Specimens were collected from two locations: the Mount Manon, Atalante, and Volcano A mud volcanoes at the base of the BAP (~4700–4900 m) at 13°50′N (Le Pichon et al., 1990; Olu et al., 1997) (C. methanophila), and the JMVF on the ultraslow-spreading Arctic Mid-Ocean Ridges at the Jan Mayen fracture zone at approximately 71°15′N and 005°45′W (Pedersen et al., 2010; Schander et al., 2010) (all other cladorhizids) (Figure 1). The mud volcano field at the base of the BAP features massive methane emission from diapiric structures with associated seep fauna mostly belonging to bivalves with chemosynthetic symbionts in additions to large bush-like aggregations of C. methanophila (Olu et al., 1997). Samples from the JMVF were taken mainly from the Troll Wall site (~71.25°N), a part of the rift valley close to Jan Mayen, with an additional specimen from an area of ultradiffuse venting in the rift valley characterized by biofilms dominated by Mariprofundus ferrooxydans (Pedersen et al., 2010) (Figure 2). Cladorhizids are common in the vicinity of venting sites at the JMVF, but despite the presence of free-living chemoautotrophic microorganisms at these sites (e.g., Lanzén et al., 2011) the lack of any large aggregations of biomass like those found at the BAP, could indicate that cladorhizids here benefit more indirectly from the increased prey availability at these sites (Vacelet, 2006) rather than directly deriving nutrition from chemoautotrophic symbionts.

Specimens used for microbiome sequencing were collected primarily using the Bathysaurus ROV during the University of Bergen Centre for Geobiology (CGB) R/V “G.O. Sars” 2011, 2012, and 2014 cruises to the JMVF, and with the Jason ROV during the 2012 R/V “Atlantis” 21-02 cruise to the base of the BAP. Isotope sampling included additional specimens from CGB “G.O. Sars” 2008–2014 cruises. The majority of specimens were either preserved by snap-freezing samples in liquid nitrogen (LN2), or in RNAlater. RNAlater specimens were as a rule kept at normal freezer temperatures (-20°C) during storage. We were also interested in seeing if a couple of ethanol preserved samples were viable for microbiome sequencing, and for some specimens, we ran parallels from part of the specimen fixed in 96% ethanol, and for two specimens, we sequenced from ethanol preserved material only. A list of cladorhizid specimens used for microbiome sequencing and host phylogenetic analysis is given in Table 1. Specimens used for isotope analysis are listed together with the results of the analysis in Table 2.

Stable isotope analyses (δ¹³C and δ¹⁵N) were conducted at the University of Bergen. Samples were dried at 80°C for 24 h in glass vials and grounded to powder using a glass pestle. Lipids were removed by adding 7% methanol in dichloromethane for 2 h to the dried sample and then removed using a glass Pasteur pipette and again dried for 24 h at 80°C. Inorganic carbon was removed by adding 0.5 M HCl to the samples. The time for the inorganic carbon to dissolve varied and the HCl was kept at minimum 5 min or until the reaction was finished. Acid waste was washed away by carefully adding and removing water to the sample until the pH reached 6–7. Finally the samples were dried and weighed in tin capsules and measured using a Delta V Plus isotope ratio mass spectrometer connected to a Flash EA 1112 elemental analyzer (Thermo Scientific). Isotope ratios are expressed in delta notation as ‰ difference in ¹³C/¹²C and ¹⁵N/¹⁴N isotope ratios compared to Pee Dee Belemnite (PDB) and air N2, respectively. Samples were calibrated to internationally acknowledged C and N isotope ratio standards.

Specimen subsamples were examined and visible contaminants removed and rinsed several times using Milli-Q water in a sterile environment. Extraction of grinded tissue was done using either the E.Z.N.A. Mollusc DNA kit (Omega Bio-tek) or the Qiagen Blood and Tissue kit. A water sample from the JMVF was filtered through a 0.22 μm cellulose filter that was cut into small pieces before extraction. Spicules were removed from the extract after lysis before adding ethanol.

In order to establish the phylogenetic position of C. methanophila, partial ribosomal 28S rRNA, mitochondrial gene cytochrome c oxidase subunit I (COI) and nuclear gene asparagine-linked glycosylation protein 11 (ALG11) were sequenced to allow comparison with existing cladorhizid sequences in GenBank from Hestetun et al. (2016b), which already includes other cladorhizids part of our study. A maximum likelihood analysis using RAxML 8.0.20 (Stamatakis, 2014) was performed on a concatenated dataset with the resulting ML most likely tree pictured with rapid bootstrap values (BS) indicated. Amplification procedure and analysis are identical to that of Hestetun et al. (2016b), where they are described in further detail.

