Independent radiation of snailfishes into the hadal zone confirmed by Paraliparis selti sp. nov. (Perciformes: Liparidae) from the Atacama Trench, SE Pacific

Snailfishes are among the most rapidly radiating families of marine fishes, resulting in a global distribution from the coastal intertidal to deep subduction trenches. The true diversity and distribution of deep-water snailfishes, particularly at hadal depths (>6000 m) and in the Southern Hemisphere, remain uncertain due to the rarity of samples. Here, we present the snailfish diversity at near-hadal and hadal depths in the Atacama Trench, which runs along the southwest coast of South America. Using free-fall baited cameras and traps, we documented at least three species of hadal snailfishes between 5920 and 7608 m based on distinct morphologies. One snailfish specimen was recovered from 6714 m, which we describe herein as Paraliparis selti sp. nov., based on a combined morphological and molecular taxonomic approach (16S, COI, and Cyt-b). Paraliparis selti sp. nov. is morphologically distinct from described snailfishes due to a combination of high number of; vertebrae (65) particularly the abdominal vertebrae (12), dorsal fin rays (60), anal fin rays (52), and caudal fin rays (8); comparatively low number of pectoral fin rays (18) which forms a deep notch with two widely spaced non-rudimentary rays. Micro-CT was used to minimise dissection of the specimen and to provide a digital holotype. Paraliparis selti sp. nov. highlights the importance of the Liparidae at hadal depths and provides evidence for at least two independent radiations of snailfishes into the hadal zone.


Introduction
Snailfishes (Family Liparidae Gill, 1861) are a diverse teleost group. Over 400 species of snailfishes are found in every ocean. Closely followed by the Ophidiidae (Gerringer et al. 2021), the Liparidae hold the widest known bathymetric range of any vertebrate family, from the intertidal to the deep subduction trenches (Chernova et al. 2004;Fricke et al. 2020). At least 200 snailfish species are present in the Southern Hemisphere (Stein and Chernova 2002;Stein 2012), and of those, about 150 species are known from the Subantarctic and Antarctic waters (Chernova et al. 2004;Chernova and Prut'ko 2011;Stein 2012). Only six genera (~20 species) are known from the eastern Pacific off the west coast of South America (Stein 2005). These species represent a distinct Chilean group of liparids (Stein et al. 2001) including the endemic, monotypic genus, Eknomoliparis (Stein et al. 1991).
Despite the challenges to sample depths deeper than 6000 m, snailfishes are well-documented at hadal depths (Linley et al. 2016). Between 6000 to 8143 m, snailfishes dominate the hadal ichthyofauna in at least ten trenches worldwide (Jamieson et al. 2021). Snailfishes are exceptionally welladapted to the hadal environment, which is characterised by low temperature (1-3°C), extremely high pressure, and seismic activity Gerringer 2019). Currently, all described hadal snailfishes are ascribed to two genera (Linley et al. 2016;Gerringer 2019;Jamieson et al. 2021). In the Southern Hemisphere, Notoliparis Andriashev, 1975 contains N. kermadecensis Nielsen, 1964 andN. stewarti Stein, 2016 from the Kermadec Trench and N. antonbruuni Stein, 2005 from the Atacama Trench, the latter described from a single badly damaged specimen captured at 6150 m depth. Pseudoliparis Andriashev, 1955 is known from Northern Hemisphere and contains P. belyaevi Stein, 1998 andP. amblystomopsis Andriashev andStein, 1998 from the Japan Trench, and P. swirei Gerringer and Linley, 2017 from the Mariana and Yap Trenches. In addition to these truly hadal species, near-hadal snailfishes are also known from the genus Careproctus Krøyer, 1862, specifically Careproctus sandwichensis Andriashev and Stein, 1998 from 5435 to 5453 m in the South Sandwich Trench, Southern Ocean. Further, at least six hadal liparid species have been observed on camera but lack voucher specimens needed for description (see Jamieson et al. 2021).
Although the importance of snailfishes at hadal depth is well-established, the true diversity and distribution of deepwater snailfishes remains uncertain due to comparatively lower sampling intensity and infrequent observations, coupled with new species being often described from single specimens in poor physical condition (Duhamel et al. 2010;Stein 2012). The visual identification of hadal snailfishes is particularly challenging, due to the frailty of the animals, the importance of internal characters such as vertebral counts and tooth morphology in identifying species, and the difference between in situ and surface environmental conditions. For example, when a hadal specimen is recovered at the surface, their gelatinous layer, or subdermal extracellular matrix, begins to quickly deteriorate, changing the body proportions of the animal considerably (Gerringer et al. 2017a). The limited number of specimens, coupled with physical changes from recovery and preservation, result in a very different appearance between in situ imaging and type specimens (Gerringer et al. 2017a). Recent molecular studies have uncovered systematic issues within the Liparidae, with multiple described clades such as Careproctus and Paraliparis Collett, 1879 being polyphyletic (Orr et al. 2019). Further, Notoliparis and Pseudoliparis, while geographically isolated, are closely related both morphologically and genetically and should likely be synonymized (Gerringer et al. 2017a;Orr et al. 2019).
In 2018, one specimen of an undescribed hadal liparid was collected from 6714 m depth by a baited trap in the Atacama Trench, southeast Pacific Ocean. Cameras recorded two additional hadal snailfish species with distinct morphologies between 5920 and 7608 m. The collected individual is only the second hadal liparid recovered in this region, the first being N. antonbruuni. Herein, we describe this species as Paraliparis selti sp. nov. using a combined morphological and molecular taxonomic approach. We show that this species belongs to the genus Paraliparis and represents an independent evolutionary lineage in the hadal zone, distinct from all currently described hadal snailfishes.

