Recycling pathways in cold-water coral reefs: Use of dissolved organic matter and bacteria by key suspension feeding taxa

Cold-water coral (CWC) reefs are one of the most diverse and productive ecosystems in the deep sea. Especially in periods of seasonally-reduced phytodetritus food supply, their high productivity may depend on the recycling of resources produced on the reef, such as dissolved organic matter (DOM) and bacteria. Here, we demonstrate that abundant suspension feeders Geodia barretti (high-microbial-abundance sponge), Mycale lingua (low-microbial-abundance sponge) and Acesta excavata (bivalve) are able to utilize 13C-enriched (diatom-derived) DOM and bacteria for tissue growth and respiration. While DOM was an important potential resource for all taxa, utilization of bacteria was higher for the sponges as compared to the bivalve, indicating a particle-size differentiation among the investigated suspension feeders. Interestingly, all taxa released 13C-enriched particulate organic carbon, which in turn may feed the detritus pathway on the reef. Especially A. excavata produced abundant (pseudo-)fecal droppings. A second stable-isotope tracer experiment revealed that detritivorous ophiuroids utilized these droppings. The high resource flexibility of dominant reef suspension feeders, and the efficient recycling of their waste products by the detritivore community, may provide important pathways to maintain the high productivity on cold-water coral reefs, especially in periods of low external food supply.

. Potential recycling pathways on cold-water coral reefs, and experimental investigation. (a) Suspension feeders on the reef may exploit enhanced concentrations of dissolved organic matter (DOM) and bacteria, transfer it to tissue biomass, and utilize it for respiration and the production of particulate waste such as sponge detritus or bivalve (pseudo-)feces. Particulate waste material may be recycled by reef detritivores. (b) The indicated hypothesized 'recycling' pathways (green and blue arrows) were verified in two laboratory stable isotope 13 C-tracer experiments, using the artificially 13 C-enriched substrates DOM (diatom-derived), bacteria, and A. excavata (pseudo-)feces.
Here, we qualitatively evaluate the potential retention and subsequent recycling of DOM, bacteria and bivalve (pseudo-)feces within CWC reef communities (Fig. 1b), using a two-step experimental approach (Fig. 2). In stable isotope tracer experiment 1 (Fig. 2a), we studied the utilization of two substrates, 13 C-enriched DOM (diatom-derived for logistical constraints, as explained below) and 13 C-enriched bacteria by abundant CWC reef suspension feeders. Next to the HMA sponge G. barretti, the LMA sponge M. lingua and the bivalve A. excavata were chosen for this study, to (1) test the original hypothesis that HMA sponges, with their high abundance of microbes, are better-suited for DOM acquisition compared to other suspension feeders 21,23,24 , and (2) investigate whether bivalves, like sponges, can utilize bacteria as resource. The three taxa were fed in the laboratory with each substrate. Subsequently, the specimens were closed-cell incubated in filtered deep water without added DOM/ bacteria, to measure (a) their utilization of substrate-derived tracer-C for respiration and detritus/(pseudo-)feces production (see tracer-C fluxes, i.e. 13 C in Fig. 1b), and (b) their total respiration, and their total production of detritus/(pseudo-)feces and DOC waste (total-C fluxes, i.e. 13 C + 12 C). For (a), we traced 13 C in the dissolved inorganic carbon ( 13 C-DIC) and particulate organic carbon ( 13 C-POC) released by the animals (tracer-C respiration, tracer POC release). For (b), we measured oxygen consumption, and total POC and DOC release. Finally, we traced 13 C in the animal tissue (tracer-C incorporation). During the preparatory and experimental phase, we observed a particularly high production of (pseudo-)fecal droppings by A. excavata, which we considered  13 C-enriched bacteria ('Bac'). Subsequent closed-cell incubation, to measure O 2 fluxes, total-C and tracer-C fluxes from concentration changes in O 2 (sensor) and between start and end water samples for DIC (dissolved inorganic carbon), POC (particulate organic carbon), and DOC (dissolved organic carbon). (b) Feeding of A. excavata with 13 C-enriched diatoms, collection of bivalve (pseudo-)fecal droppings and feeding of those to reef ophiuroids.

