Integrated Taxonomy for Halistemma Species from the Northwest Pacific Ocean

During a survey of the siphonophore community in the Kuroshio Extension, Northwest Pacific Ocean, a new Halistemma Huxley, 1859 was described using integrated molecular and morphological approaches. The Halistemma isabu sp. nov. nectophore is most closely related morphologically to H. striata Totton, 1965 and H. maculatum Pugh and Baxter, 2014. These species can be differentiated by their nectosac shape, thrust block size, ectodermal cell patches and ridge patterns. The new species’ bracts are divided into two distinct types according to the number of teeth. Type A bracts are more closely related to ventral bracts in H. foliacea (Quoy and Gaimard, 1833) while Type B bracts are more similar to H. rubrum (Vogt, 1852). Each type differs, however, from the proximal end shape, distal process and bracteal canal. Both of the new species’ morphological type and phylogenetic position within the genus Halistemma are supported by phylogenetic analysis of concatenated DNA dataset (mtCOI, 16S rRNA and 18S rRNA). Integrated morphological and molecular approaches to the taxonomy of siphonophores showed a clear delimitation of the new species from the congeners. Halistemma isabu sp. nov. is distributed with the congeners H. rubrum, H. cupulifera, H. foliacea and H. striata in the northwestern Pacific Ocean.


Introduction
Siphonophores, belonging to the class Hydrozoa Owen, 1843, live in oceans and are mostly holoplanktonic organisms [1]. Approximately, 190 species have been described as siphonophores, though there are likely more in the marine environment [2]. Siphonophores have gelatinous bodies and form unique colonies throughout their life [3]. Each individual in a colony has a variety of functions and morphological characteristics [4]. Siphonophorae are traditionally classified into three suborders (Calycophorae Leuckart, 1854; Cystonectae Haeckel, 1887; Physonectae Haeckel, 1888) depending on the presence of a gas bubble float, called a pneumatophore, and medusoids, known as nectophores [1,5,6]. Approximately 65 species with pneumatophores and nectophores are recognized as Physonectae [7]. Physonects's descriptions are often incomplete because Siphonophores have fragile gelatinous bodies and many research records are based on dissociated specimens, where only a few nectophores or bracts were collected [1,8]. Morphological changes throughout their life cycle, phenotypic plasticity and cryptic species are also responsible for erroneous species diversity estimations. Molecular methods can be used for species identification, reconstructing life cycles, and clarifying phylogeny [2,[8][9][10][11][12][13].
Molecular data show that Cystonectae and Calycophorae are monophyletic taxa while Physonectae is polyphyletic [2,8]. Some families, such as Apolemiidae Huxley, 1859 and Pyrostephidae Moser, 1925, are not included in the clade unifying physonects and Calycophorae. Most physonects, however, can be combined into the monophyletic clade Euphysonectae. This includes Agalmatidae, Physophoridae

Sample Collections
Plankton samples were collected in October 2019 during a R/V ISABU cruise in the Kuroshio Extension ( Figure 1). We used a Multiple Opening/Closing Net and Environmental Sensing System (MOCNESS, 1 × 1 m 2 , mesh size: 200 µm) towed between 200 m and the surface. Immediately after towing, we split plankton samples into 1/2 aliquots using a Folsom plankton splitter. For morphological observation, one aliquot was fixed in neutralized 5% formalin solution and stored at room temperature. For future molecular analysis, we fixed the other aliquot in 99.9% ethanol and stored at 4 • C. All material was deposited at the National Institute of Biological Resources (NIBR), Korea's invertebrate collection.
In this study, we describe a new Halistemma species found in the Northwest Pacific Ocean. Our description uses morphological characteristics and a three-gene molecular approach.

Sample Collections
Plankton samples were collected in October 2019 during a R/V ISABU cruise in the Kuroshio Extension ( Figure 1). We used a Multiple Opening/Closing Net and Environmental Sensing System (MOCNESS, 1 × 1 m², mesh size: 200 µm) towed between 200 m and the surface. Immediately after towing, we split plankton samples into 1/2 aliquots using a Folsom plankton splitter. For morphological observation, one aliquot was fixed in neutralized 5% formalin solution and stored at room temperature. For future molecular analysis, we fixed the other aliquot in 99.9% ethanol and stored at 4 °C. All material was deposited at the National Institute of Biological Resources (NIBR), Korea's invertebrate collection.

