Photosymbiotic ascidians from oceanic islands in the tropical Pacific as candidates of long-dispersal species : morphological and genetic identification of the species

Following the increase in seawater temperatures due to global warming, tropical species may expand their geographic ranges toward higher latitudes, and thus are noteworthy as potential introduced species. Among photosymbiotic ascidians, Trididemnum cyclops Michaelsen, 1921 and Diplosoma simile (Sluiter, 1909) have broad ranges in the tropical Pacific, including oceanic islands. Considering the taxonomic difficulties in identifying didemnid species, it is important to verify the species identification of specimens from oceanic islands and to examine genetic differences between them and conspecifics from continental islands. We examined zooid morphologies and partial sequences of the mitochondrial COI gene of T. cyclops and D. simile from oceanic islands in the Pacific (Bonin Islands and Hawai'i) to compare with those from the Ryukyu Archipelago (Japan), continental islands, and Caribbean Panama. The specimens from the oceanic islands were confirmed to be T. cyclops and D. simile. We assume that the colonies in oceanic islands were originally derived from the tropical western Pacific region, where the species richness is much higher than that in oceanic islands. The wide geographic range of these species suggests that T. cyclops and D. simile have long-range dispersal capabilities and broad environmental tolerances and may expand their distributions toward higher latitudes as seawater temperature increases due to global warming.


Introduction
Non-indigenous ascidians cause serious problems worldwide (Lambert 2007).Outbreaks of introduced ascidians often change native biota on large scales and cause huge losses in aquaculture.Following increases in seawater temperatures due to global warming, tropical species could expand their distribution ranges toward higher latitudes, and some species potentially become invasive in non-native regions.Among these, photosymbiotic ascidians are high-risk candidates because their cyanobacterial symbionts are known to produce various toxins (e.g., Donia et al. 2011), and the host ascidians are probably subjected to little or no predation.Additionally, species with large distribution ranges are supposed to have long-range dispersal capabilities and broad environmental tolerances, and thus would be powerful candidates for becoming invasive.Therefore, it is reasonable to further examine photosymbiotic ascidians with large distribution ranges.
In tropical and subtropical waters, about 30 ascidian species are known to harbor cyanobacterial symbionts such as Prochloron and Synechocystis (Lewin and Cheng 1989; Hirose et al. 2009a).These are all colonial tunicates belonging to the family Didemnidae, which is the largest family in the class Ascidiacea (Chordata: Tunicata) and probably includes many undescribed species, particularly in tropical waters (Kott 2005).Microscopic examination is often necessary for species identification of didemnid species, because the zooids are very small (sometimes measuring less than 1 mm in length) and simplified, resulting few morphological characters for species identification.Moreover, the colonies often inhabit cryptic sites, such as the undersides of, or crevasses in, coral limestones or dead coral fragments.Accordingly, few distribution records exist for many species (Kott 2001;Monniot and Monniot 2001).
The distribution ranges of photosymbiotic ascidians differ from species to species.For example, Didemnum molle (Herdmann, 1886) is a conspicuous photosymbiotic species because of its dome-shaped colony and its occurrence in exposed sites.This species is widely distributed from the western Indian Ocean to Fiji, and from about 30ºN to more than 30ºS (Kott 2001), but has never been recorded from some oceanic islands situated far from continents and other island groups, e.g., French Polynesia (Monniot and Monniot 1987), the Bonin Islands (Hirose et al. 2007), andHawai'i (Abbot et al. 1997).In contrast, Trididemnum cyclops Michaelsen, 1921 andDiplosoma simile (Sluiter, 1909) have considerably broader distributions than D. molle in the Pacific: the former species has been recorded from French Polynesia and the Bonin Islands, and the latter has been recorded from French Polynesia, the Bonin Islands, and Hawai'i (Monniot and Monniot 1987;Abbot et al. 1997;Hirose et al. 2007).Recently, colonies collected from Caribbean Panama were identified as D. simile based on their partial mitochondrial DNA sequences as well as morphological features (Hirose et al. 2012).These D. simile colonies may have been introduced from the Pacific.Vargas-Ángel et al. (2009) reported a recent outbreak of D. simile in American Sãmoa, where overgrowth by ascidian colonies threatens native scleractinian corals.As mentioned above, species identification of photosymbiotic didemnids is often problematic due to the shortage of morphological characters for species discrimination.For instance, we have described six Diplosoma species as new species from the Ryukyu Archipelago (Japan) since 2005, using the number of stigmata of branchial sac as a novel character for taxonomy, and the recognition of the species was also supported by molecular phylogeny analyses (see Hirose and Hirose 2009).Considering the taxonomic difficulties, it is important to verify whether the specimens from oceanic islands are actually conspecific to those from continental islands, integrating morphological and molecular methods.
In general, the species richness in tropical Indo-Pacific Oceans is highest in the Indonesian-Philippine region and steadily decreases eastward across the Pacific, consistent with dispersal from a center of high diversity (e.g., Roberts et al. 2002;Mora et al. 2003).This suggests that T. cyclops and D. simile may have greater dispersal capabilities than other photosymbiotic didemnid species, making them potential candidates expanding their distribution ranges toward higher latitudes following the global warming of seawater.In the present study, we examined zooid morphologies and the partial sequences of the mitochondrial cytochrome oxidase subunit I gene of D. simile from oceanic islands in the Pacific (Bonin Islands and Hawai'i) and compared them with those from Okinawa Island (Ryukyu Archipelago, Japan), a continental island, and from Caribbean Panama.

