Contrasting evolutionary processes during Quaternary climatic changes and historical orogenies: a case study of the Japanese endemic primroses Primula sect. Reinii

Abstract

Background and Aims

Recent biogeographic studies have illustrated that Quaternary climatic changes and historical orogenies have driven the development of high levels of biodiversity. In this context, phylogenetic niche conservatism may play a role as a major precursor of allopatric speciation. However, the effects of niche evolution on the diversification patterns of plant species under rapid habitat changes are still unknown. Here, Primula section Reinii, one of the few primroses endemic to the Japanese Archipelago, was investigated. This study aimed to clarify the phylogenetic position and relationships of section Reinii, interpret the biogeographic and diversification patterns of this group and gain a better understanding of the role of climatic niche evolution in the Japanese endemic primroses.

Methods

Dated phylogeny for Primula section Reinii is presented based on the sequences of six chloroplast genes and one nuclear gene. Biogeographic history was reconstructed using statistical dispersal–vicariance analysis. Macro-evolutionary modelling of the climatic niche was combined with biogeographic inferences.

Key Results

Section Reinii was shown to be monophyletic based on chloroplast and nuclear sequences. Fossil-calibrated dating analysis estimated that this section diverged from its sister taxon, sect. Cortusoides, around 1.82 million years ago, and intraspecific diversification occurred within the last million years. This time frame was characterized by Quaternary climatic oscillations and uplift of high mountains in Japan. Biogeographic inference suggested that this section originated at the northern end of the Japanese Archipelago and then dispersed southward to other islands. Models of climatic niche evolution indicated that the closely related species P. reinii and P. tosanensis have contrasting niche optima and rates of niche evolution.

Conclusion

Our results highlight that spatiotemporal heterogeneity in the Japanese islands may play a significant role in the biogeographic history of Japanese endemic primroses. Contrasting evolutionary processes found in closely related species illustrate the effects of climatic niche evolution on species’ diversification patterns.

INTRODUCTION

Biodiversity hotspots around the world are concentrated in mountain and island areas (Myers et al., 2000). Determining what processes generated the biodiversity in these hotspots is an essential aim of studies in evolutionary biology, ecology and conservation biology (Evans et al., 2009). Recent biogeographic studies based on molecular data have illustrated that Quaternary climatic change and historical orogenies served as major drivers of high biodiversity (e.g. Richardson et al., 2001; Hoorn et al., 2010; Wen et al., 2014; Favre et al., 2015; Mastretta-Yanes et al., 2015; Yang et al., 2016). The alteration of topography and past climatic changes associated with mountain uplift can cause fragmentation of the distribution ranges of species, which can lead to reduced gene flow between isolated populations (Ren et al., 2016). This process can lead to allopatric divergence and eventually speciation (Rice and Hostert, 1993), which may cause historical orogenies to harbour high levels of biodiversity.

The dynamics of species’ distribution ranges tend to be constrained by phylogenetic niche conservatism, i.e. the tendency of species to retain similar ecological niches over evolutionary time scales (Peterson et al., 1999; Wiens, 2004; Wiens and Donoghue, 2004). Wiens (2004) argued that phylogenetic niche conservatism, manifested by the failure of a species to adapt to novel ecological conditions, is an important component of the initial splitting of allopatric lineages. The importance of phylogenetic niche conservatism in speciation has been supported by many studies (e.g. Pyron and Wiens, 2013). Thus, phylogenetic niche conservatism appears to be a major precursor to allopatric speciation during Quaternary climatic perturbations. On the other hand, another model predicted that niche evolution (i.e. the expansion of niche breadth or specialization for new conditions) rather than niche conservatism might have facilitated speciation during climatic oscillations (Hua and Wiens, 2013). Recent biogeographic studies have also suggested that climatic niche evolution played a significant role in the speciation and diversification patterns of many groups (e.g. Graham et al., 2004; Pfenninger et al., 2007; Evans et al., 2009; Ren et al., 2015). Therefore, a better understanding of speciation is needed; in particular, the role of climatic niche evolution and diversification patterns under climate oscillation and historical orogenies must be elucidated. Previous studies into the effects of historical orogenesis and climatic change on diversification have been conducted in continental mountain chains, including the Andes (Hoorn et al., 2010) and the Qinghai–Tibet Plateau (QTP) (Fravre et al., 2015), but this relationship remains poorly understood on islands (but see Merckx et al., 2015).

The Japanese Archipelago comprises four main islands (Hokkaido, Honshu, Shikoku and Kyusyu) located on the eastern edge of the Eurasian continent (Fig. 1A) and represents one of the world’s 34 biodiversity hotspots (Mittermeier et al., 2005). The orogenies of the archipelago occurred mainly during the last few million years (Yonekura et al., 2001). Several high mountains in central Honshu experienced rapid uplift during the Quaternary (Yonekura et al., 2001; Fig. 1B). The uplift of the mountain ranges had a profound effect on the biota of the Japanese Archipelago, as it dramatically altered the climate, leading to the largest snowfalls in the world on the Japan Sea side and a dry climate regime on the Pacific Ocean side (Uemura, 1989; Gansert, 2004). Thus, the archipelago harbours multiple climate regimes, and the modern spatial patterns of plant diversity have been filtered by topography, precipitation and temperature (Kubota et al., 2014). Previous molecular phylogeographic studies have indicated that past geo-climatic events in the archipelago sculptured intraspecific genetic differentiation and contributed to the establishment of high biodiversity among plants (e.g. Fujii and Senni, 2006), insects (Tojo et al., 2017) and vertebrates (Yoshikawa et al., 2008) in the region. Nevertheless, the effects of climatic niche evolution on speciation and diversification patterns in this region over evolutionary time scales are not well understood. A practicable framework based on macro-evolutionary modelling combined with biogeographic analysis introduced by Ren et al. (2015) may provide insight into the relationship between niche evolution and diversification patterns.

