Introduction

Mangrove forests are the most critical component of coastal and estuarine ecosystems covering over 150,000 km2 in 123 countries across the tropics and the subtropics1. They provide tremendous ecological services for human beings, however, deforestation of mangrove forests are ongoing worldwide by anthropogenic as well as climate changes2,3. Conservation of mangrove forests is becoming a global concern, and local, regional and nation-wide conservation activities are prominent in the world, however, our understandings on the global mangrove are limited because studying mangrove across nations and continents over pantropical distribution was difficult4.

The present pantropical distribution of mangrove plants could be due to the long-distance dispersal of their diaspores (seed, fruit, or propagule) by ocean currents5. Late Cretaceous fossil evidence of mangroves6,7 suggested the presence of mangrove over the world at that time. However, the evolutionary process and regional diversification of mangrove species to form the present pantropical distribution of mangrove forests are poorly understood. Uncovering the formation process of the pantropical distribution of mangroves will help deeper understandings of the global mangrove and provide us new insight into future changes of the mangrove ecosystem under global climate change.

Previous studies inferred a historical process of the pantropical distribution formation by considering the interactions between sea-dispersal and geological history based on the distribution patterns and fossil records of mangroves5. Duke et al.8, in particular, suggested constructing the phylogenetic history of a pantropical genus Rhizophora L. (Rhizophoraceae), which is a key taxon in revealing the global and disjunctive distribution patterns of mangroves. The genus Rhizophora dominates most of the mangrove forests in both Indo-West Pacific (IWP) and Atlantic-East Pacific (AEP) regions9,10, and the fossil records of the genus and family are relatively abundant owing to the distinct shape of the propagules. By studying the fossil records of the genus in detail, a scenario in which two lineages of Rhizophora diversified for IWP and AEP lineages owing to the closure of the Tethys Seaway in the Oligocene was proposed5. This scenario was the most comprehensive one at that time; however, it was merely a deduction based on fossil records and geological history.

Many mangrove researchers tried to understand the most intriguing question, that is, the historical process of formation of the world mangroves, focusing on the genus Rhizophora by using molecular markers11,12,13,14,15,16. Divergence times between the species in IWP and AEP regions have been estimated by using various molecular markers and sampling coverages. Lo et al.14 suggested that the divergence time between IWP and AEP lineages of Rhizophora was approximately 48 (± 3) Ma (million years ago) using cpDNA, ITS, and ISSR markers based on samples collected from a relatively wide range of distribution, however, they used limited numbers of AEP samples and no samples from West Africa. Chen et al.15 estimated the divergence time to be approximately 12.7 Ma (95% HPD (Highest Posterior Density), 10.4 Ma-15.5 Ma) using five nuclear genes based on less number of samples and used only two AEP samples from West Africa. Although it was not stated in the paper at all, a supplementary data of Xu et al.16 implied that the divergence time could be 10.8 Ma (95% credible interval 9.2 Ma – 12.8 Ma) using 590 single-copy genes extracted from genomic data based on only one representative sample from the AEP region. Even though Chen et al.15, as well as Xu et al.16 provided more reasonable divergence time of the two potential lineages of Rhizophora based on longer nucleotide sequences than any other previous studies, their study was based on only one or two representative samples from AEP Rhizophora species without showing the most critical basis for the global distribution, that is, the monophyly of AEP and IWP lineages respectively.

To understand the comprehensive picture of the phylogeographic process of Rhizophora, phylogenetic studies including all species as well as samples covering distinct lineages and wide distribution range are essential. Multiple samples from the wide distribution range of the AEP region are particularly important because the monophyly of the AEP lineages has not been clarified by nuclear DNA data, even though the presence of considerably distinct lineages across the American continents and West Africa was reported based on cpDNA and microsatellite data13. Numbers of distinct lineages were differentiated on both Pacific and Atlantic sides of the American continents, but none of the previous studies on the global phylogeny of Rhizophora included those distinct lineages in their phylogenetic analyses.

