Introduction

Seamounts are isolated subaquatic mountains of (mostly) volcanic origin that rise at least 1000 m above the surrounding seafloor1. With a global abundance of ~ 10,0002 to ~ 125,0003 seamounts, they cover a minimum ~ 8,000,000 km22 and form one of the largest biomes on our planet4. Seamounts are often hotspots of deep-water biodiversity5,6,7,8 and can support higher species abundances than surrounding continental margins and continental slopes9. This phenomenon is known as the ‘seamount oasis hypothesis’9 that asserts that benthic invertebrates occur in higher densities and biomasses on seamounts than in other habitats in the deep sea9. Seamounts can also influence the overlying water column and affect the microbial community10,11, phytoplankton12, zooplankton13, and ultimately large fish1, which is known as the ‘seamount effect’13.

Regional variations in benthic assemblages among seamounts can be driven by differences in latitude14,15,16, longitude17, food supply18, water depth15,16,19,20,21, or distance from shore22. Structuring factors can also be region-specific, for instance Boschen et al.23 identified magnetivity (i.e., a proxy for hydrothermal activity24) as main driver of differences in the benthic composition among three seamounts in New Zealand. Overall, many factors known to drive biological community variability in the deep sea are related with water depth (i.e., temperature, pressure, oxygen concentration, or food-availability), which restricts most benthic fauna to a limited bathymetric range25. As a result, available habitat for a given population or community can be fragmented across the seamounts (and continental slopes, e.g.,26) within a region27, particularly in areas with high variability in water depth at summits. In addition, most deep-sea benthic fauna are thought to exhibit a biphasic life cycle between the release of pelagic planktotrophic or lecithotrophic larvae (i.e., respectively plankton-, or yolk-feeding, e.g.,28) to the water column for dispersal and subsequent settlement on the substrate for benthic/sessile adult stage. As such, connectivity between species populations on different seamounts is thought to be largely controlled by regional factors affecting larvae transport, development, or resilience, such as food availability and temperature within the water column, hydrographic retention mechanisms, or the presence of suitable habitat where propagules ultimately settle27. However, regional gradients and larval dispersal dynamics can be strongly modulated by small-scale processes across seamounts23,27,29,30. Consequently, integration of local to regional observational scales is often essential to accurately assess spatial patterns in seamount communities.

Variations in benthic assemblages within seamounts have been related to changes in seabed composition (e.g., hard substrate availability)16,19,23,31, slope16,32, currents32,33, or food supply32; all factors that typically covary with water depth19,23,33,34. As a result, community composition tends to be depth-stratified within seamounts6,27. However, essential niche requirements, such as a minimum rate of food supply35 or the presence of hard substrate36,37 for many sessile taxa, can be strongly modulated by the topographical complexity of a seamount in interaction with the surrounding water masses38. Similarly, aggregations of framework-building fauna can generate a habitat for other species, such as is the case with cold-water coral reefs35,39 and sponge grounds40,41. Acting as ecosystem engineers, they enhance local habitat heterogeneity rates and thereby increase alpha diversity rates42. Given this wealth of possible drivers operating locally, and cumulative effects from their interactions, benthic assemblages within a seamount tend to exhibit rapid shifts in composition across space43, usually reflecting the ranges of one or more environmental gradients that define the boundaries of different habitat types.

There are likely more than 300 seamounts beneath Arctic waters44, but only a few have been the subjects of ecological investigations. Sponge grounds (i.e., sites where these sponges reach densities of 0.5–1 m−2 (still image surveys45) to 0.03–0.1 m−2 (video surveys40)) appear to be one of the most commonly found seamount habitats in high latitudes, as they have also been observed on N Atlantic seamounts (e.g., 40°–75° N latitude belt;46,47). Roberts et al.48 suggested that short-timescale environmental variability combined with the generally nutrient rich sub-surface water masses generated at the interface of intermediate and deep-sea water masses49 might potentially enhance the development of sponge aggregations and other megafauna (e.g., ascidians, cnidarians, echinoderms, and demersal fish) at the Schulz Bank seamount (73.5° N) on the Arctic Mid-Ocean Ridge. However, very little is known about the diversity of benthic communities in seamounts at higher Arctic latitudes (e.g., > 75° N; see also50). Hence, it remains unclear what the key factors structuring Arctic seamount communities across different spatial scales are, e.g., do these differ from those shaping lower latitude seamount communities?

The sponge order Tetractinellida includes deep-sea species that occur in cold water masses of the Arctic and North Atlantic49. Their distribution ranges from seamounts in the central Arctic Ocean50,51 and Nordic Seas52,53 to Norwegian fjords54,55,56, the Mid-Atlantic Ridge46, along the continental shelves of Europe, Canada and the US40,47,57,58,59, around Iceland and the Faroe Islands60,61 to New England seamounts in the central North Atlantic62. Sponge grounds usually exhibit an increased associated benthic diversity and biomass when compared with adjacent non-sponge habitats41. Sponges typically enhance the complexity of habitats by increasing the (three-dimensional) hard surface area available for other fauna to interact (e.g., settle, reproduce, and/ or hide37,63,64,65,66). For instance, at the Schulz Bank seamount large Tetractinellida sponges like Geodia sp. and Stelletta sp. typically have ascidians or other sponges (such as the encrusting sponge Hexadella dedritifera) growing along their edges67. Such dense sponge community act not only as ecosystem engineers, but they also play an important role in nutrient and matter cycling. By filtering large volume of water (up to 2000 L m−2 d−1 in case of Geodia sp.40), they efficiently retain pico-nanoplankton cells and process dissolved compounds acting as sink and/or source of nutrients and organic matter.