Fig. 1.

(A) Location and topography of the Japanese Archipelago. The four main islands, Hokkaido, Honshu, Shikoku and Kyushu, are shown in bold. (B) Altitude-through-time curve for the Japanese mountains (modified after Yonekura et al., 2001).

Here, we investigated the biogeographic history of Primula sect. Reinii (Primulaceae), which comprises only four species endemic to Japan (Richards, 2003). These primroses occur on all four main islands of the archipelago, mainly on wet, shaded rocks (usually basic and/or ultramafic) on mountains on the Pacific Ocean side (Fig. 2). Due to the low migration and dispersal abilities characteristic of edaphic endemic plants (e.g. Damschen et al., 2012), these primroses are thought to have been strongly affected by geo-climatic events, mainly the formation and fragmentation of mountainous regions. Thus, this plant group is an ideal arena for evaluating the relationship between geo-climatic events and climatic niche evolution in the biodiversity hotspot of Japan.

Fig. 2.

(A) Presumed distribution range of Primula section Reinii species. Black shadows represent the shape and relative size of leaves for all eight taxa. (B) P. takedana, (C) P. hidakana, (D) P. reinii and (E) P. tosaensis. All scale bars correspond to 2 cm.

The phylogenetic position of sect. Reinii within the Primulaceae remains unresolved. Pollen morphology and chromosome number suggest that sect. Reinii should be attributed to subgenus Auganthus, which encompasses most primrose species found in eastern Asia (Richards, 2003). Recent molecular phylogenetic studies have suggested that sect. Reinii is closely related to sects Cortusoides and Bullatae (Mast et al., 2001; Richards, 2003), although the statistical support for these relationships was very low. Additionally, the monophyly of sect. Reinii is also uncertain. Taxa in sect. Reinii are diploid (2n = 24) perennial herbs (Richards, 2003) and have been classified into four species and five varieties based on morphological characteristics including the shape of the leaf, trichome, corolla and rhizome: P. takedana Tatew, P. hidakana Miyabe & Kudo, P. reinii Franchet & Savat, with varieties reinii, myogiensis Hara, kitadakensis (Hara) Ohwi, rhodotricha (Nakai & Maek) Yamaz and finally P. tosaensis Yatabe with varieties tosaensis and brachycarpa (Hara) Ohwi. The distribution ranges of these taxa are allopatric (Fig. 2): the geographic distributions of P. reinii and P. tosaensis are sympatric in the central Honshu and Kii mountains, but their ranges can be segregated according to their habitat altitude, approx. 1500–2800 m and 500–1500 (2000) m, respectively. Nevertheless, their taxonomic circumscription has also remained unclear to date. For instance, local endemic taxa within the P. reinii complex (var. reinii, kitadakensis, myogiensis and rhodotricha) are classified based on leaf shape (Fig. 2), which can lead to confusion because primrose leaves exhibit great variation in the wild. Although hybridization abilities among Reinii species have not been reported, intrasectional hybridization in the wild is common among Primula (e.g. Woodell et al., 1969; Zhu et al., 2009; Ma et al., 2014). Such hybridization events among Primula frequently led to incongruence between species and gene trees (e.g. Ren et al., 2015). Therefore, to investigate the influence of climatic niche evolution on speciation and diversification patterns in this primrose group, it is necessary to elucidate their phylogenetic position within Primula and the relationships among species.

The aims of this study are to (1) elucidate the phylogenetic relationships within sect. Reinii using multiple chloroplast and nuclear DNA sequences; (2) assess the effects of historical orogenies and associated climatic changes during the Quaternary on the biogeographic history of this section; and (3) better understand the role of climatic niche evolution in the diversification and speciation of the Japanese endemic primrose group.

MATERIALS AND METHODS

Taxon sampling and sequencing

We collected a total of 38 samples representing all eight taxa in Primula sect. Reinii (Supplementary Data Table S1). These samples contained multiple individuals per taxon, and widespread species were collected over their entire distribution ranges. Additionally, a total of 12 samples representing three species, P. sieboldii, P. jesoana and P. kisoana in sect. Cortusoides, which grow in Japan, were included as outgroup samples based on the large-scale phylogenetic tree of Primula developed in recent studies (Mast et al., 2006; Yan et al., 2010; de Vos et al., 2014). Genomic DNA was isolated from silica gel-dried leaves using the cetyltrimethyl ammonium bromide (CTAB) method (Doyle, 1990) and stored at 4 °C until use.

