Molecular and Morphological Species Boundaries in the Gorgonian Octocoral Genus Pterogorgia (Octocorallia: Gorgoniidae)

Most gorgonian octocoral species are described using diagnostic characteristics of their sclerites (microscopic skeletal components). Species in the genus Pterogorgia, however, are separated primarily by differences in their calyx and branch morphology. Specimens of a morphologically unusual Pterogorgia collected from Saba Bank in the NE Caribbean Sea were found with calyx morphology similar to P. citrina and branch morphology similar to P. guadalupensis. In order to test morphological species boundaries, and the validity of calyx and branch morphology as systematic characters, a phylogenetic analysis was undertaken utilizing partial gene fragments of three mitochondrial (mtMutS, cytochrome b, and igr4; 726bp total) and two nuclear (ITS2, 166bp; and SRP54 intron, 143bp) loci. The datasets for nuclear and mitochondrial loci contained few phylogenetically informative sites, and tree topologies did not resolve any of the morphological species as monophyletic groups. Instead, the mitochondrial loci and SRP54 each recovered two clades but were slightly incongruent, with a few individuals of P. guadalupensis represented in both clades with SRP54. A concatenated dataset of these loci grouped all P. anceps and P. guadalupensis in a clade, and P. citrina and the Pterogorgia sp. from Saba Bank in a sister clade, but with minimal variation/resolution within each clade. However, in common with other octocoral taxa, the limited genetic variation may not have been able to resolve whether branch variation represents intraspecific variation or separate species. Therefore, these results suggest that there are at least two phylogenetic lineages of Pterogorgia at the species level, and the atypical Pterogorgia sp. may represent an unusual morphotype of P. citrina, possibly endemic to Saba Bank. Branch morphology does not appear to be a reliable morphological character to differentiate Pterogorgia species (e.g., branches “flat” or “3–4 edges” in P. guadalupensis and P. anceps, respectively), and a re-evaluation of species-level characters (e.g., sclerites) is needed.


Introduction
Therefore, a morphological phylogenetic hypothesis places the Saba Pterogorgia sp. morphotype as a putative sister species to P. citrina, but convergent with P. guadalupensis with respect to branch morphology (Fig 2). Molecular markers for lower-level phylogenetic studies in octocorals have been useful (e.g., [15,25]), and attempts to utilize mitochondrial loci (e.g., mtMutS+COI+igr1) for molecular barcoding of octocoral species have yielded practical, albeit limited, results [26] despite the relative lack of mitochondrial sequence variation found in anthozoans [27] and the dearth of nuclear markers available for octocorals [28]. Therefore, despite the lack of a clear "barcoding gap" (i.e., maximum intraspecific and minimum interspecific genetic overlap), some octocoral species may be distinguished using mitochondrial genetic distances or nucleotide characterbased criteria.
In addition to the mitochondrial loci available, very few nuclear markers have been utilized for species-level diagnoses in gorgonian octocorals. The most common are the ribosomal transcribed spacer (ITS) regions, which have been shown to exhibit suitable variation for inter-and intra-generic phylogenetic analyses [29][30][31]. However, because nuclear ribosomal loci are multi-copy, sequence variants within individuals may confound the phylogenetic signal of this marker. An SRP54 intron (referred to simply as SRP54) is a presumed single copy nuclear marker that has been used for among-and within-species analyses in both octocorals and scleractinian corals [32][33][34]. Pairwise sequence divergences suggest that SRP54 may adequately resolve congeneric relationships within these groups. However, significant length variation in intron size across taxa, and the technical difficulty of PCR-amplifying SRP54 among disparate taxonomic groups, has hindered the assimilation of this marker into comprehensive phylogenetic analyses of octocorals [28].