In this study, we used seabed imagery to investigate regional (between seamounts) and local (within seamounts) variations in benthic megafauna communities across the Langseth Ridge in the Central Arctic (86.6°–86.9° N; approximate water depth 600–2000 m). The Langseth Ridge is a chain of three seamounts (Central Mount, Karasik Seamount, and Northern Mount), which summits are dominated by dense aggregations of mobile sponges (a striking and previously unforeseen trait;51)50. We hypothesized that seabed composition (Fig. 1, five potential habitats: H1, dense sponge grounds; H2, mats of extensive polychaeta tubes and sponge spicules covered with sulfide precipitates; H3, sediment with gravel; H4, bare rock; and H5, mixed or undominated substrate) plays a large role in structuring megabenthic assemblages, i.e. delineating abrupt seabed community variations within arctic seamounts. In addition, we describe and discuss how common processes triggered by decaying sponges, e.g. “the sponge loop”68,69,70, might enhance local and regional habitat variations.

Figure 1
figure 1

Habitat types identified over the three seamounts: (H1) dense sponge grounds of Tetractinellida gen. indet. sponges extending on top of sponge-produced spicule mats (partly comparable to habitat category d in50), (H2) mats of Serpulidae indet. and Siboglinidae indet. tubes intermixed with sponge spicules covered with sulfide precipitates (includes habitat category b in50), (H3) sediment with gravel, (H4) bare rock, and (H5) mixed substrate (equivalent to habitat type c in50). The white bar represents 50 cm.

Full size image

Results

Micro- and macrohabitat types

Most of the seafloor at the Central Mount (CM) was covered with bare rock (habitat type H4, 56% seabed coverage; Table S1, Fig. 1) and mixed substrate (H5, 30.9% seabed coverage; Table S1, Fig. 1), i.e., a mixed assemblage of sponge grounds with spicule mats, mats of polychaete tubes, sand with gravel, and/ or bare rocks. Habitat type H2, i.e., mats of Serpulidae gen. indet. and Siboglinidae gen. indet. tubes intermixed with sponge spicules and covered with sulfide precipitates, was not observed. The Karasik Seamount (KS) was dominantly covered by H4 (73.4%; Table S1) and the Northern Mount’s (NM) seabed consisted to 56.7% of habitat type H5 and to 30.6% of habitat type H4 (Table S1).

Quantitative assessment

Variations in faunal density

Megabenthos density exhibited substantial variations across the different areas investigated, both at the regional (between seamounts) and at the local (between habitats) scales. Mean faunal density at KS (mean density: 110,442 ind. ha−1; CI 95%: 86,253 − 139,541 ind. ha−1) was substantially larger than at CM (mean density: 55,745 ind. ha−1; CI 95%: 43,305–71,984 ind. ha−1) (Fig. 2a), whereas the assemblages in NM exhibited a large variability, ranging from 34,162 to 119,762 ind. ha−1 (in ca. 1000 specimen samples). Local assessments revealed that the high variability observed at NM was predominantly caused by large differences in faunal density between H4 (mean density: 113,118 ind. ha−1) and H5 (mean density: 14,123 ind. ha−1) (Fig. 2b). Densities were consistently smaller in H5 compared to H4 areas, but substantially different densities were also found between habitats of the same type across different seamounts (Fig. 2b), suggesting the existence of faunal density drivers operating at both local and regional scales. Taxon-specific densities specifically for H4 and H5 are presented in Table S2.

Figure 2
figure 2

Regional variations in (a) megabenthos density (ind. ha−1), (c) taxa richness S (in ca. 1,000 specimens), (e) 1/Simpson index D (in ca. 1000 specimens), and local variations in (b) megabenthos density (ind. ha−1), (d) taxa richness S (in ca. 1000 specimens), and (f) 1/Simpson index D (in ca. 1,000 specimens) across the three Arctic seamounts (CM Central Mount, KS Karasik Seamount, NM Northern Mount) investigated. Bars indicate mean values across bootstrap-like sample sets surveyed in each study area (a, c, e) and in the predominant habitat types (H4 and H5) within each study area (b, d, f). Error bars represent CI 95%.

Full size image

Variations in diversity

No substantial variations in taxa richness were observed across the different areas investigated, neither regionally nor locally (Fig. 2c–d). In contrast, heterogeneity diversity (i.e., 1/D, an index more sensitive to taxa evenness) was substantially higher at KS than at CM, whereas the assemblages at NM exhibited a large variability for this parameter (Fig. 2e). Similar to faunal density, local assessments revealed that the high variability observed at NM was predominantly caused by large differences in heterogeneity diversity between H4 (mean 1/D: 1.5 effective taxa) and H5 (mean 1/D: 3.7 effective taxa) (Fig. 2f). In turn, no major differences were observed in heterogeneity diversity between H4 and H5 areas at CM nor at KS (Fig. 2f). However, heterogeneity diversity in both H4 and H5 areas from KS were consistently higher than in respective H4 and H5 areas from CM, again suggesting the existence of diversity drivers operating at both local and regional scales.

Variations in assemblage composition

A total of 15 invertebrate epibenthos taxa and 4 fish taxa were identified in the image set across the three seamounts studied (Fig. S1). At KS 15 invertebrate and 3 fish taxa were observed, whereas 14 invertebrates and 4 fish taxa were detected at CM. At NM all taxa observed on the other seamounts were also found (Table S3).

Multivariate analyses showed substantial variations among the assemblages of different areas investigated, both at the regional (between seamounts) and at the local (between habitats) scales. Non-metric multidimensional scaling (nMDS) ordination of regional assemblage composition data readily distinguished the bootstrap-like samples from the three study areas, particularly those from KS and NM (Fig. 3a), as assemblage dissimilarity was higher between these two areas (βBC: 32.1%) than between CM and the other two seamounts (βBC: 24.6–25%). A much larger variation was found, however, within the assemblages of NM than within both CM and KS assemblages (Fig. 3a). Local assessments revealed that the high within-sample variability observed at NM was predominantly caused by a high dissimilarity between the assemblages in H5 and H4 (βBC: 31.3%; Fig. 3b). The latter exhibited a high resemblance with the assemblage from H4 areas at CM (βBC: 22.3%). In contrast, dissimilarity between the assemblages of H4 and H5 was less pronounced within CM (βBC: 17.3%), and almost inexistent within KS (βBC: 13.8%; Fig. 3b, overlapping confidence intervals), suggesting a stronger control of regional drivers at CM and KS areas, and a stronger control of local drivers at NM area.