Sampling
As part of the HADES-ERC (European Research Council) project, cruise SO261 on the RV Sonne in March 2018, two identical free-fall landers were deployed 18 times between the depths of 2537 and 8052 m (Table 1; Fig. 1; full deployment list in Supplementary Table 1). The landers were equipped with purpose-built high-definition cameras and LED lighting. The cameras were powered by a 12V lead-acid battery (SeaBattery, Deep Sea Power and Light, San Diego, CA). Thirty second recordings were taken every 1.5 min, with an 8 megapixel still image taken between videos. Pressure and temperature were recorded at 30-s intervals using a separate sensor (RBRDuet; RBR Ltd, Canada). Specimens were collected using acrylic cylinders 20 cm in diameter and 100 cm long with inverted funnel entrances of 5-cm diameter. The cameras and traps were baited with locally sourced mackerel (Scombridae).
Additional snailfish species observed by in situ filming, without voucher specimens, are referred to by their morphotype and site locator tag as used in Jamieson et al. (2021) based on the open nomenclature (ON) qualifiers as described in Sigovini et al. (2016) and video-based classification standardisation as described in Horton et al. (2021).

Holotype processing
Upon collection, the holotype was immediately put on ice and transferred to a 4°C cold room for processing. Two fin-clips were taken from the distal portion of the right pectoral fin and preserved in 95% ethanol for genetic analysis. The specimen was photographed in an aquarium containing chilled seawater using a Canon EOS750D and Tamron 90 mm F/2.8 macro lens with a Hoya polarising filter.
The holotype was fixed in 4% buffered formalin for 4 months, before being transferred to 75% ethanol for final storage. The specimen was transitioned to ethanol with 24-h soak periods in freshwater, 25% ethanol, and then 50% ethanol. The right pectoral girdle was dissected (preserving the left side intact) and stained using alizarin red S and alcian blue (Fisher Scientific) (Taylor 1967;Dingerkus and Uhler 1977). Table 1 Lander deployments where snailfishes were observed arranged by increasing depth in meters. Latitude and longitude are given in decimal degrees. Snailfishes were not observed at deployments shallower than 5920 m or deeper than 7608 m. Snailfish observations are presented as N max (T arr ), where N max is the maximum number observed in the same frame and T arr is the first arrival time in minutes following lander touchdown. Data in bold denotes the deployment the holotype was captured. Morphotype names of the other snailfish follow Jamieson et al. (2021)  Post-preservation images were taken using the same photographic equipment with Helicon Remote and processed as an image stack using Helicon Focus (HeliconSoft 2017). The holotype was imaged using X-ray micro-computed tomography (micro-CT) at the Natural History Museum, London, using a Nikon Metrology HMX ST 225 (Nikon metrology Europe NV, Leuven, Belgium) with a tungsten reflection target head. Experimental conditions comprised a source voltage of 90 kV and current of 111 A using a 0.5-mm aluminium filter; the object-detector distance of 58.7 mm and the sourcedetector distance of 1115.1 mm generated data with a voxel resolution of 10.0 μm. As the specimen was not fitting in the vertical field of view with the magnification setting, 5 acquisitions were necessary, moving the sample with the motorised manipulator on the vertical axis. Each acquisition consisted of 3142 projections of 1415 msec exposure each, with a frame averaging of 2. Each individual acquisition was reconstructed using CT pro 3D 3.1 (Nikon metrology Europe NV, Leuven, Belgium), generating 32-bit volume. All 5 acquisitions were merged following the protocol described in Butler et al. (2022) and exported as a single stack of 16-bit tiff images. To improve the signal-to-noise ratio and reduce file size, the dataset was reduced with a 2×2×2 binning. Hypural plate and posterior fin rays were also observed using a Leica DMi8 inverted microscope and DFC295 camera.
Type material is deposited at the Museo Nacional de Historia Natural, Santiago, Chile (MNHNCL ICT 76227).

