Genomic Markers Reveal Introgressive Hybridization in the Indo-West Pacific Mangroves: A Case Study

Biodiversity of mangrove ecosystems is difficult to assess, at least partly due to lack of genetic verification of morphology-based documentation of species. Natural hybridization, on the one hand, plays an important role in evolution as a source of novel gene combinations and a mechanism of speciation. However, on the other hand, recurrent introgression allows gene flow between species and could reverse the process of genetic differentiation among populations required for speciation. To understand the dynamic evolutionary consequences of hybridization, this study examines genomic structure of hybrids and parental species at the population level. In the Indo-West Pacific, Bruguiera is one of the dominant mangrove genera and species ranges overlap extensively with one another. Morphological intermediates between sympatric Bruguiera gymnorrhiza and Bruguiera sexangula have been reported as a variety of B. sexangula or a new hybrid species, B. × rhynchopetala. However, the direction of hybridization and extent of introgression are unclear. A large number of species-specific inter-simple sequence repeat (ISSR) markers were found in B. gymnorrhiza and B. sexangula, and the additive ISSR profiling of B. × rhynchopetala ascertained its hybrid status and identified its parental origin. The varying degree of scatterness among hybrid individuals in Principal Coordinate Analysis and results from NewHybrids analysis indicate that B. × rhynchopetala comprises different generations of introgressants in addition to F 1s. High genetic relatedness between B. × rhynchopetala and B. gymnorrhiza based on nuclear and chloroplast sequences suggests preferential hybrid backcrosses to B. gymnorrhiza. We conclude that B. × rhynchopetala has not evolved into an incipient hybrid species, and its persistence can be explained by recurrent hybridization and introgression. Genomic data provide insights into the hybridization dynamics of mangrove plants. Such information can assist in biodiversity assessment by helping detect novel taxa and/or define species boundaries.


Introduction
Mangrove forests consist of an important group of woody plants occupying coastal zone habitats. Global distributions of mangroves are mainly influenced by temperature [1], restricting them to warm tropical and subtropical latitudes in the Indo West Pacific (IWP) and Atlantic East Pacific (AEP) regions. Although these plants and associated organisms and habitats constitute one of the world's most productive ecosystems [2], extant mangrove taxa worldwide remain incompletely described and poorly identified, which limits our understanding of mangrove biodiversity and evolutionary relationships among the major constituents. In addition to convergent evolution in morphology, frequent appearance of new taxonomic entities through natural hybridization may have contributed to the difficulty in mangrove biodiversity assessment.
Natural hybridization is common in plants and plays a very important role in evolution as a source of novel gene combinations and a mechanism of speciation [3][4][5][6][7][8][9][10]. However, recurrent hybridization and introgression result in gene flow between species which could reverse the process of genetic differentiation among populations required for speciation. To understand the dynamic evolutionary consequences of hybridization, more genetic studies are needed to compare the hybrids and their parental species at the population and genomic level. Such studies are also important for accurately classifying and managing biodiversity.
Conventional approaches to hybrid identification are primarily based on their morphological intermediacy between parental species. However, many morphological characters are under environmental influences and some true hybrids may not always display an intermediate phenotype if the diagnostic traits are not controlled by codominant genes or genes of additive effects [11]. The inadequacy of morphological approach alone in studying natural hybridization is well recognized [12 and references therein]. More recent studies have employed various molecular technologies for identifying natural hybrids, for detecting introgression, and for studying hybrid speciation [e.g., [10][11][12][13][14].
Bruguiera Savigny is one of major mangrove genera of Rhizophoraceae. The genus consists of only six species which are divided into two groups according to flower size and pollinator type [1]. The two large-flowered species, Bruguiera gymnorrhiza (L.) Lamk. and Bruguiera sexangula (Lour.) Poir., have wide distributions in the Indo-West Pacific region and frequently occur in sympatry throughout Southeast Asia and tropical Australia. Morphological similarities between the two species have often led to identification uncertainties [15], and the presence of intermediate forms in sympatric populations further contributes to taxonomic difficulties. Ko (1978) treated the intermediate forms occurring in China as a variant of B. sexangula, and named it Bruguiera sexangula var. rhynchopetala Ko [16]. The observations of several morphological attributes that are intermediate between B. gymnorrhiza and B. sexangula suggest that this new taxon is most likely an interspecific hybrid, and a hybrid species name, Bruguiera 6rhynchopetala (Ko) X. J. Ge et N. C. Duke, was thus proposed to denote its putative hybrid origin [17]. Based also on morphological observations, hybridization and introgression between B. gymnorrhiza and B. sexangula in Sri Lanka was speculated but yet to be confirmed [18]. In addition, the frequency of hybridization and extent of introgression in different geographical locations is unknown for the genus.
Various molecular markers have been explored for genetic investigation of plant hybridization. Of which, a large number of inter-simple sequence repeat (ISSR) markers can be easily generated across the genome for comparative analysis of the putative hybrids and their parental species. Wolfe et al. (1998) have shown that ISSR markers could offer a high degree of resolution to relationships and patterns of introgression than other types of molecular data used in hybridization studies [11]. In this study, we investigate natural hybridization in Bruguiera over a wide geographical area in the IWP region using ISSR together with sequence markers. Specifically, we address the following questions: (1) Do molecular data support the morphology-based field identification of Bruguiera 6 rhynchopetala? (2) Are B. gymnorrhiza and B. sexangula the only parental species involved in hybridization where other congeners also occur in sympatry? (3) What is the direction and extent of introgression at each of the examined geographical locations? (4) Has Bruguiera 6 rhynchopetala genetically differentiated from the parental species to deserve a separate hybrid species status?

