Extensive Hybridization and Introgression between Melastoma candidum and M. sanguineum

Natural hybridization can lead to various evolutionary outcomes in plants, including hybrid speciation and interspecific gene transfer. It can also cause taxonomic problems, especially in plant genera containing multiple species. In this study, the hybrid status of Melastoma affine, the most widespread taxon in this genus, and introgression between its putative parental species, M. candidum and M. sanguineum, were assessed on two sites, Hainan and Guangdong, using 13 SSR markers and sequences of a chloroplast intergenic spacer. Bayesian-based STRUCTURE analysis detected two most likely distinct clusters for the three taxa, and 76.0% and 73.9% of the morphologically identified individuals of M. candidum and M. sanguineum were correctly assigned, respectively. 74.5% of the M. affine individuals had a membership coefficient to either parental species between 0.1 and 0.9, suggesting admixture between M. candidum and M. sanguineum. Furthermore, NewHybrids analysis suggested that most individuals of M. affine were F2 hybrids or backcross hybrids to M. candidum, and that there was extensive introgression between M. candidum and M. sanguineum. These SSR data thus provides convincing evidence for hybrid origin of M. affine and extensive introgression between M. candidum and M. sanguineum. Chloroplast DNA results were consistent with this conclusion. Much higher hybrid frequency on the more disturbed Guangdong site suggests that human disturbance might offer suitable habitats for the survival of hybrids, a hypothesis that is in need of further testing.


Introduction
The roles of hybridization in plant evolution have been well recognized in the past few decades (reviewed in [1]). Interspecific hybridization can lead to hybrid speciation by allowing adaptation to extreme or novel habitats [2], reinforcement of premating isolation as a response to maladaptive hybridization [3,4], and adaptive gene transfer through introgression [5]. Thus, studies of hybridization are important for understanding interspecific relationships and evolutionary processes. It is particularly interesting to study reproductive isolation in zones of hybridization and introgression, where species can maintain their identity in the face of extensive gene flow. Hybridizing species can exchange neutral or adaptive alleles, while genes underlying local adaptation or related to reproductive isolation will be impeded [6]. Differential introgression between hybridizing species, which refers to locusspecific patterns of introgression across the genomes, offers excellent opportunities to investigate the contribution of ecological processes to speciation [7][8][9].
Hybridization can also cause taxonomic problems. Interspecific hybrids can have parental, intermediate and even novel traits [10], and they may be given the status of species, subspecies or just varieties by taxonomists (e.g. [11,12]). The situation is more complex in hybrid swarms, where a continuous range of morphological traits can be observed. So keeping hybridization in mind should enable better decisions regarding taxonomic treatment.
Taxonomic problems are common in many plant genera with multiple species. One such case is the genus Melastoma L. (Melastomataceae). Melastoma is centered in Southeast Asia and extends to India, southern China, and northern Australia [13]. It was previously estimated that this genus comprises approximately 50-100 species [14,15], but only 22 species are recognized in the latest revision by Meyer [13]. Many species of Melastoma have a relatively high degree of overlap in geographic distributions and flowering periods [14], and largely shared pollinators [16][17][18][19], offering ample opportunities for hybridization. Artificial crosses have been successfully made between some species of Melastoma, and the hybrids exhibit vigorous growth (S. Dai, unpublished data). Recently, we have identified Melastoma intermedium as a natural hybrid between M. candidum and M. dodecandrum based on sequences of two nuclear genes [11]. These results indicate that reproductive isolation between some Melastoma species is not complete.
Three taxa of Melastoma, namely, M. affine, M. candidum and M. sanguineum, are very common and often sympatric in Southeast Asia and southern China. M. candidum and M. sanguineum differ markedly in indumentums of leaf and hypanthium (a cup-shaped structure which bears the sepals, petals, and stamens), which are the most important morphological traits for species delimitation in Melastoma, as mentioned by Meyer [13]. In habitat, M. candidum, a light demanding opportunist, often occurs in open fields, grasslands and roadsides. In contrast, M. sanguineum prefers shady environments, and is usually found in the edge of forest understory. M. affine, initially published as a new species [20], is the most widely distributed taxon in this genus, ranging from southern China to northern Australia [13,14]. Together with M. candidum and M. normale, it was incorporated into M. malabathricum in the English version of Flora of China [21]. As M. candidum and M. normale are in fact distinct species based on morphological, phenological and molecular data (T. Liu et al. unpublished data), the taxonomic treatment is not reasonable and we will use the species names for Melastoma in this study according to the Chinese version of Flora of China [14]. M. affine has many morphological traits intermediate between M. candidum and M. sanguineum ( Figure 1). These morphologically intermediate traits in M. affine, including trichomes in leaf, branchlet and hypanthium, leaf and bract shape, show a continuous range, with two ends resembling those of M. candidum and M. sanguineum, respectively. It is frequently found in disturbed habitats, where it is often more abundant than M. candidum and M. sanguineum. Based on its morphological intermediacy and overlapping distribution with M. candidum and M. sanguineum, we propose that M. affine might represent an interspecific hybrid between M. candidum and M. sanguineum. A continuous range of morphological intermediacy suggests that introgression might occur between the two species.
Molecular markers allow hybridization studies with greater precision than morphology-based approaches [22]. They can be used to test the hybridization hypothesis and to infer the extent of introgression. In this study, we aimed to address the following two questions: 1) Is M. affine a hybrid between M. candidum and M. sanguineum? 2) If so, is there extensive introgression between these two species? To this end, we used 13 SSR markers to genotype multiple individuals of the three Melastoma taxa in two locations, where they are sympatric. In addition, we also sequenced one chloroplast intergenic spacer in the three taxa to see the introgression pattern in the chloroplast locus.

