Introduction

Towards the end of the nineteenth century, ship-based dredging expeditions of the HMS Porcupine, Lightning, and Challenger started to reveal the occurrence of sponges at water depths up to a 1000 m1,2. Since then, many sponge species have been found to be abundant in various deep-sea ecosystems, such as cold-water coral reefs and sponge grounds3,4,5. In the North-Atlantic Ocean, massive ball-shaped demosponges from the Geodiidae family form large aggregations known as “osturs” or “cheese bottoms”6, while thin encrusting species occur in high diversity and abundance on the Rockall Bank west of Scotland and the Porcupine Bank west of Ireland7. Hexactinellids form glass sponge grounds, either monospecific or as part of a more diverse community, such as the Schulz Bank on the Mid-Atlantic Ridge8 and the Vazella pourtalesii grounds on the Scotian Shelf9,10. Despite the fact that the deep-sea is generally considered to be barren and food-limited, these deep-sea sponge ecosystems are known to be hotspots of biodiversity and organic carbon cycling10,11,12.

Nevertheless, the ecological role of sponges in the deep-sea is still relatively undescribed, although an increasing body of literature is starting to reveal their importance. They promote biodiversity by increasing habitat complexity10,13,14 and provide foraging and nursery grounds for mobile taxa, such as fish15,16. Moreover, deep-sea sponges are important in benthic-pelagic coupling of resources17,18. However, while deep-sea ecosystems are increasingly threatened by anthropogenic impacts, such as oil and gas exploration19, mining20,21, and fisheries22,23, many questions concerning the physiology and ecological functioning of sponges are still unanswered. A particularly relevant question is how the deep-sea sponge holobiont, i.e. the animal host and its numerous and diverse community of associated microorganisms24,25, takes up, processes, and releases nutrients in the seemingly food-deprived deep-sea environment.

In general, sponges are efficient filter-feeders that utilize a wide variety of particulate food sources ranging from bacterio- and phytoplankton to detrital particles, and even zooplankton18,26,27. Moreover, in recent years it has become increasingly evident that many sponges, including dominant North Atlantic deep-sea species28,29, use dissolved organic matter (DOM) as main food source30. DOM constitutes the largest pool of reduced carbon in the ocean, and is therefore the largest potential marine organic food source31. However, DOM appears to represent only a minor fraction of the food intake of most invertebrates32, and is generally considered to be recycled through bacterial processing33. Consequently, when various shallow water sponge holobionts were found to feed predominantly—often 70 to > 90% of their daily carbon intake—on DOM34,35,36, this capacity was attributed to their microbial symbionts37,38,39. Sponges with high abundances of associated microbes (HMA; species-specific microbial communities with up to four orders of magnitude higher bacterial concentrations than the surrounding sea water) were suggested to be better equipped to utilize DOM, than those with low microbial abundances (LMA; community composition and abundances of microbes similar to surrounding sea water)39,40,41. However, recent evidence from complementary studies using different approaches suggests that microbial abundance may not determine the capacity of sponges to feed on DOM, since many LMA sponges were found to take up DOM29,35,36,42 and some even at higher uptake rates than HMA sponges28,43,44. Furthermore, DOM has been shown to be assimilated by both sponge host and bacterial symbiont cells42,44, ultimately visualized at a (sub)cellular level45,46. However, the initial uptake and possible translocation between host and symbionts has yet to be established.

The capability to utilize dissolved food sources might be especially relevant in highly food-deprived environments, such as the deep-sea, where food availability is considered to primarily dependent on a vertical flux of sinking particular organic matter and local bacterioplankton production47,48. Indeed, first evidence of four dominant North-Atlantic sponge species, including one massive vase-shaped LMA hexactinellid, one encrusting sheet-shaped LMA and two massive ball-shaped HMA demosponges, showed that none of these species could acquire sufficient carbon from particulate organic matter (POM) food sources alone to meet their respiratory demand. Instead, all tested species relied on DOM to balance their metabolic requirements28. However, no information is yet available on the role of the animal host versus its symbionts in the processing of particulate and dissolved organic food sources, which hampers our understanding of different strategies these deep-sea sponges may have to acquire food.

We therefore studied how deep-sea sponges from different phylogenetic classes, and with high and low abundances of microbial symbionts process dissolved and particulate food sources. We traced the assimilation and respiration of two 13C- and 15N-enriched food sources (DOM and bacteria) by three different (i.e. based on phylogeny, abundances of associated microbes, morphology) dominant North Atlantic deep-sea sponge species—Geodia barretti (Demospongiae, HMA, massive), Vazella pourtalesii (Hexactinellidae, LMA, massive), and Hymedesmia paupertas (Demospongiae, LMA, encrusting)—using ex situ incubations. Subsequently, phospholipid-derived fatty acid (PLFA) biomarker analysis was performed to elucidate the role of the sponge host and the associated microbes in the utilization of the two isotope-tracer food sources. The use of PLFA biomarkers in combination with isotope tracer experiments has been shown to successfully provide insight in differential processing of isotope-tracer food sources by sponge host cells and bacterial symbionts in various studies e.g., Refs.41,42,44.

