Ammonium is the preferred source of nitrogen for planktonic foraminifer and their dinoflagellate symbionts

The symbiotic planktonic foraminifera Orbulina universa inhabits open ocean oligotrophic ecosystems where dissolved nutrients are scarce and often limit biological productivity. It has previously been proposed that O. universa meets its nitrogen (N) requirements by preying on zooplankton, and that its symbiotic dinoflagellates recycle metabolic ‘waste ammonium’ for their N pool. However, these conclusions were derived from bulk 15N-enrichment experiments and model calculations, and our understanding of N assimilation and exchange between the foraminifer host cell and its symbiotic dinoflagellates remains poorly constrained. Here, we present data from pulse-chase experiments with 13C-enriched inorganic carbon, 15N-nitrate, and 15N-ammonium, as well as a 13C- and 15N- enriched heterotrophic food source, followed by TEM (transmission electron microscopy) coupled to NanoSIMS (nanoscale secondary ion mass spectrometry) isotopic imaging to visualize and quantify C and N assimilation and translocation in the symbiotic system. High levels of 15N-labelling were observed in the dinoflagellates and in foraminiferal organelles and cytoplasm after incubation with 15N-ammonium, indicating efficient ammonium assimilation. Only weak 15N-assimilation was observed after incubation with 15N-nitrate. Feeding foraminifers with 13C- and 15N-labelled food resulted in dinoflagellates that were labelled with 15N, thereby confirming the transfer of 15N-compounds from the digestive vacuoles of the foraminifer to the symbiotic dinoflagellates, likely through recycling of ammonium. These observations are important for N isotope-based palaeoceanographic reconstructions, as they show that δ15N values recorded in the organic matrix in symbiotic species likely reflect ammonium recycling rather than alternative N sources, such as nitrates.

N-enrichment experiments and model calculations, and our understanding of N assimilation and exchange between the foraminifer host cell and its symbiotic dinoflagellates remains poorly constrained. Here, we present data from pulse-chase experiments with 13 C-enriched inorganic carbon, 15 N-nitrate, and 15 N-ammonium, as well as a 13 C-and 15 N-enriched heterotrophic food source, followed by TEM (transmission electron microscopy) coupled to NanoSIMS (nanoscale secondary ion mass spectrometry) isotopic imaging to visualize and quantify C and N assimilation and translocation in the symbiotic system. High levels of 15 N-labelling were observed in the dinoflagellates and in foraminiferal organelles and cytoplasm after incubation with 15 N-ammonium, indicating efficient ammonium assimilation. Only weak 15 N-assimilation was observed after incubation with 15 N-nitrate. Feeding foraminifers with 13 C-and 15 N-labelled food resulted in dinoflagellates that were labelled with 15 N, thereby confirming the transfer of 15 N-compounds from the digestive vacuoles of the foraminifer to the symbiotic dinoflagellates, likely through recycling of ammonium. These observations are important for N isotope-based palaeoceanographic reconstructions, as they show that δ 15 N values recorded in the organic matrix in symbiotic species likely reflect ammonium recycling rather than alternative N sources, such as nitrates.
carbon requirements for the foraminifera host cell and symbiont growth [3]. In oligotrophic regions, dissolved inorganic nitrogen (e.g. NH þ 4 or NO 2À 3 ) is scarce and often limits biological productivity [4]. Bulk δ 15 N data from cultured foraminifera, measured after 15 N-spiked experiments, showed N-compound translocation from the endosymbionts to the foraminifera host cell [5]. When nitrate is not limiting in the environment, model calculations revealed that up to 57% of the foraminiferal N could be translocated from the symbionts, and even up to 90-100% for model calculations taking into account nitrogen uptake by the symbionts from the recycled ammonium pool. However, several studies have concluded that symbiont-bearing foraminifera must meet their nitrogen requirements via capture and assimilation of zooplankton prey N, because the concentrations and diffusion rates of dissolved ammonium or nitrate are insufficient to explain their growth rate. In this regard, δ 13 C values measured on different amino acids from the symbiotic species Orbulina universa are consistent with two isotopically distinct sources of carbon and nitrogen, isotopically heavy metabolic carbon and nitrogen from its symbionts and relatively lighter carbon and nitrogen from the diet [6,7]. It is also likely that planktonic foraminifer can assimilate inorganic nitrogen from seawater if it is available [2,5,8] because ammonium assimilation has already been reported in a benthic species with inactive kleptoplasts [9].