Figure 3
figure 3

nMDS plots showing regional (between seamounts) and local variations (between habitat types) in faunal assemblage composition. (a) Regional assessment based on 100 randomly selected bootstrap-like samples for each study area (seamounts: CM Central Mount, KS Karasik Seamount, and NM Northern Mount) (nMDS stress: 0.09). (b) Local assessment based on 100 randomly selected bootstrap-like samples for the two dominant habitat types (H4 = bare rock, H5 = mixed substrate) in each study area (nMDS stress: 0.12). Ellipses represent CI 95%.

Full size image

Faunal assemblage at different seamounts

Central Mount

The faunal assemblage at CM was clearly dominated by sponges of the order Tetractinellida gen. indet. (84.9% of all fauna; 23,070 ind. ha−1 sponges with < 8 cm diameter, 24,348 ind. ha−1 sponges with > 8 cm diameter; Fig. 4a, Table S3). The second, third, and fourth most abundant taxa were, respectively: the shrimp Bythocaris sp. indet. (3502 ind. ha−1; 6.27% of all fauna), the brittle star Ophiostriatus striatus sp. inc. (1133 ind. ha−1; 2.03% of all fauna), and the polychaetes Apomatus globifer sp. inc./Hyalopomatus claparedii sp. inc. (1071 ind. ha−1; 1.92% of all fauna) (Fig. 4a). All other fauna accounted for 4.83% of the total faunal density.

Figure 4
figure 4

Invertebrate epibenthic megabenthos and fish densities at the Central Mount (CM), Karasik Seamount (KS), and Northern Mount (NM). (a) All fauna (ind. ha−1) observed at the different seamounts, (b) all fauna (ind. ha−1) that was observed being physically attached to Tetractinellida gen. indet. or walking/ crawling on top of Tetractinellida gen. indet., (c) all fauna (ind. ha−1) that was physically associated with Hexactinellida or was crawling on top of Hexactinellida.

Full size image

Co-occurrence network analysis of the faunal assemblage at CM showed no specific co-occurrences with the fish Gadidae fam. indet. and Liparidae fam. indet., Hexactinellida sponges, nor the gastropod Buccinidae fam. indet. (Fig. S2a). The analysis further indicated that Tetractinellida sponges co-occurred with a lower than expected frequency with the bivalve Hyalopecten frigidus (plt = 0.034) and with a higher than expected frequency with the anemone Hormathiidae (pgt = 0.000) and polychaetes (A. globifer/H. claparedii, pgt = 0.000) (Table S4).

A total of 852 specimens (1.53% of all fauna) were found associated with sponges, i.e., either as attached sessile epifauna or as mobile epifauna crawling over these, at CM. Among these, only five specimens were found associated with Hexactinellida sponges, with the remainder found on Tetractinellida specimens. Taxa most frequently associated with Tetractinellida gen. indet. were the shrimp Bythocaris sp. indet. (212 ind. ha−1; 25.1% of fauna associated with Tetractinellida gen. indet.), the anemone Hormathiidae gen. indet. (185 ind. ha−1; 21.8% of fauna associated with Tetractinellida gen. indet.), and the starfish Tylaster willei sp. inc. (180 ind. ha−1; 21.2% of fauna associated with Tetractinellida gen. indet.) (Fig. 4b–c,e–f).

Karasik Seamount

The faunal assemblage at KS was dominated by sponges of the order Tetractinellida (66.4% of all fauna; 33,422 ind. ha−1 sponges with < 8 cm diameter, 39,860 ind. ha−1 sponges with > 8 cm diameter; Fig. 4a,d) and the shrimp Bythocaris sp. indet. (22.9% of all fauna; 25,284 ind. ha−1; Fig. 4a,d, Table S4). The remaining faunal assemblage was predominantly composed by the mysid Neobirsteiniamysis inermis sp. inc. (4360 ind. ha−1), and the polychaete Macellicephalinae gen. indet. (2054 ind. ha−1) and A. globifer sp. inc./H. claparedii sp. inc. (1391 ind. ha−1). All other fauna accounted for 2.86% of the total faunal density.

The results of the co-occurrence network analysis of the assemblage at KS indicated that the brittle star Ophiostriatus striatus sp. inc., the fish Lycodes sp. indet., and Hexactinellida sponges had no specific co-occurrence with any other taxon in the faunal community (Fig. 5b). Tetractinellida sponges were observed to co-occur with gastropods Buccinidae fam. indet. at a lower frequency than expected by chance (plt = 0.014), and with polychaetes (Serpulidae and Siboglinidae, pgt = 0.039; A. globifer/H. claparedii, pgt = 0.000;) at a frequency higher than expected (Table S4).

Figure 5
figure 5

(a) Images of different stages of sponge decay observed across the three seamounts. (SD1) Healthy big Tetractinellida gen. indet. sponge, (SD2) Tetractinellida sponge partly covered with a white/blue microbial mat, (SD3) whitish-grey/bright yellow, collapsed sponge, (SD4) polychaetes covering the sponge remains. (b) Density of big sponges in different stages of decay/bleaching (in ca. 1000 specimens) across the three Arctic seamounts (CM Central Mount, KS Karasik Seamount, NM Northern Mount). For plotting purposes, densities of SD1 are shown in the left part of panel (b) and densities of SD2–SD4 are presented in the right part of panel (b).