Seven chloroplast DNA regions (rbcL, matK, rps16, trnH–psbA, trnL–F, trnT–L and trnD–T) and an internal transcribed spacer (ITS) region of nuclear ribosomal DNA were amplified and sequenced using published primers (White et al., 1990; Taberlet et al., 1991; Olmstead et al., 1992; Ooi et al., 1995; Oxelman et al., 1997; Sang et al., 1997; Anderberg et al., 1998; Honjo et al., 2004). The PCR mix contained 10× Ex Taq Buffer (Takara Ex Taq; Takara, Kusatsu, Japan), 0.2 mm dNTP (Takara), 0.25 U of Taq polymerase (Takara), 0.2 μm of each primer, sterile water and 60 ng of template DNA in a total volume of 10 μL. The PCR amplification sequence was an initial 5 min of denaturation at 94 °C and 35 cycles consisting of 30 s of denaturation at 94 °C, 30 s of annealing at 50 °C and a 1 min extension at 72 °C, with a final 10 min extension at 72 °C. After amplification, PCR products were purified using ExoStar (GE Healthcare Ltd, Little Chalfont, Buckinghamshire, UK). All sequencing reactions were performed using the BigDye Terminator Cycle Sequence Ready Reaction Kit ver. 3.1 (Applied Biosystems, Foster City, CA, USA) and sequenced on an ABI Prism 3100 genetic analyser (Applied Biosystems). All sequences were checked and assembled on an AutoAssembler (Applied Biosystems) and aligned with Muscle in MEGA5 (Tamura et al., 2011) using default settings with minor manual adjustment. All sequences were submitted to the DNA database GenBank/DDBJ/EMBL under accession numbers LC198584–LC198618 and LC199515–LC199864.

Phylogenetic reconstruction

For phylogenetic reconstruction, three data sets were prepared for this study, the first of which was used to assess the phylogenetic position and monophyly of sect. Reinii. Three chloroplast regions (matK, rps16 and trnL–F) from a single sample per species in sect. Reinii were used to create the first data set. Three chloroplast gene sequences in 35 accessions from other related taxa (subgen. Auganthus, subgen. Carolinella, subgen. Aleurita, subgen. Auriculastrum, subgen. Pinnatae and subgen. Primula) were extracted from GenBank (Supplementary Data Table S2) following the most recent classification of Primula (Richards, 2003). Based on a recent large-scale phylogenetic study of Primula (Mast et al., 2006), Androsace alpina and Soldanella sublanata were also included in the first data set as outgroups. The second data set, consisting of six chloroplast genes (matK, rps16, trnH–psbA, trnL–F, trnT–L and trnD–T), was used to elucidate phylogenetic relationships within sect. Reinii. The third data set was comprised of the ITS region of nuclear DNA. Other closely related taxa (P. jesoana, P. sieboldii and P. kisoana) were included as outgroups in the second and third data sets. All chloroplast regions were combined into a single sequence using SequenceMatrix v1.7.8 (Vaidya et al., 2011).

For maximum likelihood (ML) and Bayesian inference (BI) analyses, the best nucleotide substitution model for the three data sets was selected using the Akaike information criterion (AIC) by jModelTest v2.1.4 (Darriba et al., 2012). For the first data set, the GTR + G model was selected for all DNA regions. Meanwhile, for the second and third data sets, the GTR + I + G model was selected for trnH–psbA, trnL–F and the ITS; the GTR + G model was selected for trnD–T and trnT–L; the GTR + I model was selected for matK; and the HKY + I + G model was selected for rps16. ML analyses were performed with PhyML v3.0 (Guindon et al., 2009) using the nearest-neighbour I=interchange (NNI) branch-swapping algorithm and 10³ bootstrap replicates to estimate node support values. BI analyses were conducted using the Markov chain Monte Carlo (MCMC) method implemented in BEAST v1.8.3 (Drummond and Rambaut, 2007). Each MCMC was run for 10⁸ generations, and trees were sampled every 1000th generation. The first 10 % of trees were discarded as burn-in, and convergence was checked by Tracer v1.6 (Rambaut and Drummond, 2009). A maximum clade credibility (MCC) tree was reconstructed using TreeAnnotator in BEAST.

Divergence time estimation

We used the chloroplast data set to estimate divergence time using a secondary calibration method, as performed in Ren et al. (2015). Since age estimations obtained from secondary calibrations are subject to bias and error (Sauquet et al., 2012; Sauquet, 2013), we addressed this concern by specifying prior probability densities for the root age estimation. We avoided overdiscussion of estimated dates in integrating these dates and geo-climatic events. Divergence time analysis was carried out in BEAST. The fossil record of Primula riosiae from 15.97 million years ago (Mya) during the Miocene (Czaja, 2003) was used to calibrate the divergence of Primula and Soldanella. The root age of sect. Reinii was estimated in an independent dating analysis from a more taxonomically comprehensive four gene (rbcL, matK, rps16 and trnL–F), 15 taxon chloroplast DNA sequence data set (Supplementary Data Table S3). These sequences were downloaded from GenBank and aligned using Muscle in MEGA5. Total alignment length was 4303 positions, of which 340 were parsimony informative, 647 were variable but parsimony uninformative, and 3316 were constant characters. Nucleotide substitution model selection was performed as described above. The GTR + I model was selected for rps16 and trnL–F, the GTR + G model was selected for matK, and the GTR + I + G model was selected for rbcL. The large-scale dating analysis and fossil calibration were conducted in accordance with the method of Ren et al. (2015). After analysis, the first 10 % of trees were discarded as burn-in, and the root age distribution for sect. Reinii was estimated using Tracer.

