Subcellular view of host–microbiome nutrient exchange in sponges: insights into the ecological success of an early metazoan–microbe symbiosis

Background Sponges are increasingly recognised as key ecosystem engineers in many aquatic habitats. They play an important role in nutrient cycling due to their unrivalled capacity for processing both dissolved and particulate organic matter (DOM and POM) and the exceptional metabolic repertoire of their diverse and abundant microbial communities. Functional studies determining the role of host and microbiome in organic nutrient uptake and exchange, however, are limited. Therefore, we coupled pulse-chase isotopic tracer techniques with nanoscale secondary ion mass spectrometry (NanoSIMS) to visualise the uptake and translocation of 13C- and 15N-labelled dissolved and particulate organic food at subcellular level in the high microbial abundance sponge Plakortis angulospiculatus and the low microbial abundance sponge Halisarca caerulea. Results The two sponge species showed significant enrichment of DOM- and POM-derived 13C and 15N into their tissue over time. Microbial symbionts were actively involved in the assimilation of DOM, but host filtering cells (choanocytes) appeared to be the primary site of DOM and POM uptake in both sponge species overall, via pinocytosis and phagocytosis, respectively. Translocation of carbon and nitrogen from choanocytes to microbial symbionts occurred over time, irrespective of microbial abundance, reflecting recycling of host waste products by the microbiome. Conclusions Here, we provide empirical evidence indicating that the prokaryotic communities of a high and a low microbial abundance sponge obtain nutritional benefits from their host-associated lifestyle. The metabolic interaction between the highly efficient filter-feeding host and its microbial symbionts likely provides a competitive advantage to the sponge holobiont in the oligotrophic environments in which they thrive, by retaining and recycling limiting nutrients. Sponges present a unique model to link nutritional symbiotic interactions to holobiont function, and, via cascading effects, ecosystem functioning, in one of the earliest metazoan–microbe symbioses. Video abstract Supplementary Information The online version contains supplementary material available at 10.1186/s40168-020-00984-w.


Background
Sponges are increasingly recognised as key ecosystem engineers in many aquatic habitats, playing important roles in ecological processes, such as habitat provision and nutrient cycling [1,2]. As filter-feeders par excellence [3,4], the ecological success of sponges largely depends upon their ability to capture and transform a suite of organic and inorganic nutrients. Their unique and varied diet is related to the (inter)activity of the sponge host and its abundant and diverse microbial community, collectively termed the sponge holobiont. Sponges are considered one of the oldest extant metazoans, evolving more than 600 mya [5], and sponge-microbe associations are likely amongst the earliest of metazoan-microbe symbioses [6]. These symbioses are widely assumed to be mutualistic, but apart from a few notable exceptions [7][8][9][10], the nature of many of these beneficial interactions (mutualism, commensalism) has not been experimentally validated [6,11].
The translocation or exchange of nutrients is a common feature of beneficial nutritional symbiosis [12]. The prevailing notion is that the sponge microbiome plays an important role in sponge health and nutrition [13] by extending the metabolic repertoire of the host [14,15]. A wealth of -omics-based approaches have highlighted the metabolic potential of sponge symbionts, with putative benefits for the host ranging from chemical defence via the production of secondary metabolites [16,17], supplying nutrition through the provision of fixed carbon or essential vitamins and amino acids [18,19], to the recycling of host waste products [20,21]. However, despite its identification as a priority research area, experimental evidence for many of these putative metabolic interactions is lacking [11]. The best characterised examples of beneficial sponge symbioses are between sponges and photoautotrophic symbionts, including cyanobacteria [22][23][24] and dinoflagellates (family Symbiodiniaceae) [7,25]. Numerous studies have demonstrated the importance of these autotrophic symbionts in contributing to host nutrition, fitness, and growth through the translocation of photosynthetically fixed carbon and inorganic nitrogen [9,10,26,27]. However, not all sponges contain abundant photoautotrophs and many rely predominantly on heterotrophic feeding on organic matter to meet their nutritional requirements. Furthermore, the benefits that microbial symbionts receive from their interaction with the host are less welldefined, but they are generally expected to profit from a nutrient-rich habitat within the sponge body. For example, in the deep-sea sponge Geodia barretti, microbially mediated nitrogen transformations were suggested to be fuelled by metabolic waste products of the host [28], but without direct evidence.
Sponges are opportunistic feeders capable of ingesting a wide spectrum of particulate organic matter (POM), but specialise in capturing cells < 10 μm, such as bacterio-and phytoplankton [13]. In the last few decades, it has become clear that dissolved organic matter (DOM) is also a major component of the diet (between 50 and 97 %) of many sponges, spanning tropical, shallow-water to cold, deep-sea species (reviewed in [2]). DOM represents the largest reservoir of organic carbon in the ocean [29] and is a relatively inaccessible food source for many other multicellular heterotrophic organisms. Furthermore, sponges convert DOM to POM-a more bioavailable food source for many organisms-via the so-called 'sponge loop' [30,31], thereby aiding in the retention and recycling of nutrients in marine benthic ecosystems. Initially, it was assumed that DOM consumption by sponges was solely mediated by microbial symbionts, and therefore directly related to symbiont abundance [13,32]. DOM-feeding was expected to be largely limited to high microbial abundance (HMA) sponges, who host symbiont densities 2-4 orders of magnitude higher than their low microbial abundance (LMA) counterparts [33][34][35]. However, compound-specific (e.g. stable isotope) tracer studies have shown that both sponge cells and associated microbes are involved in organic matter assimilation [36,37], and recent nanoscale secondary ion mass spectrometry (NanoSIMS) studies have confirmed that both host choanocytes (sponge filtering cells) and symbionts can directly assimilate DOM [8,38]. It has been hypothesised that microbially assimilated DOM will be translocated to the host [39], while hostprocessed POM will be recycled by microbial symbionts [13], but these metabolic interactions have not yet been tested.
Here, we investigated the incorporation and processing of key heterotrophic dietary components by sponges at the holobiont and host cell and symbiont levels. We combined pulse-chase isotopic tracer techniques with NanoSIMS in order to trace the incorporation and fate of 13 C-and 15 N-labelled DOM and POM by the HMA sponge Plakortis angulospiculatus and the LMA sponge Halisarca caerulea (Fig. 1) over time. The use of Nano-SIMS analysis allowed us to target the prokaryotic community and host cells of our sponges in hospite with subcellular resolution. The objectives of the study were to (i) examine the role of host cells and symbionts in organic matter incorporation and (ii) test for translocation between host and symbionts over time. This enabled us to disentangle complex interactions between host and microbiome in the uptake and exchange of heterotrophically acquired nutrients.