Full size image

A total of 6,244 specimens (5.66% of all fauna) were found associated with sponges, i.e., either attached, crawling over or feeding on these, at KS. Among these, only 39 specimens were found associated to Hexactinellida sponges and the rest were on Tetractinellida sponges. Taxa most frequently associated with Tetractinellida gen. indet. were the shrimp Bythocaris sp. indet. (2975 ind. ha−1; 47.9% of fauna associated with Tetractinellida gen. indet.), the mysid Neobirsteiniamysis inermis sp. inc. (1129 ind. ha−1; 18.2% of fauna associated with Tetractinellida gen. indet.), the soft coral Nephtheidae gen. indet. (623 ind. ha−1; 10.0% of fauna associated with Tetractinellida gen. indet.), and juveniles of the sponge Schaudinnia rosea sp. inc. (527 ind. ha−1; 8.49% of fauna associated with Tetractinellida gen. indet.) (Fig. 4b–c,e–f).

Northern Mount

The faunal assemblage at NM was dominated by the sponge of the order Tetractinellida gen. indet. (73.9% of all fauna; 28,825 ind. ha−1 sponges with < 8 cm diameter, 20,054 ind. ha−1 sponges with > 8 cm diameter; Fig. 4a,d; Table S3) and the bryzoan Cyclostomatida fam. indet. (7469 ind. ha−1; 11.3% of all fauna). The remaining faunal assemblage was predominantly composed by the brittle star Ophiostriatus striatus sp. inc. (2380 ind. ha−1), the shrimp Bythocaris sp. indet. (2272 ind. ha−1), the starfish Tylaster willei sp. inc. (1435 ind. ha−1), and the polychaete Macellicephalinae gen. indet. (973 ind. ha−1). All other fauna accounted for 4.01% of the total faunal density.

Co-occurrence network analysis of the faunal assemblage at NM predicted no specific interactions between Hexactinellida sponges and any other megafaunal taxon at the seamount (Fig. S2b). Tetractinellida sponges co-occurred with a higher than expected frequency with anemones (Edwardsiidae gen. indet., pgt = 0.000; Nephtheidae gen. indet., pgt = 0.000), with polychaetes (Serpulidae and Siboglinidae, pgt = 0.000; A. globifer/H. claparedii, pgt = 0.000; Macellicephalinae gen. indet., pgt = 0.000), and with the bryozoan Cyclostomatida fam. indet. (pgt = 0.000) (Table S4). In comparison, Tetractinellida sponges at NM were observed to co-occur with a lower than expected frequency with fish (Liparidae fam. indet., plt = 0.004; Lycodes sp. indet., plt = 0.013; Rhodichthys regina inc., plt = 0.000) (Table S4).

A total of 1484 specimens (2.25% of all fauna) were found associated with sponges, i.e., either attached, crawling over or feeding on these, at NM. Among these, only 26 specimens were found associated to Hexactinellida sponges with the rest on Tetractinellida gen. indet. sponges. The taxa most frequently associated with Tetractinellida gen. indet., were the bryozoan Cyclostomatida fam. indet. (445 ind. ha−1, 30.5% of fauna associated with Tetractinellida gen. indet.), the starfish Tylaster willei sp. inc. (303 ind. ha−1, 20.8%), other small Tetractinellida gen. indet. specimens (167 ind. ha−1, 11.4%), and anemone Hormathiidae gen. indet. (161 ind. ha−1, 11.1%) (Fig. 4b–c,e–f).

Density of decaying sponges and their bulk and compound-specific isotope composition

At all three seamounts Tetractinellida gen. indet. sponges were observed in four different stages of decay (Fig. 5a). These decay stages ranged from healthy sponges (95.2–95.5% of all big Tetractinellida sponges; Fig. 5a SD1, b), to sponges that turned white/blue indicating the coverage with microbial mats (0.09–0.22% of all big Tetractinellida sponges; Fig. 5a SD2, b), to a whitish/bright yellow, collapsed sponge (1.79–2.54% of all big Tetractinellida sponges; Fig. 5a SD3, b), to dense mats of polychaetes covering the sponge remains (1.62–3.79% of all big Tetractinellida sponges; Fig. 5a SD4, b). Most sponges of decay stages SD1, SD2, and SD3 were observed at habitat type H4 at KS, while dense mats of polychaetes on top of sponge remains (SD4) were mostly observed at H4 at CM (Table S5).

To infer possible trophic interaction, in this study we used bulk isotopes (δ13C and δ15N) and phospholipid fatty acids (PLFAs) as organism-specific markers for the identification of food sources. At the Karasik Seamount, big Tetractinellida gen. indet. sponges had stable isotope values of (mean ± standard deviation) − 18.2 ± 0.22‰ δ13C and 8.16 ± 0.51‰ δ15N (Geodia hentscheli; n = 4), − 18.4 ± 0.15‰ δ13C and 8.43 ± 0.78‰ δ15N (Geodia parva; n = 7), and − 18.3 ± 0.36‰ δ13C and 8.42 ± 0.28‰ δ15N (Geodia rhaphidiophora; n = 3)50. Decaying sponges of stage SD2 had a stable isotopic composition of − 20.2 ± 0.45‰ δ13C and 4.82 ± 0.23‰ δ15N (n = 3) and the microbial mat that covered the sponges in decay stage SD2 had a stable isotopic composition of − 17.9 ± 0.08‰ δ13C and 10.8 ± 0.28‰ δ15N (n = 2). These microbial mats contained PLFAs that were to 74% bacteria-specific and 19% were sponge-specific PLFAs (Table S6).