Secondary calibration was conducted based on the root age estimate of sect. Reinii obtained from large-scale dating analysis. The calibration point for the root age of sect. Reinii was set to a γ prior, with a shape of 6.0, a scale of 0.38 and an offset of 0.27. We applied the same partition delimitation and nucleotide substitution model for each chloroplast region used in BI analysis. MCMC analysis was run for 10⁸ generations and sampled every 1000 generations using an uncorrelated log-normal relaxed clock model and a birth–death tree. After the initial 10 % of trees were discarded as burn-in, an MCC tree with mean divergence times and a 95 % highest posterior density (HPD) interval on nodes was illustrated using TreeAnnotator. FigTree v1.4.2 (Rambaut, 2009) was used to display the MCC tree and divergence times.

Ancestral area reconstruction

Statistical dispersal–vicariance analysis (S-DIVA) was performed using RASP v3.2 (Yu et al., 2015) to estimate the ancestral areas and biogeographic events of sect. Reinii. We used the BI phylogenetic trees constructed from our second data set and defined five geographic areas for all samples, i.e. region A, Hokkaido region; region B, Sea of Japan region, the Sea of Japan side of Honshu including the Hida mountain range; region C, Kanto region, the Pacific Ocean side of Honshu including the Kanto Mountains; region D, central Honshu region, the Pacific Ocean side of central Honshu including the Kiso, Akaishi and Tanzawa mountains; and region E, south-western region, including the Pacific Ocean side of south-western Honshu, Shikoku and Kyushu (see Figs 1 and 4). These boundaries were defined with reference to the classical biogeography of Japanese plant species (Hotta, 1974; Maekawa, 1998) and were roughly justified by several phylogeographic studies (e.g. Iwasaki et al., 2012). To account for phylogenetic uncertainly, we used 20 000 randomly chosen trees from BEAST output data for S-DIVA analysis. The analysis was performed using default settings, and the maximum number of areas for all nodes was set to two.

Fig. 4.

Divergence time, ancestral range and historical biogeographic event estimates for the Japanese endemic primrose section Reinii. Node ages and biogeographic inferences were estimated based on the chloroplast data set using BEAST (Drummond and Rambaut, 2007) and statistical dispersal–vicariance analysis (S-DIVA) implemented in RASP (Yu et al., 2015), respectively. The grey bars on the nodes indicate the 95 % highest posterior density. The map represents biogeographic regions used for S-DIVA analysis. Pie charts at internal nodes represent the marginal probabilities for each possible ancestral area.

Evolution of niche preferences

We tested for differences in the evolution of climatic niche preferences among sect. Reinii species using plastid phylogenies and multiple models implemented in the R package OUwie (Beaulieu et al., 2012). We focused on the three major clades within sect. Reinii: hidakana, tosaensis and reinii clades (Fig. 3). Nineteen bioclimatic variables were used as continuous traits for this test. These variables were extracted from WorldClim (http://www.worldclim.org/current; Hijmans et al., 2005) using the point sampling tool in Quantum GIS v2.14 (http://qgis.osgeo.org). Principal component analysis (PCA) was performed using the R statistical program; this method summarizes species’ climatic parameters and climatic scores on two principal component axes (i.e. PC1 and PC2) for model comparison.

Fig. 3.

Maximum clade credibility tree of section Reinii derived from Bayesian inference analysis. (A) A plastid tree based on six chloroplast genes: matK, rps16, trnL–F, trnH–psbA, trnT–L and trnD–T. (B) A nuclear tree based on the internal transcribed spacer (ITS) sequence. Maximum likelihood bootstrap values and Bayesian posterior probabilities (PP) are shown above the major branches. Thicker branches indicate bootstrap values ≥80 and PP ≥90. Outgroup species are indicated in bold.

We fitted the seven following unique macro-evolutionary models using the R package OUwie (Beaulieu and O’Meara, 2012): (1) a Brownian motion (BM) model of evolution with single rate parameter σ² (BM1; Felsenstein, 1985); (2) a BM model with group-specific rate parameters (BMS; O’Meara et al., 2006); (3) an Ornstein–Uhlenbeck (OU) model with a single parameter for the rate (σ²) and strength of selection (α) towards an optimal value without being affected by different groups (OU1; Hansen, 1997); (4) an OU model with separate optimum parameters (θ) for each group (OUM; Butler and King, 2004); (5) an expanded OU model in which both the rate (σ²) and optima (θ) parameters could differ between groups (OUMV; Beaulieu et al., 2012); (6) an OU model that assumes different groups with different optima (θ) and strength of selection (α) (OUMA; Beaulieu et al., 2012); and (7) a more expanded OU model in which all three parameters [optima (θ), rate (σ²) and strength of selection (α)] could differ between groups (OUMVA; Beaulieu et al., 2012). To account for uncertainly in the estimated values, stochastic mapping for all model tests was run ten times for each of the 100 trees sampled randomly from the BEAST posterior distribution of trees using the R package Phytools (Revell, 2012). Finally, we calculated the average Akaike weights (AICw; Burnham and Anderson, 2002) and the lower (2.5 %) and upper (97.5 %) quantiles of the AICw distribution for each model to determine the best fit model, defined as the model with the highest AICw values.

