Recent colonization by a coastal plant of inland habitats at an ancient freshwater lake, Lake Biwa: multilocus sequencing and a demographic history of Lathyrus japonicus (Fabaceae)



Tatsuo Ohtsuki, Graduate School of Human and Environmental Studies, Kyoto University, Yoshida-nihonmatsu-cho, Sakyo-ku, Kyoto 606-8501, Japan. Tel: +81-75-753-6858; Fax: +81-75-753-6894 ; E-mail:


Ancient lakes have been recognized as “long-term isolated islands” in terrestrial ecosystems. Lake Biwa, one of the few ancient lakes that formed around 4 million years ago, harbors many coastal species that commonly inhabit seashores. The beach pea, Lathyrus japonicus, is a typical coastal species of this freshwater lake, where morphological, physiological, and genetic differentiations have been reported between Biwa and coastal populations. Whether Biwa populations were isolated for long periods throughout Pleistocene climatic oscillations and subsequent range shifts is unclear. We assessed population genetic structure and demography of beach pea in this ancient freshwater lake using the sequences of eight nuclear loci. The results of STRUCTURE analyses showed evidence of admixture between Biwa and coastal populations, reflecting recent gene flow. The estimated demographic parameters implemented by the isolation with migration model (IM model) revealed a recent divergence (postglacial period) of Biwa populations, with some gene flow from Biwa to coastal populations. In addition, Biwa populations were significantly smaller in size than the ancestral or coastal populations. Our study suggests that a Holocene thermal maximum, when transgression could allow seeds from coastal plants to access Lake Biwa, was involved in the origin of the Biwa populations and their genetic divergence. Thus, coastal populations might have migrated to Lake Biwa relatively recently. Our study concluded that ancestral migrants in Lake Biwa were derived from small founding populations and accelerated genetic isolation of Biwa populations during short-term isolation.


Ancient lakes, which are defined as those in continuous existence for millions of years, have been recognized as “long-term isolated islands” in terrestrial ecosystems, and exhibit high degrees of biodiversity and endemism (Martens 1997). Numerous evolutionary studies on island systems have been performed due to the abundant intra- and/or interspecific diversity in these small land masses and have provided novel examples of evolutionary patterns (Losos and Ricklefs 2009). As these long-lived lakes are home to landlocked species, the origins and phylogenetic relationships of these endemic species in ancient lakes constitute a biogeographical enigma (e.g., Baikal and Caspian seals, McLaren 1960). On close scrutiny, however, most ancient lakes reveal both steady limnological aging processes and complex histories owing to drastic geological and climatic changes. During the isolation of the predecessors of the present-day lakes, environmental fluctuations (including abiotic and biotic factors) significantly influenced the differentiation of biota between the inland migrants and the original populations (Stager and Johnson 2008). Consequently, fine-scale phylogeographical studies are required to examine patterns of historical geographical fragmentation or gene flow restriction. In particular, application of demographic models of intra- and interspecific diversification to ancient lakes can be used to evaluate evolutionary and/or geological histories of ancient migrants in the basins.

