Genetic and morphological analyses of Gracilaria firma and G. changii (Gracilariaceae, Rhodophyta), the commercially important agarophytes in western Pacific

Many studies classifying Gracilaria species for the exploitation of agarophytes and the development of the agar industry were conducted before the prevalence of molecular tools, resulting in the description of many species based solely on their morphology. Gracilaria firma and G. changii are among the commercially important agarophytes from the western Pacific; both feature branches with basal constrictions that taper toward acute apices. In this study, we contrasted the morpho-anatomical circumscriptions of the two traditionally described species with molecular data from samples that included representatives of G. changii collected from its type locality. Concerted molecular analyses using the rbcL and cox1 gene sequences, coupled with morphological observations of the collections from the western Pacific, revealed no inherent differences to support the treatment of the two entities as distinct taxa. We propose merging G. changii (a later synonym) into G. firma and recognize G. firma based on thallus branches with abrupt basal constrictions that gradually taper toward acute (or sometimes broken) apices, cystocarps consisting of small gonimoblast cells and inconspicuous multinucleate tubular nutritive cells issuing from gonimoblasts extending into the inner pericarp at the cystocarp floor, as well as deep spermatangial conceptacles of the verrucosa-type. The validation of specimens under different names as a single genetic species is useful to allow communication and knowledge transfer among groups from different fields. This study also revealed considerably low number of haplotypes and nucleotide diversity with apparent phylogeographic patterns for G. firma in the region. Populations from the Philippines and Taiwan were divergent from each other as well as from the populations from Malaysia, Thailand, Singapore and Vietnam. Establishment of baseline data on the genetic diversity of this commercially important agarophyte is relevant in the context of cultivation, as limited genetic diversity may jeopardize the potential for its genetic improvement over time.

Introduction name, G. firma, is proposed to be adopted for the species considered as G. changii. The genetic diversity of G. firma in the western Pacific was also inferred. The validation of specimens passing under the name of G. changii as G. firma would deconvolute a small part of the Gracilaria taxonomy and somewhat ease the transfer of cultivation knowledge across the region.

Ethics statement
The samples are not endangered or protected species. No specific permits were required for this study, as the specimens were not collected from any national parks or protected areas. Consent was granted from the owner for specimen collection from cultivation farm.

Taxon sampling and sample processing
Specimens collected from Malaysia (n = 5), Singapore (n = 4), Thailand (n = 3), Vietnam (n = 4), Taiwan (n = 11) and the Philippines (n = 17) were identified as G. changii, G. fisheri and G. firma (Fig 1). Species identification was based on reference to regional literature [2-3, 5, 12, 15-16] as well as the samples' geographical provenance, with detailed morpho-anatomical analyses whenever possible. Scanned images of the type specimens for G. firma, G. changii and G. fisheri (Fig 2) were obtained from the Herbaria of the Institute of Oceanology, Academia Sinica [now Chinese Academia of Sciences] (AST), and the B.P. Bishop Museum, Honolulu, Hawaii (BISH), for morphological comparison with the specimens examined in this study. The isotype sample preserved in formalin was too old and fragmented for making any sections to obtain useful information. Attempt to collect samples of G. firma from the type locality, Xuwen County, Guangdong Province of China was unsuccessful. Representative G. changii samples were collected from the type locality in Penang, Malaysia (PSM10466 and PSM10481). Pulau Korea Besar is the local name given to an islet situated east of Gelugor in the Middle Bank, Penang, which is representative of the location where G. changii was typified. All specimens were dried and preserved as vouchers, either as a herbarium press or in silica gel from which a subsample of material was taken for DNA sequencing, and a subsample for some of the specimens were preserved in 5% formalin in seawater for morphological study. Collection information and GenBank accession numbers for the rbcL and cox1 gene sequence of the specimens examined in this study are listed in S1 Table, with the samples used for morphological analyses indicated by voucher number in boldface. Hand sections were stained with Wittmann's aceto-iron-hematoxylin-chloral hydrate [17] and mounted in 50% Hoyer's mounting medium [18]. Photomicrographs were taken on an Olympus BX51 microscope with a Q-imaging digital camera (Burnaby, BC, Canada), and habit views were taken on a Nikon D300 camera (Tokyo, Japan).