Results
Utilization of the substrates DoM and bacteria. All three investigated taxa utilized the substrates DOM and bacteria. Geodia barretti and A. excavata evidently incorporated DOM-and bacteria-tracer-C into their tissue (Fig. 3a). Tissue samples of M. lingua were unfortunately lost, so incorporation could not be evaluated for this species. However, all investigated taxa, G. barretti, A. excavata and M. lingua, respired DOM-and bacteria-tracer-C, as shown by their production of 13 C-DIC during the closed-cell incubations (Fig. 3b).
Geodia barretti and A. excavata incorporated DOM-tracer-C at a similar rate (Fig. 3a), while G. barretti showed a higher average incorporation of bacteria-tracer-C than the bivalve, but this difference was not significant (Wilcoxon, p > 0.05, Supplementary Table S1). Results, however, indicate that G. barretti incorporated ~100% of the bacteria-tracer-C during the 6-h feeding period. Its bacteria-tracer-C incorporation rate is therefore likely an underestimate due to substrate depletion.
Both species incorporated DOM-tracer-C at a higher rate than bacteria-tracer-C (Wilcoxon, p < 0.05 only for A. excavata). It is, however, important to note that the DOM concentration in our experiment was seven times higher than the bacteria concentration, consistent with the concentration difference on the reefs in deep Norwegian fjords 24 .
Mycale lingua showed the highest tracer-C respiration rate for both substrates as compared to the other two taxa ( Fig. 3b; Kruskall-Wallis, p < 0.05 only for bacteria). Geodia barretti and A. excavata showed lower, similar respiration rates of DOM-tracer-C and bacteria-tracer-C. DOM-tracer-C was respired at a higher rate than bacteria-tracer-C (Wilcoxon, p < 0.05 only for A. excavata). production of particulate and dissolved organic matter. During the closed-cell incubation in filtered seawater following the substrate exposure period (i.e. without added DOM/bacteria), all experimental animals (besides previously bacteria-fed M. lingua) released particulate organic carbon (Figs. 4a, 5, total POC release). The released POC contained tracer-C from the previously consumed substrates DOM and bacteria (Figs. 4b, 5, tracer POC release), demonstrating that all investigated taxa had transformed parts of the substrate-C into detrital waste. Acesta excavata was characterized by a high release of total-and tracer POC (Fig. 5), which (largely) consisted of (pseudo-)fecal droppings (ca. 15 droppings bivalve −1 d −1 ). The droppings had a low buoyancy and visibly accumulated on the floor of the incubation chambers. Geodia barretti, however, had a low total-and tracer POC release (Figs. 4,5), indicating a comparatively low production of particulate detritus (often termed 'sponge detritus' in sponge loop studies 34,36 ). Total and tracer POC release by M. lingua was highly variable. DOM-fed M. lingua showed a high and variable total and tracer POC release. Bacteria-fed M. lingua in three out of four replicates took up total POC from the incubation water (Fig. 4a, background POC in the mix of filtered and unfiltered www.nature.com/scientificreports www.nature.com/scientificreports/ reef water). Tracer-POC, i.e. POC with C originating from the bacteria-or DOM-substrate, by contrary, was released (Fig. 4b).
During the incubations in filtered deep-sea water (DOC concentration of 73.5 μM), A. excavata and M. lingua showed a net release of DOC (Figs. 4c, 5), while G. barretti showed a detectable DOC uptake.
(pseudo-)feces transfer to ophiuroids. 13 C from the artificially-enriched diatom substrate (Skeletonema marinoi) could be traced in all parts of the experimental food chain (Fig. 6a), i.e. A. excavata tissue after consumption of the diatoms, the collected bivalve (pseudo-)fecal droppings, and the tissue of the CWC reef ophiuroids after exposure to (only) the A. excavata (pseudo-)fecal droppings. The four ophiuroids incorporated 37% of the (pseudo-)fecal 13 C produced by two bivalves during the four experimental days (Fig. 6b).  . Total-C fluxes (in black) include the animals' total-C respiration, total POC and total DOC release, and together add up to their total-C turnover. Negative values for DOC/POC release indicate a net uptake. Tracer-C fluxes (in red/blue) refer to the C which the animals utilized from the substrates DOM/bacteria, for tissue incorporation ('Tracer-C incorp') i.e. tissue growth, tracer-C respiration, and tracer POC release.

