Tissue- and cellular-level allocation of autotrophic and heterotrophic nutrients in the coral symbiosis – A NanoSIMS study

Corals access inorganic seawater nutrients through their autotrophic endosymbiotic dinoflagellates, but also capture planktonic prey through heterotrophic feeding. Correlating NanoSIMS and TEM imaging, we visualize and quantify the subcellular fate of autotrophic and heterotrophic C and N in the coral Stylophora pistillata using stable isotopes. Six scenarios were compared after 6h: autotrophic pulse (13C-bicarbonate, 15N-nitrate) in either unfed or regularly fed corals, and heterotrophic pulse (13C-, 15N-labelled brine shrimps) in regularly fed corals; each at ambient and elevated temperature. Host assimilation of photosynthates was similar under fed and unfed conditions, but symbionts assimilated 10% more C in fed corals. Photoautotrophic C was primarily channelled into host lipid bodies, whereas heterotrophic C and N were generally co-allocated to the tissue. Food-derived label was detected in some subcellular structures associated with the remobilisation of host lipid stores. While heterotrophic input generally exceeded autotrophic input, it was more negatively affected by elevated temperature. The reduced input from both feeding modes at elevated temperature was accompanied by a shift in the partitioning of C and N, benefiting epidermis and symbiont. This study provides a unique view on the nutrient partitioning in corals and highlights the tight connection of nutrient fluxes in symbiotic partners.


34
The immense biodiversity and productivity of tropical coral reefs exists in a marine environment that 35 is generally characterized by low nutrient levels and plankton concentrations. Coined as "Darwin's 36 paradox", the investigation into why corals can thrive in areas with high wave action and low nutrients 37 has directed the scientific focus towards nutrient fluxes and cycles within the overall reef framework 38 and within individual hermatypic corals (1, 2). The coral animal has access to the inorganic nutrient 39 pool in the surrounding seawater through endosymbiotic photosynthesizing microalgae (mainly 40 Symbiodinium sp.) and via parts of its microbial community, such as cyanobacteria and diazotrophic 41 bacteria (3,4). Intracellular Symbiodinium sp. cells photosynthesize in the light and fix CO 2 into triose 42 phosphate compounds that can serve as C skeletons for subsequent processes, such as inorganic N 43 assimilation (5). The CO 2 -fixation rate in Symbiodinium is very high and C is rapidly assimilated into 44 sugars and amino acids within seconds to minutes (6, 7). The translocation of part of these soluble 45 photosynthates from the symbiont to the coral tissue happens across the entire colony surface, 46 depending on local symbiont density. While it has been demonstrated that coral autotrophy can be 47 sufficient to meet the metabolic C demands of the coral animal (8,9), coral heterotrophy (e.g. hunting 48 for prey) is also an important component of coral nutrition, especially in cryptic and mesophotic 49 habitats or turbid water environments, where light becomes a limiting factor (10-12). Upon loss of 50 autotrophic input, as is the case in bleached corals, heterotrophy becomes fundamental for survival and 51 recovery of the animal (13,14). Under normal, healthy conditions, the relative contribution of 52 autotrophy and heterotrophy to coral nutrition is modulated by the environment and may vary between 53 species (15). Importantly, the intracellular microalgae, in addition to assimilating inorganic nutrients 54 from the seawater, also allow for a reassimilation of metabolic 'waste' products of the animal host (e.g. 55 NH 4 + and CO 2 ) (16). Hence, the symbiotic nature of the coral animal ensures both access to organic 56 and inorganic nutrients and recycling and conservation of acquired nutrients in an oligotrophic 57 environment. 