Genomic DNA was extracted from silica gel-preserved samples or herbarium vouchers using an i-genomic Plant DNA Extraction Mini Kit (iNtRON Biotechnology Inc., South Korea) according to the manufacturer's recommendations. PCR amplifications of each marker were carried out using primers F7/RrbcS start or F7/R1434 [19][20][21] for rbcL and COXI43F/ COXI1549R [22] for cox1, with the PCR conditions described in [23]. PCR products were cleaned using the LaboPass Gel & PCR purification kit (Cosmo Genetech, Seoul, South Korea) and sequenced commercially in both directions using the amplification primers. DNA of the isotype of G. firma was also extracted, but it failed to amplify.

Molecular analyses
Overlapping bidirectional reads of each sample for each marker were edited and assembled into a consensus sequence using ChromasPro v.1.5 (Technelysium Pty Ltd., Queensland, Australia). Electropherograms were manually examined and checked for the occurrence of multiple peaks upon encountering ambiguous nucleotides in the sequence reads. Both rbcL and cox1 datasets were aligned using ClustalX v.2.0 [24] with the default gap extension/opening parameters, and the alignments were trimmed with BioEdit v.7.0.5.3 [25]. Additional sequences of Gracilariaceae, particularly those of Gracilaria sensu lato nested within Hydropuntia Montagne complex [26], were downloaded from GenBank and included in the phylogenetic analyses along with the sequences newly generated from this study. A large number of  the cox1 sequences for Gracilariaceae available in GenBank are only about 600-700 base pairs (bp) to serve as molecular barcode for species identification. Alignments for the rbcL and cox1 genes used in the phylogenetic analyses consisted of 1070 and 1118 bp, respectively, after trimming. Identical sequences were removed from the alignment and tree reconstructions were made using one haplotype by country. Trimmed alignments used for phylogenetic analyses are available from the authors upon request. Phylogenetic reconstruction for each dataset was inferred using maximum likelihood (ML) and Bayesian analysis. Maximum likelihood tree searches were implemented in PhyML 3.0 online execution version [27], based on the GTR+G+I model automatically selected by the program using the Akaike Information Criterion for the rbcL and cox1 datasets. Branch support was evaluated using the SH-like approximate Likelihood Ratio Test (aLRT) implemented in PhyML with 1000 bootstrap replicates. Bayesian inference (BI) was conducted with MrBayes v.3.2.6 [28]. The best-fitting substitution model, GTR+G, was selected for both datasets based on the Bayesian Information Criterion implemented in Kakusan v.3.0.2009.03.08 [29]. Bayesian analyses were initiated with a random starting tree and two parallel runs, each of which consisted of running one cold chain and three hot chains of Markov chain Monte Carlo (MCMC) iterations for 2×10 6 generations. The trees in each chain were sampled every 200 th generation. The convergence of the two MCMC runs to a stationary distribution was determined by looking at the standard deviation of split frequencies (which were always less than 0.01) and by the convergence of the parameter values in the two independent runs. The first 200 trees were discarded as burn-in, and the remaining trees were used to calculate a 50% majority rule tree and to determine the posterior probabilities for both datasets. The rbcL and cox1 phylogenies were rooted with Gracilariopsis species, based on the phylogenetic relationships inferred from global searches for the Gracilariaceae [26,30]. The absolute distances and uncorrected pairwise distances for the rbcL and cox1 sequences of the specimens analyzed in this study and related reference taxa were computed with MEGA6 [31].