The sibling relationship between Rhizophora samoensis in South Pacific Islands in the IWP region and Rhizophora mangle in the AEP region has long been an enigmatic question discussed based on their morphological similarity10, and have recently been answered by molecular analyses using cpDNA sequences and microsatellite13. Rhizophora samoensis has an identical haplotype with the one of R. mangle in the Pacific coast of the American continents, which suggests at least a single long-distance migration occurred across the East Pacific. This unusual trans-pacific long-distance dispersal might have provided the opportunity for the two genetically distinct lineages, IWP and AEP lineages, to meet in the South Pacific Islands. Previous studies reported the presence of a putative hybrid species, R. × selala, between the two distinct lineages12,14,17,18,19, however, the phylogeographic process of the hybrid formation and hybridization patterns across the South Pacific Islands are still in question.

In this study, to obtain the most comprehensive scenario of historical processes that shaped the present pantropical distribution of mangrove forests, we performed the following analyses based on global sampling, including all species of Rhizophora as well as regional representative populations over the world. We performed 1) phylogenetic analyses using both cpDNA and nuclear DNA markers to obtain robust phylogeny to test the monophyly of the AEP lineage of Rhizophora in particular, 2) phylogeographic analyses to understand the historical process over the broad distribution range using the same molecular markers, and 3) estimation of divergence time by cpDNA phylogeny to give a robust time scale of mangrove diversification followed by ancestral range estimation to know the formation history of the distribution range. In addition, to understand the patterns of hybrid formation between AEP and IWP lineages of Rhizophora at the South Pacific. We also performed 4) population genetic analyses by comparing cpDNA and nuclear DNA sequences plus microsatellite genotypes. By incorporating the results obtained from all the above analyses, we propose an updated historical scenario of the formation process of the pantropical distribution of Rhizophora.

Results

Sequence variability

cpDNA–-The aligned sequences of the four cpDNA regions were 3912 bp containing 39 polymorphic sites, which produced 12 haplotypes in 89 samples of Rhizophora (Table 1, Table S1). Five and eight haplotypes were found in the AEP and IWP regions, respectively, and only the haplotype CD was shared between the two regions (Table S2, Fig. 1). The haplotype diversity was higher in IWP species than AEP species regardless of the inclusion or exclusion of R. samoensis and the three putative hybrid species from the analyses (Table 2). The nucleotide diversity and theta were higher in IWP species in all including analyses, but these values were similar in excluding analyses of R. samoensis and the putative hybrid species (Table 2).

Table 1 Localities of sampled populations of Rhizophora and Bruguiera species and sample size for chloroplast (NCP) and nuclear DNA (NNUC) analyses.
Full size table
Figure 1
figure 1

Trees of maximum parsimony topologies and geographical distributions of the haplotype/allele from combined chloroplast DNA sequences (a), and nuclear DNA sequences (b, CesA and c, G3pdh) of Rhizophora species. The ML analysis yield congruent MP topology (not shown). Numbers above branches indicate bootstrap probabilities greater than 50% for parsimony (left) and likelihood (right), respectively. Colored circles on the right of OTU indicate DNA sequences shown on the maps (d–f) indicating localities of sampled populations. Colored pie chart on the map indicate the frequency of chloroplast DNA haplotypes (d), and nuclear DNA alleles (e, CesA and f, G3pdh) of Rhizophora species except for hybrid species. Abbreviations in brackets in the tree and along pie chart on the map indicate species names: man, R. mangle; rac, R. racemosa; sam, R. samoensis; sty, R. stylosa; api, R. apiculata; muc, R. mucronata; × ha, R. × harrisonii, × la, R. × lamarckii, × se, R. × selala.

Full size image
Table 2 Genetic diversity parameters estimated by chloroplast and nuclear DNA sequences.
Full size table

CesA–-The aligned sequences of the CesA were 555 bp long containing 42 polymorphic sites, which produced 15 alleles in 397 samples of Rhizophora (Table 1, Table S1). Six and ten alleles were found in AEP and IWP species, respectively (Table S2, Fig. 1). The allelic diversity was higher in IWP species, but the nucleotide diversity and theta tended to be higher in AEP species in both analyses (Table 2).

G3pdh–-The aligned sequences of the G3pdh were 668 bp, there were, however, a ca. 60 bp ambiguously aligned region because of poly T and C complex. Therefore, we excluded the complex region for further analysis. As a result, 608 bp was aligned in G3pdh containing 31 polymorphic sites, which produced 14 alleles in 397 samples of Rhizophora (Table 1, Table S1). Seven and eight alleles were found in AEP and IWP species, respectively (Table S2, Fig. 1). The allelic diversity was similar in AEP and IWP species, and the nucleotide diversity was higher in IWP species in both the analyses (Table 2).