Sponge collection
This study was conducted at the Caribbean Research and Management of Biodiversity (CARMABI) field station, on the island of Curaçao (12°12′N, 68°56′W) between June and July 2018. The common Caribbean encrusting sponge species Plakortis angulospiculatus (HMA) and Halisarca caerulea (LMA) were collected from the fringing reefs in front of CARMABI and at station 'Buoy 1' (12°07′28.65′′N, 68°58′23.23′′W), located on the leeward side of Curaçao. These species were selected as they are good examples of both ends of the HMA-LMA spectrum (see Fig. 1), are easy to maintain and produce good histological sections. Plakortis is a cosmopolitan genus [40] and P. angulospiculatus is a widespread Caribbean species with available bacterial community data [41,42]. For H. caerulea, an increasing body of information on physiology and ecology [30,43], transcriptomics [44] and the microbial community [45], is available. Sponge individuals were collected by SCUBA diving at water depths of between 12 and 30 m and were removed from rock faces using a dive knife (P. angulospiculatus) or chiselled from dead coral plates and cleared of epibionts (H. caerulea) and cut into~3-4 cm 3 specimens with at least two functioning oscula (i.e. outflow opening; active pumping tested with fluorescein dye). Sponges were directly transferred to the wet-lab facilities of CARMABI and maintained in 100-L flowthrough aquaria, supplied with water pumped from the adjacent reef at 10 m depth (flow rate approx. 3 L/min). Sponges were acclimatised for between 1 and 2 weeks and checked regularly to remove debris.