RESULTS

Phylogenetic reconstruction

Sequence characteristics of the ITS and chloroplast DNA (matK, rps16, trnH–psbA, trnL–F, trnT–L and trnD–T) in each data set are summarized in Supplementary Data Table S4. In the first data set, the total aligned length of chloroplast DNA was 3256 bp, containing 1094 variable sites. For the second and third data sets, the total aligned lengths of chloroplast and ITS sequences were 5760 and 796 bp, containing 442 and 86 variable sites, respectively.

The phylogenetic trees reconstructed in this study clarified the phylogenetic position and monophyly of Primula sect. Reinii. The maximum clade credibility tree reconstructed from the first data set is shown in Supplementary Data Fig. S1. The tree topologies obtained from ML and BI analyses were congruent. All five samples of Primula sect. Reinii were added to the large group of subgenera Auganthus and Carolinella, which constituted a clade with strong support (ML bootstrap value, ML = 85 %; Bayesian posterior probability, PP = 1.00). Section Reinii was resolved as the sister group to sect. Cortusoides. The clade comprising sect. Reinii and two species of sect. Cortusoides (P. sieboldii and P. jesoana) was strongly supported in Bayesian analysis (PP = 1.00), while ML analysis weakly supported this arrangement (ML = 62 %). Other phylogenetic relationships within Primula revealed in this study were generally congruent with those previously inferred from plastid trees (Yan et al., 2010; Liu et al., 2015).

The phylogenetic relationships within the section were reconstructed from nuclear and chloroplast DNA sequences. A chloroplast tree based on all 38 samples of Primula sect. Reinii reconstructed from the second data set is shown in Fig. 3A. The tree topologies were congruent in both ML and BI analyses. Sect. Reinii was strongly supported as monophyletic (ML/PP = 96/1.00) and as a sister clade to P. jesoana (ML/PP = 90/1.00). Two geographically close species from Hokkaido, P. hidakana and P. takedana, formed a clade. The hidakana clade was resolved as monophyletic with strong support (ML/PP = 99/1.00), but the monophyly of each species was not supported, as P. takedana appears to be nested within P. hidakana. A robust sister relationship between P. reinii and P. tosaensis was identified in the chloroplast tree, which received strong support in each analysis (ML/PP = 90/0.99). Primula reinii and P. tosaensis were well supported as monophyletic (ML/PP = 78/1.00 and 99/1.00, respectively). With the exception of P. reinii var. reinii and P. tosaensis var. brachycarpa, four varieties were well supported as monophyletic, including the reinii and tosaensis clades. Overall, the relationships among and within species were well supported in the chloroplast tree.

Figure 3B shows the BI tree reconstructed from the third data set using ITS sequences (the ML tree is shown in Supplementary Data Fig. S2). In contrast to the chloroplast data set, the clade containing sect. Reinii was weakly supported as a monophyletic group in the nuclear tree (ML/PP = 76/0.83), while the sister relationship between sect. Reinii and P. jesoana was not supported (ML/PP = 27/0.45). The sister relationship between the reinii and hidakana clades was found in two different trees, but it was strongly supported only in the ML tree (ML/PP = 86/0.68). On the other hand, the clade comprising P. hidakana and P. takedana was strongly supported (ML/PP = 100/1), as it was in the chloroplast tree. The monophyly of the three species, P. takedana, P. tosaensis and P. reinii, was strongly supported in both ML and BI analyses, while that of P. hidakana was supported as monophyletic only in Bayesian analysis (PP = 1.00). Overall, the relationships within species remain unresolved because the topologies of the ML and BI trees were incongruent, and the nodes were not very well supported.

Divergence time estimation

Dating analyses indicated that all divergence events within Primula sect. Reinii occurred during the Quaternary (Fig. 4). Fossil-calibrated dating analyses indicated that Primula sect. Reinii diverged from P. jesoana about 1.82 Mya (95 % HPD: 0.65–3.46 Mya). The crown age of four species, P. hidakana, P. takedana, P. reinii and P. tosaensis, was 1.36 Mya (95 % HPD: 0.48–2.65 Mya). Separation between P. tosaensis and P. reinii was estimated to have occurred about 1.1 Mya (95 % HPD: 0.35–2.09 Mya). Although P. takedana was nested within P. hidakana, the first divergence point between the two species was estimated to be 0.36 Mya (95 % HPD: 0.10–0.69 Mya). Intraspecific diversification in P. reinii and P. tosaensis occurred during the past million years (Fig. 4).

Ancestral area reconstruction

The S-DIVA analysis suggested, with high confidence, that the origin of Primula sect. Reinii was in the Hokkaido region (A, 97 %; Fig. 4). The ancestral range of this section was estimated to include the geographically distant areas of regions A and D (89 %). The common ancestral area of two geographically close species, P. reinii and P. tosaensis, was inferred to be region D (92 %). These results suggested that: (1) the most recent common ancestor of the four species, P. hidakana, P. takedana, P. tosaensis and P. reinii, dispersed from region A to D; (2) two closely related species, P. tosaensis and P. reinii, colonized region C and/or E from D; and (3) four dispersal and five vicariance events were identified in this section.

