A New Species of the Deep-Sea Sponge-Associated Genus Eiconaxius (Crustacea: Decapoda: Axiidae), With New Insights Into the Distribution, Speciation, and Mitogenomic Phylogeny of Axiidean Shrimps

Eiconaxius Bate, 1888 is a genus of axiid shrimps exclusively associated with deep-sea hexactinellid sponges. Due to its special morphology and habitat, Eiconaxius is taxonomically and ecologically controversial. Based on material recently collected from seamounts in the northwestern Pacific, a new species of Eiconaxius is described. Intraspecific morphological and genetic variation and host specificity were evaluated. The complete mitochondrial genome of the new species was sequenced to explore the systematic status of Eiconaxius and some other axiidean taxa. Our analyses showed that differentiation of the new species occurs both allopatrically and sympatrically, probably resulting from the interaction of geographical isolation and deep water current movement, rather than from adaptation to different hosts. In addition, species of Eiconaxius are suggested to have wider ranges of distribution and host than expected. The reconstructed mitogenomic phylogeny supported merging Eiconaxius into Axiidae, and recognized most axiidean families, except that Strahlaxiidae was suggested to be paraphyletic. However, more comprehensive taxon sampling is still needed to resolve the explicit internal relationships among Axiidea.

Mitochondrial genomes have been widely used to resolve deeper phylogenetic relationships and are promising to settle systematic disputes. The typical metazoan mitochondrial genome is a closed circular molecule of 15-20 kb, usually comprising 37 genes, including 13 protein coding genes (PCGs), two ribosomal RNA (rRNA) genes, and 22 transfer RNA (tRNA) genes (Boore, 1999). Mitochondrial genomes include more genetic information than incomplete mitochondrial (e.g., COI, 12S rRNA, 16S rRNA) and nuclear genes (e.g., 18S rRNA, 28S rRNA, H3, NaK, PEPCK) and have the advantages of rather conserved gene content, easily accessible nature, and diverse evolutionary rates among different segments (Shen et al., 2017;Sun et al., 2018b;Li et al., 2019). Besides, the gene order and the RNA secondary structure can provide additional useful evolutionary information (Macey et al., 1997;Roehrdanz et al., 2002). As the technology and affordability of next-generation sequencing (NGS) matures, phylogenetic studies of decapods based on whole mitochondrial genome sequences have recently become popular (Tan et al., 2015(Tan et al., , 2017(Tan et al., , 2018a(Tan et al., ,b, 2019Basso et al., 2017;Cheng et al., 2018;Sun et al., 2018aSun et al., ,b, 2019a. Currently, mitochondrial genomes of 11 species of Axiidae from five families, Axiidae Huxley, 1879, Callianassidae Dana, 1852, Callichiridae Manning and Felder, 1991, Eucalliacidae Manning and Felder, 1991, and Strahlaxiidae Poore, 1994, are available. However, only a single representative of Axiidae Calocaris macandreae Bell, 1846 has been reported, and no mitochondrial genome of the genus Eiconaxius has been determined. This makes evaluation of the monophyly of Axiidae and the systematic status of Eiconaxiidae using mitochondrial genomes unachievable. During recent cruises to seamounts in the northwestern Pacific, conducted by the Second Institute of Oceanography, Ministry of Natural Resources (SIOMNR) and Institute of Oceanology, Chinese Academy of Sciences (IOCAS), five shrimps of the genus Eiconaxius and their sponge hosts were collected from Caiwei and Weijia Guyots in the Magellan Seamount Chain and from one unnamed seamount on the Caroline Ridge. This new material provides an opportunity to present: (1) descriptions of a new species of Eiconaxius; (2) assessment of the intraspecific divergence of the new species between populations using two mitochondrial (16S rRNA and COI) genes; (3) discussion of host specificity and the role it plays in the diversification of deep-sea commensal axiid shrimps; (4) evaluation of the systematic status of Eiconaxius and other families of Axiidea based on mitogenomic analyses. Our study provides new insights into the distribution, speciation, and phylogeny of axiidean shrimps and other deepsea crustaceans.

Specimen Collection
Three specimens of Eiconaxius were collected during survey cruises near Caiwei Guyot, Magellan Seamount Chain in 2014 (one ovigerous female), and near Weijia Guyot, Magellan Seamount Chain in 2016 (one ovigerous female, one mature male), using RV Xiangyanghong 9. These three axiid shrimps and their sponge hosts were captured by the Chinese manned submersible Jiaolong. Two more ovigerous females were collected during a survey cruise from two sites on an unnamed seamount on the Caroline Ridge in 2019 using the RV Kexue. These two and their sponge hosts were captured by the remotely operated vehicle (ROV) Faxian. All specimens were immediately fixed and preserved in 75% ethanol after being photographed on board. When the specimens were unloaded and carried to the laboratory, fresh 75% ethanol was replaced. The specimens are deposited in the Sample Repository of the Second Institute of Oceanography (SRSIO), Ministry of Natural Resources, Hangzhou, China, and the Marine Biological Museum, Chinese Academy of Sciences (MBMCAS), Qingdao, China.
To resolve the taxonomic status of the undescribed Eiconaxius and to explore interspecific genetic divergence, tissue of nine other species of Eiconaxius was sampled from collections in Museums Victoria, Melbourne, Australia (NMV), and National Institute of Water and Atmospheric Research, Wellington, New Zealand (NIWA). One individual from another deep-sea sponge associated axiid species, Spongiaxius novaezealandiae (Borradaile, 1916), was included in the 16S rRNA dataset, with an aligned length of 559 bp, and in the COI dataset with an aligned length of 591 bp ( Table 1). The collection locations of all the axiid specimens used in the present study are shown in Figure 11A.

