The community composition data underlying this study were collected in the CBL, north of Alaska (7–78°N, 158–165°W) onboard USCGC Healy in July–August 2016 (Figure 1). The CBL extends from the Chukchi shelf into the Canada Basin, covering a depth gradient from ∼300 to 3000 m. It consists of the Northwind Ridge and Chukchi Plateau where stations were grouped as “mid-depth” (486–1059 m), and of the isolated Northwind Basin where stations were grouped as “deep” (1882–2610 m) (Figure 1 and Table 1; Jakobsson et al., 2008; Mayer et al., 2010). Waters of Arctic, Atlantic, and Pacific origins interact in the CBL study area with Pacific-origin water comprising the Polar Mixed Layer and upper halocline (McLaughlin et al., 2004; Steele et al., 2004; Woodgate, 2013). The lower halocline is of Atlantic origin, arriving from Fram Strait and the Barents Sea (Woodgate and Aagaard, 2005). Underneath it and characterizing the “mid-depth” stations is the Atlantic water layer (McLaughlin et al., 2004; Woodgate et al., 2007; Bluhm et al., 2015), while “deep” stations are in the Arctic Ocean deep-water layer originating from the Greenland Sea and spreading across the Eurasian Basin to the Canada Basin (Bluhm et al., 2015).

Epifauna (including invertebrates and demersal fishes) was sampled with the ROV Global Explorer (Oceaneering International), which performed a photographic survey of the seafloor at ten stations (Table 1), as described in Zhulay et al. (2019). 24-megapixel still images were collected with a downward-looking DSSI DPC-8800 digital camera along transects every 5–8 s. Four digital laser pointers, one located at each corner of a fixed distance of a 10-cm square, served as a size reference for the imaged area and size of organisms at four stations (stations 1, 6, 7, and 8), after which they stopped functioning. In addition, epifauna was sampled with a single trawl sample at six stations (stations 1, 2, 3, 9, 10, and 13, Table 1) using a 3.05 m modified plumb staff beam trawl (Abookire and Rose, 2005) equipped with a 7 mm mesh net with 4 mm in the cod end. Ca. 30 min hauls at ∼1.5 knots speed over ground were taken with bottom time estimated from a time depth recorder (Star Oddi) affixed to the net. Organisms were sorted, identified to the lowest possible taxonomic level, and counted. Select taxonomic vouchers were further identified by expert taxonomists (see section “Acknowledgments”) and taxon names were verified using WoRMS (http://www.marinespecies.org/, on September 10, 2020). The proportional abundance of each taxon was then calculated for each trawl station.

At each station, near-bottom water temperature and salinity were measured with a SBE9/11 + CTD at ∼20 m above the bottom. Sediment surface samples (0–1 cm) from box core samples were taken and frozen at −20°C for later determination of grain size composition, carbon and nitrogen content, and concentration of sediment chlorophyll a and phaeopigments (Zhulay et al., 2019). Sediment grain size was analyzed from samples pre-treated with HCl and H₂O₂, to remove calcium carbonate and organic material, on a Beckman Coulter Particle Size Analyzer LS 13320 at the Geology Laboratory of UiT The Arctic University of Norway in Tromsø. Sediment organic carbon and nitrogen (%) were determined on a Costech ESC 4010 elemental analyzer at the stable isotope facility at the University of Alaska Fairbanks (UAF). The C/N ratio, an indicator of food quality with higher values indicating lower food quality (e.g., Dorgelo and Leonards, 2001; Iken et al., 2010), was then calculated for each station. Concentrations of sediment chlorophyll a and phaeopigments (μg pigment/g dry sediment) were measured on a Turner Designs TD-700 fluorometer after pigment extraction with 5 ml of 100% acetone for 24 h in the dark at −20°C at UAF. The fluorescence of the sample was read before and after acidification with HCl (final concentration of HCl was 0.003 N) for determination of phaeopigments (Arar and Collins, 1997; Jeffrey and Welschmeyer, 1997).