An accurate assessment of diversity on Saba Bank is crucial for effective conservation polices [35] and the monitoring of the effects of environmental change or disturbance over time [36]. In order to better quantify the evolutionary diversity within Pterogorgia, molecular phylogenetic analyses were used to test morphological species hypotheses among individuals of the three described species of Pterogorgia and the unusual morphotype from Saba Bank. Partial gene fragments of mitochondrial mtMutS, cytochrome b, and igr4, along with two hyper-variable nuclear loci, ITS2 and SRP54, were used to determine (1) whether calyx and branch morphology are valid systematic characters in Pterogorgia, and (2) whether there is genetic evidence for the unusual Pterogorgia sp. morphotype from Saba Bank as a new (phylogenetic) species.

Sample Collection
Pterogorgia spp. samples from Saba Bank (USNM1122721-USNM1122730) were obtained by permission from the Department of Invertebrate Zoology at the Smithsonian National Museum of Natural History, Washington, DC, USA, originally collected as part of a bio diversity assessment of the area (see [24]). Samples collected from southeastern Florida (PCFK1-PCFK6; PAFK1-PAFK5; PAFL6) were deposited into the permanent collection of the Coral Reef Conservation Research Laboratory, University of Miami, USA (Table 1). Collection permits were issued by the National Park Service (Biscayne National Park) and the Florida Fish and Wildlife Conservation Commission. Sub-samples of three individuals of Pterogorgia citrina, four individuals of P. guadalupensis, one individual of P. cf. anceps, and two individuals Cladogram depicting hypothesized phylogenetic relationships among Pterogorgia species using morphological characters. P. anceps and P. guadalupensis share the synapomorphic character "calyces within a common groove". Autapomorphic characters, branches "3-4 edges in cross section" define P. anceps while branches "flat in cross section (7mm+ in width)" separate P. guadalupensis. P. citrina and the morphotype from Saba Bank share the synapomorphic character "distinct calyces". P. citrina contains the autapomorphic character of branches "flat to slightly oval in cross section", while the Saba Bank morphotype contains branches "flat in cross section (7mm+ in width)". Drawings adapted from [7,23,57]. of a putative new morphospecies of Pterogorgia, were used from Saba Bank. In order to incorporate variation within each species across its geographic range, six individuals of P. citrina and P. anceps were used from reefs in south eastern Florida. Samples of P. guadalupensis were only available for analysis from Saba Bank. In total, nine individuals of P. citrina, seven of P. anceps, five of P. guadalupensis, and two of a putative new morphospecies of Pterogorgia were utilized for phylogenetic analysis ( Table 1). All samples were collected using SCUBA or by freediving. Branch tips of~3cm were cut from all colonies and stored in 10mL vials of 95% ethanol.

Molecular Analyses-DNA Extraction, PCR, Cloning and Sequencing
Total genomic DNA was extracted from each sample using a modified organic protocol [37]. Final elution was in 100μL of TE buffer and samples were stored temporarily at -20°C and permanently at -80°C. The Polymerase Chain Reaction (PCR) was used to amplify two nuclear loci (the internal transcribed spacer 2 [ITS2] region of ribosomal DNA, and the signal recognition particle 54 [SRP54] intron), and partial gene regions of three mitochondrial loci (mtMutS; cytochrome b; and intergenic region 4 [igr4]). Primer (oligonucleotide) sequences for ITS2, 5.8S-436: 5'-AGC ATG TCT GTC TGA GTG TTG G -3' and 28S-663: 5'-GGG TAA TCT TGC CTG ATC TGA G -3' [31] amplified a 210bp product (~166bp with primers removed) with partial segments of 5.8S and 28S regions at the 3' and 5' ends, respectively. Primer sequences for SRP54 intron, SRP54F: 5'-ATG GGT GAY ATY GAA GGA CTG ATW GAT AAA GTC AA -3' and SRP54R: 5'-TTC ATG ATG TTY TGG AAT TGY TCA TAC ATG TC -3' [38] amplified a 206bp product (~143bp with primers removed) with small segments of the exon on either end of the intron. The three mitochondrial loci were amplified with two primer sets. The 5' end of mtMutS (~620bp used for analyses) was amplified using ND42599F: 5'-   PCR-amplifications used 10pmol of each primer, 200μM dNTPS, 2mM MgCl 2 , and 1μL of genomic DNA (varying final elution concentrations) using 