For haplotype parsimony network analysis, the rbcL sequences from GenBank of G. changii from Thailand (JQ026049) and G. 'blodgettii' from China (JQ407695) were aligned with those newly generated in this study. While the newly generated cox1 sequences formed a well-supported monophyletic clade with several GenBank entries for G. changii and G. 'blodgettii' from Malaysia, the Philippines and China, network analysis for cox1 was performed only with the sequences generated in this study and published sequences of G. changii [32][33], which encompassed specimens collected from the type locality in Middle Bank, Penang, Malaysia. The cox1 sequences for G. 'blodgettii' and G. changii from China (JQ407591) and the Philippines (KX017516, KT779908, KT779928 and KX017515) were not included in the network analysis, as the rather short consensus (less than 200 bp after trimming) that can result from alignment was not informative enough. The rbcL and cox1 alignments were trimmed to 1070 and 834 bp, respectively. A parsimony network was constructed for each alignment in TCS v.1.2.1 [34] using a default connection limit of 95%. Estimates of haplotype and nucleotide diversity based on rbcL and cox1 were calculated for the overall dataset, as well as populations from each country considered in the haplotype parsimony network analyses except for China which was represented only by a single rbcL sequence, by using DnaSP v.5.10.01 [35].

Molecular analyses
rbcL gene. The rbcL gene produced an unambiguously aligned dataset for 38 sequences which included 1070 bp, of which 370 were variable and 291 were parsimony-informative. Both phylogenetic inference methods (ML and BI) recovered a fully supported ingroup of Gracilaria ( Fig 3A) made up of four distinct clades (I, II, III and IV) with poor to moderate ML bootstrap support (BS) but high Bayesian posterior probability (PP), each corresponding to 'Atlantic' Hydropuntia (subgroup III), Gracilaria sensu stricto (subgroups IV, V and VIII), 'Pacific' Hydropuntia (subgroup II) and a putative new genus (subgroup I) identified in [26]. The topology derived using the ML method was incongruent with that of BI. Clade I formed a poorly supported sister clade with Clade II in the phylogeny inferred using ML analysis, but it formed a weakly supported monophyletic assemblage of Hydropuntia (PP = 0.87) with Clade III in the phylogeny derived from BI (data not shown). The samples identified as G. firma from Taiwan, Vietnam and the Philippines, G. changii from Malaysia and Singapore, and G. 'fisheri' from Thailand formed a strongly supported monophyletic clade (BS = 97% / PP = 1.00), along with reference sequences of G. 'blodgettii' from China (JQ407695) and G. changii from Thailand (JQ026049) in Clade I. They differed from each other by a maximum of 0.6% pairwise differences, suggesting conspecificity of these taxa. These specimens were most closely related to an unnamed Malaysian 'Hydropuntia' species (AB859153) and a G. 'fisheri' sample from the Philippines (JQ026026), to which they differed by 4.0-4.8%. The interspecific divergence of G. firma/ G.changii/G. 'fisheri' (from Thailand) analyzed in this study and other congeners within Clade I ranged from 7.4% to 9.3%.