Morphology Observation
On board, the shrimps and their sponge hosts were photographed using a Canon EOS-1D Digital Single Lens Reflex camera. In the laboratory, the specimens were measured and illustrated under a Zeiss SteREO Discovery V8 stereomicroscope. Carapace length (cl) was measured from the tip of the rostrum to the posterior end of the carapace. Total length (tl) was measured from the tip of the rostrum to the posterior end of the telson, with body stretched. The diagnosis was derived from the same DELTA database (Dallwitz, 2018) of species and characters of Eiconaxius as that used by Poore (in press).

DNA Extraction, PCR Amplification, and Sanger Sequencing
Total genomic DNA of the specimens was extracted from a small piece of muscle tissue (5∼10 mg). DNA was extracted using a QIAamp DNA Micro Kit (Qiagen, Hilden, Germany) and then eluted in 50 µL of sterile distilled H 2 O (RNase free), and stored at −20 • C. Polymerase chain reaction (PCR) amplification was carried out in a reaction mix containing 5 µL of template DNA, 25 µL of Premix Taq TM (Takara, Otsu, Shiga, Japan), 1 µL of each primer (10 mM), and sterile distilled H 2 O to a total volume of 50 µL. Mitochondrial 16S rRNA and COI genes were amplified using the primers 16S-ar/br (Simon et al., 1994) and LCO1490/HCO2198 (Folmer et al., 1994), respectively, with the following thermal profile: initial denaturation for 3 min at 94 • C, followed by 35-40 cycles of denaturation at 94 • C for 30 s, annealing at 48 • C for 40 s, extension at 72 • C for 30 s, and a final extension at 72 • C for 10 min. PCR products were purified using the Wizard TM SV Gel and PCR Clean-UP System (Promega, Madison, WI, United States) before sequencing. The purified PCR products were bidirectionally sequenced using the same forward and reverse primers for PCR amplification with ABI 3730XL DNA Analyzer (Applied Biosystems, Foster City, CA, United States).

Next-Generation Sequencing, Mitochondrial Genome Assembly, and Annotation
One µg of purified total genomic DNA of the holotype was fragmented and used to construct a paired-end library (insert size 300∼500 bp) using TruSeq TM Nano DNA Sample Prep Kit (Illumina, United States). The library was sequenced on the Illumina Hiseq 4000 platform (2 × 150 bp paired-end reads) by BIOZERON Co., Ltd (Shanghai, China). Prior to assembly, raw reads were filtered to remove the reads with adaptors, the reads with a quality score below 20 (Q < 20), the reads containing a percentage of uncalled based ("N" characters) equal or greater than 10% and the duplicated sequences. Then the mitochondrial genome was reconstructed using a combination of de novo and reference-guided assemblies, and the following three steps were used to assemble mitochondria genomes. First, the filtered reads were assembled into contigs using SOAPdenovo 2.04 (Luo et al., 2012). Second, contigs were aligned to the reference mitochondrial genome of Calocaris macandreae (KC107812) using BLAST, and aligned contigs (≥80% similarity and query coverage) were ordered according to the reference genome. Third, clean reads were mapped to the assembled draft mitochondrial genome to correct the wrong bases, and the majority of gaps were filled through local assembly.
The mitochondrial genome was annotated using the MITOS2 webserver (Bernt et al., 2013). Locations and sizes of the proteincoding genes (PCGs) were identified by Open Reading Frame Finder (ORFfinder) available on NCBI with the invertebrate mitochondrial code. Transfer RNA (tRNA) genes were identified by tRNAscan-SE 2.0 webserver (Chan and Lowe, 2019), and their secondary structures were predicted and visualized using Forna (Kerpedjiev et al., 2015). Ribosome RNA (rRNA) genes were delineated by rRNAmmer 1.2 webserver (Lagesen et al., 2007). All the gene predictions were reconfirmed by comparing nucleotide or amino acid sequences with those of published mitochondrial genomes of Axiidae using the Basic Local Alignment Search Tool (BLAST). The frequencies of both amino acids and codons, and the relative synonymous codon usage (RSCU), were calculated using MEGA 6.06 (Tamura et al., 2013). The assembly of the mitochondrial genome was verified by comparison with the 16S rRNA and COI sequence obtained from the foregoing Sanger sequencing. The circular mitochondrial genome map of the new axiid species was generated by CGView Server (Grant and Stothard, 2008).