0.6U of Taq polymerase (GoTaq, Promega, Inc.) and manufacturer provided buffers. Sterile DNA-grade water (VWR, Inc.) was added to make a final reaction volume of 25μL. Thermal cycler reaction parameters included an initial denaturing step of 94°C for 3 min, followed by 35 rounds of 94°C for 1 min, an annealing temperature of 55°C for mtMutS, cytochrome b/igr4 and ITS2, or 45°C SRP54 for 1 min, and extension at 74°C for 1 min. A final extension step of 72°C for 7 min completed the reaction. PCR products for the nuclear loci, ITS2 and SRP54, were cloned using pGEM-T Easy Vector Systems (Promega, Inc.) following manufacturer protocols. Successfully transformed and cloned bacterial colonies were used as template DNA for PCR by picking the plated bacterial colony with a pipette tip and inserting it into a PCR-ready reaction. M13 primers (M13F 5'-TGT AAA ACG ACG GCC AGT-3') and M13R 5'-CAG GAA ACA GCT ATG ACC-3'), which flanked the insertion region of the vector, were used to amplify the cloned inserts. Eight clones for each locus (ITS2 and SRP54) were sequenced for each individual. Duplicate clones found for each individual were discarded in later analyses. PCR products for mtMutS and cytochrome b/igr4 were directly sequenced. Sequencing was performed using an ABI 3130xl automated DNA sequencer. Sequence contigs were collated and edited using Geneious Pro 4.8.5 [42].

Phylogenetic Analyses
Nuclear and mitochondrial loci were analyzed both separately and jointly (as a concatenated data set). Phylogenetic relationships for each nuclear locus, ITS2 and SRP54, were first examined using separate data sets for each marker. One contained all unique sequence clones found for each individual (therefore, one individual could contain more than one sequence in a tree). The other contained a single consensus sequence for each individual. Consensus sequences were created by reducing variable positions found among clones to ambiguity codes for each individual. This allowed for concatenation with the sequences of the mitochondrial loci, which were not cloned. Because the ITS2 clones contained intra-individual variation that matched interspecific variation, and, thus, contained minimal phylogenetic information, they were excluded from the concatenation of all loci. In total, phylogenetic reconstruction was performed on six data sets (S1 Bayesian Inference (BI) was performed using MrBayes 3.1.2 [46], Maximum Parsimony analyses were completed using POY 4.1.2.1 [47], and Maximum Likelihood (ML) analyses were conducted using RAxML v.8 [48]. All Bayesian analyses consisted of two independent runs with four chains, with trees sampled every 100 th generation. The parameters of the evolutionary models used for each locus were set to match the outputs obtained from jModelTest. For among-site rate variation (+G) and proportion of invariable sites (+I), the options "gamma" and "propinv" were used. When the model required a proportion of the sites as invariable, with the remaining sites drawn from a gamma distribution (+G+I), "invgamma" was used. Concatenated alignments were partitioned into character sets for each locus, and the model of sequence evolution chosen by jModelTest applied to each partition (i.e., locus). All parameters (tratio, statefreq, shape, pinvar) were unlinked to allow each partition to have its own set of parameters, and the overall rate was allowed to differ across partitions by setting the rate prior (prset ratepr) to "variable" for all partitions. Markov chain Monte Carlo (MCMC) runs were carried out for 1 to 5 million generations for all datasets (ITS2 consensus-1 million; ITS2 unique clones-4 million; mitochondrial concatenation, SRP54 consensus and unique clones, and total concatenation-5 million). Convergence for each run was determined when the average standard deviation of split frequencies was <0.01 and the potential scale reduction factor (PSRF) was 1.00. Posterior probabilities were calculated by setting the "burn-in" to 25%, which was sufficient to eliminate trees sampled prior to the runs becoming stationary. MP using POY was applied to all datasets using the Simple Search option with static homology, which utilizes a combination of TBR and SRP branch swapping and random sequence addition with 100 repetitions. Node support was assessed by 1,000 bootstrap replicates for ML and MP. Bremer decay values were calculated using TreeRot v.3 [49].