Determining the source(s) of nitrogen for foraminifera associations is particularly important when interpreting planktonic foraminifera 15 N/ 14 N data from fossil shells [10]. In these shells, foraminifera-bound δ 15 N (FB-δ 15 N) was measured by extracting organic matrix protein N that is tightly sequestered within the foraminiferal calcite chamber walls. Ren et al. [11,12] demonstrated that the dinoflagellate bearing foraminifers, O. universa, Trilobus sacculifer, and Globigerinoides ruber record similar FB-δ 15 N values in modern ocean and recent sediments from the past 40 ka, and their FB-δ 15 N values are similar to the δ 15 N of nitrate sources. By contrast, in the co-occurring species Globigerinella siphonifera (=G. aequilateralis), which hosts small coccoid pelagophyte symbionts [13], FB-δ 15 N values were systematically 1-2‰ higher compared with the dinoflagellate-bearing species even though the four species appear to feed on similar zooplankton prey [14]. A clearer understanding of the nitrogen uptake and protein synthesis pathways will help determine whether such offsets represent trophic effects or species differences in symbiont-host physiology.
Inorganic carbon assimilation by the symbiotic dinoflagellates of O. universa has been quantified in laboratory experiments using microelectrode techniques [3,8]. More recently, LeKieffre et al. [15] used a combination of transmission electron microscopy (TEM) and nanoscale secondary ion mass spectrometry (NanoSIMS) quantitative imaging to document inorganic carbon flow from symbiont photosynthetic assimilation through lipid translocation to the foraminifera cell in a time-series experiment.
Here, we present results from a second TEM-NanoSIMS time-series experiment using seawater with 15 N-enriched NH 4 , NO 3 , and 13 C and 15 N-enriched food (Artemia, brine shrimp) to explore the different nitrogen assimilation pathways in the O. universa-symbiont system. The combination of TEM and Nano-SIMS was used in combination with stable isotope incubations to follow the assimilation of carbon and nitrogen from different sources within the foraminiferal cell and its symbiotic microalgae across a full day/night cycle. The main objective of this study was to visualize and quantify nitrogen assimilation and translocation between dinoflagellate symbionts and O. universa host cell when provided with different nitrogen sources to better understand the nitrogen flow dynamics in symbiotic associations.

Material and methods (a) Sampling
All specimens of Orbulina universa used in the experiments were hand-collected by scuba divers from surface waters (2-10 m depth), approximately 1-2 km offshore Santa Catalina Island (California, USA) in the San Pedro basin on 25 August 2015. Individuals were collected in glass jars (120 ml) and transported within an hour to the University of Southern California, Wrigley Marine Science Center. In the laboratory, each individual was measured, checked for vitality (i.e. chambers filled with cytoplasm), and for the presence of symbiotic dinoflagellates and/ or food particles in the spines (see electronic supplementary material, table S1), and then transferred to a new glass jar filled with filtered (0.45 µm) seawater. Both 'juvenile' (presphere, trochospiral shell) and 'adult' (specimens with a final spherical chamber) specimens were used in order to obtain sufficient replicates for the experimental matrix (see electronic supplementary material, table S1). Except for the feeding experiment (Exp. 3), all foraminifera were fed an unlabelled one-day-old hatched Artemia nauplius at 16.00 on the day the specimens were collected to maximize their survival rate during the experiment phase. This includes specimens that were collected with food particles in their spines.
(b) Isotopic enrichment of Artemia salina 15 N-and 13 C-labelled A. salina used in Exp. 3 were prepared according to the protocol described by Krueger et al. [16] detailed in the electronic supplementary material, text S1.