DNA extraction and PCR amplification
To support the description of this species, we applied an integrative taxonomic approach that included analysis of three mitochondrial barcoding regions, 16S rRNA (16S), cytochrome oxidase I (COI), and cytochrome b (Cyt-b). Total genomic DNA was extracted from a fin clip of the single individual of P. selti sp. nov. using a standard phenolchloroform approach. The three partial mitochondrial genes were amplified using published primers sets (Supplementary Table 2). PCR reaction mixes contained 0.2mM each dNTPs, 2.5mM MgCl 2 , 0.5μM each primer, 0.5U of Taq DNA polymerase (Bioline), 10ng DNA template in 1× NH 4 buffer (Bioline) in a total reaction volume of 20μl. PCR conditions were as follows: initial denaturation at 95°C for 2 min, followed by 35 cycles of denaturation at 95°C for 30s, annealing at 48°C (16S) or 52°C (COI, Cyt-b) for 30s, extension at 72°C for 1 min, before a final extension step at 72°C for 2 min. PCR products were purified enzymatically using ExoSAP-IT® (USB, Cleveland, OH, USA) as described in Bell (2008) and quantified by direct comparison with lambda DNA size standards on a 2% TBE agarose gel. Sequencing was undertaken with an ABI 3730xl automated DNA sequencer (MWG Eurofins, Germany).

Phylogenetic analyses
With the three barcoding markers, we constructed three datasets to infer the genetic relationship of P. selti sp. nov. to other snailfishes. Due to the absence of Paraliparis species with publicly available 16S and Cyt-b sequences, we largely focused on a COI dataset to investigate the placement of P. selti sp. nov. within the Liparidae. The Liparidae COI dataset (605 bp) comprised of the 56 species of Careproctus, Elassodiscus, Liparis, Paraliparis, Pseudoliparis, Nectoliparis, Notoliparis, and Rhinoliparis with available sequence data ( Table 2). Lumpfish (Family Cyclopteridae Bonaparte, 1831) are a commonly used outgroup for Liparidae phylogenies due to their close relationship with the family (Gerringer et al. 2017b;Orr et al. 2019). Here, we selected the smooth lumpfish, Aptocyclus ventricosus (Pallas, 1769), as the outgroup. In addition to the Liparidae COI dataset, we constructed two datasets for all snailfish species with available sequence data at each loci (16S, COI, Cyt-b, and concatenated). The first dataset comprised of a reduced sequence length with 10 Liparidae species, and the second dataset comprised the full sequence length with six Liparidae species. The methods and phylogenies constructed for other loci are presented in the Supplementary Information.
Electropherograms were viewed in MEGA X (Kumar et al. 2018) where primer sequences and any ambiguous bases were trimmed. Sequence identity was confirmed using NCBI BLASTn (Altschul et al. 1990). Nucleotide sequences for COI and Cty-b were translated into amino acid sequences to check for the presence of stop codons. Nucleotide alignments were constructed on the webPRANK server (Löytynoja and Goldman 2010) and confirmed by eye. The optimal evolutionary model for each dataset was identified by jModelTest 2.1.6 (Darriba et al. 2012) using both the Akaike information criterion (AIC) and the Bayesian information criterion (BIC). Both AIC and BIC identified the same best-fit models for the Liparidae COI dataset, Tamura and Nei 1993 model with gamma distribution and proportion of invariable sites (TN93+G+I) (Kimura 1981;Hasegawa et al. 1985;Tamura and Nei 1993).
Topologies were inferred using both maximum-likelihood and Bayesian approaches using PhyML v.3.1 (Guindon et al. 2010) and the Bayesian evolutionary analysis by sampling trees (*BEAST) software package v1.10.4 (Suchard et al. 2018), respectively. Maximum-likelihood analyses were  (2019) conducted with a neighbour-joining starting tree and using nearest neighbour interchange branch swapping using the model of sequence evolution estimated by JModelTest2 but with the parameters estimated by PhyML. The stability of nodes was assessed from bootstrap support based upon 10,000 iterations. Bayesian analyses were run for 40,000,000 generations and sampled every 100,000 trees with an uncorrelated relaxed clock (Drummond et al. 2006). The respective models of sequence evolution estimated by jModelTest2 were used but with the parameters estimated by *BEAST. Two independent runs were performed, and outputs were assessed with Tracer v1.7 to ensure convergence (effective sample size (ESS) >200). The two independent runs were combined in LogCombiner v1.8.4. with TreeAnnotator v1.8.4, the first 4,000,000 trees were discarded as burn-in where the partition frequencies among the remaining trees gave the posterior probabilities used to provide an estimate of clade credibility. Trees were visualised using FigTree v1.4.3 (Rambaut 2012) and annotated using Inkscape 0.92.2 (The Inkscape Team 2017). We paired the Liparidae COI phylogeny with species depth ranges to evaluate depth relationships between clades ( Table 2). The species depth ranges were compiled from FishBase (Froese and Pauly (2021) and checked in the primary literature (Chernova et al. 2004;Chernova 2008;Steinke et al. 2009;Dettai et al. 2011;Orlov and Tokranov 2011;Stein 2012;Linley et al. 2016;Stein 2016;Robertson et al. 2017;Solomatov and Orlov 2018;Gerringer 2019;Orr et al. 2020). While the species used within the Liparidae COI dataset are a good representation of known snailfishes, the absence of deep abyssal species is most likely an artefact of reduced sampling effort at these depths (Gerringer 2019).  Table 6 Common name -Blue Atacama snailfish Material examined HOLOTYPE -MNHNC ICT 76227, 76 mm standard length, juvenile, Atacama Trench, 21.74162°S, 71.25775°W, 6714 m depth, RV Sonne, cruise SO261, 24th March 2018 (Fig. 2).