Discussion
Our study provides direct evidence that CWC reef sponges and bivalves are capable to retain and recycle DOM and bacteria, for tissue growth, metabolism, and detritus production. We will estimate the nutritional importance of the substrates for the suspension feeders. Further, we find that the bivalve A. excavata, more than the investigated sponges, acts as important detritus producer on CWC reefs, and demonstrate the utilization of its (pseudo-) fecal waste by detritivorous ophiuroids. DoM utilization. Sponges, in particular high-microbial-abundance (HMA) sponges, are considered as the dominant DOM consumers on both warm-and cold-water coral reefs 36,44 . Their close association with heterotrophic bacteria, which absorb dissolved substances more efficiently than invertebrates 45 , supposedly gives the HMA sponges an advantage in DOM utilization 21,23,46 . In the present study, the DOM incorporation rate of HMA sponge G. barretti is in the lower range of other cold-and warm-water sponges 36,47,48 . Surprisingly however, the CWC reef bivalve A. excavata and the low-microbial-abundance (LMA) sponge M. lingua show comparable or even higher rates of DOM-incorporation and/or metabolization (Figs. 3 and 5). The ability of marine invertebrates other than HMA sponges to utilize DOM as a substrate could indicate a direct DOM uptake in the animal cells, e.g. via specific membrane transporters for monosaccharides, amino acids or fatty acids, or via pinocytosis, i.e. the ingestion of liquid by membrane vesicle budding 19,49 . Several studies have provided evidence for DOM-C incorporation in both sponge and bacteria cells 36,47,50 . Nevertheless, CWC reef LMA sponges including M. lingua have a complex microbiome comparable to that of HMA sponges 51 , and the gills of A. excavata are inhabited by heterotrophic bacteria of the order Oceanospirillales 52 , which could likewise facilitate the indirect DOM uptake by the animal-microorganism holobiont.
In this experimental approach, we provided the animals with elevated concentrations of comparatively labile, artificial (diatom-derived) DOM. This DOM source may be taken-up at higher rates than natural DOM 50,53 . At the same time however, CWCs and other reef fauna actively release DOM 9,11 , and thereby increase the concentration of labile DOM on the reef 10 above the typically low deep-sea DOC concentrations (<50 μM 8 ). In a recirculating setting, Rix et al. 36 demonstrated the utilization of CWC-derived DOM by the encrusting sponge Hymedesmia coriacea 36 . Unfortunately, a similar approach with CWC-derived DOM was logistically not feasible here, and the required quantities of 13 C-enriched CWC mucus can realistically not be produced in a laboratory setting.
A recent study on Red Sea sponges has suggested a threshold concentration of 79 μM DOC, below which sponges cannot access DOC 54 . In our study however, G. barretti showed a net DOC uptake at a natural deep-fjord DOC concentration (73.5 μM DOC, Fig. 4). DOC uptake below the suggested threshold demonstrates that G. barretti is very efficient in exploiting this resource. Mycale lingua and A. excavata, by contrary, shift to a net DOC release at this low DOC concentration. Hence, these taxa likely profit only from enhanced DOM availability in specific reef microhabitats, e.g. in close vicinity to mucus-producing CWCs 9,36 , where specifically M. lingua frequently occurs 15 .
Bacteria utilization. The CWC reef sponges incorporate and metabolize bacteria at a higher rate than the bivalve (Fig. 3). This is expected as sponges are known to retain bacteria with a near 100% efficiency 24,55,56 , while bivalves tend to target larger particles such as phytoplankton or phytodetritus (>4-7 μm 26 ). Acesta excavata has an exceptionally high clearance rate for this larger particle size spectrum 57 . Specialization of suspension feeders on a certain particle size originates from the morphology and function of the filtration structure. The choanocyte-collar filter of sponges, a 0.06-2 μm-sized gasket-and mucus-sealed mesh 24,55,56 , efficiently retains bacteria-sized particles, while larger particles are phagocytosed by surface and canal-lining pinacocyte cells 19 . Bivalves, by contrary, do not have a 'stiff ' mechanical filter, but a 'paddling' mesh of latero-frontal cilia 58 , which most efficiently traps larger particles, but allows the 'by-catch' of smaller bacteria, without being specialized on them.
