Recent origin and semi-permeable species boundaries in the scleractinian coral genus Stylophora from the Red Sea

Reticulate evolution, introgressive hybridisation, and phenotypic plasticity have been documented in scleractinian corals and have challenged our ability to interpret speciation processes. Stylophora is a key model system in coral biology and physiology, but genetic analyses have revealed that cryptic lineages concealed by morphological stasis exist in the Stylophora pistillata species complex. The Red Sea represents a hotspot for Stylophora biodiversity with six morphospecies described, two of which are regionally endemic. We investigated Stylophora species boundaries from the Red Sea and the associated Symbiodinium by sequencing seven DNA loci. Stylophora morphospecies from the Red Sea were not resolved based on mitochondrial phylogenies and showed nuclear allele sharing. Low genetic differentiation, weak isolation, and strong gene flow were found among morphospecies although no signals of genetic recombination were evident among them. Stylophora mamillata harboured Symbiodinium clade C whereas the other two Stylophora morphospecies hosted either Symbiodinium clade A or C. These evolutionary patterns suggest that either gene exchange occurs through reticulate evolution or that multiple ecomorphs of a phenotypically plastic species occur in the Red Sea. The recent origin of the lineage leading to the Red Sea Stylophora may indicate an ongoing speciation driven by environmental changes and incomplete lineage sorting.

Scientific RepoRts | 6:34612 | DOI: 10.1038/srep34612 non-sister species independently evolve the same phenotype) or morphological stasis (two sister species acquire the same phenotype from a common ancestor), and the occurrence of phenotypic plasticity (the capacity of a single genotype to produce different phenotypes in response to varying environmental conditions) 29,30 . Genetic surveys within different coral genera, such as Stylophora 14,15,17 , Acropora 16,24,25 , Pocillopora 26,27 , and Seriatopora 28 , have revealed unexpected cryptic diversity in both sympatric and allopatric populations and over relatively small (e.g., Great Barrier Reef 28 , American Samoa 26 , or Western Australia 27 ) and relatively large (e.g., Indo-Pacific Ocean 14,16,17,24 ) areas. Phenotypic plasticity is a trait often associated with corals and poses many challenges for the reliable identification of these animals 29 . Environmental factors can shape coral colonies and greatly increase intraspecific morphological variation, making the delimitation of species boundaries a difficult task 29,30 .
Zooxanthellate scleractinian corals of the genus Stylophora are widely distributed and abundant throughout the tropical and sub-tropical coral reef communities of the Indo-Pacific, from the Red Sea to French Polynesia 31 . Their branching coralla and growth forms are highly plastic in relation to environmental gradients dynamically shaping the colonial architecture 32 . Stylophora corals are relatively easy to maintain in aquaria and they can produce asexual propagules. The combination of these features has led to the use of Stylophora corals, and in particular of S. pistillata, as a key model system for scleractinian reproduction 33 , physiology 34 , phenotypic plasticity 32 , coral-dinoflagellate symbiosis 35 , and transcriptome 36 studies. Nevertheless, recent genetic surveys based on a combination of mitochondrial and nuclear loci have revealed the presence of cryptic divergence and four evolutionary distinct lineages within S. pistillata across its entire distribution range 14,17 . These data caution that general conclusions arising from comparative investigations of the "lab-rat" S. pistillata might be biased by the inclusion of different cryptic entities into experimental designs 14 . Indeed, molecular phylogenies demonstrated the existence of a single homogenous and highly-connected species across the eastern Indian Ocean and the entire Pacific Ocean (clade 1) 14,37 , and at least three distinct entities within the western Indian Ocean and the Red Sea (clades 2, 3, and 4) 14,15,17 . Although the phenotypic plasticity of S. pistillata is well documented and greatly contributed to the taxonomic confusion that characterised the genus 31,32 , three deeply divergent genetic lineages of S. pistillata (clades 1, 2, and 4) 14 showed similar skeletal morphology and a comparable range of phenotypic variation as result of morphological stasis over a period of 30-50 million years 14,17 . On the contrary, clade 3 corresponded unambiguously to S. madagascarensis and it is morphologically recognisable from the other three groups at the corallite level 15 . Moreover, differences in associated algal endosymbionts (Symbiodinium) were detected among the four lineages 14 , suggesting that different regional environments might influence the ecology of this symbiosis 38,39 .