Morphological Analysis
For morphological observation, specimens were sorted from 5% neutralized formalin plankton samples using a Live Insect Forceps (26029-10, Fine Science Tools Inc., Foster City, CA, USA) and a stereomicroscope (Olympus SZX7, Tokyo, Japan). We placed sorted specimens in 20 mL glass vials with the formalin solution at room temperature. We observed specimens on a Petri dish (Ø: 5 cm) filled with 5% neutralized formalin solution under the stereomicroscope. We identified all specimens using descriptions and illustrations from Kirkpatrick and Pugh (1984) [17], Mapstone (2009) [18], Totton (1954) [19], Totton and Bargmann (1965) [1] and Pugh and Baxter (2014) [16] and used Pugh and Baxter (2014) [16] terminology. Digital photographs of specimens were taken using a digital camera (Olympus PEN Lite E-PL3, Tokyo, Japan) connected to the stereomicroscope with side lights on the dark field. We took the digital photographs at various focal points and performed multi-focus stacking using Helicon Focus 7 software (Version: 7.5.1). We cropped the photographed objects and moved them to a black background using the Levels tool in Adobe Photoshop CS6 software (Version: 13.0). Measurements were taken using AxioVision Imaging System 4.8 software (Version: 4.8.1.0) and illustrations were drawn using a drawing tube on the stereomicroscope.

Species Delimitation and Phylogenetic Analysis
Finch TV software (Version: 1.4.0) read and cut sequences with low-quality signals, and once inspected, ClustalW [27] aligned forward and reverse sequences using default parameters embedded in Molecular Evolutionary Genetics Analysis 7 software (MEGA7, Version: 7.0.26). BLAST [28] analyses of the GenBank database were used to check that the sequences were Siphonophorae in origin and not contaminants. We acquired the mtCOI, 16S rRNA and 18S rRNA Halistemma sequences and other Siphonophorae sequences and used GenBank for the analysis. We constructed molecular analysis of the newly acquired sequences and the 142 Agalmatidae sequences (mtCOI: 37 and 16S rRNA: 19) and Siphonophorae (18S rRNA: 86) downloaded from GenBank (Table S1). The outgroup Physalia physalis (Linnaeus, 1758) was selected as belonging to Cystonectae (mtCOI and 16S rRNA) and the outgroup Neoturris breviconis (Murbach and Shearer, 1902) belonging to order Anthoathecata Cornelius, 1992 (18S rRNA). We calculated the genetic distances between alignment sequences using the Kimura 2-parameter (K2P) model [29] with complete deletion by MEGA7 software [30]. For Halistemma, we calculated the K2P genetic distances between the sequences and each of the three markers (mtCOI: 9, 16S rRNA: 10 and 18S rRNA: 6) (Tables S2-S4). In addition, we calculated K2P genetic distances for intraspecies, interspecies, intragenus and intergenus based on each of the three markers (mtCOI: 36, 16S rRNA: 19 and 18S rRNA: 13) within the Agalmatidae sensu stricto (Table S5). Maximum Likelihood (ML) and Bayesian Inference (BI) trees were drawn for phylogenetic analyses. The three single-gene ML phylogenetic trees' datasets included agalmatid sequences (mtCOI: 47, 16S rRNA: 26) and Siphonophorae (18S rRNA: 90). The concatenated BI and ML trees' datasets used 12 sequences from three concatenated Agalmatidae markers (mtCOI, 16S rRNA and 18S rRNA). To find the best-fit evolutionary model for the ML and BI analysis, we ran a Model Selection program using the MEGA7 software [30] and jModelTest software (Version: 2.1.7) [31,32] with the Akaike information criterion (AIC) [33][34][35]. Before creating the Nexus file, a FastGap (Version: 1.2) [36,37] was used to calibrate the Gap after alignment. The ML method used the best fit model proposed by Akaike information criterion with nearest neighbor interchange. This generated 1000 bootstrapping replicates for the phylogenetic tree reconstruction [38]. A BI tree was constructed using MrBayes software (Version: 3.2.6) [39][40][41] following likelihood parameters nst = 6 and rates = invgamma. We ran a Markov Chain Monte Carlo (MCMC) using the following parameters: ngen = 1,000,000, nchains = 4, samplefreq = 100, savebrlens = yes and printfreq = 1000. The BI trees were constructed using the "sump" command with burin = 250 to summarize parameters and the "sumt" command with burin = 250 to summarize the tree.