nutritional importance of DoM and bacteria. The high fraction of DOM-and bacteria-tracer-C used for tissue growth (tracer-C incorporation), as compared to tracer-C respiration and tracer-POC release (Fig. 5), suggests that both substrates are of good nutritional quality for G. barretti and A. excavata. The present study, however, only provides a snapshot of the tracer-C utilization, as the partitioning of resources may change over time. Maier et al. 59 measured tracer-C utilization of cold-water coral Lophelia pertusa at multiple time points, and demonstrated that, like G. barretti and A. excavata, the corals initially retained the majority of the acquired tracer-C in their tissue, from where it was utilized along with previously stored C.
All taxa utilize DOM-tracer-C at higher rates than bacteria-tracer-C (Fig. 5), which likely relates to the higher DOM concentration in the experiment. In the deep Norwegian fjords, the concentration of DOM-C is likewise higher than the concentration of bacteria-C 24 . Given the different substrate concentrations, it is difficult to identify substrate preferences for each of the taxa. Nevertheless, comparatively high bacteria-tracer-C utilization by G. barretti (100% incorporation) and M. lingua (high tracer-C respiration), in spite of the lower concentration, indicates that the sponges efficiently prey on bacteria. Correspondingly, a study by Leys et al. 24 indicates that sponges are optimally-suited to exploit bacteria, but meet a large fraction of their C demand by naturally more abundant DOM. Acesta excavata, by contrary, incorporated bacteria-tracer-C at a lower rate compared to G. barretti, but DOM-tracer-C at a comparable rate. This suggests that the bivalve prefers other particulate substrates, such as phytodetritus, but can likewise meet its remaining nutritional demand by DOM consumption. The reliance of all species on DOM and bacteria remains to be tested under in situ conditions. fuelling of the detritus food chain. Sponges are considered particularly important for the recycling capacity of warm-and cold-water coral reefs, due to their efficient (partial) transformation of DOM to sponge detritus, which fuels the detritivore food web [34][35][36] . The present results indicate that this so-called sponge-loop could be mediated by other suspension feeders which are abundant in CWC reef communities. All investigated sponges and bivalves partially transformed DOM (and bacteria) to organic particulates (>0.7 μm, tracer POC release, Fig. 4), and could hence mediate a detritivore recycling loop.
Nevertheless, compared to encrusting sponges, which release up to 40% of the assimilated DOM-derived tracer-C as sponge detritus 34,36 , the tracer POC release of G. barretti, M. lingua and A. excavata represents a small sink (<3%) of the utilized tracer-C (Fig. 5). The comparatively short incubation time may partly explain this low conversion. However, G. barretti also has a comparatively low total detritus production, and releases only 0.03% of its tissue-C d −1 as POC (Fig. 5). The low detritus production of G. barretti, accompanied by high investment of retained resources in tissue growth, matches the recently reported difference between emergent and encrusting sponges 60 . While emergent species like G. barretti can allocate a majority of assimilated C in three-dimensional tissue growth, their encrusting relatives are restricted to space-limited, two-dimensional growth, and may therefore invest the C over-supply in high cell turnover, cell shedding and detritus production 60 . The sponge loop on warm-and CWC reefs may thus be supported mostly by encrusting sponges, and hence be spatially confined to dead-coral-framework cavities 34,61 .
M. lingua shows a higher POC release (up to 2.9% tissue-C d −1 ), but the high variability indicates that this could partly be a measurement bias, as this very fragile species may be prone to tissue loss even when handled very carefully.
