Description and molecular phylogeny of a new species of Phoronis (Phoronida) from Japan, with a redescription of topotypes of P. ijimai Oka, 1897

Abstract We describe Phoronis emigi sp. n. as the eighth member of the genus based on specimens collected from a sandy bottom at 33.2 m depth in Tomioka Bay, Amakusa, Japan. The new species is morphologically similar to P. psammophila Cori, 1889, but can be distinguished from the latter by the number of longitudinal muscle bundles in the body wall (56–72 vs. 25–50 in P. psammophila) and the position of the nephridiopores (situated level with the anus vs. lower than the anus in P. psammophila). Using sequences of the nuclear 18S and 28S rRNA genes and the mitochondrial cytochrome c oxidase subunit I (COI) gene, we inferred the relationship of P. emigi to other phoronids by the maximum likelihood method and Bayesian analysis. The analyses showed that P. emigi is closely related to P. hippocrepia Wright, 1856 and P. psammophila Cori, 1889. We describe the morphology of the topotypes and additional material for P. ijimai Oka, 1897. Neither our morphological observations of P. ijimai, nor the phylogenetic analyses based on 18S and COI sequences, contradicts that P. vancouverensis Pixell, 1912 is conspecific with P. ijimai, a synonymy that has long been disputed.


Introduction
Phoronids, or horseshoe worms, are exclusively marine, sedentary, vermiform animals with a crown of ciliated tentacles, the lophophore, used in suspension feeding. They comprise the small phylum Phoronida, which currently contains two genera, Phoronis Wright, 1856 andPhoronopsis Gilchrist, 1907, with seven and three species, respectively (Emig 2007). Phoronid species are morphologically well defined, primarily on the basis of the arrangement and pattern of the body-wall musculature, nephridia, and lophophore in adults (e.g., Emig 1974Emig , 1979Emig , 1982. They produce characteristic actinotroch larvae, and most species have a cosmopolitan distribution (Emig 1982, Zimmer 1991. For over the last half century, no new species of phoronids have been established, although the current species diversity is likely to have been underestimated (Santagata and Zimmer 2002), with Phoronis pallida Silén, 1952 and Phoronopsis californica Hilton, 1930 being the most recently described valid species in each genus (Silén 1952, Hilton 1930. More recently described nominal species have been regarded as invalid, junior synonyms of older names based on morphological concordance: Phoronis svetlanae Temereva & Malakov, 1999 as synonymous with P. ijimai Oka, 1897(Emig 2007, and Phoronopsis malakhovi Temereva, 2000 with Phoronopsis harmeri Pixell, 1912(Emig 2003. Since DNA sequence data have been obtained for almost all valid species in the phylum (e.g., Santagata and Cohen 2009, and references therein), sequences from Phoronis svetlanae and Phoronopsis malakhovi would have helped either to discriminate these species from congeners or to corroborate the proposed synonymies.
One of the unsettled taxonomic issues in phoronid systematics is whether or not P. ijimai Oka, 1897 (type locality: Misaki, Japan) is conspecific with P. vancouverensis Pixell, 1912 (type locality: Vancouver, Canada). Emig (1971aEmig ( ,b, 1974Emig ( , 1977Emig ( , 1982Emig ( , 2007 synonymized these two nominal species based on similarity in various anatomical features in adults. Santagata and Zimmer (2002), however, avoided drawing a definitive conclusion on this synonymy, arguing that the late and competent larval stages described by Zimmer (1964) for P. vancouverensis were not recorded for P. ijimai in developmental observations by Ikeda (1901) and Wu and Sun (1980). Most of the DNA sequences from species in this complex currently deposited in GenBank are registered under the name P. vancouverensis, and all are derived from specimens collected in the northeastern Pacific, at localities closer to Vancouver than to Misaki: Friday Harbor, WA (Fuchs et al. 2009, Sperling et al. 2011; Monterey, CA (Cohen 2000, Mallatt andWinchell 2002); and Los Angeles, CA (Erber et al. 1998). For some sequences, the locality of origin is not reported in GenBank (Halanych et al. 1995, Passamaneck and Halanych 2006, Bourlat et al. 2008). On the other hand, no sequence data have been reported for P. ijimai, either from its type locality or a reasonably close locality in the northwestern Pacific. Undoubtedly, this has in part contributed to the continuing dispute over synonymy.
In this paper, we 1) describe a new phoronid species from Japan, which differs from all the previously known species in adult morphology; 2) reconstruct the phylogeny of representative phoronids, including the new species, based on DNA sequences of the nuclear 18S and 28S rRNA genes (hereafter, 18S and 28S, respectively), and the mitochondrial cytochrome c oxidase subunit I gene (COI); 3) describe topotypes of P. ijimai from Misaki, Sagami Bay, and discuss the synonymy with P. vancouverensis in the context of adult morphology and the molecular phylogeny; and 4) provide a key to the Japanese phoronid species.