Lake Biwa, a Japanese freshwater lake, harbors many coastal plants that normally inhabit seashores (e.g., Calystegia soldanella [Convolvulaceae], Vitex rotundifolia [Verbenaceae], Lathyrus japonicus [Fabaceae], Arabidopsis kamchatica subsp. kawasakiana [Cruciferae], Raphanus sativus var. raphanistroides [Cruciferae], and Pinus thunbergii [Pinaceae]; Kitamura 1968). This lake, forming approximately 4 million years ago (MYA), is widely recognized as one of the world's few ancient lakes because it is thought to have persisted as a result of the subsidence of a fault (Takaya 1963; Meyers et al. 1993; Kawabe 1994). The coastal plants were postulated to have migrated to the inland lake from coastal populations (Takaya 1963; Kitamura 1968). Some paleogeographic information indicates that Lake Biwa was connected to the Seto Inland Sea at Osaka Bay via a short canal mostly filled with seawater (Seto Inland Sea Basin; Fig. 1; Takaya 1963). Takaya (1963) inferred that populations of the coastal plant Pinus thunbergii (black pine) in Lake Biwa migrated during the middle of the Pleistocene as few paleogeographic fossil records exist (e.g., 1.1–0.4 MYA; Kawabe 1994; Meyers et al. 1993). However, this ancient lake harbors a large number of species (595 animal and 491 plant species; Mori and Miura 1990) with a comparatively low proportion (6%) of endemics (vs. ca. 54% endemics in Lake Baikal and ca. 56% endemics in Lake Tanganyika; Martens 1997). Possibly, this lake has not remained constant over geological time and has repeatedly reconnected with and become isolated from the Seto Inland Sea at Osaka Bay. Lake Biwa has connected with Osaka Bay only via the Seta-Yodo River (Fig. 1); 75 km separates Lake Biwa from Osaka Bay, and the lake is only 85 m above sea level (Ministry of Land, Infrastructure, Transport and Tourism, 2007). Lake Biwa lies in close geographic proximity and shares a related origin and past hydrological connections with the bay, which might result in floral sharing via this river. In addition, sea level fluctuations during the Last Glacial Maximum (LGM; 29.000–19.000 yr bp) and the Holocene are thought to have greatly affected the present distribution of coastal species (Taberlet et al. 1998; Kadereit et al. 2005) (e.g., during the LGM, the eustatic sea level was 125 ± 5 m lower than at present, while during the Holocene [between ca. 7 and ca. 2–1 kyr bp], it was 3–5 m higher than at present [Kevin et al. 1998]). Consequently, coastal plants might have invaded Lake Biwa recently when the sea level rose.

Figure 1.

Locations of populations of Lathyrus japonicus in this study. (A) Locations in the Japanese Archipelago and sampling localities in Russia (15) and South Korea (14). (B) Map of Lake Biwa with five extant populations (1–5) with the blue color, Seta-Yodo River, Osaka Plain, and Osaka Bay. Detailed information on these populations is provided in Table S1. The positions where Lake Kawachi and the Ogura-ike Marsh were once located are marked with a broken line with the gray color. The Osaka Plain is represented with the soil color.

Previous phylogeographic studies of L. japonicus (Ohtsuki et al. 2011) and C. soldanella (Noda et al. 2011), found in both Lake Biwa and coastal site, detected unambiguous genetic differentiation between the Lake Biwa's and coastal populations using chloroplast DNA (cpDNA) and nuclear DNA microsatellite (nDNA SSR) loci. Those also detected significantly lower genetic diversity in Lake Biwa's populations than in coastal populations. These results were likely to have been caused by founder events or genetic drift in the inland populations during the long-term isolation of Lake Biwa (Noda et al. 2011; Ohtsuki et al. 2011).

However, given that the Lake Biwa populations harbored a single haplotype of cpDNA sharing with the coastal populations in L. japonicus (Ohtsuki et al. 2011), such a shared haplotype might be interpreted as recent divergence between inland and coastal populations. Because if inland populations have been isolated from coastal populations for long time, they were expected to have unique haplotypes. In addition, the evidence that there are no endemic plant species in of Lake Biwa's lakeshore seems to suggest that the lakeside populations have immigrated from coastal area in recent time and gene flow occurred between inland and seashore populations until recently. Therefore, these previous studies could lead to an incorrect interpretation of the isolation history of Lake Biwa, as few paleogeographic data exist (Takaya 1963). To understand the biogeographic history between coastal and inland lake plants, further studies using multilocus sequencing to estimate the divergence time and population demographic history are needed. Recent advances in population genetics have allowed the estimation of demographic parameters throughout the process of population divergence (e.g., “isolation with migration” model [IM model]: Nielsen and Wakeley 2001; Hey and Nielsen 2004; MiMAR:Becquet and Przeworski 2007).