Acesta excavata could alternatively support the detritivore food chain, with a substantial release of POC as (pseudo-)fecal droppings. The bivalves occur in dense clusters of up to 23 individuals m −2 on the reefs (T. Kutti, unpublished data), and produce 2.6 to 6.3 μmol (pseudo-)fecal POC ind −1 h −1 , depending on the substrate type and food concentration (Supplementary Data S2). The estimated (pseudo-)fecal POC release of 60 to 144 μmol C m −2 h −1 is comparable to the particulate mucus release by CWC L. pertusa 9 (117 μmol C m −2 h −1 ). Due to their low buoyancy, the (pseudo-)fecal droppings sink fast (personal observation), and may accumulate below the A. excavata clusters on vertical and overhanging parts of CWC reefs. We show that the bivalve droppings are readily consumed by reef detritivorous ophiuroids. In this bivalve-driven recycling loop, the ophiuroids recycle 37% of the 4 d-(pseudo-)fecal waste of two bivalves for their own 4 d-tissue growth (Fig. 6). Their tissue incorporation (or assimilation) of 37% of the consumed detritus is in the range of other invertebrates detritivores [62][63][64] , and retains a significant amount of waste material in the live reef community. The quantitative importance of coral-derived DOM and bacteria to support this bivalve-driven recycling loop remains to be investigated, but we argue that C recycling appears to be an ubiquitous feature of main CWC reef components.
Recycling pathways on cWc reefs. As ecosystem engineers, cold-water corals do not only alter their physical environment by creating a three-dimensional reef framework 62 , but also their biogeochemical environment. Their release of mucoid DOM elevates the DOM concentration in the reef water, and stimulates bacterial growth 9,10 . Our experimental study demonstrates that abundant reef suspension feeders, including sponges and bivalves, are able to retain concentrated, labile (diatom-derived) DOM and bacteria, and recycle it to biomass. Hereby, elevated DOM concentrations seem to profit all taxa, while the HMA sponge G. barretti can even access natural DOM at low, ambient deep fjord concentrations. Further, sponges preferably exploit bacteria, while A. excavata shows lower utilization of this resource. The bivalves likely prefer larger phytodetrital particles, indicating a niche separation between the suspension feeders based on particle size. We further show that particulate organic waste of the suspension feeders, specifically the substantial amount of (pseudo-)feces released by A. excavata, is consumed (recycled) by detritivorous reef ophiuroids. Efficient resource exploitation, and the close link between the suspension-feeding and the detritivore food web, may provide mechanisms for those deep-sea ecosystems to maintain high productivity, especially in the long periods of low phytodetritus availability between the settling of the spring plankton blooms 5,7 . The animals were transported in ambient water in cooling boxes to the nearby aquarium facilities of the IMR Austevoll field station. They were maintained for two months in a 1080 L-tank with a flow-through of unfiltered deep fjord water. The deep fjord water was pumped from 165 m depth from an adjacent fjord arm, which is known to harbour both species (renewal rate: 1967 ± 58 L h −1 ; mean ± SD; temperature: 8.2 ± 0.2 °C, salinity: 35.1 ± 0.1‰, pH: 8.0 ± 0.1). For G. barretti, small whole individuals of 4-5 cm in diameter were used in the experiment. This is in contrast to the explant approach, which has commonly been used to study this species [65][66][67] . The advantage of non-explant, natural G. barretti specimens is their presence of oscula, an intact aquiferous system, and hence a natural pumping activity 24 which was confirmed by fluorescein dye. Ophiuroids of the genus Ophiacantha (at least partly detritivorous 68 ) were picked just prior to the experiment from pieces of coral framework (collection as A. excavata). Mycale lingua is a very fragile sponge, which cannot be kept in aquaria for long periods (personal observation). The experimental work on this species was therefore conducted onboard during RV GO SARS cruise 2016110 to Hola reef (Norway, 68°54′N, 14°23′E, 260 m depth) in July 2016. Mycale lingua was collected by ROV (AEgir 6000, NORMAR) and maintained onboard in a 1000 L-tank filled with in situ reef water, in a temperature-controlled room (7.5 °C). Water circulation was maintained with submersible pumps. Half the water was exchanged every 1-2 days. Only actively pumping sponges with open oscula were used in the experiment.