Phylogenetic Analysis
For the 16S rRNA and COI gene sequences obtained from Sanger sequencing, sequence chromatograms were checked using CHROMAS 2.23 (Technelysium Pty Ltd) by eye. The forward and reverse sequence fragments were assembled by CONTIG EXPRESS (a component of Vector NTI Suite 6.0, Life Technologies, Carlsbad, CA, United States). The homologous sequences were aligned by MAFFT version 7 webserver (Katoh et al., 2019) with default parameters, and manually trimmed to the same length for all the taxa. The Kimura's 2-parameter genetic distances were calculated using MEGA 6.06 (Tamura et al., 2013).
Phylogeny was reconstructed based on the mitochondrial genomes of the new species and those of 59 decapod species belonging to two suborders, 10 infraorders, with three species of Stomatopoda and one species of Euphausiacea as outgroups (Supplementary Table S1). The nucleotide sequences of 13 PCGs were aligned by MAFFT version 7 webserver (Katoh et al., 2019) with default parameters. GBlocks v0.91b (Castresana, 2000) was used to eliminate the highly divergent and poorly aligned segments of each gene (GBlocks parameters: minimum length of a block = 5; allowed gap positions = with half). Then the trimmed alignments were concatenated into a single dataset consisting of 13 PCGs using Sequence Matrix 1.8 (Vaidya et al., 2011), and each gene was treated as separate data partition in the subsequent analyses. Phylogenetic relationships were inferred from the concatenated dataset using both maximum likelihood (ML) and Bayesian inference (BI) methods. For ML analyses, the best-fit substitution models (including FreeRate heterogeneity models) and partition schemes were inferred by ModelFinder (Kalyaanamoorthy et al., 2017) implemented in IQ-TREE 1.6.10 (Nguyen et al., 2015). The ML tree was reconstructed using IQ-TREE, and branch support was assessed by performing Bayesian-like transformation of aLRT (aBayes) test (Anisimova et al., 2011) as well as ultrafast bootstrap (BP) with 1,000 replicates (Hoang et al., 2018). For BI analyses, the best-fit substitution models and partition schemes were inferred by PartitionFinder 2.1.1 (Lanfear et al., 2017) with the 'greedy' algorithm according to the Bayesian information criterion. Bayesian analysis was conducted using MrBayes 3.2.7a (Ronquist et al., 2012). Two independent runs were carried out with four Markov Chains for 20,000,000 generations started from a random tree, with sampling every 1,000 generations. The average standard deviation of split frequencies and the likelihood values were monitored, to determine whether the two runs converged onto the stationary distribution. The first 25% (5,000) trees generated prior to the achievement of stationarity of the log-likelihood values were discarded as burn-in. The remaining trees were used to construct the 50% majority rule consensus tree and to estimate the posterior probabilities (PP). The effective sample size (ESS) values for all sampled parameters were diagnosed by Tracer 1.7.1 (Rambaut et al., 2018) to ensure convergence. The phylogenetic trees and node labels were visualized using FigTree 1.4.3 (Rambaut, 2016). Finally, all newly obtained 16 rRNA, COI, and mitochondrial genome sequences were submitted to the GenBank database.

Gene Order Analysis
We mapped all mitochondrial gene orders on to the phylogeny for comparison. Additionally, the putative ancestral state of the pancrustacean ground pattern and the mitochondrial genome order of the new species were pairwise compared to predict the mitochondrial genome rearrangement events (e.g., gene reversals, transpositions, reverse transpositions, tandem duplication random loss [TDRL]) using Common interval Rearrangement Explorer, heuristically exploring mitochondrial rearrangements based on common intervals (CREx; Bernt et al., 2007).

Remarks
For the most recent diagnosis of Eiconaxius see Poore (2017).

Diagnosis
Rostrum tapering evenly, 1.3 times as long as wide. Submedian gastric carinae U-shaped, diverging widely from base of median carina. Major cheliped, merus lower margin with 2 prominent teeth in distal half; palm wider distally than at midpoint; palm upper margin denticulate; fixed finger cutting edge with notch and blunt tooth in proximal half; dactylus cutting edge with basal molar, notch and straight beyond. Minor cheliped, palm upper margin denticulate, fingers almost as long to longer than upper margin of palm; fixed finger cutting edge denticulate.