Lathyrus japonicus (beach pea) is a diploid perennial herb (Fig. 2; Brightmore and White 1963) that commonly occurs in temperate coastal areas of Asia, Europe, and North and South America. Its extensive range is explained by seed dispersal by currents and the ability of the seed to remain viable while floating in seawater for up to 5 years (Brightmore and White 1963). This species is easy to design polymerase chain reaction (PCR) primers and obtain nuclear gene sequences because we can access accumulated expressed sequence tag (EST) information from the Fabaceae. Consequently, beach pea is suitable for elucidating the demographic history of coastal plants that are isolated in Lake Biwa, using multilocus sequencing analyses.

Figure 2.

Picture of beach pea at coast site (by Tatsuo Ohtsuki).

In this study, we applied the “isolation with migration” model (IM model), based on multilocus sequences for the inland and coastal populations of L. japonicus, and: (1) the divergence time of populations of beach pea in Lake Biwa, (2) the historical gene flow between the inland and coastal regions, and (3) the population sizes of the inland and coastal populations. We assessed whether coastal plants colonized inland habitats during the long period of isolation, from 1.1 to 0.4 MYA (Kawabe 1989, 1994; Meyers et al. 1993) or recently, during the LGM. We also examined the level of historical gene flow through the Seta-Yodo River and a founder event with ancestral migrants in Lake Biwa at the time of divergence.

Material and Methods

Lathyrus japonicus leaf materials were sampled throughout most of its distribution range in the Japanese Archipelago, as well as from populations in Eastern Russia and South Korea. Details of the locations of the populations and the number of individuals used for the analysis are shown in Figure 1 and Table S1. In total, 36 individuals from 15 populations were used for sequencing of multilocus genes, covering all populations of Lake Biwa and a wide range of coastal populations. Total DNA was extracted from silica gel-dried leaves using the cetyl trimethyl ammonium bromide method based on Doyle and Doyle (1990).

PCR amplification and sequencing of nuclear loci

In the preliminary analysis, 61 nuclear loci were sequenced using four to eight samples of the inland and coastal populations to detect nuclear DNA polymorphisms. Forty-four regions were amplified using primers developed by Choi et al. (2006), which are universally applicable for the Fabaceae. In addition, 17 regions were amplified using primers developed based on the sequences of genes and/or ESTs from Fabaceae obtained from GenBank. Details of the primers used for the preliminary analysis of nDNA are listed in Table S2.

The PCR was performed with 40 cycles of 30 sec at 94°C, 30 sec at 54°C, and 1 min at 72°C in a total reaction volume of 10 μL containing 6.75 μL sterilized water, 0.08 mmol/L dNTP mixture, 0.25 U TaKaRa Ex Taq (TaKaRa Biotechnology Co., Otsu, Japan), 1.0 mmol/L reaction buffer (TaKaRa Ex Taq), 0.2 μmol/L of each primer, and 1.0 μL of template DNA (10 ng/μL). The PCR products were visualized on 1.0% tris acetate EDTA-agarose gels stained with ethidium bromide, and purified with glass powder using the GeneClean II Kit (Qbiogene, Solon, OH). The products were sequenced directly in both directions using the standard methods of the BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Foster City, CA). The sequences were aligned using AutoAssembler software (Applied Biosystems).

We sequenced 50 (ca. 27,000 bp) of 61 loci and found no polymorphisms in 26 loci and failed to obtain clear electropherogram for 16 loci. Ultimately, we successfully sequenced and obtained polymorphisms for eight loci (Table 1): ATCP, GDCP, SUSY, CRY2B, Locus 29, Locus 31, PAT1, and Locus 61. Locus information and a list of the primers are shown in Table S3. According to a recent simulation study (Knowles and Carstens 2007; Strasburg and Rieseberg 2010), eight loci are appropriate for estimating demographic parameters under the IM model. The sequences of five loci (ATCP, GDCP, SUSY, CRY2B, and PAT1) were determined for all 36 samples from 15 populations, and those of Locus 29, 31, and 61 were determined for 31, 33, and 35 individuals, respectively, due to unclear electropherograms.