Materials and Methods
preparation of labelled substrates. Diatoms (Skeletonema marinoi, culture collection of the Royal Netherlands Institute for Sea Research, NIOZ) were cultured axenically on F/2-culture medium in 6 batches www.nature.com/scientificreports www.nature.com/scientificreports/ of twelve 1 L-flasks, with 2 mM NaHCO 3 (99 atom% 13 C), under a 12 h light −12 h dark cycle. Diatoms were harvested after 3 weeks 59 . Diatom cells were collected on a 0.45 μm-cellulose acetate filter, flushed into centrifuge tubes with artificial seawater, and concentrated by centrifugation. The diatom pellet was rinsed with ca. 1 L artificial seawater to remove residual medium, centrifuged and lyophilized. One part of the diatoms was used for DOM extraction for experiment 1, another was used in experiment 2. DOM was extracted in two batches from 2 g dry diatoms. Diatom cells were therefore lysed in ultrapure water. Cellular particulates were removed by centrifugation (4000 rpm). The supernatant DOM solution was filtered over 0.22 μmsterile filters, and lyophilized. Mixed bacteria cultures were obtained in two batches from sediment (Oosterschelde mudflats, Netherlands), inoculated in 0.6 L unfiltered, aerated seawater with 0.8 M glucose, 0.8 M ammonium chloride, and yeast extract (17 °C, dark). 3 d-old culture medium was inoculated to new medium (8.3 M glucose, 1.875 M ammonium chloride, yeast extract), which after 3 d was transferred to the final M63 medium with 13 C-glucose (U-13 C 6 , 99 atom%) as C source. The bacteria were harvested after 3 d, by two-step centrifugation (2000 rpm to remove large cells/ aggregates; 8500 rpm to obtain the small 1 μm-diameter cells). Individual bacteria pellets were rinsed with filtered seawater (FSW, 0.7 μm, 50 mL) to remove residual medium, and suspended in 1.5 mL FSW. All substrates were stored at −20 °C. Subsamples (diatoms: 1.5 mg, n = 3; DOM: 1 mg, n = 3; bacteria: 100 μL dried suspension, n = 2) were analyzed for C content and δ 13 C on an elemental analyser coupled to an isotope ratio mass spectrometer (EA-IRMS, Flash 1112, DELTA-V, THERMO, double resistors for measurement of highly 13 C-enriched samples). L-glutamic acid (USGS40, USGS41), 13 C-enriched glucose and bicarbonate were used as reference materials. The fractional 13 C abundance of the substrates was: F 13 diatoms : 29.2%, F 13 DOM : 24.4 to 25.5%, F 13 bacteria : 94.7 to 96.5%). The EA-IRMS was thoroughly cleaned between analysis of highly 13 C-enriched substrates and other samples (see below).

Experiment 1: DOM and bacteria utilization
feeding. Geodia barretti, A. excavata and M. lingua specimens were fed separately with either DOM (238 to 240 μM C) or bacteria (34 to 35 μM C). Substrate concentrations were chosen high enough to ensure detectable isotope enrichment in metabolic products and tissue, but low enough to still resemble CWC reef concentrations 10 . The experimental DOM-C concentration was seven times higher than the bacteria-C concentration, a factor difference which is realistic for CWC reefs in the deep Norwegian fjords 24 . The animals were placed in 4.8 L-plexiglass incubation chambers (Fig. 2a, n = 4 per substrate, except for n = 3 for DOM-fed G. barretti and M. lingua), filled with fresh 0.35 μm-filtered deep fjord water (A. excavata, G. barretti, pumped from deep fjord) or 0.7 μm-filtered deep water (M. lingua, collected above reef with Niskin bottles). A magnetic stirrer in the chamber lid ensured mixing. The DOM and bacteria substrates were dissolved/suspended in 40 mL filtered deep water. Colloids were removed by forcing the bacteria solution through a 0.8 mm-syringe needle, and 0.22 μm-filtration of the DOM solution. The respective substrate was injected through a port in the lid. To maintain a stable temperature, the plexiglass incubation chambers were partially submerged in a tarp-covered (dark) 1080 L-flow-through tank (8.2 ± 0.2 °C); for the on-board experimental work on M. lingua in a 100 L-tank in a dark, 7.5 °C -temperature-controlled room. For each taxon, the feeding