(c) Experimental design
Three isotopic labelling experiments were initiated the day after collection. In Exp. 1, specimens of O. universa were incubated for 18 h in seawater that was double-spiked with 15 NH þ 4 and H 13 CO À 3 . For Exp. 2, specimens of O. universa were incubated for 18 h in seawater that was double-spiked with 15 NO À 3 and H 13 CO À 3 . In Exp. 3, specimens of O. universa were maintained in ambient seawater and fed small pieces of 15 N-and 13 C-enriched A. salina, and then followed for eight additional hours (figure 1). Experiments 1 and 2 were initiated at 13.00 local time (corresponding to the onset of the period of maximum symbiont photosynthetic rate [2]), which we refer to as t = 0 h. Twenty-four specimens were transferred into 22 ml borosilicate glass scintillation vials (one specimen per vial) which were filled with 0.45 µm filtered seawater spiked by the addition of 2 mM NaH 13 CO 3 (99% 13 C; Cambridge Isotopes Laboratory Inc.) and either 10 µM 15 [17].
For Exp. 3, three specimens were transferred into 22 ml scintillation vials (one specimen per vial) which was filled with 0.45 µm filtered seawater, and fed a small piece of 15 N-and 13 C-enriched A. salina that had been thawed from the freezer at 11.00 local time (t 0 = 0 h for this experiment). The Artemia pieces were gently pipetted onto the O. universa spines at which time the foraminifera transported the food particles to the surface of the shell where it royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 287: 20200620 ingested the food via the production of food vacuoles (see https:// www.youtube.com/watch?v=BYQNt52tiVU). The three specimens used in this experiment trapped the food particles in their spines within a few minutes.
Following specimen transfer into labelled solutions or following labelled Artemia feeding, the vials were capped and immersed in water baths at 22°C under artificial light (Sylvania F24T12 'Cool White' fluorescent lights). Irradiance levels exceeded 350 µmol photons m −2 s −1 , which is the P max light saturation threshold for symbiont photosynthesis in this species [2,18]. Specimens were maintained on a 12/12 h light/dark cycle, similar to natural conditions, with lights off at 19.00 local time (dark period), and lights back on at 07.00 the next day. At pre-defined sampling times (t = 1 h, 6 h, 12 h, or 18 h for experiments 1 and 2, and t = 8 h for Exp. 3 (figure 1)), three specimens were moved into vials for TEM fixation. Three additional individuals were maintained in unspiked filtered seawater and fixed before the experiments to serve as unlabelled control specimens.

(d) Chemical fixation and TEM observation
At each sampling time point, selected O. universa specimens were individually transferred into 0.5 ml micro-centrifuge tubes and processed exactly as described by LeKieffre et al. [15] (see electronic supplementary material, text S2 for details).

(e) NanoSIMS analysis
NanoSIMS settings and analysis procedures follow that described by LeKieffre et al. [15,19] (see electronic supplementary material, text S2 for details), except that both 13/12 C and 15/14 N ratios were measured. Quantified 13 C/ 12 C and 15 N/ 14 N ratios were obtained as follows: NanoSIMS image processing was carried out as described in LeKieffre et al. [15,19] using the software LOOK@NANOSIMS [20] (see electronic supplementary material, text S2 for details). Isotopic enrichments were calculated for the foraminifer cytoplasm, dinoflagellate, and cellular compartments. For each condition, time point, and type of cellular compartments, the average 13 C-and 15 N-enrichments were calculated based on three replicates (unless otherwise specified).

(f ) Statistics
Statistical analysis was performed as described in [15] (see electronic supplementary material, text S3 for details). Briefly, statistical analysis was carried out on the total set of ROIs (regions of interest) for each time point using a linear mixed-effect model taking into account pseudo-replication effects, followed by a Tukey multiple comparison test. The results of the linear mixedeffect model are listed in the electronic supplementary material, text S4. Comparisons of δ 13 C of the different cell compartments between Exp. 1 and Exp. 2 were performed with t-tests for each cell compartment and each time point. The results of the t-tests are listed in the electronic supplementary material, table S2. All statistical analysis were performed with the RSTUDIO software [21] with the significance level set to α = 0.05 (i.e. p-value < 0.05).

