The presence of a cryptic barrier in the West Pacific Ocean suggests the effect of glacial climate changes on a widespread sea‐dispersed plant, Vigna marina (Fabaceae)

Abstract Ocean currents are an important driver of evolution for sea‐dispersed plants, enabling them to maintain reciprocal gene flow via sea‐dispersed diaspores and obtain wide distribution ranges. Although geographic barriers are known to be the primary factors shaping present genetic structure of sea‐dispersed plants, cryptic barriers which form clear genetic structure within oceanic regions are poorly understood. To test the presence of a cryptic barrier, we conducted a phylogeographic study together with past demographic inference for a widespread sea‐dispersed plant, Vigna marina, using 308 individuals collected from the entire Indo‐West Pacific (IWP) region. Chloroplast DNA variation showed strong genetic structure that separated populations into three groups: North Pacific (NP), South Pacific (SP) and Indian Ocean (IN) (F′CT among groups = 0.954–1.000). According to the Approximate Bayesian computation inference, splitting time between NP and SP was approximately 20,200 years (95%HPD, 4,530–95,400) before present. Moreover, a signal of recent population expansion was detected in the NP group. This study clearly showed the presence of a cryptic barrier in the West Pacific region of the distributional range of V. marina. The locations of the cryptic barrier observed in V. marina corresponded to the genetic breaks found in other plants, suggesting the presence of a common cryptic barrier for sea‐dispersed plants. Demographic inference suggested that genetic structure related to this cryptic barrier has been present since the last glacial maximum and may reflect patterns of past population expansion from refugia.


| INTRODUC TI ON
Oceans are important corridors for estuarine and coastal plant species as they enable the exchange of individuals and help to maintain genetic uniformity with remote populations across distributional ranges.
Many plant species growing in coastal environments produce diaspores (mainly seeds and fruits) specifically adapted to long distance dispersal (LDD, Cain, Milligan, & Strand, 2000) by ocean currents (Ridley, 1930). As oceanic dispersal (sea-dispersal) is often the most effective mode of long distance seed dispersal (Harwell & Orth, 2002), many sea-dispersed plants have extremely wide distributional ranges in the tropics and subtropics worldwide (Tomlinson, 1986). Thus, low genetic differentiation among populations could be expected due to their extended gene flow by LDD of diaspores across the distributional range (Kudoh & Whigham, 1997;Nilsson, Brown, Jansson, & Merritt, 2010). For example, genetic variation in not only maternally inherited chloroplast DNA (Takayama, Kajita, Murata, & Tateishi, 2006) but also bi-parentally inherited nuclear DNA (Takayama, Tateishi, Murata, & Kajita, 2008) showed an absence of population differentiation between the Pacific and Indian Oceans in Hibiscus tiliaceus, suggesting high gene flow among populations and the absence of genetic barriers across their distribution. Alternatively, it has been suggested that LDD of sea-dispersed plants is more limited than initially expected due to land and water barriers as well as the mobility and survival of diaspores during dispersal (Duke, Lo, & Sun, 2002;Triest, 2008). Indeed, clear genetic structure across land masses has been observed in certain mangroves (e.g. see a review by Triest, 2008). For example, recent studies clarified that the Malay Peninsula acts as a genetic barrier in Bruguiera gymnorhiza (Minobe et al., 2010;Urashi, Teshima, Minobe, Koizumi, & Inomata, 2013) and Ceriops tagal (Huang et al., 2012), and that the Central American Isthmus is a genetic barrier in Hibiscus pernambucensis and Rhizophora mangle (Takayama et al., 2006;Takayama, Tamura, Tateishi, Webb, & Kajita, 2013). Even when no clear land barrier exists, clear genetic structures were reported for mangrove species of Rhizophora (Takayama et al., 2013) as well as Avicennia (Mori, Zucchi, & Souza, 2015), in which the South Equatorial Current likely acted as a barrier to shape genetic structure in the Atlantic region.