A subset of the useable ROV images of the sea floor were manually analyzed from each station (39–180 per station, 940 images in total) (Table 1, Zhulay et al., 2019). Image processing and analyses were performed with the ImageJ¹ (Rasband, 2009). Taxa were identified to the lowest possible level based on a combination of morphological features visible on the ROV imagery, the voucher collection from trawls, and additional identifications by taxonomic experts (see acknowledgments). Taxa that could not be identify were excluded from this analysis due to difficulties related to assigning trait modalities to these organisms. All taxa present on the images were counted per image and proportional abundance of taxa per station was calculated. Rocks larger than two cm were counted and the average number of rocks per picture was included in the statistical analyses as an environmental factor.

Nine commonly used traits represented by a total of 39 modalities were chosen for the present analysis following established definitions by Bremner et al. (2006), Costello et al. (2015), Degen et al. (2018), and Sutton et al. (2020) (Table 2). The traits reflected morphology (adult size, body form), behavior (living habitat, mobility, adult movement, feeding habit, substrate affinity) and life-cycle characteristics (larval development and reproduction) (Table 2; reviewed by Martini et al., 2020a). For the purpose of this study, modalities of larval development trait were based on the published concept for Arctic traits analysis (Degen and Faulwetter, 2019) that assumes that planktotrophs disperse farther than lecithotrophs although many exceptions are known to occur in the deep sea (Young, 2003). Every trait was coded for every taxon identified based on: (1) observations made from trawl-collected material during the cruise and/or from ROV images (i.e., traits directly measured in situ, also referred as realized traits, c.f. Martini et al., 2020a) for size, body form, adult movement, living habit, and substrate affinity or (2) information inferred from published literature (referenced in Supplementary Table 1), online traits databases (e.g., polytraits, Faulwetter et al., 2014; the Arctic Traits Database, Degen and Faulwetter, 2019) and relevant web pages (e.g., FishBase², Sea Life Base³) (i.e., traits acquired from other sources, also referred as potential traits, c.f. Martini et al., 2020a) for the rest of the traits. The size of organisms was measured on board from specimens collected at each station in the trawl samples or from the ROV images with the ImageJ software (Rasband, 2009). Size measurements from ROV images were possible at the four stations where the digital laser pointers were functioning and where the positioning of a given organism was suitable for those measurements. The average adult size of a given species across all stations was used for the analysis. As information about biology and behavior of many epifaunal taxa in the Arctic deep sea remains limited or is at times non-existent at the species level, coding of these taxa was conducted based on information available for closely related species in the same genus or family (following, e.g., Faulwetter et al., 2015; Rand et al., 2018; Sutton et al., 2020). In a few cases, modalities common at even higher taxonomic rank (such as direct development in Peracarida) were applied. For coding a “fuzzy coding” procedure (Chevenet et al., 1994) was used, resulting in a “traits by taxon” matrix (Supplementary Table 2). The “fuzzy coding” procedure allows taxa to be coded with multiple modalities to different degrees using a 0–3 code, with 0 indicating no affinity, 1 and 2 indicating partial affinity, and 3 indicating the highest affinity for a given modality. This approach was proposed to account for variation encountered within a species (Chevenet et al., 1994) and when incorporating information from species in the same genus or family (Charvet et al., 2000). While including higher taxonomic levels may introduce uncertainty to the results, their suitability has been documented for biological traits analysis as well as taxonomic community analysis, especially in multivariate analyses that showed that the functional structure of communities could be conserved (Bournaud et al., 1996; Bowman and Bailey, 1997; Dolédec et al., 1998). To give the same weight to each taxon and trait, the fuzzy codes (0–3) were converted to proportions for each trait modality totaling to 1 (e.g., Bolam et al., 2017).