Phylogenetic relationships

FindModel identified the best model for each gene: GTR + gamma for cpDNA, Tamura-Nei + gamma for CesA, and Hasegawa-Kishino-Yano + gamma for G3pdh. The results of MP and ML showed congruent topology for the monophyly of each AEP and IWP species except for R. samoensis and R. × selala (Fig. 1). Rhizophora samoensis had an identical haplotype and allele with AEP species clade, and R. × selala had the same cpDNA haplotype with AEP species and same nuclear alleles with the AEP and IWP clades in hetero. Rhizophora mangle and R. racemosa G. Mey. did not make its own clade within the AEP clade. Rhizophora apiculata Blume formed a monophyletic clade with R. × lamarckii Montrouz., and R. stylosa Griff. and R. mucronata Lam. formed a monophyletic clade together in the IWP clade.

Minimum spanning network of haplotypes

All three Minimum Spanning Networks constructed using haplotypes of cpDNA, CesA, and G3pdh showed clear separation of IWA and AEP groups (Fig. S2). A haplotype possessed by Rhizophora species (colored in orange: GH in cpDNA, F in CesA, G in G3pdh) was found only from the Atlantic region and was placed at the basal position of the AEP group in each network.

Global population genetic structure

Global distribution of cpDNA haplotype and nuclear allele in the six non-putative hybrid species are shown in Fig. 1. The details of haplotype and allele frequency are in Table S2. The patterns of nuclear alleles in putative hybrid species were shown in Table 3 and Table S3. There are only one cpDNA haplotype (CD) and CesA and G3pdh alleles (both B) which can be observed in both AEP and IWP species. These haplotypes and alleles were commonly and geographically-widely found in the species in the AEP region; however, they were found only in R. samoensis and R. × selala in the IWP region.

Table 3 Localities of sampled populations of Rhizophora samoensis, R. stylosa, and R. × selala in New Caledonia and Fiji, and sample size for PCR–RFLP and microsatellite analyses.
Full size table

In the AEP region, each haplotype and allele were widely distributed. A remarkable genetic differentiation between the Pacific and the Atlantic side were found in cpDNA haplotypes, although there are few overlaps between the two regions in terms of both nuclear DNA alleles. Populations on the Pacific coast in Mexico and Costa Rica had the alleles which widely dominant in the Atlantic region. In the IWP region, there are several haplotypes and alleles which were widely distributed or unique to a single locality.

Divergence times

To estimate divergence time for deeper nodes within Rhizophora, we conducted phylogenetic analyses with two selected cpDNA regions (rbcL and atpB-rbcL intergenic region) and two nuclear genes combining with the available data set deposited in DDBJ (DNA Data Bank of Japan). The most recent common ancestor of Rhizophora and sister mangrove genera, Ceriops Arn. and Kandelia (DC.) Wight et Arnott, was estimated to be 36.0 Ma (95% HPD 34.5–38.2 Ma: Fig. 2) by using cpDNA data including 27 OTUs gathered in the Bayesian analysis with the two calibration points. The split between AEP and IWP clades of Rhizophora was estimated to be 10.6 Ma (5.4–16.3 Ma). In the AEP clade, the split between R. racemosa and R. mangle/R. racemosa/R. samoensis clades were estimated to be 3.0 Ma (0.5–6.0 Ma) in the cpDNA analysis. In the IWP clade, the split between R. apiculata and R. stylosa/R. mucronata clades were estimated to be 7.4 Ma (3.1–12.2 Ma). The estimation time of the split between AEP and IWP clades estimated using the first calibration point only was 10.3 – 34.0 Ma (2.2 – 53.7 Ma) depending on the data set (Table S4).

Figure 2
figure 2

Chronograph and biogeographic history for Rhizophora and outgroup taxa based on the relaxed-clock Bayesian MCMC methods in BEAST and BioGeoBEARS in RASP using chloroplast DNA sequences. Error bars on nodes indicate 95% highest posterior densities around the mean dates. Arrows indicate fossil calibration points described in Methods. Probability of ancestral states are shown in pie chart. *Rhizophora samoensis is distributed in the IWP region but has an identical haplotype with AEP species.