Recent studies have reported genetic structure within oceanic regions where no apparent barriers exist, suggesting that inconspicuous barriers in the marine environment may exist (Duke et al., 2002). For example, Wee et al. (2015) detected clear genetic differentiation between South East Asian (Japan, Vietnam, Philippine and Indonesia) and Oceanian (Fiji, Vanuatu and New Caledonia) populations in Rhizophora stylosa and Guo et al. (2016) detected similar genetic structure in Rhizophora apiculata. As the locations of these genetic breaks might correspond to the boundaries of oceanic currents, Wee et al. (2015) suggested that oceanic circulation patterns might have acted as "cryptic barriers." However, the presence of cryptic barriers has not yet been well examined.
Firstly, previous studies (Guo et al., 2016;Wee et al., 2015) did not cover the entire species distributional range and the species studied by them are not widely distributed across the Indo-West Pacific (IWP) region, which is one of the main biogeographical regions for mangroves. Extensive and detailed sampling is needed to test the presence of cryptic barriers, especially in the West Pacific region (sensu Wee et al., 2015). Additionally, extensive sampling may enable the evaluation of the strength of cryptic barriers in comparison to known geographical barriers (land barriers). For a deeper understanding of the process by which genetic structure related to cryptic barriers is formed, a comparative approach is required.
Secondly, to understand cryptic barriers in detail, it is important to consider past population demography, such as past distribution changes, land barriers, LDD and ocean circulation, in addition to examining population genetic structure. Although the past demographic history of sea-dispersed plants has not yet been well examined, recent advances in population genetics with approximate Bayesian computation (ABC) enable a deeper understanding of the past demographic dynamics of species, in terms of evolutionary parameters such as effective population size, time scale of divergence and population size change (Bertorelle, Benazzo, & Mona, 2010). By taking this approach we can deepen our understanding of the processes forming cryptic barriers.
Some adaptations for oceanic dispersal are recognized in this species, such as remarkable salt tolerance (Tomooka, Kaga, Isemura, & Vaughan, 2011), seed buoyancy and viability in sea water (Nakanishi, 1988). Indeed, this species has a wider distributional range than any other tree species of mangroves (e.g. B. gymnorhiza and R. stylosa) in the IWP region. Therefore, this species can be a model species to test the presence of cryptic barriers across the IWP region. Vigna marina has a subspecies, V. marina ssp. oblonga (Benth.) Padulosi, but we did not use it in this study as it is only distributed in West Africa and has been suggested to be not so closely related (genetically) to V. marina (Sonnante, Spinosa, Marangi, & Pignone, 1997).
The aim of this study was to reveal the presence and process of formation of a cryptic barrier in Vigna marina in two ways: (a) to evaluate the genetic diversity and population genetic structure throughout the species distributional range using chloroplast DNA (cpDNA) and show if clear genetic structure caused by a cryptic barrier exists within the west Pacific region, and (b) to infer the past demographic history of V. marina in this region using the ABC approach to assess the processes of formation of the cryptic barrier that shaped the genetic structure.