(a) Ultrastructure observations
All O. universa observed in this study exhibit the same ultrastructural features described previously [15,[22][23][24]. Specimens have symbiotic dinoflagellates containing a nucleus with condensed chromosomes, two peripheral chloroplasts with stalked pyrenoids surrounded by starch, and accumulated starch grains within the cell cytosol for symbiotic dinoflagellates sampled at the end of the light phase (t = 6 h), whereas vacuolarized dinoflagellates in the foraminifer endoplasm at the end of the dark phase (t = 18 h) were nearly starch-free (figure 2 and electronic supplementary material, figures S1 and S2). Accumulation of 13 C-labelled C is first seen in the dinoflagellate starch surrounding the chloroplast pyrenoids (electronic supplementary material, figure S2) after 1 h of incubation. After 6 h of incubation (at the end of the light phase), numerous 13 C-labelled starch grains can be seen in the dinoflagellate symbiont cells, with δ 13 C values of approximately 10 000-40 000‰. At this time, 13 C-enrichment is also observed in organelles throughout the dinoflagellate cell. The 13 C-enrichment levels remain stable during the first 6 h of the dark phase (between t = 6 and t = 12 h; figure 3). By the end of the dark phase at t = 18 h, the δ 13  In the foraminiferal cell, a weak 13 C-enrichment can be seen in the foraminiferal lipid droplets starting at t = 1 h (figure 3 and electronic supplementary material, figure S2). After 6 h in light, the foraminiferal endoplasm exhibits 13 C-labelled hotspots, mostly in lipid droplets, electron-opaque bodies, and the fibrillar bodies. δ 13 C values approach 1300‰, 800‰, and 400‰, respectively. NanoSIMS images taken on specimens from the dark phase at t = 12 h reveal a greater proportion of 13 C-enriched lipid droplets than during the light phase of the experiment (t = 1-6 h). Here, lipid droplets reach values between 1500 and 2500‰. It is noteworthy that dinoflagellate lipids and the lipid droplets observed along their periphery or   figure S3). The electron-opaque body (small electron-dense circular vesicle of approx. 500 nm diameter) δ 13 C remain stable from t = 6 to 18 h (through the dark phase). By contrast, the 13 C-enrichment of the foraminiferal cytoplasm increases significantly between t = 6 and 18 h, reaching approximately 850‰. Protein-rich fibrillar bodies [25,26] are found throughout the cytoplasm (electronic supplementary material, figure S2), and a 13 C-enrichment in some, but not all of them could also be detected after t = 6 h with values ranging between 300 and 500‰ (figure 3). In fibrillar body structures displaying 13 C uptake, the 13 C-enrichment appears to increase during the first period of the dark phase (t = 6-12 h), and then remains stable between t = 12 and 18 h with values ranging from 700 to 1200‰.

(c) Inorganic 15 N-assimilation
NanoSIMS data reveal a significantly greater 15  N-enrichment is not observed in the dinoflagellates until after 6 h of exposure to 15     . Measurements were made for symbiotic dinoflagellates, foraminiferal cytoplasm, fibrillar bodies, electron-opaque bodies, and lipid droplets. Error bars represent one standard deviation (n = 3, except values marked with an asterisk: n = 2). Baseline δ 13 C and δ 15 N values for the NanoSIMS control specimens that were grown in unlabelled seawater are δ 13 C = 0 ± 50‰ and δ 15 N = 0 ± 51‰, (n = 2; ± 3σ).
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 287: 20200620 Artemia fragments (Exp. 3) is filled with numerous digestive vacuoles (figure 4) that are characterized by an ovoid to circular shape, and a diameter ranging from 1 to 5 µm. These digestive vacuoles display both 13 C-and 15 N-labelling (figure 4), confirming that their content is 13 C-and 15 N-enriched fragments of Artemia; the only source of 13 1 and figure 4). However, their 13 Cenrichments were similar to the unlabelled control specimens.