Evolution of niche preferences

The 19 bioclimatic variables extracted from WorldClim (Supplementary Data Table S5) were summarized using PCA (Supplementary Data Table S6). The first two axes (PC1 and PC2) explained 52.4 and 26.9 % of the variation, respectively. PC1 correlated strongly and positively (r > 0.80, P < 0.001) with temperature seasonality (Bio4, bioclimatic variables of WorldClim) and had strong negative relationships with annual mean temperature, minimum temperature of the coldest month and mean temperature of the coldest quarter (Bio1, Bio6 and Bio11). PC2 had a strong positive correlation with mean diurnal range (Bio2) and a strong negative one with precipitation during the driest month, precipitation in the driest quarter and precipitation in the coldest quarter (Bio14, Bio17 and Bio19).

We fitted seven models of climatic niche evolution using the species’ climatic parameters obtained from PC1 and PC2 (Supplementary Data Table S7). The best model for PC1, based on AICw, was OUMV (AICw = 0.584). BMS was the second-best model for PC1 (AICw = 0.223). The OU1 model was supported only for PC2 (AICw = 0.551). Three parameters, niche optimum (θ), rate of niche evolution (σ²) and strength of selection (α), were estimated for each of the three groups (hidakana, reinii and tosaensis) from all supported models (Supplementary Data Table S8). For PC1, the model-averaged parameters were estimated from the OUMV and BMS models. The niche optima (θ) for the three groups, hidakana, reinii and tosaensis, were 3.10, −0.73 and −2.41, respectively (Fig. 5). The rate of evolution (σ²) was fastest in the tosaensis clade (σ² = 3101), which was four to eight times faster than those in the other two groups (hidakana: 379, reinii: 707; Fig. 5). The strength of selection was identical for all groups (α = 320). For PC2, the OU1 implying that niche evolution is not random was the model with the best fit. This model assumes a single optimum for all groups; thereby, the different groups shared a common parameter (see Supplementary Data Table S8).

Fig. 5.

Niche evolution parameters for three clades estimated by macro-evolutionary modelling. Niche optimum (θ) and niche rate (σ²) were estimated for PC1 based on two supported models (OUMV and BMS). The strength of selection (α) was identical for all groups (see Supplementary Data Table S2) and is not shown.

DISCUSSION

Phylogenetic position of sect. Reinii and its interspecific relationships

The phylogenetic position and monophyly of Primula sect. Reinii were unclear in previous studies (e.g. Mast et al., 2001, 2006; Yan et al., 2010; de Vos et al., 2014; Liu et al., 2015); however, our analyses recognized sect. Reinii as a robust monophyletic group and a sister taxon to sect. Cortusoides in subgen. Auganthus, which included almost all sections (Supplementary Data Fig. S1). The monophyly of sect. Reinii was also supported by both chloroplast and nuclear trees based on multiple individuals from each taxon (Fig. 3). Among the sister section Cortusoides, P. jesoana was nested in the root of sect. Reinii, as suggested by Kitamura et al. (1957) based on morphological characteristics.

This study recognized three robust clades: P. takedana and P. hidakana (clade hidakana), all varieties of P. reinii (clade reinii) and all varieties of P. tosaensis (clade tosaensis). However, the topology of the three clades was incongruent in the chloroplast and nuclear trees: clades reinii and tosaensis made a robust clade in the plastid tree, while the nuclear tree supported a clade of reinii and hidakana at a low Bayesian posterior probability (Fig. 3). The chloroplast phylogeny agreed well with the morphological characteristics of the rhizome: members of clade hidakana share a creeping rhizome, while a creeping rhizome does not develop in clades reinii and tosaensis (Richards, 2003). Incongruence between chloroplast and nuclear phylogenies could be attributed to hybridization, introgression and/or incomplete linage sorting, considering the most common cases in the phylogenetic studies of Primula (e.g. Liu et al., 2015; Ren et al., 2015). The specific cause of the incongruence could be identified by determining the correct genealogy using multiple nuclear markers or larger data sets [e.g. single nucleotide polymorphisms (SNPs) based on genomic data].

The samples of P. hidakana and P. takedana, species endemic to the Hokkaido region, clustered with strong support in both the chloroplast and nuclear trees (Fig. 3). Although each species was recognized as monophyletic in the nuclear tree, P. takedana was nested within P. hidakana in the chloroplast tree. Given their distribution range and habitat environments, incongruence between the chloroplast and nuclear phylogenies may be attributed to the past chloroplast capture: the current distribution range of the sister species of P. takedana and P. hidakana is restricted to northern and southern edges, respectively, across the serpentine-rich metamorphic belt in the Hokkaido region (see Fig. 1), and these two species might have been in contact, leading to hybridization. The north–south extending serpentine belt might have served as a corridor of migration for the edaphic endemics during the glacial period. This range shift scenario is supported by our date estimation that the two sister species diverged 0.36 Mya (Fig. 4). Further examination using ecological niche modelling will be needed to support this scenario.