| Plant sampling and DNA extraction
We sampled Vigna marina in thirty-four localities and the sample size at each location ranged from one to 31 individuals, with a total TA B L E 1 Locality, sample size (N), and values of the genetic diversity parameters: numbers of observed haplotype (S), haplotype diversity (H) and nucleotide diversity (π) in 21 populations of Vigna marina Note. Tajima's D (Tajima, 1989), and Fu and Li's F and D (Fu & Li, 1993) were calculated for populations with cpDNA variations (NA = not analyzed, *p < 0.10, **p < 0.02). We treated some closely located populations (<ca. 100 km) as a single population. Group is the one indicated by SAMOVA (see Figure 1a).
sample size of 308 individuals (Supporting Information Table S1). To avoid sampling closely related individuals, each individual was selected randomly and separated from each other by at least 10 m.
Collected leaf samples were dried by silica gel prior to DNA extraction. For the following data analyses, we treated some closely situated locations (< ca. 100 km) as a single location, resulting in a total of 21 locations (Table 1). Genomic DNA was extracted from dried leaf tissue using the cethyltrimethyl ammonium bromide (CTAB) based extraction method (Doyle & Doyle, 1987).

| Genetic diversity, phylogeny and population structure
The haplotype diversity (H), nucleotide diversity (π), Tajima's D (Tajima, 1989), and Fu and Li's F and D (Fu & Li, 1993) were calculated over all locations, for each of the three SAMOVA groups (see below) and for each location separately using the program DNASP ver. 5.10 (Librado & Rozas, 2009). The haplotype network was constructed and edited using the statistical parsimony procedure implemented in NETWORK ver. 4.613 (www.fluxus-engineering.com) following the median-joining method (Bandelt, Forster, & Röhl, 1999). To evaluate population group structure, we used the spatial analysis of molecular variance (SAMOVA; Dupanloup, Schneider, & Excoffier, 2002) algorithm. SAMOVA is based on a simulated annealing procedure that aims to maximize the F CT and the proportion of total genetic variance due to differences between groups of populations in an analysis of molecular variance (AMOVA), and to define groups of populations that are geographically homogeneous and maximally differentiated from each other. Here, the algorithm was performed by SAMOVA software (Dupanloup et al., 2002), which detected the number, K, of groups giving the largest F CT value. The K was user-defined and set between two and five with 500 independent simulated annealing processes in each run. Pairwise genetic differentiation described as F CT and its standardized value F' CT (Meirmans & Hedrick, 2011) were calculated for all possible pairs of SAMOVA groups using GENALEX 6.5 (Peakall & Smouse, 2012). To evaluate genetic relationships among populations, we reconstructed a neighbor-joining (NJ) tree based on genetic distance (D A ) (Nei, Tajima, & Tateno, 1983) among populations and tested the statistical confidence of the NJ tree topology based on 1,000 bootstraps using Populations 1.2.30 beta software (Langella, 2007). The NJ tree was reconstructed on a topographic map using GeoMapApp (http://www.geomapapp.org/; Ryan et al., 2009) and GenGIS2 (Parks et al., 2009).

| ABC1: Infer population demographic history
To evaluate the demographic history of the three SAMOVA groups, we tested seven simple population divergence scenarios (Figure 1c).
The scenarios were as follows: Scenario The definition of the three populations tested in ABC1 using DIYABC. (a) Results of the spatial analysis of molecular variance (SAMOVA) and pairwise F′ CT value among three groups. (b) The population Neighbor-joining (NJ) tree reconstructed on a topographic map. Each branch was colored based on the results of SAMOVA (Pop1 = orange, Pop2 = blue, Pop3 = green). (c) The seven scenarios tested in ABC1 and the summary of the estimated parameters of the most-likely scenario (= scenario 2). In all scenarios, t# represents time scale measured in number of generations and N# represents effective population size of the corresponding populations (Pop1, 2, 3, "a" before divergence) during the relevant time period (e.g. 0-t1, t1-t2) Scenario

| ABC2: Infer population size change in the North Pacific group
To examine the past transition of the effective population size of the North Pacific group (NP), which had enough haplotype diversity for within group demographic analysis, we tested three simple population demographic scenarios ( Figure 2).
The scenarios were as follows: Scenario We changed the priors of the maximum population size and the maximum values of time scale from 10,000 (default value) to 100,000 to obtain better posterior distributions based on the results from the pilot runs (Supporting Information Table S2). The mutation model was selected using the program JMODELTEST ver. which measures the discrepancy between model and real data.

| Genetic diversity and population structure
In total, 29 haplotypes were detected and high levels of genetic diversity (haplotype diversity, H = 0.864 and nucleotide diversity, π = 1.870 × 10 −3 ) were observed throughout the species distributional range (Table 1). The haplotype diversity, H, ranged from 0.000 to 0.911 and the nucleotide diversity, π, ranged from 0.000 to 2.040 × 10 −3 . High values of haplotype and nucleotide diver-

| Inference of past demographic history
In ABC1 Figure S1 and Table S3). The estimated values of 39 summary statistics did not show significant differences between the observed and simulated data based on the posterior distributions (Supporting Information Table S4). The PCA showed that the observed data were located around the center of the cluster of points of the simulated data based on the posterior distributions (Supporting Information Figure S2), suggesting that scenario 2 was a good fit for the observed data.
In ABC2, the highest value of posterior probability was pre- and 11.40 (95%CI: 7.81 × 10 −1 -1.99 × 10 1 ), respectively (Supporting Information Figure S3 and Table S3). The estimated values of eight summary statistics did not show significant differences between the observed and simulated data based on the posterior distributions (Supporting Information Table S4). The PCA suggested that scenario 2 was a good fit for the observed data (Supporting Information Figure   S4).