Across the three seamounts, several taxa were observed in association with Tetractinellida sponges in different stages of decay, such as the asteroid Tylaster willei sp. inc., and the shrimps Bythocaris sp. indet. and Neobirsteiniamysis inermis sp. inc. Additionally, polychaetes of the family Macellicephalinae gen. indet. crawled over or fed upon decaying Tetractinellida sponges at the Karasik Seamount and byrozoans Cyclostomatida fam. indet. were associated with decaying Tetractinellida sponges at the Northern Mount. At the Karasik Seamount, these asteroids had a stable isotopic composition of − 13.4 ± 6.05‰ δ13C and 10.84 ± 1.53‰ δ15N (n = 2)50 and the total PLFA pool of a starfish collected from the top of a Tetractinellida sponge consisted to 48% of bacteria-specific PLFAs, to 33% of algae-specific PLFAs and to 6% of sponge-specific PLFAs (Table S6). Shrimps had stable isotope values of − 21.7 ± 1.71‰ δ13C and 12.4 ± 0.89‰ δ15N (n = 2)50. Unfortunately, due to lack of bulk material, no PLFAs were extracted from shrimps associated with the sponges, so no information about their PLFAs composition is available.

Discussion

The seafloor at the three seamounts of the Langseth Ridge consisted mainly of bare rock, sand, and gravel along with a mix of biogenic structures composed of reef-forming sponge grounds, spicule mats, and polychaete tubes. Our results showed that the megafaunal densities and assemblage composition, but not taxon richness, differed at regional (between seamounts) and local (within seamount, between habitats) scales across the studied area. Demosponges of the order Tetractinellida numerically dominated the assemblages across the three seamounts, possibly owing to their unique capacity to source carbon directly from the refractory matter on the seabed50, which likely makes these mostly (if not fully) independent from the water column food-supply in such a low primary productivity area71,72. Shrimps (Central Mount, Karasik Seamount) and bryozoans (Northern Mount) were the other most abundant taxa encountered, yet by far not as abundant as by Tetractinellida sponges present in a much smaller area. Here, we discuss the potential processes causing the observed variations in megabenthic composition at different scales. We further elaborate on the functional role Tetractinellida sponges play in the high Arctic seamount ecosystem and describe a new possible pathway of the sponge loop as a potential additional mechanisms for recycling organic matter in this food-deprived ecosystem.

Variations in assemblage composition observed in regional assessments were likely related to inherent differences in megabenthos density across the three seamounts. This difference in densities could be correlated with the height of the seamounts: depth, or more precisely, the strong covariation of key factors (i.e., food supply and temperature) with increasing depth73 has been widely highlighted as major proxy for deep-sea benthic abundance and biomass74,75. As such, and in line with our results, many studies have shown how depth-related variations in population densities can yield markedly distinct benthic communities in seamounts18,22,76. Water temperature and current strength may be other drivers of the variations in faunal abundance observed. For instance, water temperatures measured at the Karasik Seamount (0.66 °C) and the Northern Mount (0.68 °C) were higher than at the Central Mount (0.23 °C)77. In contrast, current velocity measured during the cruise were generally weak (< 0.1 cm s−1) with a predominantly westwards component and no evidence of associated upwelling currents78. It is hence more plausible that food supply and temperature decreases with depth have a stronger influence on the observed variations in megabenthic abundance than the overlying current dynamics. It is noted that bottom currents and hydrographic processes can typically exhibit periodic or seasonal increases, leading to enhanced food supply rates (e.g.,79), particularly in interaction with the complex topography of seamounts38. However, we rule out the possibility that high densities of bryzoans at the Northern Mount was related to increased seasonal currents due to the sluggishness of the current over the year, and the year-round ice cover.

Variations in assemblage composition at local scale appears to be clearly driven by the existence of different habitats. Habitat type H4 (bare rock) and H5 (mixed substrate) covered between 87 and 89% of the seamount areas studied, whereupon H4 supported a relatively denser community compared to H5. This difference in megabenthic densities was partly related to variations in morphotype composition between H4 and H5 at the Northern Mount and the Central Mount. For instance, in both seamounts, only very few brittle stars were observed across bare rocks (H4), whereas they were very abundant across the mixed substrate seafloor areas (H5). The brittle star Ophiostriatus striatus is an opportunistic deposit feeder that was observed grazing upon fresh and detrital ice algae in the Nansen Basin close to the Gakkel Ridge during the minimum sea ice extent in 201280,81. The mixed substrate (max thickness of spicule mat: 15 cm50) may trap settled particles50 which could subsequently serve as a food source for brittle stars and other deposit-feeding fauna. This would explain the very low densities of this brittle star across bare rocks where potentially increased hydrodynamics together with the lack of tridimensional structure provided by the spicule mat might prevent detritus accumulation. However, it remains unclear why it is almost absent from the Karasik Seamount as Zhulay and colleagues observed uncommon swimming behavior in the species82, which might facilitate the connectivity between the Central Mount and the Karasik Seamount. This suggests that a combination of regional and local environmental differences likely causes the variability between the megabenthos assemblages at the Northern and Central Mount and that at the Karasik Seamount.

Besides the habitat types H4 and H5, also H1 (Tetractinellida sponge grounds) and H3 (sediment with gravel) were observed at all seamounts, whereas H2 (mats of polychaete tubes) was found only at the Northern Mount and the Central Mount. However, owing to the primarily exploratory nature of the research cruise to investigate the geological, geochemical, and biological processes of the active hydrothermal vent at the Gakkel Ridge83 and seamounts at the Langseth Ridge in 201650,51,83, the surrounding topography and community structure were not well known before the expedition. Therefore, no previous information was available to design a series of seabed image surveys that could grant a balanced sampling effort between habitat types (i.e. fully unknown prior to the expedition). Our bootstrap-based assessment allowed the reduction of this study limitation (i.e., unbalanced sampling effort across different habitat types) by focusing on the variability associated to different ecological estimators rather than in the actual estimations (e.g., mean values), which can be a robust way for instance, to infer ecological patterns in opportunistic deep-sea datasets (e.g.,84,85), yet was only conceived here as a preliminary approach. In this regard, the comparably smaller image sample size for H1, H2, and H3 did not allow for reliable statistical comparison of these ones with the more dominant H4 and H5. Thus, based on our preliminary work, future studies aimed at acquiring a better understanding of the community ecology and composition in this area shall now be able to appropriately design benthic image surveys, i.e., yielding even sampling effort across each of the, now characterized, Langseth Ridge habitat types.