Diagnosis
We place P. selti sp. nov. within the genus Paraliparis based on the following characteristics: ventral disk absent, pseudobranchs absent, single pair of nostrils present, single suprabranchial pore present, coronal pore absent (Stein 2012), and six branchiostegal rays present (Kai et al. 2020). Lower pectoral fin lobe does not form a single filament. The hypural plate is not divided and is fused with the terminal vertebral column. The head lacks barbels or head flaps.
It differs from its congeners through a complex of characters; high number of ;vertebrae (65) particularly the abdominal vertebrae (12), dorsal fin rays (60), anal fin rays (52), and caudal fin rays (8); comparatively low number of pectoral fin rays (18) which forms a deep notch with two widely spaced non-rudimentary rays; mouth horizontal; Jaw teeth simple and not uniserial; colour in life, blue-black anteriorly becoming pale posteriorly, blue pigment becoming dusky when preserved, peritoneum black.  Table 4). Head: Snout is deep, rounded, slightly projecting pre-preservation. Snout became bluntly rounded following preservation (Fig. 2, Fig. 3). Subrostral fold is present but does not cover the entire upper lip. The eye is low and does not touch the dorsal contour of the head. Sub-orbital distance long, 90% of the eye diameter to the closest part of the upper lip. The mouth is horizontal (upper jaw symphysis below the lower margin of the eye) and subterminal with the lower tooth plate behind the upper plate but strongly overlapping (Fig. 4). The jaw is relatively deep. Oral cleft reaching below the centre of the eye. Teeth are caniniform with the larger teeth towards the inside of the mouth. Premaxillary with prominent symphysis, in 8 rows of upwards of 9 teeth (Fig. 5a, Supplementary Figure 2). The mandibular tooth rows are less distinct (Fig. 5b). Dorsal pharyngeal jaw with well-developed teeth appearing to come together vertically in a rocking motion (Fig. 5c), a ventral pharyngeal jaw was not found and is not visible in Fig. 5 (Supplementary Figure 3). Six branchiostegal rays on each side, anterior two spaced from blade-like posterior rays. Pore-like single nostrils, level with the upper margin of the pupil. Pores are large, nasal pores 2, infraorbital pores 5, preoperculomandibular pores 7, temporal pores 2 (comprising 1 postorbital and 1 suprabranchial pore). Coronal pore is absent. Two oval chin pores are present, opening directly and separately on the skin surface with a crescent-shaped skin fold anteriorly but not extending posterior of the pores, interspace slightly wider than pore width, and unpigmented. The lateral chin profile is slanted at roughly 45°. Gill opening is restricted, completely above the pectoral fin, and covered by the opercular flap. The opercular flap is triangular with a blunt tip and moderate dorsally notched anterior base (Fig. 4). Body: The dorsal-most pectoral fin ray is in line with the ventral edge of the orbit. Upper pectoral fin lobe rays gradually decrease in length. There are no rudimentary pectoral fin rays, with 13 upper lobe rays, 2 notch rays, and 3 lower lobe rays. The pectoral girdle is long, extending almost to the mandibular symphysis (Fig. 4b). Rays of the upper pectoral fin lobe are close dorsally, becoming increasingly widely spaced following the 10th ray. The pectoral fin notch is wide with two equally spaced notch rays. The first ray of the lower pectoral fin lobe is twice as far from the second as the second is from Fig. 2 Paraliparis selti sp. nov. holotype, 6714 m, Atacama Trench. 83 mm total length, 75.9 mm standard length. a Freshly recovered. b Changes to anterior profile following preservation in 4% buffered formaldehyde and transfer to 75% ethanol the third (Fig. 6). The rays of the lower pectoral fin lobe have more segments (5) than those of the notch (2). The number of segments in the intact upper lobe rays is unknown. Dorsal and anal fin pterygiophores present to at least the last three vertebrae. Hypural plate fused but with a visible cleft between the plates. Twelve abdominal vertebrae, the anterior most two lacking haemal arches, last three with short haemal spines but distinct from the much longer haemal spines of the 53 caudal vertebrae. No pleural ribs. The first dorsal ray inserts between the neural spines of the 5th and 6th vertebrae, one or two interneuralia (free pterygiophors) are between vert 4 and 5 (Fig. 3d, Supplementary Figure 1). The anus is in line with the posterior edge of the opercular spine. Colouration: Body blue-black anteriorly becoming cream posteriorly in life. Peritoneum black. Skin partially transparent and subdermal extracellular matrix present but thin, potentially due to immaturity based on in situ observations of this morphotype. Preserved specimen dusky to anal fin origin, body pale in alcohol with black peritoneum (Fig. 3a).  Only a single round dorsal radial was observed (Fig. 6). However, the ventral radials are often reduced in Paraliparis (Andriashev 1998) and the absence of calcification of the radials may prevent them from being visible following clearing and staining (Stein 2005). As there is some doubt as to if all radials have been observed, the number and position of them will not form part of the diagnosis. Fenestrae are not present and are rare within the genus (Orr et al. 2019). Distal radials were also not observed, consistent with other Paraliparis species (Orr et al. 2019).