Measurements
See Table 1 for detailed measurements.

Nectophore Descriptions
Figures 2 and 3. The nectophore is an inverted bell-shape with axial wings and thrust block (Figure 2a,b). The axial wings have a petal-shape that ends sharply proximal outwards (Figure 2(a_1)). The sides of the nectophore are thick. The thrust block is high, big, thick and well-developed, rising above the axial wings ( Figure 2(b_1)). Basic ridge pattern includes lateral, vertical lateral, upper lateral and lower lateral ridges. Upper lateral ridges (Figure 2(c_1)) make a hook-shaped curve in the proximal part and down the slope to the medial line of the nectophore (Figure 2c). There are many folds along the curve (Figure 2(c_2)). Both upper lateral ridges approach one another, forming a Y-shape ( Figure 2(d_1)). In the proximal part of nectophore, before the Y-shape, there are many slanted folds facing to the center. These ridges continue pronounced in the nectophore's distal part towards the ostium level ( Figure 2(a_2)). They then curve out as slightly raised lines and become parallel to the upper margin of the ostium before petering out on the surface of the nectophore. The upper lateral ridge pairs are incomplete at both distal ends. The lateral ridges form a distinct line overhanging the nectophore's surface ( Figure 2(d_2)). The lateral ridges extend from near the ostium to the upper lateral ridges on both sides. Ostium regions of lateral ridges are located at the same level as the distal ends of upper lateral ridges. The proximal end of lateral ridges is split into three branches near the upper laterals, two of which are complete, and one is deficiently bent in a central direction ( Figure 2(c_3)). There are incomplete oblique ridges between the upper lateral ridges and the lateral ridges ( Figure 2(d_3)). There are long oval planes from the proximal apex of incomplete oblique ridges to the intersection point of Y-shape. The complete vertical lateral ridges run obliquely down from the upper lateral ridge to join with the lower lateral ridges. There are a pair of ectodermal cell patches on the triangular facet on each side between the upper and lower lateral and vertical lateral ridges of the nectophore. Some or all of these patches were destroyed on most nectophores. The lower lateral ridges pass along the lower lateral border of the nectophore. In the proximal part, the lower lateral ridges meet the upper lateral ridges in the axial wing. In the distal part of nectophore, the lower terminate distally on the lower lateral borders of the ostium (Figure 2(b_2)). The ostial regions are wider than the width of the mouth plate. The mouth plate is deeply divided, forming two round plates that overlap extensively in the mid-line. There are two stairs in nectosac towards the ostium (Figure 2(b_3)).
The nectosac surface has a scale pattern. The mantle canal starts at the internal pedicular canal's proximal end, the ascending branch of the mantle canal goes to the proximal direction to the trust block center (Figure 2(e_1)). The descending branch of the mantle canal goes in a distal direction toward the lower side before the nectosac level (Figure 2(e_2)). The ascending branch is about 2/3 of the total mantle canal. The descending branch of the mantle canal has a club at the proximal end. The pedicular canal arises from the mantle canals' intersection and goes down toward the central apex of the nectosac (Figure 2(e_3)). The stem attachment point is located in the proximal end of the internal pedicular canal (Figure 2(e_4)). The external pedicular canal was not preserved in the present material. At the central apex of the nectosac, four radial canals originate from the distal end of the internal pedicular canal, extend along the lateral surfaces of the nectosac. The upper canal passes straight down to the ostium level following the upper surface of the nectosac. The lower canal passes straight down to the ostium, descending vertically along the center of the nectophore's lower side. The lower canal has distal diverticula that go down to a point near the mouth plate and end in a club-shaped swelling. Lateral radial canals on the left and right sides are looped to the nectosac (Figure 2(a_3)).     