The main habitat types of the deep Arctic Ocean are ridges, seeps, hydrothermal vents, and deep basins filled with soft sediment81,86,87,88,89,90,91. The megabenthic community observed at the Langseth Ridge differs substantially from the community at seeps and hydrothermal vents whose fauna is characterized by chemosynthetic taxa, such as siboglinid polychaetes87,88,89,92 and gastropods91, and from soft-sediment communities. These contain mostly the phyla Echinodermata, Cnidaria, Porifera, and Arthropoda86,93,94, and show a bathymetric distribution with a lower shelf/upper slope community (characteristic taxa: brittle star Ophiocten sericeum, bivalve Yoldiella solidula), a lower slope community (characteristic taxa: bivalve Bathyarca frielei, polychaete Galathowenia fragilis), and an abyssal community (characteristic taxa: polychaete Anobothrus laubieri, sea cucumber Kolga hyalina)95. Instead, the sponge grounds on the Central Mount, the Karasik Seamount, and the Northern Mount resemble partly the sponge ground community at the Schulz Bank seamount. The megafaunal community at this seamount comprised 20 taxa and was dominated by the Tetractinellida sponges Geodia parva and Stelletta rhaphidiophora67. Further abundant sessile taxa were ascidians, anthozoans (Gersemia rubiformis), other Demospongiae (Lissodendoryx (Lissodendoryx) complicate, Hexadella dedritifera), and Hexactinellida sponges67. Mobile taxa like echinoderms and fish were also observed, but they occurred at lower densities67.

Tetractinellida sponges at the Langseth Ridge host diverse taxa, such as juvenile sponges, anthozoans, byrozoans, or the polychaetes A. globifer sp. inc./H. claparedii sp. inc. The latter colonized the edge of Tetractinellida sponges as they may benefit from the water fluxes generated by the pumping activity of the sponges, as well as from the particle detritus expelled by them as a source of food for the epi-endobiota96. In comparison, only few specimens were found associated to hexactinellid sponges: several sponge specimens, bryozoans, and anthozoans. The difference in the number of associated specimens between the Hexactinellida and Tetractinellida sponges can be related to their different morphology (papillate/globular and massive, respectively) and the spicule “fur” produced by Tetractinellida that facilitate the epifauna settlement61. Additionally, Tetractinellida sponges at the Northern Mount, Karasik Seamount, or Central Mount had a large variety of mobile fauna associated with them, such as starfish, shrimps, and mysids. These starfishes either predate on the Tetractinellida sponges or graze upon sponge detritus, as sponge-specific PLFAs were detected in the analyzed starfish. Such predatory or detrital transfer of sponge-derived particulate organic matter (POM) to echinoderms has been measured in ex-situ pulse-chase incubation experiments by Bart et al.68. Using their experimental design with Geodia barretti and brittle stars, the authors were not able to differentiate between the so-called ‘deep-sea detrital sponge loop’ and the ‘deep-sea predatory sponge loop’68. In the “detrital sponge loop”, G. barretti released POM which is subsequently taken up by brittle stars69. On the other hand, in the ‘predatory sponge loop’, spongivores directly predate upon G. barretti.

In this study, we observed an additional pathway of the predatory deep-sea sponge loop that might function as follows (Fig. 6): an unknown chemical, physical, or biological cue triggers the decay of putatively healthy Tetractinellida sponges (δ13C value: − 18.2 to − 18.4‰)50. At the Tisler cold-water coral reef in Norway, for instance, a mass mortality event of G. barretti was partially related to temperature heat shocks97,98. During this event, the decaying sponges turned blue and black97, whereas (potentially bacterially-induced) diseased G. barretti from Korsfjord in Norway had a brown/black discoloration and disintegrated, fouling sponge tissue98. At the Langseth Ridge, decaying Tetractinellida sponges are covered by microbial mats and attract predators/ spongivores, such as starfishes. The starfish predates on the sponge as observed at the Schulz Bank seamount99 and/or grazes upon the microbial mat (δ13C value: − 17.9‰) covering the decaying sponge (δ13C value: − 20.2‰) as indicated by sponge-specific fatty acids and the higher δ13C value of − 13.4‰50. We hypothesize that feces produced by the starfish are subsequently recycled by the cryptic community living hidden in the sponge spicule-polychaete tube mats, such as amphipods, tanaidaceans, gastropods, and polychaetes83. The sponge spicule-polychaete tube mats may provide similar ecological and biogeochemical functions like ‘dead’ cold-water coral framework at Haas Mount in the North Atlantic100. There, dead framework increases resource retention and recycling like a “filtration-recycling factory”100. Maier et al.100 estimated that dead coral framework and living corals at Haas Mount had nearly equal contributions to the total oxygen consumption at the reef. At the Langseth Ridge, the cryptic community in the sponge spicule-polychaete tube mats likely releases dissolved organic matter (DOM) which can be taken up by the Tetractinellida sponges. Indeed, Bart et al.68,101,102 and Maier et al.103 measured dissolved organic carbon (DOC) uptake by Geodia sp. in ex-situ incubation experiments. This uptake of decaying sponge-derived DOM by healthy sponges would also support the relatively young age of Tetractinellida sponges that Morganti et al.50 reported in their study and related to the potential assimilation of relatively young DIC. We therefore propose such ‘deep-sea sponge loop’ as additional mechanism to re-cycle organic carbon matter in this ecosystem, facilitating the presence of such dense and abundant sponge community in the Langseth Ridge.

