Re-Examination of the Phylogenetic Relationship among Merulinidae Subclades in Non-Reefal Coral Communities of Northeastern Taiwan

: Species identiﬁcation for spawning corals relies heavily on morphology. Recent molecular phylogenetic approaches have demonstrated the limits of traditional coral taxonomy based solely on skeletal morphology. Merulinidae is considered a complex taxonomic group, containing 24 genera and 149 species. This family is one of the most taxonomically challenging and its taxonomy has largely improved in recent studies. However, studies of the phylogeny of Merulinidae are constrained by limited geographic scales. In Taiwan, merulinid corals are dominant in non-reefal communities on northeast coasts and they consistently spawn between summer and fall. This study is a ﬁrst attempt to establish a molecular database of merulinid corals in this new area, including a volcanic island (Kueishan Island), and provide information about sexual reproduction. We analyzed 65 specimens, including 9 genera and 28 species collected from Taiwan using one mitochondrial marker (COI: cytochrome c oxidase subunit 1 gene) and three nuclear markers (ITS: nuclear ribosomal internal transcribed spacer, 28S rDNA D1 and D2, and histone H3) to re-examine phylogenetic relationships and search for new species. Overall, 58 COI sequences, 59 for ITS, 63 for 28S, and 62 histone sequences were newly obtained from the collected specimens. The reconstructed molecular tree demonstrates that all the specimens and reference sequences we examined are clustered within Merulinidae. Subclades A, B, C, D/E, F, G, H, and I are congruent with previous studies. However, Astrea curta is separated from the other congeneric species, Astrea annuligera (XVII-B), which is a sister to Favites and deﬁned as a new subclade K. In addition, two new species ( Paragoniastrea deformis and Paragoniastrea australensis ) were discovered for the ﬁrst time in Taiwan, and we deﬁned them as a new subclade J. In addition, A. curta , P. auastralensis , and P. deformis are all hermaphroditic spawners and released bundles in July. This study greatly improves the accuracy of biodiversity estimates, systematic taxonomy, and reproduction for Taiwan’s coral ecosystem.


Uncovering Taxonomic Progress
The identification of scleractinian corals based solely on morphology is challenging because some scleractinian species can exhibit environment-correlated variations in morphology, i.e., Ecomorphs [1]. In addition, species display phenotypic plasticity across their distribution, making it difficult to rely on shared morphological features to identify them [2,3]. Therefore, it is important to combine morphological and molecular characteristics to improve the accuracy of the determination of evolutionary relationships.
The traditional classification of scleractinia into seven suborders was out of date [4][5][6][7]. Given the comprehensive study of the entire taxon with morphological and molecular approaches, the scleractinian corals can be generally divided into three major groups: basal, robust and complex [8]. Furthermore, they are separated into 21 clades (I-XXI) [8,9]. Many scleractinian corals at family and genus were revised or remained unclear taxonomic position (Scleractinia incertae sedis  Veron & Pichon, 1980, the last of which was redefined from the conventional Euphyllia ancora, E. yaeyamaensis, and E. divisa [11]. In addition, the genera Nemenzophyllia, Physogyra, and Plerogyra were removed from Euphylliidae because they formed a separate clade with Blastomussa (clade XIV) [9,12].

Taiwan Taxonomy and Species Diversity
Taiwan is located at the center of the Philippine-Japan Island arc at a latitude of 21.90 • N to 25.3 • N, crossing from the Tropic of Cancer close to the northern tip of the Coral Triangle [22]. To date, 317 scleractinian coral species have been reported in Taiwan and display a latitudinal gradient of decreasing species diversity from south to north [23,24]. In addition, coral assemblages contain 21 genera covering 87% of the total number of genera of merulinid corals and 89 species covering 60% of the total number of merulinid species [25]. Taxonomic phylogenetic studies of scleractinian corals collected from Taiwan are very limited [9,11,[26][27][28][29]. In addition, biogeographical integration is needed on a larger scale. For example, Polycyathus chaishanensis (Caryophyllidae) was proposed to be endemic to Taiwan [27]. Later, this species was also found to inhabit Indonesia, based on molecular evidence [30]. Euphyllia ancora has been a model species for studies on sexual reproduction [31] and its genus was recently revised to Fimbraphyllia [11]. This revision created an important foundation on the convergent and divergent functionalities of genes and compared functional genes among the cnidarians underlaying precisely the phylogenetic position of the studied species.

