Discovery of a colossal slickhead (Alepocephaliformes: Alepocephalidae): an active-swimming top predator in the deep waters of Suruga Bay, Japan

A novel species of the family Alepocephalidae (slickheads), Narcetes shonanmaruae, is described based on four specimens collected at depths greater than 2171 m in Suruga Bay, Japan. Compared to other alepocephalids, this species is colossal (reaching ca. 140 cm in total length and 25 kg in body weight) and possesses a unique combination of morphological characters comprising anal fin entirely behind the dorsal fin, multiserial teeth on jaws, more scale rows than congeners, precaudal vertebrae less than 30, seven branchiostegal rays, two epurals, and head smaller than those of relatives. Mitogenomic analyses also support the novelty of this large deep-sea slickhead. Although most slickheads are benthopelagic or mesopelagic feeders of gelatinous zooplankton, behavioural observations and dietary analyses indicate that the new species is piscivorous. In addition, a stable nitrogen isotope analysis of specific amino acids showed that N. shonanmaruae occupies one of the highest trophic positions reported from marine environments to date. Video footage recorded using a baited camera deployed at a depth of 2572 m in Suruga Bay revealed the active swimming behaviour of this slickhead. The scavenging ability and broad gape of N. shonanmaruae might be correlated with its colossal body size and relatively high trophic position.

Remarks. Narcetes shonanmaruae sp. nov. is similar to N. erimelas in anal fin position being entirely behind the dorsal fin. However, our species differs from all the other Narcetes species based on the following morphological  Etymology. The species epithet shonanmaruae is a feminine noun in Latin, referring to the ship 'Shonan maru' from which the type materials were caught, honouring the vessel's considerable contribution to deep-sea fish research in the area. The proposed Japanese vernacular name is 'Yokozuna Iwashi' . This species belongs to the family Alepocephalidae, which is referred to as 'Sekitori Iwashi' in Japanese: 'Sekitori' meaning a sumo wrestler and 'Iwashi' meaning a sardine, thereby implying a massive sardine. The term 'Yokozuna' refers to the highest rank in sumo wrestling in Japan. Accordingly, we propose the name 'Yokozuna Iwashi' as being indicative of the large body size and the high trophic position of the newly described species. English vernacular name: Yokozuna Slickhead.
Distribution. Currently only known from Suruga Bay at depths deeper than 2100 m.
Phylogenetic placement of Narcetes shonanmaruae. Mitogenomic DNA sequences for the N. shonanmaruae holotype (SH8-69) and paratype 1 (SH8-43), in addition to those for three species of Conocara and Talismania antillarum, were determined in the present study and deposited in nucleotide sequence databases under the accession numbers shown in Supplementary Table S1. Each translated amino acid sequence was aligned with homologues from 73 other teleost species (75 operational taxonomic units, OTUs, in total, accession numbers shown in Supplementary Table S1). The evolutionary models selected for the maximum likelihood (ML) analysis of each amino acid sequence of 13 mitochondrial protein-coding genes are shown in Supplementary Table S2. The aligned positions of the 13 proteins from each species combined had a total length of 3790 amino acids. The phylogenetic position of N. shonanmaruae determined from our ML analysis was distinct from that of all other alepocephalid fish (Fig. 4). The phylogenetic analyses showed that N. shonanmaruae is a sister species of Narcetes erimelas known from the Pacific, Eastern Central Atlantic, and Indian Ocean at depths of between 1300 and 2740 m 10,13 (Fig. 4). The genetic distances between the COI gene sequences of N. shonanmaruae and Narcetes erimelas are 0.017-0.018, which are greater than the intraspecific values in N. shonanmaruae (0.000-0.002). The bootstrap value in the ML analysis (94%) indicated the monophyly of this clade. The monophyly of the order Alepocephaliformes was supported by a bootstrap value of 100%. The family Alepocephalidae was not monophyletic due to the phylogenetic position of Bathylaco nigricans that fell into a single clade with four species from the family Platytroctidae, and the bootstrap value did not strongly support monophyly (Fig. 4).