Table 1. Summary of nucleotide polymorphisms and genetic divergence of eight loci between inland and coastal populations of Lathyrus japonicus
LocusAligned size (bp)Largest nonrecombining blockLake Biwa populationsCoastal populationsDivergence
No. of seqs. S π Rm No. of seqs. S π Rm K
  1. S, number of segregating (polymorphic) sites; π, average number of pairwise nucleotide differences per site; Rm, estimate of the minimum number of recombination events; K, the genetic divergence between inland and coastal populations.

ATCP 4173233040.0037614250.0051520.00645
GDCP 428428300004210.0008700.00056
SUSY 246246300004220.0015700.00283
CRY2B 415415300004220.0015600.00132
Locus 29 2911083090.0118623290.0134930.01480
Locus 31 7282823040.00099036150.0048720.00382
PAT1 3591593040.0031814260.0053410.00503
Locus 61 3533533010.0006804240.0046500.00399
Mean 404.6289.3302.750.002560.5405.50.0046910.00343

The sequences of alleles were determined and polymorphisms were determined probabilistically using the PHASE software (Stephens et al. 2001) in DnaSP ver. 5.0 (Librado and Rozas 2009) for sequences with two or more heterozygous sites. In subsequent analyses, only those haplotypes with a probability >0.60 were used (Harrigan et al. 2008).

Population structure analyses

To compare the population structure between the Lake Biwa and coastal populations, Bayesian clustering was conducted with STRUCTURE ver. 3.1 (Pritchard et al. 2000) using multilocus allele combinations. An admixture model was used, and a correlate of allele frequencies among populations was assumed. Typically a burn-in of 10,000–100,000 is more than adequate (Pritchard et al. 2000). Then, the probability of assigning the haplotypes of individuals into clusters was estimated using 3.0 × 105 iterations, following a burn-in period of 1.0 × 105 iterations. The number of clusters (K) was set from 1 to 15, and all runs were replicated 20 times to test the stability of the results. Additionally, we selected coastal populations (populations 6–15) and conducted a similar analysis to look for distinct clusters within the coastal populations. The number of clusters (K) was set from 1 to 15. The symmetric similarity coefficient (SCC: H’) was calculated between all pairs of runs for the same K using CLUMPP to determine the appropriate number of clusters (K) (Jakobsson and Rosenberg 2007). In addition, the most likely number of clusters was estimated according to the model value (ΔK) based on the second-order rate of change with respect to the K of the likelihood function following Evanno et al. (2005) to determine the best fitting value of K.

Diversity and neutrality analyses

For coastal versus inland populations, we calculated the number of segregating sites (S), the average number of pairwise nucleotide differences per site (π; Nei 1987), and genetic divergence between inland and coastal populations (K) for each locus. We estimated the minimum number of recombination events (Rm) within the eight loci using the four-gamete test (Hudson and Kaplan 1985). To evaluate whether each locus followed neutral equilibrium, Tajima's D (Tajima 1989), D*, and F* of Fu and Li (1993) were estimated and their deviations from neutrality were evaluated by 10,000 coalescent simulations. These summary statistics were calculated using DnaSP ver. 5.0 (Librado and Rozas 2009).

Estimation of population demographic parameters and testing of population divergence models