Animals
The sampling of ascidian colonies was carried out by snorkeling in shallow subtidal areas (<3 m) in the Bonin Islands, Hawai'i, Taiwan, and Ryukyu Archipelago, from 2007 to 2012 (Figure 1).   1. Bn, Bonin Islands; Fu, Fukui; Hw, Hawai'i; Ka, Kagoshima; Ok, Okinawa; Pa, Caribbean Panama; Tw, Taiwan.and D. watanabei colonies were collected from Yobuko-hama in Tonakijima Island (Ryukyu Archipelago) on 21 March 2012 (Figure 1).Some colonies were anesthetized with menthol and 0.37 M MgCl 2 for approximately 2 h and then fixed with 10% formalin-seawater for microscopic observation.Some pieces from colonies were also preserved in 99% ethanol for DNA extraction.

Microscopy
Formalin-fixed specimens were dissected under a binocular stereomicroscope for species identification.Zooids isolated from the fixed materials were observed under a light microscope equipped with differential interference contrast (DIC) optics.Several micrographs were combined to increase the depth of field using the post-processing image software Helicon Focus Pro 4.2.8 (Helicon Soft, Ltd., Kharkov, Ukraine).Calcareous spicules in the tunics of T. cyclops colonies were photographed under a light microscope and the diameters of 100 spicules were measured using ImageJ 1.44 (http://imagej/nih.gov/ij).

DNA extraction, amplification, and sequencing
Tissue samples were preserved in ethanol at -30°C.Genomic DNA was extracted from about 30-40 zooids using a DNeasy Tissue Kit (Qiagen K.K., Tokyo, Japan) following the manufacturer's protocol.Part of the mitochondrial COI gene was amplified using UroCox1-244F (5′-CATTTWT TTTGATTWTTTRGWCATCCNGA-3′) and Uro Cox1-387R (5′-GCWCYTATWSWWAAWACA TAATGAAARTG-3′), and the amplification and sequencing of the partial COI fragments followed Hirose et al. (2009c).The PCR primer sequences were originally constructed by Yokobori (Hirose et al. 2009c) based on the sequence comparisons of tunicate COI genes, since amplification of the so-called 'Folmer's region' of the COI gene (at the 5' end of the cytochrome c oxidase subunit 1 region) was often unstable in our specimens.Although the primers amplify a different region from the 'Folmer's region' of the COI gene, more sequence data were available in the public domain for the 'Yokobori's region' than for the 'Folmer's region' of photosymbiotic didemnid species.At least five clones were randomly selected and sequenced for each sample.The sequences determined in this study have been deposited in the DNA databases of GenBank, European Molecular Biology Laboratory (EMBL), and DNA Data Bank of Japan (DDBJ) under the accession numbers shown in Table 1.