Description of Female Holotype
Body robust, integument solid, surface generally glabrous. Rostrum triangular in dorsal view, 0.15 times as long as carapace, 1.25 times as long as wide at base, apex acute, slightly directed upwards, overreaching antennular peduncle article 2; lateral margins each with 5 small, shallow teeth, continuous with lateral carinae; ventral margin unarmed. Carapace glabrous, with gastric region inflated; cervical groove faint; median carina shallow, entire, unarmed, reaching middle of rostrum; submedian gastric carinae U-shaped, diverging widely from base of median carina, short; lateral carina unarmed, diverging posteriorly, extending over anterior 0.1 of carapace length; pterygostomian angle rounded, with five obsolete marginal denticles.
Telson sub-oval, 0.75 times as long as maximum width at midlength, lateral margin arcuate, each side armed with 12 uneven teeth, posterolateral angle obtuse, defined by 2 small teeth, posterior margin truncate, with median tooth; dorsal surface concave, with scattered long simple setae, posterolateral and posterior margins setose.
Eyestalk well developed, not reaching middle of rostrum; cornea globular, without pigment. Antennular peduncle article 2 half-length of article 1, almost reaching end of rostrum; articles 2, 3 subequal in length, 1.2 times as long as wide; flagella 1.5 times as long as carapace. Antennal peduncle article 1 unarmed; article 2 with blunt distal tooth on ventral surface, triangular blade elongate, acute, reaching distal margin of article 4; article 3 with acute ventromesial tooth; article 4 4.0 times as long as wide; article 5 half-length of article 4; scaphocerite slender, acute, reaching middle of article 5; flagellum 0.7 times as long as carapace.
Minor (left) cheliped shorter, more slender than major, 1.4 times as long as carapace, palm 0.6 times width of major; coxa with distal tooth on mesial margin; basis unarmed; ischium compressed laterally, 1.2 times as long as broad, with prominent triangular tooth, 4 denticles on lower margin, upper margin unarmed; merus 2.4 times as long as ischium, compressed laterally, upper margin slightly arcuate, unarmed, lower margin with 3 small teeth at mid-length; carpus 0.6 times as long as merus, lower distal angle unarmed; palm slightly inflated, subequal in length to merus, upper margin 1.2 times as long as wide, sharply carinate, with 8 subequal obsolete teeth, lateral and mesial surfaces with few scattered tubercles in distal half, lower lateral carina sharp, extending to tip of fixed finger, with row of 8 teeth proximally, distolateral and distomesial margins oblique, unornamented; fingers forming deep sub-triangular cavity defined by sharp longitudinal carina on mesial surface when closed; fixed finger as long as palm, nearly straight, distally slightly upturned, opposable margin with row of small triangular teeth along whole length; dactylus 1.2 times as long as palm, distally curved, upper margin sharply carinate, unarmed, lateral surface with blunt longitudinal carina on midline, opposable margin with several tiny denticles in distal half, small proximal notch.
Uropod as long as telson; peduncle stout, unarmed; endopod 2.3 times as long as broad, lateral margin armed with 8 (left), 10 (right) teeth over distal half, distal angle slightly produced, posterior margin straight, dorsal surface with faint longitudinal ridge and scattered long simple setae; exopod 1.7 times as long as broad, shorter than endopod, lateral margin armed with row of 18 (left), 16 (right) teeth over distal two-thirds, distal angle bifid or almost so, posterior margin almost straight, distally strongly convex, dorsal surface with faint longitudinal ridge; exopod and endopod both with long plumose setae along mesial and posterodistal margins.

Non-type ovigerous female (MBM 304668)
Generally similar to holotype. Rostrum slightly directed upwards, reaching to middle of antennular peduncle article 2, lateral margins each with 2 small, shallow teeth. Telson lateral margin armed with 9 (left), 12 (right) teeth subequal in size. Major (right) cheliped ischium lower margin with 2 teeth, upper margin unarmed; merus lower margin with 2 sharp teeth, upper margin with small distal tooth and midlength tooth; carpus lower margin with distal tooth, upper margin unarmed; palm upper margin with 7 sharp subequal teeth, lower-lateral carina with row of 6 teeth proximally, distolateral and distomesial margins unarmed; dactylus upper margin with tooth at proximal quarter. Minor (left) cheliped ischium lower margin with tooth, upper margin unarmed; merus lower margin with 2 sharp teeth, upper margin with 2 small teeth distal to midlength; carpus lower margin with distal tooth, upper margin unarmed; palm with 5 sharp subequal teeth, lower-lateral carina with row of 5 teeth proximally, distolateral margin with prominent bifid triangular tooth and small additional tooth, distomesial margin with 3 small teeth; fixed finger with lower mesial ridge unarmed; dactylus upper margin unarmed.

Non-type ovigerous female (MBM 304669)
Generally similar to holotype. Rostrum slightly directed upwards, reaching to middle of antennular peduncle article 2, lateral margins each with 2 small, shallow teeth. Telson lateral margin armed with 11 (left), 13 (right) teeth subequal in size. Major (left) cheliped ischium lower margin with 3 teeth, upper margin unarmed; merus lower margin with 3 sharp teeth, upper margin with small distal tooth and midlength tooth; carpus lower margin with distal tooth, upper margin unarmed; palm upper margin with 7 sharp subequal teeth, lower-lateral carina with row of 5 teeth proximally, distolateral and distomesial margins unarmed; dactylus upper margin with tooth at proximal third. Minor (right) cheliped ischium lower margin with 3 teeth, upper margin unarmed; merus lower margin with 3 sharp teeth, upper margin with 2 small teeth distal to midlength; carpus lower margin with distal tooth, upper margin unarmed; palm with 6 sharp subequal teeth, lower-lateral carina with row of 5 teeth proximally, distolateral margin with prominent bifid triangular tooth and small additional tooth, distomesial margin with 3 small teeth; fixed finger with lower mesial ridge unarmed; dactylus upper margin with tooth at half length.