Distribution
Known only from the Atacama Trench (Fig. 1) between 5913 and 7616 m depth.

Etymology
Named for "selti", meaning blue in the almost extinct Kunza language of the indigenous peoples of the Atacama Desert. The waters over the Atacama Trench are especially productive, due in part to the mineral enrichment by dust deposits from this neighbouring Atacama Desert and aided by intense upwelling along the continental slope.

Comparisons
In addition to its significantly deeper habitat in the Atacama Trench, P. selti sp. nov. can be distinguished from shallowliving south-eastern Pacific Ocean Paraliparis in having two widely spaced notch rays in the pectoral fin, with a total of 18 pectoral fin rays and a small and restricted gill opening positioned completely above the pectoral fin (Stein 2005). Paraliparis selti sp. nov. is also distinct from shallower Paraliparis from the west coast of Central and South America in having a horizontal mouth, rather than the oblique mouth of P. debueni Andriashev, 1986, P. molinai Stein et al., 1991, P. angustifrons Garman, 1899, and P. membranaceus Günther, 1887. Of the deeper-living Pacific Paraliparis, P. selti sp. nov. has fewer pectoral rays than P. latifrons Garman, 1899 (22); P. fimbriatus Garman, 1892 (24); P. angustifrons Garman, 1899 (37); P. attenuatus Garman, 1899 (24); P. mawsoni Andriashev, 1986 (23-24) and P. plicatus Stein, 2012 (21). It also has fewer pectoral fin rays than are known from Australian Paraliparis (Stein et al. 2001). It has slightly more than P. thalassobathyalis Andriashev, 1982 (15-17) but is best separated from P. thalassobathyalis by having more caudal rays (8 vs. 5). Paraliparis selti sp. nov. has similar counts and proportions to the North Atlantic's P. calidus Cohen, 1968 andP. bathybius (Collett, 1879) but differs from the former in having more abdominal vertebrae (12 vs. 10) and the latter by having fewer notch rays (2 vs. 3-4). Paraliparis selti sp. nov. is distinct from P. paucidens Stein, 1978 andP. neelovi Andriashev, 1982 in that P. selti sp. nov. does not have uniserial teeth on either jaw. Paraliparis selti sp. nov. also differs from genetically similar species in having fully developed pectoral notch rays, as opposed to the fully rudimentary notch rays of P. garmani Burke, 1912, P. neelovi, P. voroninorum Stein, 2012, P. stehmanni Andriashev, 1986, P. longicaecus, P. camilarus, P. plicatus, and P. nullansa Stein, 2012 and the singular or partial rudimentary rays of P. posteroporus and P. parviradialis Stein, 2012. Fig. 4 Anterior morphology of preserved P. selti sp. nov. MNHNC ICT 76227, external left, internal micro-CT scan structure right: a dorsal, b lateral, and c ventral surfaces. Oral jaws (premaxilla and dentary) are highlighted in lilac, upper pharyngeal jaws in green, otoliths (saggittae large, lapilli, and asterisci also shown) in blue, and branchiostegal rays in red. Some low-density bone structures such as the neurocranium are present in the specimen but not shown at this density threshold. The scale bar represents 5 mm Fig. 5 3D rendering of the dentition in P. selti sp. nov. MNHNC ICT 76227 from CT data. a upper premaxillary; b lower dentary bone; c upper pharyngeal teeth. The density threshold has been adjusted in each to provide maximum detail. The scale bar represents 1 mm Morphologically, P. selti sp. nov. most strongly resembles Antarctic rather than Southern Pacific species. The relatively high counts for vertebrae, dorsal, and anal rays and a relatively low number of pectoral fin rays is common within what Andriashev (2003) called the long-tailed Paraliparis. Paraliparis selti sp. nov. most closely resembles the bathyal P. orcadensis Matallanas & Pequeño, 2000 but has fewer dorsal (60 vs. 68) and anal (52 vs. 60) fin rays, P. orcadensis is also described as uniform grey with nearly back snout and fins rather than P. selti sp. nov.'s variegated blue colour.
