Integrative taxonomy reveals a rare and new cusk-eel species of Luciobrotula (Teleostei, Ophidiidae) from the Solomon Sea, West Pacific

Abstract. With six valid species, Luciobrotula is a small genus of the family Ophidiidae, commonly known as cusk-eels. They are benthopelagic fishes occurring at depths ranging from 115–2300 m in the Atlantic, Indian, and Pacific Oceans. Among them, Luciobrotula bartschi is the only known species in the West Pacific. Three specimens of Luciobrotula were collected from the Philippine Sea, Bismarck Sea, and Solomon Sea in the West Pacific during the AURORA, PAPUA NIUGINI, and MADEEP expeditions under the Tropical Deep-Sea Benthos program, and all of them were initially identified as L. bartschi. Subsequent examination with integrative taxonomy indicates that they belong to two distinct species, with the specimen collected from the Solomon Sea representing a new species, which is described here. In terms of morphology, Luciobrotula polylepis sp. nov. differs from its congeners by having a relatively longer lateral line (end of the lateral line below the 33rd dorsal-fin ray) and fewer vertebrae (abdominal vertebrae 13, total vertebrae 50). In the inferred COI gene tree, the two western Pacific species of Luciobrotula do not form a monophyletic group. The genetic K2P distance between the two species is 13.8% on average at the COI locus.


Introduction
formalin and later transferred to 70% ethanol for long-term preservation. The specimens were deposited in the ichthyological collections of the NTUM and ASIZP.

Morphological examination
The three specimens collected in this study and four voucher specimens of Luciobrotula bartschi (ASIZP 0070170, ASIZP 0063749, ASIZP 0066071, and ASIZP 0075076) from the East China Sea and South China Sea (Fig. 1) were morphologically examined (see below and Appendix). Methods for measuring, counting, and general terminology followed Nielsen (2009). Specimens were measured with a dial caliper to the nearest 0.1 mm. Internal osteological characters were examined through radiographs. Color pattern was based on freshly collected specimens and photos, with additional information provided after preservation.

DNA data collection
Total genomic DNA was extracted from each tissue using a commercial DNA extraction kit and a robot (LabTurbo 48 Compact System extractor, Taigene Biosciences Corp., Taipei, Taiwan) following the manufacturer's protocols. The cytochrome c oxidase subunit I (COI) gene was chosen as a marker for molecular examination of the specimens. A polymerase chain reaction (PCR) was used to amplify the European Journal of Taxonomy 750: 52-69 (2021) target gene fragment using the universal fish primers provided in Ward et al. (2005). PCR was carried out in a 25 µl volume containing 9 μl sterile distilled water, 0.5 μl of each primer (10 µM), 12.5 μl of EmeraldAmp MAX HS PCR Master Mix (TaKaRa), and 2.5 μl of DNA template (around 10~20 ng). The thermal cycling profile for amplification consisted of an initial denaturation stage (95°C, 60 sec), followed by 35 cycles each with a denaturation step (95°C, 30 sec), an annealing step (51°C, 30 sec), and an elongation step (72°C, 40 sec), before a final extension stage (72°C, 7 min). The successfully amplified products were then purified using the AMPure magnetic bead cleanup protocol (Agencourt Bioscience Corp.) and sequenced by Sanger sequencing at Genomics BioSci and Tech (Taipei). The same primers used for PCR were also used for sequencing; only the forward COI primer was used.

Sequence alignment and phylogenetic analysis
The obtained COI sequences were viewed and edited using CodonCode Aligner ver. 7.2.1 (CodonCode Corporation, Dedham, MA, USA) and were then aligned with eight other homologous sequences of Luciobrotula species retrieved from GenBank (NCBI, Nation Center for Biotechnology Information) (n = 7) and BOLD (The Barcode of Life Data Systems) (n = 1) ( Table 1) using the automatic multiplealignment program MUSCLE (Edgar 2004). MEGA X (Kumar et al. 2018) software was further used to manage the compiled dataset and compute pairwise distances of compared sequences with the Kimura-2-Parameter model (K2P) (Kimura 1980). The phylogenetic analysis was conducted based on the compiled COI dataset using the maximum likelihood method (ML) with the nucleotide substitution model GTR+G as implemented in the software RAxML ver. 8.0.4 (Stamatakis 2014). Nodal support was assessed with bootstrapping (Felsenstein 1985) under the ML criterion, based on 1000 pseudoreplicates. Neobythites bimarginatus Fourmanoir &Rivaton, 1979 andNeobythites stigmosus Machida, 1984 were used as outgroups to root the inferred COI tree.

Species delimitation analysis
The same COI gene dataset was used in three DNA-based species delimitation analyses, Automatic Barcode Gap Discovery (ABGD) (Puillandre et al. 2012), Bayesian based Poisson Tree Processes (bPTP) (Zhang et al. 2013), and Character-Based DNA Barcoding (CBB) (Desalle et al. 2005). ABGD is a tool for detecting significant differences between intra-and interspecific variation (barcode gap) by examining pairwise genetic distances. Operational taxonomic units (OTUs) or putative species were redefined through the analytical algorithm. The analysis was performed at the web interface (https://bioinfo.mnhn.fr/abi/public/abgd/abgdweb.html) with the default value (1.5) for relative gap width (X), and the intraspecific divergence (P) value (0.001 to 0.1) with 20 steps under the K2P distance.
bPTP is a method for delimiting species based on a rooted phylogenetic tree, and mutations are modeled as speciation or branching events. Here, we used the inferred COI gene tree (see above) as the input tree. The bPTP settings are: number of MCMC generations = 100 000; thinning = 100; burn-in = 0.1; and seed = 123. The analysis was performed at the web interface available from https://species.h-its.org/.
CBB is a method derived from the standard DNA barcoding approach (Hebert et al. 2004;Ward et al. 2005). Instead of simple cut-off distance thresholds for delimiting the species presented in the DNA dataset, the putative species are identified through the presence or absence of discrete and unique nucleotide substitutions within the DNA sequences of taxa (Desalle et al. 2005;Rach et al. 2008;Brower et al. 2010;Guimarães et al. 2020). Here, MEGA X software and inferred COI gene tree were used to determine apomorphic nucleotide sites in L. polylepis sp. nov. at the COI locus. Numbering of the determined nucleotide sites starts from the first nucleotide of the gene defined through sequence alignment with the complete COI sequence retrieved from the whole mitochondrial genome of Neobythites unimaculatus Smith & Radcliffe 1913 (AP018428: 5544-7094).
The congruent results from bPTP and ABGD analyses were considered to be primary support of the OTUs (i.e., inferred potential species); other criteria (CBB result, morphological evidence, etc.) were also used for final validation of delimited species.

Molecular phylogeny and species delimitation
The COI dataset comprised 13 aligned sequences including three newly obtained sequences from the collected specimens, three additional sequences of L. bartschi from the South China Sea and Western Australia, five sequences of L. coheni from the Eastern Pacific, plus two outgroup sequences ( Table 1). The length of the aligned sequences of the dataset is 618 bp. Figure 2 shows the phylogenetic tree inferred from the ML analysis based on the dataset. The monophyly of the genus Luciobrotula is strongly supported (bootstrap value = 98%), and ingroup sequences form three clades or lineages among which two contain sequences from the two known species (Fig. 2). While two of our newly obtained sequences (ASIZP 0913925 and PNG1082) fall into the L. bartschi clade, the third one (PNG2363) appears to be a previously unknown lineage. Advanced species delimitation analyses with ABGD and bPTP based on the same COI dataset reveal a congruent result with a prediction of three OTUs, corroborating the phylogenetic finding (Fig. 2). The delimited OTUs (or inferred species) are genetically distinct from each other. The unknown lineage is distinct from others by 37 unique nucleotide sites based on CBB analysis. The average genetic distances measured using the K2P model among them are from 0.130 to 0.138 at the COI locus. Further morphological examination on the specimens indicates that the features of the sample collected from the Solomon Sea (PNG2363) are unique among all known Luciobrotula species (see below), and we validate it herein as a new species.  Table 2 Diagnosis Luciobrotula polylepis sp. nov. is morphologically distinct from all congeners by the following combination of characters: lateral line ending below 33 rd dorsal-fin ray; dorsal-fin rays 86, anal-fin rays 70, precaudal vertebrae 13, total vertebrae 50; gill rakers 17 (3 long rakers and 14 dentigerous plates); longest gill raker 2.1% SL; height of posterior margin of maxilla 3.2% SL; distance from the snout to end of lateral line 60% SL; one interorbital pore and four occipital pores.

Differential diagnosis
The new species is most similar to L. brasiliensis because both share the low number of vertebrae. It differs from L. brasiliensis by having a much longer lateral line (ending at the 33 rd dorsal-fin ray vs ending at the 2 nd dorsal-fin ray), a slightly more posterior position of the anal-fin origin (first anal ray below dorsal ray no. 22 vs first anal ray below dorsal ray no. 17), more pectoral-fin rays (32 vs 26), more gill rakers (17 vs 13-14), longer gill raker on first arch (2.1% SL vs 1.2% SL).
It differs from L. nolfi by having a slightly longer lateral line (ending at the 33 rd dorsal-fin ray vs ending at the 27 th -31 st dorsal-fin ray), slightly more anterior position of the anal-fin origin (first anal-fin ray below the 17 th vertebra vs first anal-fin ray below the 19 th -20 th vertebrae), smaller head (23.9% SL vs 24.5-28.0% SL), and relatively deeper body (16.3% SL vs 12.5-15.0% SL).
It differs from L. lineata by having a much longer lateral line (ending at the 33 rd dorsal-fin ray vs ending at the 2 nd dorsal-fin ray), fewer dorsal-fin rays (86 vs 92), more pectoral-fin rays (32 vs 26), shorter pelvic-fin rays (10.9% SL vs 15.0% SL) and longer gill raker on the first arch (2.1% SL vs 0.7% SL). A detailed comparison between the new species and other congeners is provided in Table 2.
Along the COI gene, the following apomorphic sites are unique nucleotides from the only specimen of L. polylepis sp. nov. examined here; these nucleotide sites can be used for the molecular diagnosis of the species to differentiate it from L. coheni and L. bartschi examined in this study. Nos Longest gill filament on anterior arch

Etymology
The name polylepis is derived from the Greek 'poly', meaning 'many' or 'numerous', and 'lepis', meaning 'scales', in reference to the much longer lateral line and therefore more lateral line scales compared with L. bartschi, the only congener distributed in the West Pacific.

Description
Measurements and counts of the holotype given in Table 2. Body elongate with tapering caudal portion, snout and head slightly depressed; eye small and round, horizontal eye diameter about half of snout length. Mouth large, oblique; upper jaw reaching a vertical through the posterior margin of orbit,  (2021) posterior part vertically much extended, slightly protruding beyond lower jaw when mouth closed. Boomerang-formed vomer; palatine, and upper and lower jaw with many small, close-set, rather blunt teeth in several irregular rows; fang-like teeth absent in both jaws. One median and a pair of two large basibranchial tooth patches. Anterior nostril with low rim and placed midway between upper lip and posterior nostril, with small rounded flap rising from anterior rim. Posterior margins of preopercle, interopercle, and subopercle rounded, without spine. First gill arch with four finely dentigerous plates on upper branch, one long raker on the angle, and lower branch with two long rakers interspaced with 10 dentigerous plates (Fig. 4D); gill filaments ca 100, the longest about half as long as longest gill raker; pseudobranchial filament damaged, unavailable count.
Sensory pores are found all over head (Fig. 4A-B). Supraorbital with group of eight pores behind eye, five pores immediately above eye, and five small pores in a row on tip of snout, larger pore between flaps on tip of snout, and above each nostril, one interorbital pore, four occipital pores, six suborbital pores and four mandibular pores, 10 small pores close to lower jaw, between this row and mandibular having four small pores, and finally a row of six pores above posterior mandibular, two pores behind posterior end of maxilla, and preopercle with six pores.
Sagittal otolith is elongate and thin, about 2.5 times as long as high. Sulcus divided into ostium and cauda. Cauda is about ⅔ of ostium (Fig. 4C). Due to the damaged anterior rim, the presence of an ostial channel could not be ascertained.
Body, top of head, and opercle covered with small cycloid scales, with ca 72 scales in oblique line from origin of anal fin forwards and ca 111 scales from upper part of gill slit to base of caudal fin; single lateral line originating at upper angle of opercle and extending posteriorly in straight line placed about midway between midline and profile of body, ending below 33 rd dorsal-fin ray. Dorsal-fin origin above end of pectoral fin; anal-fin origin at about mid-body of fish, pectoral fin placed medially and pelvic fin reaching one third from base to anal fin.  (2021) Third neural spine pointed, length of first spine half as long as second spine (Fig. 3C), neural spines of posterior 10 pre-caudal vertebrae with blunt tips and broad bases, 4 th -11 th precaudal vertebrae with broad bases and depressed neural spines, 7 th -13 th precaudal vertebrae with parapophyses, and pleural ribs on 3 rd -6 th precaudal vertebrae. Epipleural ribs hard to observe.
Head brown; body brownish-yellow with bluish-brown abdomen (Fig. 3A). Dorsal, pectoral, anal, and caudal fins black. Color of preserved specimen similar to that of fresh specimens, the head and body uniformly brown with dark bluish-brown abdomen (Fig. 3B).

Distribution
Possibly endemic to waters off Papua New Guinea; the only known specimen was collected on the SE continental slope of New Britain Island, Papua New Guinea, at depths of 430-620 m (Fig. 1).

Key to all known species of Luciobrotula Smith & Radcliffe, 1913
(modified from Nielsen 2009) recorded from Papua New Guinean waters, first by Nielsen & Møller (2008) (n = 1), and later by Fricke et al. (2014) (n = 1). However, the specimen of 'L. bartschi' (NTUM 10054) examined by Fricke et al. (2014) represented a misidentification. Upon our reexamination, we found that it possesses a combination of characters (a copulatory organ and the caudal fin fused with the dorsal and anal fins) that matches fishes from another ophidiiform family, the Bythitidae Gill, 1861 (Møller et al. 2016). Nevertheless, the six samples of L. bartschi examined in this study were all collected from sites within the reported range of the species (Fig. 1). The new species is possibly endemic to Papua New Guinea, as it is so far known from its type locality only. These two western Pacific species appear to share a similar bathymetric range (430-620 m vs 400-2283 m depth) and habitat (mud bottom), and both are found in Papua New Guinean waters. Certainly, their distribution and ecology require more investigations.
Specimens of Luciobrotula seem to be rare. Despite intensive sampling efforts from either local organizations in Taiwan or international expeditions through the TDSB for over a decade, only a few specimens were made available for scientific investigations. The difficulty in sampling has limited our understanding of biodiversity, phylogeny, biogeography, and ecology of deep-sea fishes such as those from the rare genus Luciobrotula or others (e.g., Chelidoperca Boulenger, 1895) (Lee S.-H. et al. 2019). In spite of that, in this study we successfully uncover the hidden diversity of the Luciobrotula in the West Pacific using an integrated approach in taxonomy and conduct the first phylogenetic study of Luciobrotula. From the inferred phylogenetic tree, the two western Pacific species of Luciobrotula are shown to be distantly related to each other despite their geographic proximity. Our preliminary phylogenetic result also indicates that the species (L. bartschi and L. coheni) sharing a similar morphology may not be closely related.