Differential processing of dissolved and particulate organic matter by deep-sea sponges and their microbial symbionts

Deep-sea sponges create hotspots of biodiversity and biological activity in the otherwise barren deep-sea. However, it remains elusive how sponge hosts and their microbial symbionts acquire and process food in these food-limited environments. Therefore, we traced the processing (i.e. assimilation and respiration) of 13C- and 15N-enriched dissolved organic matter (DOM) and bacteria by three dominant North Atlantic deep-sea sponges: the high microbial abundance (HMA) demosponge Geodia barretti, the low microbial abundance (LMA) demosponge Hymedesmia paupertas, and the LMA hexactinellid Vazella pourtalesii. We also assessed the assimilation of both food sources into sponge- and bacteria-specific phospholipid-derived fatty acid (PLFA) biomarkers. All sponges were capable of assimilating DOM as well as bacteria. However, processing of the two food sources differed considerably between the tested species: the DOM assimilation-to-respiration efficiency was highest for the HMA sponge, yet uptake rates were 4–5 times lower compared to LMA sponges. In contrast, bacteria were assimilated most efficiently and at the highest rate by the hexactinellid compared to the demosponges. Our results indicate that phylogeny and functional traits (e.g., abundance of microbial symbionts, morphology) influence food preferences and diet composition of sponges, which further helps to understand their role as key ecosystem engineers of deep-sea habitats.

www.nature.com/scientificreports/ 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 zooplankton 18,26,27 . Moreover, in recent years it has become increasingly evident that many sponges, including dominant North Atlantic deep-sea species 28,29 , use dissolved organic matter (DOM) as main food source 30 . DOM constitutes the largest pool of reduced carbon in the ocean, and is therefore the largest potential marine organic food source 31 . However, DOM appears to represent only a minor fraction of the food intake of most invertebrates 32 , and is generally considered to be recycled through bacterial processing 33 . Consequently, when various shallow water sponge holobionts were found to feed predominantly-often 70 to > 90% of their daily carbon intake-on DOM [34][35][36] , this capacity was attributed to their microbial symbionts [37][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 DOM 29,35,36,42 and some even at higher uptake rates than HMA sponges 28,43,44 . Furthermore, DOM has been shown to be assimilated by both sponge host and bacterial symbiont cells 42,44 , ultimately visualized at a (sub) cellular level 45,46 . 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 production 47,48 . 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 requirements 28 . 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 13 C-and 15 N-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. Processing of tracer ( 13 C-and 15 N-) DOM (left) and bacteria (right) by deep-sea sponges. Dark red bars represent tracer carbon (C DOM or C bac ) assimilation rates, light red bars represent tracer carbon respiration rates and blue bars represent tracer nitrogen (N DOM or N bac ) assimilation rates in µmol C or N food source mmol C or N −1 sponge d−1. HMA high microbial abundance sponge species, LMA low microbial abundance sponge species.
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 N bac mmol C and N −1 sponge d −1 ), and the two demosponges H. paupertas (0.3 ± 0.3 and 0.7 ± 0.3 µmol C and N bac mmol C and N −1 sponge d −1 ) and G. barretti (0.2 ± 0.1 and   Relative assimilation and respiration of 13 C-tracer DOM (left) and bacteria (right). Dark red bars represent relative tracer carbon (C DOM or C bac ) 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.
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)).

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 evidence 28,30,49 that DOM-processing is not restricted to high microbial (HMA) sponges as is commonly suggested 40,41,50 , 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  www.nature.com/scientificreports/ 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 sponges 40,41,50 . 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 C DOM mmol C sponge d −1 , 0.4-1.7 µmol N DOM mmol N sponge 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 C DOM mmol C sponge d −1 , 2.0 ± 2.0 µmol N DOM mmol N sponge 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 CO 2 fixation, such as Nitrospinae and Chloroflexi [53][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) bacteria 18,42 . Still, DOM potentially constitutes a much larger proportion of their daily diet than bacteria 28 , mainly due to the order of magnitude higher ambient concentration of DOM-derived C and N in seawater 35 . 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 coralliophaga 56 . Considering the high dissolved inorganic nitrogen (DIN) efflux rates found in many sponges from both tropical and deep-sea habitats 54,57,58 , 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 vitamins 59 , 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 sugars 60,61 , which can be rapidly respired. Glucose, for example, was found to be almost exclusively respired by sponges 35,56 . 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 demands 62 .
It is important to mention that both bacteria and DOM used as food sources in this study were laboratorymade. This could affect our results as bacteria grown in culture may have different C-and N-contents compared to bacteria naturally occurring in seawater 63 . 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 refractory 44,64 . 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 colleagues 29,44 . 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 longchained, sponge-specific PLFAs from shorter precursor PLFAs 65 . A recent study by Achlatis et al. 46  www.nature.com/scientificreports/ 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 DOM 45,46 and these cells may be partly lost as detrital waste through shedding 41,45,66 . Cell loss was not monitored here, due to the methodological complexity, yet should be incorporated into future (tracer) metabolic studies on deep-sea sponges.

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 www.nature.com/scientificreports/ 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 13 C and 15 N according to de Goeij et al. 42 . In short, prefiltered seawater containing natural bacterial communities was concentrated and added to M63 medium 71 . As C source, 1 g L −1 13 C-glucose (glucose D U-13C6 99%, Cambridge isotopes CLM-1396, Eurisotop) was added and (NH 4 ) 2 SO 4 in the original recipe was replaced by 1.2 g L −1 15 N-NH 4 Cl as N source (99% 15 N, 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 13 C-and 15 N-labelled food sources. All sponges were allowed to acclimatize for a minimum of 1 week prior to the incubation experiments 72 . 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 devices 43 . 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 × 10 6 labelled bacteria mL −1 for V. pourtalesii (approximately 16 µmol L −1 bacterial carbon (BC)) and 0.5 × 10 6 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 13 C: 12 C and 15 N: 14 N ratios.
To quantify respiration of the labelled food sources, duplicate water samples for dissolved inorganic 13 carbon (DI 13 C) were taken with acid-washed polycarbonate syringes, directly after adding the labelled substrate (t 0 ) and at the end of each incubation (t end ) (δ 13 DIC). 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% H 3 PO 4 and flushed with helium for 5 min, after which 0.7 mL of sample was added. Na 2 CO 3 (2 and 10 mmol L −1 ) and Li 2 CO 3 (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, 13 C-Na 2 CO 3 and Li 2 CO 3 (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, 13 C: 12 C and 15 N: 14 N ratios, are expressed in standard delta notation as: where R is the ratio of 13 C: 12  www.nature.com/scientificreports/ and Total assimilation and respiration rates for C and assimilation rates for N were calculated by multiplying the excess fractional abundance (E sample ) by the total C org or N org content (μmol) of the tissue.
Concentration and isotopic composition of individual PLFAs was determined with a gas chromatographcombustion-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 (A PLFA ), the peak area of the standard C19:0 (A 19:0 ), and carbon amount of the standard C19:0 (C 19:0 ): 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 ( 13 C) 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 13 C content in the extracted phospholipids.
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 6 77 with the PER-MANOVA + add-on 77 . 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 met 78 . 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).

Data availability
All datasets will be made publicly available upon publication through the meta-data record PANGAEA.