To determine the demographic history that followed the divergence of the inland and coastal populations, we used the IM model (Nielsen and Wakeley 2001; Hey and Nielsen 2004) and estimated demographic parameters using the program IMa (Hey and Nielsen 2007). IMa uses a Markov chain Monte Carlo (MCMC) method to estimate the posterior probability densities of six demographic parameters, including the divergence time (t), bidirectional migration rates (m1 and m2), and the effective population sizes of the ancestral (θA) and descendent populations (θ1 and θ2) (Fig. 3). The IM model is based on several assumptions, including the neutrality of the loci, no recombination within the loci, free recombination across the loci, and no migrations from an unsampled third species to the focal species (Nielsen and Wakeley 2001; Hey and Nielsen 2004). Then we applied the largest nonrecombining block of eight loci to remove recombinations within the loci (Table 1). The running of IMa involved two steps, the M mode and L mode. First, functions of the model parameters were estimated in M mode. The run was performed using 10 independent chains under Metropolis coupled with a geometric heating scheme with the parameters g1 = 0.997 and g2 = 0.997 and with a prior probability density (m1 = 55, m2 = 500, q1 = 0.5, t = 0.2). The M-mode run was performed twice with different random seeds to check for convergence and the marginal posterior distribution and the maximum likelihood estimates (MLEs) of demographic parameters were predicted by running IMa in L mode. After all runs converged under the full model that was maximized over all parameters to generate an estimate of the posterior density and the six demographic parameters were estimated, the fit of the data was tested against simpler models using the nested model approach in the L mode of IMa. Based on the likelihood of nested models, a likelihood ratio test was conducted to determine whether the data fit a simpler model or the full model (Hey and Nielsen 2007). To estimate the levels of gene flow after the divergence of the inland and coastal populations, we tested the full model against models in which one or both gene flow parameter(s) (m1, m2) were set to zero. In addition, models with equal effective population sizes among species and ancestral populations were tested. An infinite site model (IS, each mutation occurs at a unique site; Kimura, 1969) of mutation was applied to seven loci (ATCP, GDCP, SUSY, CRY2B, Locus 31, PAT1, and Locus 61) and the Hasegawa–Kishino–Yano model (HKY, transition vs. transversion rates and base frequencies can be unequal; Hasegawa et al. 1985) was applied to Locus 29 because Locus 29 was not used in the IS model.

Figure 3.

Six demographic parameters in the full “Two-population Isolation with Migration” model, applied to the Lake Biwa and coastal populations.

Demographic parameters were scaled using the neutral mutation rate (μ). The mutation rates of nuclear genes in L. japonicus are unknown. Thus, we used the synonymous substitutions of the whole genome of Glycine max (Fabaceae), which belongs to the same family as L. japonicus (6.1 × 10−9 substitutions synonymous site−1 year−1; Schmutz et al. 2010) because substitution rates estimated from whole-genome sequences are generally as reliable as fossil data (e.g., Chen et al. 2010; Li et al. 2010). Then we used the geometric mean (1.588 × 10−6 substitutions/locus/year) based on the above rate for rescaling parameters of the beach pea. In addition, the beach pea is a perennial and can be fertile within 4 years after germination (Brightmore and White 1963), a generation time of 4 years was applied to scale population sizes.


Sequencing analysis

We sequenced eight nuclear loci with a total length of 3237 bp. The length of the aligned sequences for each locus ranged from 246 to 728 bp (Table 1). All sequences with different alleles were deposited in the DNA Data Bank of Japan (DDBJ; AB675956–AB676029). We also determined the largest nonrecombining block of eight loci with a total length of 2314 bp, because the IM implementation allowed no recombination within the loci (Table 1).

Comparisons of sequence polymorphisms and divergences between the inland and coastal populations are shown in Table 1. Nucleotide diversity was variable across the loci (inland populations, π = 0.0000–0.0119; coastal populations, π = 0.0009–0.0135). The inland populations tended to have smaller numbers of polymorphic sites and recombination events, and lower nucleotide diversities, as compared to those of the coastal populations. All of the polymorphisms of the inland populations were shared among the coastal populations. No locus showed a significant deviation from neutral expectations in Tajima's D or Fu & Li's D* and F* statistics (Table 2).

Table 2. Results of neutrality tests for eight loci between inland and coastal populations of Lathyrus japonicus
LocusLake Biwa populationsCoastal populations
Tajima's DFu & Li's D*Fu & Li's F*Tajima's DFu & Li's D*Fu & Li's F*
ATCP 1.477430.806151.149811.548800.910301.27881
GDCP  –0.843840.558970.73887
SUSY  –−0.144970.921510.70549
CRY2B  –0.730410.765790.87562
Locus 29 1.951321.058021.530501.387561.050801.39782
Locus 31 −1.003690.806150.340690.662241.038011.07751
PAT1 −0.302970.949810.682721.396551.021911.32296
Locus 61 −0.082370.594480.469011.806421.021911.46730