In situ recording of P. selti sp. nov. showed a high degree of variation in size and colour/pattern. Generally, however, P. selti is deep blue anteriorly becoming pale posteriorly in a blotchy and irregular manner (Fig. 7). Paraliparis selti sp. nov. possessed the widest bathymetric range of the snailfishes observed, from 5920 to 7608 m. The arrival time (T arr ) and maximum number (N max ) suggest that P. selti sp. nov. are most abundant around 6700 m depth (Table 1). They appear attracted to the mackerel bait but would not feed on it, instead feeding on the gathered amphipod swarm. No features were observed that differentiated recordings of this morphotype from the collected specimen, which forms the holotype for P. selti sp. nov. described herein.
The two additional snailfish morphs recorded in situ but not collected possess features which distinguish them from P. selti sp. nov. Liparid sp. indet. 2-PCT is a larger and more  robust morph. It is dusky pink in colour with a darker peritoneum visible through the fish's translucent skin (Fig. 8a). From the image data, it was possible to observe a welldeveloped ventral suction disk, a posterior anus near the anal fin origin, and a blunt and broad-based opercular spine. Photos also revealed that the pectoral fins of Liparid sp. indet. 2-PCT comprise at least 15+3+4 rays. The lower pectoral fin lobe rays are free and the tips of the notch rays are partially free. The notch is moderately deep with more widely spaced rays. The pectoral fins do not reach to the end of the peritoneum. The snout is short, about double the eye diameter, and the profile is rounded. The head and body are broad and gelatinous with thin and loose skin. This morph would often settle directly on the bait and suction feed on the amphipod swarm, becoming visibly engorged during the recordings.
The third snailfish morph seen in the Atacama Trench, Liparid sp. indet. 3-PCT, was easily distinguished by its large wing-like upper pectoral lobes, large head with small eyes, and prominent, sloped snout (Fig. 8b). This pale morph has been previously recorded in the Atacama Trench as 'unidentified liparid from 7050 m' in Jamieson (2015) and 'The Perú-Chile Snailfish' in Linley et al. (2016). The morphotype possesses at least 19+2+4 pectoral rays. The rays of the lower lobe are thick and free, curving slightly anteriorly. The rays of the notch are more widely spaced than those of the upper lobe; the notch is shallow and not clearly distinct from the lobes. The tips of the upper lobe and notch rays are free. The snout is prominent, and there is an abrupt change to the profile at the orbit. The centre line of the snout forms a ridge and may be due to an enlarged ethmoid (nasal) bone, as is found in the genus Eknomoliparis. The snout is more than five times the eye diameter. The body cavity is short, less than the head and pectoral fin lengths. This morph tended not to approach the bait as closely as the other morphotypes and would hunt amphipods in the periphery.