Sampling
A sediment sample was obtained with a Smith-McIntyre grab having an aperture of 25 cm × 25 cm, from a sandy bottom at 33.2 m depth (32°32'27"N, 130°03'17"E) in Tomioka Bay, Amakusa, Kumamoto, Japan (Fig. 1A, 1B) on 26 November 2009 by Keiichi Kakui, Hiroshi Yamasaki, and Shushi Abukawa on board the research and training vessel Seriola of the Amakusa Marine Biological Laboratory (AMBL), Kyushu University. The sediment was agitated and stirred in a bucket with seawater and the supernatant was decanted; specimens suspended in the supernatant were collected with a sieve having a 0.3-mm mesh size. Of the 560 specimens obtained, most were fixed in 10% formalin seawater, and the rest were placed directly in 99% EtOH.

Morphological observation
Measurements of the lophophore and body size were taken from digital photographs with ImageJ 1.37v software (Rasband 1997, Abramoff et al. 2004. For observation of internal morphology, specimens were dehydrated in an ethanol series, cleared in n-butanol, embedded in paraffin, sectioned at a thickness of 5-6 μm, and stained with hematoxylin-eosin (HE). DeltaViewer 2.1.1 software (Wada et al. 2005) was used to construct three-dimensional images of the nephridium. All the type and voucher specimens have been deposited in the National Museum of Nature and Science, Tsukuba, Japan (NSMT).

DNA extraction and PCR amplification
Total genomic DNA was extracted from one of the ethanol-fixed specimens of the new species, as well as one of the topotypes of P. ijimai (NSMT-Te 881), using a DNeasy Blood and Tissue Kit (Qiagen), following the manufacturer's protocol. The 18S gene was amplified with three primer sets: 1F/4R, 3F/18sbi, and 18Sa2.0/9R (Giribet et al. 1996, Whiting et al. 1997. The 28S fragment was amplified with primer set LSU5/LSU3 (Littlewood 1994). The COI fragment was amplified with the primer pair LCO1490/HCO2198 (Folmer et al. 1994). PCR reactions were performed with ExTaq (TaKaRa). Conditions for hot-start thermal cycling were 2 min at 94°C; 35 cycles of 45 sec at 94°C, 45 sec at 50°C, and 90 sec at 72°C; and 7 min at 72°C. PCR products were visualized on a 1% agarose gel and purified according to the method of  Oka, 1897 at Misaki, Sagami Bay D enlargement of Hiroshima Bay, with the solid circle indicating an additional collecting site for P. ijimai at Etajima. Boom et al. (1990) with some modifications Tachi 2009, Kobayashi et al. 2009). Cycle sequencing was performed with BigDye Terminator 3.1 (Life Technologies). The PCR primers were used for sequencing reactions, together with two additional 28S primers, D2F (Littlewood 1994) and a truncated version (Thollesson and Norenburg 2003) of 28z (Hillis and Dixon 1991). Both product strands were sequenced with an ABI 3130 Genetic Analyzer (Life Technologies). Chromatograms were edited and overlapping sequence fragments were assembled by using ATGC 4.0.6 (GENETYX). The sequences have been deposited with DDBJ/EMBL/GenBank under accession numbers AB621913-AB621915 for the new species and AB752304-AB752305 for P. ijimai (Table 1).

Morphological analyses
From the literature (Emig 1974, Santagata andCohen 2009) and our own data, we tabulated 32 morphological and reproductive characters (Suppl. material 1) among 11 phoronid species. Based on this data matrix, we performed three different analyses using Mesquite version 2.75 (Maddison and Maddison 2011): 1) a cluster analysis with  (2000); e Giribet et al. (2000); f Cohen and Weydmann (2005); g Mallatt and Winchell (2002); h Halanych et al. (1995) single-linkage method based on distances between taxa calculated from the data matrix; 2) a morphology-based cladistic analysis; and 3) a most-parsimonious reconstruction of ancestral characters. For the cladistic analysis, a heuristic search was conducted with tree length criterion and rearrangement by subtree pruning and regrafting (SPR); all trees were rooted with Phoronis ovalis Wright, 1856 as the outgroup based on the results of Santagata and Cohen (2009). The ancestral character reconstruction was carried out based on the maximum-likelihood tree based on concatenated COI-18S-28S dataset (see below) for the 21 adult morphological characters.