Six haplotypes were inferred from the partial rbcL gene and they were connected by a maximum of six mutational steps ( Fig 3B). Apart from the samples from Singapore and Vietnam, which were represented by two haplotypes, samples from other countries consisted of only a single haplotype: R1 for the Malaysian and Thai samples, R4 for the Philippine samples and R5 for the Taiwanese samples ( Fig 3B, S1 Table). The haplotype R1 was represented by G. changii from Malaysia and Singapore, G. firma from Vietnam and G. 'fisheri' from Thailand, whereas haplotypes R2 and R3 were exclusively represented by G. changii from Singapore and G. firma from Vietnam, respectively. Haplotypes R2 and R3 differed from haplotype R1 only by a single base pair. Haplotype R5 was more closely connected to haplotype R4, with two mutational steps, than to the haplotypes R1-R3 from the mainland-bound countries, with four to six mutational steps. Representing G. 'blodgettii' from China, haplotype R6 was connected to haplotypes R3 and R4 by three mutational steps each ( Fig 3B, Table 1). The overall haplotype and nucleotide diversities of the Gracilaria populations in the western Pacific region examined in this study based on the rbcL gene were 0.748 ± 0.033 and 0.00216 ± 0.00016, respectively. While samples from Vietnam and Singapore were both represented by two haplotypes, the former recorded a higher haplotype and nucleotide diversity (0.667 and 0.00062) compared to the latter (0.500 and 0.00047). cox1 gene. The topologies of the phylogenies for 39 taxa inferred using ML and BI methods are largely congruent. A strongly supported group that corresponds to Clade I in the rbcL phylogeny was recovered in the cox1 phylogeny ( Fig 4A). The relationships among the remaining Gracilaria taxa were unresolved as hard polytomies in the Bayesian tree and with very poor nodal support from both analyses. All samples sequenced in this study formed a fully supported monophyletic clade with the GenBank entries for G. changii from Malaysia, Thailand and the Philippines, along with the Chinese and Philippines G. 'blodgettii'. These monophyletic taxa differed from each other by 0-2.4% pairwise distance over the 1118-bp alignment. These specimens were most closely related to an unnamed Malaysian 'Hydropuntia' species as well as G. 'changii' and G. 'fisheri' from the Philippines, to which they differed by 7.1-8.2%. Interspecific divergence of other samples of Atlantic origin within the G. caudata clade to the G. firma/ G. changii/G. 'fisheri' analyzed in this study ranged from 11.7% to 16.3%.
Forty cox1 gene sequences generated from this study and five published sequences representing different haplotypes of G. changii [32][33] were collapsed into seven haplotypes of 834 bp with 13 variable sites that were connected by a maximum of eight mutational steps ( Table 2). Haplotype C1 was shared among G. firma from Vietnam, G. 'fisheri' from Thailand and G. changii from Malaysia and Singapore, whereas haplotype C5 was shared with the Vietnamese G. firma and Malaysian G. changii. The remaining haplotypes C2, C3 and C4 were unique to Malaysian samples, each differing from haplotype C1 by one to three mutational steps. The haplotypes C6 and C7, each representative of G. firma from the Philippines and Taiwan, were more closely related to one another than they were to the haplotypes C1-C5 that represent the mainland-bound countries. The former differed from one another by three mutational steps, while they differed from the latter by five to eight mutational steps (Fig 4B, Table 2). The overall haplotype and nucleotide diversities of the populations examined in this study based on cox1 were 0.766 ± 0.038 and 0.00364 ± 0.00024, respectively. In contrast to the populations from Thailand, Singapore, the Philippines and Taiwan that showed no variation, the samples from Malaysia and Vietnam recorded a haplotype diversity of 0.800 and 0.500, and nucleotide diversity of 0.00224 and 0.00120, respectively.

Morphological and anatomical observations
Samples from Taiwan, Vietnam and the Philippines displayed morpho-anatomical features that conformed to the circumscription for G. firma (Fig 5). Their identity was also confirmed by morphological comparison with the holotype and isotype of G. firma (Fig 2A and 2B). Samples from Malaysia and Singapore were identified as G. changii based on geographical provenance, morphological comparison with the holotype, and molecular evidence. Although our sterile G. changii collections precluded anatomical observation of the reproductive structures, they showed identical rbcL and cox1 sequences as those collected from the type locality in Penang, Malaysia. The voucher samples of the Thai samples were lost and not morphologically examined; therefore, they were only provisionally assigned as G. 'fisheri' based on their geographical provenance. In addition to the molecular data, careful comparison of the morphological observations of type specimens (Fig 2) and the species descriptions [3,5] (Table 3) revealed that G. firma and G. changii showed no inherent differences to support their treatment as distinct species. In accordance with the principle of priority in nomenclature, we propose to reduce G. changii to G. firma, and the morpho-anatomical descriptions are given below.
