively (Table S1). This suggests that these two individuals are likely to be F1 hybrids of the corresponding species. By excluding these putative hybrids, a consistent clustering of individuals of the same species was obtained among replicate runs (H′ = 1.00), in which each Phyllodoce species and K. fragrans was unambiguously distinguishable (Fig. 3b).
Species relationship based on phylogenetic analyses using multilocus sequences
Based on both entire sequences and partial sequences without recombination, the species tree using E. asiatica as the outgroup showed that P. nipponica formed a clade with the four widespread Phyllodoce species (Fig. 4; pp = 0.99/1.00). In addition, P. nipponica was sister to the monophyletic group of the remaining four species (Fig. 4; pp = 0.98/0.83). By contrast, the relationships among P. caerulea, P. empetriformis, P. aleutica and P. glanduliflora were not resolved clearly (pp < 0.4) for sequences either with or without recombination (Fig. 4b,c).
Demographic history of Phyllodoce
The MLEs of the divergence time of P. nipponica from each of the widespread species were older than those of each pair of widespread species (Analysis 1: Tlow = 162–390 thousand yr (ka), Thigh = 238–574 ka; Analysis 2: Tlow = 91–139 ka, Thigh = 134–204 ka; Table 1), while 95% HPD values showed overlap between them (Analysis 1: Tlow = 67–761 ka, Thigh = 99–1120 ka; Analysis 2: Tlow = 33–673 ka, Thigh = 49–990 ka; Table 1). Regarding P. nipponica (Analysis 1), the population sizes of ancestral species were smaller than those of the extant species except for the case of P. empetriformis (MLE: θ0 = 0.176, θ1 = 0.257, θA = 1.595; Table 1). By contrast, population sizes of ancestral species were larger than those of the extant species across the widespread species (Analysis 2) except for the pair of P. aleutica and P. caerulea (MLE: θ0 = 0.397, θ1 = 0.257, θA = 0.003; Table 1). The MLEs of the migration rates were mostly low (m =0.001), which means that the highest peak of the posterior probability distributions was at m =0, indicating that no migration occurred between the pairs. By contrast, for the pairs of P. aleutica with P. caerulea, P. empetriformis or P. glanduliflora, posterior distributions of migration rates peaked at more than zero (m >0; Table 1), indicating that gene exchange had occurred after speciation.
Table 1. Maximum likelihood estimate (MLE) and 95% highest probability density (HPD) of demographic parameters of the isolation with migration (IM) model
m0 > 1
m1 > 0
t, divergence time of the pairs of species; Tlow and Thigh, scaled divergence time based on slower (5.3 × 10−9 substitutions per site per year) and faster (7.8 × 10−9 substitutions per site per year) substitution rates, respectively (ka; thousand yr); θ0 and θ1, population sizes of former and latter species in each pair, respectively; θA, population size of an ancestral species of the pairs of species; m0 > 1 and m1 > 0, migration rates backwards in time from former to latter and latter to former species, respectively.
Phyllodoce aleutica versus P. nipponica
P. caerulea versus P. nipponica
P. empetriformis versus P. nipponica
P. glanduliflora versus P. nipponica
P. aleutica versus P. caerulea
P. aleutica versus P. empetriformis
P. aleutica versus P. glanduliflora
P. caerulea versus P. empetriformis
P. caerulea versus P. glanduliflora
P. empetriformis versus P. glanduliflora
A previous phylogenetic study based on cpDNA and nuclear sequence variation revealed that P. nipponica, an endemic to the Japanese archipelago, formed a clade with P. caerulea and represented a rather derived taxon in Phyllodoce (Gillespie & Kron, 2013). In the present study, the clade containing P. caerulea and other widespread Phyllodoce species did not include P. nipponica, whose sequences were newly determined (Fig. 2). Incomplete lineage sorting in P. nipponica, the occurrence of introgression from P. nipponica to P. caerulea or artificial errors could result in P. nipponica exhibiting polyphyly. However, the samples of P. nipponica sequenced in the present study are representative of those used in previous phylogeographical studies (Ikeda & Setoguchi, 2007, 2013), where no specifically divergent cpDNA haplotypes were detected across c. 160 samples from most of its range. In addition, the samples lacked any signature of genetic admixture with other species (Fig. 3). Therefore, we consider that the present samples of P. nipponica have yielded sequences that correctly reveal the phylogenetic position of this species.
Sequences of multiple nuclear loci unambiguously showed that P. nipponica represents a lineage distinct from either of the four widespread Phyllodoce species or K. fragrans (Fig. 4). Although the phylogenetic relationships among the widespread species were not resolved unambiguously, their clade was sister to P. nipponica, indicating that P. nipponica was not derived from either P. caerulea or other widespread species. This finding precludes the possibility that P. nipponica is an endemic species that was derived from P. caerulea or P. aleutica at the southern margins of their ranges in East Asia. Instead, our results support the hypothesis that the persistence of an ancestral species in the Japanese archipelago led to the divergence of P. nipponica.