A total of 106 invertebrate and fish taxa were used for BTA, of which 53 taxa occurred in the ROV images and 77 taxa in trawls with 26 taxa common to both sampling gears. In addition to the “traits by taxon” matrix, “taxa by stations” (i.e., presence/absence of taxa or proportional abundances of taxa at each station) and “traits by stations” (i.e., trait composition at each station, obtained by multiplying the “taxa by stations” and “traits by stations” matrices and indicated by presence/absence or proportional abundance weighted scores) matrices were generated (following Beauchard et al., 2017; Degen et al., 2018). Three “taxa by stations” (Supplementary Tables 3–5) and three “traits by stations” (Supplementary Tables 6–8) matrices were compiled for subsequent analysis: presence/absence of taxa based on ROV and trawl samples combined across all stations, and one each with proportional abundances acquired from either ROV or trawl samples. Proportional abundance was chosen over absolute abundance due to the above-mentioned failure of the laser pointers that made it impossible to calculate absolute abundances for all ROV stations. Proportional abundance was also used for trawl samples for consistency.

To test our first hypothesis, namely that epifaunal organisms in the CBL are predominantly small-sized, non-sessile deposit-feeders or scavengers with equal representation of direct and indirect development, we included the four traits: size, larval development, adult movement and feeding habit. The “traits by stations” matrix was used to test this hypothesis; it was based on presence/absence data of taxa collected with both the ROV and the trawl (Supplementary Table 6).

To test the second hypothesis, namely that a difference in functional structure exists between mid-depth and deep communities, all nine traits were used. To investigate differences in the overall functional composition of epifauna and visualize potential differences between these two depth strata, we applied a fuzzy correspondence analysis (FCA, Chevenet et al., 1994). FCA is an extension of a regular correspondence analysis that is suitable for fuzzy coded traits data of discrete variables (Chevenet et al., 1994). FCA is based on the “traits by stations” matrix (Supplementary Tables 7, 8) and identifies and visualizes traits and their modalities contributing most to the difference in the functional structure among stations (Bremner et al., 2006), and provides correlation ratios of each trait along the fuzzy principal axes, representing the amount of variance of a certain trait modality explained by a given axis (Chevenet et al., 1994). Correlation ratios greater than 10% were considered as the traits contributing most to variation among the stations following Conti et al. (2014). A Kruskal–Wallis test was used to test for significant differences in proportional abundance of trait modalities between mid-depth and deep stations using the “traits by stations” matrix from ROV data (Supplementary Table 7). Only one trawl sample was available from deep stations preventing statistical comparisons.

We then visualized which of the available environmental variables explained most of the variation in the functional structure of the epifaunal communities using a canonical correspondence analysis (CCA) performed on the “traits by stations” matrices for ROV (Supplementary Table 7) and trawl (Supplementary Table 8) samples. For environmental data, we included water depth, bottom water salinity and temperature, grain size composition, number of rocks in ROV images, concentration of benthic pigments in sediment (phaeopigments and Chl a), carbon content in sediment, and C/N ratio. A forward selection procedure was used to identify environmental variables explaining most of the variability in the trait-by-station data. These variables were then used in the model, whereas other factors were overlaid on the plots as passive factors. The significance of the models and environmental variables were tested with Monte Carlo permutation tests (Oksanen et al., 2013).

As part of hypothesis two, we tested for differences in functional diversity and redundancy between depth strata for both ROV and trawl-based data. FD was estimated using Rao’s quadratic entropy (Rao’s Q), which is a measure of trait dissimilarity (Rao, 1982; Botta-Dukát, 2005). Rao’s Q ranges from 0 to 1, where 0 means low FD (i.e., communities are the same in their biological trait profiles) and 1 means high FD (i.e., communities are unique in their biological trait profiles) (Van der Linden et al., 2016). FD was calculated based on the “traits by taxon” (Supplementary Table 2) and “taxa by stations” (Supplementary Tables 4, 5) matrices.

FR is the relationship between FD and species diversity (Ricotta et al., 2016), and was calculated as the ratio of FD to the taxonomically based Simpson index (D, calculated using equation (1)).

(1) D = 1 - ( ( Σ n ( n - 1 ) / N ( N - 1 ) ) ,

where n is the total number of organisms of a particular species, and N is the total number of organisms of all species. In order to obtain a regularly increasing index, the formula was converted to: 1–(FD/D) (Van der Linden et al., 2016). FR defines to which degree different species represent the same ecosystem functions (Petchey and Gaston, 2006; de Bello et al., 2007) and ranges from 0, where all species have different trait-categories, to 1, meaning all species display the same trait-categories (de Bello et al., 2007). A Kruskal–Wallis test was used to check for significant differences in FD and FR between mid-depth and deep ROV stations.