Results

Bulk tracer processing of dissolved organic matter (DOM) and bacteria

Both the algal-derived dissolved and bacterial-derived particulate organic food sources were assimilated and respired by all three sponge species (Fig. 1).

Figure 1
figure 1

Processing of tracer (13C- and 15N-) DOM (left) and bacteria (right) by deep-sea sponges. Dark red bars represent tracer carbon (CDOM or Cbac) assimilation rates, light red bars represent tracer carbon respiration rates and blue bars represent tracer nitrogen (NDOM or Nbac) assimilation rates in µmol C or Nfood source mmol C or \({\text{N}}_{{{\text{sponge}}}}^{ - 1}\) d−1. HMA high microbial abundance sponge species, LMA low microbial abundance sponge species.

DOM processing rates and assimilation efficiency

The high microbial abundance (HMA) sponge Geodia barretti showed significantly lower average uptake rates of carbon from the dissolved food source (CDOM) (0.7 ± 0.6 µmol CDOM mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1; mean ± SD throughout text unless stated otherwise) than the low microbial abundance (LMA) species Hymedesmia paupertas and Vazella pourtalesii (2.9 ± 0.8 and 3.5 ± 0.7 µmol CDOM mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1, respectively) (Fig. 1, Table 1). These differences were mainly caused by the order of magnitude lower CDOM respiration rates of the HMA species (0.09 ± 0.05 µmol CDOM mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1) compared with the two LMA species H. paupertas and V. pourtalesii, respectively (4.7 ± 0.8 and 1.4 ± 0.4 µmol CDOM mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1, Fig. 1). No significant differences were found in nitrogen (NDOM) uptake between G. barretti (0.4 ± 0.3 µmol NDOM mmol \({\text{N}}_{{{\text{sponge}}}}^{ - 1}\) d−1), H. paupertas (1.7 ± 0.8 µmol NDOM mmol \({\text{N}}_{{{\text{s ponge}}}}^{ - 1}\) d−1), and V. pourtalesii (0.8 ± 0.2 µmol NDOM mmol \({\text{N}}_{{{\text{sponge}}}}^{ - 1}\) d−1) (Table 1). Despite possessing the lowest processing rates, the HMA sponge G. barretti showed a higher assimilation-to-respiration efficiency (77%) compared to H. paupertas and V. pourtalesii (50 and 32%, respectively), although only the difference between G. barretti and V. pourtalesii was significant (Fig. 2, Table 2).

Table 1 Results of the pairwise comparisons for two-factor Monte Carlo PERMANOVAs testing for differences in total dissolved organic carbon and nitrogen (CDOM and NDOM) and bacterial carbon and nitrogen (Cbac and Nbac) processing rates between the deep-sea sponge species Geodia barretti (Demosponge, HMA), Hymedesmia paupertas (Demosponge, LMA) and Vazella pourtalesii (Hexactinellid, LMA).
Figure 2
figure 2

Relative assimilation and respiration of 13C-tracer DOM (left) and bacteria (right). Dark red bars represent relative tracer carbon (CDOM or Cbac) assimilation, light red bars represent relative tracer carbon respiration, as % of total processing of each food source per sponge species tested. HMA high microbial abundance sponge species, LMA low microbial abundance sponge species.

Table 2 Results of the pairwise comparisons for the two-factor Monte Carlo PERMANOVAs testing for differences between the deep-sea sponge species Geodia barretti (Demosponge, HMA), Hymedesmia paupertas (Demosponge, LMA) and Vazella pourtalesii (Hexactinellid, LMA), in the assimilation-to-respiration efficiencies of dissolved organic carbon (CDOM) and bacterial carbon (Cbac).

Bacteria processing rates and assimilation efficiency

For the bacterial food source, significant differences in processing rates were found between the hexactinellid V. pourtalesii, showing the highest total bacterial C and N processing rates (1.1 ± 0.4 and 1.6 ± 0.5 µmol C and Nbac mmol C and \({\text{N}}_{{{\text{sponge}}}}^{ - 1}\) d−1), and the two demosponges H. paupertas (0.3 ± 0.3 and 0.7 ± 0.3 µmol C and Nbac mmol C and \({\text{N}}_{{{\text{sponge}}}}^{ - 1}\) d−1) and G. barretti (0.2 ± 0.1 and 0.6 ± 0.1 µmol C and Nbac mmol C and \({\text{N}}_{{{\text{sponge}}}}^{ - 1}\) d−1) (Fig. 1, Table 1). Respiration rates were low, 0.04 ± 0.04, 0.09 ± 0.11 and 0.07 ± 0.02 µmol Cbac mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1 for G. barretti, H. paupertas and V. pourtalesii, respectively, and comparable between the three species. The hexactinellid V. pourtalesii also showed the highest assimilation-to-respiration efficiency for the bacterial food source (97%), compared with demosponges H. paupertas and G. barretti (68 and 84%), although this efficiency was only significantly different between V. pourtalesii and H. paupertas (Fig. 2, Table 2).

C:N ratios of assimilation of DOM and bacterial food source

Averaged over all sponge species, C:N ratios for the assimilation of the DOM food source were significantly higher than for the bacterial food source (1.2 ± 0.3 versus 0.4 ± 0.2, respectively (t = 7.0, df = 17, p < 0.0001)).