Pulse-chase experiment with isotopically labelled tracers
To track the incorporation and fate of carbon (C) and nitrogen (N) over time, as important metabolic 'currencies' for DOM and POM nutrition [30,31], a pulsechase experiment was conducted using 13 C-and 15 N-labelled DOM and POM. Preparation of the isotopically labelled substrates is detailed in Additional file 1. Sponge individuals were incubated independently with DOM and POM over a 3-h pulse phase (0-3 h) and then transferred to flow-through aquaria with non-labelled water for the 45-  POM (to determine background enrichment levels). Individual sponges were transferred to air-tight 2-L incubation chambers (see [30] for description of chambers) filled with 0.7-μm GF/F filtered seawater (FSW; 47 mm, Whatman). Chamber lids were equipped with a magnetic stirring device to ensure constant mixing and water flow during the incubations. Isotopically labelled DOM or POM was injected via syringe and the lid placed on each chamber ensuring no headspace. Food sources were added to give a final concentration of~90 μM dissolved organic carbon (DOC) and~1 × 10 6 bacteria cells/mL in the chambers (see Additional file 1). An optical oxygen probe was inserted through an airtight port in the chamber lid and dissolved oxygen concentration (DO) was measured continuously (OXY-4 mini, PreSens) to ensure sufficient oxygenation and to monitor respiration throughout the incubations. Respiration rates were 48 ± 31 (mean ± SD) and 104 ± 59 μmol O 2 /g DW/h for the HMA sponge P. angulospiculatus and LMA sponge H. caerulea, respectively. Chambers were placed in a flowthrough (3 L/min) aquarium to maintain ambient reef temperature and sponges were incubated in the dark during the pulse phase to exclude photosynthesis by photoautotrophs. At each sampling time-point, sponges were removed from the incubation chambers (pulse phase) or flow-through aquaria (chase phase) and rinsed in 0.7-μm FSW followed by Milli-Q water; chase sponges were first rinsed in label-free seawater before transfer from chambers to flow-through aquaria. Sponge planar surface area and thickness were then quantified from scaled photographs using ImageJ [46]. Three pieces of tissue per specimen (technical replicates) were removed using a biopsy punch (P. angulospiculatus, 2 mm ø, PFM Medical UK) or sterile scalpel blade (H. caerulea) and fixed in 2.5 % (v/v) glutaraldehyde + 1 % (w/v) paraformaldehyde in PHEM buffer (1.5× PHEM (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgSO 4 .7H 2 O), and 9 % (w/v) sucrose, pH 7.4) for electron microscopy and NanoSIMS analysis. The remaining tissue was transferred to pre-weighed cryovials using a sterile scalpel blade and stored at -20°C for bulk C and N content and stable isotope analysis (for details see Additional file 1).

Sample preparation for electron microscopy and NanoSIMS
Tissue samples were fixed for 12 h at 4°C, rinsed three times with PHEM buffer (1.5× PHEM and 9 % (w/v) sucrose) and post fixed for 1.5 h with 1 % (w/v) osmium tetroxide in Milli-Q water. Samples were dehydrated in a graded series of ethanol and embedded in EPON Araldite. Embedded tissue was sectioned perpendicular to the surface of the sponge. Ultrathin (120 nm) and semithin (500 nm) sections were cut using a Reichert Ultracut S microtome. Ultrathin sections were transferred to finder grids (Electron Microscopy Sciences, Hatfield, PA, USA), stained with uranyl acetate and lead citrate, and imaged at 100 kV accelerating voltage using a Philips CM10 transmission electron microscope (TEM). These high-resolution images provided an initial characterisation of the tissue structure of both sponges, particularly regarding symbiont density and location (Fig. 1b, e). Semithin sections were transferred to silicon wafers, stained as above, and imaged with a Zeiss Sigma field emission scanning electron microscope (SEM) at 8 kV. Electron microscopy was performed at the Electron Microscopy Centre Amsterdam (EMCA). Regions of interest were identified by SEM and sample maps made to guide NanoSIMS analysis. One replicate per species and food source from incubations at T 0 , T 0.25 , T 0.5 , T 3 , and T 48 was selected for NanoSIMS analysis.

NanoSIMS analysis
To visualise the subcellular fate of incorporated 13 C and 15 N within sponge tissue, areas imaged by SEM were subsequently imaged with a NanoSIMS 50 ion probe (CAMECA, Paris, France) at the Centre for Microscopy, Characterisation and Analysis (University of Western Australia, Perth). Sections were gold coated (10 nm), then bombarded with a 16 keV Cs + primary ion beam to detect the negative secondary ions 12 C 12 C, 13 C 12 C, 12 C 14 N, 12 C 15 N and 31 P (for details, see Additional file 1). Between 5 and 8 different areas were scanned per sample. Images were processed using the OpenMIMS software plugin (National Resource for Imaging Mass S p e c t r o m e t r y , h t t p s : / / g i t h u b . c o m / B W H C N I / OpenMIMS/wiki) for Fiji. Mass images were drift corrected, aligned and stacked, and presented as huesaturation-intensity (HSI) images of the 13 C/ 12 C and 15 N/ 14 N ratios. Enrichment of 13 C and 15 N were quantified for the following regions of interest (ROI): (I) choanocytes, (II) all cells of the mesohyl (including archaeocytes and amoebocytes, but excluding spherulous cells which were present in H. caerulea only), and (III) sponge-associated microbes (Fig. 1). Additionally, areas of the mesohyl devoid of host cells (circles of < 1 μm diameter) and spherulous cells of H. caerulea were also included in the analysis and separately categorised, but will not be discussed in detail herein due to generally low enrichment throughout the pulse-chase experiment. All ROI were manually drawn over the NanoSIMS maps of 12 C 14 N ions, using 31 P and SEM images as a reference to identify cell types. The ROI selected for analysis captured the majority of enrichment in all areas of tissue scanned; only on few occasions were unidentifiable enriched cells excluded from analysis. A total of 14, 979 ROI were analysed across 20 sponge individuals (Additional file 2: Table S1).
Unlabelled control sponge samples were analysed to obtain natural abundance ratios for each ROI category. Extracted isotopic ratios (R sample ) for each ROI were multiplied by a correction factor, CF: where R sample IRMS yeast represents ratios obtained from bulk isotope ratio analysis of Saccharomyces cerevisiae (yeast), and R sample yeast from ratios obtained from daily scans of yeast using NanoSIMS (see Additional file 1). Isotopic ratios were corrected and converted to 13 C and 15 N atom fractions (Atom%), F, which gives the amount of a specific atom (e.g. 13 C) as a percentage of the total number of atoms (e.g., 13 C + 12 C): A summary of 13 C and 15 N atom fractions and the number of ROI analysed are given in Additional file 2: Table S1. Individual ROI were deemed enriched in 13 C and 15 N if extracted Atom% values exceeded three times the standard deviation of mean isotopic values of corresponding ROI categories of unlabelled controls. The proportion of enriched versus nonenriched ROI populations was determined for choanocytes, mesohyl cells, and sponge-associated microbes (Additional file 2: Figure S1).