Stomach content analysis. Stomach contents were analysed from two individuals of N. shonanmaruae (SH8-69 and SH8-43), both of which contained chyme. In addition, the stomach of the holotype (SH8-69) contained a pair of fish otoliths (~ 5 mm in diameter). However, due to the advanced state of degradation, it was not possible to identify the species of prey fish from these otoliths.
DNA barcoding targeting prey mitochondrial COI genes was performed using chyme collected from the holotype. After removal of low-quality reads, chimeric sequences, and singletons from all detected reads, ~ 200,000 reads were obtained and were clustered in three different OTUs (Supplementary Table S3). OTU1 (85%   www.nature.com/scientificreports/ appearance in the total reads) was derived from N. shonanmaruae despite the use of blocking PCR primers designed to reduce amplification of the host DNA. OTU2 (9% appearance in the total reads) showed highest identity (91%) with the mitochondrial COI gene of the genus Bassozetus in the family Ophidiidae. OTU3 (6% appearance in the total reads) showed highest identity (88%) with the fungal genus Cortinarius (Agaricomycetes; Agaricomycetidae).
Compound-specific isotope analysis of amino acids. The nitrogen isotopic composition of amino acids and the trophic positions (TPs) estimated from the values of glutamic acid and phenylalanine are shown in Table 3 and Fig. 5. The estimated TPs of the two examined N. shonanmaruae individuals were both 4.9, which was the highest among the predatory fishes examined in this study. The nitrogen isotopic compositions of phenylalanine from N. shonanmaruae (#15 and #16) (-3.0‰ and 0.1‰, respectively) were slightly lower than those of the other fish species examined (#1-14) (+ 3.9‰, SD = 1.5, n = 14) (Fig. 5).
In situ observation of Narcetes shonanmaruae using a baited camera.  Table S4, Supplementary Video S1). A 12-s sequence of highdefinition video footage showed that this N. shonanmaruae individual swam into the video frame from the right side and then changed direction towards a position out of view of the camera lens (Supplementary Video S1). The fish vigorously beat its tail fin when veering. Its total length estimated from the video image was more than 100 cm. Individuals of the pudgy cusk-eel Spectrunculus grandis and lithodid crab Paralomis sp. were also observed in the video recordings, as shown in Supplementary Video S1.

Discussion
In this paper, we describe a novel slickhead fish species, which was collected from deep-water sites within Suruga Bay, Japan. To the best of our knowledge, this species is the largest among species in the family Alepocephalidae, and its trophic position is one of the highest among marine organisms reported globally to date 13,16,17 .   www.nature.com/scientificreports/   Table 3. www.nature.com/scientificreports/ Maximum standard lengths (SL) of 77 species from the family are presented in FishBase and a previous study 10,13,16 (summarized in Supplementary Table S5). A frequency distribution of SL in Alepocephalidae clearly shows that N. shonanmaruae and two other slickheads (Alepocephalus bairdii and A. agassizii) are much larger compared to other species in the family, exceeding 100 cm in SL (Supplementary Fig. S17). The length range with highest frequency in SL is 20-40 cm, and the average maximum SL is 35.3 cm (n = 78, SD = 21.5, including the present species), which demonstrate the large size of N. shonanmaruae.