Sponge host and symbiont tracer processing of DOM and bacterial food source

PLFA profiles of deep-sea sponges and food substrates

The three investigated sponge species possess distinct PLFA profiles (Fig. 3). Total PLFA carbon was on average 0.7 ± 0.6% (V. pourtalesii, n = 6), 0.9 ± 0.4% (G. barretti, n = 11) and 1.9 ± 0.6% (H. paupertas, n = 7) of the total sponge holobiont carbon biomass. The HMA sponge G. barretti showed the highest relative contribution of bacteria-specific PLFAs (62.7 ± 3.0%; depicted in red in Fig. 3), which is significantly higher than LMA sponges H. paupertas (8.6 ± 6.9%) and V. pourtalesii (17.7 ± 23.9%), respectively (Table S4). In contrast, the proportions of sponge-specific PLFAs were at least four times higher in the LMA species H. paupertas and V. pourtalesii (68.7 ± 11.1% and 58.0 ± 52.1%, respectively) compared to the HMA sponge G. barretti (16.4 ± 3.0%) (Table SS4. The sponge-specific PLFA biomarkers (depicted in blue in Fig. 3) of the demosponges H. paupertas and G. barretti contained typical very-long chained demosponge PLFAs, such as C26:2, C28:2, C26:3 and C28:3. Sponge-specific PLFAs of the LMA hexactinellid sponge V. pourtalesii were dominated by one very long-chained PLFA (C30:3ω7). A considerable part of the PLFAs of all three species could not be assigned exclusively to either bacterial or sponge origin (20.9–27.2%) and are depicted in grey in Fig. 3. An overview of the relative abundance of each PLFA per species is presented in Supplementary Table S3 online.

Figure 3
figure 3

Phospholipid-derived fatty acid (PLFA) profiles and total percentages of bacteria-specific, sponge-specific, and unspecific PLFAs of deep-sea sponges. Bacteria-specific (red), sponge-specific (blue), and unspecific PLFAs (grey) are shown for (A) Geodia barretti (n = 11), (B) Hymedesmia paupertas (n = 7), and (C) Vazella pourtalesii (n = 6). Values are presented as mean (± SD) percentages of the total PLFA content. Stacked bars depict the total relative contributions of bacteria-specific (red), sponge-specific (blue), and unspecific (grey) PLFAs for each species. HMA high microbial abundance, LMA low microbial abundance.

PLFA profiles of both food sources: two algal-derived DOM types (cyanobacteria-DOM fed to V. pourtalesii, and diatom-DOM fed to G. barretti and H. paupertas) and bacteria, are shown in supplementary Fig. S3 and Table S3. DOM obtained from the lyophilized cyanobacterium Agmenellum quadruplicatum contained mainly C16:0 (38.4%) and C16:1 (36.9%). The diatom-derived DOM also contained mainly C16:1 (64.8%), C16:0 (12.7%), and the typical diatom biomarker C20:5ω3 (10.2%). The bacterial culture only contained C16:0 (22.4%), C18:1 (59.9%) and C18:0 (17.7%) in detectable amounts.

Assimilation of DOM and bacterial food source in host and symbionts

The stable isotope-enriched tracer DOM and bacterial food source were clearly assimilated into PLFAs of all three deep-sea sponge holobionts (Fig. 4). Demosponges G. barretti (HMA) and H. paupertas (LMA), showed assimilation rates of DOM into total PLFAs of one order of magnitude higher (0.02 ± 0.01 and 0.02 µmol ± 0.01 CDOM mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1, respectively) than for bacteria (0.002 ± 0.001 and 0.003 ± 0.002 µmol Cbac mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1, respectively). In contrast, the opposite was found in the hexactinellid LMA V. pourtalesii, with one order of magnitude lower assimilation rates of DOM into PLFAs (0.0004 µmol CDOM mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1) compared to the bacterial source (0.003 ± 0.002 µmol Cbac mmol \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1) (Fig. 4). ). For DOM-fed sponges, the majority of labelled PLFAs found in G. barretti (62.9%), for H. paupertas (66.9%) and for V. pourtalesii (55.0%) originated from PLFAs present in the food sources (e.g., C14:0, C16:0, C16:1, C18:0, C18:1, C20:5ω3, depending on the source, indicated with the arrows in Fig. 4). For the bacterial food source, the dominant fatty acids C16:0, C18:1 and C18:0 accounted for 24.5% (G. barretti), 34.8% (H. paupertas) and 89.3% (V. pourtalesii) of the total tracer PLFA incorporation. This indicates that the major fatty acids of both food sources were directly ingested by the sponges. Additionally, both food sources were traced in PLFAs that were not originally present in the food source, indicating de novo synthesis of these PLFAs by the sponge holobionts. Most of these newly synthesized PLFAs are either bacteria-specific (i/aiC15:0, iC17:0, Cy-C17, Cy-C19, Me-C16:0, Me-C18:0) or non-specific (C16:1) in all sponge species and irrespective of food source (8–45% of total tracer carbon assimilation depicted in red, Fig. 4, Supplementary Fig. S4 online). Sponge-specific PLFAs only contributed marginally to the newly synthesized PLFA fraction (0–8% depicted in blue, Fig. 4, Fig. S4), irrespective of food source and sponge species. However, a large part of the labelled PLFAs were neither bacterial- nor sponge specific (54–91% depicted in grey, Fig. 4, Supplementary Fig. S4 online), and could thus be de novo synthesized by either the host or its symbionts. Interestingly, bacteria-fed sponge specimens showed a higher percentage of labelled sponge-specific PLFAs compared to DOC-feed specimens in all tested sponge species (G. barretti: 4.5 vs. 8.8%, H. paupertas: 2.8 vs. 6.0%, V. pourtalesii: 0.0 vs. 0.2%). However, the combined effect of species and food sources was not statistically significant (df = 2, F = 0.359, pperm = 0.78).

