When photosynthesis was active, [1-¹³C]-pyruvate produced ¹³C-labelling patterns qualitatively similar to those previously observed in corals exposed to ¹³C-labelled bicarbonate in seawater, with preferential accumulation in the pyrenoid and starch granules of the dinoflagellates, and translocated ¹³C-labelled lipids in the coral host tissue^10,13,15 (Fig. 1a). As expected, when photosynthesis was inactive, i.e., in the light + DCMU and night treatments, no ¹³C-assimilation was detected in either the symbionts or the host lipid bodies (Figs. 1b and  2c, d). However, weak and diffuse ¹³C-labelling was observed in both the oral epidermis and gastrodermis in all incubation conditions (Fig. 2a, b). This may result from [1-¹³C]-pyruvate redirected through gluconeogenic pathways, but this background-labelling level is insignificant compared to those obtained with [2,3-¹³C]-pyruvate under similar conditions.Fig. 1

Corals were incubated with differentially labelled pyruvate, in the presence or absence of the photosynthetic inhibitor DCMU, for 12 h. Scanning electron microscopy (top row) images and their correlative NanoSIMS image (bottom row) are shown for corals labelled with a [1-¹³C]-pyruvate in the light; b [1-¹³C]-pyruvate in the light + DCMU; c [2,3-¹³C]-pyruvate in the light, and d [2,3-¹³C]-pyruvate in the light + DCMU. Circles labelled with ‘S’ show the position of the algal symbionts in their hosts’ oral gastrodermis, while unmarked circles show the position of host lipid bodies. Note that only a couple of host lipid bodies are circled for illustration purposes; these ROIs do not reflect the real abundance of lipids in the tissue, all of which were included in the analysis. See text for an explanation of the differential labelling patterns.

Incubation with [2,3-¹³C]-pyruvate resulted in strong, heterogeneously distributed ¹³C-labelling of the host tissues in all conditions (Figs. 1c, d and  2a, b). The similarity of ¹³C-assimilation levels in the epidermis (seawater-facing) and gastrodermis (coelenteron-facing) in the light + DCMU and night treatments supports the assumption that tissue layers have access to, and assimilate pyruvate equally (Supplementary Fig. 1), consistent with the recent observation of ubiquitous (pyruvate-transporting) monocarboxylate transporters in cnidarians¹⁶. The total absence of ¹³C-labelling in the symbionts in the light + DCMU incubation (Figs. 1d and 2d) was expected, because dinoflagellate algae lack the pyruvate dehydrogenase complex required to breakdown pyruvate into acetyl-CoA¹⁷. They are thus unable to use pyruvate directly to fuel their own TCA cycle, or produce lipids.

The [2,3-¹³C]-pyruvate incubations permit several novel observations. Comparison between anabolic ¹³C-assimilation in the light versus the light + DCMU and night incubations, permits quantification of the symbiont boost to host anabolism via translocation. In our experiments, translocation increased ¹³C-incorporation by 13–22% in the epidermis (Fig. 2a), 40–58% in the gastrodermis (Fig. 2b), and 135–169% in gastrodermal host lipid bodies, relative to the light + DCMU and night incubations (Fig. 2c). ¹³C-labelling in the host lipid bodies in the light + DCMU and night incubations (i.e., when the recycling of metabolic ¹³CO₂ is either blocked or inactive) has to derive from host de novo lipid synthesis. Comparison of the ¹³C-labelling levels in the light + DCMU and night treatments with those in the light, thus allows us to estimate how much host de novo lipid synthesis contributes to the total holobiont lipid reserve, when the symbiosis is functioning. We found this to be ~40% under the conditions used in this study. Finally, comparison of ¹³C-assimilation in the light + DCMU and night incubations permits to determine whether host anabolism itself exhibits diel rhythmicity; we found no statistically significant differences in the tissue regions studied (Fig. 2a–c) and thus no evidence for diel rhythmicity in anabolic turnover.

To test for differences in ¹³C-labelled pyruvate assimilation between different regions of interest (ROIs) in the coral polyp, data were first analysed by two-way ANOVA with region of interest and colony as factors (Supplementary Table 1). A significant interaction between these two-factors was detected, meaning that anabolic turnover patterns differ between polyps from different colonies (Fig. 3). Therefore individual polyps were analysed separately (Supplementary Table 2). The mesentery filaments (MFs) always had the highest levels of anabolic turnover, while the tentacles or pharynx had the lowest (Fig. 3), which is reasonable considering the localization, function and cellular composition of these structures. MFs are located in the corals gastric cavity, where they play an important role in cleaning the polyp¹⁸ and in the digestion of food and pathogens¹⁹. They contain numerous cell types including (digestive) phagocytes, (secretory) gland cell morphotypes, mucocytes, and epithelio-muscular cells^20,21. Correlated electron microscopy (EM) revealed ¹³C was preferentially incorporated in large gland cells (Fig. 4b, white arrows), which contained heterogeneously labelled matter.Fig. 3