Deep-sea gigantism, which was first described for a plaice species Pleuronectes platessa showing a size increase with depth 18 , has previously been reported in various marine taxa, e.g. the giant squid Architeuthis dux, the giant deep-sea isopod Bathynomus giganteus, the Japanese spider crab Macrocheira kaempferi, the bluntnose sixgill shark Hexanchus griseus, and the oarfish Regalecus glesne 19 . Although there is currently no generally accepted explanation for the deep-sea gigantism observed in fishes, it has been proposed that, in crustaceans, this phenomenon is a consequence of larger cell size that develops in response to lower temperatures 20 , as has also been proposed for other groups 21 . In crustaceans, deep-sea gigantism may also in part reflect decreases in temperature leading to a longer lifespan and thus larger sizes for species with indeterminate growth 20 . Among fish taxa, scavenging species, identified as those that regularly appear at baited cameras, increase in size with depth, whereas non-scavenging species decrease in size over the same depth range 22,23 . The gigantism of deepsea fish could be interpreted in terms of a response to differences in the characteristics of food supply by the two groups 22 . Scavenging species depend on large, randomly distributed packages of carrion falling from the shallower waters, and a larger body size permits greater swimming abilities and endurance, thereby allowing the fish to move efficiently between occasional feeding events in the deep sea 23 . Trophic level and body size are also believed to be positively correlated across all the species examined, and morphological constraints associated with gape limitation may play a prominent role in determining body size 24 . Deep-sea alepocephaliform fishes are highly diverse, particularly within the family Alepocephalidae (slickheads), which consists of more than 90 species recorded globally [25][26][27] . Slickheads have been suggested to be important consumers of gelatinous zooplankton [28][29][30] , and generally show lower trophic positions than that of other fishes living in the same environments 31,32 . Thus, the feeding ecology and the trophic level of N. shonanmaruae would appear to be unique in this family. The scavenging ability (as implied by capture using a longline) and broad gape of N. shonanmaruae ( Supplementary  Fig. S16f) might be correlated with its colossal body size and relatively high trophic position.
The estimated trophic positions (TP = 4.9) of the two individuals for which stomach contents were analysed were the highest among the predators examined in the present study and are relatively high compared with the values obtained in previous studies 17, [33][34][35][36] . The bluntnose sixgill shark H. griseus is known as an apex predator/ scavenger in the deep sea [37][38][39] , whereas the false catshark Pseudotriakis microdon is a large predatory shark that reaches 3 m in length and feeds on a wide variety of prey, including teleost and chondrichthyan fishes, cephalopods, and marine mammals 40 . The trophic positions determined for H. griseus (4.2 and 4.3) and P. microdon (4.3) in the present study are all lower than that of N. shonanmaruae. Since trophic positions estimated from stable isotope analysis provide long-term dietary information 41,42 , these results indicate that N. shonanmaruae commonly feeds on high-trophic prey (TP = ca. 4). Nielsen et al. 17 compiled amino acid stable nitrogen isotope ratios from 359 marine species covering four trophic levels (primary producer, herbivore, omnivore, and carnivore) from the literature. We calculated all the TPs in this previous study using Eq. The nitrogen isotopic compositions of phenylalanine from N. shonanmaruae were slightly lower than those of the other fish species examined (Fig. 5). Such differences are within intra/inter-species variations collected from same regions. Intra-species variations of phenylalanine δ 15 N values were reported from a largescale blackfish Girella punctata collected sympatrically in Sagami Bay, Japan (the values ranging from 3.7 to 8.7), but the TPs were nearly consistent (TP ranging from 2.7 to 3.1) because the nitrogen isotopic compositions of glutamic acid fluctuated according to those of phenylalanine 36 . The phenylalanine δ 15 N values were also variable between planktonic/ pelagic taxa collected at a same site 100 km north of the island of Oahu in the North Pacific Subtropical Gyre 43 , ranging from − 4.8 to 7.5, which is comparable to the values found in the results of the present study (Fig. 5).