Discussion (a) Carbon assimilation by the symbionts and translocation to the host cell
The dynamics of photosynthetic C-fixation and starch accumulation by symbiotic dinoflagellates and subsequent transfer of organic C to the O. universa cell were described in a previous NanoSIMS pulse-chase experiment [15]. These authors reported that within 45 min, 13 C-labelled starch accumulated around the chloroplast pyrenoids and by the end of the 6-h pulse phase, numerous 13 C-labelled starch granules could be observed within the dinoflagellate cytosol. During the dark phase, 13 C-labelled starch was converted to lipid in the symbiont cell and then transferred to the foraminifera cytoplasm, where it was assimilated and distributed throughout the cytoplasm. The present work supports these previous observations (figure 3 and electronic supplementary material, figure S2) and the pathway for carbon translocation from symbionts to planktonic foraminifera, which has been estimated to contribute about 40% of the total C budget of the host-symbiont complex [1]. In other symbiotic associations involving dinoflagellates, carbon translocation to the host occurs through the movement of compounds such as glycerol, glucose, lipids, organic acids, or amino acids [27][28][29][30]. However, the study of the holobiont transcriptome in symbiotic radiolarians, another marine planktonic protozoan, revealed that most of the microalgal genes involved in glucose, fatty acids, or glycerol production were not upregulated, suggesting that amino acids might be the major form of C translocation from symbionts to the host [31]. In the present study, small lipid droplets are seen in the foraminiferal cytoplasm in direct contact with the symbiosome, in between the symbiosome and the dinoflagellate membrane, and within the dinoflagellate cytosol during the dark phase (electronic supplementary material, figure S3). These observations further support our hypothesis that lipids are an important vehicle for C transfer across the dinoflagellate and symbiosome membrane in planktonic foraminifera, but does not exclude the possibility that a concomitant transfer of C also occurs in soluble form, e.g. as amino acids. The slightly higher 13 C-enrichment observed after 1 h in the dinoflagellates incubated in Exp. 2 compared to those from Exp. 1 could result from different metabolic pathways for metabolite release. Dinoflagellates incubated in Exp. 1 had access to excess ammonium, while those incubated in Exp. 2 only had access to nitrate, which they did not assimilate (see further in the Discussion). When deprived of a usable nitrate source, dinoflagellates switch their carbon metabolism to more lipid production rather than N-containing compounds (e.g. amino acids and therefore proteins, nucleic acids) [32]. A similar pathway in the O. universa symbiotic dinoflagellates could explain the higher 13 C-enrichment observed in symbionts from Exp. 2 (accumulation of the 13 C in their starch granules and/or lipid droplets) while in Exp. 1, more soluble N-compounds (e.g. amino acids) could have been produced and lost during sample preparation.
Note that the 13 C-enrichment differences observed at t = 1 h for the lipid droplets between the Exp. 1 and the Exp. 2 specimens could be due to high inter-individual variability, as one of the three specimens incubated in Exp. 1 exhibited much higher 13 C-values than the two other specimens. Therefore, the analysis of a larger number of specimens is recommended for future experiments.