Molecular phylogeny
We checked validity of the yielded COI sequences to prevent the isolation of nuclear encoded mitochondrial psuedogenes (NUMTS) instead of true mitochondrial sequences before phylogenetic analyses. We regarded the consistently yielded fine single peaks for all the analysed sites in chromatograms and including neither indel nor stop codon as the criteria for judging the safely rejection of the possibility for the contamination of NUMTS. The COI, 18S, and 28S sequences obtained for the new species were aligned with those from other phoronids deposited in GenBank (Table 1) using Clustal W ( Thompson et al. 1994) implemented in Seaview 4.2.5 (Gouy et al. 2010) and/or MEGA 5.05 (Tamura et al. 2011). The alignment was performed gene by gene, before concatenated data sets were generated. These sequences were analyzed both independently and as concatenated data sets.
Maximum likelihood (ML) analyses was performed with MEGA 5.05. For ML, the best-fit model for all data sets determined by the AICc implemented in MEGA 5.1 was GTR+G+I (general time reversible [Tavaré 1986] with gamma-distributed rates and invariant rates among sites). Optimal ML trees were found by a nearest neighbor interchanges (NNI) search, starting with a tree topology generated by the BIONJ method (Gascuel 1997) using maximum composite likelihood (MCL) distances (Tamura et al. 2004). One-thousand bootstrap pseudoreplicates were analyzed to obtain nodal support values.
Bayesian analyses were performed by using MrBayes 3.1.2 (Ronquist and Huelsenbeck 2003). The best-fit substitution model was GTR+G+I model, determined from AICc tests in MrModeltest 2.3 (Nylander 2004) and PAUP* 4.0b10 (Swofford 2003). A Markov-Chain Monte-Carlo (MCMC) search was performed with four chains, each of which was run for 1,000,000 generations. Trees were sampled every 100 generations, and those from the first 250,000 generations were discarded as burn-in, ensuring that a stable likelihood had been reached. Trace files generated by Bayesian MCMC runs were inspected in TRACER 1.5.0 (Rambaut and Drummond 2007) to check that the number of sampling generations and effective sample sizes were large enough for reliable parameter estimates. A consensus of sampled trees was computed, and the posterior probability for each interior node was obtained to assess the robustness of the inferred relationships.
Distribution and habitat. Phoronis ijimai is widely distributed in the North Pacific, along the coasts of North America, Canada, Japan, and Russia, including the Sea of Japan (Emig 1971a, 1974, Emig and Golikov 1990, Temereva and Malakhov 1999. Phoronis ijimai has been reported from hard substrates such as rocks, bivalve shells, and wood, and also from a sandy bottom; it often forms dense populations, up to about 15,000 individuals per m 2 (Emig 1974).
Remarks. Our topotype material of P. ijimai collected from Misaki perfectly agrees with previous morphological accounts of this species (Oka 1897, Emig 1971a, 1974 in the following characters: 1) the long nephridial papilla and the large anal funnel of  the nephridium, 2) the small diameter of the two giant nerve fibers, 3) the number of longitudinal muscles in the right oral and both anal coeloms, and 4) the brooding of embryos on lophophoral organs. These characters also agree with the description of P. hippocrepia, but differ in 1) the large number of longitudinal muscles in the right oral coelom, and 2) the single chamber in the ascending branch of the nephridium. Our topotypes of P. ijimai also match the description of P. vancouverensis (Pixell 1912, Emig 1971a, 1974. While our specimens have slightly fewer longitudinal muscles in the right anal and left oral coeloms compared to the original description of P. vancouverensis by Pixell (1912) and the revised description of P. ijimai by Emig (1974), respectively, the numbers are within the range of variation in P. ijimai (Emig 1974). The topotypes had fewer tentacles, probably due to the smaller size of the body and lophophore.
Etymology. The specific name, a masculine noun in the genitive case, is in honor of the French researcher Dr. Christian C. Emig for his remarkable contributions to lophophorate systematics.
Gonads not observed in any of our specimens; sex could thus not be determined. Distribution and habitat. Phoronis emigi is known only from a sandy bottom in northern Tomioka Bay, Amakusa, Japan, where we detected densities of up to about 90 individuals per 100 cm 2 . We observed no chitinous tubes after agitation and decantation during sampling, but the tubes would be fragile and might have been lost.
Remarks. Phoronis emigi sp. n. is morphologically most similar to P. psammophila Cori, 1889, with which it has in common 1) a long ascending branch of nephridium that is more than three times the length of the descending branch, 2) a single nephridial funnel, with the aperture situated at the tip of the descending branch, 3) a single giant nerve fiber situated on the left side, and 4) two lateral mesenteries. Phoronis emigi differs from P. psammophila in the number of longitudinal muscle bundles in the body wall (56-72 vs. 25-50 in P. psammophila) and the position of the right nephridiopores (at the same level as the anus vs. lower than the anus in P. psammophila) (cf. Andrews 1890, Selys-Longchamps 1907, Marsden 1959, Long 1960, Emig 1968, 1971b, 1979. Naturally, P. emigi is morphologically similar to, but distinct from, the nominal Phoronis architecta Andrews, 1890, which is regarded as a junior synonym of P. psammophila (Emig 1971b(Emig , 1974. Based on the descriptions by Andrews (1890) and Brooks and Cowles (1905), Emig (1971bEmig ( , 1974 noticed that P. psammophila and P. architecta are morphogically identical, with the exception of the differences in larval brooding type and the presence of nidamental gland. Subsequently, Emig (1977) found that P. psammophila shows a sympatric occurrence with Phoronis muelleri in the type locality of P. architecta; therefore, he concluded that the larval brooding type and the absence of nidamental gland of P. architecta described in Brooks and Cowles (1905) came from a specimen of P. muelleri. On the other hand, some researchers have suggested the need of reexamination of the synonymy (Stancyk et al. 1976, Santagata andZimmer 2002). Although we could not observe the larval brooding type of P. emigi, the present species is clearly different from any of these species, P. psammophila, P. muelleri, and nominal P. architecta, in the adult morphologies such as number of longitudinal muscle bundles.
The lack of gonads in our specimens was probably due to breeding seasonality. The breeding period of phoronid species previously studied is generally from spring to autumn (Rattenbury 1953, Emig 2003, whereas our material was collected at the end of November. Our specimens were likely in the post-breeding condition, following spawning and the relaease of embryos.