These genetic data corroborated the traditional thought, based on morphological criteria 19,20,31 , that Stylophora displays its peak of diversity in the western and northern Indian Ocean. In particular, Stylophora corals from the seas around the Arabian peninsula show remarkable variability in colony morphology and growth form, with a total of six morphospecies assumed to live in this area, namely S. pistillata, S. subseriata, S. danae, S. kuehlmanni, S. mamillata, and S. wellsi 31 . Interestingly, S. danae and S. kuehlmanni seem to be endemics of the Gulf of Aden and the Red Sea, whereas S. mamillata and S. wellsi are restricted to the Red Sea 20,31 , suggesting that the ancestral species of the genus Stylophora originated in the Red Sea 10 . On the one hand, genetic and morphological data demonstrated that S. danae, S. kuehlmanni, and S. subseriata from the Gulf of Aden belong to clade 4 14 and likely represent ecomorphs of S. pistillata determined by variation in wave movement and light intensity 15,20 . On the other hand, S. mamillata and S. wellsi display encrusting growth forms with knobby-lobbed verrucae distinguishing them from other Stylophora species 19,20 but no genetic data from these two species are available to date. Moreover, S. mamillata grows on shaded reef slopes between 20 and 40 m depth, S. wellsi occurs in very shallow water of exposed fringing reefs with strong swell and water action 20 , and S. pistillata lives in a wide range of habitats and depths 19 . The latter species shows high phenotypic plasticity, exhibiting thick, short branches when growing in shallow, wave-exposed environments and developing slender and anatomising colonies in deeper, protected water 19,31 .
In this study, we attempted to assess whether the Red Sea represents a biodiversity hotspot for the coral genus Stylophora, integrating new sequence data from this region with previously published phylogenies 14,15,17,37 . A large collection of Stylophora samples from different localities along the Saudi Arabian Red Sea (spanning about 2,000 km coastline) was obtained, including colonies from each of the six morphospecies reported to co-occur in the region. Their phylogenetic relationships and the amount of genetic differentiation were investigated, with a particular focus on the Red Sea endemic morphospecies S. mamillata and S. wellsi 31 . We sequenced three mitochondrial and three nuclear DNA regions previously employed to detect cryptic speciation in S. pistillata 14,15,17,37 , as well as one plastid locus from the associated symbiotic dinoflagellates Symbiodinium 40 . Our specific aims were to define the genetic boundaries and isolation among S. mamillata, S. wellsi, and S. pistillata in the Red Sea and to evaluate the timing of the origin of the extant Red Sea Stylophora endemics. On the basis of the obtained data, we discuss whether S. mamillata, S. wellsi, and S. pistillata might represent multiple ecomorphs of a single phenotypically-plastic species or a species complex in the early stages of speciation. The possible roles of phenotypic plasticity, reticulate evolution, introgressive hybridisation, and incomplete lineage sorting are evaluated, and possible causes of the recent diversification of these morphospecies in the Red Sea are proposed.

Results
Phylogenetic and haplotype network analyses. We obtained sequence data from three mitochondrial regions, namely the barcoding portion of the cytochrome oxidase I gene (COI), the putative control region (CR), and an open reading frame of unknown function (ORF), along with three nuclear loci, namely the internal transcribed spacer 1 (ITS1) and 2 (ITS2) of ribosomal DNA and the heat shock protein 70 gene (HSP70). Sequences were obtained for each locus from all 103 Stylophora colonies sampled across the Red Sea with the exception of three specimens that were not genetically characterised at the ITS1 and ITS2 regions (Tables 1 and 2, Data S1). The single-locus phylogenies based on COI and CR clustered all the analysed Stylophora specimens from the Red Sea within clade 4; these further clustered with published sequences of S. pistillata from the Gulf of Aden and the Scientific RepoRts | 6:34612 | DOI: 10.1038/srep34612 Red Sea (Fig. 1). Notably, the COI phylogeny showed that Stylophora is not monophyletic due to the inclusion of Seriatopora (Fig. 1) as previously reported 14 . The main discrepancy among the three proposed mitochondrial phylogenies was related to the phylogenetic topology inferred from ORF, where clade 4 was split into two main lineages. Of these two lineages, the most divergent one was composed only of colonies from both the Red Sea and the Gulf of Aden identified as belonging to the "S. pistillata" complex without any S. mamillata and S. wellsi samples (Fig. 1C). Mitochondrial haplotype networks suggested that all specimens of S. mamillata and S. wellsi shared a single ancestral haplotype for the CR locus that was also found in most "S. pistillata" complex samples from the Red Sea, while the two former species showed separated haplotypes for the ORF region (Fig. S1). For both mitochondrial loci, sequences of "S. pistillata" complex from clade 4 presented several different haplotypes, some of which were highly divergent from the majority of the remaining haplotypes.