For 16S rDNA sequencing, we targeted the V4–V5 regions using the universal prokaryote primers 519F (5′-CAGCMGCCGCGGTAA-3′) (Øvreås et al., 1997) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) (Herlemann et al., 2011). Ion Torrent libraries were constructed as described in detail in Jørgensen and Zhao (2016). In short: all DNA extracts were PCR amplified in triplicates and subsequently pooled and purified using QIAquick PCR purification kit (Qiagen). The purified DNA was used in a second PCR in order to attach barcodes and adaptor sequence. Resulting amplicons were purified using AMPure XP bead Purification Kit (Agencourt), following the manufacturer’s protocol, before all samples were pooled in equimolar concentrations (26 pmol). Prior to sequencing quantification was done using a Quantus Fluorometer. Amplicons were sequenced applying the Ion Torrent PGM Personal Genome Machine (PGM) platform technology (Life Technologies) with a 318 chip. The thermal PCR programs used were as follow: first round PCR: each reaction (20 μl) contained 10 μl 2x HotStarTaq^®master mixture (Qiagen), 0.2 μl of each primer (100 μM stock), 2 μl template, and ddH₂O. The PCR program was initiated with a hot start activation step for 15 min at 95°C followed by 27 cycles of 95°C for 30 s, 56°C for 30 s, and 72°C for 30 s. Second round PCR used same program but only seven cycles and contained in each reaction (25 μl) contained 12.5 μl 2x HotStarTaq^®master mixture (Qiagen), 0.2 μl 806r-B-Key (100 μM stock), and 2 μl 519f MID primer (10 μM stock), with 5 μl of purified PCR products from first-round amplification as the template, according to the Ion Torrent protocol.

Filtering (truncate length 200, max expected errors 0.5), chimera detection, removal of singletons, and OTU clustering (97% identity) were done using the UPARSE pipeline with USEARCH 8.1.1861 (Edgar, 2013). Taxonomic assignment and phylogenetic tree construction was done with the Green Genes database v 13_8 using QIIME (Caporaso et al., 2010), and 16S rRNA amplicon data was submitted to the Sequence Read Archive as biosamples with accession numbers SAMN05722969 and SAMN05722971–SAMN05722981, under BioProject accession number PRJNA341449. Bray–Curtis pairwise dissimilarity index values, cluster and MDS analyses were calculated using R package vegan, and data were visualized using R packages ggplot2 and pheatmap. C. methanophila Sanger sequences for specimens J633-5 (Cmet1) and J636-3 (Cmet4) were uploaded to GenBank with accession numbers KX815331–KX815335.

Stable isotope analysis of δ¹³C and δ¹⁵N was done on 15 samples representing five cladorhizid species (Figure 3; Table 2). The results show that all JMVF cladorhizids had isotope ratios within a quite narrow range of around -22 to -24‰ for δ¹³C and 8.5 to 10.5‰ for δ¹⁵N. The BAP C. methanophila specimens were found to have very light δ¹³C and δ¹⁵N ratios of around -60 to -66‰ for δ¹³C and 3.5 to 5.2‰ for δ¹⁵N, clearly separating it from the other cladorhizid specimens (Mann–Whitney rank-sum p-value = 0.001465).

The Cladorhizidae phylogenetic analysis recovered the two sequenced C. methanophila specimens as a close sister taxon to C. gelida with high support (BS = 100). C. gelida is a common Cladorhiza species with a wide biogeographical distribution and eurybathyal depth range, and is also included in the microbiome dataset here. Other North Atlantic arbuscular Cladorhiza species (C. abyssicola Sars, 1872, C. oxeata Lundbeck, 1905, C. corticocancellata Carter, 1876, and C. tenuisigma Lundbeck, 1905) were also recovered as close relatives within the genus (Figure 4).

Our results represent the first comparative study of microbiome composition in carnivorous sponges, expanding on the results of the single-species carnivorous sponge study by Dupont et al. (2013) as well as providing a reference for comparison against larger studies of non-carnivorous sponges like Thomas et al. (2016) and deep-sea sponges such as Kennedy et al. (2014).