Figure 4
figure 4

Incorporation of DO13C and bacterial 13C tracer into sponge-specific, bacteria-specific, and unspecific phospholipid-derived fatty acids (PLFAs). Values are presented as percentages of total CDOM and Cbac tracer incorporated into each individual PLFA (mean ± SD). Bacteria-specific (red), sponge-specific (blue), and unspecific PLFAs (grey) are shown for (A) Geodia barretti (DOM: n = 3, bacteria: n = 4), (B) Hymedesmia paupertas (DOM: n = 3, bacteria: n = 3), and (C) Vazella pourtalesii (DOM: n = 1, bacteria: n = 3). Green arrows indicate the presence of the fatty acid in the DOM or bacterial source fed to the sponge. HMA high microbial abundance, LMA low microbial abundance. Total PLFA incorporation rate is presented in the box in the upper right corner of each panel in µmol CDOM or \({\text{C}}_{{{\text{sponge}}}}^{ - 1}\) d−1 (mean ± SD).

Discussion

Here, we show the processing (i.e. assimilation and respiration) of a dissolved (DOM) and a particulate (bacteria) food source for a selection of three dominant deep-sea sponges from two different phylogenetic classes that differ in functional traits, such as the abundance of microbial symbionts and morphology. Our results corroborate recent and increasing evidence28,30,49 that DOM-processing is not restricted to high microbial (HMA) sponges as is commonly suggested40,41,50, but that low microbial abundance (LMA) sponges are capable of processing DOM at even higher rates than HMA species. However, the HMA sponge in this study, Geodia barretti, did show the highest assimilation-to-respiration efficiency of DOM, i.e. relatively more food is put into new biomass instead of lost through respiration. Overall, assimilation-to-respiration efficiencies were lower for DOM than for bacteria. For the particulate food source (bacteria), not bacterial abundance (i.e. HMA or LMA), but phylogenetic class distinguished its processing rates, with highest rates found in the hexactinellid compared to the two demosponges. The hexactinellid Vazella pourtalesii also showed a very high assimilation-to-respiration efficiency (97%) and higher processing rates into compound-specific phospholipid fatty acids (PLFAs) for bacterial food compared with DOM, whereas the demosponges G. barretti and Hymedesmia paupertas showed the opposite. Both tracer food sources were, within the time frame of the incubation experiments, foremost assimilated into shorter-chained microbial symbiont specific PLFAs by all sponges, and not in the very-long-chained sponge host-specific PLFAs. However, this does not quantify symbiont and host processing, since the majority of DOM and bacterial food tracers were found in non-host/symbiont-specific PLFAs. Based on the observations made here on a limited number of tested species, we cannot draw conclusions on the exact drivers (e.g., phylogeny, abundance and composition of microbiome, morphology) in the processing of dissolved and particulate food sources. To fully understand how phylogenetic and anatomical differences affect food processing by deep-sea sponges, more studies on a broader spectrum of sponge types are needed.

Bulk processing of DOM and bacteria as food source

It is commonly assumed that sponges with high microbial abundances are better equipped to take up dissolved food than LMA sponges40,41,50. However, DOC processing rates (i.e. the sum of assimilation and respiration) for the LMA species studied here (H. paupertas and V. pourtalesii) are four to five times higher than for the HMA sponge G. barretti, mainly due to higher respiration of the DOM source. Overall, assimilation rates (0.6–1.4 µmol CDOM mmol Csponge d−1, 0.4–1.7 µmol NDOM mmol Nsponge d−1) are in the same range as the only reported rates of (deep-sea coral-derived) DOM processing for the deep-sea LMA sponge Hymedesmia coriacea (1.7 ± 1.6 µmol CDOM mmol Csponge d−1, 2.0 ± 2.0 µmol NDOM mmol Nsponge d−129. Thus, despite differences in microbial abundances or DOM sources (i.e. diatom-, cyanobacterial-, coral-derived DOM), DOM-tracer assimilation rates seem to be comparable among deep-sea sponges.

The assimilation-to-respiration efficiencies indicate that DOM is assimilated most efficiently by the HMA sponge G. barretti, even though at lower rates. The complex interaction between sponge host and its abundant community of microbial symbionts might result in an uncoupling between uptake, assimilation, and respiration, leading to the very low respiration rates of DOM found for G. barretti. For example, respired C from the DOM source may be fixed by microbial symbionts into organic matter via chemoautotrophy. In fact, Van Duyl et al.51,52 showed that sponge holobionts are capable of fixing inorganic C. Additionally, G. barretti is known to possesses multiple associated bacteria phyla that support CO2 fixation, such as Nitrospinae and Chloroflexi53,54,55.