| Presence of a cryptic barrier in the West Pacific
This study clearly showed strong genetic structure within the dis- On the other hand, most of the major haplotypes were highly shared among populations within each of the NP or SP clusters, suggesting substantial gene flow among populations within each genetic cluster. In addition, the AMOVA suggested only a low proportion of variation was due to differences among populations within clusters (9.17%; Table 3). These results imply that despite V. marina having the ability for LDD, gene flow between the NP and SP regions has been historically prevented even though no geographical barrier exists.
Interestingly, the level of genetic differentiation in V. marina across the West Pacific cryptic barrier is as strong as across the Malay Peninsula region (F' CT = 0.954 to 1.000; Figure 1a). This highlights that cryptic barriers should be recognized as just as important as other geographic barriers when evaluating the phylogeographic history and genetic diversity of V. marina. Generally, the Malay Peninsula is one of the most important present and historical geographical barriers (Duke et al., 2002;Voris, 2000) forming current patterns of phylogeographic structure and genetic diversity in several mangroves (Duke et al., 2002;Huang et al., 2012;Urashi et al., 2013) and other marine species (Alfaro, Karns, Voris, Abernathy, & Sellins, 2004;Gaither et al., 2011)  A clear genetic break in the West Pacific Ocean has also been observed in some other sea-dispersed plants: Rhizophora stylosa (Wee et al., 2015), R. apiculata (Guo et al., 2016), Xylocarpus granatum (Tomizawa et al., 2017) and Sonneratia alba , and in these cases the locations of the genetic breaks roughly corresponded to the one determined for V. marina. The presence of common genetic structure in multiple widespread sea-dispersed species indicates the presence of a common cryptic barrier. Generally, widespread sea-dispersed plants, including mangroves and "pantropical plants with sea-drifted seeds" (Takayama et al., 2006), often have common genetic structures because they usually share some ecological features (such as habitat, species distribution, seed dispersal strategy) and phylogeographic history. These species are also likely to share common divergence histories relating to geographic and cryptic barriers. In this study, the ABC approach was employed to elucidate the historical processes that formed the clear genetic structure related to the West Pacific cryptic barrier observed in V.
marina. Although these inferences merely represent the population demographic history relating to V. marina, they may also reflect common historical processes shared by multiple widespread sea-dispersed plants distributed in the IWP region.

| Formation of a cryptic barrier in V. marina
The demographic inference estimated that the splitting time be-  (2009)  TA B L E 3 Results of the analysis of molecular variance (AMOVA) performed considering the three cpDNA population groups defined in the spatial analysis of molecular variance (SAMOVA; Dupanloup et al., 2002) in the South China Sea during the LGM period. It is well kwon that richness of alleles or haplotypes is expected to be higher in refugial areas than areas recolonized after the glacial period (Comps, Gömöry, Letouzey, Thiébaut, & Petit, 2001;Petit et al., 2003).  (Tomizawa et al., 2017); Sonneratia alba ) in relation to the LGM.
These findings suggest that not only the widespread temperate forest plants, but also the sea-dispersed tropical plants have been influenced by glacial climate change. By studying other sea-dispersal plants in the same way, and also by using higher resolution markers such as multi-locus nuclear DNA, we may obtain further support for this idea.
In this study, ABC analyses explored the population demographic history in relation to refugia during LGM or the past ice age previous to the LGM. Although statistical uncertainty remains (e.g., time scale, generation time, broad 95% CI range and assumed model, reviewed in detail see Tsuda, Nakao, Ide, & Tsumura, 2015), the ABC inferences used in this study were useful to explore the processes by which the cryptic barrier in widespread sea-dispersed plants is formed. Regarding the model assumed in the ABC, DIYABC does not consider gene flow after divergence, and it may bias the inferred temporal parameters (Tsuda et al., 2015). However, as discussed previously, genetic differentiation among the examined regions was quite high (F′ CT = 0.954 to 1.000), suggesting highly restricted gene flow among them. Thus, as far as the demographic history among the three regional groups is concerned, the bias in this study is likely limited. In addition, as gene flow after divergence wasn't considered, divergence times might have been underestimated (Tsuda et al., 2015).
However, the main discussion here would not be changed, with the population split still occurring before the LGM.

| CON CLUS ION
In this study, the presence of a cryptic barrier in the West Pacific distributional range of a widespread sea-dispersed plant, V. marina, was clearly detected. The locations of the cryptic barrier observed in V. marina corresponded to the genetic breaks found in other plants, suggesting the presence of a common cryptic barrier for sea-dispersed plants in the West Pacific region. Demographic inferences using the ABC approach suggested that genetic structure related to the cryptic barrier has been present since the last glacial period and may reflect patterns of past population expansion from refugia. These results suggest that despite high LDD ability that can homogenize genetic structure, present phylogeographic patterns of sea-dispersed plants have been strongly affected by glacial climate changes, as reported in other land plants.