The class level microbiome community structures observed in this study are similar to previously published filter-feeding sponge microbiomes, with a dominance of Gammaproteobacteria and Alphaproteobacteria (Hentschel et al., 2012; Thomas et al., 2016). Other major groups in our study are Bacteroidetes and Thaumarchaeota, and Bacteroidetes was found to be the dominating group in L. cupressiformis (Figure 6). Thaumarchaeota are often found as a secondary group in several studies (Taylor et al., 2007; Thomas et al., 2016). Reports of Bacteroidetes are less common, but have been reported for deep-sea sponges (Kennedy et al., 2014), including the previously sequenced microbiome of the carnivorous sponge L. hypogea (Dupont et al., 2013).

At higher taxonomic resolution (family and genus level), large variation in the different species microbiomes is evident (Figure 7). This is consistent with previous surveys reporting on high degrees of host specificity among sponge microbiomes (Hentschel et al., 2012; Thomas et al., 2016). Moreover, some of the results from sequencing multiple replicates from the same specimen (Figure 5) show considerable variation also within each sample, especially for Thiohalorhabdales. While it is possible that the fixative has partly influenced the results in this particular case here (see section “Run Statistics”), the results are in accordance with an uneven distribution of microbial organisms within the host sponges, as previously reported through TEM observations by Riesgo et al. (2007).

Interestingly, an unknown Thiohalorhabdales, and Oceanospirillales, both from the SUP05 clade as well as others, were abundant across all species in the dataset (Figure 7; Table 4). The only described species of Thiohalorhabdales is Thiohalorhabdus denitrificans, an obligate chemolithoautotrophic sulfur-oxidizing bacterium from hyper-saline habitats (Sorokin et al., 2008). Thiohalorhabdales have been reported in low concentrations from sea water with abundance correlated with nitrite concentration (Frank et al., 2016). The Oceanospirillales SUP05 group is also associated with sulfur oxidation, partly at vent and seep sites (Anantharaman et al., 2013). It has been reported both as free-living bacteria and as an endosymbiont (Marshall and Morris, 2013).

We found small to moderate amounts of Thaumarchaeota Marine Group 1 reads belonging to Nitrosopumilus in parts of our dataset (Figure 7; Table 4), mainly in the case of C. gelida. This group was also found in moderate quantities in the JMVF water sample. Members of Nitrosopumilus are common in seawater and grow by oxidizing ammonium to nitrite (Könneke et al., 2005). Interestingly, sponge symbiont archaeal nitrification has been reported from several studies (e.g., Hoffmann et al., 2009; Liu et al., 2012; Radax et al., 2012), and Nitrosopumilus was also found to be the dominating archaeon in part of the specimens in the deep-sea sponges in Kennedy et al. (2014), implying a possible role for this group within carnivorous sponges.

Sulfur- and ammonia-oxidizing symbionts seem to be common in non-carnivorous deep-sea sponges (Nishijima et al., 2010; Arellano et al., 2013), implying that chemolithoautotrophy could be a widespread supplementary source of nutrition for deep-sea sponges in general (Kennedy et al., 2014). Though the taxonomic affinity of the major putative sulfur-oxidizing groups in our dataset (Thiohalorhabdales, Oceanospirillales) were partly different than that of Kennedy et al. (2014) (chiefly Chromatiales, but also some Oceanospirillales and others), this raises the possibility that partial nutrient acquisition from sulfur oxidation could be common in carnivorous sponges as well, though as evidenced by the isotope signatures from our study (Figure 3) such activity would be a supplement to carnivory rather than the major mode of nutrition.

Other significant groups present in the dataset were identified as associated with several marine bacterial taxa: For the Alphaproteobacteria, the two major OTUs were somewhat evenly split between an unknown Rhodobacteraceae and the marine genus Octadecabacter. The Bacteroidetes were, excepting an unknown Flavobacteriaceae in L. cupressiformis, almost exclusively Polaribacter, a genus containing several gas vacuolated, marine bacterial species (Gosink et al., 1998) (Figure 7; Table 4). (For Methylococcales, see section “Trophic Considerations and Extent of Methanotrophy in Carnivorous Sponges”).