All statistical analyses were performed using the software R (R Core Team, 2017) with the package ade4 (Dray and Dufour, 2007) for the FCA and calculation of FD index, and the package vegan (Oksanen et al., 2013) for the CCA analyses. A schematic representation of the hypotheses tested and methods used to test the hypotheses, along with figure and table numbers representing results of the tests, is given in the Supplementary Figure 1.

The body size modality “small-medium” was the most frequent in the epifauna across the study area, while size “large” was rarest (Figure 2A). The most frequent larval development was “direct” followed by “lecithotrophic,” while the occurrence of the modality “planktotrophic” was much lower (Figure 2B). Non-sessile adult movement modalities combined were more frequent than sessile forms. Individually, “crawlers” were dominant, though “sessile” was second most frequent and since this study focused on epifauna, “burrowers” were expectedly least frequent in the data set (Figure 2C). Feeding habit “predators” was most frequent, followed by “suspension feeder.” The least frequent modalities of feeding habit were “parasite/commensal” and “subsurface deposit feeder” (Figure 2D).

The FCA showed substantial variation in functional composition across all stations. The first two axes of the FCA accounted for 81.9% of the variability in distribution of trait modalities, with 51.3% for the first and 30.6% for the second axis (Figure 3A). Most of the variation along the first axis was explained by body form (BF) (31%), reproduction (R) (20%), living habit (LH) (14%), feeding habit (FH) (18%), and substrate affinity (SA) (19%) (Table 3).

Trait modality composition generally differed between mid-depth and deep stations (Figure 3A). Deep stations 7, 12, and 13 were located on the upper right hand side of the FCA plot and corresponded to higher proportional abundance of modalities “swimming” (MV4), “sexual-internal fertilization” (R3) and “planktotrophic larval development” (LD1) (Figures 3A,B). Mid-depth stations 6, 9, and 10, located on the lower left hand side of the FCA plot, were characterized by higher proportional abundance of “upright body form” (BF5), affinity for hard substrate (SA2) and “predators” (FH5) (Figures 3A,B). Variation along the second axis was driven mostly by living habit (LH) (32%) and mobility (MO) (10%) (Table 3). These traits separated mid-depth stations 1 and 2 from the rest of the stations (Figure 3A), and the two stations were characterized by higher proportional abundance of modalities “low mobility” (MO2), “sessile” (MV1), “deposit feeding” (FH1), and “tube-dwelling” (LH3) (Figure 3B). Deep station 11 and mid-depth station 8 were not differentiated from the remaining stations based on the traits used. In general, results for the trawl data supported those from the ROV (Supplementary Figure 2 and Supplementary Table 9). The difference was that higher proportional abundance of modalities “sessile” (MV1) and “small size” (S1) was observed at the single deep trawl station, while mid-depth trawl stations were characterized by high proportional abundance of modalities “free-living” (LH1), “crawlers” (MV3), “dorsoventrally compressed” (BF3), and “medium size” (S3) (Supplementary Figure 2).

Results of the Kruskal–Wallis test indicated significant differences between mid-depth and deep stations for body form (BF), feeding habit (FH), reproduction (R), and lifestyle and mobility traits [adult movement (AM), living habit (LH), and mobility (MO)]. Deep stations were characterized by significantly higher proportional abundance of the body form “globulose” (BF1), feeding habits “suspension feeder” (FH3) and “opportunist/scavenger” (FH4), reproduction “sexual-internal fertilization” (R3), and higher proportional abundance of “free-living” (LH1), “highly mobile” (MO4), and “swimmer” (MV4) modalities (Kruskal–Wallis test; Table 4 and Figure 4). Mid-depth stations had significantly higher proportional abundance of the body form “upright” (BF5), living habit “tube-dwelling” (LH3), and mobility “slow movement” (MO2) (Kruskal–Wallis test; Table 4 and Figure 4). The Kruskal–Wallis test was not run on trawl data because only one deep station was sampled by trawl, but boxplots for the trawl data indicated generally similar patterns to those from the ROV data. Differences included higher proportional abundance of modalities “attached” and “sessile” at the single deep trawl station than at deep ROV stations, and higher proportional abundance of modalities “free-living” and “crawling” at the mid-depth trawl stations (Supplementary Figure 3).