Population genetic structure

In the Bayesian clustering, highly consistent configurations of individual assignments were detected at = 2 (H’ = 0.99; Fig. 4) and the ΔK was the highest at = 2 (Fig. 4). From STRUCTURE, the clustering pattern was highly consistent across 20 independent runs, suggesting that introgressions might have occurred from coastal populations to inland ones when = 2. This seems more reasonable if we consider that the inland populations are derived from the coastal populations (Fig. 4). As the coastal populations distributed widely from Russia to Japan showed the same patterns of assignment proportion, regardless of locality (= 2–4; Fig. 4), we considered the coastal populations as one group.

Figure 4.

(A) Assignment of haplotypes of individuals for two clusters (K = 2) and two to four clusters (K = 2–4) in coastal populations implemented by STRUCTURE. Each vertical bar represents an individual and its assignment proportion. Regions and population numbers are shown under the bars. (B) Similarity coefficient (SCC: H’), and ∆K from the STRUCTURE analysis.

Estimation of demographic history by coalescent-based analyses

The MLEs and the 90% highest posterior density (90% HPD) of the population genetic parameters estimated by IMa are shown in Table 3. The time of the split between inland and coastal populations was estimated as = 0.0065 (90% HPD = 0.001–0.052; Table 3). After calibration with the mutation rate (6.1 × 10−9 substitutions site−1 year−1), the estimated divergence time between the inland and coastal populations was 4025 (90% HPD = 681–31,890) years bp (Table 3), coinciding with the end of the Holocene thermal maximum (HTM). The marginal distribution of divergence time, T, harbored a sharp peak, but did not drop to zero in sufficiently high values (Fig. 5).

Table 3. Maximum-likelihood estimates (MLEs) of demographic parameters of IM model between inland and coastal populations
  θ 1 θ 2 θ A m 1 m 2 t 2Nem 1 2Nem 2 N 1 N 2 N A T (year)
  1. The 90% highest posterior density (HPD) intervals of each parameters are shown in parentheses. Demographic quantities, N1, N2, NA, and T, were converted based on a mutation rate estimate of 6.1 × 10−9 substitutions site−1 year−1. θ1, θ2, θA: effective population size of Lake Biwa populations, Coastal populations and Ancestral population, respectively; m1, m2: population migration rate from Coastal populations to Lake Biwa populations forward in time and vice versa; t: time since population divergence; 2Nem1, 2Nem2: the effective rate at which genes come into a population per generation from Coastal populations to Lake Biwa populations and vice versa; N1, N2, NA: the estimate effective population size of Lake Biwa population, Coastal population, Ancestral populaation, respectively.



(90% HPD)



























(90% HPD)



























(90% HPD)

























Figure 5.

(A) Marginal distribution of the posterior probability of divergence time (× years), (B) migration rates, and (C) effective population sizes (× 1000) estimated by IMa analyses conducted between inland and coastal populations of Lathyrus japonicus. All parameters were scaled by a mutation rate of 6.1 × 10−9 substitutions synonymous site−1 year−1.

The model without gene flow after divergence of the inland and coastal populations was rejected (= 0.071; Table 4), suggesting introgression and/or recent gene flow (Table 3). The estimated 2Nem (the effective rate at which genes come into a population per generation; Wilkinson-Herbots, 1998) was nonzero, while the probability distributions did not show sharp peaks (Fig. 5). The amount of gene flow was asymmetric; gene flow from inland populations to coastal populations was greater (m2 = 36.25, 90% HPD = 2.75–217.8, 2Nem2 = 1.0927) than that in the opposite direction (m1 = 0.0275, 90% HPD = 0.028–43.86, 2Nem1 = 0.0003; Tables 3 and 4).

Table 4. Tests of nested models for no migration (a) and different effective population sizes (b)
 Lake Biwa populations – Coastal populations
Modellog(P)df2LLR P
  1. Log(P) is the posterior probability of the model given data, 2LLR = 2 × (log(P)full model−log(P)nested model), df is the difference in the number of parameters between the nested and full models, and the P value is the probability of achieving the test statistic (2LLR) by chance under the null model. The models with > 0.05 are shown in boldface.