Full size image

Biogeographic inference

Biogeographic stochastic mapping analyses in BioGeoBEARS yielded the BAYAREA + J model as the best- fitting model for the phylogeny of all Rhizophoraceae species (Table S5). The ancestral ranges of IWP and AEP were Indo-Malesia and Australasia (EF) (ML probability = 0.60) and East America and West Africa (BC) (ML probability = 0.26) (Fig. 2). The estimated ancestral range of all AEP species was East America and West Africa (BC) (ML probability = 0.92) (Fig. 2).

Hybridization patterns in the South Pacific Islands

The results of PCR–RFLP showed that all individuals of R. stylosa and R. samoensis including their hypocotyls possessed their specific genotypes in cpDNA and nuclear DNA, and all individuals of R. × selala possessed a particular combination of cpDNA haplotype, R. samoensis genotype, and nuclear DNA genotypes, R. stylosa and R. samoensis in hetero, which can be expected of F1 hybrid genotype (Tables S2 & S3). All propagules had the same genotypes as their mother tree of R. stylosa or R. samoensis. The results of STRUCTURE based on microsatellite data were shown in Fig. 3 and Fig. S1. The value of log-probability began to plateau at K = 4 and the highest delta K value was returned at K = 2 so that we showed K = 2 and K = 4 as most probable numbers of the cluster. In K = 2, the two clusters represented R. stylosa and R. samoensis, respectively, and both clusters can be admixed fifty-fifty in three populations of R. × selala. In K = 4, each of the two clusters represented R. stylosa or R. samoensis mostly depending on the locality. Red and green clusters were found in R. stylosa from New Caledonia or Fiji, and blue and orange ones in R. samoensis. Two clusters found in each locality were admixed in R. × selala.

Figure 3
figure 3

Results of STRUCTURE of Rhizophora species in the South Pacific Islands. Vertical columns represent individual plants, and the heights of bars of each color are proportional to the posterior means of estimated admixture proportions. The number of cluster K = 2 and K = 4 are shown.

Full size image

Discussion

Phylogeny and timing of diversification of Rhizophora

This study provided the most comprehensive phylogeny that explains the present pantropical distribution of a mangrove genus Rhizophora. We included distinct lineages from a wide range of the AEP region and showed that all AEP samples were in the AEP clade that is sister to the IWP clade in all analyses using multiple DNA markers (Fig. 1). Although trans-oceanic dispersal and frequent hybridization have been reported in the genus Rhizophora, this result ensures an adequate discussion of the divergence time between the AEP and IWP lineages of Rhizophora based on the phylogenetic tree.

The divergence time between the AEP and IWP lineages was 10.6 Ma by using cpDNA and broad outgroup taxa (Fig. 2) and 11.0–11.5 Ma by a combination of three different regions (cpDNA, CesA, and G3pdh) (Table S4) which are similar to the ones, 12.7 Ma by five nuclear genes15 and 10.8 Ma by 590 coding genes16 and far different from 48 (± 3) Ma by cpDNA and ITS14. Because our study based on global sampling showed the monophyly of the AEP lineage and similar estimates for the divergence time between the IWP and AEP linages, we consider that the estimates of Xu et al.16 also can be reasonable even though they used only one sample from the AEP region.

The separation of the AEP and IWP lineages reasonably coincide with the closure of the Tethys Seaway. The biogeographic stochastic mapping analyses suggested that the common ancestor of the AEP and IWP lineages was separated into two ancestral ranges 10.6 Ma: the Atlantic region (BC) for the AEP lineage, and the Indo-Pacific region (E, F, and EF) for the IWP linage (Fig. 2). The separation can be attributed to the formation of the land barrier at Tethys Seaway that separated the Atlantic and Indo-Pacific regions around 11 to 20 Ma20,21,22,23,24. The common ancestor of the AEP lineage at the Atlantic (BC) was suggested to expand the range to the Pacific regions (A and ABCF) in the later diversification (Fig. 2). Although there were only three extant AEP lineages of Rhizophora in the dated tree (Fig. 2) due to less length of sequence data, longer length of cpDNA sequence showed the presence of five haplotypes (AB, CD, E, F, and GH) (Fig. 1a, Fig. S2), and three haplotypes (CD, E, and F) were exclusively found in the Pacific region (Fig. 1a,d). This remarkable segregation between the Pacific and Atlantic populations can be attributed to the closure of the Isthmus of Panama at 3 Ma25 after the range expansion of Rhizophora species from Atlantic to the Pacific.