58 Despite the autotrophic capability of scleractinian corals, which originated in the middle to late 59 Triassic (17)(18)(19), reef-building symbiotic corals have retained highly developed feeding mechanisms 60 connective coenenchyme tissue and post-fixed in 1% [v/v] osmium tetroxide in water for 1 h. We 165 exclusively analysed the coenenchyme tissue between polyps in order to avoid data artefacts due to the 166 individual polyps catching different amounts of prey within the pulse period. All used histological 167 terminology follows Peters (22). Post-fixed samples were dehydrated, embedded in Spur resin blocks, 168 microtomed and post-stained for TEM, following established protocols (7). Transmission electron 169 microscopy (TEM) was carried out at 80 or 100 kV with a Philips CM 100 and a Tecnai 12 (FEI) 170 transmission electron microscope, respectively. While ultrathin sections (70 nm) were mounted on 171 carbon formvar-coated Finder copper grids and imaged for correlative TEM and NanoSIMS, semi-thin 172 sections (500 nm) were mounted on glass slides and directly imaged with NanoSIMS in order to cover 173 larger areas and to acquire sufficient data for statistical analysis. 174 Suitable areas were analysed for 13 C-and 15 N-enrichment using a NanoSIMS 50L ion microprobe. 175 Sections were gold-coated and bombarded with a 16 keV primary Cs + ion beam. Raster scans of the 176 area of interests were performed with a beam focus spot size of around 150 nm (6 layers). Image 177 resolution was set to 256x256 pixels with a beam dwell time of 5 ms per pixel. Image size for semi-178 thin sections was set to 40x40 µm, which provided a complete cross section of the surface body wall 179 epidermis and gastrodermis with sufficient resolution for subcellular features to be clearly resolved. 180 Image sizes for correlative TEM and NanoSIMS varied. For semi-thin sections, approximately 10-15 181 images were obtained per replicate in order to cover enough area to obtain enrichment values for 40-182 60 symbiont cells and their surrounding host tissue. In case multiple sections from the same sample 183 block were required to obtain enough cells, sections were taken at least 20 µm apart to avoid cutting 184 and imaging another layer of the same symbiont cells. The secondary ions 12 C 2 -(mass 24), 13 C 12 C -185 (mass 25), 12 C 14 N -(mass 26), and 12 C 15 N -(mass 27) where simultaneously counted in electron 186 multipliers at a mass resolution of about 10000, enough to resolve all potential interferences. 187

Data treatment and normalisation 188
NanoSIMS data were processed using the L'IMAGE software (created by Dr. Larry Nittler, Carnegie 189 Institution of Washington). Maps of 13 C/ 12 C and 15 N/ 14 N ratio distributions were derived from drift-190 corrected image ratios of 13 C 12 Cand 12 C 2 -, and 12 C 15 Nand 12 C 14 N -, respectively. TEM and 191 corresponding NanoSIMS images were used to visually identify enriched subcellular structures. Note 192 that shown TEM images were adjusted for contrast. In order to permit easy comparison between 193 different NanoSIMS images, the enrichment levels in 13 C and 15 N for all presented images are 194 expressed using the δ-notation, which expresses the relative difference between the measured isotope 195 ratio in a labelled tissue (R sample ) and the measured natural isotope ratio in the same tissue type 196 (R reference ) in parts per thousand: 197 Regions of interest (ROIs) were drawn on ratio images to assess average enrichment in each type of 199 compartment. The following ROIs were defined and manually drawn on each image based on contour 200 outlines visible on the 12 C 14 Nimage and/or the isotopic ratio images: Symbionts (cross-sectional 201 diameter >3 µm), extra-algal lipid bodies (outside of symbionts, but contained within the 202 symbiosome), host gastrodermis (excluding symbionts), host epidermis, and host lipid bodies 203 (diameter >1 µm). Each NanoSIMS image thus provided data points for multiple symbiont and lipid 204 ROIs and one data point for gastrodermis and epidermis, respectively. All NanoSIMS images were 205 slightly smoothed (smooth width of 3 pixels). 