Three outgroups were used for each dataset (except for ITS2 alignments, which used two), and consisted of species in different genera within the same family as Pterogorgia (Gorgoniidae). Gorgonia ventalina and Pseudopterogorgia acerosa were used in the mitochondrial and SRP54 datasets, and Pinnigorgia flava was used in the ITS2 dataset. A species from a separate family (Plexauridae), Eunicea flexuosa, was used in all datasets.

Mitochondrial Loci
Initially, the phylogenetic information from each mitochondrial locus, mtMutS, cytochrome b and igr4, was explored individually. Each locus contained minimal sequence variation among the three species of Pterogorgia and the morphotype from Saba Bank. The trimmed alignment for mtMutS contained 620bp with only 9 variable positions, 7 of which were parsimony-informative. The partial sequence of cytochrome b contained 66bp with 1 variable position that was also parsimony-informative. The igr4 alignment was 40bp in length and also contained 1 variable parsimony-informative site. Initial phylogenetic analyses with each individual mitochondrial locus recovered similar but weakly supported trees. Because of the limited variation among the three mitochondrial loci, a concatenated mitochondrial dataset was used.
Phylogenetic reconstruction using the concatenated mitochondrial dataset recovered a tree with two clades that were well-supported with ML and MP analyses, but were deficient with Bremer support and BI (posterior probabilities 81 and 85) (Fig 4). All of the P. anceps and P. guadalupensis individuals grouped within a clade, and all of the P. citrina and Pterogorgia sp. individuals grouped in a sister clade. There was no inter-or intra-specific sequence variation among the individuals grouped within each clade, with the exception of P. citrina (SB723). No geographic variation was found among individuals of the same species collected from eastern Florida and Saba Bank.
Nuclear Loci SRP54 intron. Clones from each individual yielded one to four unique SRP54 sequences, and their alignment contained 7 parsimony-informative and 17 variable but parsimony-uninformative characters. Phylogenetic reconstruction recovered a tree with two primary clades ( Fig 5A). None of the three morphospecies of Pterogorgia, nor the Saba morphotype, were resolved monophyletically. However, and unlike with the mitochondrial loci (Fig 4), two P. guadalupensis individuals (SB725 and SB727) grouped within a poorly-supported clade that consisted of the P. citrina and Pterogorgia sp. individuals. The remaining P. anceps and P. guadalupensis individuals grouped in a well-supported sister clade. None of the Pterogorgia spp. clustered by geographic location. SRP54 consensus sequences (5 parsimony-informative characters) for each individual similarly recovered a phylogeny with two sister clades (Fig 5B). Similar to the unique clones tree for SRP54 (Fig 5A), one of the clades contained all of the individuals of P. anceps and three of the five individuals of P. guadalupensis. However, support was low with BI. Unlike the unique clones tree, the clade that grouped the P. citrina and Pterogorgia sp. individuals together with two individuals of P. guadalupensis (SB725 and SB727) was well-supported.
ITS2. Clones from each individual yielded one to four unique ITS2 sequences, and their alignment contained 5 parsimony-informative and 22 variable but parsimony-uninformative characters. Phylogenetic reconstruction using these sequences recovered a largely unsupported tree with a single clade that contained clones from some P. anceps and P. guadalupensis individuals (S2A Fig). The remaining sequences of P. citrina, Pterogorgia sp., and P. guadalupensis were unresolved. ITS2 consensus sequences for each individual (2 parsimony-informative characters in the alignment) recovered a similarly unresolved tree with a single poorly supported clade that grouped clones from most P. anceps and P. guadalupensis individuals (S2B

Mitochondrial and Nuclear Loci
A concatenated dataset containing fragments of the mitochondrial loci mtMutS, cytochrome b, and igr4, and the nuclear locus SRP54 (14 parsimony-informative and 2 variable but parsimony-uninformative characters) recovered a tree with two well-supported sister clades (BI was relatively low for the P. citrina-Pterogorgia sp. clade-posterior probability 88) (Fig 6). Similar to the mitochondrial analyses, the P. anceps and P. guadalupensis individuals grouped together in one clade. However, unlike the other datasets, P. guadalupensis was paraphyletic within this clade. A sister clade contained all of the P. citrina and Pterogorgia sp. individuals. There was no support for the separation of P. citrina and the Pterogorgia sp. morphotype.