According to the fossil record, some extant arctic-alpine species were also present in the early Pleistocene and survived throughout the Pleistocene (Matthews & Ovenden, 1990; Birks, 2008). Therefore, the divergence of P. nipponica might be dated to the Tertiary or early Pleistocene. However, contrary to such an early origin hypothesis, the estimated divergence time indicates a mid-Pleistocene divergence of P. nipponica from each of its widespread relatives (MLE = 162–574 ka; 95% HPD 67–1120 ka; Table 1). Given that this estimation assumed a substitution rate (5.3–7.8 × 10−9) based on A. thaliana (Ossowski et al., 2010) and the general rate in angiosperms (Wolfe et al., 1987), a more specific rate is required to estimate accurately the date of divergence of P. nipponica. In addition to the substitution rate, there is a caveat regarding the IM model that we applied to estimate the divergence time. Because the precise phylogenetic relationship was not resolved among the widespread species, we applied the model assuming a sister relationship between P. nipponica and each of the widespread species. This pairwise approach violated the assumption of the IM model that assumes a speciation history between the most closely related species (Nielsen & Wakeley, 2001; Hey & Nielsen, 2004). Nevertheless, the species pairs represented a part of the genealogy between P. nipponica and the clade of the widespread species. Because we applied plausible pairs of species, the present estimations should capture an approximate timing of the divergence of P. nipponica. As the climate is known to have changed extensively during the mid- and late Pleistocene (c. 900 ka) (Gibbard & van Kolfschoten, 2004), range shifts would have occurred more extensively than during previous periods. Thus, extensive changes in distribution during the mid- and late Pleistocene may have been involved in the colonization as well as the endemism of arctic-alpine plants in East Asia.
A previous study of C. nipponica inferred a more recent divergence history for this Japanese endemic alpine plant (c. 208 ka; Ikeda et al., 2012) and an even more recent origin was inferred for S. nishdae and S. rebunshirensis (c. 70 ka; DeChaine et al., 2013). Thus, the timing of the southwards migration of arctic-alpine plants into East Asia and the subsequent divergence of endemics, especially in the Japanese archipelago, varies among species. Contrary to the previous suggestion of a recent divergence from widespread species, our study found that the Japanese endemic P. nipponica represents an independent lineage from its widespread relatives distributed in East Asia, northern North America and the Arctic. This finding suggests that population persistence and subsequent endemism of arctic-alpine plants have occurred in the Japanese archipelago, the southernmost part of their ranges in East Asia. Consequently, the results of our molecular investigation provide evidence in support of the long-standing hypothesis of the evolution of arctic-alpine plants that some species persisted as ancient relicts in the less glaciated region of East Asia over a long period of time during the Pleistocene (Hultén, 1937).
Clearly, the phylogenetic and biogeographical history of Phyllodoce needs further investigation. First, because the present study did not include P. breweri and K. leachiana, both endemics to southern parts of North America, the basal divergence of members of the Phyllodoce–Kalmiopsis clade as well as their dispersal history between East Asia and North America remains to be resolved. The more basal position of K. fragrans relative to P. nipponica enabled us to hypothesize that P. nipponica probably originated from ancestral species in North America that dispersed into East Asia as suggested for Saxifraga sect. Trachyphyllum species (DeChaine et al., 2013).
In addition to the basal species relationships within Phyllodoce, the phylogenetic history among its widespread species was not resolved (Fig. 4). With the caveats discussed above regarding divergence time, these widespread species could plausibly have originated recently (MLE = 91–204 ka; 95% HPD 26–1507 ka; Table 1). The extensive climate change in the late Pleistocene (Gibbard & van Kolfschoten, 2004), leading to range shifts as well as geographical isolation, might have facilitated their allopatric speciation. Because including or excluding recombination within sequences was irrelevant to the ambiguity of phylogenetic relationships, forces other than recombination, such as incomplete lineage sorting across loci and/or migration between species, may have caused the low resolution. The present species trees were estimated by incorporating multispecies coalescence and by applying multiple loci. Therefore, incomplete lineage sorting alone is unlikely to explain the ambiguous relationships. Alternatively, gene exchange between divergent species following Pleistocene range shifts could have influenced phylogenetic relationships among these widespread species. Indeed, migration between some pairs of species was detected (Table 1). Although these migration rates might not be estimated correctly because of the violation of the IM model mentioned above, gene exchange after speciation could plausibly have complicated the genealogies among these relatives. Further analyses including not only multiple samples of all Phyllodoce and Kalmiopsis species but also phylogenetic inferences incorporating post-divergence gene flow are necessary to unravel the entire phylogenetic history of the Phyllodoce–Kalmiopsis clade and to obtain a complete picture of its biogeographical history.
We thank Ken Marr for editing the English of this paper, commenting on it and collecting samples in Alaska and Canada, N. Hashimoto for supporting DNA experiments, M. Robert and K. Amsberry for kindly providing leaf tissues of Kalmiopsis fragrans, the Editor and anonymous referees for their helpful comments, the October Hill Foundation for providing funds to collect samples from Alaska, and the National Museum of Nature and Science for funding provided to H.I. In addition, this work was supported by JSPS KAKENHI grant nos. 22405013 and 23657015.