Statistical analysis
Statistical analysis of the bulk tissue data was performed using SPSS (software V25). The effect of time on incorporation of DOM-or POM-derived 13 C and 15 N during the 3-h pulse phase was quantified using linear regression models for each species (Additional file 2: Table S2), and differences between regression coefficients were compared using univariate ANOVA in SPSS. Data met the assumptions of linearity, normality, homoscedasticity and sample independence.
NanoSIMS was foremost used here as an observational tool to visualise the uptake and transfer of organic 13 C and 15 N over time, and presentation of these results is limited to observed trends, since one sponge individual per time-point was analysed. However, as replication using NanoSIMS occurs at the single-cell level and not holobiont level (e.g. [7,47,48]) we present statistical analysis of the extracted single-cell data in Additional file 2: Tables S3 and S4.

Bulk incorporation of organic matter by sponge holobionts DOM
The sponge species P. angulospiculatus (HMA) and H. caerulea (LMA) showed significant enrichment of DOMderived 13 C and 15 N into their tissue over time (Fig. 2a, b; Additional file 2: Table S2). Stable isotope enrichment increased linearly during the 3-h pulse phase (regression analysis, all p < 0.001; Additional file 2: Table S2) in the sponge holobionts and gradually decreased during the subsequent chase phase when sponges were returned to label-free seawater (Fig. 2a, b). The rate of 13 C incorporation during the pulse phase was significantly higher for the LMA species H. caerulea than for the HMA species P. angulospiculatus (ANOVA, F = 10.54, df = 1, p = 0.003) and the inverse relationship was true for 15 N incorporation (ANOVA, F = 5.5, df = 1, p = 0.027).

POM
Both P. angulospiculatus and H. caerulea showed a linear increase of POM-derived 13 C-and 15 N-enrichment into their tissue during the 3-h pulse phase (regression analysis, all p < 0.001; Additional file 2: Table S2), but the HMA species P. angulospiculatus incorporated POM at a significantly lower rate than the LMA species H. caerulea for both 13 C and 15 N (ANOVA, F = 11.13 and p = 0.003, F = 11.34 and p = 0.002 for 13 C and 15 N, respectively) and enrichment was low across the pulsechase experiment (Fig. 2c, d). Halisarca caerulea incorporated 13 C and 15 N in a pattern similar to that seen with DOM: a rapid increase during the 3-h pulse and gradual decrease during the subsequent chase phase (Fig. 2c, d).

Single-cell analysis of organic matter incorporation
NanoSIMS analysis revealed substantial 13 C-and 15 Nenrichment derived from both isotopically labelled food sources into host cells and sponge-associated microbes of P. angulospiculatus and H. caerulea. Comparison of 13 C/ 12 C and 15 N/ 14 N ratio images with SEM micrographs enabled the patterns of enrichment to be ascribed to specific cells and subcellular structures within the sponge tissue (Fig. 1) and tracked through time.