Plant sampling
Fresh young leaves were collected from individual trees in four mangrove forests, including one site in Hainan Island of South China Sea, one site in North Sulawesi of Indonesia, and two sites in northeastern Australia. The sampled trees were targeted to include diverse morphs of parental species and their putative hybrids present at each site (Table 1). Hybrids were identified based on their unique intermediacy or a combination of morphological characteristics of the putative parents. All individual samples were assigned a field collection identification number and kept separately for genetic analyses.
The sample size varied among sites depending on availability of identifiable hybrids in each mangrove forest. On the northeastern coast of Hainan Island, B. gymnorrhiza and B. sexangula partly overlap in Dongzhai Mangrove Nature Reserve, and frequent occurrence of intermediate forms at this site resulted in a discernible hybrid zone. In contrast, the hybrid forms were relatively rare or difficult to ascertain in the Indonesian population located in North Sulawesi, as in the two coastal river populations in northeastern Australia, the Embley River (ER) and the Johnstone River (JR) of Queensland. In addition to B. gymnorrhiza and B. sexangula, two other congeneric species, Bruguiera cylindrica (L.) Bl. and Bruguiera parviflora (Roxb.) Wight & Arn. ex Griff., coexist in the Embley River mangrove forest. Samples of all four Bruguiera species and putative hybrids were taken from this site for comparative analysis to determine parentage involved in hybridization.

DNA extraction and ISSR amplification
Fresh young leaves were taken and stored individually at 4uC or dried and preserved with silica gel in plastic bags until DNA extraction. All dry leaf samples were kept in an electronic auto-dry cabinet (WEIFO, Taiwan). Total genomic DNA of each individual sample was extracted using a modified method of [19] for fresh leaves, and DNeasy Plant Mini Kit (QIAGEN) was used for silica gel-dried leaves to improve DNA yield and quality.
A large number of ISSR primers of Set No. 9 (Biotechnology Laboratory, University of British Columbia) were initially tested for PCR amplification. Twelve of the tested primers were selected for use based on the repeatability of banding patterns ( Table 2). PCR amplifications were carried out in a 20-mL volume containing 20 ng of DNA template, 2.0 mL of 106 reaction buffer (10 mM Tris-HCl, pH9.0, 50 mM KCl, 0.1% TritonX100), 2.5 mM of MgCl2, 1.5 U of Taq DNA polymerase, 0.2 mM of dNTP and 0.2 mM of each single primer. Amplification was performed in an MJ Researcher PTC-200 TM programmable thermal controller under the following conditions: heat denaturation at 94uC for 5 min followed by 35 cycles of 30 sec at 94uC, 45 sec at 49uC, 90 sec at 72uC, and a final 7 min extension at 72uC. The amplified fragments were separated by electrophoresis on 2% agarose gels. The gels were stained with ethidium bromide and visualized under UV light and recorded with the aid of a gel documentation system (Gel Doc 1000&2000, Bio-Rad). A 3-kb DNA ladder (MBI Fermentas) was used as a molecular weight marker for comparing amplified fragment size across gels.

