Morphological species identification

In order to identify morphospecies, sclerites from different parts of the studied colony were obtained by dissolving the tissues in 10% sodium hypochlorite, followed by careful rinsing in distilled water. Particular care was taken to examine polyp sclerites when applicable, as these play a major role in the taxonomy of the speciose genus Sinularia (e.g. McFadden et al. 2009; Ofwegen and Benayahu 2012). When necessary, sclerites were prepared for scanning electron microscopy as follows: they were carefully rinsed with double-distilled water, dried at room temperature, coated with gold, and examined with a Jeol 6480LV electron microscope operated at 10 kV. Identification of species was facilitated by comparisons with permanent sclerite preparations of type material (when available) kept in the Steinhardt Museum of Natural History, Tel Aviv University, Israel (SMNH) and the Netherlands Center for Biodiversity, Naturalis, Leiden (RMNH). The identified specimens are deposited at SMNH (collection numbers preceeded by ZMTAU) (Table 1).

Molecular species identification

Extraction of DNA from ethanol-preserved tissue samples, PCR amplification, and sequencing of the mtMutS, COI (including the adjacent intergenic region, igr1) and 28S rDNA barcoding regions followed the protocols published in McFadden et al. (2011 2014). The L-INS-i method in MAFFT (Katoh et al. 2005) was used to align sequences to reference datasets consisting of previously published sequences for the relevant genera (e.g. McFadden et al. 2006 2009 2014; Benayahu et al. 2012). Pairwise measures of genetic distance (Kimura 2-parameter) among sequences were computed using MEGA v.5 (Tamura et al. 2011). jModelTest2 (Darriba et al. 2012) was used to select appropriate models of evolution for maximum likelihood (ML) analyses that were run using GARLI 2.0 (Zwickl 2006). Analyses were run using mtMutS alone (Sinularia), mtMutS concatenated with igr1 + COI (all taxa), and 28S rDNA alone [Cladiella, Klyxum, Nephtheidae, Xeniidae (for list of genera of these families see Table 1)]. Bayesian analyses used MrBayes v. 3.2.1 (Ronquist et al. 2012) with the same evolutionary models as ML, run for 5,000,000 generations (or until standard deviation of split partitions < 0.01) with a burn-in of 25% and default Metropolis coupling parameters.

Identification of Sinularia species using pure characteristic attributes (PuCAs)

A total of 27 nominal species of Sinularia plus at least one potentially new, undescribed species were identified based on the reconciliation of morphological with molecular data (Table 1). Phylogenetic analyses of mtMutS plus igr1 + COI identified Sinularia specimens to clade and sub-clade (see McFadden et al. 2009) with high support (data not shown). Within sub-clades, however, distinct morphospecies often shared identical or nearly identical mtMutS sequences. Minimum pairwise genetic distances (Kimura 2-parameter) among morphospecies ranged from 0.5-1.1% among five species in sub-clade 4D, and from 0-1.8% among eleven species belonging to sub-clade 5C. Intraspecific genetic distances as high as 0.8% were observed in several morphospecies. As a result of this substantial overlap between intra- and interspecific genetic distances, there was no barcoding gap that could be used as a threshold value for delimiting species reliably. Morphospecies whose mtMutS sequences differed by just a single nucleotide (genetic distance ~0.1%) could, however, be distinguished unequivocally if that difference represented a pure characteristic attribute (PuCA). All 12 species belonging to clade 4 and 9 of 13 species belonging to clade 5 had simple or compound PuCAs that distinguished them from all other species (Fig. 2). For example, the presence of a C at nucleotide position 245 of mtMutS is a simple PuCA that distinguished S. lochmodes (Fig. 3A) from all other species belonging to clade 5. Within sub-clade 5C, however, individuals of four morphospecies (S. abrupta: Fig. 3B, S. acuta: Fig. 3C, S. penghuensis: Fig 3D and S. slieringsi) could not be distinguished reliably using PuCAs as they often shared identical mtMutS haplotypes. In addition, some colonies identified morphologically as S. exilis had mtMutS haplotypes that were identical to those of S. erecta (Figs. 2; 3E). For those morphospecies that lacked diagnostic PuCAs or for which there was an unresolved incongruence between morphology and molecular barcode, the barcode nonetheless narrowed the initial identification to a small subset of species within a clade, after which we relied on morphological characters for the final species identification.

Open in a separate window

Fig. 2.

Fig. 2. Pure characteristic attributes (PuCAs) diagnostic of species in Sinularia clades 4 and 5 from Dongsha Atoll. Number above each column indicates nucleotide position in 735-nt mtMutS sequence relative to the start codon (ATG = positions 1-3). Nucleotides highlighted in gray represent species-diagnostic PuCAs; nucleotides highlighted in blue represent clade-diagnostic PuCAs. N = number of specimens sequenced. IUPAC ambiguity codes (e.g. Y, R) denote intraspecific polymorphisms.