Bracts are leaf-shape with a sharp proximal end and a round distal end. The proximal ends are thicker than the distal ends. Bracts are slightly curled in the lower direction. A number of uncertain and irregular wrinkles appear on the centerline. On the distal end, there are one or two lateral teeth and a distal process (Figure 4(a_1,2)). There are a pair of ectodermal cell patches in the middle of the bracts, whose size and position vary considerably. In many cases, these patches were damaged. The proximal part of the bracts has a concave neck and narrowed end. The bracteal canal is elongated in the longitudinal direction of bract (Figure 4(a_3)). The bracteal attachment lamella, about 2/3 length of the bract is attached on the lower surface along part of the mid-line below the bracteal canals (Figure 4(a_4)). The bracteal canal is clearer than the bracteal attachment lamella and becomes sharper as it goes to both ends. The bracteal attachment lamellas are often damaged. Bracts are leaf-shape with a sharp proximal end and a round distal end. The proximal ends are thicker than the distal ends. Bracts are slightly curled in the lower direction. A number of uncertain and irregular wrinkles appear on the centerline. On the distal end, there are one or two lateral teeth and a distal process (Figure 4(a_1,2)). There are a pair of ectodermal cell patches in the middle of the bracts, whose size and position vary considerably. In many cases, these patches were damaged. The proximal part of the bracts has a concave neck and narrowed end. The bracteal canal is elongated in the longitudinal direction of bract (Figure 4(a_3)). The bracteal attachment lamella, about 2/3 length of the bract is attached on the lower surface along part of the mid-line below the bracteal canals (Figure 4(a_4)). The bracteal canal is clearer than the bracteal attachment lamella and becomes sharper as it goes to both ends. The bracteal attachment lamellas are often damaged.
or two lateral teeth and a distal process (Figure 4(a_1,2)). There are a pair of ectodermal cell patches in the middle of the bracts, whose size and position vary considerably. In many cases, these patches were damaged. The proximal part of the bracts has a concave neck and narrowed end. The bracteal canal is elongated in the longitudinal direction of bract (Figure 4(a_3)). The bracteal attachment lamella, about 2/3 length of the bract is attached on the lower surface along part of the mid-line below the bracteal canals (Figure 4(a_4)). The bracteal canal is clearer than the bracteal attachment lamella and becomes sharper as it goes to both ends. The bracteal attachment lamellas are often damaged.    Although there are common features, bract shape is quite variable. There are ectodermal cell patches and lateral teeth for each specimen which we divided into two types (Type A: Figures 4a and 5a; Type B: Figures 4b-f and 5b-f). Type A has only one lateral tooth. It is broad and has a broad distally lateral indentation with an uneven outline near the distal process (Figure 5a). Type B has an elongated shape with a pair of lateral asymmetrical teeth in the distal part and distal process ( Figure  5b, d-f). These bracts have two oblique furrows on the upper surface that run from the lateral sides of the bracts towards the middle line in the distal direction and form oblique flaps. A small pointed proximal process is separated by a constriction. One tooth is usually round and the other tooth more acute. The edge line between the tooth and the distal process sometimes has an uneven outline including bumps, cusps and small indentations (Figure 5b). Bract 5C has one evident tooth and small hillock on the other side. This bract lacked oblique furrows. Although there are common features, bract shape is quite variable. There are ectodermal cell patches and lateral teeth for each specimen which we divided into two types (Type A : Figures 4a and 5a; Type B: Figures 4b-f and 5b-f). Type A has only one lateral tooth. It is broad and has a broad distally lateral indentation with an uneven outline near the distal process (Figure 5a). Type B has an elongated shape with a pair of lateral asymmetrical teeth in the distal part and distal process (Figure 5b,d-f). These bracts have two oblique furrows on the upper surface that run from the lateral sides of the bracts towards the middle line in the distal direction and form oblique flaps. A small pointed proximal process is separated by a constriction. One tooth is usually round and the other tooth more acute. The edge line between the tooth and the distal process sometimes has an uneven outline including bumps, cusps and small indentations (Figure 5b). Bract 5C lacked oblique furrows.