Individual polyps (n = 3) originating from corals incubated with [2,3-¹³C]-pyruvate in the light + DCMU treatment, were imaged to determine whether host anabolism alone (i.e., without the symbiont contribution) differs between regions of the polyp. Four regions of interest were identified: aboral tissue layers, mesentery filaments, pharynx and tentacle. The relative anabolic assimilation of ¹³C (denoted by majuscule letters) differed substantially between regions of interest within individual polyps. Shown are schematic representations of the allocation of ¹³C (expressed relative to the highest enrichment level measured) and the quantitative data (mean ± SE, n = 5 images per structure per colony).

The tentacles also contain mucocytes and epithelio-muscular cells to facilitate their retraction²². However, as in the case of the pharynx, they are dominated by columnar support epithelial cells and nematocytes; stinging cells involved in prey capture²¹. ¹³C-enrichment was not observed in mucocytes, nor in nematocytes (Figs. 1c and  4d), suggesting that once formed, these structures require little anabolic maintenance. It is worth noting that both the acclimation period and the isotopic-labelling experiments were conducted in filtered seawater to eliminate prey from the water and prevent the confounding effects of heterotrophy on host metabolism^15,23. It is possible that this led to reduced anabolic activity in these areas, and that regeneration or repair would have been higher if an element of feeding had been added to the experimental design.

Another interesting observation was the presence of ¹³C-rich hotspots in all of the ROIs studied (Fig. 4). Some of the hotspots (particularly those present in the MFs; Fig. 4b) were contained in vesicles, however most hotspots appeared to be free in the host tissue and outside the cell nuclei (Figs. 1 and  4); their molecular identity remains to be determined.

In this study, we show that pyruvate is an effective isotopic marker of anabolism in photosymbiotic holobionts. When combined with NanoSIMS isotopic imaging, isotopically labelled pyruvate provides quantitative information at the tissue- and single cell level and can be applied to disentangle complex host–symbiont metabolic interactions. Of particular current interest, the method can be used to throw light on how reef-building corals respond to temperature-induced bleaching: i.e., when corals expel or lose their symbiotic algae²⁴. An advantage of this tool is that it is broadly compatible with bulk tissue analyses, such as high-performance liquid chromatography²⁵ or gas chromatography–mass spectrometry⁸, which lack spatial resolution but provide additional important information about the molecules being formed, or the specific anabolic pathways that are active.

Three Stylophora pistillata mother colonies were collected in August 2018 at 8 m depth from a coral nursery situated adjacent to the Inter-University Institute for Marine Sciences (Eilat, Israel). The corals were fragmented, mounted on numbered plugs and placed in separate tanks in the Red Sea Simulator (RSS) aquarium system²⁶, where they were left for a month to recover from any handling stress incurred and to acclimate to ambient conditions (Supplementary Fig. 2a). Corals were not fed during acclimation to eliminate the potentially confounding effect of heterotrophy on host metabolism^15,23.

12 h isotopic pulses were conducted in 250 mL glass beakers, set atop a submersible magnetic stir-plate, which was placed in a flow-through aquarium. Day (light) and night incubations were conducted in ambient thermal conditions (26 ± 1 °C) in accordance with the diel light cycle in Eilat (day: 06:30–18:30). Light incubations were conducted under natural, but shaded light (mean: 144 ± 230 μmol photons m⁻² s⁻¹) conditions (Supplementary Fig. 2b) and used [1-¹³C]-pyruvate or [2,3-¹³C]-pyruvate (Cambridge Isotope Laboratories Tewksbury, MA, USA), with and without the photosynthetic inhibitor 3-(3,4-dichlorophenyl)-1,1-dimethylurea (DCMU). Night incubations used [1-¹³C]-pyruvate and [2,3-¹³C]-pyruvate only. Pyruvate (500 mmol stock prepared in distilled water) was added at a concentration of 1 mM. This concentration was deemed sufficient, because it produced detectable levels of labelling in NanoSIMS images in preliminary trials. DCMU (stock dissolved at 0.01% in ethanol) was added at 10 µM; a common concentration used to block photosynthesis in corals²⁷. A separate experiment using fragments from the same mother colonies as those used in the isotopic-labelling experiments was conducted to ensure this concentration did not affect the respiration (and thus metabolic functioning) of the coral host (Supplementary Fig. 3).