https://doi.org/10.1371/journal.pone.0182176.t001 ( Fig 5A-5C) to scarcely branched (Fig 5D). The branching is mainly irregular, sometimes alternate or secund in portions, and up to four orders or more. Branches are terete, always constricted at the base, either only slightly or deeply (Fig 5A-5D, arrowheads), and taper gradually toward acute or obtuse apices. Obtuse apices occur when thallus tips are broken, and two or more fine, short branchlets tend to regenerate from a single broken tip (Fig 5B-5D, arrows). Thalli texture is firm and cartilaginous. Color ranges from light or brownish green to olive black when fresh. Transition of cell size from cortex to medulla is gradual in the thalli of higher branching orders (Fig 5E-5G) and abrupt in the primary or secondary branches (Fig 5H). The thallus consists of one to two layers of pigmented cortical cells measuring 7.5-12.5 μm long by 3.8-7.5 μm wide, two to three layers of subcortical cells measuring 47.5-87.5 μm long by 25-40 μm wide, and five to nine layers of lightly staining medullary cells, which are irregularly subspherical to polygonal, measuring from 60-100 μm to 180-250 μm in diameter, and up to 530 μm in older thalli (Fig 5H).
Reproductive structures. The gametophytes are dioecious. Tetrasporophytes and gametophytes are isomorphic. Cystocarps are borne over the surfaces of the fertile thalli. Mature cystocarps are hemispherical and slightly or not constricted at the bases and 0.8-1.1(-1.8) mm in diameter. Carpogonial branches and early postfertilization stages were not seen in the specimens examined. After presumed fertilization, fusion cells (Fig 6A, arrowhead) begin to form by the progressive incorporation of sterile branches and neighboring vegetative cells to the cells of the carpogonial branch. The formation of the pericarp and fusion cell is more or less complete prior to gonimoblast initiation. As the gonimoblasts develop, the pit connections between the pericarp cells at the level of the fusion cell break down to initiate a schizogenous cavity (Fig 6A), which enlarges as the pericarp becomes thicker, and the gonimoblasts grow ( Fig 6B). The cells at the cystocarp floor or inner pericarp start to form multinucleate tubular nutritive cells that extend to the lower part of the outer pericarp (Fig 6B, arrow). At maturity, the innermost cells of gonimoblasts are vacuolated and united by numerous secondary pit connections, whereas the outer gonimoblast cells differentiate into tightly packed small cells bearing short straight chains of carposporangia ( Fig 6C). Gonimoblast cells at the cystocarp floor pit-connect to the inner pericarp cells secondarily. Inconspicuous multinucleate traversing filaments connecting only the inner pericarp cells and gonimoblast cells are formed at the cystocarp floor (Fig 6C-6E, arrows). The fusion cell is persistent in mature cystocarp (Fig 6E, Numbers in parentheses after taxa denote the number of specimen with identical sequences from each country. (B) Haplotype network for G. firma based on the alignment of partial cox1 genes consisting of 834 bp. Circle size is proportional to haplotype frequency. Lines connecting the haplotypes represent single base pair mutations. Small black circles represent undetected or hypothetical haplotypes. Haplotypes are colored according to the geographic origin as shown in the key.
https://doi.org/10.1371/journal.pone.0182176.g004    (Fig 6E). The cystocarp sometimes fails to develop beyond the fusion cell-stage and generate gonimoblasts, but the pericarp continues to enlarge as in developing cystocarps (Fig 6F). Thalli with aborted carposporophytes also bear cystocarps of normal size and form. Reproductive structures are scattered over the thallus except the basal and extreme apical portions. Spermatangial parental cells are initiated from outer cortical cells and are arranged in depressions that form the spermatangial conceptacles ( Fig 6G). Mature spermatangia are 3-5 μm long by 1.5-2.3 μm wide, scattered over the surface of the deep verrucosa-type conceptacles with an average depth of 50 μm up to 120 μm (Fig 6H-6I). Tetrasporangia are initiated superficially from terminal cells cut off by oblique, longitudinal cell division of the outer cortical cells. They enlarge and are surrounded by elongated cortical cells (Fig 6J). Mature tetrasporangia measure up to 40 μm long by 20 μm wide and divide twice to produce four cruciately arranged tetraspores.