Figure 6
figure 6

Conceptual model of the sponge loop likely present at the seamounts of the Langseth Ridge. The individual components of the sponge loop are the following: (1) Putatively healthy Tetractinellida sponges, (2) microbial mat covering a sponge, (3) asteroids predating upon the microbial mat covering a Tetractinellida sponge/ directly upon the sponge, (4) mat of polychaete tubes intermixed with sponge spicules hosting a cryptic microbial and faunal community. Illustrations by Tanja Stratmann.

Full size image

In conclusion, this study presents a detailed description of megabenthos assemblages at the northernmost seamounts explored so far. Interestingly, taxa richness did not differ between seamounts and habitats. While the megabenthos community composition showed substantial differences at regional and local scale, likely driven by intrinsic seamount characteristics (water temperature and depth) and distinct habitats, respectively. The Northern Mount had the highest density of bryozoans, which were almost absent in other seamounts and a more pronounced difference in megabenthic composition between the bare rock and mixed substrate habitats. So far, there is no evidence of particular processes, such as the increase of bottom currents or the different hydrographic conditions at the Northern Mount for explaining such distinctive features when compared to the other two seamounts. Further video and/or image transects at the individual seamounts are required to assess the megabenthos communities inhabiting the three other classified habitats.

Using bulk and compound-specific stable isotope analysis of phospholipid-derived fatty acids (PLFA) from Tetractinellida sponges, microbial mats, and starfish, this study showed the uptake of sponge-specific PLFA by starfishes that either originate from predation on the sponge (‘deep-sea predatory sponge loop’) or the uptake of sponge-derived detritus (‘deep-sea detrital sponge loop’). Starfish could also graze upon the microbial mat covering the decaying sponges. In either case, it is hypothesized that the feces of starfish are recycled by the cryptic community living in the sponge-spicule-polychaete tube mat and converted to DOM. This DOM may be subsequently taken up again by ‘healthy’ Tetractinellida sponges.

Like cold-water corals, sponges play an important role in habitat forming as ecosystem engineers and their spicules intermixed with the polychaete tubes create a perfect matrix for a “filtration-recycling factory”.

Materials and methods

Study area

The Langseth Ridge is a permanently ice-covered underwater mountain ridge in the central Arctic Ocean that stretches approximately 125 km from 87° N, 62° E to 85° 55′ N, 57.45′ E83,104 (Fig. 7). It is comprised of three summits, the Central Mount (CM), the Karasik Seamount (KS), and the Northern Mount (NM). The CM has its summit at 86° 47.83′ N, 61° 54.52′ E where its maximum elevation reaches to 722 m below the sea surface83. This seamount has a gradually increasing slope from 3300 m to its point of maximum elevation83. Its slope on the western side is steeper than the one on the eastern side where the slope drops to 4500 m water depth83. The tallest mountain on the Langseth Ridge is the KS which summit is located at 86° 43.0′ N, 61° 17.6′ E and reaches to 2500 m above the seafloor (i.e., 585 m below the sea surface83)104. The NM is located at 86° 51.86′ N, 61° 34′ E and has a maximum elevation of 631 m below the sea surface83. This seamount has a steep slope from its peak towards the Gakkel Ridge rift valley in the north at 4000 m depth83.

Figure 7
figure 7

Map of the Northern Mount, Central Mount, and Karasik Seamount along the Langseth Ridge with all Ocean Floor Observation and Bathymetry System (OFOBS) deployments. The station numbers correspond to the OFOBS transect numbers in Table S7. The map was created using ArcGIS software, version 10.5 (https://www.arcgis.com).

Full size image

Bottom water at the three seamounts had a temperature range between − 0.02 °C (CM) and 0.22 °C (NM) and a salinity of 34.9 PSU83. Oxygen concentration in the water was only measured at the NM and amounted to 322 µmol L−1105. Nitrite was not detected in bottom water and ammonium was only found in KS’s bottom water (0.02 µmol L−1)105. Phosphate concentrations ranged in bottom water from 0.67 to 0.68 µmol L−1, nitrate concentration was between 12.0 and 12.7 µmol L−1, and silicate ranged from 5.53 to 5.86 µmol L−1105.

Seabed image collection and processing

Image collection

The high-resolution digital photo camera (CANON EOS 5D Mark III, modified by iSiTEC for underwater applications) of the towed Ocean Floor Observation and Bathymetry System (OFOBS)106 was used to take still images of the seafloor of the three different seamounts. OFOBS was deployed four times (= four transects; Fig. 7 and Table S7) during the RV Polarstern cruise PS101 in the central Arctic Ocean (chief scientist: Prof. Dr. Antje Boetius)83. During each deployment, OFOBS was towed 1.5–2.5 m above the seafloor at a speed of < 1 knot and photographs were taken every 20 s to avoid overlap between images. The area of each image was calculated using three laser points on the seafloor that were organized in an equilateral triangle (distance between points: 0.5 m) as reference for scaling (area per image: mean ± SE: 8.43 ± 0.17 m2). A total of 3162 photographs were collected, from which 2099 (17,691 m2 of seabed) were used (Table S7), as only bright images collected within the aimed altitude range were selected for analysis. All photographs were loaded into the open-source software “Program for Annotation of Photographs and Rapid Analysis (of Zillions and Zillions) of Images” PAPARA(ZZ)I107.

Habitat classification

To classify macro- and microhabitat108 types at the different seamounts, it was recorded for each image whether \(\ge\) 75% of the seafloor was covered by dense sponge grounds of Geodia sp. indet./Stelletta sp. indet. and sponge spicule mats (habitat type H1; includes parts of habitat category d in50), by mats of Serpulidae indet. and Siboglinidae indet. tubes covered with sulfide precipitates (habitat type H2; corresponds partly to habitat category b in50), by sediment with gravel (habitat type H3), or by bare rock (habitat type H4). When the seafloor was covered to \(\le\) 75% by one of the four habitat types and therefore consisted of an assemblage of sponge ground and spicule mats, mats of polychaete tubes, sediment with gravel, and/ or bare rock, it was classified as mixed substrate (habitat type H5; habitat types a and c in50). The five habitat types described in this study are shown in Fig. 1.