During the isotopic-labelling experiments, water changes were performed and fresh isotopic labels were added every 3 h to ensure stable water chemistry. At the end of the labelling experiment, the apical tip of each coral fragment was removed and a 1 cm coral piece was clipped off and immersed in fixative (0.5% formaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer with 0.6 M sucrose, pH 7.4–7.6) for 24 h at room temperature¹⁰. Pieces were washed and transferred to 0.1 M phosphate buffer containing 0.5 M ethylenediaminetetraacetic acid (EDTA), where they were stored at 4 °C, until their calcium carbonate skeletons were fully dissolved (1–2 weeks).

Samples were dissected into small tissue pieces containing a single polyp and post-fixed for 1 h at room temperature with 1% osmium tetroxide in 0.1 M phosphate buffer. The samples were then dehydrated in a graded series of ethanol (50%, 70%, 90%, and 100%), and embedded in Spurr resin blocks. Thin (200 nm) and semi-thin (500 nm) sections were cut using a 45° Diatome diamond knife and mounted on round glass slides (10 mm) for scanning transmission electron microscopy (GeminiSEM 500, Carl Zeiss Microscopy GmbH, Jena, DE), or NanoSIMS imaging.

All NanoSIMS images (40 × 40 μm, 256 × 256 pixels, 5 ms pixel⁻¹ dwell time, five layers) were obtained using a 16 keV Cs⁺ primary ion beam, focused to a spot-size of about 120 nm. Secondary ions (¹²C₂⁻, ¹³C¹²C⁻) were simultaneously counted in individual electron-multiplier detectors, with a mass resolving power of ~9000 (Cameca definition). Isotopic ratios were formed from drift-corrected ion images using the ratio of ¹³C¹²C⁻ to ¹²C₂⁻ and expressed as parts-per-thousand (‰) deviation relative to an isotopically unlabelled coral tissue sample prepared and analysed in an identical manner.

Two separate experiments were performed: (1) A proof-of-concept experiment, designed to quantify the different labelling patterns produced by [1-¹³C]-pyruvate and [2,3-¹³C]-pyruvate, across all experimental incubations (i.e., light, light + DCMU and night). (2) Examination of anabolic variation in different tissue regions of the coral polyp. These two experiments are presented and discussed separately below.

1. Proof-of-concept experiment: Analysis was limited to the oral tissue layers of the coenosarc tissue (i.e., epidermis and gastrodermis). In these two layers, it was possible to define four ROIs using the L’IMAGE software (Dr. Larry Nittler, Carnegie Institution of Washington). These ROIs were: host epidermis, host gastrodermis (excluding lipid bodies and dinoflagellates), host lipid bodies, and dinoflagellate cells. ROIs were drawn using the ¹²C¹⁴N⁻ images. The epidermis and gastrodermis are separated by the mesoglea and are thus easy to separate in these images. Dinoflagellate cells are also easily identified by their size and internal structure²⁸. Lipid bodies are distinguished by their size and drop-like appearance and their low ¹²C¹⁴N⁻ signal (and hence relatively low N:C ratio, e.g. Fig. 1); these identifications have been tested numerous times against TEM images. A minimum of five images were taken of each experimental treatment (30 symbiont cells), producing a dataset that contained 138 images (Supplementary Table 3).

To test for differences in ¹³C-labelled pyruvate assimilation between [1-¹³C] and [2,3-¹³C]-pyruvate, separate Student’s t-tests (Bonferroni-correction: α = 0.004) were performed for each experimental treatment (light, light + DCMU, night) and each region of interest (epidermis, gastrodermis, host lipid bodies, symbiont; Supplementary Table 4). To test whether treatment influenced ¹³C-labelled pyruvate assimilation within a region of interest, a restricted maximum-likelihood model was used, with treatment (light, light + DCMU, night) as a fixed factor and colony as a random factor (Supplementary Table 3). If treatment was found to be significant, Tukey HSD tests were performed to identify where the differences lay. Where necessary, data were square root transformed (Supplementary Table 5) prior to analysis in order to meet the assumptions of normality (Shapiro–Wilk W test) and homogeneity of variance (Levene’s test).

2. Examination of anabolic variation in the coral polyp: Tissue layers in four regions of the polyp were identified as areas of interest. These were the aboral tissues (comprised of the aboral gastrodermis and the calicodermis), the tentacle, the pharynx, and the mesenteries filaments. These structures were selected because they contain highly specialized cell types adapted to their biological function. Five NanoSIMS images (same settings as above) were taken per structure for each polyp, generating a dataset of 60 images (Supplementary Table 6).

To test for patterns in pyruvate-derived ¹³C-assimilation between structures from different colonies, data were first analyzed by two-way ANOVA with structure (aboral tissues, tentacle, pharynx, mesentery filament) and colony (1–3) as fixed factors (Supplementary Table 1). Data were then split by colony and re-analyzed by one-way ANOVA and Tukey HSD tests to identify where differences lay (Supplementary Table 2). In all cases, assumptions of normality and variance equality were met.

Further information on research design is available in the Nature Research Reporting Summary linked to this article.