Our results provided clear evidence of the recent trans-pacific migration of R. mangle from the Pacific side of the New World (AEP) to the South Pacific Islands (IWP)13,14. The haplotype possessed by R. samoensis from the South Pacific Islands and R. mangle from the Pacific side of the New World was identical both in cpDNA (CD) and nuclear DNA markers (B in both markers) (Fig. 1). Our data also confirmed that R. × selala in the South Pacific Islands was a F1 hybrid asymmetrically formed by R. stylosa as father lineage and R. samoensis as mother lineage. In addition, our microsatellite data suggested that independent events of hybrid formation occurred recurrently in New Caledonia and Fiji (Fig. 3). These results suggest that R. × selala are F1 hybrids between the species from IWP and AEP lineages. Although hybrids are recurrently formed when the two lineages meet, they are sterile F1 hybrids because of strong reproductive isolation caused by long separation of parental lineages for about 11 Ma. Our results supported the previous observation that R. × selala had low pollen fertility and abnormal tetrads18,19.

Updated scenario for the formation of the distribution pattern of Rhizophora

According to our findings and some previous studies, we finally updated the historical scenario behind the formation of the distribution pattern of the mangrove genus Rhizophora, proposed by Duke et al.5 We illustrated a synthetic scenario of the historical process with the tectonic events and changes of global temperature that shape the present pantropical distribution of the mangrove genus Rhizophora in Fig. 4 and as followings:

Figure 4
figure 4

Historical scenario of diversification of Rhizophora. Concurrent significant tectonic events, global temperature, and fossil occurrences have been listed for comparison with the result of phylogenetic analysis in this study. The patterns of global temperature and fossil occurrences were modified from previous studies7,30.

Full size image

Eocene—Oligocene (50–23 Ma)

Rhizophora originated during the Eocene and was distributed worldwide by the Middle and Late Eocene evidenced by several comprehensive assessments of fossil records6,7,26,27,28,29. Our biogeographic analysis supported that the origins of extant Rhizophora were mostly in IWP region and further spread to AEP region (Fig. 2). The mangrove habitat might have been more widely extended in longitude than at present, because the average temperature of the earth was 4–12 °C higher30. In fact, the pollen of mangrove palm, Nypa, was reported from the Early Eocene layer in Tasmania31 and New Zealand32, where no mangrove habitat occurs at present. The wider longitudinal distribution area of mangrove might be achieved by the migration of propagules in the east–west direction between the Old World and New World, via the southern route around the African continent. Furthermore, ocean currents through “Tethys Seaways (Mediterranean Sea)” also could work as a dispersal corridor between the two regions (Fig. 4, "Globalization of Rhizophora").

Miocene (23–10 Ma)

By geological approaches, the closure of Tethys Seaways was estimated from 11 to 20 Ma20,21,22,23,24 which is consistent with our estimation of the divergence time ca. 11 Ma for the deep phylogenetic break between IWP and AEP lineages. The closure of the Tethys Seaway created a land barrier between the IWP and AEP regions33. In addition, Mid-Miocene cooling occurred at the same time30 would narrower the distribution range to the equator and prevented the migration of propagules via the southern route around the African continent (Fig. 4, "IWP & AEP separation").

Late Miocene to Present (10 Ma-)

The global cooling at Late Miocene and Pleistocene might have promoted a decrease in the population size for many mangrove species via sea level drop, and separation to regional populations may lead allopatric speciation in Rhizophora. In the AEP region, both R. mangle and R. racemosa expanded their distribution across the American continents from the Atlantic region via migration through the Panama Seaway. After the closure of the Panama Seaway at 3 Ma25, genetic differentiation between the Pacific and Atlantic populations of both species occurred. In the IWP region, the R. apiculate, R. stylosa, and R. mucronata might form their present distribution through different demographic histories in each species despite having sympatric distributions today34,35. Finally, eventual trans-Pacific dispersal of R. mangle happened from American continents to the South Pacific. The extremely long-distance dispersal caused secondary contact between distinct lineages of Rhizophora, which can independently produce the sterile F1 offspring in the South Pacific Islands (Fig. 4, "Trans-Pacific dispersal from AEP to IWP").