Purpose of This Research
The phylogeny of Merulinidae reconstructed in Huang et al. [16] was based on samples/taxa from Australia, Singapore, Japan, and the Philipines in the Pacific Ocean and the Atlantic Ocean. Taiwan is located in the Pacific Ocean; it is an important stepping stone between the Philippines and Japan. Merulinid corals are major spawning members and consistently spawn between summer and fall in non-reefal coral communities in northern Taiwan [32]. These spawning corals are important for maintaining local recruitment, providing heterogenetic materials to the local population and connecting across different populations. However, some convergent macro-morphological characteristics make it challenging to identify some genera in the field, such as Goniastrea (ceroid form), Favites (ceroid and plocoid forms), and Diploastraea (plocoid form) [24,33]. In addition, species identification for spawning corals based on morphological criteria in the fields and underwater photographs is difficult because the polyps are deformed when "the mature sperm and eggs move to the mouths of polys" (i.e., bundle setting).
As mentioned above, these challenges can be resolved by molecular approaches, as was demonstrated by Huang et al. [10]. For example, ceroid forms of Goniastrea, Diploastraea, and Favites are clearly separated in subclades A, B, and F based on phylogenetic reconstruction using multiple loci [10]. Therefore, Chen et al. [32] identified the species to the genus level of each specimen using molecular approaches and the BLAST tool [34]. Subsequently, specimens were identified to species level using the morphology of their skeletons. The established molecular database of merulinid corals can provide further insight into the phylogenetic relationships among the subclades of Merulinidae. The objectives in this present study were to: (1) establish a molecular database of spawning corals of Merulinidae from Taiwan, which have not been studied before; (2) re-examine the phylogenetic relationship between the specimens collected from northern Taiwan and the merulinid corals in previous studies, using phylogeny reconstructions based on multiple loci; and (3) record any new species or subclades we might find in this region.

Sample Collection
Chen et al. [32] demonstrated that the spawning season for merulinid corals is July to August, from 2014 to 2016, in northeast Taiwan. Merulinid corals with bundle-setting behavior and released bundles still attached outside of the mouths of polyps were collected at night by scuba diving. Some corals were collected at two offshore islands and their sexual reproductive behavior was observed using histological approaches [32]. A total of 65 specimens from four sites were chosen for this study: 26 specimens from Pitoujiiao (25 • (Figure 1). Coral fragments were collected by using chisels and hammers and separated into two parts. One was fixed in 90% ethanol for molecular analysis. The other was bleached in sodium hypochlorite until the tissue was entirely removed, rinsed in freshwater, and air-dried for the morphological analysis.

Sequence Management, Alignment, and Matrix
The raw forward and reverse sequences were edited and assembled into consensus sequences by the CodonCode Aligner V6.0.2 program (CodonCode Corporation Dedham, MA, USA). To exclude sequences amplified from zooxanthellae, the consensus sequences obtained were used to perform the BLAST searches (http://www.ncbi.nlm.nih.gov/BLAST/, accessed on 2 April 2020) [34]. The sequences obtained from the collected specimens of spawning corals in northern Taiwan were deposited into the NCBI GenBank (accession numbers in Supplementary Table S1). Newly obtained sequences for COI (n = 58), ITS (n = 59), 28S (n = 63), and histone (n = 62) were combined with sequences retrieved from public sources (Table S1).
All sequences for each gene were automatically aligned with the accurate alignment option (E-INS-i) in MAFFT v.7 ([35]; http://mafft.cbrc.jp/alighment/server/, accessed on 10 January 2021) under default parameters. The resulting multiple sequence alignments were translated into inferred amino acid sequences as a guide for inferred gap placement between coding regions using Se-Al v.2.0a11 [36]. The amino acid residue and nucleotide were manually adjusted to minimize the gaps. PAUPRat software v.3.1 [37] on the CIPRES Science Gateway (http://www.phylo.org, accessed on 10 January 2021) [38] was used to calculate descriptive statistics (sequence variations and informative sites) for the compared sequences of each gene.

Molecular Datasets
Some sequences were not obtained because the gene failed to amplify during PCR. Operational taxonomic units (OTUs) were created for each gene for the phylogeny reconstruction. The phylogeny reconstructions were conducted based on the combined gene matrix (COI, 28S, ITS, and histone); IGR (noncoding intergenic region between COI and the formylmethionine transfer RNA gene) was ignored because the overall alignment was not similar enough.
The clades distant from the merulinid corals (XVIII-XXI) were included for the phylogenetic inference following Huang et al. [10]. These sequences were comprised of three species of Lobophylliidae (Moseleya latistellata, Acanthastrea echinata, and Lobophyllia corymbosa), three species of Faviidae (Montastraea multipunctata, Favia fragum, and Mussa angulosa), and one species of Plesiastreidae (Plesiastrea versipora). Phylogeny reconstructions were created for each gene, along with four combined datasets based on maximum likelihood and Bayesian analyses.