Discussion
The mitochondrial and nuclear markers used in this study each demonstrated minimal genetic variation at the interspecific level in Pterogorgia. For example, among the mitochondrial fragments used, cytochrome b and igr4 each contained only one parsimony-informative site, and mtMutS contained seven. However, the fragments used were relatively short in length, which may have led to a limited number of phylogenetically useful sites. Nevertheless, although interspecific variation was minimal and there was no distance-based "barcoding gap" among the Pterogorgia morphospecies, there were variable positions and/or indels (indels with SRP54 only) among each locus which consistently separated P. citrina and Pterogorgia sp. from P. anceps and P. guadalupensis, and which could serve as nucleotide character-based "barcodes" Phylogram generated by a concatenated mitochondrial data set (mtMutS + cytb + igr4) using Bayesian Inference (BI), maximum likelihood (ML), and maximum parsimony (MP). Node support for BI, ML and MP is shown from left to right, with Bremer decay values in parentheses. Taxa in bold are individuals whose placement differed between the mitochondrial and SRP54 (nuclear) phylogenies (see Fig 5).
Although the nuclear loci (ITS2 and SRP54) contained more overall variation, much of it was uninformative and consisted of intra-individual and/or unshared single nucleotide differences that may have been artificially generated by cloning [50][51]. These uninformative sites may have contributed to the lower node support observed in the P. citrina-Pterogorgia sp.-P. guadalupensis clade of the SRP54 unique clones phylogeny (Fig 5A). When these substitutions are accounted for, SRP54 did not contain any more interspecific variation than the mitochondrial loci. Therefore, although it is considered hyper-variable, SRP54 may not be useful as a reliable species-level marker among other gorgonian octocoral taxa. With ITS2, although it has been useful as a species-level marker in other gorgonian taxa (see Introduction), with Pterogorgia it was less informative than the mitochondrial loci at the interspecific level, and was not able to recover a supported phylogeny. Consequently, these data fundamentally consist of two loci, one mitochondrial (the combined mtMutS+cytochrome b+igr4) and one nuclear (SRP54).
SRP54 is considered a single copy gene [32], yet three to four unique sequences were recovered from clones from individuals of each Pterogorgia species, indicating that there may be multiple copies of this gene. However, a third or fourth SRP54 sequence variant was only recovered once among individuals (S1 Table), suggesting that Taq error might explain some of these cloned sequence variants. Sequencing more clones per individual would help determine whether SRP54 is indeed multi-copy.
The mitochondrial and SRP54 data generated slightly conflicting phylogenies. The difference was the placement of two of the five P. guadalupensis individuals (SB725 and SB727). For example, the mitochondrial loci grouped all of the P. guadalupensis and P. anceps individuals in a single clade, and all of the P. citrina and Pterogorgia sp. individuals in a sister clade (Fig 4). Node support for BI, ML and MP is shown from left to right, with Bremer decay values in parentheses. A method with node support <50 contains "-". In (a) clones are coded by a "c" (clone) followed by the number of the clone relative to the total unique clones found for that individual, and arranged by taxonomic affinity for clarity. Taxa in bold are individuals whose placement differed between the SRP54 (nuclear) and mitochondrial phylogenies (see Fig 4). However, SRP54 grouped the P. guadalupensis individuals SB725 and SB727 within a sister clade that contained all of the P. citrina and Pterogorgia sp. individuals (Fig 5B). A concatenated dataset of these markers recovered all of the P. guadalupensis individuals in a clade together with the P. anceps individuals but in an unresolved subclade (Fig 6). Nevertheless, this result was likely influenced by the relatively longer and more informative mitochondrial dataset, whose loci have been shown to be useful at the interspecific level [26]. Alternatively, the discrepancy between the nuclear and mitochondrial phylogenies may be explained by competing hypotheses based on contrasting evolutionary histories of each marker/incomplete lineage sorting [52] or hybridization/reticulate evolution [53][54]. For example, in the scleractinian coral genus Acropora, molecular analyses suggest that hybridization has resulted in several species complexes (or syngameons) [55].