Methods

Plant collections

We collected leaf samples from 56 populations of all eight species including three putative hybrid species across the IWP and AEP regions for phylogenetic analyses (Table 1). Upon collection, leaf samples for DNA extraction were put in reclosable plastic bags and dried with silica gel. We also collected 15 populations of R. stylosa, R. samoensis, and R. × selala in New Caledonia and Fiji for investigation of hybridization pattern in the South Pacific Islands (Table 3). In addition, a part of the hypocotyl of propagule for DNA extraction was also collected from R. samoensis and R. stylosa in Fiji (Table 3). Twenty-nine populations from the AEP region and nine populations form the IWP region using in this study are the same as those used in a previous study13. Identification of the species was done following the keys in Tomlinson10 and at least one voucher specimen from each population and species was deposited in the Herbarium at the Herbarium of University of Ryukyus (URO).

DNA extraction and sequencing

Genomic DNA was extracted from the dried leaf and hypocotyl tissue using the cetyltrimthyl ammonium bromide (CTAB) extraction method36 and purified using GENECLEAN III Kit (MP Biomedicals, Solon, USA). PCR amplification and sequencing were performed using 1 to 18 individuals from each population (Table 1). Sequences of four cpDNA regions, atpB-rbcL intergenic spacer (IGS), trnS-trnG, rpl16 intron, and rbcL gene, were determined with the same protocol as in the previous study13. Other available sequences from Rhizophoraceae were obtained from GenBank accessions (Table S1). In addition, a part of rpl16 intron sequence was determined in 165 individuals of R. stylosa, R. samoensis, and R. × selala, and 64 hypocotyls of the propagule of R. stylosa and R. samoensis in New Caledonia and Fiji. Two nuclear DNA regions, glyceraldehyde 3 phosphate dehydrogenase (G3PDH), and cellulose synthase (CesA) homolog were amplified using primers reported in Strand et al.37 and Cronn et al.38. Sequencing reaction was performed with using primers GPDX7F37 and G3PDH-788R1 (5′-CAATGAAGTCTGTGGATACCAA-3′) for G3PDH, and CesA-1150F (5′-CCACCTGAGCAGCAGATGGAAG-3′) and CesA-1800R (5′-ACGACAGTTGAAAGTGGCTGTGC-3′) for CesA. In cases where the phenogram of the two nuclear DNA regions showed duplicated peaks in more than two sites, we segregate the alleles of the PCR amplicon with single-strand conformation polymorphism (SSCP) analyses39 and re-sequenced after acrylamide gel extraction.

Phylogenetic inference

Genetic diversity parameters were estimated using DnaSP 5.1040. Phylogenetic analyses were performed using the Wagner maximum parsimony (MP) and Maximum Likelihood (ML) methods based on three different matrices, the combined sequence data of four cpDNA, G3PDH and C3esA, separately. Bruguiera gymnorrhiza was used as outgroup for all of the Rhizophora species according to previous phylogenetic study11. MP and ML analyses were conducted using PAUP*41 and RAxML42, respectively. For the ML analyses, FindModel (http://hcv.lanl.gov/content/hcvdb/findmodel/) was used to find the best fitting model for nucleotide sequence evolution.

Minimum spanning network

We used PopART43 to construct a minimum spanning network44 using the alignments of haplotype and allele sequences of cpDNA, G3PDH and C3esA.