Open in a separate window

Fig. 3.

Fig. 3. Underwater photographs of Dongsha Atoll octocorals: (A) Sinularia lochmodes, (B) S. abrupta, (C) S. acuta, (D) S. penghuensis, (E) S. erecta and (F) Lobophytum rigidum.

Phylogenetic relationship between Sarcophyton and Lobophytum

The second-most speciose genera of soft corals found at Dongsha Atoll were the alcyoniids Lobophytum (7 species) and Sarcophyton (6 species). The phylogenetic relationship between these two genera is currently unclear, as previous molecular analyses have divided their species among three distinct clades, each supported by diagnostic morphological characters (McFadden et al. 2006). A combined analysis of mtMutS and COI sequences obtained for specimens from Dongsha unequivocally assigned individuals to each of these three previously designated clades (‘Sarcophyton’, ‘Lobophytum’ and ‘mixed’) (Fig. 4). However, six colonies identified as S. crassocaule and L. rigidum (Fig. 3F) belonged to a well- supported, genetically distinct fourth clade that was sister to the [‘Sarcophyton’ + ‘Lobophytum’] clade; a reference specimen of S. crassocaule from Indonesia also belonged to this fourth clade, although it was genetically distinct from the S. crassocaule in Dongsha. Four species belonging to the ‘Sarcophyton’ clade (S. cinereum: Fig. 5A, S. elegans, S. nanwanensis: Fig. 5B, S. trocheliophorum: Fig. 5C, D) were morphologically and phylogenetically distinct from one another and had unique barcode sequences that closely matched reference material of those species. Within the ‘Lobophytum’ clade, five species were identified based on morphology: L. catalai, L. crassum: Fig. 5E, L. hirsutum, L. legitimum, and L. pauciflorum: Fig. 5F. Colonies identified as L. crassum, however, belonged to four genetically distinct sub-clades, each of which also included reference material previously identified as L. crassum as well as some other species (Fig. 4). Two ‘mixed’ clade species were identified, S. ehrenbergi (Fig. 6A) and L. hsiehi (Fig. 6B). Barcode sequences for S. ehrenbergi varied slightly among individuals but most haplotypes matched existing reference material of that species; we were unable to obtain sequence data from any specimens of L. hsiehi.

Open in a separate window

Fig. 4.

Fig. 4. Maximum likelihood tree based on concatenated mtMutS and igr1 + COI sequences (1552 bp) for the soft coral genera Lobophytum and Sarcophyton. Circled letters indicate clades: M, “mixed” clade; C, “Crassocaule” clade; S, “Sarcophyton” clade; L, “Lobophytum” clade (see McFadden et al. 2006). Numbers above nodes: ML bootstrap values; below nodes: Bayesian posterior probabilities. Boldface taxon labels: reference sequences from other studies; *specimen from Penghu Archipelago (Benayahu et al. 2012). Co: ZMTAU accession number.

Open in a separate window

Fig. 5.

Fig. 5. Underwater photographs of Dongsha Atoll octocorals: (A) Sarcophyton cinereum, (B) S. nanwanensis, (C) S. trocheliophorum, (D) S. trocheliophorum, (E) Lobophytum crassum and (F) L. pauciflorum.

Open in a separate window

Fig. 6.

Fig. 6. Underwater photographs of Dongsha Atoll octocorals: (A) Sarcophyton ehrenbergi, (B) Lobophytum hsiehi, (C) Litophyton sp., (D) Sinularia ceramensis, (E) S. brassica, (F) S. flexibilis.

Character-based DNA barcodes as a tool for octocoral identification

The identification of shallow-water Indo- Pacific octocorals to the species level is extremely challenging, particularly within speciose genera such as Sinularia (e.g. Benayahu et al. 2012; McFadden et al. 2014). Morphospecies are distinguished primarily on the basis of differences in colony growth form and the distribution and shape of sclerites within the tissues (e.g. Fabricius and Alderslade 2001; McFadden et al. 2009). A high level of taxonomic expertise is required to evaluate the latter, and the degree to which such morphological characters may vary intraspecifically has been assessed for only a few Sinularia species (e.g. S. brassica: Benayahu et al. 1998; S. leptoclados: Verseveldt 1980; Ofwegen et al. 2013; S. polydactyla: Verseveldt 1980; Ofwegen et al. 2016). Moreover, most of the earlier studies were based on limited numbers of specimens, did not examine the entire range of intraspecific variation in morphological characters, and lack supporting molecular data. Although the use of molecular barcode markers facilitates the identification of taxa without the need for extensive taxonomic expertise, the single-gene markers currently available are inadequate to discriminate closely- related species within most genera (McFadden et al. 2006 2009 2011 2014). For instance, in a previous survey of octocorals in Palau, application of genetic distance thresholds using the mtMutS DNA barcode discriminated 80% of Sinularia morphospecies, while the overall concordance between identifications based independently on morphology vs. barcode sequences was 77.6% (McFadden et al. 2014). When the molecular barcode was combined with assessment of colony growth form; however, 96% of specimens were identified to the species level. In the present study we used a similar combined approach, reconciling morphological characters with barcode sequences in order to discriminate taxa and assign appropriate binomials.