  θ1, θ2, θA, m1, m2 = 0−3.237312.1393 0.071
θ1, θ2, θA, m1 = 0, m2−3.087011.8386 0.086
θ1, θ2, θA, m1 = 0, m2 = 0−10.9793217.6232<0.05
θ1=θ2 θA m1 m2−5.540516.7456<0.05
θ1=θA θ2 m1 m2−4.243514.1516<0.05
θ1 θ2=θA m1 m2−4.553114.7707<0.05

We also found that the effective population sizes of the inland populations were significantly smaller than those of the coastal populations (Tables 3 and 4; Fig. 5). The estimated effective population size of the inland populations was N1 = 719 (90% HPD = 174–3154), nearly one-third that of the coastal populations was N2 = 2333 (90% HPD = 449–11,286). The effective sizes of the ancestral population tended to be larger than those of the descendent populations, although the posterior probabilities of NA = 33,570 showed a wide, flat distribution (90% HPD = 17,033–64,807).


Recent origin of inland populations and biogeographic history

The present multilocus analysis could reconcile the difficulties in explaining the demographic history of inland populations of Lathyrus japonicus. The estimated divergence time (ca. 4000 yr bp, 90% HPD: ca. 700–32,000; Table 3) corresponds to the postglacial period, contradicting the previous inference of long-term isolation of inland (Lake Biwa) and coastal populations (Noda et al. 2011; Ohtsuki et al. 2011) of L. japonicus and Calystegia soldanella. The sea level around the Japanese Archipelago was ca. 130 m higher at the maximum transgression, around 6000 yr bp, than at present (Izeki 1976; Kimura 1996). In particular, the Osaka Plain (Fig. 1) had been largely submerged in the Seto Inland Sea (ca. 3500 yr bp). Following gradual formation of the alluvial plain of the Yodo River system (Yasuhara et al. 2002), wetland environments commonly covered vast areas of the Osaka Plain. Scattered lakes and marshes (e.g., Lake Kawachi and the Ogura-ike Marsh; Fig. 1) formed a hydrological connection between Lake Biwa and Osaka Bay via the Seta-Yodo River. Additionally, this river was only 85 m above sea level (Yasuhara et al. 2002; Ministry of Land, Infrastructure, Transport and Tourism, 2007), allowing the colonization of coastal species into the Biwa populations.

This biogeographic history is dependent on the divergence time, based on the assumption of substitution rate for the scaling. However, even a slower mutation rate (e.g., 1.0 × 10−9 substitutions synonymous site−1 year−1) resulted in a recent divergence (ca. 24.000 yr bp). These results robustly discard the hypothesis that inland populations of coastal plants in Lake Biwa were isolated for long periods throughout Pleistocene climatic oscillations. However, due to the very widely credibility intervals (HPD: ca. 700–32,000; Table 3), we cannot statistically exclude the possibility that the genetic divergence occurred during last glacial period. Nonetheless, during the last glacial period, global sea level was 60–90 m below the present (Yokoyama et al. 2001; Siddall et al. 2003), which caused a distance between Lake Biwa and Set Inland Sea. Therefore, the beach pea was thought to hardly have a chance to invade to Lake Biwa. Given the palaeogeographical findings, our conclusion is that the colonization history of Lake Biwa occurred recently, likely involving the change in sea level following postglacial period, especially HTM (11,000–5000 yr bp; Wanner et al. 2008).