Overall, these data did not support the Pterogorgia sp. from Saba Bank as a separate species, or even P. anceps and P. guadalupensis (two longstanding morphological species that are generally uncontested in the literature [7,23,56]) as distinct species. However, two well-supported clades were recovered among the Pterogorgia spp. with the mitochondrial and concatenated datasets, suggesting that there are at least two lineages within the genus at the species level. One lineage consists of the morphospecies P. anceps and P. guadalupensis, and the other contains Pterogorgia sp. and P. citrina. This phylogenetic arrangement is in general agreement with what is hypothesized based on morphology, albeit less resolved (see Fig 2). It confirms the morphological relatedness of P. anceps and P. guadalupensis, but cannot differentiate among them. For example, the character "calyces within a common groove" is supported as a synapomorphy of the P. anceps-P. guadalupensis clade. However, the separation of the two morphospecies based on branch morphology (e.g., branches "flat" or "3-4 edges" in cross section) was not supported despite conspicuous differences between them. Bayer [23], after examining material collected from the Gulf of Mexico, commented that P. guadalupensis "is perfectly distinct and worthy of recognition. I have been unable to find specimens of P. anceps that grade into it". Since then, however, unusual specimens of P. anceps have been collected from Panama and the Bahamas with aberrant morphologies (Juan Sanchez, per. comm.). In addition, the specimen of Pterogorgia cf. anceps (SB721) collected from Saba Bank appeared most like P. anceps, but had branches whose morphology approached that of P. guadalupensis.
Similarly, the morphological character "distinct calyces" was supported as a synapomorphy of the P. citrina-Pterogorgia sp. clade. However, branch morphology among P. citrina and Pterogorgia sp. (e.g., "oval" or "flat") was not supported as a species-level character. Therefore, the large (>7mm) flat branches of the novel Pterogorgia sp. may represent an unusual morphological type of P. citrina, possibly unique to Saba Bank, within a larger genetic lineage of P. citrina that encompasses various branch morphologies. In this case, the branch morphology of the novel Ptergorgia sp. is convergent with those of P. guadalupensis.
Together, these results suggest that, in common with other octocoral taxa [26,28], the variation exhibited by these molecular markers may not be sufficient to distinguish morphospecies within each of the phylogenetic clades recovered. Despite this, it is clear that branch morphology may not be reliable for differentiating Pterogorgia species. Morphological differences observed in branch morphology among members of the two recovered lineages may instead represent intraspecific colony variation. A reanalysis of sclerite morphology among the morphotypes that fall within each of the two genetic lineages of Pterogorgia is required in order to more accurately delineate Pterogorgia species, and, ultimately, reevaluate the taxonomy of the genus. This, together with more robust geographic sampling and the incorporation of additional nuclear and/or population-level markers, will help to further illuminate the molecular and morphological divide within Pterogorgia.  Fig. ITS2 phylogenies. Phylograms generated from ITS2 sequences for (a) unique clones found for each individual and (b) consensus sequences for each individual using Bayesian Inference (BI), maximum likelihood (ML), and maximum parsimony (MP). Node support for BI, ML and MP is shown from left to right. A method with node support <50 contains "-". In (a) clones are coded by a "c" (clone) followed by the number of the clone relative to the total unique clones found for that individual. (PDF) S1 Table. SRP54 clones. The number of clones found for each individual for SRP54. (PDF)