Phylogenetic analyses
There were 16 partial COI gene sequences (three Trididemnum sequences and 11 Diplosoma sequences) obtained in this study.Partial COI gene sequences data of six Trididemnum species (eight sequences) and eight Diplosoma species (12 sequences) were retrieved from GenBank (Table 1).These were all available COI gene sequence of the 'Yokobori's region' in Trididemnum and Diplosoma.A total 36 COI sequences (11 Trididemnum sequences and 25 Diplosoma sequences) were used for phylogenetic analyses, in which Didemnum molle (AB433947), Lissoclinum bistratum (AB433977) and L. punctatum (AB 433976) were used as the outgroup (Table 1).Initial alignments were performed using MUSCLE (Edgar 2004), and these were then inspected by eye and manually edited.Base frequencies, pairwise base differences, and genetic distances were calculated using PAUP* 4.0 Beta 10 (Swofford 2003).The following analyses were performed on the aligned DNA sequences of the partial COI gene: maximum likelihood (ML) using the October 2008 version of TREEFINDER (Jobb 2008); and phylogenetic analysis based on Bayesian inference (BI) using MrBayes 3.12 (Ronquist and Huelsenbeck 2003).To select an appropriate nucleotide substitution model, Modeltest ver.3.7 (Posada and Crandall 1988) was used with PAUP.For data with all three codon positions, the general time reversible model with invariant sites and gamma distribution (GTR+I+G) was selected as the best model using  Akaike's information criterion (AIC).For the BI analysis, a Markov chain Monte Carlo (MCMC) analysis was run for 1,000,000 generations, and trees were built at 100-generation intervals (burn-in = 2500).Statistical support for the ML tree was evaluated using a nonparametric bootstrap test with 1000 resampling events.

Trididemnum cyclops from the Bonin Islands
Small oval colonies were about 5 mm in diameter.Colonies were green due to Prochloron cells distributed in the peribranchial and cloacal cavities, while the colonial margin and bottom were white due to calcareous spicules in the tunic (Figure 2A).The thorax of zooids were strongly contracted in our specimens (Figure 2B).The thorax contained three rows of stigmata.The anterior end of the endostyle was always pigmented and referred as the "pigment cap."Gonads were not found in the specimens.Stellate spicules were embedded in the tunic (Figure 2C), and their diameters were 29 µm on average (standard deviation = 6.0; maximum diameter = 48 µm).

Diplosoma simile from Okinawa Island, the Bonin Islands, and Hawai'i
Colonies were thin sheets and entirely green due to Prochloron cells distributed in the peribranchial and cloacal cavities (Figure 3A-D).Colony colors varied among colonies even within the same collection site: some colonies inhabiting shaded sites were bluish, while colonies from shallow, exposed sites often had white dots of pigment cells in the tunic.No spicules were present.Appearances of the thoraxes differed somewhat among zooids depending on the extent of contraction in the fixatives (Figure 3E-H).The thorax had four rows of stigmata.Six stigmata were in the first (top) to third rows and five stigmata were in the fourth row (bottom).The retractor muscle emerged from the underside of the thorax.The thickness of the retractor muscle varied among specimens.There were no spicules in the present species as well as the other Diplosoma species.