(b) Inorganic nitrogen assimilation and exchanges between the symbionts and the host cell
One of the robust observations from this study is that the symbiotic dinoflagellates display a clear preference for ammonium over nitrate assimilation. Ammonium-15 N accumulation was up to 40 times higher than nitrate accumulation after 18 h of incubation (figures 2 and 3). The 15 N-enrichment in all the foraminiferal compartments investigated (cytoplasm, electron-opaque, and fibrillar bodies) was much higher in Exp. 1 ( 15 NH þ 4 ) compared to Exp. 2 ( 15 NO À 3 ). In Exp. 1, ammonium assimilation into the dinoflagellate nucleus and chloroplasts as well as the protein-rich foraminiferal fibrillar bodies (figure 2) was visible already after 1 h of incubation. By t = 6 h, the 15 Ntracer was distributed throughout the foraminiferal cytoplasm, but was more concentrated in the symbiont and foraminifera organelles associated with protein and nucleotide synthesis, in addition to small electron-opaque bodies (figures 2 and 3). In all the cell compartments investigated including the dinoflagellates, the 15 N concentration increased throughout the duration of the experiment (light and dark phases). It is noteworthy that, contrary to inorganic carbon where only the photosynthetic dinoflagellates are capable of C-assimilation [15], ammonium could be assimilated directly by the foraminiferal cell. Ammonium assimilation via a cytoplasmic pathway was suggested for benthic kleptoplast (i.e. stolen chloroplasts) species from aphotic habitats, based on the observation of ammonium assimilation in the dark while the kleptoplasts were photosynthetically inactive [9]. Thus, the increase of 15 N-enrichment in the foraminiferal compartments observed over the 18 h incubation likely results from a combination of a transfer of 15 N-compounds from the dinoflagellates to the foraminiferal host cell and assimilation of 15 N-ammonium by one or more foraminiferal cytoplasmic pathways.
Generally, photosynthetic organisms such as algae display a tight coupling of inorganic C and N uptake, with photosynthesis providing the carbon skeleton necessary for N assimilation [33]. In our experiment, δ 15 N increased in the dinoflagellate cytoplasm during both the light and dark phases, the foraminifera symbionts are capable of assimilating ammonium N throughout the entire diurnal cycle examined. Dark assimilation of ammonium has been reported previously in different dinoflagellate species [32], and dinoflagellate symbionts in corals assimilate nitrate in the dark even after corals were acclimated to darkness for 24 h [34]. These observations demonstrate the ability of symbiotic dinoflagellates to royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 287: 20200620 assimilate N during the night in the absence of carbon skeletons supplied by photosynthetic processes. Although corals store nitrogen in highly N-concentrated uric acid crystals within the dinoflagellate cells, [34], we have not observed such crystal structures within the dinoflagellate cytoplasm in O. universa (figure 3 and electronic supplementary material, figure S4). Nevertheless, we note that dinoflagellates are known to be able to store large amounts of N within their cells, potentially in the form of free amino acids [32], which in our case would have been lost during sample preparation.

(c) Nitrate assimilation inhibited by ammonium in symbiotic microalgae
Orbulina universa has a wide habitat range, ranging from the tropics through the oligotrophic gyres to the subtropical convergences and equatorial and coastal upwelling zones. This species complex has the broadest temperature tolerance among extant planktonic foraminifera species with a range from 9 to 30°C [35,36]. In oligotrophic regions, ammonium is generally scarce but is thought to serve as the main source of N for phytoplankton because of its high turnover rate [37][38][39]. Ammonium assimilation results in a better energy yield than nitrate assimilation [40] and thus nitrate uptake is inhibited with increasing ammonium concentration in photosynthetic organisms [41][42][43], including dinoflagellates [44]. Higher assimilation of ammonium over nitrate has been observed in other photosynthetic organisms such as cyanobacteria (Prochlorococcus and Synechococcus) [45]. The gene expression involved in N assimilation depends on the primary form of N that is available to the photosynthetic organisms [46][47][48]. In Thalassicolla nucleata, a planktonic radiolarian found in association with the photosynthetic symbiotic dinoflagellate Brandtodinium sp., holobiont transcriptome analysis revealed a dramatic downregulation of the genes involved in nitrate reduction while the genes involved in ammonium assimilation were all upregulated [31].