DOM
Within 15 min, sponge-associated microbes of both sponge species began to assimilate DOM-derived 13 C and 15 N. During this time, 19 and 51 % of the spongeassociated microbes of P. angulospiculatus and H. caerulea became enriched in 15 N, respectively (Additional file 2: Figure S1). Nearly half of choanocyte cells (i.e. the sponge filtering cells; 43 and 51 % for 13 C and 15 N, respectively; Additional file 2: Figure S1) of the HMA species P. angulospiculatus became enriched during the first 15 min, with 'hotspots' of incorporation apparent in the cells (Fig. 3a, e), while incorporation of DOM into choanocytes of the LMA species H. caerulea was low (5 and 4 % of cells enriched for 13 C and 15 N, respectively; Additional file 2: Figure S1). After 30 min, numerous highly localised areas of 13 C and 15 N appeared in the apical tip of choanocyte cells of the LMA species (Fig. 3j,  n), demonstrating direct incorporation of DOM. For both sponges, there was substantial enrichment of 13 C and 15 N into choanocytes and sponge-associated microbes at the end of the 3-h pulse, and choanocytes contained numerous large (1-2 μm) enrichment hotpots (Fig. 3). While some of these hotspots remained at the end of the chase phase, 13 C and 15 N became more homogenously distributed throughout choanocyte cells (Fig. 3).
The average 13 C and 15 N isotopic enrichment of choanocytes decreased during the chase phase in the HMA species P. angulospiculatus, while average enrichment in choanocytes of the LMA species H. caerulea remained the same for 13 C and decreased slightly for 15 N (Fig. 4ad; Additional file 2: Table S1). A concurrent increase in the average isotopic enrichment of sponge-associated microbes occurred during the chase phase for both sponges, indicating translocation of C and N from choanocyte cells to sponge symbionts (Fig. 4a-d). Moreover, the percentage of sponge-associated microbes enriched in 15 N increased from 86 to 95 % between 3 and 48 h in the HMA species, and from 58 to 88 % in the LMA species (Additional file 2: Figure S1). Mesohyl cells of the LMA species H. caerulea incorporated DOM-derived 13 C and 15 N during the 3-h pulse, and average enrichment further increased during the chase phase (Fig. 4b,  d), indicating translocation of C and N from choanocytes to these cells. Enrichment of mesohyl cells in the HMA species P. angulospiculatus generally remained low throughout the pulse-chase (Additional file 2: Table S1).

POM
As reflected in the bulk tissue data, incorporation of POM-derived 13 C and 15 N by the HMA species P. angulospiculatus was low across the pulse-chase (Fig. 5a-h). Sparse incorporation of labelled food bacteria can be seen in choanocytes of this HMA species at multiple time-points, with few individual food bacteria being phagocytosed (red arrows, Fig. 5b, e, f). Incorporated 15 N became homogenously distributed throughout choanocytes at the end of the chase period (Fig. 5h). Average 13 C-and 15 N-enrichment of spongeassociated microbes remained low across the time series (Additional file 2: Table S1). The overall low incorporation rates of POM by P. angulospiculatus and variability between samples precludes reliable interpretation of potential translocation between cell types during the chase period.
A markedly different pattern was observed for the LMA species H. caerulea. Rapid incorporation of POM-  Figure S1). Individual labelled food bacteria were visibly phagocytosed by choanocytes during this time (Fig. 6d-f). Conversely, incorporation of POM-derived 13 C and 15 N into sponge-associated microbes was low during the early pulse incubations, indeed, after 30 min, 13 C-and 15 N-enrichment was detected in only 2 and 32 % of microbes, respectively (Additional file 2: Figure S1). At the end of the 3-h pulse, substantial incorporation of 13 C and 15 N was evident in choanocytes and sponge-associated microbes of the LMA species (Fig. 5k, o; Additional file 2: Table S1).  Table S1 Concentrated hotspots of enrichment were visible in choanocytes, but 15 N had generally become dispersed throughout these cells (Fig. 5o). The presence of 13 C and 15 N in sponge-associated microbes after 3 h suggests translocation of metabolites from choanocyte cells, as symbionts are unable to phagocytose POM (see "Discussion" section). Both 13 C and 15 N became more uniformly distributed throughout the choanocytes during the chase period (Fig. 5l, p).
The average 13 C and 15 N enrichment of choanocytes in the LMA species H. caerulea decreased slightly during the chase period (Additional file 2: Table S1), although this trend was masked in Fig. 4 because enrichment at 3 h was highly variable, and included highly enriched choanocytes not captured in the boxplot distributions (Fig. 4f, h). The average isotopic enrichment of sponge-associated microbes increased during the chase period, and the proportion of enriched microbes also increased between 3 and 48 h, from 30 to 50 % for 13 C, and 83 to 100 % for 15 N (Additional file 2: Figure S1). Mesohyl cells of the LMA sponge became enriched in 13 C and 15 N during the 3-h pulse, and average enrichment further increased during the chase phase (Fig. 4f, h). The increased isotopic enrichment of sponge-associated microbes and cells of the mesohyl during the chase phase indicates translocation of C and N from choanocytes to these cells, since the majority of DOM incorporation was by choanocytes during the 3-h pulse (Fig. 5k, o) and their isotopic enrichment did not increase during the chase period.