Phylogenetic analysis
Eight haplotypes were obtained from the partial COI gene sequences of 11 specimens from six Trididemnum species.The average base frequencies of all three codon positions from the eight haplotypes of Trididemnum specimens were 28.9% A, 9.7% C, 16.9% G, and 44.5% T, and the A+T composition frequency was 73.4%.The aligned COI sequences were 401 bp long, without gaps (insertions/deletions). Figure 4 shows a BI consensus tree using the GTR+I+G substitution model.The strict consensus ML under the GTR+I+G substitution model is not shown.The sequence of T. nubilum from Okinawa Island was identical to that of the specimen from Kume Island.The sequences of T. paracyclops from Ishigaki and Kume islands were the same as the sequence of T. cyclops from Hahajima.The monophylies of T. cyclops and T. paracyclops individually were not supported, although the monophyly of T. cyclops + T. paracyclops was Twenty-one haplotypes were obtained from the partial COI genes of 25 specimens from eight Diplosoma species.The average base frequencies of all three codon positions from 21 haplotypes of Diplosoma specimens were 31.2%A, 10.5% C, 16.7% G, and 41.7% T, and the A+T composition frequency was 72.9%.The aligned COI sequences were 401 bp long, without gaps (insertions/ deletions).Sequence differences within each species were lower (0-11.0%;0-44/401 bp) than interspecific sequence differences (10.2-22.4%;41-90/401 bp). Figure 5 shows a BI consensus tree using the GTR+I+G substitution model.The monophyly of each Diplosoma species was supported by BPP and high bootstrap values, except D. simile in which monophyly was not strongly supported (BPP = 0.63; 55% in ML).
The monophyly of D. simile specimens from the Pacific Ocean (Okinawa Island, the Bonin Islands, and Hawai'i) was supported by BPP and high bootstrap values.