Both morphological analyses and DNA barcoding of stomach contents showed N. shonanmaruae to be piscivorous in its feeding habits. In this regard, although a pair of fish otoliths was found in the stomach of the holotype, prey species identification was not possible due to the severe degradation of these otoliths. Even though the stomachs of both analysed individuals were almost empty, small amounts of chyme were detected. Such a paucity of stomach contents is common among deep-sea fish that tend to regurgitate their stomach contents upon being landed 44 . DNA meta-barcoding of the stomach chyme also supported a piscivorous habit. A sequence similarity search for sequences amplified from the chyme indicated the prey item to be an ophidiid fish (Family Ophidiidae), which comprise a dominant group of deep-sea demersal fishes containing 218 known species 45,46 . This result is consistent with a predatory lifestyle. N. shonanmaruae possesses large, widely spaced gill rakers (Fig. 2f,  Supplementary Fig. S16c). In general, gill rakers play the role of entrapment structures to retain planktonic prey from a volume of engulfed water and are thus common in planktivorous filter feeding fish 47,48 . Increasing the number of gill rakers enhances crossflow filtering and the closely spaced gill rakers also limit the escape possibilities of small prey. However, a dense gill raker apparatus is more likely to become clogged by sediments than sparser gill rakers, and foraging in the muddy bottom of the profundal zone most likely requires other gill raker adaptations. Accordingly, a high number of long gill rakers is common in planktivorous fish species, whereas www.nature.com/scientificreports/ benthic species usually display a lower number of shorter gill rakers [49][50][51][52] . The rakers of N. shonanmaruae seem to be too sparse to support filter feeding, which is inconsistent with a planktivorous feeding habit.
Cannibalism of this slickhead might be possible, because the most dominant OTU from DNA meta-barcoding of the stomach chyme was the sequences of N. shonanmaruae even though a blocking primer against the N. shonanmaruae sequence was applied. However, several previous studies showed amplification of host sequences from stomach contents in spite of using specific blocking primers against the host 53,54 . In addition, there is no record of smaller individuals of this slickhead at any depths in Suruga Bay (Y. F., unpublished data), which implies difficulty in smaller N. shonanmaruae individuals being the staple diet for adults. Somewhat incongruously, OTU3 showed high homology with several species of Agaricomycetes (Fungi) and Oomycetes (Stramenopiles). As these are not considered to form part of the N. shonanmaruae diet, it is assumed that they could be experimental contaminants of terrestrial origin 55,56 . Alternatively, they might be unknown parasitic organisms present in the digestive apparatus of slickheads or environmental DNA included in the stomach sample.
Our video observations revealed that N. shonanmaruae is an active swimmer (Supplementary Video S1). In the same video sequence, we observed the most dominant fishes living at the depth (Spectrunculus grandis, Simenchelys parasitica, and Coryphaenoides acrolepis), which are taeniform or anguilliform species that do not move rapidly 57,58 . In contrast, the newly discovered slickhead-being fusiform and possessing a narrowed but robust caudal peduncle with a relatively large emarginate caudal fin-was observed veering vigorously with a single tail stroke and rapidly disappearing from the video frame. Such a body plan and swimming behaviour are consistent with the stomach contents (i.e. fish) and a high trophic position, all indicating that this slickhead is probably a top predator in the deepest part of Suruga Bay.
Each year, many new species of fish are discovered, most of which are relatively small 13,59 . In contrast, discoveries of large bony fishes from marine environments have been relatively rare in recent years. Since the 1800s, Suruga Bay has been one of the most studied bays in Japan 4-6 , and has yielded many marine organisms described as new species [60][61][62] . Deep-sea fishing, including that using longlines and bottom trawls, continues to be practiced in Suruga Bay, and therefore the deep-sea fauna inhabiting this bay is relatively well documented [7][8][9] . The discovery of a colossal slickhead from this bay was thus completely unexpected. In the present study, we conducted two longline searches at depths greater than 2171 m during two separate cruises and collected two individuals from each line. This high encounter probability might be due to a combination of the research method used and depth searched. Longline fishing is generally conducted at shallower depths (shallower than 1000 m in Suruga Bay) because of the difficulty of line control and limited profitable catches. Scientific trawls and dredges have sometimes been conducted at greater depths, although it is difficult to collect large, fast swimmers. In this regard, obvious differences in size and species selectivity of longlines and trawl nets have been reported 63 , and it is highly likely that further longline-based surveys in deep waters will reveal the actual diversity of deep-sea predators, not only in Suruga Bay but also in many marine environments worldwide.