When intraspecific and interspecific genetic distance values are both low and overlap extensively, as they often do in Sinularia and other octocoral genera (McFadden et al. 2011 2014), methods of species discrimination that rely on arbitrary threshold values may fail when even a small amount of intraspecific variation or sequencing error is present. Character- based barcodes, however, are more robust to such noise, as species-diagnostic nucleotide characters (pure characteristic attributes, PuCAs) can be distinguished from characters that vary intraspecifically, and only the former can be used to identify species (DeSalle et al. 2005; Rach et al. 2008; Zou et al. 2011; Bergmann et al. 2013; Paknia et al. 2015). For example, Sinularia ceramensis from Dongsha Atoll exhibited intraspecific variation at 8 of 735 nucleotide positions in the mtMutS barcode sequence, leading to intraspecific genetic distances that ranged from 0-0.7% and overlapped with some interspecific genetic distances. All individuals of S. ceramensis, however, shared diagnostic PuCAs at two nucleotide positions in mtMutS (T at position 204 and C at position 294) (Fig. 2), allowing them to be discriminated unequivocally from all other species regardless of the noise contributed by intraspecific variation. Unlike threshold-based methods; however, application of character-based barcodes requires that a sufficient number of individuals of a species be sequenced in order to distinguish species-diagnostic PuCAs from nucleotides that vary intraspecifically.

Although application of a character- based rather than a genetic distance-based barcode allowed better discrimination of species of Sinularia that differed from one another by only 1-2 diagnostic nucleotides (genetic distances < 0.3%), we nonetheless encountered distinct morphospecies that shared identical mtMutS sequences and therefore could not be distinguished using either a character- or distance-based approach. Among the species belonging to the confusing 5C (“leptoclados”) sub-clade of Sinularia, S. acuta, S. abrupta, S. penghuensis, and S. slieringsi shared identical mtMutS haplotypes, and S. wanannensis differed from them at only a single nucleotide position (Fig. 2). Further investigation of those morphospecies using multilocus next-generation RADseq methods supports the genetic distinctions among them, however, and also suggests that a nuclear gene barcode marker such as 28S rDNA—perhaps in combination with mtMutS—may better discriminate closely-related Sinularia species (unpub. data).

Regardless of whether a character- or threshold-based approach is used to discriminate taxa, the use of molecular barcodes to assign binomials relies on the establishment of a database that links sequence data to validated reference specimens (Puillandre et al. 2011 2012). A limitation to the establishment of such a database for octocorals is that sequences of type colonies along with their appropriate taxonomic descriptions (i.e. including SEM images of their entire suite of sclerites) are usually only available for recently described species. For most taxa, any reference sequences that are currently available derive from freshly collected (non-type) material, as old museum specimens often fail to yield DNA of a quality suitable for sequencing. Even if the newly collected reference material was identified by a skilled taxonomist, it has not always been compared directly to the respective type material and may not have been obtained from the type locality. Such practices introduce sources of error into the sequence database, and may result in molecular barcodes being assigned to reference specimens that have been mis-identified. These taxonomic errors are propagated further when the barcode sequence is then used to assign the incorrect binomial to additional material. Unfortunately, such errors can be difficult to detect and correct, especially when type material is not available for comparison, because it has either been lost or destroyed or its depository is not known. For most octocoral taxa, taxonomic revisions accompanied by barcoding of type material or other validated exemplars (including neotypes) will be necessary before it will be possible to use molecular barcodes alone to assign bionomials with a high degree of confidence.

Acknowledgments: This study was made possible by a grant to YB from the Taiwanese Ministry of Science (MOST) to conduct octocoral surveys in Dongsha. We thank the staff members of Dongsha Atoll National Park, Dongsha Atoll Research Station (DARS), and Biodiversity Research Center, Academia Sinica (BRCAS) for assistance during the field work. We thank V. Wexler for assistance with graphics, M. Weis for technical assistance, and A. Gonzalez, L. Mattson and M. Blevins for assistance with DNA sequencing. N. Paz is acknowledged for skillful editorial assistance. SWD, PH, NHI, AC and CSM were supported in part by the Howard Hughes Medical Institute Undergraduate Science Education Program award #52007544 to Harvey Mudd College.