Plant Materials
Our sampling sites were located in Baihualing Mountain, Qiongzhong, Hainan and Zhuhai Campus of Sun Yat-sen University, Zhuhai, Guangdong. At both locations, no specific permissions were required for scientific research. No endangered or protected species were involved at the two locations. There is little human disturbance in Baihualing Mountain area. Zhuhai Campus of Sun Yat-sen University was founded in 1999 and our sampling site is located at the boundary of the secondary forest and an ornamental lawn (Zoysia tenuifolia). At this site, the habitats have been severely disturbed because partial forests were removed for the construction of the lawn, and shrubs and weeds are removed every year. For short, we will use Hainan and Guangdong for the two sites hereafter. On both sites, three taxa of Melastoma, namely, M. candidum, M. sanguineum and M. affine are common and sympatric. Plants were provisionally identified according to the diagnostic morphological characteristics described in the Chinese version of Flora of China. Briefly, M. candidum has densely puberulous trichomes in the leaves, and appressed scales on the branchlets and hypanthiums, while M. sanguineum has glabrous leaves and spreading, hispid trichomes on the branchlets and hypanthiums (Fig. 1). In addition, the two species differ strikingly in leaf shape, with lanceolate leaves and five veins for M. sanguineum, and elliptic to elliptic-ovate leaves and usually seven veins for M. candidum. M. affine has many intermediate morphological traits, including trichomes in leaf, branchlet and hypanthium, and leaf and bract shape, between M. candidum and M. sanguineum ( Figure 1). For convenience, we will directly use the species names based on the morphological identification although some of them were later genetically assigned as other categories. Over 20 individuals of each taxon were sampled on both sites (Table 1). Leaves from each individual were collected and stored in plastic bags with silica gel until DNA extraction.

SSR Genotyping
Total genomic DNA was extracted from dried leaf tissues using the CTAB method [23]. We genotyped all of the individuals sampled from both locations with 14 nuclear microsatellite markers previously developed in Melastoma dodecandrum [24]. PCR conditions followed Liu et al. [24]. Forward SSR primers were labeled with fluorescent dye FAM or HEX and fragment length polymorphisms were visualized on an ABI 3730 DNA analyzer and scored using GeneMapper 3.7 (Applied Biosystems, Foster City, CA, USA). One SSR locus, in which more than two peaks were observed in some individuals of M. candidum, was excluded in the following analysis. The SSR genotyping data were available in Table S1.

SSR Data Analysis
We used the Bayesian approach implemented by STRUC-TURE [25] to assign individuals to species and identify admixed individuals from multilocus genotype data. This Bayesian clustering program identifies the K (unknown) genetic clusters of origin of the sampled individuals and assigns the individuals simultaneously to the genetic clusters by calculating the posterior probability. To determine the optimal clusters (K = 1,10) for all samples, ten independent runs for each K value were conducted with10 5 burnin and 10 6 Markov Chain Monte Carlo (MCMC) iterations using the admixture and correlated allele frequency models. The K value with the highest posterior probability was estimated using DK statistics [26]. A threshold membership coefficient (qi) value of 0.10 was adopted to distinguish between purebred individuals (0, q i ,0.10 or 0.90,q i ,1.0) and hybrids (0.10,q i ,0.90) [27,28].
We also analyzed the thirteen-locus genotype data with the program NewHybrids [29] based on statistical model-based Bayesian methods. The software considers six genotype categories: pure species A, pure species B, F1 hybrid, F2 hybrid, and the F1 backcross to pure species A or pure species B, with the results of estimated posterior probability to assign each individual to one of the six genotypic classes. We assigned each individual to the most likely (probability.0.50) NewHybrids genotypic class. Individuals for which no single class had greater than or equal to 0.50 probability were not assigned to a class.
We first analyzed the SSR data from the two sampling sites jointly. Because species of Melastoma show substantial genetic differentiation among populations from different regions (T. Liu et al. unpublished data), samples from Hainan and Guangdong were also analyzed separately with the two programs.