The chloroplast phylogeny identified multiple samples of all three varieties in P. reinii var. kitadakensis, myogiensis and rhodotricha, as monophyletic, whereas var. reinii was paraphyletic (Fig. 3A). Distribution of these three varieties was confined to a narrow range with limestone, serpentine and granite geology, implying long-term isolation accompanied by differentiation of morphological factors, such as leaf shape (Fig. 2). On the other hand, var. reinii might have experienced allopatric differentiation across regions D and E.

Biogeographic history of Japanese endemic primroses

The present results provide a possible explanation for the biogeographic history of Japanese endemic primroses based on the dates of past geo-climatic events. Biogeographic reconstruction using S-DIVA and fossil-calibrated dating analyses indicated that Primula sect. Reinii originated in the Hokkaido region (A) during the early–middle Pleistocene (approx. 1.82 Mya) and then migrated southward through central Honshu (D) to Kanto (C) and/or the south-western (E) regions (Fig. 4). The common ancestral range of four species, P. takedana, P. hidakana, P. reinii and P. tosaensis, was inferred to be in the Hokkaido and central Honshu regions around 1.36 Mya, in the middle Pleistocene. This time frame is characterized by Quaternary climatic oscillations and the uplift of the high mountains in central Honshu (Kaizuka and Murata, 1969; Yonekura et al., 2001), suggesting that the diversification of sect. Reinii was interlinked with geo-climatic events in the Japanese Archipelago. Previous phylogeographic studies have suggested that the high mountains in central Honshu acted as refugia for arctic–alpine plants during interglacial periods and, thus, sculptured the north–south differentiation in the Japanese Archipelago via vicariance events (Fujii et al., 1997, 1999; Fujii and Senni, 2006; Ikeda and Setoguchi, 2007; Ohsawa and Ide, 2011). Primula hidakana, the most basal species in the phylogenetic tree of the sect. Reinii, occurs above the tree line (Richards, 2003), similar to those arctic–alpine plants on Hokkaido Island. Thereby, Reinii species would have experienced glacial range expansions and post-glacial retreats to refugia in high mountains and/or rocky habitats (limestone, serpentine and/or granite rocks) in central Honshu during climatic oscillations. Thus, our results suggest a migration scenario in which sect. Reinii originated in Hokkaido and migrated to southerly regions of the Japanese Archipelago during the mid–late Pleistocene.

Inter- and intraspecific differentiation would have been promoted by topographical changes accompanied by climatic oscillations and allopatric population isolations during the past million years. This suggestion is supported by S-DIVA analysis, which indicated that most biogeographic events affecting this section were concentrated in this period (Fig. 4). Reinii species prefer cool, moist and rocky habitats, and it is plausible that mountain uplift and climate events increased their distribution range and led to lineage divergence. A similar time window of divergence has also been reported in other plant groups in Japan, such as Ainsliaea (Mitsui et al., 2008), Asarum (Takahashi and Setoguchi, in press), Rhododendron (Watanabe et al., 2017) and Schizocodon (Higashi et al., 2015). Recent papers suggested that the plant diversity and endemism in the archipelago have expanded as a result of allopatric speciation processes, which led to the region-specific abiotic conditions, as characterized by insularity and topographic and climatic heterogeneities (Kubota et al., 2017; Watanabe et al., 2017). Our results agree with this suggestion and highlight the importance of topographic and climatic changes during the Quaternary in promoting the high level of plant diversity in the area.

Influence of niche evolution on diversification patterns

The results of macro-evolutionary modelling indicated a clear differentiation of climatic niches among the three phylogenetic clades of sect. Reinii. The hidakana, reinii and tosaensis clades exhibited heterogeneous rates of niche evolution (Fig. 5). Wiens and Donoghue (2004) proposed that the interplay between niche conservatism and niche evolution would prove to be a major theme in the biogeographic history of many clades. However, studies in insular habitats have often indicated that niche conservatism is one important factor in species diversification (e.g. Merckx et al., 2015). Our results differ from those of previous studies conducted in insular habitats and suggest that both the conservatism and the evolution of climatic niches may explain the biogeographic and evolutionary patterns of Japanese endemic primroses.

The modelling results based on PC1, which indicate a temperature gradient and are supported by the OUMV and BMS models (Fig. 5), suggest that each group has different niche optimal values (θ) and rates of niche evolution (σ²). The model-averaged niche optimal values of the three clades could be clearly distinguished (Fig. 5), corroborating that their habitat preferences are differentiated within the islands of Japan.

The hidakana clade, which includes two species occurring in the Hokkaido region, indicated the slowest rate of niche evolution and preferred the coolest climate among the three clades (Fig. 5). These results would be ascribed to their current distribution ranges, which were restricted to the northern end of the Japanese islands (Fig. 2). The evolutionary rate of the hidakana clade observed in the macro-evolutionary modelling was consistent with the biogeographic history reconstructed in this study. In particular, the lineage divergence between Hokkaido and Honshu Islands was sculptured by the uplifts of high mountains in central Honshu and climatic oscillation in the early–middle Pleistocene, and might have been reinforced by the slowest rate of climatic niche evolution. Thus, the phylogenetic niche conservatism in the hidakana clade may play a significant role in the initial north–south lineage divergence within the section via vicariance events.