Morphological analyses
In the resulting cladogram from the cluster analysis (Fig. 12A), three major clades were retrieved: 1) Phoronopsis harmeri + Ph. californica + Ph. albomaculata; 2) Phoronis emigi + P. psammophila + P. muelleri + P. pallida; and 3) P. hippocrepia + P. ijimai + P. australis. It shows the morphological similarity of the new species P. emigi with P. psammophila, sharing 16 adult morphological characters. Phoronis emigi also resembles P. muelleri and P. pallida, with which it shares 15 and 12 characters, respectively ( Fig. 12A; Suppl. material 1). We conducted another cluster analysis without nephridial characters (eliminating character 6-14 in Suppl. material 1) to test the influence of the large amount of nephridial characters. In the resulting cladogram (Appendix 1 - Supplementary Fig. S1A), the same three major clades mentioned above were also obtained, although the topology between/ within the three clades changed.
Our cladistic analysis yielded 57 equally parsimonious trees. The majority-rule consensus tree of those (Fig. 12B) did not resolve the relationship between P. emigi, P. psammophila, P. muelleri, and P. pallida; these four species formed a large clade together with Phoronopsis spp., with low consensus frequency value (68.4%). Another clade including three species (P. australis + P. ijimai + P. hippocrepia) appeared as a sister group to this large clade; P. australis formed a clade with P. ijimai (89.5% in consensus frequency), to which P. hippocrepia was the sister taxon (79.0% in consensus frequency). A parsimony tree without nephridial characters (Appendix 1 - Supplementary Fig. S1B) was almost identical to the tree including nephridial characters, except that P. emigi appeared as sister to Phoronopsis (85.0% in consensus frequency), and P. ijimai formed a clade with P. hippocrepia (67.0% in consensus frequency).