Divergence time estimates

To test temporal aspects of phylogeographic breaks, we inferred divergence times using a relaxed-clock method in the program BEAST v1.6.245. Two chloroplast regions, rbcL and atpB-rbcL, were used to conduct comprehensive analyses by including compatible data from other genera in Rhizophoraceae in previous phylogenetic study11. In addition, two nuclear regions, CesA and G3pdh obtained in this study were used for the dating analyses. We estimated divergent times using 12 different combinations of data and taxon because the nucleotide diversity and available sample of the outgroup taxa were different depending on the DNA regions. As previously reported13,14, Rhizophora species had experienced hybridization and shared haplotype/genotype partially, hence the sorting of haplotypes/genotypes into a sample increased the number of OTUs which shared identical haplotype or genotype. We first conducted a multispecies coalescent analysis46 in the dating analyses using the combination of different genes. The multispecies coalescent analysis estimates a species (taxon) tree based on unlinked multi-locus sequence data, taking into account that gene trees are incorporated in a shared species (taxon) tree by following the stochastic coalescent process. For the analysis that included multiple regions, three OTUs from Rhizophora samples, such as R. stylosa—R. mucronata, R. apiculata, and R. mangle – R. racemosa were used for the coalescent analysis. The analysis was conducted using the program BEAST v1.6.245. We used the fossil record to place priors on the crown age of a node, arising of mangrove clade (Bruguiera Lam., Kandelia, Ceriops, and Rhizophora) in Rhizophoraceae. In the review of mangrove fossil record7, fruits and pollen of ancestors very close to modern mangrove genera in Rhizophoraceae existed in the Early Eocene (50 Ma), but not in the Palaeocene (55 Ma) when the fossil record of another mangrove genus, Nypa, is abundant worldwide. Therefore, we set a lognormal prior offset of 50 Ma with a mean and standard deviation of 0.5 for the first calibration point. In addition, ancestral Rhizophora fossil has been reported to the late Eocene (33.9–38 Mya)47. We set a lognormal prior offset of 34 Ma with a mean and standard deviation of 0.5 for crown age of a node of a monophyly clade of Ceriops, Kandelia, and Rhizophora for the second calibration point. We used the two calibration points for the cpDNA analyses using broad species in Rhizophoraceae. On the other hand, we only used the first calibration point for 12 different combinations of data and taxon due to the limitation of the taxon coverage to use the second calibration point. We conducted three independent searches of 10 × 107 generations sampling every 1000 generations under the models selected by FindModel in each genetic region. The results log files from the three runs were combined by the program LogCombiner v1.7.545. We confirmed that ESS values of all estimated parameters after 20% burn-in were > 200 in the program Tracer v1.545. We also generate a maximum clade credibility tree using 20% burn-in trees obtained in three runs by LogCombiner v1.7.5 and TreeAnnotator v1.7.545 and presented the mean and 95% highest posterior density (HPD) of node ages by FigTree v1.448.

Biogeographic inference

To infer the ancestral range for Rhizophora, we used the ML method implement in BioGeoBEARS 0.2.149 in the RASP 4.2 package50. We defined the geographical ranges as A = West America, B = East America, C = West Africa, D = East Africa, E = Ind-Malesia, and F = Australasia according to the previously reported definitions of geographic ranges of Rhizophora5. We estimated ancestral ranges and biogeographical events on the BEAST MCC trees of Rhizophoraceae species based on two cpDNA regions (Fig. 2), using a likelihood-based framework in BioGeoBEARS that allows for testing various biogeographical models. The best model was selected using AIC corrected for sample size (AICc). We rerun the BioGeoBEARS analysis on 100 randomly sampled post-burn- in BEAST chronograms. The number of maximum areas per ancestral range was set to four.

PCR–RFLP and microsatellite analysis

To reveal hybridization patterns in the South Pacific Island, PCR–RFLP and microsatellite genotyping were conducted in 165 individuals of R. × selala, R. samoensis, and R. stylosa in New Caledonia and Fiji, and 64 hypocotyls of propagule of R. samoensis and R. stylosa in Fiji (PCR–RFLP only). For PCR–RFLP, the amplified G3PDH and CesA fragments in the 165 individuals and 64 hypocotyls were digested with restriction enzymes, MspI (CCGG) and Sau3A (GATC), respectively. Based on the sequence data, we expected the length of DNA fragments of R. stylosa after this treatment to be ca. 10, 30, 670 bp in G3PDH and 50, 50, 70, 390 bp in CesA, and those of R. samoensis to be 10, 700 bp and 50, 70, 440 bp. Digested DNA fragments were detected 2% agarose electrophoresis, and determined genotypes of two nuclear DNA regions. In addition, four microsatellite markers (RM50, RS19, RS33, and RS59) developed by Takayama et al.51,52 were selected for the analysis of hybridization patterns in New Caledonia and Fiji, because the four markers can be constantly amplified in both AEP and IWP species. We determined microsatellite genotypes in 165 individuals of R. × selala, R. samoensis, and R. stylosa in New Caledonia and Fiji. PCR amplification and genotyping were performed according to the previous study13. Genetic structure was evaluated by the Bayesian clustering method using STRUCTURE 2.3.353,54,55. Markov chain Monte Carlo (MCMC) searches consisted of 100,000 “burn-in” steps, followed by 100,000 iterations. Twenty replicate runs were performed at each K from 1 to 10 and calculated the log-probability and the delta K proposed by Evanno et al.56, given a certain value of K for seeking the best-fit K value for the data.