time was chosen as long as possible, to increase the chance of successful 13 C-labelling. At the same time, feeding was stopped before the O 2 saturation dropped below 80% to prevent adverse low-oxygen effects (A. excavata 12.5 ± 0.5 h, G. barretti 6.2 ± 0.2 h, M. lingua: 7.4 ± 1.3 h). The oxygen concentration was therefore monitored with a FireSting O 2 logger (TeX4, Pyro Science). After feeding, the water in the feeding chambers was exchanged with filtered deep water to remove the residual 13 C-labelled substrate. Between the feeding and the subsequent closed-cell incubation (see below), all chambers and sampling material were cleaned with 2% HCl. closed-cell incubation. After the feeding (0.5 to 2.5 h), the animals were closed-cell incubated without food ( Fig. 2a), to measure their total respiration, POC and DOC release, and the metabolization of the food substrates, as tracer-C respiration and tracer POC release. The animals were incubated in 1.3 L/4.8 L-plexiglass chambers, depending on their size. For each taxon, the incubation time was chosen as long as possible to detect the targeted C and O 2 fluxes. At the same time, the incubations were stopped before the O 2 saturation dropped below 80% to prevent adverse low-oxygen effects (A. excavata: 11 ± 1.6 h, G. barretti: 5.8 ± 1.2 h, M. lingua: 5.4 ± 1.2 h). Geodia barretti and A. excavata were incubated in fresh 0.35 μm-filtered deep fjord water, M. lingua in a mix of unfiltered and 0.7 μm-filtered deep water (filtration: glass fiber filters, i.e. GFF; unfiltered:filtered = 1:49; water collection see above). The incubation set-up was nearly identical to the feeding set-up, but incubation chambers were closed airtight and free from air bubbles. O 2 consumption (respiration) rates of the incubated animals were derived from the O 2 concentration decrease during the incubations, measured with a continuously logging FireSting probe fitted through the chamber lid. The release (production) of DIC (including 13 C-DIC), POC (including 13 C-POC) and DOC by the animals was measured as the increase in the respective concentration between a start and an end water sample. In the case of G. barretti and A. excavata, the start water samples were taken from an additionally-prepared chamber (no animal), and end water samples from each animal-chamber at the end of the incubation. In the case of M. lingua, the start samples were taken directly from each animal-chamber, and the removed water refilled with a known amount of 0.7 μm-filtered water, in which DIC, POC and DOC concentrations were also measured. DIC and DOC water samples were taken by glass syringe (2%-hydrochloric-acid [HCl]-cleaned). DIC samples were filled in 10 mL-headspace vials, fixed with 10 µL of a saturated mercury chloride solution and stored at 4 °C. DOC samples were filtered over pre-combusted (450 °C, 4 h) GFFs into acid-cleaned, pre-combusted amber vials (40 mL). DOC samples were acidified to pH 2 with concentrated HCl, and stored in the dark at 4 °C. For POC samples, a larger water volume (POC: 2 to 4.2 L, for M. lingua: 0.5 to 1 L) was filtered over pre-combusted, pre-weighed GFFs (per sample: one to three filters, i.e. subsamples), which were dried up to constant weight (40 °C), and stored dark at −20 °C. Two 'no organism'-controls were run in parallel to each taxon-food-combination, to correct the animal O 2 consumption, DIC, POC and DOC release rates. www.nature.com/scientificreports www.nature.com/scientificreports/ Animal sampling. After the incubations, i.e. after a total experimental time of 14 to 25 h, the animal volume was measured via water displacement in a graduated beaker. The tissue of sponges and bivalves (shells removed) was thoroughly rinsed with filtered seawater to remove non-ingested DOM/bacteria. The animal samples were lyophilized, and stored frozen (−20 °C). Additional samples of unfed A. excavata (n = 9) and G. barretti (n = 3) served to measure background isotope values. The M. lingua tissue samples were unfortunately lost, and data cannot be reported.