Sequencing of the Chloroplast trnS-trnG Region
The chloroplast trnS-trnG region was amplified using the universal primers trnS and trnG [30] for all of the sampled individuals. Amplification conditions were as follows: 1 cycle at 94uC for 4 min; 28 cycles at 94uC for 45 s, at 56uC for 45 s, at 72uC for 2 min; followed by 1 cycle at 72uC for 8 min. The purified PCR products were sequenced with the amplification primers in an ABI 3730 DNA analyzer with the BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, California, USA). All the sequences were aligned and compared in SeqMan (DNASTAR Inc., Madison, Wisconsin, USA). The sequences obtained in this study have been deposited in GenBank with accession numbers KF737398-KF737408.

SSR Profiles
We genotyped 13 SSR loci for 46, 50 and 47 individuals of M. candidum, M. sanguineum and M. affine, respectively, which were sampled from two locations, Hainan and Guangdong (Table 1)

STRUCTURE Analysis
We first analyzed the SSR data from the two sampling sites jointly. The STRUCTURE analysis for the three taxa yielded a highest delta K value for K = 2 (Table A and  Hainan -The STRUCTURE analysis for the three taxa from Hainan indicated that two genetic clusters were most likely (Table  B and  Guangdong -Again, the STRUCTURE analysis indicated that two genetic clusters were most likely (K = 2) for the three taxa on the Guangdong site (Table C and

NewHybrids Analysis
We conducted the NewHybrids analysis for the two sampling sites jointly and for each sampling site separately. The results of joint (Table S2, Figure S1) and separate (Table S3 and S4, Figure 3) analyses were slightly different. Because species of Melastoma exhibit regional differentiation, assignment with data Hainan -The NewHybrids analysis found that on the Hainan site, of all the individuals assigned, the putative parental species, M. candidum and M. sanguineum, were correctly assigned except five individuals of M. sanguineum (Figure 3, Table S3). Among the five individuals of M. sanguineum, three and one were assigned as backcross hybrids to M. sanguineum and F2 hybrid, respectively, and one was not assigned to a class.
Among the 24 individuals of M. affine, six and three were assigned to F2 hybrids and backcross hybrids to M. sanguineum, respectively, and three were not assigned because no single class had more than or equal to 0.50 probability. Interestingly, among the remaining 12 individuals, 8 and 4 were assigned as M. candidum with high (p.0.90) and modest (0.54,p,0.90) probability, respectively. None of the sampled M. affine individuals were assigned to F1 hybrids.
Guangdong -The NewHybrids analysis found that most individuals of the putative parental species, M. candidum (18/22) and M. sanguineum (19/27), were correctly assigned on the Guangdong site ( Figure 3, Table S4). Eight individuals of M. sanguineum and three individuals of M. candidum were assigned to F2 hybrids. One individual of M. candidum was not assigned to a class.
For M. affine, most individuals (14/23) were assigned to F2 hybrids. Three and one individuals were assigned to M. candidum and M. sanguineum, respectively. The remaining five individual was  (Table 2). We designated the two chlorotypes as C-type and G-type. For M. candidum, all individuals sampled in Guangdong and 19 of 24 individuals sampled in Hainan were C-type, while the remaining 5 individuals were Gtype. For M. sanguineum, all individuals collected in Hainan and only 3 of 27 individuals collected in Guangdong were G-type, while the remaining 24 individuals were C-type.
For M. affine, all but 2 individuals had either C-type or G-type, and the remaining 2 individuals from Hainan had a unique 5 bp insertion, which formed one mutation step from the G-type (Table 2). Specifically, 12, 10 and 2 individuals of M. affine from Hainan had C-type, G-type and the unique type, respectively. In Guangdong, 17 of 23 individuals were C-type and the remaining 6 had G-type.