Molecular Phylogenetic Analysis
The maximum likelihood (ML) trees of each partition were reconstructed with raxml-GUI v.2.0 [39] using the best model (GTR+I+G). The five datasets, including three nuclear genes (ITS, 28S, and histone H3), one mitochondrial gene (COI), and a combined gene dataset, were partitioned based on coding position. The combined gene datasets were conducted with five independent runs, and the tree with the best ML scores was selected as the final tree. Nodal support was assessed by bootstrapping, and only the nodes with ≥70 [40] based on 1000 pseudo-replicates were shown.
Bayesian inference (BI) was carried out in MrBayes v.3.2.6 [41]. PartitionFinder was used to select the best partition scheme and accompanying substitution model, according to the Bayesian information criterion [42]. The best-fit substitution model was determined by ProtTest3. Two Monte Carlo Markov chains (MCMCs) were run for 4 × 10 6 million generations in two simultaneous runs, each with four different chains. The convergence of the estimates was checked by the standard deviation of split frequencies and by monitoring the likelihood score over time using Tracer v.1.6 [43]. Trees were sampled every 1000 generations, with the first 2500 (25%) discarded as "burn-in." The remaining sampled trees were collected to construct a 50% majority-rule BI consensus tree. Nodal support from BI was assessed, and only nodes with ≥0.90 posterior probabilities (PPs) were shown.

Characteristics of the Gene Data
In the 65 specimens collected from Taiwan, 58 newly obtained COI sequences, 59 newly ITS sequences, 63 for 28S, and 62 for histone H3 sequences were obtained for the first time (Table S1). Examining the individual gene dataset, the aligned COI sequence was 744 base pairs (bp) long, with 180 variable and 81 parsimony informative sites. That of 28S was 865 bp, with 316 variable and 147 parsimony informative sites. That of ITS was 1249 bp, with 662 variable and 459 parsimony informative sites. That of histone H3 was 344 bp, with 109 variable and 82 parsimony informative sites.

Results of the Analysis Matrix
The dataset comprised a total of 3202 bp and 186 OTUs (123 OTUs from references and 65 OTUs from the present study). ML and BI methods (using raxmlGUI and MrBayes, respectively) were used to reconstruct the phylogenies for the combined dataset. The results from the partitioned ML analysis and BI conducted with the combined dataset were congruent ( Figure 2). The ML analysis yielded a log-likelihood value of −28,909.928401 and the BI analysis yielded (−3.069135 × 10 4 , −3.083239 × 10 4 ).

Favia fragum
Caulastraea tumida (G61875) Figure 2. Phylogenetic tree of merulinid corals and their allies based on the combined gene dataset inferred with the maximum likelihood method using the GTR+G model. Molecular subclades within Merulinidae (XVII) are defined as being A to I following Huang et al. [10]. The other two novel clades (J and K) are defined in this study. Branch lengths are proportional to inferred nucleotide substitutions. Numbers at the nodes represent bootstrap values (only ≥70 shown) from the maximum likelihood method and posterior probability (only ≥0.9 shown) from the Bayesian inference. Bold branches on the tree indicate statistically robust nodes. The spawning month of specimens in Chen et al. [32] are in brackets.

The Phylogenetic Tree
The spawning specimens we examined were all nested within Merulinidae (taxa in bold font in Figure 2). These specimens were placed in five subclades B, C, E, F, and G but not in the subclades A, D, H, or I. The following were nested in subclade B: one specimen each of Astrea annuligera, Coelastrea palauensis, Dipsastraea rotumana, Dipsastraea mathaii, and Favites valenciennesi and two specimens of Coelastrea aspera collected from Kueishan Island; one specimen of Dipsastraea favus collected from Pitoujiiao and two from Kueishan Island; one specimen of Dipsastraea lizardensis collected from Longdong. Two Cyphastrea chalcidicum specimens, collected from Longdong, were nested in subclade C. One specimen each of Mycedium mancaoi, Mycedium robokaki, and Pectinia paeonia, collected from Longdong, and one Mycedium elephantotus, collected from Pitoujiiao, were nested in subclade E. Subclade F consisted of one specimen each of Favites valenciennesi and Favites flexuosa, collected from Pitoujiiao, and two Favites magnistellata, collected from Pitoujiiao and Kueishan Island. All the Favites pentagona specimens collected from Kueishan Island, Pitoujiiao, and Longdong were clustered with F. pentagona from Singapore and the Philippines. Subclade G consisted of two Platygyra pini, five Platygyra ryukyuensis, three Platygyra verweyi, and one Platygyra sinensis, collected from Longdong; one species each of Platygyra lamellina and Platygyra daedalea, collected from Pitoujiiao; and two Favites stylifera from Pitoujiiao and Longdong. Six Paragoniastrea deformis, collected from Pitoujiiao, four Paragoniastrea australensis specimens from Pitoujiiao, and one from Longdong were clustered into subclade J. Subclade K was a monophyletic clade, consisting of three specimens of Astrea curta, collected from Keelung Island, and two from Pitoujiiao, which were clustered with those from the Great Barrier Reef and the Philippines (BP:98, PP:1).