The NO 3 half-inhibition concentration of NH 4 was estimated for two dinoflagellate species at around 1.3 μM [49],  so we assume that the natural NH 4 concentration measured in the surface waters of San Pedro basin (0.2 µM) is not sufficient to inhibit NO 3 assimilation by the symbiotic dinoflagellates. Rather, we suggest that in the O. universa symbiotic system, the host foraminifer provides sufficient recycled ammonium to its dinoflagellates such that similar gene expression regulation likely occurs which would result in the inhibition of the nitrate assimilation pathway. Indeed, in our experiments, all the foraminifera were fed the day before the experiment, and the feeding experiment (Exp. 3) showed that prey were still being digested 6 h after the feeding event. Thus, it is likely that recycled ammonium was still available to the symbionts and host in the foraminifera incubated in Experiment 2 ( 15 NO À 3 ). We also cannot exclude the possibility that the foraminifera might be able to excrete ammonium from N-stored pools. Nevertheless, the ability of the O. universa dinoflagellates to efficiently use ammonium N whenever available is a powerful evolutionary adaptation for this protist association to tightly recycle bioavailable nitrogen in oligotrophic environments, regardless of its source from food or dissolved inorganic N.
(d) Nitrate assimilation via feeding on prokaryotes?
Despite the fact that O. universa was incubated in filtered (0.45 µm) ambient seawater, we could not avoid incidental bacterial contamination or prevent prokaryotes from being introduced into the culture vessel, if they were not already present on the foraminiferal spines and rhizopodia. Indeed, in Exp. 2 ( 15 NO À 3 ), the highly 15 N-enriched vesicles we observe in the O. universa cytoplasm (electronic supplementary material, figure S5) have a size range between 0.5 and 1 µm and could therefore be prokaryotes. As these vesicles were in various states of degradation and always enclosed within degradation vacuoles, it is unlikely that they are endosymbionts. These prokaryote-like vesicles could have assimilated 15 NO À 3 from the seawater while residing outside the foraminiferal cytoplasm in the spines or rhizopodia. In this case, they could have been ingested by the foraminifer after the experiment began. The digestion of these 15 N-enriched prokaryote-like vesicles could result in the production of recycled 15 N-compounds that, in turn, were either directly transferred to the dinoflagellate, or excreted in their microenvironment and reacquired by the symbionts. This process does not necessary involve a remineralization step into ammonium as symbiotic dinoflagellates are also able to assimilate organic nitrogen (dissolved amino acids, urea) [30,34,50]. Thus, we suggest that the weak 15 N-nitrate assimilation observed in the foraminiferal cell and its dinoflagellates could also be the result of digestion of these 15 N-labelled vesicles ( potential prokaryotes) rather than exclusively through direct nitrate uptake by the symbiotic dinoflagellates.
(e) Nitrogen assimilation derived from prey capture by the foraminiferal host The two foraminifera fed with 13 C-and 15 N-labelled Artemia (Exp. 3) contain cytoplasm filled with digestive vacuoles (figure 4). The lower 13 C-and 15 N enrichment values observed for specimen 153 could be due to a lower enrichment of the 13 C-and 15 N-labelled Artemia particle it was fed. Indeed, 13 Cand 15 N-enrichments were measured on a lyophilized aliquot, but did not take into account enrichment differences between individuals or tissue components [16]. Regardless, in both specimens the symbiotic dinoflagellates were found labelled with 15 N after 8 h. Since all the 13 C-and 15 N-labelling in this experiment must come from prey digestion, the observation of 15 N-enriched dinoflagellates in the foraminiferal cytoplasm attests to the transfer of 15 N-compounds between the digestive vacuoles and the symbiotic dinoflagellates. This transfer could happen via a direct intracellular transfer of N-compounds from the foraminifer to the dinoflagellate, or more likely, via symbiont assimilation of recycled ammonium excreted into the extracellular microenvironment of the foraminifera spines since the symbionts were primarily on the spines and not in vacuoles within the foraminifera cytoplasm during this experiment. Additional experiments are required to more completely document the diurnal dynamics of foraminiferal food-derived N assimilation, ammonium remineralization, and subsequent excretion/uptake by the symbiotic dinoflagellates. However, we show here that symbiotic dinoflagellates do use recycled N derived from food digestion by the host. These observations confirm the model proposed by Uhle et al. [5] that most of the nitrogen used by the O. universa symbionts comes from recycled ammonium excreted by the foraminifera, rather than direct nitrate assimilation by the dinoflagellates. This mechanism is likely used by other species of dinoflagellate bearing planktonic foraminifera and could be common among other symbiotic systems as it was also observed in other symbioses (e.g. cnidarians, sponges; e.g. [51][52][53]). In coral symbiosis, studies estimated that symbiotic dinoflagellates could meet up to 90% of their N requirements through N recycling [52,53].