The phasing of heterozygous nuclear loci was consistent and multisite heterozygotes were resolved with confidence, showing P > 0.8 for all the analysed Stylophora individuals. We always detected a maximum of two predominant ITS1 and ITS2 sequences in each of the collected specimens, and therefore these two nuclear loci were considered as single-copy nuclear genes (as is HSP70) and each individual was considered either homozygous or heterozygous for these two DNA regions 37,41 . Nuclear loci were more diverse than mitochondrial ones in terms of haplotype number, haplotype diversity, and nucleotide diversity (Table 1). Although ITS1 15,17 , ITS2 17 , and HSP70 ( Fig. S2) resolved S. pistillata as a complex of four deeply divergent clades, the nuclear loci grouped the Stylophora morphospecies colonies from the Red Sea in a single lineage (clade 4) (Figs 2 and 3). The nuclear haplowebs revealed pervasive allele sharing among Stylophora morphospecies from the Red Sea included in clade 4 (Figs 2 and 3). No clear clusters or cohesive groups of individuals sharing a common allele pool were detected across Stylophora morphospecies; alleles belonging to each species were highly dispersed across the three nuclear haplowebs. A clear overlap of intra-and interspecific genetic distances was observed in each of the three nuclear loci (Table 3).
Symbiodinium clade association. High-quality Symbiodinium sequences of a ~300 bp portion of the plastid psbA gene (psbA) were obtained from a subset of Stylophora specimens (n = 64), without showing any signal of polymorphisms 40 . Based on the psbA analysis, all the sampled colonies of Stylophora from the Red Sea harboured either Symbiodinium clades A or C (Fig. 4). In detail, Symbiodinium associated to S. mamillata belonged exclusively to clade C. Stylophora wellsi and "S. pistillata" complex hosted either clades A (n = 22) or C (n = 17), and their Symbiodinium composition seemed to be unrelated to the north-south latitudinal gradient of the Red Sea ( Fig. 4).

Population genetic and recombinant analyses.
The amount of genetic isolation and differentiation among "S. pistillata" complex, S. mamillata, and S. wellsi was estimated using the hierarchical Analysis of Molecular Variance (AMOVA) analysis and pairwise F ST comparisons, assuming each species as a single population and using each of the three individual nuclear loci. The AMOVA results showed no genetic structuring and isolation among species for ITS2 and HSP70 whereas F CT was significant for ITS1 (Table 4). Indeed, only a small fraction of the genetic variation was explained by the species grouping (19.34% for ITS1, 6.29% for ITS2, and 9.12% for HSP70); most variance occurred within or among individuals (Table 4). Estimated pairwise F ST values were not significantly different from 0 in all comparisons except between the "S. pistillata" complex and S. mamillata based on HSP70 (Table 5). Collectively, the AMOVA analysis and the pairwise F ST values suggested that the three Stylophora species were genetically indistinguishable from each other and that limited genetic isolation has occurred.
An analysis of potential recombinant events was conducted to test for hybridization signals among the sequenced nuclear loci, but no evidence of recombination was observed in the three analysed nuclear loci.
Divergence time estimation. The multi-locus time-calibrated phylogeny of the genus Stylophora was inferred from the concatenated mitochondrial and nuclear datasets for a total of 5,410 bp (Fig. 5). Divergence times between the main Stylophora clades and their 95% highest posterior density (HPD) intervals were estimated using the earliest fossil record of the genus from Santonian and Oman, i.e. the appearance of Stylophora octophyllia 65.5-70 Ma 42,43 . The group containing the three Stylophora species from the Red Sea ("S. pistillata" complex, S. mamillata, and S. wellsi) was estimated to be the most recent lineage within the entire genus, originating 2.51 Ma

Discussion
In this study we evaluated the importance of the Red Sea as a biodiversity hotspot for the coral genus Stylophora, including specimens representing all six morphospecies reported to occur in sympatry in this region ( Fig. 6) 19,31 .