Molecular Phylogenetic Analysis
We obtained 15 new sequences of Halistemma isabu sp. nov.: six for mtCOI, six for 16S rRNA, and three for 18S rRNA (Table S1). We recovered the high-rank taxa phylogeny within Siphonophorae using a 18S rRNA-tree ( Figure S1). Monophyletic Cystonectae (ML = 98) is a sister clade to Codonophora (the clade including all other siphonophores, ML = 99) and monophyletic Calycophorae (ML = 93) placed within the paraphyletic Physonectae. However, relationships between genera and species were obscured within the 18S rRNA-tree because of most nodes' low support values. Therefore, other genetic markers were used to analyze Agalmatidae sensu stricto phylogeny. Other single-gene trees (mtCOI and 16S rRNA, Figures S2 and S3) and the concatenated tree ( Figure 6) produced congruent topologies; differences between trees are related to differences in taxon sampling. Node supports were lower in the single-gene trees while the concatenated dataset provided well-resolved and supported relationships. We recovered the genus Halistemma as monophyletic, but only two species were included in the concatenated dataset. H. isabu sp. nov. formed a clade with H. rubrum (Vogt, 1852) (PP = 0.95; ML = 88). Other unidentified Halistemma specimens from GenBank (GQ119989, GQ119991, KJ816934) were only included in either mtCOI or 16S rRNA datasets. They grouped with the H. rubrum and new species clade. Genus Halistemma was a sister clade to the group comprising Agalma and Athorybia. Genus Agalma was recovered as a paraphyletic group within the BI phylogenetic tree due to Athorybia rosacea (Forsskål, 1775) convergence with Agalma okenii Eschscholtz, 1825 (PP = 1). This paraphyly, however, was not supported within the ML phylogenetic tree. Finally, Nanomia was a sister taxon to the Halistemma, Agalma and Athorybia group.

Agalmatidae Sensu Stricto Genetic Diversity
The K2P genetic distances between the new species' specimens ranged from 0 to 0.002 using the mtCOI dataset and from 0 to 0.011 using the 16S rRNA dataset (Tables S2 and S3). The K2P genetic distances between Halistemma species, including the new species, ranged from 0.118 to 0.124 in mtCOI and from 0.113 to 0.147 in 16S rRNA (Tables S2 and S3). Interspecies K2P genetic distances in mtCOI ranged from 0.138 to 0.150 for Agalma and from 0.168 to 0.172 for Nanomia. Interspecies K2P genetic distances in 16S rRNA ranged from 0.059 to 0.108 for Agalma. The intergenus divergence in mtCOI was minimal (0.134) between Agalma-Halistemma and maximal (0.212) between (Halistemma-Nanomia).
The intergenus divergence in 16S rRNA was minimal (0.083) between (Agalma-Halistemma) and maximal (0.510) between (Agalma-Nanomia). The K2P genetic distances between the new species showed low values. It is usually difficult to match species based on nectophore and bract morphology (unless the entire colony is collected). However, the low difference (0.000-0.002) shows that they are the same species. The larger K2P distances between specimen from GenBank (H. sp., H. sp. 1, H. rubrum) and collected specimen for the mtCOI gene (0.118-0.124) and for 16S rRNA (0.113-0.147) are more typically seen between species. Within Hydrozoans, Zheng et al. (2014) [23] showed the mean congeneric K2P distance based on mtCOI sequence data ranged from 0.045 to 0.243, while K2P distance based on 16S rRNA sequence data ranged from 0.062 to 0.236. The congeneric K2P distances among other long-stem agalmatid species (Agalma, Halistemma, Nanomia) ranged from 0.098 to 0.172 for the mtCOI gene and from 0.059 to 0.149 for the 16S rRNA gene. These estimates cannot be considered conclusive due to the lack of GenBank data. We can, however, consider the genetic divergences seen between H. isabu sp. nov., and other Halistemma species. These are similar to interspecific distances among other agalmatids.