206 In order to be able to accurately contrast the input from autotrophic and heterotrophic nutrients, the 207 measured enrichment in each ROI was normalized to the isotopic labelelling of the respective source 208 (seawater or brine shrimp) as outlined in the following. For subsequent calculations, the measured 209 isotopic ratios were converted into 13 C and 15 N atom fractions, F (in atom percent, AP, when expressed 210 as F x 100): 211 The atom ratios for the resin embedded heterotrophy food source were determined with F Het_C = 213 0.04772±0.01155 and F Het_N = 0.33772±0.01360 (N=6). Atom fractions of 13 C and 15 N in the spiked 214 seawater (F SW_C and F SW_N ) were calculated taking into account the relative concentrations and atom 215 fractions of nitrate and C in seawater and the added spike. Assuming standard atom fractions for C 216 (F PDB : 0.01111) and N (F N : 0.00366) (44) in the natural seawater and ambient concentrations of 2.2 217 mM total dissolved inorganic C and 0.3 µM nitrate (40), the following equations express the final 218 seawater atom fractions in 1 L of seawater after the addition of the spike (F SW ). Atom fractions for the 219 added spike were 0.98 for both NaH 13 CO 3 and K 15 NO 3 , as indicated by the supplier (Sigma-Aldrich, St 220 Louis, MO, USA). 221 #$ % = (' () *+% , -.*/ )"(' (.+12 *+% , -(.+12 *+% ) ' () *+% "' (.+12 *+% = (3.3 , 5.5 )"(3.5 , 5.67) Using the measured atom fraction from labelled and unlabelled Artemia in resin, the corrected APE for 227 the heterotrophic coral samples was calculated for each element as: 228 In essence, these APE for each type of ROI express the proportion of the total C or N pool that has 230 been replaced with newly assimilated C and N from either food source. Assuming steady state 231 conditions of the C and N pool, i.e. assuming a negligible change in total biomass in the imaged 232 coenenchyme areas within the pulse period, APE is a measure of the relative C and N turnover in each 233 ROI over 6 h in the light. In general, it should be noted that conventional preparation of biological 234 samples for NanoSIMS involves a loss of most soluble compounds and restricts observations mainly 235 to the anabolic part of metabolism. Thus, the presented values should be interpreted as minimal 236 estimates for tissue C and N assimilation. 237 Since APE only represent the relative turnover of a compartment without considering the actual size of 238 this compartments, a semi-quantitative measure of C and N assimilation was calculated by multiplying 239 size of compartment (cross-sectional area A, in µm 2 ) with relative turnover in this compartment (APE 240 in %) for each element. This "integrated area fraction" (IAF) for each type of compartment allows an 241 assessment of relative C and N partitioning across the main three tissue compartments (sym: symbiont, 242 gastr: gastrodermis, epi: epidermis) of the surface body wall within the 6 h pulse, where the total 243 assimilated C or N pool is composed of: 244 Note, that IAF is a semi-quantitative measure based on the reasonable assumption that the C/N 246 elemental ratio in these compartments is similar, consistent with NanoSIMS images (data not shown). 247 For a fully quantitative comparison, one would have to determine the total C-and N-pool in each 248 compartment separately. Note that IAFs and the resulting partitioning budget only considers the 249 structurally incorporated C and N; i.e. it does not consider losses due to mucus shedding or respiration. 250

Statistical analysis 251
Considering that ROI enrichment values contain two sources of variability, resulting from the cutting 252 level within the cell (affecting for example the abundance of starch granules in a particular symbiont 253 ROI) and the true biological variability as result of size, age and physiological state of individual cells, 254 we decided to treat the ROI data points as raw data points and include biological replicate as a factor 255 in the analysis, rather than treating multiple ROIs from one biological replicate as technical replicates 256 and reducing them to a single value. Thus, we sought to balance the number of data points for the 257 same types of ROI between the three biological replicates for statistical considerations. Note that the 258 published unfed autotrophy dataset from Krueger,Horwitz (39) was reused here and contrasted with 259 other feeding conditions. Depending on the specific question, the three datasets (Aut [U], Aut [F], Het 260 [F]) were tested individually or jointly with multifactorial ANOVAs for each ROI type. In case of 261 insignificant interaction terms, the model was reduced to a minimal adequate model. Significant 262 interaction terms were analysed with Tukey HSD post hoc tests. How regular host feeding affects the 263 autotrophic assimilation at different temperatures in each compartment was tested with a Three-Way 264 ANOVA with temperature, feeding acclimation, and replicate as factors (Aut [U] vs. Aut [F]; Table  265 S1). Secondly, the heterotrophy dataset (Het [F]) was tested for the effect of temperature (Table S2). 