In all sponges, assimilation-to-respiration efficiencies appeared to be highest for the bacterial food source. Especially the hexactinellid V. pourtalesii showed a stunning 97% assimilation-to-respiration efficiency. This corroborates earlier findings that sponges, and hexactinellids in particular, are very efficient in filter-feeding and assimilating (tracer) bacteria18,42. Still, DOM potentially constitutes a much larger proportion of their daily diet than bacteria28, mainly due to the order of magnitude higher ambient concentration of DOM-derived C and N in seawater35. The different assimilation-to-respiration efficiencies of DOM versus bacteria suggest that food sources may serve different purposes for sponge nutrition as was previously hypothesized by Refs.28,56. This is further corroborated by the significantly lower C:N ratios of bacterial assimilation compared with the assimilation of DOM for all species, which indicates that sponges differentially process C and N from these sources, i.e. process relative more N from particulate over dissolved food sources. Preferential assimilation of bacterial N has also been found for the cold-water sponge Spongosorites coralliophaga56. Considering the high dissolved inorganic nitrogen (DIN) efflux rates found in many sponges from both tropical and deep-sea habitats54,57,58, preferential assimilation of N might facilitate the sponge’s maintenance of stoichiometric homeostasis. Additionally, bacteria contain high fractions of essential constituents, such as amino acids, fatty acids, and vitamins59, which are essential building blocks for anabolic processes. In contrast, algal-derived DOM, such as the here used diatom- and cyanobacteria-DOM, contains a relatively high fraction of neutral sugars60,61, which can be rapidly respired. Glucose, for example, was found to be almost exclusively respired by sponges35,56. Therefore, we hypothesize that DOM may serve as the main energy source for deep-sea sponges to sustain their minimal energetic requirements, while supplementation with bacteria and other high-quality particulate food sources is essential to support anabolic processes (e.g., somatic growth, reproduction, and cell turnover), particularly during episodic food pulses after phytoplankton blooms. This also explains why sponges can be found in areas with locally enhanced particle supply, even where this supply of particulate organic matter (POM) alone is not enough to sustain their minimal respiratory demands62.

It is important to mention that both bacteria and DOM used as food sources in this study were laboratory-made. This could affect our results as bacteria grown in culture may have different C- and N-contents compared to bacteria naturally occurring in seawater63. Furthermore, the composition and bioavailability of freshly produced, artificial DOM (lysed diatom/cyanobacterial cells) may differ from the marine DOM pool, particularly in the deep-sea, where DOM is presumed to be largely refractory44,64. Nevertheless, all deep-sea sponges used in this study proved capable of utilizing dissolved food sources, irrespective of the abundance of associated microbes. Yet, the processing of DOM within sponge holobionts differed between species.

Processing of dissolved and particulate food sources into host and symbiont PLFAs

Our results on the assimilation of labelled DOM and bacteria into sponge host and symbiont PLFAs suggests that within sponge holobionts, DOM is primarily assimilated by bacterial symbionts, which corroborates with earlier findings by Rix and colleagues29,44. However, we cannot be conclusive here for three reasons: (1) A large portion of DOM assimilated into PLFAs (54–91%) could not be assigned to either host- or symbiont-specific biomarkers. (2) Assimilated DOM may have been incorporated into sponge cells, but not (yet) metabolized into PLFAs. For deep-sea sponges, it might simply take longer than the 24–48 h duration of the incubations to synthesize long-chained, sponge-specific PLFAs from shorter precursor PLFAs65. A recent study by Achlatis et al.46 visualized at a subcellular level that sponge cells, in contrast to bacterial cells, first store carbon in other cellular components or in other lipids than PLFAs. (3) Sponge-mediated assimilation may have been underestimated, since part of the newly synthesized sponge-specific PLFAs may have been already lost through cell turnover. Evidence is accumulating in tropical sponges that choanocytes are the dominant sponge cells that process DOM45,46 and these cells may be partly lost as detrital waste through shedding41,45,66. Cell loss was not monitored here, due to the methodological complexity, yet should be incorporated into future (tracer) metabolic studies on deep-sea sponges.

De novo synthesis of PLFAs confirms the active processing of labelled food sources, rather than the simple re-use of source PLFAs. For example, the labelled bacteria-specific PLFAs Cy-C17:0, Me-C16:0 and Me-C18:0 were not present in the bacterial food source, but could only be found in the sponge-holobiont at the end of the incubation, thus indicating de novo synthesis by sponge-associated bacteria. In the same incubations, also labelled sponge-specific PLFAs (e.g., C22:1ω7, C24:1ω9, i-C25:1ω7, C25:1, C28:2ω7) were found in all tested deep-sea sponge species after the incubations. Similarly, incubations with labelled DOM as food source resulted in the synthesis of labelled bacteria-specific (e.g., i/aiC15:0, i-C17:0, Me-C16:0, Cy-C17:0, Me-C18:0) and sponge-specific PLFAs (G. barretti: i-C25:1ω7, C26:2ω7, Me-C26:2ω7; H. paupertas: C26:2ω7, C27:3, C27:1, C28:3ω7, C28:2ω7), that were not present in the DOM substrate provided. This confirms that while 13C from bacteria is more efficiently assimilated into PLFAs by the sponge holobiont, also DOM can serve as source material for the de novo synthesis of bacterial- and sponge-specific PLFAs. Thus, both associated bacteria and the sponge host are involved in the processing of particulate (i.e. bacteria) and dissolved food (i.e. DOM).