Phylogenetic analyses
Paraliparis selti sp. nov. was successfully characterised across three gene amplicons where 1476 bp were resolved for 16S, 1399 bp for COI, and 1007 bp for Cyt-b. These sequences have been annotated and deposited into GenBank and can be accessed under GenBank accession numbers MN423252 for 16S, MN422493 for COI, and MN422494 for Cyt-b.
More broadly within the liparids, the COI phylogeny suggests that Liparis is monophyletic while both Paraliparis and Careproctus do not form monophyletic genera (Fig. 9). Instead, there are three distinct Paraliparis sub-clades, which are polyphyletic because of Rhinoliparis. Careproctus is split into two distinct lineages by Paraliparis and Rhinoliparis, with a clade of C. attenuates, C. continentialis, C. georginus, and C. longipectolis being sister to Paraliparis. Further, the larger COI gene topology resolves the placement of other hadal snailfishes, Pseudoliparis and Notoliparis, relative to P. selti sp. nov. While P. selti sp. nov. is firmly placed within the Paraliparis, Pseudoliparis, and Notoliparis are rooted basal to the Careproctus, Paraliparis, and Rhinoliparis lineages (0.9 Bayesian posterior probability, unsupported ML bootstrap support).

Morphological characters of Clade M
The COI phylogenetic analysis (Fig. 9) placed P. selti sp. nov. within Clade M, as defined by Orr et al. (2019). Clade M is dominated by Antarctic species (14 of the 19 species), but surprisingly contains the Northern Atlantic species P. garmani; P. bathybius (Collett, 1879) and the Northern Pacific P. paucidens Stein, 1978; in addition to an undescribed species. They are all found deeper than~600 m with most species captured 1500-2000 m deep. The tooth shape is consistently simple (or unknown in two cases) but is dominated by uniserial tooth patterns