Biological analysis

Megabenthic fauna (> 1 cm size; Fig. S1) visible on the photographs were annotated and identified to the lowest taxonomic hierarchy possible (morphotype [mtp]: typically Genus or Family level) based on previous image collection, published by81. Tetractinellida sponges were identified by spicule analysis in50 by Prof. Dr. Hans Thore Rapp. The taxonomic nomenclature of the morphotypes presented follows109.The life-habit of specimens was recorded whenever these were found attached (sessile fauna) or crawling (mobile fauna) on other specimens (i.e., generally sponges). As it is not possible to distinguish between Geodia sp. indet. and Stelletta sp. indet. sponges on seabed images, these specimens were identified as Tetractinellida gen. indet. (Fig. S1H). The latter were annotated based on their diameter size as “Tetractinellida gen. indet. ‘big’” (diameter > 8 cm) and “Tetractinellida gen. indet. ‘small’” (diameter < 8 cm). Polychaetes of the family Serpulidae indet. and Siboglinidae indet. (Fig. S1M) were not annotated individually, but as patches. The polychaetes A. globifer sp. inc/H. claparedii sp. inc (Fig. S1N) that were observed associated with Tetractinellida gen. indet. specimens were annotated as present/absent, i.e., they were annotated once per image when they were present. The uncertainty of the image-based identifications was indicated following the recommendations by109 for standardization of the open taxonomic nomenclature.

Quantitative data analysis

Patterns in diversity and distribution of faunal assemblages were quantitatively assessed at two scales: (i) regionally (scale: 10s km), across the three different seamounts (CM, KS, and NM), and (ii) locally (scale: 100s m), between the two habitats with the largest seafloor coverage (H4 and H5; encompassing 94% of all specimens and 88% of all the seabed area surveyed) across the three seamounts. In each case, megabenthos specimen data were pooled for each study area or target stratum, e.g., per seamount or habitat type, and then resampled using a modified form of bootstrapping110.

Resampling methods provide robust estimates of variability and confidence intervals of sample parameters111,112, and are particularly well suited to analyze seabed image data obtained from survey designs that lack true sample replication (see e.g.,84,113) like in this case with four OFOS transects across three seamounts due to the exploratory nature of the research cruise (Fig. 7 and Table S7). To implement the bootstrap, image data were randomly resampled with replacement until a minimum of 1000 megabenthos specimens were encompassed, and that process was repeated 10,000 times for each target stratum. This process yielded bootstrap-like samples (bootstrap generated sub-samples) with fixed specimen count size, ranging in total seabed cover from 72 to 490 m2, to minimize the potential effect of variable faunal densities in the estimation of ecological parameters.

A range of ecological parameters were calculated from each set of bootstrap-like samples to compare the assemblages from different target strata. Patterns in abundance were assessed by estimation of numerical density (ind. ha−1), whereas diversity was assessed by estimation of taxa richness (S, in ca. 1000 specimens) and Simpson’s index (D, in ca. 1000 specimens)114.

Variations in assemblage composition were assessed by non-metric multidimensional scaling (nMDS) ordination of bootstrap-like samples, based on the Bray–Curtis dissimilarity (or beta-diversity, βBC) measure115 calculated using square-root-transformed faunal density. Mean values of each parameter in each target stratum were calculated, along with corresponding C.I. 95% based on the simple percentile method110. All analyses were performed using a custom R116 script using multiple functions of the vegan package117.

Variations in ecological parameters between study areas were reported by comparing CI 95% (i.e., the upper limit of a given estimate must be lower than the lower limits of the estimate that is compared to). Such cases are significant at p < 0.05, but the true (undetermined) p-value will, necessarily, be considerably lower.

Co-existence of taxa inhabiting the same seamount were investigated by a probabilistic model of co-occurrence118. For this purpose, records of taxa densities for each annotated seafloor image per seamount were converted into presence-absence records in order to perform the probabilistic taxa co-occurrence analysis for all images of a single seamount combined using the cooccur package119 in R. The resulting co-occurrence table reports the probability p of two taxa co-occurring at the same seamount with a higher frequency pgt or lower frequency plt than observed118. When pgt < 0.05, two taxa co-occur at a higher rate than expected by chance, and when plt < 0.05, two taxa co-occur at a lower rate than expected by chance119. For the CM dataset, 145 taxa pairs were investigated and 45 pairs (24% of all combinations) were excluded from the analysis because their co-occurrence was expected to be < 1. For the KS, 135 taxa pairs were analyzed and 55 pairs (29% of all combinations) were excluded, and for the NM, 147 taxa pairs were studied and 43 pairs (23% of all combinations) were discarded.

Bulk and compound specific isotope analysis of microbial mats and sponge-associated fauna

Samples were obtained during RV Polarstern cruise PS101 from September to October 2016. Decaying sponges and bacteria mat were collected using push cores and the starfish observed on top of the sponge was collected using “Nereid Under Ice” remotely operated vehicle (NUI ROV). Samples were immediately stored on retrieval to the surface at − 20 °C for isotope analyses. Bulk stable isotope composition (δ13C, δ15N) of freeze-dried, pulverized star fish, microbial mat observed on top of a decaying sponge (Fig. 5a, SD2), and decaying sponge was measured on an elemental analyzer (EM) coupled with a Isotope Ratio Mass Spectrometer (IRMS) as described in50. PLFAs were extracted from freeze-dried, pulverized star fish and microbial mat following a modified Bligh and Dyer extraction120 as described in detail in the protocol by de Kluijver121,122.