Conclusion

Deep-sea sponges are capable of assimilating C and N from both dissolved and particulate food sources, but differentially process the two types of food. Contrary to the conventional view, the LMA sponges tested here, H. paupertas and V. pourtalesii processed DOM at higher rates than the HMA sponge G. barretti, but at lower assimilation-to-respiration efficiencies. For bacteria, the highest assimilation-to-respiration efficiency and processing rates were found in the hexactinellid LMA species V. pourtalesii. This sponge also showed a higher incorporation rate of bacteria over DOM into PLFAs, opposite of the rates found for the demosponges G. barretti and H. paupertas. We hypothesize that, in general, bacteria are more efficiently assimilated and serve as an N source to support anabolic processes of deep-sea sponges, while DOM primarily serves as energy source to sustain maintenance metabolism of the sponge holobiont. Our results further indicate that the phylogenetic class and functional traits, such as abundance of microbial symbionts, of sponges influence their food preferences and diet composition, which further helps to understand the role of sponges as key ecosystem engineers of the deep-sea.

Materials and methods

Study areas, sponge collection, and maintenance

This study used the following deep-sea sponge species: Vazella pourtalesii (Hexactinellidae, LMA), Geodia barretti (Demospongiae, HMA) and Hymedesmia paupertas (Demospongiae, LMA) (Fig. S1, Supplementary Table S1 online). All sponges were analysed by transmission electron microscopy to determine species and LMA/HMA status [Refs.58,67, Hans-Tore Rapp personal observations, data not shown]. Sponges were collected by remotely-operated vehicle (ROV) during two research cruises in 2016 and 2018. Whole V. pourtalesii individuals were collected attached to their rocky substrate at ~ 300 m water depth, during the Hudson cruise 2016–2019 (September 2016) at the Emerald Basin (43°59′49.0″N 62°46′15.7″W), an area with a large monospecific V. pourtalesii population68. Sponges were kept in the dark in a 1000-L holding tank and transported without air exposure to the Bedford Institute of Oceanography, Dartmouth, Nova Scotia, Canada. In the lab, sponges were kept in the dark in a 1000-L flow-through holding tank, through which sand-filtered seawater from the Bedford Basin was continuously pumped at 7 L h−1. A chiller was used to maintain water temperature at 8 °C. Whole G. barretti and H. paupertas individuals were collected attached to their rocky substrate during the G.O. Sars cruise 2018108 (August 2018) in the western Barents Sea (70°47′13.9″N 18°03′23.8″E), an area dominated by large G. barretti individuals69. These sponges were kept on board the research vessel in the dark in 20-L flow-through tanks in a climate room at 6 °C. Seawater was directly pumped in from a water depth of 6 m at 30 L h−1. All (13C- and 15N-) tracer DOM incubations with H. paupertas and G. barretti were performed on board the ship and tracer bacteria incubations were performed at the University of Bergen, Norway. Sponges were transported without air exposure. In Bergen, sponges were kept in a dark climate room (8 °C) in multiple 20-L flow-through aquarium systems. Flow originated from unfiltered water pumped from 200 m water depth from the outer fjord near Bergen at ~ 50 L h−1 with a temperature ranging from 6–8 °C.

Preparation of 13C- and 15N-labelled food sources

13C- and 15N-enriched dissolved (algal-derived) and a particulate (i.e. bacterial) organic food source tracer were administered to the three sponge species: individuals of each species received either the dissolved organic (DOM) or the bacterial food source (see below for details on replication) A detailed culture and extraction protocol for both food sources can be found in Supplementary Materials and Methods online. For V. pourtalesii, tracer DOM was extracted from 1 g 13C- and 15N-labelled lyophilized algal cells (Agmenellum quadruplicatum) (Cambridge isotopes CNLM-455-1, Eurisotop). For G. barretti and H. paupertas, tracer DOM was extracted from axenic diatom (Phaeodactylym tricornicum) cultures on F/2 medium70 amended with 80% 15N-NaNO3 (Cambridge isotopes NLM-157, Eurisotop) and 100% 13C-NaHCO3 (Cambridge isotopes CLM-441, Eurisotop). Different algal-derived DOM-sources were used because in 2016, time constraints did unfortunately not allow us to prepare diatom algal batch cultures before the experiments with V. pourtalesii were performed, therefore DOM was extracted from pre-labelled algae, directly obtained from Cambridge isotopes. Both DOM-sources were obtained by lysing the algal cells and filtering the lysate over a 0.2 µm polycarbonate filter. The filtrate was collected, lyophilized, and analysed by EA-IRMS for C and N content and isotopic composition. Before adding the DOM to the incubations, aliquots of 5 mL were made by dissolving the lyophilized DOM in MilliQ. Tracer bacteria were pre-labelled with 13C and 15N according to de Goeij et al.42. In short, prefiltered seawater containing natural bacterial communities was concentrated and added to M63 medium71. As C source, 1 g L−1 13C-glucose (glucose D U-13C6 99%, Cambridge isotopes CLM-1396, Eurisotop) was added and (NH4)2SO4 in the original recipe was replaced by 1.2 g L−1 15N- NH4Cl as N source (99% 15N, Cambridge isotopes NLM-467-5, Eurisotop). Labelled bacteria were concentrated and resuspended in 0.2 µm filtered seawater before dividing in aliquots and storing at 4 °C.