Molecular phylogeny
In this study, most of the sites for both 18S and 28S were unambiguously aligned; therefore, we used the entire region excluding gap sites for our phylogenetic analyses. For the COI dataset, we used all the codon positions in our phylogenetic analyses.
The 18S dataset comprised 1756 bp aligned sites, with 208 variable sites, for 15 ingroup taxa. In the resulting ML tree (Fig. 13A) (log L = −4104.32), not all nodes are resolved or well supported. Phoronis emigi appears in a polytomous clade along with P. architecta (= psammophila) and a large, weakly supported clade that includes P. ijimai and nominal "P. vancouverensis" from California. Japanese P. ijimai is the sister taxon to nominal "P. vancouverensis" from California, with high nodal support (100/1.0). These species are embedded in a clade otherwise containing only P. australis from various localities, with Spanish P. australis the sister taxon to the ijimai/"vancouverensis" clade (nodal support, -/0.96). The Bayesian tree (log L = −4371.60) was identical in topology to the ML tree.
The 28S dataset comprised 1065 bp aligned sites, with 333 variable sites, for 13 ingroup taxa. Most nodes in the ML tree (Appendix 1 - Supplementary Fig. S2) (log L = −3898.29) are resolved, and many have high nodal support. Phoronis emigi forms a clade with P. australis from New Caledonia with moderate to high nodal support (97/0.71). Phoronis australis appears as polyphyletic, with nominal "P. vancouverensis" comprising the sister taxon to a well-supported but polytomous clade containing P. australis from Australia and Japan, and P. muelleri. We did not obtain a 28S sequence for P. ijimai, which is thus missing from this analysis. The resulting Bayesian tree (log L = −4601.76) is topologically identical with the ML tree, but the clade containing P. emigi and New Caledonian P. australis is supported by lower Bayesian posterior probability (0.71).
The COI dataset comprised 621 bp aligned sites, with 253 variable sites, for 12 ingroup taxa (the tree was rooted with P. ovalis, which was the basal phoronid in all trees rooted with brachiopods). The resulting ML tree (Fig. 13B) (log L = −3633.85) is completely resolved, but with variable nodal support. The sister taxon to Phoronis emigi is P. architecta (= psammophila) rather than New Caledonian P. australis as in the 28S ML tree. The two P. australis samples inlcuded in the analysis form a clade with high support (96/1). Phoronis ijimai and nominal "P. vancouverensis" group together with high support (98/1), with this clade forming the sister group (nodal support, 59/0.84) to (Phoronopsis harmeri + Ph. viridis). Phoronopsis appeared polyphyletic, with Ph. californica the sister taxon to all other phoronids except P. ovalis. The resulting Bayesian tree (log L = −3772.71) was identical in topology to the ML tree.
The concatenated 18S-28S dataset comprised 2819 bp aligned sites, with 537 variable sites, for 13 ingroup taxa. The ML tree (Fig. 14A) (log L = −8247.64) was identical in topology to the 28S ML tree (Appendix 1 - Supplementary Fig. S2), except the unresolved trichotomy of AU and JP P. australis and P. muelleri in the latter is resolved in the 18S-28S tree. The Bayesian tree (log L = −9181.86) differs from the ML tree in that P. emigi forms a clade with P. hippocrepia, with New Caledonian P. australis the sister group to this clade.
The concatenated 18S-28S-COI dataset comprised 3440 bp aligned sites, with 555 variable sites, for 11 ingroup taxa (the tree was rooted with P. ovalis). The resulting ML Figure 13. A Maximum-likelihood tree for 15 phoronid samples based on 18S data; three brachiopod species (Novocrania anomala, Discinisca cf. tenuis, and Glottidia pyramidata) are included as outgroup taxa B maximum-likelihood tree for 12 phoronid samples based on COI data; the tree is rooted with Phoronis ovalis. The scale bars indicate branch length in substitutions per site. Nodal support values are presented as the ML bootstrap value followed by the Bayesian posterior probability; only values >50% and 0.50, respectively, are shown. tree ( Fig. 14B) (log L = −10594.85) differs from the 28S and 18S-28S trees in several ways. The sister taxon to P. emigi is P. hippocrepia (nodal support, 55/0.98) rather than New Caledonian P. australis. The positions of New Caledonian P. australis and P. architecta (= psammophila) are different in the 18S-28S-COI ML tree, but these changes in topology appear to some extent due to the omission of P. pallida from the 18S-28S-COI dataset. The topology within the "P. vancouverensis" / P. australis / P. muelleri clade also differs between 18S-28S-COI ML and the other trees that include 28S. The 18S-28S-COI Bayesian tree (log L = −10802.56), was identical to the ML tree in topology.