Hybrid Origin of M. affine
The status of M. affine was controversial. It was initially published as a new species [20], but was recently incorporated into M. malabathricum with M. candidum and M. normale in the English version of Flora of China [21]. Based on intermediate morphology between, and overlapping geographic distribution with, M. candidum and M. sanguineum, we propose that it may represent a hybrid between M. candidum and M. sanguineum.
However, there is also an alternative hypothesis that the three taxa of Melastoma are one large panmictic species complex or morphologically variable species. With SSR data from two separate sampling sites, we can test this hypothesis directly. If this hypothesis holds, it is expected that M. candidum and M. sanguineum will cluster by sampling site rather than by species when SSR data from the two sampling sites were analyzed jointly, because more gene flow is expected for species or populations in sympatry. However, our STRUCTURE analysis showed that clustering of individuals from the two sampling sites is by species rather than by sampling site. Therefore, this hypothesis is rejected. After excluding the alternative hypothesis, we then focus on testing the hypothesis of hybrid origin of M. affine. Based on the Bayesian approach implemented by STRUCTURE, the putative parental species, M. candidum and M. sanguineum, formed two genetic clusters and many individuals of M. affine showed genetic admixture between the two clusters. These individuals were further assigned as F2 hybrids or backcross hybrids by the NewHybrids program. Thus we provide compelling evidence for hybrid origin of M. affine. It should be noted that some individuals of M. affine genetically assigned as M. candidum may result from repeated backcross to M. candidum and the number of our SSR markers might be insufficient to identify them. The chloroplast data are consistent with hybrid origin of M. affine. M. affine has identical chloroplast trnS-trnG sequences with M. candidum and M. sanguineum except that two individuals (4.3%) have a unique chloroplast haplotype on the Hainan site. This may be due to unsampled polymorphism in the parental species or new mutations occurring in M. affine.
Interestingly, none of the M. affine individuals we sampled are F1 hybrids. The phenomenon that all assigned hybrid individuals were non-F1 hybrids has also been observed in other cases [31,32]. In general, F1 formation is a rare event, but once fertile F1 hybrids are produced, formation of later-generation hybrids (F2, F3 and backcross hybrids) is much more straightforward [31,33].
Introgressive Hybridization between M. candidum and M. sanguineum M. candidum is naturally distributed in Japan, South China, and northern Vietnam, while M. sanguineum has wider distribution in South China, Burma, Thailand, Malay Peninsula, Sumatra, Borneo and Moluccas [13,14]. As geographic distribution of M. candidum and M. sanguineum shows extensive overlap, introgressive hybridization should be geographically extensive. Interestingly, in addition to occurring in the overlapping regions of parental species, M. affine even extends to Australia [13,14], where both parental species do not exist. M. affine also occurs in Taiwan, where one of the parental species, M. sanguineum, has not been found [34]. Birds can mediate the seed dispersal of M. affine to new places [35]. The successful radiation of M. affine throughout almost the whole range of Melastoma suggests that this species has a successful colonizing strategy. Hybrid vigor may help M. affine colonize and adapt to new habitats where one or both parental species fail to do so. M. affine in Australia may represent a hybrid species since it can sustain independently without parental species, further studies, however, are required to test this hypothesis.

A Good System for Studying Adaptation and Speciation
M. candidum, a light demanding opportunist, usually occurs in open fields, grasslands and roadsides. In contrast, M. sanguineum prefers shady environments, and can usually be found in the edge of forest understory. Obviously, the two species show differential adaptation to their own habitats. The trichomes in the leaves of the two species also reflect differential adaptation to light conditions. While M. sanguineum has glabrous and ceraceous leaves, which can reflect excessive sunlight back, the leaves of M. candidum are densely covered with puberulous trichomes, showing adaptation to intense sunlight. Interestingly, trichome in leaves is also a diagnostic trait for the two species.
Natural selection at loci under differential adaptation can prevent gene flow between species. Selection may maintain some degree of phenotypic species integrity despite extensive admixture of other genomic regions [36,37]. We have not observed any of the 13 SSR loci as diagnostic markers between the two species, suggesting most loci in the genomes could be exchangeable between the two species. It is likely that genomic scan for strong differentiation could identify loci linked to differential adaptation and/or reproductive isolation. With ultrahigh throughput, the next generation sequencing technologies can be used for this purpose at the genomic level. Moreover, it takes only two to three years for Melastoma species from seed germination to the onset of flowering based on our observations. The short generation time in Melastoma is advantageous for QTL mapping. Taken together, Melastoma should represent a good study system to dissect the genetic basis of adaptive traits and speciation.

The Influences of Anthropogenic Disturbance on Hybridization
With different light requirements, the two Melastoma species are ecologically isolated. Ecological differentiation between closely related species can reduce the occurrence of natural hybridization. Anthropogenic disturbance, however, can offer opportunities for hybridization and suitable habitats for the survival of hybrids [38]. The sampling location in Guangdong was destroyed for the campus construction and is subjected to yearly removal of shrubs and weeds, while there is relatively little anthropogenic disturbance in the sampling location in Hainan. Thus, anthropogenic disturbance on the Guangdong site is more serious than that on the Hainan site. Our SSR analysis reveals a hybrid frequency of 34.7% (25/72) on the Guangdong site, much higher than that on the Hainan site (18.3%; 13/71). This suggests anthropogenic disturbance could provide more suitable habitats for the survival of hybrids in Melastoma. However, samples from only two locations are assessed in our study, and more samples are required to generalize the influence of anthropogenic disturbance on hybridization between the two Melastoma species. File S1 Analysis of appropriate K value for the SSR data of three Melastoma taxa. The analyses were conducted for the combined Hainan and Guangdong data sets (A), Hainan data set (B) and Guangdong data set (C), respectively. (XLSX)