The habit of C. methanophila is arbuscular, with a great number of filaments extending in all directions from the branches. Together with the types and sizes of its skeletal spicules, the morphology is similar to and suggests a close affinity to a group of Cladorhiza species from the North Atlantic. This group includes the three species C. abyssicola (type species of the genus), C. corticocancellata and C. gelida, also examined here, as well as C. iniquidentata Lundbeck (1905), C. oxeata, and C. tenuisigma (Vacelet and Boury-Esnault, 2002; Hestetun et al., in press). The phylogenetic relationships and systematic classification of the carnivorous sponges were investigated in a recent study (Hestetun et al., 2016b), which included most of the North Atlantic cladorhizid species, and sequences from that study were used here with the addition of C. methanophila in order to recover its phylogenetic position.

In view of the morphological similarities, the recovery of C. methanophila as a close sister taxon to the other arbuscular North Atlantic Cladorhiza species in the dataset (Figure 4) is not very surprising. C. methanophila has also been reported from the 15°20′N fracture zone segment of the Mid-Atlantic Ridge (Hestetun et al., 2015), and has a known depth range of 2600–4900 m. Most other arbuscular Cladorhiza species have a somewhat shallower distribution (mainly 200–2000 m), but the closest sister taxon, C. gelida, has a eurybathyal range, reported from 500 to 3500 m (Hestetun et al., in press). Given the lack of any evidence of methanotrophic symbiosis in other closely related cladorhizids in the dataset here, a likely hypothesis is that methanotrophy in C. methanophila is a rather recent event, possibly occurring as an adaptation to a vent or seep locality either at the BAP or the Mid-Atlantic Ridge.

As deep-sea sponges, cladorhizids are often found in greater concentrations in the enrichment zones surrounding vent and seep systems, where prey is comparatively abundant. However, the large size and great abundance of the shrub-like aggregations of C. methanophila at the BAP greatly exceeds normal concentrations of carnivorous sponges, even at other vent sites. Though captured prey items show that C. methanophila still retains the ability to acquire nutrients through prey capture (Vacelet and Boury-Esnault, 2002), this suggests that this sponge is able to derive a large amount of nutrients from its chemoautotrophic association (Vacelet et al., 1995; Olu et al., 1997) (compare Figure 2 for a visual impression).

Results showed that an undescribed Methylococcales OTU was recovered in large abundance from C. methanophila specimens, supporting earlier findings describing symbiosis between this species and methanotrophic symbionts. The isotope results, in particular the very light δ¹³C ratios of -60.5 to -66.0‰, are very strong indicators of a symbiotic origin for organic carbon in the host sponge, and consistent with δ¹³C ratios reported in other symbioses from e.g., Gulf of Mexico vent (-51 to -57‰) (Childress et al., 1986) and seep (-74‰) (Cavanaugh et al., 1987) (-45.4 to -39.6‰ and -68.4 to -78.8‰) (MacAvoy et al., 2002) mussels. Surprisingly, they are also lighter than ratios reported for C. methanophila from the BAP by Vacelet et al. (1995) (-48.4 to -48.8‰), which could be related to changes in source carbon signature or precise sampling location.

The high sponge concentrations at the BAP, previous TEM and isotope values (e.g., Vacelet et al., 1995), and microbiome and isotope results from our study (Figure 3; Table 2), clearly show that this species obtains a majority of nutrients from its symbiont relationship rather than through prey capture. The only other sponge where methanotrophic symbiosis has been reported is Myxilla (Ectyomyxilla) methanophila Maldonado and Young (1998). However, here the dominating methanotroph is Methylohalomonas (Chromatiales, Ectothiorhodospiraceae) (Arellano et al., 2013).

In contrast to C. methanophila, close relatives including C. abyssicola, C. corticocancellata, and C. gelida, did not have isotopic ratios suggesting a large contribution from chemolithoautotrophic symbiosis (Figure 3) (see Sweetman et al., 2013 for other JMVF biota). Gammaproteobacteria in groups related to known sulfur-oxidizing bacteria were prominent in the microbiome across different cladorhizids in this study. Hydrogen sulfide is present in high concentrations in the high-temperature fluids venting from white smokers at the Troll Wall site at the JMVF (Baumberger, 2011; Dahle et al., 2015). However, sulfur oxidation, which seems to be a common feature in deep-sea sponges, may, if indeed it signifies chemolithoautotrophy, account for a smaller portion of the nutrient uptake than the methanotrophic symbiosis in C. methanophila. Thus, chemolithoautotrophic symbiosis as a dominant form of nutrient acquisition within carnivorous sponges seems to be an isolated case rather than a more widespread phenomenon in currently reported species.