Agalmatid Family Phylogeny Based on Molecular Evidences
Our  [2] with higher support values using the concatenated sequences and three markers. With the exception of the concatenated ML Agalmatidae tree, the three trees, mtCOI, 16S rRNA (ML), and concatenated (BI), had exactly the same topology (Halistemma-Agalma-Athorybia composed a big monophyletic clade). However, the concatenated BI tree shows Agalma as a paraphyletic group with a high support value (PP = 1). In contrast, the concatenated ML tree supports Agalma as a monophyletic group (ML = 77). This suggests that Agalma and Athorybia's phylogenic position is unclear. Therefore, we excluded Athorybia from the genetic distances analysis. The single-gene tree also has limitations in revealing distinct topology. In order to get a clearer phylogenetic tree, we need to use additional markers while adding more data on the other Halistemma and Siphonophorae.

Morphological Comparisons of Halistemma Species
Pugh and Baxter (2014) [16] provided a diagnosis of the genus Halistemma. "Diagnosis: Agalmatid siphonophores with the characteristic arrangement of ridges on the nectophores; basically consisting of pairs of upper and lower lateral, lateral, and vertical lateral. Characteristic sinuous arrangement of the lateral radial canals on the nectosac. Adult bracts of at least two different types. In development, the colonies pass through a distinctive post-larval nectalia stage. Tentilla with single terminal filament, often ending in a cupulate process. Involucrum vestigial or covering only first one or two spirals of cnidoband." Halistemma species are different mainly in nectophore, bract and tentilla morphology [16]. Since tentilla were absent in the specimen we collected, we focused our attention on nectophore and bract morphology. We compared the new species with other congeneric species based on Pugh and Baxter (2014) [16]'s description (Table 2). Halistemma nectophores are different in vertical ridge number, ridge completeness, trust block size compared to axial wing size, and mouth plate presence. General nectophore shape varied from ovate-triangular in H. rubrum and H. transliratum Pugh and Youngbluth, 1988 to triangular with a step (T-shape) in H. cupulifera Lens and van Riemsdijk, 1908, H. striata Totton, 1965 The mantle canal's ascending branch is slightly longer than the descending branch in H. striata, and the ascending branch is approximate twice the descending branch length in H. transliratum and the new species. The radial canals are typical for all species. The upper and lower canals are straight, and the lateral canals have a similar looped pattern. In H. transliratum, the lateral radial canals are often irregular and occasionally small diverticula branch off at various places along their length. All radial canals fall into the ostial ring canal. In the new species, however, small diverticulum issue from the lower canal and runs distally within mouth plate.
Characteristics may differ depending on the specimen's maturity. Pugh and Baxter (2014) [16] studied young Halistemma species' nectophores. The trust block's size to axial wing ratio changes during the nectophore maturation. In young nectophores, the central thrust block is usually very flat and undeveloped so that the relatively large axial wings extend well beyond. During maturation, the upper lateral ridges' proximal part and axial wings diverge to the nectophores sides leaving room for the protruding trust block. The very young nectophoral bud clearly shows the ridge pattern in H. rubrum. The ridge pattern is usually present in young nectophores and does not change during maturation. The lateral ostial processes are prominent in young nectophores (in H. rubrum, H. foliacea) but become relatively small in mature nectophores. The mouth plate became visible in H. cupulifera, H, striata, H. foliacea, H. maculatum mature nectophores. We assumed that our collected specimen's characteristics, such as prominent trust block, lateral ostial process absence, and Y-shaped upper lateral ridge pair, represent mature nectophores.
The new species has unique morphological features. Lateral ridge ends are split into three branches near the upper lateral ridges, two of which are complete and one which is deficiently bent in a central direction. Before the upper lateral ridges go to the nectophore's median line, they go through a hook-shaped curve. An incomplete oblique ridge pair issues from below the upper lateral ridges' hook-shaped curve in the lateral ridge's direction.
We conclude that some of the new species' characteristics are similar to H. striata and H. maculatum. For example, the ridge pattern and mouth plate shape are similar to H. maculatum except for the lateral ridges' proximal part, which are divided to three branches with additional incomplete lateral ridges. The shape of nectosac, one pair of ectodermal patch and the upper lateral ridges' Y-shape brings it closer to H. striata. Stephanomia rubra "f" by Totton (1954) [19] from Indian Ocean has a step-shaped nectosac, a hook-shaped upper lateral ridge curve and a prominent trust block. All of these characteristics are similar to specimens examined in this study. Pugh and Baxter (2014) [16] described the similarities between Totton's specimen "f" with H. maculatum. However, H. maculatum's nectosac is different from specimen "f" and H. isabu sp. nov. Totton's specimen has typical lateral ridge proximal ends as in H. maculatum while H. isabu sp. nov. nectophores have three lateral ridge proximal end branchelets.
Bracts are very variable and change during maturation. Larval bracts were studied in H. rubrum, H. transliratum, and H. maculatum's nectalia-stage. These bracts are symmetrical and elongated with a prominent distal process. Otherwise, larval bract characteristics differ significantly from adult bract characteristics. There are two pairs of teeth in larval bracts of H. rubrum and H. maculatum. The proximal teeth pair is very close to the larval bract's proximal end. H. rubrum has two ectodermal cell patch pairs in their larval bracts compared to one pair in adult bracts. The H. maculatum ectodermal cell patches' number and position differed greatly between the larval and adult bracts. Larval bracts in H. transliratum are similar to adult bracts type "A" in their elongated shape, but the morphology is simplified in the larval type compared to the adult type. Species such as H. striata and H. foliacea has young bracts descriptions. Maturing and mature bract differences were not significant and predominantly concerned shape changes. We conclude that the new species' bracts are not similar to larval bracts because of their shape asymmetry, short distal process and complex morphology.
Bract classification does not yet have uniform principles. For example, bracts A and B in H. rubrum or A and C in H. cupulifera were classified based on their lateral teeth's shape and position. Shape is an important characteristic for H. foliacea bract classification, but bract classification for the rest of the species (H. cupulifera, H. striata, H. transliratum, H. maculatum) is based mainly on teeth number of ectodermal cell patch numbers. We found two bract types in Halistemma isabu sp. nov. Type A (Figures 4a and 5a) are similar to H. foliacea's ventral bracts because they have distally broad lateral indentations. They differed however, from H. foliacea's teeth pair, proximal end shape, and bracteal canal. Bracts B of H. isabu sp. nov. are similar to H. rubrum's bracts A. In both species, the bracts have a narrowed proximal end with a concave neck, one teeth pair and one ectodermal cell patch pair. However, the new species' bracts B are more concave and have uneven distal end contours. Unlike H. maculatum's bracts A, the new species' bracts B have a small distal process and only one ectodermal cell patch pair. sampling sites. Most Halistemma species' wide geographical distribution and their coexistence in the same localities require a more accurate analysis and their phylogenetic position should be checked using molecular methods.