Discussion
This is the first study to establish a molecular database for spawning corals, an important contribution to our understanding of genetic diversity in coral communities. We sequenced 1 species from Keelung Island, 9 from Kueishan Island, 13 from Longdong, and 12 from Pitoujiiao. In total, we sequenced 9 genera and 28 species, and most subclades were consistent with those of previous studies.

Phylogentic Relationship of Merulinid Subclades
The Merulinidae are defined as monophyletic in this study, confirming previous findings [10,16]. Of four Atlantic species, Favia fragum, Mussa angulosa, Orbicella annularis, and Montastraea cavernosa, only O. annularis is nested within the subclade C and a sister to Cyphastrea spp. The genus Paramontastraea Huang et al. 2014 [10] examined in this study was also a sister to Echinopora Lamarck, 1816, and nested within subclade I.
Increasing the sequence lengths, taxon sampling, and sampling locations may improve the phylogenetic relationship among taxa. Adding new sequences of merulinid corals from Taiwan generated longer aligned sequences with which to examine the phylogenetic relationships among the subclades of Merulinidae; as a result, most subclades changed their phylogenetic positions ( Figure S1). For example, the tree topology reconstructed in Huang et al. [10] showed that Hydnophora (subclade H) is closer to Favites (subclade F). However, our reconstructed phylogenetic tree showed that the Hydnophora lineages only closer to Favites pentagona and the rest of Favites spp. are close to novel subclade K, which comprises Astrea curta and Favites russelli. In addition, subclade B shifted its position from D/E clades to subclade H and Favites pentagona.
As mentioned in Huang et al. [10,15,16], Favites pentagona and Paragoniastrea autralensis displayed polyphyletic patterns that require further investigation [10]. Paragoniastrea australensis was far from subclade A (Goniastrea spp.) and was clustered together with Astrea curta, Astrea annuligera, Astrea devatieri, and Favites russelli as a novel clade. Favites pentagona renders Favites polyphyletic in the molecular phylogeny and sister to the Favites spp. (subclade F) and subclade D/E [10]. Therefore, Huang et al. [10,16] suggested that these two species require further study with increasing sample collection from other locations. According to our new analysis, the molecular phylogenetic tree implied that P. australensis displayed a polyphyletic pattern, which is consistent with Huang et al. [10,16]. However, they were close to subclade A (Goniastrea spp.) and formed a novel subclade J (Paragoniastrea spp.). Favites pentagona formed a monophyletic pattern, which is different from Huang et al. [10,16]. They are separate from the major Favites spp. (subclade F) and are close to subclade H (Hydnophora).

Application of Molecular Phylogentic Approaches
Merulinidae corals with plocoid and ceroid forms are difficult to accurately identify to the genus level in the field because of their macro-morphological homoplasies [24,33]. In our phylogenetic analysis, all of the corals in the plocoid form were placed into subclades B, C, F, I, and/or K, including 18 Dipsastraea spp. Most of the samples we examined were either Coelastrea spp., Favites pentagona, Platygyra spp., Paragoniastrea australensis, or Paragoniastrea deformis. The genetic divergence between these four groups (subclades B, F, G, and J) may be driven by the differences in their sexually reproductive timing. Favites pentagona and P. deformis spawn in July, while Platygyra spp. and P. australensis spawn in August [32].

Conclusions
This study integrates reproduction information, morphological characteristics, and molecular phylogenetic analysis to increase our understanding of the genetic diversity of Merulinidae. Ten major subclades (A, B, C, D/E, F, G, H, I, J, and K) were reconstructed. Our study identified Paragoniastrea deformis and Paragoniastrea australensis in Taiwan for the first time. Together, the two species form the new subclade J. Astrea curta were separated from another congeneric species, Astrea annuligera (XVII-B), clustered with Favites russelli into the new subclade K. Finally, we contributed information on the species diversity of coral communities in Taiwan and fill gaps involving merulinid corals between Japan and the Philippines in the Western Pacific.