Discussion
Here, we show the uptake, processing, and transfer of dissolved and particulate food by sponge host and microbiome in one of the earliest metazoan-microbe symbioses. Consistent with recent findings [8,38], both host cells and microbial symbionts of the HMA sponge P. angulospiculatus and the LMA sponge H. caerulea were actively involved in heterotrophic feeding processes. Contrary to the expectation that microbially assimilated DOM would be translocated to the host, we found experimental evidence for the translocation of both DOM-and POM-derived C and N from host cells to symbionts over time, demonstrating that the microbiome retains important nutrients  Table S1. Note the different y axis scale between carbon and nitrogen values within the holobiont via the recycling of host waste products. Our findings highlight the competitive strategy of the filter-feeding sponge holobiont, not only by tapping into a food resource (i.e. DOM) that is relatively inaccessible to many other marine heterotrophic metazoans, but by recycling potentially limiting nutrients via interactions with its microbial symbionts.
The role of host and microbiome in organic matter uptake DOM uptake DOM was rapidly incorporated by the HMA species P. angulospiculatus and the LMA species H. caerulea, and bulk rates of incorporation into sponge tissue (1.1-1.5 μmol C tracer /mmol C sponge /h and 0.4-0.6 μmol N tracer /  Table S1 mmol N sponge /h; Additional file 2: Table S2) are in the range of rates presented for seven other tropical sponge species (0.3-1.2 μmol C tracer /mmol C sponge /h and 0.3-1.3 μmol N tracer /mmol N sponge /h) [30,37], confirming the viability of our tested sponges. These findings contribute to a growing body of evidence that DOM-feeding by sponges is not related to the abundance of spongeassociated microbes [8,38,49,50]. NanoSIMS analysis revealed that both symbionts and host choanocyte cells were actively involved in DOM uptake. The rapid 13 Cand 15 N-enrichment of symbiotic microbes (< 15 min) supports direct incorporation rather than transfer of nutrients by host cells. Indeed, incorporation of DOM by host cells was scarce during this initial time frame in the LMA species H. caerulea (Fig. 3i, m), whereas numerous enriched microbial cells were detected in regions distant from choanocyte chambers. Sponge-associated microbes may directly utilise limited amounts of DOM that enter the mesohyl through intercellular spaces between individual choanocytes or cells of the internal and external epithelia [51,52]. Previous studies have revealed 'gaps' in the dermal membrane of sponges [53] and choanocytes have been shown to regulate their intercellular spaces to allow particle entry into the mesohyl [54].
Resident microbes likely incorporated low molecular weight (LMW, < 1 kDa) DOM, which is known to be rapidly (from min to h) assimilated [55,56], as these labile molecules can readily diffuse across microbial membranes or be channelled via porins [57,58].
The potential of choanocytes to directly incorporate DOM was first highlighted in H. caerulea after rapid incorporation of the cell proliferation marker BrdU-provided as a dissolved organic compound-into these cells [59]. Incorporation of DOM by choanocytes has since been confirmed in the bioeroding sponge Cliona orientalis [8] and the Mediterranean sponges Aplysina aerophoba and Dysidea avara [38] after labelling with a range of DOM sources. Here, our time-series approach allowed us to investigate DOM processing by choanocytes over time. We found hotspots of 13 C and 15 N incorporation in the apical tip of choanocyte cells of both sponges (particularly evident in H. caerulea, Fig. 6a-c) after 15 and 30 min of DOMfeeding. This strongly supports direct incorporation of DOM by choanocytes rather than the rapid translocation of metabolites from microbial symbionts located in the sponge mesohyl, as the apical surface of the choanocyte layer faces the surrounding seawater as it is drawn into choanocyte chambers. Likely, DOM is captured and retained within the apical collar complex of the choanocyte and subsequently endocytosed [8,38]. Endocytosis of solutes can proceed via various pinocytotic pathways, including macropinocytosis or clathrin-or caveolae-mediated endocytosis [60]. The formation of membrane 'ruffles' found by Willenz and van de Vyver in 1982 [53], concurrent with the envelopment of latex beads by sponge exopinacocytes, may now be interpreted as evidence of macropinocytosis. Likewise, Laundon and colleagues [61] found macropinocytotic inclusions in choanocytes of a homoscleromorph sponge.

POM uptake
Our findings represent the first confirmation of host to symbiont translocation of POM-derived nutrients in a sponge, the LMA species H. caerulea, and provide support for the hypothesis that sponge symbionts gain nutritional benefit from their host-associated lifestyle. Uptake of POM in the LMA species H. caerulea was rapid (< 15 min; Fig. 2c, d), but NanoSIMS analysis showed that uptake was primarily confined to choanocyte cells, not microbial symbionts. Individual food bacteria were phagocytosed into large intracellular vesicles of choanocyte cells, in contrast to DOM incorporation into the apical collar via pinocytosis (Fig. 6). Substantial symbiont processing of POM did not occur until the end of the 3-h pulse, meaning that POM must first be captured, phagocytosed, and digested by choanocytes before intercellular microbes can scavenge C and N. This is consistent with current understanding that prokaryotes are incapable of phagocytosing bacteria [62]. POMincorporation by choanocytes of the HMA species P. angulospiculatus was low throughout the pulse-chase experiment, and it may be that POM forms a minor part of the natural diet of this species. In subsequent natural diet experiments, we did observe that bacterial uptake rates for P. angulospiculatus are generally low (M. Hudspith and J.M. de Goeij, personal observation). However, we cannot exclude that P. angulospiculatus selected against our cultured POM source (Vibrio caribbeanicus), as sponges can selectively feed on particles, and different species may specialise in retaining specific portions of the planktonic community [54,63,64]. Therefore, we remain cautious in interpreting the ecological relevance of these findings, and our results should be compared with other planktonic sources.
DOM and POM are differentially metabolised by the LMA sponge Halisarca caerulea NanoSIMS images showed different spatial patterns of assimilation in choanocytes of the LMA sponge H. caerulea, indicating that the processing times of the two food sources differed in this species. Numerous intracellular hotspots (1-2 μm diameter) with substantial coenrichment of C and N derived from DOM and POM could be seen within the choanocytes of both sponges after 3 h (yellow arrow heads, Figs. 3 and 5). These regions correlated with electron-dense vesicles in SEM images and represent putative endosomes or endolysosomes (DOM), and phagolysosomes (POM). However, in the POM-fed LMA species, these hotspots were in addition to a more uniform labelling of N throughout the choanocytes, unlike in DOM-fed sponges (Figs. 3o  and 5o). This suggests that within 3 h, enzymatically degraded food bacteria had been digested into small molecules that had passed into the cytoplasm and entered biosynthesis pathways [65]. The comparatively rapid dispersal of N throughout choanocytes of the LMA species H. caerulea indicates that digestion and anabolic processing of particulate food was faster than that of dissolved food, and future studies should investigate if these two food sources differentially contribute to sponge nutrition.
NanoSIMS analysis indicated that DOM-and POMderived C and N were translocated from choanocytes to mesohyl cells of H. caerulea during the chase period. This is consistent with the traditional view that digested material is passed from food capture cells to cells of the mesohyl (mainly amoebocytes) by transcytosis, whereby digestion is completed and metabolites distributed to other cell types [54]. No translocation to mesohyl cells was observed in the HMA species P. angulospiculatus within our experimental time frame, although variable and low uptake in the POM treatment may have masked potential transfer. In the LMA species H. caerulea, mesohyl cells, spherulous cells, and the mesohyl itself, showed a pronounced increase in POM-derived 15 N enrichment between 3 and 48 h ( Fig. 4h; Additional file 2: Table S1), indicating preferential utilisation of POMderived N by cells of the mesohyl and the mesohyl matrix, which consists primarily of the polypeptides collagen, galectin, and fibronectin-like molecules [66]. Taken together, our results suggest that POM represents a high-quality food source for host anabolism (which may include predation on microbial symbionts [39,67]), despite representing only a small portion of the natural organic matter pool and sponge diet [2,50].
The sponge microbiome as metabolic waste processors: commensalism or mutualism?
Long-term associations between sponges and their symbiotic microbes are frequently described as mutualistic, although experimental evidence confirming that symbionts derive benefit from their host is scarce [7,8]. Interactions have largely been addressed from the perspective of the host and the benefits received from symbiotic partners (e.g. photosynthates, vitamins, natural products [6]), whereas reciprocal benefits to symbionts are often presumed. It is generally hypothesised that symbiotic microbes profit from a high supply of nutrients and protection within the sponge mesohyl, but here we provide empirical evidence indicating that the prokaryotic communities of both P. angulospiculatus and H. caerulea directly benefit from the heterotrophic feeding activity of the host. NanoSIMS analysis indicated translocation of C and N from choanocytes to symbionts after DOMand, restricted to the LMA species H. caerulea, POMfeeding (Fig. 4). This does not mean that translocation of metabolites from symbionts to the host did not occur, but rather was not observed under the current experimental design. Alongside the NanoSIMS observations, analysis of the single-cell data confirmed that the trends of increasing isotopic enrichment of microbial symbionts and mesohyl cells during the chase phase, indicative of translocation, were statistically significant (Additional file 2: Tables S3 and S4). However, biological replication is needed to robustly test the variation at the holobiont level. Our findings complement the wealth of genomic studies that have highlighted the potential of microbial symbionts to utilise host products (e.g. [68,69]). Typical for LMA species, the microbiome of H. caerulea is largely dominated by Proteobacteria (Alpha-and Gamma-) [45] while P. angulospiculatus harbours a more diverse microbial community composed of Chloroflexi, Acidobacteria, and Actinobacteria, amongst other groups characteristic of HMA species [41]. These phyla comprise widespread and highly diverse lineages encompassing an extraordinary range of metabolic lifestyles. Thus, translocation of metabolites to symbiotic microbes may be fuelled by a multitude of different pathways, including the uptake of organic compounds released by host cells via incomplete oxidation [32] and the assimilation of host-derived metabolic waste (Fig. 7). Chemoautotrophic members of the sponge microbiome can utilise inorganic C produced during host respiration [70,75,76] and microbial N metabolism can be driven by the assimilation or remineralisation of sponge-excreted nitrogenous wastes [20,[71][72][73]77]. Indeed, ammonia-oxidising archaea and bacteria have been detected in H. caerulea [74]. Given the complexity of the sponge holobiont, further studies are needed to link microbial identity to specific pathways of exchange and could employ RNA-based stable-isotope probing or correlative fluorescence in situ hybridisation and NanoSIMS. Emerging techniques such as time-of-flight (ToF) and hybrid SIMS [78] can also be utilised to determine the exact molecular nature of exchanged metabolites.
While our results show that symbionts of P. angulospiculatus and H. caerulea benefit from host feeding, we propose that the sponge host likely gains reciprocal benefit in a mutualistic manner. Utilisation of host waste products by the microbiome is advantageous for the host because toxic metabolites (e.g. ammonium) may be depurated and/or translocated back to host cells via microbial extracellular release or cell degradation [79]. Fig. 7 Schematic overview of organic matter uptake by the HMA species P. angulospiculatus and LMA species H. caerulea and hypothesised pathways of nutrient recycling by the microbiome. The width of arrows represents expected quantitative difference in fluxes: choanocyte processing pathways are based on NanoSIMS data (this study) and putative microbially mediated pathways are based on published data (partial transcriptional or genomic evidence included) [20,28,[69][70][71][72][73][74]. Internal nitrogen recycling pathways are hypothesised to be elevated in HMA vs. LMA species [13,49]. Host amoebocyte cells digest material passed from food capture cells (not shown), and in HMA species, these cells are prolific consumers of microbial symbionts (this study, [39]) No translocation of organic matter derived C and N from choanocytes to mesohyl cells was evident in the HMA species P. angulospiculatus, but we did observe amoebocyte-like cells of this species routinely phagocytosing multiple intercellular microbes (Additional file 2: Figure S2), demonstrating that symbiont-host transfer may also occur via phagocytosis or 'farming' of symbionts. This has been observed in deep-sea sponges [39,67] and may provide an alternative source of nutrients for host cells in these oligotrophic environments. Additionally, it is plausible that the nutritional sponge host-microbiome interactions observed here are based on benefits received by the symbionts only (commensal) or that the relative benefits change according to intrinsic and external factors. Although further studies are needed to confirm if the internal recycling mechanisms observed here occur more broadly across other HMA and LMA species, we hypothesise that this is likely to be a general feature of sponge symbioses, given that the enrichment of genes related to the processing of hostderived compounds appears to be a core feature of sponge symbionts (e.g. [68,80,81]). The internal recycling of nutrients, mediated by mutualistic interactions with symbionts, would be advantageous for the sponge holobiont by limiting nutrient loss. Indeed, similar nutrient conserving mechanisms have been shown in other marine symbioses [79,82]. Ultimately, these small-scale sponge-microbe interactions drive the internal cycling of the main biological elements (C, N and P) and are significant because they translate to whole sponge functioning and can therefore impact biogeochemical processes at ecosystem scales [83].

Conclusions
Here, we visualise the role of sponge cells and microbial symbionts in the incorporation and exchange of two key dietary food sources: dissolved and particulate organic matter. Our results show that although microbial symbionts were involved in the uptake of DOM, sponge filtering cells appeared to be a major site of organic matter uptake in both the HMA sponge P. angulospiculatus and the LMA sponge H. caerulea, with translocation of C and N from host cells to microbial symbionts occurring over time. This indicates utilisation of host-derived waste products by the microbiome and provides evidence that symbionts benefit from their association with the sponge host. The interplay between the highly efficient filter-feeding host and its microbial symbiontsmost notably by tapping into resources that others cannot readily utilise (i.e. DOM)-provides the sponge holobiont with a competitive edge over other heterotrophic marine organisms and enables them to persist and thrive in the many oligotrophic environments they are abundant in (e.g., the deep sea, the Mediterranean and coral reefs). Further studies quantifying the benefits or costs of symbiotic interactions will enhance our understanding of the influence of microbes on host ecology [25], and determining the response of these symbioses (and hence holobiont) to environmental perturbation will be crucial in the face of increasing anthropogenic pressures.