266 In the third analysis, we tested whether there is a difference in the tissue-specific C and N contribution 267 from autotrophy and heterotrophy and how elevated temperature interacts with this. This was tested 268 with a Three-way ANOVA for each compartment, with temperature, feeding mode and replicate as 269 Table S3). In addition to the species' mean response, individual responses 270 of the colonies are displayed since there were usually significant differences in the C-and N-turnover 271 between replicates (raw data summary in Table S5). 272

Ultrastructural observations 274
Autotrophically fixed C in the symbionts was mainly concentrated in the primary starch ring around 275 the pyrenoid, in secondary starch granules, in lipid bodies throughout the symbiont cell, and as 276 translocated product in the lipid bodies of the host gastrodermis ( Fig. 1A, B, D, E). In addition, 277 occasionally detected extra-algal lipid droplets, located outside of the algae, but within the 278 symbiosome ( Fig. 1A, B: LB*), showed very high levels of C-turnover (8-11%). Fixed nitrate was 279 incorporated throughout the symbiont cells at high levels, outlining them against the host gastrodermis 280 ( Fig. 1C, F). Occasionally, strongly labelled 15 N hotspots were observed inside these cells. Within the 281 symbionts, the accumulation body was the structure with the lowest levels of photosynthetic C or N 282 incorporation. 283 The host tissue showed different degrees of labelling with autotrophic N and the highest enrichment 284 was usually observed in the immediate vicinity of symbiont cells, where host nuclei and nucleoli 285 showed significant N incorporation (Fig. 1C). The symbiosome membrane complex that engulfs the 286 Symbiodinium cells also incorporated a large amount of fixed nitrate (Fig. 1C, F arrow). 287 298 After 6 h of heterotrophic feeding, 13 C and 15 N derived from the digested brine shrimp was detected in 299 all coral tissue layers, including the gastrodermis of the basal body wall and the calicodermis (Fig. 2, 300 S2A-C). Enrichment was highest in the gastrodermis for both elements, where numerous gastrodermal 301 hotspots (~1-2 µm diameter) expressed strong co-labelling of C and N at both temperatures 302 (ambient/elevated: Spearman's ρ=0.75/0.81, both p < 0.0001, N=490/173). Three distinct types of 303 labelled hotspots could be differentiated in the gastrodermis: (i) round, membrane-delineated vesicles 304 that contained homogenous labelled material (pink arrows, Fig. 2, 3D, 4), (ii) elongated, clearly 305 defined electron-dense structures, that were of cup shape and in direct contact with the lipid bodies 306 (blue arrows, Fig. 2, 3E, 4, 5), and (iii) oval or elongated patches, which were difficult to distinguish 307 from the surrounding tissue apart from being electron-lucent (green arrows, Fig. 2, 3F, 4). With regard 308 to their heterotrophic isotopic signature, vesicles and electron-lucent patches formed a continuum, 309 whereas the cup-shaped structures tended to be more heavily enriched in 15 N relative to 13 C ( Fig. 4D-310 F). 311 A few isotopic hotspots were also found in the host epidermis, representing mostly small vesicles, but 312 with a much lower level of labelling than the gastrodermal vesicles ( Fig. S2D were also detected in the symbiont after 6 h. Here, N-labelling throughout the symbiont cell was 318 homogenous, although occasional small vesicle-like hotspots were observed (Fig. S3). Detectable C-319 labelling was limited to starch granules and the primary starch sheet around the pyrenoid ( Fig. 2A, B, 320 D, E). 321 Average symbiont cell size in hospite across the three replicates was 7.3±0.1 µm (mean±SE, total 322 N=316) at ambient temperature and they occupied 30-40% (N=160) of the gastrodermal cross-323 sectional area. A significant increase in average cell size was detected at elevated temperature, further 324 increasing when combined with feeding (temperature x feeding acclimation: F 1,605 =4.21, p=0.0407*). 325 Average symbiont cell diameter in fed corals at elevated temperature was with 8.