Significantly lower FD and higher FR were found for deep ROV stations compared with mid-depth stations (p = 0.03 for both, Figure 5). Though the lack of station replication prevented statistical analysis for the trawl samples, the trends in FD and FR were similar to the ROV results, though the difference in FD and FR between the two depth strata was less distinct (Supplementary Figure 4).

Results of the CCA for the ROV samples showed that depth and sediment organic carbon content were the most important factors explaining variability in the distribution of trait modalities (Table 5). These environmental factors explained 64% of the total variation. Depth was positively associated with “dorsoventrally” compressed (BF3) and “globulose” (BF1), “internal fertilization” (R3), “free-living” (LH1), “swimming” (MV4), and “suspension-feeding” (FH3), modalities (Figure 6). High carbon content was positively associated with the modalities “vermiform” (BF2), “sexual-external fertilization” (R2), “sexual-brooding” (R4), “tube-dwelling” (LH3), and “surface deposit-feeding” (FH1) (Figure 6).

Results of the CCA for trawl samples also indicated depth as the most important factor influencing the functional composition of epifaunal communities, with temperature also being a significant factor (Supplementary Table 10). Both factors were used to constrain the CCA, resulting in 92.9% of the total variation explained. In contrast to the ROV data, depth was positively associated with “sessile” (MV1) and “attached” (LH6), and substrate affinity “hard” (SA2). Temperature was positively associated with “predators” (FH5) and “opportunist/scavenger” (FH4) modalities, and substrate affinity “biological” (SA3) (Supplementary Figure 5).

Body size of organisms affects many ecological functions including energy and nutrient cycling, and secondary production (Degen et al., 2018). One of the most common characteristics of deep-sea benthos is the small size of most species (Rex and Etter, 1998). Our study results are consistent with this paradigm and hence with our hypothesis, in that the second smallest (small-medium size, 10–50 mm) organism category had the highest occurrence, while large organisms (≥50 mm) had the lowest occurrence in epifaunal communities across the CBL. Low occurrence of the smallest size category is unsurprising given that we targeted epibenthic megafauna (typically ≥10 mm). A series of studies reporting reduced average body size with depth for deep-sea meiofauna (Soltwedel et al., 1996; Soetaert et al., 2002; Kaariainen and Bett, 2006) and macrofauna (Rex et al., 1999; Kaariainen and Bett, 2006) generally support Thiel’s size-structure hypothesis (Thiel, 1975) for these groups. This decrease in body size with depth has also been found for epifauna (Rex et al., 2006; Wei et al., 2010). Opposite to this trend, some deep-sea taxa with larger body sizes than in shallow areas have also been documented (Rex and Etter, 1998), in some cases resulting in gigantism. This phenomenon, often attributed to low temperature and high oxygen availability that causes slow growth rate and longevity (Shirayama and Horikoshi, 1989), has been found for deep-sea isopods, amphipods, pycnogonids, ostracods, and anemones (Timofeev, 2001; Danovaro et al., 2014) but in our study, only the very large pycnogonid Colossendeis proboscidea could fit this concept.