Top predators play an important role in the maintenance of species diversity and of ecosystem functions [64][65][66][67] . The extinction and reintroduction of gray wolves in Yellowstone National Park is an example of the drastic change in a population of top predators affecting the entire ecosystem 68 . Similar instances are too frequent to enumerate not only in terrestrial but also in marine ecosystems 65,69,70 . "Fishing down the food web" 71 occurs even in the deep sea that is the largest habitat for life on the Earth. Trophic positions of catches from deep sea are quite high (from 3.5 to 4.5 predominantly, estimated using FishBase), and some fish resources have already been depleted due to over fishing 72 . Oceanic warming, acidification, and deoxygenation are occurring even in the deep sea 1 , which is assumed to initially affect large, predatory consumers. If the deep-sea ecosystems are severely damaged through the trophic cascade like that in Yellowstone, its influence on the global environment is immeasurable. As an initial step, there is an urgent need to conduct a broad and precise census of predators in the deep oceans to facilitate protection of natural resources of the planet and human livelihoods.

Methods
Bottom longline. In 2016, two research cruises were conducted in January-February (SH16-01) and in November-December (SH16-02) in Suruga Bay using the training vessel Shonan maru belonging to the Kanagawa Prefectural Marine Science High School. Two research longlines, SH8 and SH12, were deployed on the bottom of the Suruga Trough at the mouth of the bay on 3rd and 4th February and 22nd and 23rd November 2016, respectively (Fig. 1). The coordinates and depths of the surveyed sites are shown in Supplementary  Table S4. Both longlines were left overnight on the deep-sea floor, and then retrieved onboard.
The longlines were composed of a 4-km main line (5 mm in diameter, polyester), 400 hooks with 5-m branch lines through the main line, two 10-kg lead sinkers, and a radio buoy at each end of the main line. A half or onethird portion of mackerel was impaled on each hook as bait. Three miniature salinity, temperature, and depth loggers (DST-CTD; Star-Oddi, Garðabaer, Iceland) were attached to the start, mid, and end points of the main lines. This study was conducted in accordance with the Guidelines for Proper Conduct of Animal Experiments published by the Science Council of Japan, and the guidelines for fish experiments published by the Nature Conservation Committee of the Ichthyological Society of Japan. All the field experiments were approved by the Research Safety Committee of the Japan Agency for Marine-Earth Science and Technology.
Morphological observations and X-ray tomography scanning. The total body lengths and weights of the four alepocephaliform fish caught using longlines were measured. All the individuals were immediately frozen onboard. After completion of the cruises, additional observations and detailed measurements of individuals were performed in the laboratory. Morphological characters of this alepocephaliform fish were compared to those in previous studies 10,11,[73][74][75][76][77] . X-ray CT scanning of two individuals (holotype and paratype 1) was conducted for observations of osteological characteristics using a Discovery 750 HD CT scanner (GE Health- www.nature.com/scientificreports/ care, Waukesha, WI, USA) under the following conditions: tube voltage, 120 kV; tube current, Auto mA; section thickness, 0.625 mm; rotation time, 1.0 s; and pitch, 0.984:1. Taxonomic keys were referred from previous studies 74 .

Microfocus X-ray CT morphological analysis of otoliths. Morphological analyses of otoliths from
two individuals were conducted using a microfocus X-ray CT scanner (ScanXmate-D160TSS105; Comscantecno Co. Ltd., Kanagawa, Japan). This system applies X-rays to a sampling stage, which rotates 360 degrees, with a high-resolution setting (X-ray focus spot diameter of 0.8 µm, X-ray tube voltage of 45 kV, detector array size of 1024 × 1024, 1200 projections/360º, two-times averaging, sequential imaging, 2.0 s/projection). Spatial resolution of scanning was changed from 7 to 15 µm, depending on the size of otoliths. Reconstruction of threedimensional (3D) tomographic images was performed with ConeCTexpress (R) software (Comscantecno Co. Ltd., Kanagawa, Japan) using the convolution back-projection method. In order to avoid artefacts associated with 3D image reconstruction, noise-cancelling and ring-artefact filters were applied. Calculations of lengths, volumes, and other morphological properties were performed using Molcer Plus imaging software (White Rabbit Corp. Inc., Tokyo, Japan).