Color in Life
Body and appendages orangish translucent, tips of rostrum and chelipeds darker; cornea of eye faint yellow, opaque; mature female with ovary light blue; embryos sapphire at early embryonic stage, whitish at late embryonic stage.

Host
The specimen from Caiwei Guyot was found in the cavity of a hexactinellid sponge belonging to the subfamily Corbitellinae (Lyssacinosida, Euplectellidae) ( Figure 6E); the specimens from Weijia Guyot were found in the cavity of the hexactinellid sponge Amphidiscella sp. (Lyssacinosida, Euplectellidae, Bolosominae) ( Figure 6F), and the specimens from an unnamed seamount on the Caroline Ridge were found in the cavity of the hexactinellid sponge Farrea sp. (Sceptrulophora, Farreidae) (Figures 6G,H).

Etymology
From Latin serratus, meaning serrated, referring to the strongly serrated upper margin of the cheliped palm.

Organization and Characterization of Mitochondrial Genome
A total of 51,653,517 clean reads (7,578,946,749 bp) were generated by Illumina HiSeq sequencing with an insert size of approximate 450 bp. After assembling, a 16,195 bp circular molecule was obtained (Figure 7), which represented the FIGURE 7 | The organization of the mitochondrial genome of Eiconaxius serratus sp. nov. The full names of protein coding genes, rrnS and rrnL, are listed under abbreviations rrnS and rrnL, 12S and 16S ribosomal RNA genes, respectively; atp6 and atp8, ATPase subunit 6 and 8 genes, respectively; cox1-cox3, cytochrome c oxidase subunits I-III genes, respectively; cob, cytochrome b gene; nad1-6 and 4l, NADH dehydrogenase subunit 1-6 and 4l genes, respectively. One uppercase letter amino acid abbreviations are used to label the corresponding tRNA genes. The position of control region (CR) is indicated in the figure.  Table S1).
The complete mitochondrial genome encodes 37 genes, including 13 PCGs, 2 rRNA genes, and 22 tRNA genes (duplication of trn L and trn S ), which exhibits the same components as most other decapod mitochondrial genomes. Twenty-four genes (9 PCGs and 15 tRNAs) are encoded on the heavy (H) strand, while other 13 genes (4 PCGs, 7 tRNAs, and 2 rRNAs) are encoded on the light (L) strand ( Table 3). The base composition of the heavy strand is A = 34.69%, T = 38.01%, C = 14.26%, and G = 13.04%. The A + T content (72.70%) is distinctly higher than the G + C content (27.30%). A total of 1,614 bp of non-coding nucleotides are scattered among 19 intergenic regions varying from 2 to 1,035 bp ( Table 3). The largest non-coding region (1,035 bp) located between trn I and trn Q is identified as the putative control region (CR) according to its location in the mitochondrial genome (Figure 7). Furthermore, there are 11 overlaps between adjacent genes in the new species with a size range of 2 to 20 bp ( Table 3). The combined length of 13 PCGs was 11,148 bp, accounting for 68.83% of the entire mitochondrial genome. All PCGs started with ATD as initiation codons (9 with ATG, 2 with ATT, and 2 with ATA) and end with two conventional stop codons (TAA and TAG) ( Table 3). Among 13 PCGs, Leu (15.31%) and Cys (1.11%) are the most and the least frequently used amino acids, respectively (Supplementary Figure S1A). RSCU analysis shows that UUA (Leu, 3.80%) is the most and CGC (Arg, 0.00%) the least frequently used codons. In addition, NNW codons have a higher abundance than NNS codons (Supplementary Figure S1B). The lengths of 22 tRNA genes range from 62 to 72 bp, and all tRNA genes can be folded into classic clover leaf structures (Supplementary Figure S2). Two ribosomal RNA genes (rrnS and rrnL) are located on the L strand between nad1 and trn I , with lengths of 794 bp and 1247 bp, respectively (Figure 7 and Table 3).

Phylogenetic Analysis
The final concatenated dataset consisted of 10,629 bp (∼95.34% of the original 11,148 bp alignment) after the poorly aligned positions and the hypervariable regions were removed with GBlocks. We found that the BIC scores of edge-linked partition scheme were always better than that of edge-unlinked partition scheme. Therefore, we chose the best-fit substitution models and partition schemes selected by edge-linked partition scheme findings for subsequent phylogenetic analyses (Table 4). Tree topologies resulting from the BI and ML analyses were highly congruent and generally well supported, except for a few internal nodes in the clades of Caridea, Axiidae, and Brachyura (Figure 8 and Supplementary Figure S3). At the order level, both Stomatopoda and Decapoda were found to be monophyletic with high support values (BP = 100%, aBayes = 1.00, PP = 1.00). At the suborder level, both Dendrobranchiata and Pleocyemata were monophyletic (BP = 100%, aBayes = 1.00, PP = 1.00) and Dendrobranchiata located at the basal position of decapods with strong support (BP = 98%, aBayes = 1.00, PP = 1.00). At the infraorder level, the monophyly of each infraorder was consistently well supported in all analyses (BP ≥ 95%, aBayes ≥ 0.95, PP ≥ 0.95). Furthermore, apart from the grouping of Polychelida and Glypheidea (BP = 50%, aBayes = 0.63, PP = 0.63) and the position of Stenopodidea (BP = 83%, aBayes = 0.99, PP = 0.88), the relationships between infraorders were mostly well resolved. Although this study was not intended to investigate the familial relationships within decapod infraorders, these too were generally well supported in both analyses. Axiidae (represented by two genera) were monophyletic (BP = 89%, aBayes = 1.00, PP = 1.00) and basally positioned within Axiidea in all analyses with high support values (BP = 100%, aBayes = 1.00, PP = 1.00). Strahlaxiidae were paraphyletic, Neaxius and Strahlaxius basal to a clade consisting of Eucalliacidae, Callichiridae, and Callianassidae (BP = 100%, aBayes = 1.00, PP = 1.00). The relationships among the last three families were obscure, as the topologies derived from ML and BI analyses were different yet well supported. In the ML tree, Eucalliacidae and Callichiridae (represented by one species each) formed a sister group to Callianassidae with high support values (BP = 90%, aBayes = 1.00). While in the BI tree, Callichiridae and Callianassidae grouped together first (PP = 1.00), and then clustered with Eucalliacidae (PP = 1.00). Callianassidae (five species) were always recognized as a monophyletic group with high support values (BP = 100%, aBayes = 1.00, PP = 1.00) in our analyses.

Mitochondrial Gene Order and Rearrangements
The gene orders of mitochondrial genomes were mapped onto the phylogeny based on analyses of 13 PCG nucleotide sequences (Figure 9). Within Axiidea, four unique gene arrangements were identified. Among them, the mitochondrial gene order of Eiconaxius serratus sp. nov. was identical to that of the other axiid species Calocaris macandreae.
Compared with the pancrustacean ground pattern, at least five genes were rearranged in the mitochondrial genome of E. serratus sp. nov. The trn L (CUN) (trn L1 ), which is located between nad1 and rrnL in the mitochondrial genomes of more primitive taxa, was found between trn L (UUR) (trn L2 ) and cox2 in E. serratus sp. nov. Besides, the rearrangement of trn L1 also involved shifting between two strands, which can find parallels in Stenopodidea and Anomura. Similarly, the trn I relocated between rrnS and CR, and the trn V moved to upstream of trn G , both with a reversal. In addition, the trn D moved to upstream of trn M , while the PCG cox3 relocated between nad3 and trn A .
Based on the analysis of CREx, two alternative rearrangement scenarios were inferred, from the putative ancestral state of the pancrustacean ground pattern to E. serratus sp. nov., as a result of successive events of transposition/reverse transposition, reversal, and TDRL (Figure 10).

Morphological Differences
Eiconaxius serratus sp. nov. is uniquely diagnosed by the combination of a relatively narrow acute rostrum and two prominent distal teeth on the lower margin of the merus of the major cheliped. Many species have a similar rostrum but no others have such prominent meral teeth. Most species have a serrated meral lower margin while a few have one or two teeth or tubercles just beyond the midpoint. E. spinigera (MacGilchrist, 1905), E. rubrirostris Komai et al., 2010, andE. albatrossae Kensley, 1996 are similar (differing in having much smaller meral teeth). While the upper margin of the chelipeds in several species are obscurely denticulate, none has the strong erect teeth possessed by E. serratus sp. nov.
Of the nine species of Eiconaxius for which 16S and COI genes are available, E. serratus sp. nov. has the least genetic divergence from E. caribbaeus (Faxon, 1896) (Table 2). E. caribbaeus differs in having a rounded rostrum, weakly serrated cheliped meral margin, and without strong tooth on the upper margin of the palm of the minor cheliped. We noted several obsolete marginal denticles along the pterygostomian angle in E. serratus sp. nov., a feature easily missed by earlier authors.

Distribution and Differentiation
The 16S rRNA genetic divergence between specimens of Eiconaxius serratus sp. nov. from Magellan Seamount Chain and the seamount on the Caroline Ridge was 0.5%. Although 16S rRNA tends to have slower rates of substitution than other mitochondrial genes , normally a genetic divergence of less than 1% in the 16S rRNA gene indicates conspecificity for decapod crustaceans (Cabezas et al., 2009;Matzen da Silva et al., 2011;Lavery et al., 2014). Besides, this value was significantly lower than the averaged intrageneric genetic divergence (9.5%) of Eiconaxius. We recognize all individuals as belonging to the same species.
It is notable that high COI intraspecific divergence of E. serratus sp. nov. was observed. The divergence between the specimens from Magellan Seamount Chain and the unnamed seamount on the Caroline Ridge ranged from 8.3 to 9.3% (Table 2), indicating low connectivity between the two populations. Because adults of Eiconaxius are restricted to the cavity of deep-sea hexactinellid sponges, gene flow between populations must be mediated through larval dispersal rather than migration of adults. For one thing, the long distance (>1700 km) across the East Mariana Basin is a barrier leading to potential geographical isolation. For another, two deep-water currents in the Northwest Pacific could play an important role directing larval dispersal. The Northwest Pacific (5 • -15 • N) contains two main westward oxygen-rich water currents at 2000-3000 m depth (Kawabe et al., 2003;Kawabe and Fujio, 2010) ( Figure 11B). The southern current which flows westwards south of the Carolina Seamounts toward the Yap Trench does not concern us. The northern one bifurcates at 150 • E, just north of the Caroline Seamounts, its major branch flowing through the vicinity of the Challenger Deep and proceeding westward into the West Mariana Basin, while the minor branch swerves along west of the East Mariana Basin and flows eastward near the Magellan Seamounts. Consequently, the two current branches flowing in opposite directions could confine larval dispersal, promoting the divergence of E. serratus sp. nov. populations. For the same reason, the low divergence between the two populations from the Magellan Seamount Chain (0.2% for COI) might because the sampled regions are influenced by same branch. Interestingly, the specimens collected from two very close sites on the unnamed seamount on the Caroline Ridge showed significant COI genetic divergence (7.2%). Because these two specimens came from similar depths and hosts, neither bathymetric segregation nor host shift explains their divergence but could result from their separation at two sites on opposite flanks of the steep seamount (summit < 800 m). We speculate the seamount itself between them is a physical barrier preventing larval dispersal and connectivity.
In summary, we can confirm that this species of Eiconaxius ranges over 1500 km despite genetic discontinuity. The interaction of geographical isolation and deep-water currents contribute to this intraspecific differentiation. The high COI intraspecific divergence between specimens of E. serratus sp. nov. from two very close sites is consistent with sympatric speciation in Eiconaxius. Poore (2018) recorded four species from the eastern Caribbean Sea, and Poore (in press) recorded several species from limited areas in the southwest Pacific, even two species from the same dredge sample. While most species have been described from few specimens from small areas, others FIGURE 11 | (A) Collection locations of the axiid specimens in the present study; (B) collection locations of Eiconaxius serratus sp. nov., with schematic diagram of the deep water currents at low latitudes in the West Pacific in the upper deep layer (marked in red arrows, from Kawabe et al., 2003;Kawabe and Fujio, 2010).
Hitherto, little has been known about host specificity of species of Eiconaxius (Ortmann, 1891;Faxon, 1893Faxon, , 1896Bouvier, 1925;De Man, 1925;Kensley, 1996;Komai, 2011;Komai and Tsuchida, 2012;Poore and Dworschak, 2018). The reasons for this are various. Specimens are often collected in large numbers without their hosts, presumably because their sponge homes have been destroyed by destructive dredge sampling. Shrimps found inside hexactinellids have been identified by decapod taxonomists who have not consulted sponge taxonomists for precise identification of the host. Encouragingly, recent collections by deep-sea submersibles and ROVs have provided a better picture of the relationship between deep-sea decapods and their sponge hosts (Saito et al., 2006;Komai and Tsuchida, 2012;Komai, 2013;Jiang et al., 2015;Komai et al., 2016;Dworschak, 2016;Xu et al., 2016Xu et al., , 2017Kou et al., 2018;Poore and Dworschak, 2018). Komai and Tsuchida (2012) found Eiconaxius acutifrons Bate, 1888 associated with sponges of the family Farreidae Gray, 1872 (Hexactinellida, Sceptrulophora). Interestingly, two males were found inside the internal cavity of the same sponge. In our study, submersible and ROV successfully collected intact sponge hosts of E. serratus sp. nov. belonging to two orders (Lyssacinosida, Sceptrulophora) and at least three genera, indicating E. serratus sp. nov. has low host specificity. The diversity and number of sponges decrease significantly with greater depth, narrowing the choice of hosts for Eiconaxius. The two populations from the Magellan Seamount Chain with low genetic divergence have distinct hosts while the two populations from the Caroline Ridge with high genetic divergence share the same host. We postulate that species of Eiconaxius could associate with a variety of sponges, or even other organisms. Accordingly, we consider host shift may not play a key role in speciation of Eiconaxius, as was found for some spongicolid shrimps (Saito and Komai, 2008;Goy, 2010;Kou et al., 2018).

Mitogenomic Phylogeny
On the whole, our phylogenetic reconstruction based on nucleotide sequences of 13 PCGs recovered topologies consistent with recent phylogenomic studies in supporting Axiidea as the sister group of the Gebiidea-Anomura-Brachyura clade (Shen et al., 2013;Tan et al., 2019;Wolfe et al., 2019). Within Axiidea, Axiidae were recovered as monophyletic (albeit with only two species) and at the most basal position, consistent with previous inferences based on morphology (Poore, 1994), Sanger sequencing (Tsang et al., 2008a;Felder and Robles, 2009;Robles et al., 2009), and genomic results (Shen et al., 2013;Tan et al., 2015;Wolfe et al., 2019). In addition, the two species of Axiidae shared the same mitochondrial gene order pattern which differs from that in other axiidean families, an observation strengthening the monophyly of Axiidae. However, the only two axiid genera included, Calocaris and Eiconaxius, have been placed in other families in the past, Calocarididae and Eiconaxiidae, respectively, both of which have been synonymized with Axiidae (Poore and Collins, 2009;Robles et al., 2009). Hence, more mitochondrial genomes of taxa from Axiidae sensu lato are needed to clearly demonstrate the monophyly of Axiidae.
Previous phylogenetic studies have been unable to clarify the validity of Strahlaxiidae due to the limited taxon representatives (Tsang et al., 2008a;Bracken et al., 2009;Robles et al., 2009;Lin et al., 2012;Shen et al., 2013;Tan et al., 2015Tan et al., , 2017Sun et al., 2018bSun et al., , 2019a. Tan et al. (2019) reconstructed the phylogeny of Decapoda based on mitochondrial genomes and found that two representative genera of Strahlaxiidae were paraphyletic, as did we. These questions about the monophyly of Strahlaxiidae are also reflected in different mitochondrial gene orders. Morphological differences between the three strahlaxiid genera need to be further investigated.
The phylogeny of higher axiidean genera, so-called "Callianassoidea" was reviewed by Robles et al. (2020) who offered a new phylogeny based on analysis of four genes, mitochondrial 16S and 12S rRNA along with nuclear histone 3 and 18S rRNA, and morphology of about half the known species. The family Eucalliacidae was found to be polyphyletic in the molecular analysis but the family and its genera had well defined morphological synapomorphies. Callichiridae has ambiguous molecular support but good morphological support. Callianassidae sensu stricto and four smaller families were well supported. Our study based on few representatives does not contradict these findings. We have provided new data and highlighted the potential of mitochondrial genomes to robustly resolve the deeper relationships among Axiidea, while undeniably the present taxon coverage of this mitogenomic tree is still limited. Therefore, more comprehensive taxon samplings in future will be necessary to lead us closer to the goal of reconstructing the natural evolutionary history of axiidean shrimps.

ACKNOWLEDGMENTS
We sincerely thank Dr. Kuidong Xu (Institute of Oceanology, Chinese Academy of Sciences, Qingdao) for providing us with the deep-sea specimens collected during the M5 seamount cruise in 2019. Drs. Lin Gong (Institute of Oceanology, Chinese Academy of Sciences, Qingdao) and Chengcheng Shen (Second Institute of Oceanography, Ministry of Natural Resources, Hangzhou) for identifying the sponge host of the new axiid, and Dr. Xuwen Wu (Institute of Oceanology, Chinese Academy of Sciences, Qingdao) for offering us color photos of the new species and its sponge hosts taken shortly after collection. We are grateful to the crews of the RV Xiangyanghong 9 and manned submersible Jiaolong for their help in sampling during the Magellan Seamount Chain cruises in 2014 and 2016, and the crews of the RV Kexue for their support in sampling with ROV Faxian during the M5 seamount cruise in 2019. We want to express our appreciation to Dr. Kareen Schnabel (National Institute of Water & Atmospheric Research, New Zealand) for sending us tissue samples. Comparative material was also obtained from specimens collected during the KAVIENG 2014, BATHUS 3 and KARUBAR expeditions to the deep waters in the southwest Pacific, and the KARUBENTHOS 2015 to Guadeloupe, organized by the Muséum national d'Histoire naturelle, Paris (MNHN) and associated organizations. We also thank all the organizers, cruise leaders, scientific and technical crews, and sorters involved to making these collections available. We especially thank Philippe Bouchet, Laure Corbari, Paula Lefèvre-Martin, and Anouchka Krygelmans for help in making the collections available at MNHN and sharing them with Museums Victoria.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars. 2020.00469/full#supplementary-material FIGURE S1 | (A) Codon usage of the mitochondrial genome of Eiconaxius serratus sp. nov. Numbers to the left refer to the total number of codons; (B) the relative synonymous codon usage (RSCU) of the mitochondrial genome of Eiconaxius serratus sp. nov. Numbers to the left refer to the total number of the RSCU values. Codon families are provided on the X axis.