Conclusions
In this study, we conducted a morphological and molecular analysis of H. isabu sp. nov. found in the Kuroshio Extension of Northwest Pacific Ocean. We provided information about the new species including digital photographs, illustrations, measurements, and DNA sequencing (mtCOI, 16S rRNA and 18S rRNA). We compared the new species morphologically to Halistemma congeneric species, and using a molecular approach, confirmed that the new species formed a monophyletic clade with other Halistemma species. We also calculated Agalmatidae sensu stricto genetic distances. These data and discussions are an important contribution to the controversial Agalmatidae taxonomy research.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4441/12/11/3283/s1, Table S1: Accession numbers of sequences used in this study's phylogenetic analyses, Table S2: K2P distances of mtCOI sequences between genus Halistemma from Genbank and this study, Table S3: K2P distances of 16S rRNA sequences between genus Halistemma from Genbank and this study, Table S4: K2P distances of 18S rRNA sequences between genus Halistemma from Genbank and this study, Table S5: Estimates of average K2P distances over sequence pairs within family Agalmatidae, based on mtCOI, 16S rRNA and 18S rRNA gene data, Figure  S1: Molecular phylogenetic tree of family Agalmatidae based on 18S rRNA data using the maximum likelihood method, Figure S2: Molecular phylogenetic tree of family Agalmatidae based on mtCOI data using the maximum likelihood method, Figure S3: Molecular phylogenetic tree of family Agalmatidae based on 16S rRNA data using the maximum likelihood method.
Author Contributions: All authors contributed to the study conception and design. Sample and data collection, N.P. and W.L.; Data analysis, N.P. and A.A.P.; writing-original draft preparation, N.P.; writing-review and editing, A.A.P. and W.L. All authors have read and agreed to the published version of the manuscript.