Estimation of the divergence between P. tosaensis and P. reinii was dated to approx. 1.1 Mya, in the middle Pleistocene (Fig. 4), whereas niche evolution rates of the sister species were in obvious contrast, with faster niche evolution in P. tosaensis than in P. reinii (Fig. 5). Intraspecific niche evolution (expansion of niche breadth, without a shift) in P. tosaensis would thus be faster than that in P. reinii. The evolution of niche breadth is explained as intraspecific variation in niche occupancy (Colwell and Futuyma, 1971). Considering the current natural habitats of P. tosaensis, their fast niche evolution may be relevant to determining their elevation ranges. Indeed, in contrast to the habitats of P. reinii that are restricted to the sub-alpine zone, P. tosaensis inhabits a wide elevation range, from 500 to 2000 m. The fast niche evolution observed in P. tosaensis may be a consequence of the variable climatic niche occupancy associated with a wide elevation range. The mode and tempo of niche evolution play important roles in the adaptation of species to novel environments and affect diversification patterns during speciation and extinction (Kostikova et al., 2013). Thus, the fast niche evolution of P. tosaensis may have provided opportunities for this species to adapt to a wide range of habitats (e.g. warmer, lower elevation sites) during rapid habitat changes caused by climatic and topographic changes during the mid–late Pleistocene.

Conversely, P. reinii seems to be restricted by phylogenetic niche conservatism. Primula reinii displays much lower niche evolution rates than P. tosaensis (Fig. 5). Additionally, the results of PCA and climatic niche evolutionary modelling (Fig. 5) indicate that P. reinii prefers a cooler climate than P. tosaensis. These results suggest that migration to lower and warmer habitats may have been limited due to phylogenetic niche conservatism. On the other hand, such migration might be promoted by intraspecific diversification within P. reinii through vicariance, because P. reinii inhabits rocky and high-altitude mountain sites; thus, each population would have been isolated in ‘sky islands’. This hypothesis is supported by the plastid tree, which indicates relatively high sequence differentiation among local endemic taxa (Fig. 3A).

Thus, the sister taxa of P. reinii and P. tosaensis illustrate the contrasting range shift histories that can be explained by conservatism and climatic niche evolution on the islands of Japan. Niche evolution would have provided an opportunity for P. tosaensis to adapt to novel habitats, while phylogenetic niche conservatism reinforced intraspecific divergence within P. reinii, which comprised four varieties. Our results suggest that the spatiotemporal heterogeneity of the Japanese Islands might play a significant role in the diversification patterns of Japanese endemic primroses through climatic niche evolution. Generally, niche evolution from high- to low-elevation habitats is thought to be rare (Gamish et al., 2016), but it has been reported for other plant groups distributed in the QTP region (Myricaria in Liu et al., 2009; Primula in Ren et al., 2015; Dysosma in Wang et al., 2017). Even in Japan, Kubota et al. (2017) revealed that the strength of climatic niche conservatism differs among ferns, gymnosperms, angiosperm trees and angiosperm herbs, and indicated that the adaptation to harsh climates by angiosperm herbs has evolved independently in a variety of clades. Our study also indicates that climatic niche evolution can occur among closely related species in the islands during climatic and topographic perturbations, reinforcing the idea that ecologically divergent sister species are more likely to have originated in the Quaternary (Hua and Wiens, 2013).

In conclusion, climatic changes and historical orogenies during the Quaternary might have played a significant role in determining the speciation and diversification patterns of primroses in Japan. Our study highlights the importance of climatic niche evolution for biogeographic patterns of insular endemic plant species in the archipelago. While we focused on abiotic factors, such as temperature and precipitation, to shed light on the diversification patterns of Japanese endemic primroses, biotic factors, such as life history, effective population size, the presence of a seed bank, and the type and number of individuals, may also be involved in niche variation (Aizen and Patterson, 1990; Edwards and Westoby, 1996; Kassen, 2002; Kostikova et al., 2013). Thus, future investigations based on population genomic approaches using a highly polymorphic DNA marker, large-scale genomic data sets and ecological studies including both biotic and abiotic factors are necessary to elucidate the mechanism of fast climatic niche evolution in P. tosaensis.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Table S1: the samples used in this study. Table S2: GenBank database accession numbers for the DNA sequences of the taxa used in the present study. Table S3: GenBank accession numbers for the DNA sequences of the 15 taxa used to determine the root age of section Reinii using the fissile calibration method. Table S4: summary of information on gene regions in the three data sets used in this study. Table S5: the 19 bioclimatic variables for each sample in the three clades (hidakana, reinii and tosaensis) used for macro-evolutionary modelling regarding climatic niches. Table S6: summary of the species’ climatic parameters for each sample in the three clades (hidakana, reinii and tosaensis) used for macro-evolutionary modelling. Table S7: summary of alternative models of climatic niche evolution for section Reinii. Table S8: estimated parameters of the best-fit models for each PC axis. Figure S1: maximum clade credibility tree for subgenus Augunthus derived from Bayesian inference analyses of three chloroplast genes, matK, rps16 and trnL–F.