In contrast to expectations based on traditional taxonomy 19,31 , the molecular data reported here suggested the presence of a single highly-connected genetic unit of Stylophora in this region. These results contradicted the assumption that the Red Sea represents a biodiversity hotspot for Stylophora and supported the western Indian Ocean as centre of diversity and origin 9,10 , given the co-occurrence of at least three deeply divergent genetic entities in the latter area 14,17 . Five DNA markers (COI, CR, ITS1, ITS2, and HSP70) gave congruent results and confirmed the presence of four genetically isolated clades in Stylophora across its entire geographic distribution 14 .
However, all Stylophora corals from the Red Sea belonged to a single molecular clade together with samples from the Gulf of Aden 15 , and the Red Sea morphospecies were indistinguishable on the basis of these five variable and phylogenetically informative loci (COI, CR, ITS1, ITS2, and HSP70). Conversely, the phylogenetic topology inferred from ORF partitioned S. pistillata specimens from the Red Sea and the Gulf of Aden in two non-sister clades (Fig. 1C). The discordance among mitochondrial DNA markers has been previously discussed 15 and is potentially caused by pseudogenes in the mitochondrial genome of some corals, as similarly detected in fish 44 . This scenario may also explain the high values of haplotype and nucleotide diversity in comparison to those found in the other mitochondrial and nuclear loci (Table 1). In fact, genetic diversity among the analysed Stylophora morphospecies from the Red Sea (the interspecific genetic distances based on ITS1, ITS2, and HSP70) was among the lowest ever documented in corals, < 1% in all pairwise comparisons 45  The incongruence between morphological species delimitations and genetic species boundaries is striking and may be caused by several factors. The collected colonies of S. pistillata, S. danae, S. subseriata, and S. kuehlmanni exhibit a smooth morphological continuum with regard to the skeletal features traditionally used to identify these taxa 19,31 . For example, the branch thickness, the presence of corallite hood, and the coenosteum formation can extensively vary on a single corallum and can be shaped by environmental conditions. Considering the absence of interspecific genetic differentiation based on the six molecular loci employed in this study [14][15][16][17][24][25][26][27][28]37 , the latter three morphospecies should be considered as junior synonyms of S. pistillata, as suggested in previous studies 15,20 . The Red Sea is characterised by strong latitudinal gradients in environmental variables (e.g., temperature, salinity, and nutrients) 22 , by a great diversity of habitats and reefs 3 , and by a complex geological history 21,46 . These aspects, combined with the extreme phenotypic plasticity 29,32 and habitat generalisation 33,34 documented in S. pistillata, can possibly explain the outstanding phenotypic polymorphism of this species in the region.
Despite the lack of genetic differentiation, S. mamillata and S. wellsi show distinct colony morphologies and depth-partitioning 19,20,31 . There are at least two possible scenarios that could explain semi-permeable species boundaries among "S. pistillata" complex, S. mamillata, and S. wellsi. Stylophora mamillata and S. wellsi may be regional ecomorphs of the highly phenotypic plastic S. pistillata or they may actually be valid species. If the latter, they remain connected through genetic exchange or they may be examples of recent speciation events. The first hypothesis (i.e., S. mamillata and S. wellsi are ecomorphs of S. pistillata) is supported by the extreme phenotypic plasticity of S. pistillata in the Red Sea 29,32 . A single genotype can produce different phenotypes in response to changing environmental and selective regimes 30 , resulting in distinct morphologies influenced by depth, light, and wave action. For example, a translocated colony of P. meandrina began to grow with a morphology more similar to P. damicornis after several months 47 ; similar translocation experiments could be carried out for S. mamillata and S. wellsi. Colony morphology is known to be a misleading character in the identification of several corals 11,13,17,24,19,31 . The encrusting S. mamillata is usually characterised by nodes thought to be incipient branches, but these can sometimes grow into small clear branches. Similarly, the formation of verrucae is one of the morphological features prescribed to diagnostically identify S. wellsi among Stylophora species, but some analysed colonies of S. pistillata show this peculiar structure (Fig. 6). Under the second scenario (i.e., S. mamillata and S. wellsi are valid species), reticulate evolution through hybridisation and gene exchange or incomplete lineage sorting due to recent speciation may explain the genetic data. Although no recombinant events were detected in the obtained nuclear sequences, introgressive hybridisation and gene exchange have been extensively documented in closely related scleractinian corals. This can be promoted by extensive sympatry and  50 . These findings suggest that prezygotic isolating mechanisms in the pocilloporids are permeable and may also provide chances for introgression in Stylophora. whose morphospecies in the Red Sea may form a Stylophora syngameon (a group of species connected through genetic exchange). Moreover, it is widely accepted that hybridisation is enhanced in isolated and peripheral regions 7,51,52 , such as the Red Sea 3,21 , and novel hybrids may have advantageous reproductive abilities 48    phylogeny strengthens the hypothesis that suggested the isolation of distinct Stylophora populations as a result of the fragmentation of the Tethys Sea, which promoted a great diversification of the genus in the Indian Ocean 3 . On the basis of these results, S. mamillata, S. wellsi, and S. pistillata may represent a species complex undergoing early speciation that still shares most ancestral alleles and polymorphisms. The rapid speciation of the three recent morphospecies of Stylophora might have been promoted by the strong environmental changes encountered in the Red Sea during Pliocene and Pleistocene 21,46 , which may have favoured niche partitioning and ecological differentiation. Furthermore, the extreme phenotypic plasticity of S. pistillata might have played a crucial role in promoting speciation, creating intraspecific variations that can form the basis for interspecific diversification 30 . Furthermore, although no genetic isolation was detected among "S. pistillata" complex, S. mamillata, and S. wellsi, they might remain recognisable because disruptive selection may be occurring 53 or because hybridisation is not pervasive 5 . Indeed, the three morphospecies are ecologically differentiated, showing distinct habitat preference and depth partitioning and, in this condition, disruptive selection can contribute to the maintenance of ecological differences among species and increased phenotypic variation 49,53 .
Coevolution of the coral host and symbiont dinoflagellate Symbiodinium may play a significant role in niche specialisation, habitat partitioning, and ecological diversification of corals. It may also promote speciation events 54,55 . Previous studies on brooding corals of the genera Madracis and Agaricia across a large depth gradient  Table 3. Interspecific and intraspecific average genetic distances expressed as percent (standard deviation in brackets) for S. mamillata, S. wellsi, and "S. pistillata" complex based on ITS1, ITS2, and HSP70 sequence data. (2-60 m depth) demonstrated that their associated Symbiodinium consistently revealed patterns of host specificity and depth-based zonation 49,56 , shaping host bathymetric distribution and ecology. Stylophora corals in the Red Sea hosted either Symbiodinium clades A or C and, interestingly, all the "deep-water specialist" S. mamillata colonies harboured exclusively Symbiodinium clade C, whereas the "shallow-water specialist" S. wellsi and the generalist "S. pistillata" complex were associated with Symbiodinium clades A and C along the entire Saudi Arabian Red Sea (spanning 13° of latitude from the Gulf Aqaba to the Farasan Islands). Previous investigations of S. pistillata from the Red Sea demonstrated associations with Symbiodinium clades A, C, or A + C combinations 14 and, in particular, it shifts from hosting mainly Symbiodinium clade A in shallow waters (2-6 m depth) to Symbiodinium clade C in deeper waters (24-26 m depth) in the Gulf of Aqaba (northern Red Sea) 35 . Nevertheless, Symbiodinium   data presented in this study are not exhaustive and further detailed analyses based on ITS2 typing via next generation sequencing approaches are needed 57 to clarify if there is a specific association between S. mamillata and Symbiodinium clade C, which may suggest adaptation of this symbiosis to deep water (below 20 m depth).

Conclusions
Despite the presence of distinct colony morphologies, Stylophora corals from the Red Sea belong to a single cohesive molecular lineage and host less genetic variability compared to other regions, such as the western Indian Ocean and the Gulf of Aden, where up to three genetic clades occur in sympatry 14,15,17 . These results raise several questions concerning the evolution of the extant Styophora morphospecies in the Red Sea. Further analyses are needed in order to evaluate whether S. mamillata and S. wellsi represent either valid endemic species arising from recent speciation or whether they are simply local ecomorphs of the common "S. pistillata" complex adapted to distinct depth and light conditions. Indeed, single and multi-gene approaches may be affected by the slow evolution rate of coral mitochondrial DNA 58 and by the incomplete concerted evolution of rDNA 59 , resulting in low genetic variation levels of the analysed loci. In these cases, the application of reduced genome approaches, such as RNA-seq or RAD-tag seq, will provide a genome-wide perspective and may improve the phylogenetic resolution and species boundaries definitions, as already demonstrated in the scleractinian coral Pocillopora 44 and the octocoral Chrysogorgia 60 . A closer investigation of the reproductive modes in S. mamillata and S. wellsi may provide insights into the possible occurrence of reproductive barriers and the role of hybridisation events among these morphospecies. Finally, translocation experiments of S. mamillata and S. wellsi in different depths and environmental regimes will enhance the understanding of phenotypic plasticity and polymorphism whereas associated transcriptomic analyses might indicate which genes are involved in these mechanisms.