(f ) Fibrillar bodies: protein storage organelles for biomineralization
The results from Exp. 1 (figures 2 and 3) show that the foraminifera organelles most actively synthesizing proteins during the day are the fibrillar bodies that make up the fibrillar system [54]. These enigmatic organelles are only found in planktonic foraminifera and were first described by Rhumbler [55] who called them 'Gallertstränge' (meaning gelatinous strands). Later, light [25] and TEM [26,56] examination of planktonic foraminifera ultrastructure reported that the fibrillar bodies were composed of strands of proteins that were located primarily in the final and penultimate chambers. Hansen [26] first speculated that the fibrillar system could be linked to foraminifer buoyancy although no evidence was presented to support this hypothesis. Subsequent TEM examination of Orbulina universa ultrastructure during chamber formation [57] documented the presence of fibrillar material in the extracellular environment adjacent to a newly forming chamber, suggesting the fibrillar material is linked to biomineralization during initial chamber formation rather than serving a buoyancy function. Active uptake of 15 NH 4 by the fibrillar bodies in Exp. 1 is intriguing in that the rapid synthesis and accumulation of protein during the day must serve a fundamental and time-sensitive role in foraminifera physiology for individuals to expend so much energy in protein synthesis. Electronic supplementary material, figure S1 also shows that the fibrillar bodies are incorporating 13 C on a timescale similar to that of 15 N. Chamber formation in planktonic foraminifera is an episodic event that involves the rapid transport and organization of a polysaccharide-protein-rich organic matrix outside the foraminifera endoplasm [58]. The entire process from initial organic matrix extrusion through initial chamber wall biomineralization occurs in 1-2 h in O. universa [57]. Hence, the royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 287: 20200620 biomineralizing proteins and associated polysaccharides needed for chamber formation must be synthesized and stored within the foraminifera cell prior to a biomineralization event in order for chamber formation to occur so rapidly. We suggest that the fibrillar system is the synthesis and storage organelle for the organic matrix proteins/ polysaccharides prior to chamber biomineralization.

Conclusion
Our data show that ammonium is the major source of N used for protein and nuclear synthesis by the symbiotic dinoflagellates in the planktonic foraminifera Orbulina universa. The observation of 15 N-labelled dinoflagellate cells after feeding the foraminifera with 15 N-Artemia shows that symbiotic dinoflagellates use recycled ammonium (or other N-compounds). As O. universa typically inhabits oligotrophic waters where ammonium is scarce, recycled ammonium provided by the host is likely a key source of nitrogen for protein and nucleotide synthesis in the symbiotic dinoflagellates. These findings are of major interest for the interpretation of foraminiferabound 15 N/ 14 N ratios from fossil shells (FB-δ 15 N), as recycled ammonium does not necessarily have the same isotopic fractionation that other nitrogen sources exhibit such as nitrate.
Considering an isotopic balance within the symbiont-host system, complete uptake of recycled ammonium requires the recycled ammonium to have the same 15 N/ 14 N ratios as the external N source, e.g. the prey, yet any ammonium excretion due to incomplete uptake will cause an isotopic difference between recycled ammonium and the organic matter produced. Our evidence thus raises an interesting mechanism for interpreting the observed species FB-δ 15 N difference in fossil records. Specifically, it is possible that the non-symbiotic foraminifera, or foraminifera species bearing non-dinoflagellate symbionts, may exhibit more elevated δ 15 N in their biomass and shell-bound organic N, because of the lack of efficient N recycling processes observed between dinoflagellate symbionts and hosts.
Data accessibility. The datasets generated during and/or analysed during the current study are available in the Zenodo repository (doi:10. 5281/zenodo.3834220).