354
The effects of regular feeding and temperature on coral autotrophy 355 The symbiont and the host epidermis were the only structures for which regular feeding showed an 356 effect on the turnover of autotrophic nutrients, specifically C. Symbionts in corals that experienced 357 regular feeding showed significantly increased autotrophic C acquisition by 10±8% at both 358 temperatures (mean±SD, N=6, Fig. 6A, Table S1). Such a positive effect was also seen for epidermal 359 C turnover, albeit much smaller and only at ambient temperatures (+51±57%; N=3; Fig. 6C, Table  360 S1). In general, elevated temperature significantly lowered the turnover of autotrophic C and N in all 361 main coral compartments for both elements, except for N in the host lipid bodies (Table S1, Fig. 6). 362 Structural incorporation of autotrophic C and N at elevated temperature declined on average by 363 19±12% and 5±26% in the symbiont, irrespective of feeding state. Likewise, gastrodermal C and N 364 incorporation declined significantly (-40±29% and -29±36%), with epidermal N turnover dropping on 365 average by -33±16%. Autotrophic C turnover in the host tissue was highest in the host lipid bodies 366 with ~8% C replacement over 6 h in the light, while at the same time displaying a similar N turnover 367 as the overall gastrodermis (Fig. 6D, F, H). For host lipid bodies, elevated temperature only 368 significantly affected C (-12±17%, N=6), but not N turnover. 369  Partitioning of autotrophic C and N between symbionts and the two host tissue compartments of the 387 surface body wall was very similar between regularly fed and unfed corals, and temperature had no 388 substantial effect on this (Fig. 7). After 6 h, approximately 75% and 90% of the detected C and N was 389 located in the symbiont tissue that made up ca. 1/5 th of the cross-sectional area of the surface body 390 wall. The remaining material was mainly located in the gastrodermis and only 1-2% of the autotrophic 391 C and N pool was present in the epidermis, despite representing half of the area of the surface body 392 wall coenenchyme. The disproportional retainment of C and N in the symbionts meant also that 393 allocation of C to the gastrodermal layer was at least twice as high as N allocation (~25% vs. ~10%, 394 Fig. 7). 395

Autotrophic vs. heterotrophic C and N anabolism under ambient and elevated temperature 396
After 6 h of photosynthetic activity, the symbionts replaced on average 5-6% of their structural C 397 (from DIC) and ~2% of their N (from nitrate) under ambient conditions (Fig. 6A, E). Average 398 structural C-and N-turnover in the gastrodermis due to translocated photosynthates was with 1.0-1.6% 399 and 0.1-0.2%, respectively, approximately ten times higher than in the host epidermal layer, which 400 only showed little acquisition of photoautotrophic nutrients (Fig. 6B, C, F, G). 401 Heterotrophic nutrient input considerably exceeded autotrophic input for the gastro-and epidermal 402 compartment, especially for N, where turnover-values where one to two orders of magnitudes higher 403 under ambient temperatures (Fig. 6, Table S3). Heterotrophic N contribution to the symbiont 404 population after 6 h also exceeded its own nitrate fixation by a factor of two to three, while C 405 incorporation was approximately three times lower than the autotrophic input (Fig. 6A, E). In contrast 406 to the autotrophic feeding mode, heterotrophic C was not concentrated in host lipid bodies, which 407 showed lower C enrichment than the overall gastrodermis and received four times less C from 408 heterotrophy compared to autotrophy. 409 Elevated temperature had a distinct effect on the C-and N-metabolism of the coral compartments. 410 Temperature consistently reduced autotrophic and heterotrophic input to all compartments (Fig. 6, 411 Table 1, S1, S2), and these temperature-induced drops in nutrient assimilation were significantly larger 412 for the heterotrophic feeding mode (with the exception of C-assimilation in symbionts and host lipid 413 bodies; Fig. 6, Table 1, S3). This strong impact of temperature led to a situation where heterotrophic 414 assimilation rates dropped by 50-70% and even fell to the level of autotrophic rates observed at 415 ambient temperature in two of the compartments (C in gastrodermis and N in symbiont; Fig. 6B, E). 416 Concomitantly, significantly reduced C-and N-enrichment under elevated temperature was also 417 observed in the heterotrophic hotspots (Fig. S4). 