Discussion
Before our study, three species of phoronids had been recorded from Japan: Phoronis ijimai, P. australis, and P. psammophila. The former two were reported from Misaki (Oka 1897, Ikeda 1902, and the latter from Lake Hamana (Hirose et al. 2011). Phoronis ijimai was also reported from Akkeshi under the name P. hippocrepia (Uchida and Iwata 1955), but the taxonomic identity of this population is uncertain (Hirose et al. 2011). Bailey-Brock andEmig (2000) listed Tokyo Bay as a locality for P. pallida, with the note "coll. T. Furota", although they did not include any other details about the specimens. The known phoronid diversity in Japan thus remains low, with all specimens reported from sandy substratum. Investigations on rocky shores may yield additional species in the future.
Although the molecular phylogenetic trees (Figs 13A, 13B, 14A, 14B; Appendix 1 - Supplementary Fig. S2) produced by the various datasets differed in topology, our phylogenetic reconstructions suggest that most of the adult morphological characters Figure 14. A Maximum-likelihood tree for 13 phoronid samples based on the combined 18S + 28S data set; three brachiopod species (Novocrania anomala, Discinisca cf. tenuis, and Glottidia pyramidata) are included as outgroup taxa B maximum-likelihood tree for 11 phoronid samples based on the combined COI + 18S + 28S data set; the tree is rooted with P. ovalis. Scale bars indicate branch length in substitutions per site. Nodal support values are presented as the ML bootstrap value followed by the Bayesian posterior probability; only values >50% and 0.50, respectively, are shown. used to date in phoronid taxonomy are highly homoplastic (Fig. 15A-D), and thus phylogenetically less informative than the molecular data. According to the character matrix and the cladogram based on 32 morphological and reproductive characters among 11 phoronid species (Suppl. material 1; Fig. 12A, 12B; Appendix 1 -Supplementary Figs S1A, S1B, S3 A-D, S4 A-D), Phoronis emigi comprise a group with P. psammophila, P. muelleri, and P. pallida. In none of our molecular trees (Figs 13A, 13B, 14A, 14B), however, did these four species alone comprise a clade. In the COI tree (Fig. 13B), P. architecta (= psammophila), P. muelleri, and P. emigi comprise a clade that also includes P. hippocrepia. In the COI-18S-28S tree (Fig. 14B), P. emigi and P. architecta (= psammophila) group with P. hippocrepia, to the exclusion of P. muelleri, but no morphological or reproductive characters (Suppl. material 1; Fig.  15) appear to be synapomorphic for this clade, though character 19 (ratio of number of longitudinal muscles in oral coelom / anal coelom) in these three species is smaller than in other species of the genus except for P. ovalis, which lacks lateral mesenteries (Suppl. material 1).
Our molecular trees do not correspond with any of the subdivisions of phoronids suggested by previous researchers solely based on morphological characters (Silén 1952, Marsden 1959, Emig 1974. Within the phylum, Emig (1974) proposed five subgroups based on nephridial structure (Appendix 1 - Supplementary Fig. S5); most of these subgroups were identical to those in Silén's (1952) morphological categorization, except that Silén (1952) grouped P. psammophila with P. ijimai rather than P. muelleri. Although relationships within each group vary depending on the characters used in the analyses, our morphology-based cladograms ( Fig. 12; Appendix 1 -Supplementary Figs S1, S3, S4) mostly correspond Emig's (1974) subgroup relationships; therefore, Emig (1974) would have been classified P. emigi in his "group 3" along with P. psammophila Table 2. Pairwise genetic distances (K2P distances) based on 583 positions of COI sequences between P. ijimai, P. emigi, and the other species. The largest (P. australis JP and P. muelleri) and the lowest (P. australis NC and P. vancouverensis) interspecific distances are also listed. The analysis involved 12 phoronid sequences.  Supplementary Fig. S2), however, shows a clade comprising these three species alone. In the COI tree (Fig. 13B), these species form a clade that also includes P. hippocrepia. Our morphological and molecular results do not contradict that "P. vancouverensis" is conspecific with P. ijimai, as proposed by Emig (1971a). Although we were not able to obtain a 28S sequence for P. ijimai, in the 18S and COI trees it always formed a clade with "P. vancouverensis" accompanied by high nodal support (Fig. 13A, 13B). The Kimura (1980) 2-parameter (K2P) distance between P. ijimai and "P. vancouverensis" for 583 bp of COI was 0.07, substantially below the value of the intraspecific distance 0.115 between P. australis NC and P. australis JAPAN (Table 2). On the other hand, the interspecific distances among phoronids ranged from 0.164 to 0.287; therefore, K2P divergence factor between 0.115 and 0.164 could be a threshold for discriminating phoronid species.

Species
Taxonomic key to Japanese Phoronida