Discussion
The discovery of P. selti sp. nov. between 5920 and 7608 m represents another endemic and ecologically important hadal snailfish Linley et al. 2016;Gerringer 2019;Fig. 9 Bayesian tree showing the position of P. selti sp. nov. within the liparids using the COI barcoding region (a) with corresponding depth distributions compiled from FishBase and cross-referenced in the primary literature (b; Table 2). Bayesian posterior probabilities and Maximumlikelihood bootstrap supports are shown on branch nodes. Values less than 50% or values not supported by the alternative method are not stated or depicted by an asterisk. Names with asterisks signify revised identifications by Orr et al. (2019). Clade M within the phylogeny, as defined by Orr et al. (2019), is represented by bolded pink lines Jamieson et al. 2021). Further, this species provides insights into the evolution of snailfishes into the deep sea. The paths taken in the radiation and colonisation of the snailfishes, particularly around the Pacific Americas and Australia, are still not resolved (e.g. Orr et al. 2019). Genetic data from rare South American Pacific Ocean species, such as P. selti sp. nov., will ultimately aid our understanding of how snailfishes have evolved, adapted, and diversified into the greater depths of the deep sea. We applied molecular barcoding loci to examine the phylogenetic placement of P. selti sp. nov. within the Liparidae and confirmed its taxonomic placement based on morphology within the Paraliparis genus (Fig. 9). Mounting evidence strongly suggests that the genus Paraliparis is paraphyletic and in need of taxonomic revision (e.g. Kido 1988;Knudsen et al. 2007;Orr et al. 2019). Our molecular phylogenies support this assertion. Notwithstanding, our phylogenies did place P. selti sp. nov. within the main Paraliparis Clade M defined by Orr et al. (2019). Upon review, the members of Clade M possessed common traits. They were mostly found in the Antarctic at bathyal depths. Most show a reduction in dentition with uniserial simple teeth on at least one jaw. Counts were quite consistent, and they tended to be long-bodied, with high abdominal vertebrae counts. The inclusion of molecular data from multiple loci, such as those presented here, will help elucidate the phylogenetic relationships in Paraliparis and the family Liparidae overall. There is even mounting support for including the Liparidae within the separate Order Cottiforms, based on both morphological and molecular evidence (Imamura and Abe 2002;Wiley and Johnson 2010). To align with the currently accepted literature, we have continued to use Perciformes but acknowledge that Cottiforms should likely be adopted in the future.
The in situ behaviour of P. selti sp. nov. resembles that recorded for other hadal snailfishes (Fig. 7, Supplementary Video 3). P. selti sp. nov. are suction-feeding predators that prey on scavenging amphipods attracted to the bait (rather than a scavenger of the bait itself) consistent with other hadal snailfishes (Linley et al. 2016;Gerringer et al. 2017c;Linley et al. 2017). Amphipods were found within the holotype's buccal cavity upon recovery and the micro-CT data revealed its stomach contained many amphipods (Supplementary Figure 4). Strong pharyngeal jaws are prevalent within the hadal snailfishes and appear to be an adaptation to feeding on the abundant hadal amphipods (Gerringer 2019). Functional morphology and variation within liparid pharyngeal jaws, particularly in deep-sea representatives which tend to show a reduction in hard body parts, is well-worth continued study. With the continued adoption of techniques such as micro-CT, pharyngeal jaw morphology may become an important diagnostic character in the Liparidae and is likely to withstand the damage done by recovery and preservation.
Adaptive radiation of liparids into hadal trenches has been hypothesised to have occurred independently, with the hadal descent of an abyssal lineage in the northwest Pacific Ocean and a separate descent to the southwest Pacific Ocean (Priede and Froese 2013). In contrast, more recent genetic studies indicate that all previously known hadal snailfishes form a single clade, which is basal to other snailfishes (Gerringer et al. 2017b;Gerringer 2019;Orr et al. 2019), thus, more suggestive of a singular hadal lineage. The COI inferred phylogeny presented by this study strongly support a different case for P. selti sp. nov., which appears to represent an independent radiation into the hadal zone ( Fig. 9; Supplementary  Figure 5, 6). Those species genetically most closely related to P. selti sp. nov. are also predominantly from the Southern Ocean, where recent in situ observations by Jamieson et al. (2021) reported three hadal snailfish morphs very closely resembling those recorded in the Atacama Trench. Future research may reveal an Antarctic origin to this second independent radiation into the hadal zone. The singular specimen of P. selti sp. nov. may prove important to our understanding of both snailfish radiation and their adaptations to life at hadal depths.

Conclusion
The snailfishes are among the most rapidly radiating families of marine fishes, resulting in a worldwide distribution from the coastal waters to the hadal trenches. Due to the challenges of sampling at depths >6000 m, the full diversity of snailfishes remains unresolved. Paraliparis selti sp. nov. represents an important new species which confirms a second and independent lineage of vertebrate life adapted to hadal depths. This progresses our understanding of the colonization and diversification of snailfishes into hadal trenches and the factors that allow this family to live in the ocean's deepest zones.
London; Marcus Anders Krag and Peter Rask Møller of the Natural History Museum of Denmark. We thank Natalia Chernova of the Zoological Institute, Russian Academy of Sciences, and Sandra Raredon and Jeff Williams of the Smithsonian Institution for photographs, x-ray, and discussion of types when they were not possible to view in person. The global network of natural history museums is an invaluable scientific resource. This work would not be possible without such places, both as repositories of specimens and highly talented people.
Funding Advanced European Research Council grant HADES-ERC "Benthic diagenesis and microbiology of hadal trenches" grant (agreement number 669947) awarded to Ronnie N Glud (University of Southern Denmark). Ship time provided by BMBF, Germany.

Declarations
Conflict of interest The authors declare no competing interests.
Ethics approval All applicable international, national, and/or institutional guidelines for the care and use of animals were followed.
Sampling and field studies All necessary permits for sampling in Chile's Exclusive Economic Zone were obtained by the authors from the Servicio Hidrográfico y Oceanográfico de la Armada de Chile. The study is compliant with CBD and Nagoya protocols.
Data availability Genetic sequences from the present study are available on GenBank (MN423252 for 16S, MN422493 for COI, and MN422494 for Cyt-b). Micro-CT data is available for download through MorphoSource: https://www.morphosource.org/concern/media/ 000417515. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .