Single-cell visualization indicates direct role of sponge host in uptake of dissolved organic matter

Abstract

Marine sponges are set to become more abundant in many near-future oligotrophic environments, where they play crucial roles in nutrient cycling. Of high importance is their mass turnover of dissolved organic matter (DOM), a heterogeneous mixture that constitutes the largest fraction of organic matter in the ocean and is recycled primarily by bacterial mediation. Little is known, however, about the mechanism that enables sponges to incorporate large quantities of DOM in their nutrition, unlike most other invertebrates. Here, we examine the cellular capacity for direct processing of DOM, and the fate of the processed matter, inside a dinoflagellate-hosting bioeroding sponge that is prominent on Indo-Pacific coral reefs. Integrating transmission electron microscopy with nanoscale secondary ion mass spectrometry, we track ¹⁵N- and ¹³C-enriched DOM over time at the individual cell level of an intact sponge holobiont. We show initial high enrichment in the filter-feeding cells of the sponge, providing visual evidence of their capacity to process DOM through pinocytosis without mediation of resident bacteria. Subsequent enrichment of the endosymbiotic dinoflagellates also suggests sharing of host nitrogenous wastes. Our results shed light on the physiological mechanism behind the ecologically important ability of sponges to cycle DOM via the recently described sponge loop.

1. Background

Benthic organisms have evolved a suite of adaptations for ample feeding in nutrient-poor environments. Examples of such adaptations can be found in the symbioses that benthic invertebrates form with an array of prokaryotic or eukaryotic microorganisms, or in the mechanisms that allow efficient filter and suspension feeding [1–3]. Marine sponges are known to display both of these adaptations, and in fact, they represent the oldest metazoan–microbe interaction [4]. Many sponge species host diverse and often populous communities of associated microbes, from low-microbial-abundance (LMA) sponges, with concentrations of associated microbes similar to seawater (10⁵–10⁶ ml⁻¹), to high-microbial-abundance sponges, with 10²–10⁵ times more concentrated microbes than the surrounding seawater [5,6]. The sponge microbiome probably provides the sponge with a range of otherwise relatively inaccessible nutrients, such as inorganic nitrogen [7,8] or phosphorous [9]. Sponges also exhibit an efficient filtering system that allows them to trap suspended pico- and nano-plankton from the surrounding seawater [10].

In recent years, it has become clear that sponges have additional adaptations that allow them to effectively consume dissolved organic matter (DOM) [11,12], which is operationally defined as organic matter that will pass through a 0.2–1 µm filter pore [13]. DOM represents the largest pool of organic matter in the oceans [13]. On coral reefs, the labile (i.e. bioavailable to organisms) fraction of this pool is higher than in the surrounding oceans, mainly due to continuous replenishment of carbon-rich photosynthates released by macroalgae and corals [14,15]. In fact, dissolved organic carbon appears to be the main carbon source (approx. 56–97% of daily carbon intake) in the diet of many sponges [16]. Sponges consume DOM and shed particulate organic matter as detritus, which is then shunt to higher trophic levels [17], or store this nutrition as biomass, that is hypothesized to be predated upon by other reef organisms [18]. Similar to the microbial loop [19,20], this so-called sponge loop is being increasingly recognized as a crucial mechanism of retaining carbon on oligotrophic tropical reefs, and in other ecosystems where sponges are abundant, such as on temperate Mediterranean and cold-water reef ecosystems [21–24].

Originally, removal of DOM by sponges was tentatively attributed to their microbial endosymbionts [25,26]. Heterotrophic bacteria are efficient DOM cyclers [14,27], and thus, the primary suspects for DOM cycling inside sponges too. However, ambient seawater is rapidly pumped through the sponge canals without traversing the sponge epithelium [10,28], and so without making direct contact with the mesohyl matrix (where bacteria are hosted extracellularly) or the specialized bacteriocytes (where bacteria are hosted intracellularly) [29,30]. Also, fatty acid profiles of sponges labelled with stable isotopes have demonstrated that sponge cells may also directly participate in the uptake and assimilation of DOM [22,31,32]. The most plausible site of initial uptake in this case would be the choanocytes or primary feeding cells of sponges [11]. These cells are also responsible for phagocytosing and distributing particulate intake from the feeding chambers that they form, to the adjacent sponge structures [33]. To date, direct evidence of DOM processing in sponges at the cellular level is lacking.

Using a novel approach that combines stable isotope labelling, transmission electron microscopy (TEM) and nanoscale secondary ion mass spectrometry (NanoSIMS), we performed a pulse-chase experiment and examined an intact sponge holobiont at the cellular level to investigate (i) the capacity of choanocytes to directly process ¹³C- and ¹⁵N-labelled DOM from the surrounding seawater, and (ii) the fate of heterotrophically obtained DOM inside the sponge. As a study subject, we chose the Indo-Pacific coral-excavating sponge Cliona orientalis, which is known to take up DOM [34,35]. Cliona orientalis belongs to the LMA sponges [36], but hosts a large assemblage of eukaryotic dinoflagellates of the genus Gerakladium (formerly the metazoan-specific lineage of Symbiodinium Clade G [37]) inside specialized cells of its outer body. Energetically boosted by these photosynthetic symbionts [35,38–40], C. orientalis is an important bioeroder of coral frameworks on the Great Barrier Reef and other shallow, well-lit habitats, but the potential for reciprocal nutrition offered to the dinoflagellates by the sponge after uptake of organic material also remains unclear.

2. Methods

(a). Sponge collection

Seven individual C. orientalis sponges (‘beta’ growth form; electronic supplementary material, figure S1a) were sampled in February 2016 on the northern Heron Island reef slope, Great Barrier Reef (5–7 m water depth; 151.9302° E, 23.4326° S). From each sponge, replicate cores of 35 mm in diameter were produced using a circular pneumatic drill (electronic supplementary material, figure S1b). The cores acclimated for three weeks in outdoor flow-through seawater at the Heron Island Research Station, before being randomly assigned to three 40 l aquaria (four cores per aquarium) covered with Marine Blue 131 light filters (Lee Filters, Andover, UK). Noon irradiance and temperature were 243 ± 18 µmol quanta m⁻² s⁻¹ and 27.7 ± 0.1°C, respectively, in the aquaria during the experiment (Odyssey and HOBO pendant loggers; mean ± s.e.m. throughout text).

(b). Preparation of tracer DOM and sponge labelling

After acclimation, a pulse-chase experiment was conducted with stable isotope tracer to examine the incorporation and fate of DOM over time. Isotopically enriched DOM was extracted from a 10 mg l⁻¹ suspension of lyophilized ¹⁵N- and ¹³C-enriched cyanobacterial cells in artificial seawater (isotopic abundances of 98% and 99%, respectively, Sigma-Aldrich, USA; protocol modified from [17]). In brief, cell suspension aliquots were pelleted through centrifugation, and resuspended in milli-Q water overnight to lyse the cyanobacterial cells. The solution was filtered to separate the DOM from particulates and from cyanobacterial cell remnants (6 µm to GF/C to GF/F), and salinity was adjusted. The resulting DOM solution was diluted with 0.22 µm filtered seawater to a final concentration of 127 µmol l⁻¹ DOC (measured using a Total Organic Carbon Analyzer, Shimadzu) and 59 µmol l⁻¹ DON (calculated by persulfate digestion and flow injection analysis). The DOC concentration was double that of summer levels at the study site, but still ecologically relevant for clionaid feeding [41]. The DON concentration was five times higher than background levels, and corresponds to high primary production or coral mucus release [42].

The sponge cores were individually incubated over 3 h (starting at noon) in the dark inside continuously mixed, sealable chambers (250 ml). Dark incubations avoided influence of the photosynthetic activity of the sponge symbionts, inhibiting their uptake of inorganic carbon and reducing their related ammonium assimilation [38,43]. All chambers were temperature-regulated (26°C), and optical probes measured dissolved oxygen concentrations as a proxy for sponge respiration (OXY-10 meter, PreSens, Germany). Incubations (n = 3) with stable-isotope-enriched seawater only served as controls.

After the 3 h pulse phase, labelled sponges were returned to the label-free flow-through aquaria until the end of the 21 h chase phase (3–24 h). At each time point, three cores were removed, two of which were analysed with NanoSIMS (0, 3 and 24 h; 0 h control sponges were not exposed to enriched seawater). Subsamples of the outer photosymbiotic and inner heterotrophic body of each core (electronic supplementary material, figure S1b–f) were separated and fixed for TEM and NanoSIMS analyses following [38]. Sample fixation and embedding remove all inorganic label from the sample, and therefore, only fixed organic enrichment is monitored through NanoSIMS.

(c). Tracing of dissolved organic matter label through NanoSIMS

After fixation, the subsamples were transported to the Centre for Microscopy and Microanalysis (University of Queensland, Brisbane), where they were processed and embedded following [38]. Thin sections (120–200 nm) were mounted onto finder grids (Electron Microscopy Sciences, Hatfield, PA, USA), stained for contrast and viewed with a JEOL 1011 transmission electron microscope. For each grid, replicate raster regions (approx. 35 × 35 µm) that contained cellular structures of interest were chosen (defined below). These were tagged on a virtual map that would guide the NanoSIMS camera during subsequent label tracing.

The pre-mapped grids were imaged with a NanoSIMS ion probe (NanoSIMS 50, CAMECA, Paris, France) at the Centre for Microscopy, Characterisation, and Analysis (University of Western Australia, Perth, Australia) following the methodology detailed by Achlatis et al. [38]. NanoSIMS analysis was centred on five different areas of interest (AOI) in total. For the outer sponge subsamples, three AOI were defined:

• (I)Dinoflagellate symbionts of the genus Gerakladium (cell diameter of 7–10 µm, S in figure 2b,f,j).
• (II)Surrounding cytoplasm of the host archaeocyte-like cells that contain dinoflagellates (hereafter termed ‘dinoflagellate-hosting cells', selections of approximate diameter 5 µm, Hs in figure 2b,f,j).
• (III)Random selections of unidentified sponge or microbial cells and intercellular space of the mesohyl matrix in the same field of view as the dinoflagellates (diameter 5 µm, M in figure 2b,f,j).

For the inner sponge subsamples, two AOI were defined:

• (IV)Choanocyte or filter-feeding cells that form choanocyte chambers (cell diameter 3–4 µm, C in figure 1b,f,j).
• (V)Random selections of unidentified sponge or microbial cells and intercellular space of the mesohyl matrix in the same field of view as the choanocytes (diameter 5 µm, M in figure 1b,f,j).

Other than the above AOI, individual cells that may have occasionally appeared enriched were not systematically quantified. In total, six individual sponge cores were analysed (2 per time point), with an unequal number of raster regions and resulting AOI per raster (detailed in data table 1, available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.7wm37pvng [44]). Eighteen choanocyte chambers were examined overall, with, on average, 12 choanocytes per chamber.

Post-NanoSIMS, the OpenMIMS software (Harvard University) in Fiji platform [45] was used to calculate the isotopic values of the AOI. For an unbiased selection, firstly, the corresponding TEM maps were consulted to verify and draw the AOI onto the ¹²C¹⁴N⁻ ratio maps, which only display the naturally abundant isotopes. The drawn AOI were then superimposed on the ¹⁵N/¹⁴N and ¹³C/¹²C ratio maps (figures 1c,g,k and 2c,g,k for nitrogen, figures 1d,h,l and 2d,h,l for carbon), quantifying the mean isotopic ratios of the selected structures. Enrichment levels of the 3 and 24 h time points were expressed using delta–delta notations (Δδ¹⁵N and Δδ¹³C above the 0 h background, in ‰) as follows:

$$\mathrm{\Delta }{\delta }^{15}\mathrm{N}=\left(\frac{N_{\mathrm{mes}}}{N_{\mathrm{nat}}}-1\right)\times {10}^3$$

and

$$\mathrm{\Delta }{\delta }^{13}\mathrm{C}=\left(\frac{C_{\mathrm{mes}}}{C_{\mathrm{nat}}}-1\right)\times {10}^3,$$

where N_mes and C_mes = ¹⁵N/¹⁴N and ¹³C/¹²C ratio, respectively, measured in the AOI of the labelled sponges.

Figure 1.

Visualization of ¹⁵N- and ¹³C-DOM enrichment of the inner heterotrophic body (ochre-yellow in electronic supplementary material, figure S1b) of the bioeroding sponge C. orientalis. (a,e,i) TEM images of three selected samples before the addition of label (t = 0 h), at the end of the pulse period (t = 3 h) and at the end of the chase period (t = 24 h). The red rectangle displayed in (e) is enlarged in figure 4. (b,f,j) NanoSIMS ¹²C¹⁴N⁻ image displaying the selection of the AOI for enrichment quantification: (IV) Choanocytes (in purple, C; total measured n = 316) and (V) nearby inner-body mesohyl cells and intercellular space (in blue, M; n = 115). (c,g,k) NanoSIMS distribution of ¹⁵N/¹⁴N ratio and (d,h,l) of ¹³C/¹²C ratio over the same selected areas at each time point. The rainbow scale shows the deviation of the ratios from the natural abundance ratio, ranging from natural abundance in blue (0.0037 for ¹⁵N/¹⁴N and 2× 0.0110 for ¹³C/¹²C) to several-fold enrichment in red (approx. threefold and fivefold, respectively). Note the hotspots of carbon enrichment inside choanocyte cells at 3 h (see also figure 4). (Online version in colour.)

Figure 2.

Visualization of ¹⁵N- and ¹³C-DOM enrichment in the outer photosynthetic body (brown in electronic supplementary material, figure S1b) of the photosymbiotic bioeroding sponge C. orientalis, with a focus on the resident Gerakladium dinoflagellates. (a,e,i) TEM images of three selected samples (time points as described in figure 1). (b,f,j) NanoSIMS ¹²C¹⁴N⁻ image displaying the selection of the areas of interest (AOI) for enrichment quantification: (I) Dinoflagellate symbionts (in green, S; total measured n = 94), (II) Cytoplasm of dinoflagellate-hosting sponge cells (in yellow, Hs, n = 68) and (III) nearby outer-body mesohyl cells and intercellular space (in red, M; n = 145). (c,g,k) NanoSIMS distribution of ¹⁵N/¹⁴N ratio and (d,h,l) of ¹³C/¹²C ratio over the same selected areas at each time point. The rainbow scale is detailed in figure 1. (Online version in colour.)

N_nat and C_nat = natural background ¹⁵N/¹⁴N and ¹³C/¹²C ratio, respectively, measured in the AOI of the non-labelled control sponges.

Assuming that the average biomass (i.e. N and C content) of each AOI category remains constant over the experimental duration, delta–delta notations can be used as a proxy for the relative tracer ¹⁵N and ¹³C uptake of each AOI over time. To quantify absolute tracer uptake, and to compare this between AOI, delta–delta notations would have to be normalized to the N and C content of each of the 5 AOI individually (not done here) [46]. Either way, NanoSIMS is foremost a qualitative assessment of isotope-tracer processing and tends to underestimate absolute uptake rates, especially of carbon, due to the loss of low-molecular weight compounds that are soluble and poor in amino-groups during fixation, or the addition of external carbon during embedding [47]. These methodological biases apply equally to all sample areas measured, allowing for relative comparisons.

(d). Statistical analysis

Semiparametric permutational analysis of variance (PERMANOVA) was used to test the Euclidean dissimilarities between the Δδ¹⁵N and Δδ¹³C levels of each AOI across the treatment time points in a two-factor design [38,48]. The first factor was Time (post-pulse t = 3 h and post-chase t = 24 h) and the second was AOI [(I)–(V)]. Based on dissimilarity matrices and 4999 permutations, the pseudo-F statistic and the P_(perm) value were produced under a full model using type III sums of squares to account for the unbalanced design in terms of AOI replication. Significant interaction effects were followed by pairwise post hoc comparisons using permutational pseudo-T tests. Data from the replicate sponges per time point were pooled for each AOI. As an indication of significant enrichment of the AOIs at the t = 3 h and t = 24 h time points, the respective ¹⁵N/¹⁴N and ¹³C/¹²C ratios were compared to those of the control sponges (electronic supplementary material, table S1). All analyses were done in Primer v. 6 [49] and significance was set at the P_(perm) < 0.05 level.

3. Results

Continuous linear decrease in dissolved oxygen concentration throughout the sponge incubations (12.5 ± 0.9 µmol O₂ h⁻¹ incubation⁻¹), in combination with the visual observation of open oscula, confirmed actively pumping and respiring sponge specimens. Moreover, non-sponge control incubations showed minimal background respiration (less than 5% of sponge incubations).

The NanoSIMS analysis illustrated ¹⁵N and—to a lesser extent—¹³C enrichment in individual sponge and symbiont cells of the inner (figure 1) and outer sponge body (figure 2) of the intact holobiont (all values in data table 1 at [44]). It also detected metabolic heterogeneity within AOI populations (undetectable in bulk studies) (figure 3). The ¹⁵N/¹⁴N and ¹³C/¹²C ratios of all AOIs at t = 3 h and t = 24 h were higher than the respective background ratios of control sponges (all p_(perm) < 0.05; electronic supplementary material, table S1). Depending on the AOI, the Δδ¹⁵N and Δδ¹³C enrichment levels differed between the pulse and chase time points (two-way interaction; ¹⁵N: pseudo-F = 4.211, p_(perm) = 0.005; ¹³C: pseudo-F = 6.383, p_(perm) = 0.032; electronic supplementary material, table S2).

Figure 3.

Quantification of (a) nitrogen and (b) carbon enrichment (Δδ above background) in host and dinoflagellate cells of the bioeroding sponge C. orientalis in response to a pulse of seawater enriched in ¹⁵N- and ¹³C-DOM. Through NanoSIMS isotopic analysis, the mean ¹⁵N and ¹³C enrichments were quantified in a total of five different areas of interest (AOI). Twenty raster regions were analysed at t = 3 h and 17 at t = 24 h, and numbers above the bars show the pooled AOI measured (n = …) (details in data table 1 [44]). Asterisks denote significant differences within single AOI types from t = 3 h to t = 24 h. The box–whisker plots display the data in quartiles, with the bottom of the box corresponding to the 25th percentile, the line within the box marking the median, the top of the box corresponding to the 75th percentile and the circles denoting the means. Note the large difference in the y-axis scales of the two panels. (Online version in colour.)

(a). Inner sponge body

Seawater containing DOM will initially come into contact with the choanocyte chambers that line the canals in the inner sponge body (electronic supplementary material, figure S1e,f; figure 1). Choanocytes were enriched in ¹⁵N (5178 ± 1080‰; mean ± s.e.m.) after the 3 h pulse, showing highest mean enrichment compared to all other cell types, including the dinoflagellate symbionts (figure 3a). At the end of the chase phase, the choanocytes remained enriched in ¹⁵N (817 ± 121‰), but at lower levels compared to the end of the pulse (two-way interaction, pairwise comparison t = 3 h versus t = 24 h, p_(perm) = 0.006; figure 3a). At both pulse and chase time points, Δδ¹⁵N values of inner mesohyl cells did not exceed 100 ± 9‰ (figures 1 and 3a).

The choanocytes were also enriched in ¹³C (213 ± 41‰) after the 3 h pulse, again showing the highest mean enrichment compared to the other cell types (figure 3b). However, the ¹³C enrichment inside the choanocytes appeared to be confined to well-defined subcellular hotspots, as opposed to the ¹⁵N-enrichment, which was shown throughout the cytoplasm (figure 4). These hotspots, where Δδ¹³C values were more than 10 times higher than when calculated over the entire cell (figure 4b), were clearly located inside the choanocyte cells and were not bacterial. The ¹³C enrichment of the choanocytes was highly variable between and within scanned chambers. At the end of the 3 h pulse, half of the measured choanocytes displayed at least one ¹³C-hotspot, with an average of seven hotspots found per chamber. At the end of the chase phase, Δδ¹³C of choanocytes was lower than after the pulse phase (53 ± 7‰; p_(perm) = 0.003) and hotspots were no longer observed, with the exception of 2 out of the 71 choanocytes examined at this time point. Again, Δδ¹³C values of the inner mesohyl cells remained low (≤14 ± 3‰, figures 1 and 3b).

Figure 4.

Visualization of carbon-enrichment hotspots of the choanocyte cells of C. orientalis after 3 h of exposure to ¹⁵N- and ¹³C-enriched DOM. (a) High-resolution TEM image of choanocytes (lower magnification shown in figure 1e) and corresponding ¹⁵N/¹⁴N and ¹³C/¹²C quantification of the same areas. Choanocytes (as well as occasional unidentified host cells) actively processed nitrogen throughout most of their endocellular material, whereas carbon processing appears concentrated in individual hotspots inside the cells. (b) Quantification of the Δδ¹³C levels of all choanocyte cells examined, and the hotspots found inside choanocyte cells. Note the large difference in the y-axis scales of the two graphs. (Online version in colour.)

(b). Outer sponge body

Dinoflagellate symbionts showed a ¹⁵N-enrichment of 3011 ± 298‰ after the pulse (figures 2 and 3a), and this enrichment increased at the end of the 21 h chase to 4503 ± 555‰ (t = 3 h versus t = 24 h, p_(perm) = 0.012). Similarly, the dinoflagellate-hosting cells displayed a strong ¹⁵N-enrichment at the end of the pulse (837 ± 255‰) which more than doubled by the end of the chase (2028 ± 284‰, t = 3 h versus t = 24 h, p_(perm) = 0.004). Outer sponge mesohyl cells maintained relatively low Δδ¹⁵N values albeit also increasing over time (t = 3 h versus t = 24 h, p_(perm) = 0.004; figure 3a).

Regarding the carbon enrichment, all three targeted cell types of the outer sponge body maintained relatively low Δδ¹³C values (figure 3b).

4. Discussion

Marine sponges contribute to nutrient cycling by mass feeding on DOM (the largest available marine source of organic matter), which is a challenging feeding strategy for most other multicellular organisms [17,50]. Through the combination of isotopic labelling and NanoSIMS, we show for the first time at subcellular level that the choanocytes, the sponge filtering cells, have an active role in the processing of DOM in the excavating demosponge C. orientalis. This contradicts with the hypothesis that DOM processing is confined to prokaryote symbionts within the microbiome of sponges (e.g. [13,26,30,51]). Our results also show that a fraction of the processed matter is efficiently shared with the eukaryotic dinoflagellates present in the sponge, illustrating how life in nutrient-poor environments has adapted to efficiently recycle scarce nutrients, at the organismal as well as the ecosystem level.

(a). Sponge cells play a direct role in the first step of the sponge loop

Our NanoSIMS results illustrate that the sponge choanocytes are capable of trapping and incorporating DOM in the LMA species C. orientalis. Although most of the DOM in the oceans resides in a low-molecular weight size fraction (‘true dissolved’ fraction, less than 1 kDa), high molecular weight DOM (colloidal fraction, greater than 1 kDa) that forms the boundary between true dissolved and particulate is an important source for certain filter-feeding invertebrates [52,53]. Sponges use both fractions albeit the truly dissolved fraction was mainly found to be processed by the associated bacteria (¹³C-labelled glucose in the study in question), whereas algal-derived DOM was processed by both the bacteria and host [31]. The latter corroborates the results of our study and implies that the sponge holobiont has access to a diverse DOM pool, which is a major competitive advantage compared with organisms to which DOM is largely unavailable under oligotrophic conditions. The fact that choanocytes were the predominant DOM-enriched sponge cells in our study further supports the hypothesis of DOM uptake by both low- and high-microbial abundance sponges that share certain morphological traits (reviewed by de Goeij et al. [16]).

In our experiment, the carbon enrichment was concentrated in well-defined hotspot areas inside the choanocyte cells. This pattern allows us to suggest that DOM was taken up via pinocytosis and first stored in intracellular vesicles. Also known as solute endocytosis, or cell drinking, pinocytosis involves entrapment of extracellular fluid via invaginations of the cell membrane [54]. Pinocytosis can be achieved by either non-specific adsorption of the DOM to the microvilli extensions of the choanocyte membrane or by the more efficient involvement of specific membrane receptors. At the end of the 3 h pulse, we detected the nitrogen signal in the diluted material that by then occupied most of the choanocyte cell, whereas the carbon signal was not detectable once diluted. This implies that vesicle-stored DOM was, at least partially, already processed within the first 3 h of our experiment, spreading the fixed material inside the choanocytes. The poor carbon signal in comparison may reflect preferential burning of carbon-rich compounds to fuel the energetically costly beating of choanocyte flagellae [10,55], or abovementioned methodological biases against carbon fixation in the sample processing [47].

The enrichment levels that we report under a dark pulse may underestimate the full capacity of C. orientalis to use DOM in the light, when pumping rates are highest [56]. DOM utilization may be even higher in other LMA sponges, as pumping of C. orientalis is atypically slow for an LMA species [56], due to its reliance on autotrophy over heterotrophy [35,38,39].

(b). Sharing of waste products with dinoflagellate symbionts

Our NanoSIMS results offer one of the first insights into the fate of DOM after entering the sponge via the choanocytes. As mentioned above, part of the material may be directly metabolized by the choanocytes [10,55]. Our data show that metabolized products are then further processed reaching the dinoflagellate symbionts, implying that in C. orientalis, and potentially other phototrophic sponges, nutrients are to a certain extent cycled internally prior to being released.

Sponges typically release dissolved inorganic nitrogen (DIN), but this source can also be internally cycled by endosymbiotic microbial communities [29]. The strong nitrogen enrichment we found in the dinoflagellates points towards interception of host nitrogenous waste products as they concentrate in the outer sponge body en route to excretion, possibly enhanced as exhaled, DIN-labelled water is re-filtered by the sponge [57]. Accordingly, while the nitrogen enrichment of the choanocytes decreased over time, that of the dinoflagellates increased. Such waste products could also be shared in the form of organic molecules (urea) [58], although the comparatively trivial carbon enrichment of the dinoflagellates can be better explained by inorganic nitrogen uptake. The distinct nitrogen versus carbon signal of the dinoflagellates­ further deems direct heterotrophic feeding by the protists on the DOM source unlikely, as such feeding would necessarily result in greater similarity between the two signals. The poor carbon signal may also reflect metabolic burning rather than storage, and/or methodological biases in the fixation of carbon.

Direct, non-host-mediated assimilation of labelled DIN by the dinoflagellates is an alternative possibility [38]. However, direct assimilation is less likely here because (i) the labelling pulse was performed in the dark, whereas such assimilation is most active in the light in hospite (but not necessarily in free-living dinoflagellates [59]) as light-driven concentration gradients allow DIN to reach the dinoflagellates [43,60] and (ii) the nitrogen enrichment became higher over time, despite the chase being conducted in label-free seawater. Direct assimilation of labelled inorganic carbon by photosynthesis of the dinoflagellates was a priori inhibited in darkness.

5. Conclusion

We have examined a relatively unexplored heterotrophic pathway of an abundant and ecologically important bioeroding sponge, illustrating that sponges might be capable of efficient DOM cycling, regardless of their prokaryotic communities. Our methodology is suitable for detecting prokaryotic enrichment inside sponges (see electronic supplementary material, figure S2), and might facilitate necessary studies targeting bacteria to elucidate their comparative role in DOM uptake in hospite. Without the involvement of choanocytes, however, we would expect bacterial uptake to be limited, given that DOM needs to transverse at least one cell membrane (two for intracellular bacteria) before reaching the endosymbiotic microbes in the first place [29].

The ability of sponge cells to take up and use DOM from the surrounding seawater is not a unique feature. Animal cells in a plethora of soft-bodied marine invertebrates such as corals, echinoderms, molluscs and pogonophorans, as well as a range of marine larvae, share this ability, many at the low-DOM concentrations found naturally [50,61]. What is remarkable about sponges though is that DOM supports the lion's share of their metabolic needs (usually at more than 90% of their daily organic carbon intake), when in most other organisms, proportions of 1–20% are documented [16,50]. Mass DOM feeding of sponges may be facilitated by the evolution of a complex filtration system with a grand internal surface area lined by choanocyte microvilli, maximizing the volume of seawater that sponges (and associated bacteria) are exposed to, and the quantity of extracted material [12]. In this way, DOM turnover by sponges becomes ecologically important, in some cases meeting rates equal to daily gross primary production of entire reef systems [17].

We also show that the symbiosis between C. orientalis and its intracellular dinoflagellates meets the requirements of a mutualism (when undisturbed [62]), as dinoflagellates are rewarded for the generous supply of photosynthates that they share with their host cells [38,39]. Mutualistic symbioses between reef-building species and Symbiodiniaceae lay at the heart of coral reef growth [63]. Variations of such symbioses, however, also provide evolutionary advantages to species that thrive on reef excavation. Understanding how dominant bioeroding species sustain their metabolism has important implications for understanding their ecological success, providing insights into potential downstream consequences for their bioerosion rates inside the skeletons of the already-stressed corals that they inhabit.

Being the oldest metazoans, and the oldest metazoan–microbe partnership, sponges provide a unique model for exploring the evolution of life in the sea. Unlike terrestrial life, invertebrates in marine environments are constantly exposed to a soup of organic nutrients in dilute solution, and to a suspension of potential microbial symbionts ready to be acquired. Our study highlights some of the vital strategies that sponges have evolved to make the most out of their sedentary lifestyle in these environments. Such strategies, among others, have allowed sponges to persist through evolutionary time and to become some of the projected winners in future benthic environments [64].

Data accessibility

Isotopic data are available in the Dryad Digital Repository: https://doi.org/10.5061/dryad.7wm37pvng [44].

Authors' contributions

M.A., M.P., S.D. and O.H.-G. conceived the ideas and designed methodology; M.A. ran the pulse-chase experiment with advice from M.P., S.D. and O.H.-G.; M.A. and K.G. prepared and analysed the TEM samples; P.G. and M.R.K. led the NanoSIMS analysis; M.A., M.P., S.D. and J.M.d.G. analysed the data; M.A. and M.P. led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

Competing interests

The authors declare no conflict of interest.

Funding

This work was supported by the Australian Research Council (ARC) Laureate (grant no. FL120100066) (O.H.-G.), the ARC Centre of Excellence for Coral Reef Studies (grant no. CE0561435) (S.D. and O.H.-G.), the Australian Government Research Training Program Scholarship (M.A.) and the Holsworth Endowment by the Ecological Society of Australia (M.A.).