Methods
Coral collection and identification. A total of 106 colonies of Stylophora corals were collected along the coast of the Saudi Arabian Red Sea, between 1-40 m depth. Furthermore, four specimens of S. pistillata from Papua New Guinea (clade 1), one colony of S. pistillata from Madagascar (clade 2), and one sample of S. madagascarensis from Madagascar (clade 3) were included in the analyses. Each colony was photographed underwater and tagged (Data S1). A small portion of tissue (~2 cm 3 ) was preserved in 95% ethanol for molecular analyses while the remaining portion (~10 cm 3 of the colony) was bleached in sodium hypochlorite, rinsed with fresh water, and airdried. Morphospecies identification was achieved by examining type material and reference monographs 19,20,31 . For analyses, we considered S. pistillata, S. subseriata, S. danae, and S. kuelhmanni as part of a single lineage (indicated in the text as "S. pistillata" complex), corresponding to Stylophora morphs M and L 15 , and clade 4 14 . Indeed, these four species in the Red Sea "form a smooth continuum with regard to those skeletal structures which have been used previously to help establish them" 20 . On the contrary, S. mamillata and S. wellsi are easily morphologically distinguishable from the above four morphospecies, and were therefore treated as separate entities 20 (Fig. 6).
DNA extraction and PCR amplifications. Genomic DNA was extracted using the DNeasy Blood & Tissue kit (Qiagen, Hilden, Germany) and DNA concentration of extracts was quantified using a Nanodrop 1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA). A total of six loci were amplified and sequenced for the analysed Stylophora morphospecies 14,15,17,37 : COI, CR, and ORF from the mitochondrial genome, and ITS1, ITS2, and HSP70 gene from nuclear DNA. Symbiodinium clades of Stylophora hosts were identified using the plastid psbA minicircle (psbA) 40 . The list of primers and PCR annealing temperature is indicated in Table S1. Amplifications were performed in a 12.5 μ l PCR reaction mix containing 0.2 μ M of each primer, 1X Multiplex PCR Master Mix (Qiagen, Hilden, Germany), and < 0.1 ng DNA. PCR consisted of an initial denaturation at 95°C for 15 min, followed by 30 cycles of denaturation at 94 °C for 30 sec, annealing for 1 min, extension at 72 °C for 1 min, and a final extension at 72 °C for 10 min. All PCR products were purified with Illustra ExoStar (GE Healthcare, Buckinghamshire, UK) and directly sequenced in both directions using an ABI 3130xl Genetic Analyzer (Applied Biosystems, Carlsbad, CA, USA). All sequences generated as part of this study were deposited in the EMBL database (Data S1).
Phase determination, sequence alignment, and recombination assessment. Forward and reverse sequences were assembled and edited using Sequencher 5.3 (Gene Codes Corp., Ann Arbor, MI, USA).
Most Stylophora colonies showed double peaks and intra-individual polymorphisms from nuclear loci and were thus considered to be heterozygotes 41 . Nuclear sequences were phased using SeqPHASE 61 and Phase 62 when alleles showed the same length (n = 21 for ITS1, n = 12 for ITS2, and n = 62 for HSP70), and using Champuru 63 if the two predominant alleles were of different length (n = 40 for ITS1 and n = 68 for ITS2). In the former case, the two alleles with the highest probability (an order of magnitude greater than the other sequence pairs) were chosen whenever there were multiple possible phases. No obvious or significant differences of genetic diversity and haploweb inference were obtained using alternative phases (results not shown). Alleles of different length were detected only in the ITS1 and ITS2 regions. Phased heterozygotes were represented by both alleles in the further alignments and population genetic analyses. Alignments for each individual locus were performed using MAFFT 7.130b 64 and the iterative refinement method E-INS-i. Determination of potential recombinant events that can be interpreted as significant signals of hybridisation was carried out using RPD4 65 for each of the three nuclear loci. In particular, the algorithms RDP, GENECONV, BootScan, MaxChi, Chimaera, SiScan, LARD, and 3SEQ were investigated using the default settings in all cases.