Sponge incubations with 13C- and 15N-labelled food sources

All sponges were allowed to acclimatize for a minimum of 1 week prior to the incubation experiments72. All individuals appeared healthy during their time in the aquaria, and throughout the experiments. All oscula were open and active pumping was confirmed using fluorescent dye. None of the used sponges showed signs of tissue necrosis and no mortality occurred through the experiments. Individual sponges were enclosed in acid-washed (0.4 mol L−1 HCl) flow chambers with magnetic stirring devices43. During the experiments, chambers were kept in the dark and in a water bath to maintain a constant seawater temperature during the incubations (ranging from 6 to 9 °C depending on the incubation). Chambers were closed without trapping air in the system. During incubations, oxygen was continuously measured with an OXY-4 mini oxygen sensor (PreSens). Oxygen profiles of the sponge incubations are depicted in Supplementary Fig. S2 online.

Incubation time was based on the consumption of oxygen within the incubation chamber and the biomass of the sponge (Supplementary Tables S1, S2, and Supplementary Fig. S2, online). Different individuals were used for each food source, and each sponge received multiple pulses of labelled food sources to ensure detectable enrichment in host and symbionts. Each V. pourtalesii specimen (n = 3 for DOM, n = 3 for bacteria) was incubated for 2 × 24 h. H. paupertas (n = 3 for DOM, n = 3 for bacteria) and G. barretti (n = 3 for DOM, n = 4 for bacteria) individuals were incubated for 3 × 8 h. Labelled substrates were added with sterile syringes. DOM was added to a final concentration of 80 µmol L−1 dissolved organic carbon (DOC) for all three tested sponge species (Supplementary Table S2 online). Bacteria were added to a final concentration of 1 × 106 labelled bacteria mL−1 for V. pourtalesii (approximately 16 µmol L−1 bacterial carbon (BC)) and 0.5 × 106 labelled bacteria mL−1 for G. barretti and H. paupertas (approximately 12 µmol L−1 BC). In between subsequent incubations, water was replaced with non-labelled fresh seawater and a new pulse of tracer substrate (in aforementioned concentrations) was added. Seawater incubations without sponges (n = 3 for DOM, n = 3 for bacteria for incubations in 2017 and 2018, respectively) were performed accordingly to serve as controls.

Assimilation and respiration of labelled food sources

After the incubations, all sponges were thoroughly rinsed with 0.2 µm filtered seawater to ensure no labelled residue adhered to the outside of the sponge, and dipped in MilliQ to remove salts. Specimens were then dried (48 h at 60 °C) and dry weight (DW) determined. Then, sponge tissue was homogenized with mortar and pestle and stored in a desiccator until further analysis. Samples for organic carbon content analysis were decalcified with 4 mol L−1 HCl to remove inorganic carbon and lyophilized (24 h). Approximately 10 mg per sample (in silver capsules) was analysed on an Elemental Analyser (Elementar Isotope cube) coupled to an isotope ratio mass spectrometer (BioVision) for simultaneous measurement of organic C and total N content as well as 13C:12C and 15N:14N ratios.

To quantify respiration of the labelled food sources, duplicate water samples for dissolved inorganic 13carbon (DI13C) were taken with acid-washed polycarbonate syringes, directly after adding the labelled substrate (t0) and at the end of each incubation (tend) (δ13DIC). Samples were transferred through a 0.2 µm polycarbonate syringe filter into 3 mL exetainers without trapping air in the vial, and poisoned with 5 µL supersaturated HgCl. Samples were stored at 4 °C until further analysis. 20 mL air-tight glass vials were filled with 3 drops of 90% H3PO4 and flushed with helium for 5 min, after which 0.7 mL of sample was added. Na2CO3 (2 and 10 mmol L−1) and Li2CO3 (2 mmol L−1) served as reference standards for concentration and isotope signature. Samples were analysed by gas chromatograph-isotope ratio mass spectrometry (GC-IRMS) on a Thermo GasBench-II coupled to a Delta-V advantage. As references, 13C-Na2CO3 and Li2CO3 (LSVEC) were calibrated versus NBS-19 and NBS-18 on a Kiel-MAT253. Standard methods according to manual of the manufacturer were used. The standard deviation of the measurements was < 0.05 ‰.

To assess the assimilation and respiration of tracer food sources, 13C:12C and 15N:14N ratios, are expressed in standard delta notation as:

$$\delta^{13} C {\text{or }}\delta^{15} N \left( \permil \right) = \left( { \frac{{R_{sample} }}{{R_{ref - 1} }} } \right) \times 1000$$

where R is the ratio of 13C:12C or 15N:14N in the sample (Rsample; i.e. sponge tissue) relative to reference material: Vienna Pee Dee Belemnite for C (Rref = 0.01118) and atmospheric nitrogen for N (Rref = 0.00368). Total bulk enrichment was calculated as the excess fractional abundance of 13C or 15N (E) in the samples (Fsample) compared with their background [i.e. the natural content of 13C or 15N in the samples before adding tracer (Fbackground)]:

$$E_{sample} = F_{sample} - F_{background}$$

where F is the fractional abundance of heavy isotope (13C or 15N) calculated as:

$$F_{sample\;or\;background} = \frac{{\delta^{13} C }}{{\delta^{13} C + \delta^{12} C }} \;or\; \frac{{\delta^{15} N }}{{\delta^{15} N + \delta^{14} N }}$$

and

$$F_{sample\;or\;background} = \left( {\frac{{\delta^{13} C \;{\text{or}}\;\delta ^{15} N }}{1000 + 1}} \right) \times R_{ref}$$

Total assimilation and respiration rates for C and assimilation rates for N were calculated by multiplying the excess fractional abundance (Esample) by the total Corg or Norg content (μmol) of the tissue.

Phospholipid-derived fatty acid extraction and data analysis

Phospholipid-derived fatty acids (PLFAs) were analysed for bacteria-fed (G. barretti: n = 4, V. pourtalesii: n = 3, and H. paupertas: n = 3) as well as DOM-fed sponge specimens (G. barretti: n = 3, V. pourtalesii: n = 1 (unfortunately, there was only sufficient tissue to perform PLFA analysis for one individual), and H. paupertas: n = 3), and compared to non-labelled backgrounds of the respective sponges (V. pourtalesii: n = 2, G. barretti: n = 3, H. paupertas: n = 2). PLFAs were extracted according to a modified Bligh and Dyer Extraction protocol73,74 (see Ref.75 and Supplementary Materials and Methods online for detailed protocol).

Concentration and isotopic composition of individual PLFAs was determined with a gas chromatograph-combustion-interface isotope ratio mass spectrometer (GC-c-IRMS) consisting of a HP G1530 GC (Hewlett-Packard) connected to a delta-plus IRMS via a type-III combustion interface (Thermo Finnigan, Bremen) on an analytical non-polar column (CP-sil 5, 25 m × 0.32 mm × 0.12 µm). Fatty acid identification was carried out with GC mass spectrometry (MS) (Finnigan Trace GC) using the same column and settings as for IRMS. Results were compared with pre-existing datasets from Utrecht University based on equivalent chain length (ECL).

Data-analysis was performed using the R-package “Rlims”76. Concentration of PLFA’s within the sample were calculated based on the peak areas of the respective PLFA (APLFA), the peak area of the standard C19:0 (A19:0), and carbon amount of the standard C19:0 (C19:0):

$$C_{PLFA} = \frac{{A_{PLFA} \div A_{19:0} \times C_{19:0} }}{gs \times f} \times \frac{n}{n + 1}$$

where gs is the total amount of the sample (mg), f is the fraction of DCM recovered during the extraction, and n is the number of C-atoms in the PLFA. The last factor is to correct for the methyl group that was added during the analytic procedure.

Stable isotope ratios (13C) of individual fatty acids were calculated from FAME data by correcting for the carbon atom in the methyl group that was added during derivatization. Carbon isotope ratios were calculated relative to Vienna Pee Dee Belemnite (VPDB) and tracer carbon incorporation was quantified through determining above background 13C content in the extracted phospholipids.

$$\delta^{13} C_{PLFA} = \frac{{\left( {n + 1} \right)\delta^{13} C_{PLFA} - \delta^{13} C_{methanol} }}{n }$$

where n is the number of carbon atoms in a fatty acid. PLFAs are described as CX:YωZ, where X is the number of C-atoms in the PLFA, Y is the number of double bonds, and Z is the position of the first double bond counted from the methyl (ω) end of the molecule. Prefix “Me” indicates mid-methyl branching, prefix “i” (iso) and “ai” (anteiso) indicate a methyl group one or two carbon away from the methyl end, respectively. Prefix “Cy” indicates presence of a cyclo-ring.

Statistical analysis

All statistical analyses were conducted in PRIMER-E version 677 with the PERMANOVA + add-on77. PERMANOVAs with Monte Carlo tests were performed as this method is robust for small sample sizes and when assumptions of normality and homogeneity are not met78. Individual two-factor PERMANOVAs based on Euclidian distance, with type 3 (partial) sum of squares and unrestricted permutation of raw data (9999 permutations) were used to test for differences in total processing rates of DOM- and bacterial-derived C and N between and within species and assimilation-to-respiration efficiencies for all species per carbon source. Post hoc pairwise comparisons were carried out when species or food source were identified as a significant factor (Tables 1 and 2). To compare pooled average assimilation C:N ratios between DOM and bacteria, a Welch’s t-test was performed. Normality was visually checked with a Q–Q plot. To test for differences between species in the contribution of bacteria-specific and sponge-specific PLFAs to the total PLFA profile of each species, A one factor PERMANOVA based on Euclidian distance, with type 3 (partial) sum of squares and unrestricted permutation of raw data (9999 permutations) was used (Supplementary Table S4 online).