Discussion
The photosymbiotic ascidians from oceanic islands (i.e., the Bonin Islands and Hawai'i) were verified as T. cyclops or D. simile, based on both morphological features and molecular phylogeny inferred from the partial sequences of mitochondrial COI genes.According to Kott (1980), T. paracyclops Kott, 1980 is distinguished from T. cyclops by differences in the colony, the spicules, the proportion of zooids, the retractor muscles, the course of the gut, and the number of vas deferens coils.However, Monniot and Monniot (1987) assigned both species to T. cyclops, considering these differences to represent intraspecific variation.Quantitative investigation showed that the spicule sizes differed significantly between the two species, but the monophyly of each species was not supported by molecular phylogenies (Hirose et al. 2010).The sizes of colonies and the diameters of the spicules in the present specimens from Hahajima were within the range of those in T. cyclops.In the present phylogenetic trees based on partial COI gene sequences, T. cyclops and T. paracyclops formed a monophyletic clade, and these two species were mixed up within the clade (Figure 4).This supports the view of Monniot and Moniot (1987) regarding T. paracyclops as a junior synonym of T. cyclops.In either case, the specimen from Hahajima was identified as T. cyclops.
The species of photosymbiotic Diplosoma can be discriminated through the liberating point of the retractor muscle from the zooid and the number of stigmata in each row (Hirose and Su 2011), and the present specimens were consistent without exception with the character states of D. simile (Figure 3E-H).Furthermore, partial COI gene sequences distinguish many photosymbiotic Diplosoma (Hirose and Hirose 2009).In the phylogenetic tree, photosymbiotic Diplosoma form two clades corresponding to the liberating point of the retractor muscle (i.e., underside of the thorax and halfway down the esophagus), and all the D. simile specimens examined here formed a monophyletic clade supported by moderate Bayesian posterior provability (BI) and bootstrap value (ML) (Figure 5).While D. simile from Caribbean Panama branched at the most basal position in the clade, another region (Folmer's) of the COI gene sequences from Caribbean specimens were the same as the sequences from Okinawa Island (Hirose et al. 2012).Here, the all specimens morphologically identified as D. simile were concluded to be conspecific.
The present results indicate that T. cyclops and D. simile probably have long-range dispersal capabilities, and we assume that the colonies from oceanic islands were originally derived from the tropical western Pacific region, where the species richness is much higher than that of the oceanic islands.Since didemnid ascidians are always viviparous, active dispersal is limited in the larval period.However, didemnid larvae are relatively large among ascidian species, and often settle and metamorphose into juvenile colonies soon after their release from the mother colony.Therefore, larval dispersal would be limited in didemnids and rafting on drifting material in the sessile stage would be more important for long-distance dispersal; as Jackson (1986) stated that "Rafting is the only reasonable explanation for the existence of the vast majority of clonal species on Indo-Pacific oceanic islands."Worcester (1994) demonstrated that rafting is a successful and important mode of dispersal in a botryllid ascidian, Botrylloides sp.It is possible that artificial transport via vessels (on ship hulls and in ballast water), drifting buoys, and other artifacts are plausible vectors for the long-dispersal of didemnid species.
In many photosymbiotic ascidians, algal symbionts are transmitted from generation to generation (vertical transmission), whereby embryos are known to acquire photosymbionts from their mother colony (Kojima and Hirose 2012 and references therein).In D. simile, pre-hatching embryos collect Prochloron cells from the cloacal cavity of the mother colony using a specialized organ called the rastrum, and the rastrum, with numerous photosymbionts, is folded in the posterior part of the larval trunk (Hirose 2000).Therefore, they enjoy the benefits of the symbionts from the beginning of their life, and this may be an advantage for survival on the drifting materials.It is uncertain why some species, such as T. cyclops and D. simile, have wide distribution ranges including oceanic islands and other photosymbiotic species do not.A possible explanation is that T. cyclops and D. simile have broader environmental tolerances than other photosymbiotic species.
Will photosymbiotic ascidians expand their geographic ranges toward higher latitudes following increases in seawater temperature?The distribution ranges of photosymbiotic ascidians are supposed to be limited by temperature minima because their cyanobacterial photosymbionts are susceptible to temperatures below 20ºC (Dionisio-Sese et al. 2001), which is consistent with the distribution patterns of photosymbiotic ascidians in the Ryukyu Archipelago-Taiwan region.The number of species gradually decreases toward higher latitudes (Hirose 2013 and references therein), and 19 species have been so far recorded from the Yaeyama Islands (24-24º30'N; the southernmost island group in the Ryukyus), while only four species, including D. simile, have been found in the Osumi Islands (ca.30ºN; the northernmost island group).In Taiwan, photosymbiotic ascidians have been recorded from Kenting (ca.22ºN; South Taiwan) and Lyudao (22º40'N; island in southeast Taiwan) but not from Keelung (ca.25ºN; the northern coast of Taiwan; Hirose and Nozawa 2010; Hirose and Su 2011).The minimum winter water temperature at the sea surface is about 16ºC on the northern and northwestern coasts of Taiwan due to cold China coastal water, but is about 19ºC at Okinawa Island (26-27ºN).Note that D. simile has been widely recorded from the Ryukyus and Taiwan, as well as Indonesia, Singapore, the Great Barrier Reef, and various islands from the west and central Pacific.The distribution of T. cyclops ranges from the west Indian Ocean to the west and central Pacific and from the Ryukyus to the Great Barrier Reef.Their broad tolerance and wide dispersal abilities suggest that they have the necessary attributes to expand their distribution range.It would be possible that they will become invasive in nonnative regions, as reported for D. simile in American Sãmoa (Vargas-Ángel et al. 2009).

Figure 1 .
Figure 1.Map of the North Pacific and Caribbean Sea indicating the regions of the collection sites listed in Table1.Bn, Bonin Islands; Fu, Fukui; Hw, Hawai'i; Ka, Kagoshima; Ok, Okinawa; Pa, Caribbean Panama; Tw, Taiwan.

Figure 4 .
Figure 4. Bayesian inference (BI) consensus tree of Trididemnum spp.based on cytochrome oxidase subunit I (COI) gene sequences.The GTR+I+G model was used for the analysis.Bayes posterior probability and bootstrap probabilities larger than 50% for the maximum likelihood (ML) analyses are noted.New sequences from this study are in bold.

Figure 5 .
Figure 5. Bayesian inference (BI) consensus tree of Diplosoma spp.based on cytochrome oxidase subunit I (COI) gene sequences.The GTR+I+G model was used for the analysis.Bayes posterior probability and bootstrap probabilities larger than 50% for the maximum likelihood (ML) analyses are noted.New sequences from this study are in bold.

Table 1 .
List of Didemnidae species used for phylogenetic tree construction in this study.