Phylogenetic and haplotype network analyses. General statistics concerning the obtained sequences and the variability of the seven employed markers were calculated with DnaSP 5.10.1 66 , as reported in Tables 1  and 2. The best evolutionary model for each individual molecular locus was selected using jModeltTest 2.1.1 67 , as indicated in Table S1. Phylogenetic analyses were conducted under three criteria (Bayesian inference (BI), maximum likelihood (ML), and maximum parsimony (MP)) for each of the three individual mitochondrial regions (COI, CR, and ORF) and the nuclear HSP70 gene. For BI analyses, four Markov Chain Monte Carlo (MCMC) chains were run for 8 million generations in MrBayes 3.2.1 68 , saving a tree every 1,000 generations. The analyses were stopped when the deviation of split frequencies was less than 0.01 and all parameters were checked in Tracer 1.6 69 for effective sampling size and unimodal posterior distribution. The first 25% trees sampled were discarded as burn-in following indications by Tracer 1.6. ML topology was reconstructed using PhyML 3.0 70 using the Shimodaira and Hasegawa test (SH-like) to check the support of each internal branch. MP analysis was performed using PAUP 4.0b10 71 and a heuristic search strategy with tree-bisection-reconnection (TBR) branch swapping for 100 replicates and random stepwise addition. Node support was assessed throughout 1,000 bootstrap replicates. The median-joining network analysis implemented in Network 4.613 72 was applied to each of the three nuclear datasets and to the highly variable mitochondrial CR and ORF regions in order to evaluate relationships among haplotypes. In order to find groups of individuals sharing a common allele pool, nuclear haplonets were converted into haplowebs 73 by drawing additional connections between the two haplotypes co-occurring in heterozygous individuals using Network Publisher 2.0.0.1 (Fluxus Technology, Suffolk, UK).

Population genetic analyses.
A hierarchical Analysis of Molecular Variance (AMOVA) was performed using Arlequin 3.5.1.2 74 to determine the percentage of genetic variance explained by morphospecies clustering and the significance of population structure among and within the analysed Stylophora morphospecies Scientific RepoRts | 6:34612 | DOI: 10.1038/srep34612 (subdivided as "S. pistillata" complex, S. mamillata, and S. wellsi), assuming each of the three morphospecies as a population. The genetic differentiation among taxa was estimated by means of pairwise F ST with Arlequin 3.5.1.2, calculated with Slatkin's distance and using genetic distances corrected by a Kimura two-parameter evolutionary model. Significance was tested using 1,000 permutations and allowing a minimum P-value of 0.05. Intra-and interspecific genetic distances were calculated using DnaSP 5.10.1 under a Kimura two-parameter evolutionary model and variance was estimated with 1,000 bootstrap replicates.

Divergence time estimation.
In order to provide a provisional estimate of the divergence time of each of the Stylophora clades and of the Red Sea Stylophora morphospecies, a time-calibrated phylogenetic hypothesis was inferred under a Bayesian framework using BEAST 1.8.2 75 , based on the complete concatenated mitochondrial and nuclear dataset (5,410 bp). We specified the same six partitions as above with unlinked evolutionary models, an uncorrelated (lognormal) clock model, and a Yule tree prior. The analysis was run for 50 million generations, with a sampling frequency of 1,000. After checking adequate mixing and convergence of all runs with Tracer 1.6 69 , the first 20% trees were discarded as burn-in and a maximum clade credibility chronogram with mean node heights was computed using TreeAnnotator 1.8.2 75 . It is known that Stylophora occurred both in the Caribbean and the Indo-Pacific during the late Cretaceous but then it disappeared from the former basin during the early Miocene 20 . Because the genus occurs today only in the Indo-Pacific, we time-constrained the node leading to Stylophora spp in the tree based on the fossil record of S. octophyllia, which first appear in Santonian and Oman during the Maastrichtian around 65.5-70 Mya 43,44 .