Overall, we confirm our hypothesis that the majority of epifauna found in our study was non-sessile; most were crawlers but swimmers were also found. Not unexpectedly, burrowers were less common, given the focus of the study was epifauna. As a consequence of the ability to move organisms can escape from disturbance (natural or anthropogenic), disperse or migrate (Beauchard et al., 2017; Degen and Faulwetter, 2019) and increases the chance of finding scarce and patchy food compared to sessile or less mobile organisms. Still, movement rates of epibenthic megafauna are generally lower in the deep sea compared to shelves (Thistle, 2003; Ruhl, 2007). For example, deep-sea brittle stars and holothurians move at 1–3 cm min^–1 and 1–2 cm min^–1, respectively, compared with 15–45 cm min^–1 and 7 cm min^–1, respectively, in shallow waters (summarized in Thistle, 2003). When stimulated, for example by food, however, many deep-sea animals can move faster (Premke et al., 2006; MacDonald et al., 2010; Taylor et al., 2016). For example, we observed unusual swimming behavior in the brittle star Ophiostriatus striatus, perhaps an adaptation to access patchy food falls (Boetius et al., 2013). ROV observations such as ours, hence, increase our often scarce knowledge of traits of deep-sea taxa. Besides mobile taxa, we did also find an unexpectedly high occurrence of the modality sessile in our study, especially obvious in ROV imagery. Sessile taxa in our study area, including ascidians, sponges, stalked cirripedes and crinoids, and zoanthid and nephtyid cnidarians were in part present on the numerous drop stones providing hard substrate for these organisms (Zhulay et al., 2019).

Trophic structure of the Arctic benthic deep-sea communities is poorly studied (but see Iken et al., 2005), though feeding habits influence energy flow, nutrient cycling, secondary production, organic matter decomposition, and nutrient regeneration (Bremner, 2008; Degen et al., 2018). We do know that the major food source is organic detritus originating mostly from the upper productive zone (Fabiano et al., 2001; Thistle, 2003). This organic material often undergoes strong transformation while sinking, decreasing nutritional value and particle size (Thistle, 2003). The paradigm that deposit feeding is among the best strategies to collect and process this organic detritus efficiently (Thistle, 2003) has indeed been supported for both macrofaunal and megafaunal deep-sea communities (Kröncke, 1998; Iken et al., 2005; Rex and Etter, 2010). The deposit feeders were also common yet not dominant in our study area, only partly confirming our hypothesis. Predators and suspension-feeders, however, were more common among our epifaunal taxa. This is contrary to a trend of decreasing proportions of predatory asteroid and gastropod species with depth (Carey, 1972; Rex et al., 1990). Our findings, however, are in agreement with other studies that found that predation is in fact common in oligotrophic seas or in areas with little food input (Kröncke and Türkay, 2003; Wieking and Kröncke, 2003; Vacelet, 2008), and are supported by high nitrogen isotope values in certain taxa of our study area (Iken et al., 2005). One theory states that prey can be more easily detected in the deep sea compared to shallow water environments, as “flow in the benthic boundary layer is slow and thus chemical gradients and pressure waves produced by prey should be more persistent and provide better information for prey location” (Thistle, 2003). Facultative predation is even known for specific deep-sea species of, for example, sponges (Vacelet, 2006; Godefroy et al., 2019; Martini et al., 2020b), and bivalves (Morton, 2016; Morton and Machado, 2019), suggesting feeding modes may be unusual, highly plastic and require more study. A stable isotope-based assessment for the study area is ongoing and will provide more clarification of the species’ feeding modes and trophic levels.

The high occurrence of suspension feeding taxa among the CBL epifauna was initially surprising. Higher current and particle fluxes, proving food for suspension feeders, tend to occur on elevations and slopes (Clark et al., 2010) including the Chukchi Slope Current in the CBL (Corlett and Pickart, 2017) and were suggested to provide food for suspension feeders from the nearby Chukchi shelf to the Northwind Ridge (Bluhm et al., 2005). In addition, suspension feeders on the above-mentioned drop stones can extend above the substrate and into the benthic boundary layer, where the currents are slightly faster and carry food particles (Vogel, 1996). Besides on stones, some taxa (in particular anthozoans) were elevated above the seafloor in other ways, namely either on stalks of crinoids or polychaete tubes or, in the case of most hormathiid anemones (Hormathia spp., Allantactis parasitica) on gastropod shells (usually Colus spp.).