ISSR data analysis
The amplified ISSR fragments were recorded as presence (1) or absence (0) for each individual. The Ewens-Watterson test for neutrality [20] was performed to examine whether all the ISSR markers used in this study are selectively neutral. Principal coordinate (PCO) analysis was performed on the binary ISSR data matrices using MVSP version 3.13p [21] for each of the geographical locations. Among a variety of different measures of distance or similarity that can be used for PCO analysis, the mean character difference was found to be comparable to several other genetic distance measures for binary data matrices and thus selected for use. The mean character difference distances were measured between the samples directly, and eigen analysis of the distance matrix resulted in direct ordination of the samples. The results were displayed as a two-dimensional scatter plot for visualization of genetic relatedness among individuals at each location.
To test whether Bruguiera 6 rhynchopetala is genetically differentiated from parental species to deserve a separate taxonomic status, we examined the clustering pattern among hybrid and other sympatric species, using all 112 individuals in Neighbor-Joining (NJ) analyses. The Jaccard distance matrix, which considers shared presence but not shared absence of ISSR bands as similarity [22], was used to reconstruct the NJ tree in PAUP* 4.0 b [23]. Bootstrap support values were obtained based on 1000 replications.
To evaluate the status of hybrid individuals and examine if there is backcrossing with either parent species or intercrossing among hybrid individuals (i.e., the production of F 2 or later generation), the Bayesian method implemented in NEWHYBRIDS 1.1. [24] was employed. The six genotype classes investigated were: pure parent A, pure parent B, F 1 progeny (50% of the genome originated from parent A and 50% from parent B), F 2 progeny (50% originated from F 1 s and 25% from each of the parents A and B), backcrosses with parent A (50% originated from F 1 s and 50% from parent A), and backcrosses with parent B (50% originated from F 1 s and 50% from parent B; for detail see [25]). Analyses were performed separately for each of the study locations. Each analysis was run independently for three times, starting with a different random number of seeds and for 10 5 iterations of MCMC chains after 10 4 burn-in steps, without using any prior information on individual or allele frequency. The affinity of an individual to the respective genotype classes is assessed by posterior probability values.

Sequence markers
In addition to ISSR, sequences of the nuclear ribosomal internal transcribed spacer (ITS) and two chloroplast intergenic regions (trnG-trnS, trnH-rpl2; Table 2) were used for a subset of samples with the aim to examine biparental and maternal relationships among taxa. Results of such were used to infer the direction and extent of introgression. Amplification was performed under the following conditions: heat denaturation at 94uC for 5 min followed by 35 cycles of 30 sec at 94uC, 60 sec at 54 or 55uC, 90 sec at 72uC, and a final 7 min extension at 72uC. The purified amplification products were sequenced directly on an ABI 3100 (Applied Biosystems) automated DNA sequencer with the BigDye terminator cycle sequencing kits. All sequences were deposited in GenBank with the accession numbers presented in Table S1.
All sequences were aligned with ClustalX [26] and manually adjusted with the Sequence Alignment Editor version 1.d1 [27]. For the ITS data, phylogenetic trees were constructed using the maximum likelihood (ML) criterion in RAxML version 7.0.4 [28] and the Bayesian criterion in Mr. Bayes version 3.0b4 [29]. For Table 2. Sequences of primers used for PCR amplification of inter-simple sequence repeat (ISSR), ribosomal internal transcribed spacer (ITS) and chloroplast DNA markers. ML analyses all searches were heuristic with TBR branch swapping. The nucleotide substitution model was first determined by the Akaike Information Criterion (AIC) method with Modeltest version 3.06 [30]. The best-fitting model (GTR+G) and related parameters of the dataset were then used in the ML searches. Bootstrap support (BS) was assessed with 1,000 replicates with the rapid bootstrap algorithm implemented in RAxML [31]. For Bayesian analyses, four Markov chains each initiated with a random tree and with two independent runs each run for 10,000,000 generations, sampling every 100 th generation were conducted. Likelihood values were monitored for stationarity with Tracer v1.4.1 [32]. Trees and other sampling points prior to the burn-in cut-off were discarded and the remaining trees were imported into PAUP* v4.0b10 [23] to generate a majority-rule consensus. Posterior probability values [PP; 33] were used to evaluate support of all nodes in the Bayesian trees. Because the two chloroplast intergenic regions are linked on a haploid genome, sequences were combined and treated as a single marker for analyses. Because gaps were found to be phylogenetically informative among our studied taxa, they were coded as multistate characters with SeqState version 1.32 [34] and appended to the sequence matrices prior to the analyses. The statistical parsimony method of Templeton et al. (1992) [35] implemented in TCS v1.13 [36] was used to construct a haplotype network with the chloroplast data. Compared to phylogenetic trees, this approach appears to be more useful in resolving reticulate relationships [37]. Haplotypes were estimated based on the uncorrected p-distances above which the parsimony principle is violated with more than 5% probability. All connections were iteratively joined among haplotypes only when the parsimony has a probability of at least 0.95 of being true as determined by coalescence theory, starting with the shortest distance until all haplotypes are joined or the distance exceeds the parsimony limit [36]. Given chloroplast genome is predominately maternally inherited in flowering plants, the proportion of haplotype sharing between hybrids and parental species reflects the direction of hybridization.

Parental origin and genetic relatedness
A total of 284 ISSR marker loci were recorded which represent all fragments amplified with the 12 ISSR primers for 112 individuals belonging to five different taxa (B. gymnorrhiza, B. sexangula, B. cylindrica, B. parviflora and the hybrid B. 6 rhynchopetala) from four geographically isolated populations. The number of detected fragments differed according to taxa and geographical locations (Table 1 and 3). Genetic differentiation among populations due to geographical isolation resulted in a large number of population-specific bands or alleles within each taxon. Among all the examined taxa, B. 6 rhynchopetala was found to contain the highest number of bands despite its relatively small sample size. Over 90% of the bands detected in B. gymnorrhiza were present in B. 6 rhynchopetala, except for the Johnstone River population where only one hybrid individual was detected based on morphological criteria. This percentage of band sharing was followed by B. sexangula, B. cylindrica, and B. parviflora in descending order (Table 3), though the latter two taxa were found only in the Embley River population. Compared to B. gymnorrhiza, fewer bands detected in B. sexangula were present in B. 6 rhynchopetala (ranging from 56.9-83.9%; Table 3). Only about 50% of the bands detected in B. parviflora and B. cylindrica (Embley River, Australia) were present in B. 6 rhynchopetala (Table 3). When only taxon-specific bands were considered, majority of the bands unique to B. gymnorrhiza or B. sexangula were observed in B. 6 rhynchopetala, whereas B. parviflora or B. cylindrica bands were rarely present in B. 6 rhynchopetala.
The four Bruguiera species, B. gymnorrhiza, B. sexangula, B. parviflora, and B. cylindrica, were clearly separated along the first two axes in the scatter plots (Figure 1), consistent with their  Figure 1A). On the other hand, hybrids sampled from Hainan Island were more clustered with one another and were all positioned more or less intermediate between B. gymnorrhiza and B. sexangula along axis 1 ( Figure 1B). In contrast, the three hybrid individuals from Indonesia were all positioned closely to B. gymnorrhiza ( Figure 1C). The proportion of band sharing (Table 3) and genetic relatedness revealed in the scatter plots ( Figure 1) support the morphological hypothesis that B. gymnorrhiza and B. sexangula are the parental species of B. 6 rhynchopetala. The varying degree of relatedness between the hybrids and the two parental species suggests that introgression occurs within populations and that such introgression is mostly unidirectional, i.e., the hybrid may preferentially if not exclusively backcrosses to one parent only.

Direction of hybridization and introgression
A total of nine haplotypes labelled as A-I (Table 4) were detected based on the chloroplast data and they were distinguished from each other by 1-6 mutations in the haplotype network ( Figure 2). Five haplotypes (A, B, C, F, and I) belonged to B. gymnorrhiza and four of these haplotypes were shared with B. 6 rhynchopetala. In comparison, only one out of the three haplotypes of B. sexangula (E, G, and H) was found in B. 6 rhynchopetala. Haplotype D appears to be unique to B. 6 rhynchopetala which is not found in either of the parental taxa. This hybrid-specific haplotype could be a product of intercrossing among hybrids and fast evolutionary rate at a microsatellite site, or due to insufficient sampling of parental taxa. The chloroplast data indicate motherhybrid relationships of B. 6 rhynchopetala with both B. gymnorrhiza and B. sexangula. However, the asymmetrical pattern of haplotype sharing suggests a predominant maternal role of B. gymnorrhiza during hybridization.
Apart from the chloroplast data, nuclear data also reveal a similar pattern of genetic association between B. 6 rhynchopetala and B. gymnorrhiza. All hybrid individuals were found to be nested in the clade that contains exclusively B. gymnorrhiza in both the ITS and ISSR trees (namely clade BG; Figure 3). Surprisingly, none of the morphological hybrids were found to be closely related to B. sexangula. The clade BG is shown to be sister to B. sexangula and B. cylindrica regardless of geographical localities. Among all, B. parviflora is clearly genetically distant from the rest of the taxa.
These relationships appear to be a result of unidirectional backcrossing of hybrids to B. gymnorrhiza, which leads to significant gene introgression.

Hybrid status
NewHybrids analyses indicated that individuals of B. 6 rhynchopetala are a mix of primarily F 1 s and backcross progeny with B. gymnorrhiza, with rare presence of F 2 s and backcross progeny with B. sexangula (Figure 4). For example, hybrids from Hainan are mainly F 1 s, but B. 6 rhynchopetala from Australia (sites ER and JR) contains a mixture of backcrosses and likely F 2 s. While B. 6 rhynchopetala from the Indonesia population are considered as hybrids based on morphological features, NewHybrids analyses did not support this interpretation but indicated that these individuals may belong to B. gymnorrhiza (Figure 4). However, reamplifications with ISSR primer 818 confirmed the presence of B. sexangula-specific bands in the Indonesian hybrids. Given only six genotype classes were specified in the NewHybrids analyses, it is likely that these hybrids represent progeny after several generations of backcrossing (i.e., introgressants of advanced generations) that could no longer be detected as hybrids based on the molecular data.

Discussion
Recent advances in molecular technology have offered unprecedented opportunities for fine analysis of natural hybridization at genetic and genomic scale in plants [10,12]. Of the various genetic approaches to studying hybridization in plants, inter-simple sequence repeat (ISSR) is one of the simplest molecular methods that can be used to generate a large number of molecular markers across the genome for comparative analysis of the putative hybrids and their parental species on a broad geographical scale, as demonstrated in the present study. In addition, chloroplast and nuclear sequence markers provide an alternative means in elucidating the direction of hybridization and introgression at the species level. A combination of these molecular methods shed light on the origin and evolution of the mangrove hybrid B. 6 rhynchopetala, which was previously unclear based on morphological characters alone.

Detection of hybridization
Many studies that use molecular markers to test hybridization hypotheses rely on the expectation that hybrids show additive marker profiles; and either all, or nearly all, parent-diagnostic markers should be found in a hybrid [11,38]. For example, the hybrid status of Helianthus paradoxus is supported by its possession of   Table 4  The number of individuals included from each site and locality label can be found in Table 1.
GenBank accession numbers are provided in Table S1. Statistical parsimony network of these haplotypes is shown in Figure 2. a BS1090JR is likely a hybrid, which has the same haplotype E as some of the individuals of B. sexangula, but is grouped with B. gymnorrhiza and B. 6 rhynchopetala based on nuclear markers (see Figure 3); b BG420ER is also likely a hybrid which has the same haplotype E as some of the individuals of B. sexangula, but is grouped with individuals of B. gymnorrhiza and B. 6 rhynchopetala from the same site based on nuclear markers. doi:10.1371/journal.pone.0019671.t004  a combination of allozyme and rDNA alleles found in H. annuus and H. petiolaris [39]. The extent of polymorphism in the parental species also affects the expectation of marker additivity [40]. Complete or nearly complete additivity in hybrids is expected only if heterozygosity and polymorphism are low within parental species and/or hybrid segregation or recombination is negligible (e.g., in F 1 generation) [38,41].
In the present study, some of the ISSR primers revealed apparent band additivity in B. 6 rhynchopetala from Hainan Island (e.g., Figure S1), but the band-sharing statistics computed over all primers are more informative on genomic additivity of the hybrid in other populations. Based on the combined ISSR data of all marker loci, the hypothetical hybridization between B. gymnorrhiza and B. sexangula can be confirmed. Interspecific polymorphism at marker loci provided a large number of species-specific bands for differentiating sympatric Bruguiera species. Most of the B. gymnorrhizaand B. sexangula-specific bands were present in the hybrid genomic profile, supporting the hypothesis that they were the parental species involved in the hybridization events. Compared to ITS and cpDNA sequencing analyses, the scatter plots from PCO analysis can most effectively separate B. 6 rhynchopetala from its parental species.
The possible involvement of sympatric B. cylindrica or B. parviflora in hybridization can be eliminated as their species-specific bands were much rarer in the hybrids. However, it is noteworthy that not all ISSR bands as well as chloroplast haplotypes present in B. 6 rhynchopetala were found in their respective parental populations (Table 3 and 4). The few unique genetic features of B. 6 rhynchopetala could be due to polymorphism within each parental population (the percentage of polymorphic loci ranging from 37.50-50.32% in B. gymnorrhiza and 16.51-24.62% in B. sexangula). Relative to the large number of ISSR loci surveyed, the samples included in this study could not possibly contain all existing alleles within the natural population of the respective parental species. Furthermore, given that some hybrid individuals could be F 2 or introgressants of advanced generations (Figure 4), these B. 6 rhynchopetala-specific ISSR fragments and chloroplast haplotype could be a result of recurrent intercrossing or new mutations in the hybrids. Another explanation for those missing parental bands in B. 6 rhynchopetala could be due to the dominant nature of ISSR markers. When hybrids are screened for the presence or absence of a parental marker, it may not be found if the parental genotype was heterozygous for the dominant marker [38]. While codominant markers such as microsatellites are known to offer major advantages over dominant markers for revealing Mendelian genotypes and in discriminating hybrids in some cases [e.g., [42][43][44], the usage of microsatellites appears to be limited to only closely-related hybridizing species. Primers designed in one species may not be widely applicable across the genus when determining parental origins of a hybrid taxon.

Evidence of introgression
The variation in scatterness among hybrids in the PCO plots ( Figure 1) and results of the NewHybrids analyses ( Figure 4) clearly indicate that B. 6 rhynchopetala consists of F 1 s, F 2 s and introgrexssants. Although the hybrid samples from Hainan population are all well separated from the parental species and most of them are F 1 s (Figure 1, S1 and 4), F 2 s or progeny of backcrosses with B. gymnorrhiza apparently exist in this population (Figure 4), resulting in their closer genetic affinity to B. gymnorrhiza than to B. sexangula in the nuclear data (Figure 3). Similarly, two of the five morphological hybrids in the two Australian populations are likely F 2 s and the rest are apparently progeny of backcrosses with B. gymnorrhiza (Figure 4). In the case of the Indonesian population, although the three hybrid samples exhibit intermediate morphological characters between B. gymnorrhiza and B. sexangula, they are genetically much closer to B. gymnorrhiza than to B. sexangula (Figure 1 and 3). These hybrids could belong to more advanced generations of introgressants, which cannot be accurately detected in the NewHybrids analysis ( Figure 4).
If B. 6 rhynchopetala consisted of only F 1 s, the species-specific bands from B. gymnorrhiza and B. sexangula should have about equal presence in the hybrid genome. However, a pattern emerges that more B. gymnorrhiza-specific bands were present in B. 6 rhynchopetala, exceeding the percentage of B. sexangula-specific bands by 26-76% in our studied populations regardless of sample size (Table 3). This pattern of nonsymmetrical band-sharing provides strong evidence for a unidirectional introgression between the hybrid and B. gymnorrhiza in at least three of the four mangrove forests. A higher level of pollen dispersal from B. gymnorrhiza or a larger B. gymnorrhiza population size compared to B. sexangula could result in unidirectional introgression. Our chloroplast DNA data suggest that both B. gymnorrhiza and B. sexangula can serve as the pollen recipient (i.e., maternal parent) when the two species coexist and hybridize ( Figure 2). Although a field survey showed that the number of mature individuals of B. sexangula is about twice that of B. gymnorrhiza in the hybrid zone in Hainan, the majority of B. 6 rhynchopetala samples had the B. sexangula chloroplast genotype [45], indicating that B. gymnorrhiza acted primarily as a pollen donor in the hybridization and introgression events. In addition to asymmetrical pollination from parental species, other pre-or post-mating isolation mechanisms may exist between the hybrids and B. sexangula, which could effectively prevent backcrosses to B. sexangula in some of the populations. However, the direction of introgression may vary among geographical locations. For example, a sample from site JR of Australia was morphologically classified as B. sexangula but genetically identified as most likely an introgressant (Figure 4).

Mechanisms of reproductive isolation
Natural hybridization plays an important role in plant speciation and evolution. Knowledge of the extent of hybridization and introgression is relevant in predicting evolutionary fate of the hybrids as well as the parental species. In theory, introgressive hybridization can prevent genetic differentiation necessary for hybrid speciation, and at the same time, result in continuous gene flow from one species into the other and hence blur the previously established species boundaries. One question arises from this evolutionary dynamic -how could sympatric Bruguiera taxa remain as distinct species if hybridization and introgression occur in all geographical regions? Different flower sizes and pollination mechanisms have been reported for different Bruguiera species. The two large-flowered B. gymnorrhiza and B. sexangula are pollinated mostly by birds, and the two small-flowered species B. cylindrica and B. parviflora are pollinated mainly by insects [1]. These major differences can lead to prezygotic reproductive isolation between the two groups. Within the small-flowered species group, no intermediate morphs were found between sympatric B. cylindrica and B. parviflora. Within the large-flowered species group, similarities in floral morphology, phenology, and pollination mechanisms all facilitate hybridization. Lack of pre-or post-mating isolation mechanisms and potentially a high degree of genome compatibility between sympatric B. gymnorrhiza and B. sexangula have resulted in fertile hybrids in sites of sympatry. However, the rarity of hybrids in some of the sympatric populations, such as in Indonesia and Australia, implies that hybridization between the two large-flowered species may not be as frequent as their extensive sympatry suggests. This could be due to a shorter overlapping flowering period between the two species in these locations. The timing and duration of flowering are affected by climate and other ecological conditions which vary according to geographical locations. In northeastern Australia, B. gymnorrhiza and B. sexangula overlap in flowering time only in August and the hybrid flowers in August-September [46]. In contrast, favorable environmental conditions in Hainan permit nearly year-long flowering in both species, and hence provide more opportunities for hybridization. Moreover, the hybrids in Hainan also have a relatively long flowering time from March-June, providing ample time for introgression. Although interspecific gene flow through frequent hybridization could potentially lead to morphological convergence between B. gymnorrhiza and B. sexangula, at least in Hainan, the two species can still be separated genetically in accordance with their taxonomic identifications. This indicates that introgressive hybridization in Bruguiera has not yet resulted in convergent evolution.
Several intermediate forms of putative hybrid origin also exist in a closely related mangrove genus, Rhizophora, including Rhizophora lamarckii Montr. or Rhizophora 6 lamarckii [1], Rhizophora 6 annamalayana [47] and Rhizophora 6 selala [48]. Recent molecular studies have confirmed that R. 6 lamarckii is a hybrid between Rhizophora apiculata and Rhizophora stylosa; R. 6 annamalayana is a hybrid between R. apiculata and Rhizophora mucronata; and R. selala is a hybrid between R. stylosa and Rhizophora samoensis in the Indo-West Pacific region [49]. Despite frequent hybridization in sympatric sites, reproductive isolation between all the parental species of Rhizophora is ensured by F 1 hybrid sterility. However, this is not the case in Bruguiera, as no hybrid sterility or reduced fertility is observed for B. 6 rhynchopetala. As shown in this study, F 2 s and introgressants exist within the hybrid populations, though the frequency of hybridization and extent of introgression apparently vary among geographical locations differing in climatic and ecological conditions.

The effects of sample size
As mentioned earlier, the difference in sample sizes among sites in the present study is primarily due to difference in the frequency of hybrids that could be morphologically identified at the time of field collection. The extensive overlapping in flowering phenology in Hainan apparently facilitates hybridization between B. gymnorrhiza and B. sexangula, resulting in a high frequency of F 1 hybrids and thus a discernible hybrid zone. In contrast, F 1 hybrids are rare in the Indonesian and Australian populations and the introgressants are difficult to detect in the field. Consequently, natural hybridization between Bruguiera species has often gone undetected in most of Indo-West Pacific region where environmental conditions and flowering phenology significantly differ from those existing in Hainan. Thus our results provide the most needed genetic evidence showing that introgressive hybridization between B. gymnorrhiza and B. sexangula actually occurs over a wide range of geographical locations.
To further examine whether the difference in sample sizes has any significant impact on the results of this study, we compared the two Australian sites, ER and JR. Despite their large differences in the sample sizes of parents and morphologically identifiable hybrids, the same results and conclusions can be reached based on our genetic analyses. Even though the sample size may affect the total number of ISSR fragments detected for each taxon within each population, it does not affect the detectability of hybrids in this study. For example, only one hybrid (field identification number 1065) from site JR could be ascertained based on morphological criteria during field sampling and our genetic analysis recognized it to be an F 2 . On the other hand, sample ID 1090 from the same site, which was uncertainly assigned to B. sexangula based on its morphology, is likely to be a hybrid as indicated by its conflicting affiliations with the two parental species between nuclear and cpDNA analyses. Similarly, all morphologically identifiable hybrids from site ER were confirmed by our genetic analyses. In addition, two other samples assigned as B. gymnorrhiza in the field are shown to be introgressants using the combined nuclear and cpDNA analyses. Because some backcross hybrids, especially those introgressants of advanced generations, resemble either parent in the diagnostic morphological characteristics, they could not be recognized as hybrids during field collection. Therefore more introgressants might be detected genetically by increasing sample sizes of the putative parents. However, the results will only lend further support to our conclusions in the present study.

Conclusions
In the Indo-West Pacific region, Bruguiera 6 rhynchopetala was shown to comprise F 1 s, F 2 s and different generations of introgressants. The extent of hybridization and introgression varies among Bruguiera populations according to their geographical locations. Among various analyses, the PCO scatter plots and NewHybrids analysis based on ISSR data can most effectively distinguish the hybrid from its parental species. However, genetic affinities shown in both ISSR and ITS phylogenies indicate that B. 6rhynchopetala has not sufficiently differentiated from B. gymnorrhiza to deserve a distinct species status. In addition, ISSR data provide strong support for multiple independent origins of B. 6 rhynchopetala, as the hybrid individuals from different geographical locations form separate genetic clusters. These hybrids occur only within the parental habitats, and there is no observable ecological differentiation from the parents other than a shorter flowering period. Thus, we conclude that lack of reproductive isolation between B. 6 rhynchopetala and its parental species has resulted in introgression, and the persistence of B. 6 rhynchopetala can be accounted for by recurrent hybridization between sympatric B. gymnorrhiza and B. sexangula in the Indo-West Pacific region. Figure S1 ISSR fragments amplified by one of the 12 primers (UBC Primer No. 818; see Table 2), showing band-sharing between individuals of Bruguiera 6 rhynchopetala (lanes 12-21) and sympatric B. gymnorrhiza (lanes 2-11) and B. sexangula (lanes [22][23][24][25][26][27][28][29][30] in Hainan. Interspecific polymorphism at marker loci provided species-specific bands for differentiating sympatric Bruguiera taxa. Most of the B. gymnorrhizaand B. sexangula-specific bands were present in the hybrid genomic profile. M: DNA ladder (lane 1) was used as a molecular weight marker for comparing amplified fragment size across gels. Arrows mark the first sample of each taxon. (EPS)