DNA sequencing. Total DNA was extracted from the muscle tissue of N. shonanmaruae specimens using a DNeasy Tissue and Blood Kit (QIAGEN Japan, Tokyo, Japan). Tissue samples of four other slickhead species were provided by Academia Sinica, Taiwan (three Conocara species), and Tokai University, Japan (Talismania antillarum), and used for DNA extraction. The sample IDs of these species are shown in Supplementary Table S1. Extracted DNA was used as a template for PCR amplification to amplify the mitogenome. Two universal primer sets for the mitochondrial 16S rRNA gene 78 and the COI gene 79 were used for PCR (Supplementary Table S6).
In addition, three degenerate primer sets were designed for ATPase F0 subunit 6 (ATP6), NADH dehydrogenase subunit 4 (ND4), and cytochrome b (cytb) genes (Supplementary Table S6), based on the sequences from several alepocephalid species deposited in nucleotide sequence databases. After PCR amplification and partial sequencing of the genes, additional primers specific to N. shonanmaruae were designed to amplify five long fragments from the mitogenome (Supplementary Table S6). Specific primers used for sequencing of the mitogenomes of the four additional slickheads were designed in the same manner. PCR was preformed using an Ex Taq Kit (TaKaRa, Kyoto, Japan). All the PCR products were purified using a Wizard SV gel and PCR purification Kit (Promega KK, Tokyo, Japan). The purified fragments were used for sequencing reactions using a Big Dye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA). Sequencing was performed using an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems Japan Ltd., Tokyo, Japan). After sequencing, assembling and ambiguity editing of each sequence were conducted using CodonCode Aligner software (CodonCode Corporation, Dedham, MA, USA) to reconstruct whole mitogenome sequences.
Phylogenetic analysis. Seventy-five OTUs (six obtained in the present study and the remainder from nucleotide sequence databases) were used for phylogenetic analysis, among which those of two Osteoglossiformes species (Arapaima gigas and Osteoglossum bicirrhosum) were used as an operational outgroup (Supplementary Table S1). All the 13 protein-coding genes (ND1, ND2, COI, COII, ATP8, ATP6, COIII, ND3, ND4L, ND4, ND5, ND6, and cytb) in the mitogenome were used for phylogenetic analysis. The predicted amino acid sequences of the 13 protein-coding genes were independently aligned using MAFFT v7.164b 80 with auto parameters, followed by automatic editing of the resulting alignments using the GBLOCKS program 81,82 . An appropriate evolutionary model for each gene alignment was then selected for maximum likelihood (ML) analysis using Aminosan software 83 , based on Akaike Information Criteria (Supplementary Table S2). Amino acid sequence alignments from each gene were concatenated (3790 amino acids from 75 OTUs in total). ML analysis using the concatenated alignment of fish mitochondrial genes was performed using RAxML version 8.2.10 84 , with appropriate evolutionary models selected. ML bootstrap probability analysis was performed using the same software with 100 resamplings.