A previous study found a shared cpDNA haplotype between inland populations and coastal populations, which was inferred as an ancestral polymorphism (Ohtsuki et al. 2011). However, the recent origin estimated by the present multilocus study indicates that the shared haplotype is likely due to seed dispersal. Given that plains around lakes are frequently flooded, seeds would be prone to being washed from Lake Biwa via the Seta-Yodo River. This habitat characteristic could allow gene flow between inland and seashore populations, especially seed dispersal from inland to seashore populations. This is consistent with the present genetic admixture (Fig. 4). Furthermore, statistical tests have provided evidence of significant gene flow between the inland and coastal populations after differentiation (Table 4), supporting Biwa populations as the source. However, the IM model cannot determine whether the historical gene flow occurred during the early or late stage after population differentiation (Hey and Nielsen 2004). The current habitats of the Lake Biwa's beach pea are completely isolated inland from the lakeshore, with no possibility of seed dispersal by water. Therefore, our finding of gene flow may represent historic gene flow. Consequently, fluctuations in sea levels could allow the seeds of coastal plants to access Lake Biwa, and the unidirectional gene flow from inland to coastal populations is thought to be primarily attributable to the difference in elevation between the inland area and the sea.

The rapid fixation of polymorphisms was influenced by a founder event

Our estimated demographic parameters showed that the populations in Lake Biwa were significantly smaller than the coastal populations (Tables 3 and 4). These results suggest a founder event of the beach pea during the initial colonization. The nucleotide diversity (π) of the inland populations was also found to be smaller than that of the coastal ones, and all inland populations show fixed alleles (GDCP, SUSY, and CRY2B; Table 1) supporting this postulate. In addition, the lakeshore habitats of the beach pea in Lake Biwa is quite limited (five extant populations are remained), which can be attributed to the small population size of Lake Biwa. This history is consistent with previous phylogeographic studies indicating that the Lake Biwa populations experienced a loss of genetic diversity (e.g., the fixation of a single cpDNA haplotype and smaller allelic richness of nSSR in the Lake Biwa populations; Ohtsuki et al. 2011). Therefore, the smaller inland populations would be the result of postfounder event.

The coastal population was significantly smaller (N2 = 2333 [90% HPD = 449–11,286]) than the ancestral population (NA = 33,570 [90% HPD = 17,033–64,807]). While coastal populations potentially have an extensive range in the Japanese archipelago, the estimated parameters showed that its population is smaller than that of the ancestral one. This could be one explanation for the fact majority of loci have positive Tajima's D values, caused by structure in population size. (Table 2). The coastline of the Japanese archipelago is frequently damaged by typhoons in the summer, which often seriously affect coastal populations of L. japonicus. Thus, the present coastal populations in the Japanese Archipelago may be reduced by occasional damage by typhoons.

Evolutionary history of inland populations at Lake Biwa

Inland and coast habitats have distinct environments. Indeed, intra- and interspecific adaptive divergence has been reported inland in the yellow monkey flower Mimulus guttatus (Lowry et al. 2008). In our study system, the inland and coastal individuals of L. japonicus differ in morphology (leaf blade thickness; Ohtsuki and Setoguchi 2011) and physiology (the leaf flavonoid content; T. Ohtsuki, Y. Murai, T. Iwashina and H. Setoguchi,, unpubl. data), perhaps due to a lack of salinity stress in the Lake Biwa populations. These intraspecific differences indicate a history of isolation of the inland populations from the coastal populations. The present demographic history suggested the recent colonization involving a recent (around the HTM) founder event involving the beach pea in an inland freshwater lake. In addition, the inland populations have been influenced little by additional gene flow from coastal populations.

Further studies, including comparative genetics (such as candidate genes for salinity stress) and/or ecology (such as transplant experiments and examination of photosynthetic responses to salinity stress), will be required to unravel the adaptive divergence between inland and coastal populations.


We thank H. Higashi and H. Oh for providing the leaf materials of Lathyrus japonicus and Y. Mitsui for advice regarding the IM analysis. This work was supported by a Grant-in-Aid for Science Research (#24247013) and the Foundation of River & Watershed Environment Management to H. S. and (#24•888) to T. O. from the Japan Society for the Promotion of Science. We also thank the anonymous reviewers for helpful comments regarding the manuscript.

Conflict of interest

The authors declare no conflict of interest.

Author Contributions

All authors conceived and designed the experiments. T. O. and H. S. collected samples, T. O. performed the experiments. T. O. and H. I. analyzed the data. T. O and H. S. prepared the manuscript.