Scavengers were also represented in the study area, although not as highly occurring as reported in some deep-sea areas including the Eurasian Arctic (Klages et al., 2001; Premke et al., 2006). It is unsurprising that parasites/commensals were the least occurring feeding type in the study area given our study focused on epifaunal megafauna of which few are parasitic. Smaller external and internal parasites are in fact occurring in Arctic megafauna, especially in demersal fish (Klimpel et al., 2006), but were invisible on the ROV images. We did encountered taxa such as ribbon worms, isopods and sea leeches, which generally contain parasitic forms on fishes and arthropods (Køie, 2000; Mantelatto et al., 2003; Ravichandran et al., 2009), but we did not observe them on a potential host. Commensalism was encountered for some hormatiid anemones attached to shells of gastropods, a widespread strategy increasing probability of contact with food particles, while providing protection to the host (Buhl-Mortensen et al., 2015). Similar commensal relationships were observed for other anthozoans and the amphipod Amathillopsis spinigera that were often found in association with sessile tubeworms and stalked crinoids. Clearly, more research is needed on parasitic and commensal biotic interactions in the deep-sea.

Little is known about larval development in the Arctic Ocean in general. Recent molecular studies, however, have documented the presence of pelagic larvae of more species than previously acknowledged for the Arctic (Ershova et al., 2019) and the deep sea (Kersten et al., 2019), and detailed studies in the deep sea have added species-specific observations (e.g., Mercier and Hamel, 2008; Martinez and Penchaszadeh, 2017; Montgomery et al., 2017). This trait is important for ecological functions such as dispersal, recolonization, recovery, tolerance to stress, and link between pelagic and benthic realms (Degen and Faulwetter, 2019). In the present study, direct development dominated as a single modality, yet given the sparse species-specific literature, generalizing this conclusion for the Arctic is premature. Nevertheless, advantages of this development type include, for example, protection from various unfavorable environmental conditions in the pelagic realm and settling on unfavorable substrate, and experiencing little planktonic predation. Most important for the deep sea, juveniles are less dependent on either limited or variable food availability (Mileikovsky, 1971). Yet indirect development, including planktotrophic and lecithotrophic larvae, was almost equally prevalent in the epifaunal taxa of our study area, which, in general, supports our hypothesis. This finding is consistent with a growing number of studies documenting the occurrence of pelagic larvae in both polar waters (Schlüter and Rachor, 2001; Fetzer and Arntz, 2008; Kuklinski et al., 2013; Brandner et al., 2017; Ershova et al., 2019) and deep-sea areas (Scheltema and Williams, 2009; Kersten et al., 2019). Among larval development types, lecithotrophs were most common in our study area. This is similar to findings in the NE Greenland, the deep-sea of the NE Atlantic, and Antarctica, where more than 70% of echinoderms were found to reproduce with pelagic larvae, the majority of which were lecithotrophs (Pearse, 1994). Development with pelagic larvae allowing dispersal over broader areas is an advantage, in particular for species with limited mobility (Fetzer and Arntz, 2008; Stübner et al., 2016), which were found in high numbers in our study area. In a work by Mercier and Hamel (2008), depth-related shifts in life history strategies and a simultaneous combination of brooding and broadcast-spawning with lecithotrophic larvae were reported in a deep-sea asteroid. This finding also stresses the need to species- and habitat-specific work to help close many knowledge gaps that currently limit final conclusions on true diversity and plasticity of life-history traits in deep-sea benthos.

In summary, our investigation of functional traits of deep-sea epifauna from the CBL area generally supported our first hypothesis that small, non-sessile organisms are the most common, with a relatively equal proportion of direct and indirect (mostly through lecithotrophic larvae) development. The hypothesized predominance of deposit feeding, however, was not found in the observed species pool, though that feeding mode was more prominent in the proportional abundance-weighted data set. That modality is common in infaunal taxa, which we did not cover here (Gage and Tyler, 1991; Iken et al., 2001; Mamouridis et al., 2011). Our analysis of trait modalities highlights instead that there is no single way to live successfully under deep-sea conditions, but rather that, similar to shallower areas, multiple strategies are in fact viable.