Dietary analyses. The abdominal cavities of two N. shonanmaruae individuals (SH8-69 and SH8-43) were dissected prior to formalin fixation. Each stomach was excised, and the entire stomach contents were retrieved. The contents were visually examined, and prey items were selected from the digested materials. The chyme from one individual was used for DNA meta-barcoding analysis targeting residual prey DNA. DNA extraction, library preparation, and high-throughput sequencing were performed by Bioengineering Lab. Co., Ltd. (Kanagawa, Japan) as follows. Approximately 20 g of stomach chyme was freeze-dried and completely homogenized. DNA was isolated from the homogenized material using an ISOSPIN Blood & Plasmid Kit (Nippon Gene, Tokyo, Japan). Short fragments of COI genes were amplified using modified metazoan universal primer sets 85 , incorporating forward and reverse adapter sequences, i.e. ACA CTC TTT CCC TAC ACG ACG CTC TTC CGA TCT-GGW ACW GGW TGA ACW GTW TAY CCY CC and GTG ACT GGA GTT CAG ACG TGT GCT CTT CCG ATCT-TAHACTTCNGGGTGKCCR AAR AATCA. In this procedure, a set of blocking primers was used to prevent amplification of the COI gene from the predatory (host) fish, as described previously 86 . Using the PCR products, paired-end sequence libraries were constructed according to the tailed PCR method, and these libraries were sequenced using Illumina MiSeq (Illumina, San Diego, CA, USA) under 2 × 250 bp conditions. After removing low-quality reads, OTUs were constructed. A homology search of each OTU was subsequently performed to identify prey animals from the stomach.
Nitrogen isotope analysis of individual amino acids. The extraction and derivatization of amino acids and nitrogen isotope analyses of specific amino acids were conducted according to the methods described www.nature.com/scientificreports/ in previous studies 35,87 . Each sample was hydrolysed with 12 M HCl at 110 °C for 12 h. The hydrolysate was filtered to remove any precipitate using Pall Nanosep MF Centrifugal Devices containing a GHP Membrane (Pall, Port Washington, NY, USA), and then washed with n-hexane/dichloromethane (3:2, v/v) to remove hydrophobic constituents. The derivatization of amino acids was performed sequentially using thionyl chloride/2propanol (1:4, v/v) at 110 °C for 2 h and pivaloyl chloride/dichloromethane (1:4, v/v) at 110 °C for 2 h. The Pv/ iPr derivatives of the amino acids were extracted with n-hexane/dichloromethane (3:2, v/v). The nitrogen isotopic composition of the individual amino acids was determined by GC/C/IRMS using an Agilent Technologies 6890 N gas chromatograph (Agilent Technologies, Santa Clara, CA, USA) coupled to a Thermo Fisher Scientific Delta plus XP mass spectrometer (Thermo Finnigan, Bremen, Germany) with a GC-C/TC III interface 35,87 . The isotopic composition was reported as the δ 15 N notation relative to atmospheric N 2 (AIR). The analytical error was better than ± 0.5‰. The TP of samples (TP Glu/Phe ) was calculated from the δ 15 N values of glutamic acid (δ 15 N Glu ) and phenylalanine (δ 15 N Phe ) using Eq. (1) with β values of -3.4‰.
Baited camera observations. Baited camera observations were conducted on 26th November 2016 at depths greater than 2000 m in the mouth of Suruga Bay (Fig. 1, Supplementary Table S4). The camera system consisted of a high-definition video camera (KCN-EV7520SDI; Totsu Sangyo, Tokyo, Japan), an HD-SDI recorder (HDS-1601; Totsu Sangyo), an LED light array composed of 20 red LEDs (OSR5XNE3C1E; OptoSupply, Hong Kong) fixed with an epoxy resin 88 , an acoustic releaser (STD-301; Kaiyo Denshi, Saitama, Japan), a sinker-releasing unit (PJ-0001; Pearl Giken, Chiba, Japan), a DST-CTD logger (Star-Oddi), a current profiler (Zpulse Doppler current sensor 4930; Aanderaa, Bergen, Norway), and two batteries (9200WP-L; Daiwa, Tokyo, Japan). The camera system was deployed from Shonan maru in free-fall mode. A 30-kg ballast was released by the acoustic releaser according to the release command from the ship. Video recording was started 30 min before landing and yielded a 5-h video footage. All the data, including videos and current profiles, were recovered when the system was retrieved onboard.