Conspecificity of Gracilaria firma and G. changii
Many Gracilaria species with stipitate branches that taper toward acute apices have been described from the western Pacific, including G. firma, G. changii and G. fisheri (Fig 2). They were likened to G. blodgettii from the western Atlantic based on their superficial branch morphology [3,36]. The Atlantic entity was distinguished from its Pacific counterparts by the shallow spermatangial conceptacles of the textorii-type [5,37]. Gracilaria firma, G. changii and G. fisheri were traditionally distinguished by differences in the thallus size, cell size transition Genetic and morphological analyses of Gracilaria firma and G. changii from cortex to medulla, presence or absence of the nutritive filaments in the cystocarp cavity, as well as the shape of spermatangial conceptacles (Table 3). Molecular analyses based on the rbcL and cox1 gene sequences showed that specimens identified as G. firma, G. changii and G. 'fisheri' are very similar. The specimens examined in this study also formed a monophyletic clade with several unpublished GenBank entries for G. 'blodgettii' from China and the Philippines with full nodal support in the phylogenetic reconstructions (Figs 3A and 4A). Taxa in the monophyletic clade were most closely related to an unnamed Hydropuntia species from Malaysia as well as some sequences identified as G. changii and G. fisheri from the Philippines, of which they differed by 4.0-4.8% over 1070 bp of the rbcL and 7.1-8.2% over 1118 bp of the cox1 genes. The nucleotide divergence based on rbcL was considered well beyond the intraspecific threshold of 1.5-2.0% recommended for members of Gracilariaceae [26]. In addition, G. 'changii' and G. 'fisheri' from the Philippines were not sampled from the type locality and thus their actual identity remained unverified. On the other hand, our specimens included samples of G. firma identified after thorough morphoanatomical examination, as well as G. changii collected from the type locality in Malaysia, of which all formed a monophyletic clade in the phylogenies inferred using both rbcL and cox1 genes. These taxa showed nucleotide divergence of 0.6% and 2.4% in the rbcL and cox1 gene sequences, respectively, which are below the intraspecific nucleotide variation threshold of 1.5% and 2.6% suggested for the rbcL [26] and cox1 [38] genes, respectively, in Gracilariaceae. This implied that specimens passing under the different names of G. firma, G. changii, G. 'fisheri' and G. 'blodgettii' in the western Pacific, all characterized by the presence of stipitate branches tapering toward acute apices, may belong to a single genetic entity (Figs 3A and 4A). We were unable to examine the morpho-anatomical features of the samples provisionally designated as G. 'fisheri' from Thailand analyzed in this study, as the vouchers were lost. Nevertheless, the results suggested the conspecificity of G. firma and G. changii. We did not consider the conspecificity of G. fisheri with G. firma, since the actual identity of the Thai specimens as G. 'fisheri' cannot be ascertained by morphological observations in this study.
Our recent collections exhibit similar morphology to the type specimens (Fig 2) and the species descriptions in the literature ( Table 3). The rbcL and cox1 gene sequence analyses also support the treatment of G. firma and G. changii as a single species. Although the specimens of G. firma examined in this study are not from the type locality, species identification was confirmed by comparing the morpho-anatomical observations with previous literature [3,[11][12] and type specimens of G. firma (Fig 2A and 2B). Specimens from Vietnam, Taiwan and the Philippines typically exhibited gradual transition in cell size from cortex to medulla, cystocarps with small gonimoblast cells and spermatangial conceptacles of the verrucosa-type, all characteristic of G. firma (Table 3).
The type of transition in cell size from cortex to medulla, whether gradual or abrupt, was considered to be consistent within Gracilaria species [39]. However, we observed that the specimens of G. firma from Taiwan showed gradual transition in cell size from cortex to medulla in young branches (Fig 5E-5G) and abrupt cell size transition in mature branches (Fig 5H). A continuum in the type of cell size transition across vegetative thallus of different ages was also observed in G. blodgettii [37]. The gradation of cell size across the thallus appears to vary with age and may not be a diagnostic taxonomy character. The nutritive filaments are a prominent feature of Gracilaria sensu lato and the position of these filaments varies in species. Most of the members of Gracilaria sensu stricto have the nutritive filaments extending from gonimoblast cells to both the outer pericarp and the cystocarp floor, while these filaments are restricted to the cystocarp floor in Hydropuntia [30,40] [18], have nutritive filaments restricted to the cystocarp floor and a virtual lack of these filaments between the gonimoblasts and outer pericarp. It should be noted that G. firma was originally described to be lacking nutritive filaments between the gonimoblasts and pericarp (page 144, lines 3-4) [3]. As mentioned earlier, nutritive filaments do occur at the cystocarp floor in many species of Gracilaria, including species which are also lacking the filaments between pericarp and gonimoblasts. Detailed morpho-anatomical observations on the specimens of G. firma from our sequenced collection revealed the existence of nutritive filaments that are only weakly formed at the cystocarp floor (Fig 6C and 6D). We speculated that such nutritive filaments might have been overlooked when G. firma was first described, as the filaments are not as conspicuous as those present in other members of Gracilaria. The presence of such inconspicuous nutritive filaments connecting the gonimoblasts and cystocarp floor in G. firma echoed the scarce presence of basal absorbing filaments in G. changii [5]. The type of spermatangial conceptacles was considered as the main differentiating feature that set G. firma and G. changii apart [13], but a mixture of spermatangial conceptacles of the verrucosa-type and Polycavernosa-type had been observed within the same branchlet of G. changii and G. fisheri [2]. As such, there seems to be no absolute designation of male type in a single species, at least in G. changii and G. fisheri, where spermatangial conceptacles of the verrucosa-type and Polycavernosa-type can grade into each other. The Polycavernosa-type had been considered as a variation, and sometimes the mature manifestations, of the verrucosa-type of spermatangial conceptacles [2]. Our detailed morpho-anatomical observations, as well as a comparison of the species description in literature, showed that the reproductive anatomy of G. firma is largely congruent with that circumscribed for G. changii. Both G. firma and G. changii are characterized by branches that are constricted at the base and taper toward acute or obtuse (broken) apices, cystocarps consisted of small gonimoblast cells with carposporangia borne at the periphery and inconspicuous multinucleate tubular nutritive cells extending from gonimoblasts at the cystocarp floor, as well as spermatangia borne in deep conceptacles of verrucosa-type.
Based on the molecular data and morphological observations, we recommended the merging of G. changii and G. firma, with the oldest validly published name, G. firma Chang & Xia, being adopted for this genetic species according to the principle of priority in nomenclature. Accordingly, the species with the aforementioned morpho-anatomical features from the western Pacific (i.e., G. changii) should be treated as G. firma, and this taxonomic conclusion is made formal below. The taxonomy status of G. fisheri will be a topic for future work until more samples are available for examination.

Genetic diversity of Gracilaria firma
The specimens examined in this study exhibited morphological variations in branch size and branching pattern for the five chlorotypes (R1-R5) and mitotypes (C1, C4-C7) recovered. Some bushy specimens with more than four branching orders display very short internodes with many branches and branchlets, whereas others with a branching order of three or less can have very few or many unilateral branchlets. Occasionally, two thick branches can coalesce into one, of which the medullary tissues had degenerated and left behind a hollow cavity.
The genetic diversity estimates for G. firma were similar to those reported for Gracilariopsis chorda (Holmes) Ohmi [42] and G. vermiculophylla (Ohmi) Papenfuss [43] in the northwestern Pacific, with high haplotype (~0.7) but low nucleotide diversities (~0.002-0.003). Gracilaria firma showed apparent phylogeographic structure across the countries bordering the South China Sea, which is likely to be shaped by the oceanographic conditions in the western Pacific. The recovery of a Vietnamese sample with identical haplotypes association as the samples from Malaysia, Singapore and Thailand (R1-C1) implied genetic connectivity along the coastline of the Southeast Asian mainland (S1 Table, Fig 1). The haplotype C5, deemed unique to the remaining Vietnamese samples examined in this study, is apparently identical to a Gen-Bank entry of G. changii (JX228063) collected from the east coast of Peninsular Malaysia over 834 bp (Fig 4B). Genetic homogeneity of the G. firma populations from the mainland-bound countries could be a result of postglacial recolonization. These countries were once connected as part of the Sunda shelf which emerged with other land bridges during the Pleistocene [44]. The South China Sea that is enclosed by the Sunda Shelf may have served as one of the refugia where population isolation and differentiation occurred [45] to produce the rather high number of haplotypes recovered in the G. firma sampled along Vietnam to the Malay Peninsula (three chlorotypes, R1-R3 and five mitotypes, C1-C5). The Monsoon-driven current circulation pattern of the South China Sea and Malacca Strait may have promoted spore dispersal, thus maintaining gene flow in the populations along the coastlines of Vietnam, Malaysia and Singapore. Populations in the Philippines (R4-C6) and Taiwan (R5-C7) appeared to be genetically disconnected from the populations in the mainland-bound countries. The San Bernardino Strait that carries water from the western Pacific into the Sulu Sea may have limited the gene flow from Sorsogon in the Philippines with the Malaysian, Singaporean, Thai and Vietnamese populations that are in direct contact with the South China Sea (Fig 1), thus maintaining its unique haplotype. Several studies revealed that red algal species such as Gracilaria salicornia (C. Agardh) Dawson [46] and Gelidiella fanii S.-M. Lin [47] also showed divergence between populations from the Philippines and other mainland-bound Southeast Asian countries. Similarly, maintenance of the endemic haplotype in Taiwan could be attributed to the unidirectional northeastward circulation pattern of the Kuroshio Current that flows from the Pacific across the Luzon Strait into the Taiwan Strait, thereby restricting the gene flow between the Taiwanese population and the Southeast Asian populations (Figs 1, 3B and 4B). Cultivation of the agarophyte by means of vegetative propagation could also contribute to fixation of particular haplotypes at a regional scale [48]. The possibility of anthropogenic factors mediating the genetic connectivity of G. firma around South China Sea cannot be ruled out as the shipping traffic and material exchange have not been well documented. However, this region appeared to be the native distribution range of this species and probabilities are high that farms have been established only using local material.
The establishment of molecular markers as barcodes for rapid species identification to drive further multidisciplinary studies is only meaningful when a species is correctly assigned a validly published name to represent a single genetic species. A significant correlation between haplotype richness and the number of collection sites had been reported [49]. Intensive sampling effort to include samples from multiple distinct sites is necessary to reveal the genetic diversity of the species in the region more comprehensively. This is especially relevant in the context of agarophyte cultivation, as limited genetic diversity of crops jeopardizes the potential for sustained genetic improvement over the long term and renders them more susceptible to disease and pest epidemics [48,50]. DISTRIBUTION: Widely distributed in the tropical and subtropical regions in the western Pacific, including China [3,7,[51][52], Malaysia (as G. firma and G. changii in [7,16]; G. changii in [5,[32][33]; and G. firma in [53]), the Philippines (as G. firma and G. changii in [54]; G. firma in [7]), Thailand (as G. firma and G. changii in [7,55]; G. changii in [15]), Vietnam (as G. firma in [13]; G. verrucosa, G. fisheri and G. blodgettii in [11]), Japan [56], Taiwan (as G. confervoides in [57]; G. verrucosa in [58]; G. blodgettii in [59]; G. tenuistipitata var. liui [8]; and G. firma in [12]) Supporting information S1 Table. Collection information and GenBank accession numbers for the rbcL and cox1 gene sequences of the specimens examined in this study. Voucher number in boldface indicates sample used for morphological analyses; asterisk after voucher number indicates sample collected from type locality. (DOCX)