Experiment 2: Transfer of (pseudo-)feces to ophiuroids
Two bivalves (A. excavata) were placed in one 7-L plexiglass chamber with 0.35 μm-filtered deep water, equipped with a rotating disk in the lid, and fed with 13 C-enriched diatoms (300 μM C, Fig. 2b). After 12 h, the water was exchanged, and the bivalves were placed on a mesh. The bivalves produced distinct (pseudo-)fecal droppings which were collected after 12 h from the chamber floor with a pipette. The ophiuroids (n = 4) were placed in separate 1.2 L-plexiglass chambers with FSW. To each ophiuroid-chamber, three (pseudo-)fecal droppings were added, so that each ophiuroid was supplied with 7.2 ± 0.4 μmol C ophiuroid −1 d −1 . After 22 h, the remaining (pseudo-)fecal droppings were removed and the water was exchanged. This cycle of bivalve-feeding, dropping-collection and ophiuroid-feeding was repeated four times (total C supply: 28.8 ± 1.6 μmol C ophiuroid −1 (4 d) −1 ). For the entire experiment, the respective incubation chambers were partially submerged in the 1080 L-flow-through tank to maintain a stable temperature (8.2 ± 0.2 °C). Ophiuroids, bivalves, and samples of bivalve (pseudo-)fecal droppings, collected on GFFs (n = 4), were lyophilized and stored frozen (−20 °C). Additional samples of unfed ophiuroids (n = 3), unfed bivalves (n = 9), and non-enriched droppings (n = 5) were analysed for background isotope values.

chemical analyses and calculations
Dic, Doc and poc analysis. The DIC-δ 13 C was analysed by DIC-transformation to gaseous CO 2 via addition of phosphoric acid, and CO 2 injection on the EA-IRMS 59 via an additional injection port downstream of the combustion tube. Total DIC concentration was measured on an Apollo SciTech AS-C3. DOC concentration was measured via high-temperature catalytic oxidation on a Shimadzu TOC-VCPN, with certified reference material (Hansell Laboratory). Total POC on the GFFs, and POC-δ 13 C, was analysed on the EA-IRMS described above ('Preparation of labelled substrates').
Total-C and tracer-C fluxes. Total-C fluxes, measured in the closed-cell incubations, include the total-C respiration, estimated from the O 2 consumption, assuming a respiratory quotient of O 2 :C = 1 69 , and the release (i.e. concentration change) of POC and DOC. Tracer-C fluxes, likewise measured in the closed-cell incubations, include tracer-C respiration and tracer POC release. Tracer-C fluxes were derived from the concentration change of 13 C-DIC/ 13 C-POC (calculated from the DIC/POC concentration change and the DIC/POC-δ 13 C), divided by the F 13 of the respective substrate (see 59 for details). Total-C and tracer-C fluxes were standardized to feeding/ incubation time (hours) and tissue organic carbon content (OC, see next paragraph). tissue organic carbon and tracer-c incorporation. Acesta excavata, G. barretti, and ophiuroid samples were homogenized by pestle and mortar. Subsamples (A. excavata: 2 mg, G. barretti: 5 mg, ophiuroids: 11 mg, n = 3 per sample) were analysed for tissue organic carbon content (OC), and δ 13 C on the EA-IRMS. Ophiuroid tissue was decalcified with HCl prior to the analysis 59 . The GFFs with (pseudo-)fecal droppings were analysed as a whole on the EA-IRMS. The amount of 13 C in the animal tissue and bivalve droppings was calculated from the δ 13 C as described in detail in 59 , using Vienna Pee Dee Belemnite as standard with an isotope ratio of R = 0.0111802. Tracer-C incorporation was obtained from the amount of 13 C divided by F 13 substrates , and standardized to the feeding time (hours) and tissue OC. The OC of M. lingua (required for the standardization of C fluxes) was estimated as 0.5 · AFDM 61,70 (ash-free dry mass), their AFDM as AFDM = log(V) * 0.265 (T. Kutti, unpublished data), where V is the sponge volume. As an additional measure, we calculated the percentage of the provided tracer-C (i.e. bacteria-or DOM-tracer C added in the feeding incubations), which the animals incorporated.
Data analysis. Data are reported as mean ± standard deviation. Data analysis was performed in R 71 .
Non-parametric statistical testing was chosen to account for low replicate numbers. Detailed results of statistical tests are available as Supplementary Table S1. A Kruskal-Wallis-test with a Dunn post-hoc test (R package FSA 72 ) was applied to compare tracer-C respiration and tracer POC release between G. barretti, M. lingua and A. excavata, fed with (1) DOM and (2) bacteria. A Wilcoxon rank sum test served to compare (a) rates of tracer-C incorporation between G. barretti and A. excavata, and (b) tracer-C incorporation, tracer-C respiration and tracer POC release of each taxon between the substrates DOM and bacteria.