418 Table 1. The effect of temperature on autotrophic and heterotrophic input. Shown are the mean changes in autotrophic 419 (Aut) and heterotrophic (Het) carbon and nitrogen turnover for each compartment (mean±SD, N=3) in regularly fed [F] 420 corals (cf. Fig. 6). All compartments refer to the surface body wall of the coral coenenchyme. Pairs with asterisk indicate that 421 temperature caused a significant stronger reduction in heterotrophic than autotrophic turnover (Fig. 6 Comparison of the autotrophic and heterotrophic nutrient partitioning between the three tissue 424 compartments of the surface body wall after 6 h of incubation showed radically different pictures for 425 both feeding modes. Heterotrophy caused a strong co-allocation of C and N with almost identical 426 patterns of C and N partitioning across the three compartments (Fig. 7). While the majority of the 427 autotrophic C and N pool is retained in the symbiont after 6 h (see above), approximately two thirds of 428 the heterotrophic nutrient pool is present in the gastrodermis after 6 h under ambient conditions. The 429 remaining third is approximately equally split between symbiont and epidermal compartment, 430 matching the relative compartment size in the case of the symbiont (Fig. 7). The epidermis is the main 431 benefactor of the heterotrophic feeding mode and received a substantially larger share for both 432 elements under ambient conditions compared to autotrophy (~21% vs. ~1%; Fig. 7). The effect of 433 elevated temperature was not just restricted to a reduction in overall nutrient assimilation, but also 434 caused a considerable shift in the partitioning of nutrients, especially for the heterotrophic feeding 435 mode. Here, the shift primarily benefited the symbiont and epidermal compartment and reduced the 436 share of the gastrodermis from originally ~63% to ~50%. This pattern of increasing allocation towards 437 symbiont and epidermis to the disadvantage of the gastrodermis was also observed in the autotrophic 438 dataset of regularly fed corals, albeit much more subtle (Fig. 7). 439

448
By employing correlative TEM-NanoSIMS isotopic imaging, our study revealed the primary sites for 449 the concentration and storage of nutrients derived from auto-and heterotrophy and highlights the 450 critical role that feeding mode and seawater temperature have on the relative turnover and partitioning 451 of C and N within the different tissues of the intact symbiosis. The following discussion will almost 452 exclusively make reference to prior studies of scleractinian corals. Although sea anemones are 453 considered good coral 'model organisms' for many research questions, they are not suitable models 454 for nutritional research of the coral symbiosis. Anemones are characterized by a thick macroscopic 455 tissue and absence of a skeleton, creating different demands and utilization dynamics for C and N 456 compared to a coral animal with a thin tissue layer over a massive C skeleton sink. In the same way, 457 assimilatory performance of the symbionts, e.g. with regard to nitrogen, differs in anemones due the 458 fundamentally thicker host tissue (e.g. absence of detectable nitrate fixation [(45), personal 459 observation] due to high levels of metabolic ammonia release). Thus, we will primarily consider 460 previous studies of auto-and heterotrophy from scleractinian corals for comparison and discussion. 461 where translocated 15 N was most concentrated. Here, stronger 15 N labelling of host nuclei and nucleoli 480 was clearly visible (cf. Fig. 1C) and especially host nucleoli showed a high turnover for autotrophic 481 and heterotrophic N, likely due to the faster turnover of nucleolar ribosomal proteins relative to the 482 nuclear structure (48). Interestingly, the symbiosome membrane that encapsulates the symbionts in 483 their host-derived vacuole appeared to incorporate a substantial amount of photosynthetically derived 484 N within the pulse period. Indeed, in the small stretched membrane part highlighted in figure 1C, the 485 N turnover was with 0.58% ca. 3-6 times higher than the overall gastrodermal value (cf. Fig. 6F). This 486 rapid turnover of its proteinaceous components reveals a highly dynamic nature of the symbiosome 487 membrane that is fuelled by the photosynthate delivery of its occupant. 488

Heterotrophic C and N in the host gastrodermis 489
Nutrients derived from the digestion of the isotopically enriched brine shrimps in the polyps were 490 detectable throughout the wider coenenchyme tissue. Their particular enrichment in the gastrodermis 491 of the surface and basal body wall that directly face the gastrovascular cavity was consistent with the 492 role of the gastrovascular canals as a distribution system throughout the colony (25). The similar 13 C-493 and 15 N-enrichment values from random coenenchyme patches of three independent coral colonies 494 (Fig. 6) demonstrated a highly efficient distribution and assimilation of heterotrophic food over the 495 entire colony surface after just 6 h. Given that the provided food was of proteinaceous nature, the high 496 degree of co-labelling of C and N in the overall gastrodermis and in the different hotspots as well as 497 the similar partitioning of both elements across the compartments suggests that basic monomers 498 containing both elements (e.g. amino acids) were used directly in anabolic processes, rather than being 499 de-aminated or extensively processed involving internal reserves prior to incorporation into the tissue. 500 Nevertheless, the three observed types of hotspots reflected different stages of food processing. The 501 round membrane-delineated 15 N-hotspots (pink arrows Fig. 2-4) constitute vesicles that contained food 502 material that was endo-or phagocytosed from the gastrovascular canals and transported through the 503 host tissue. These structures are likely of transitory nature and a direct result of the ingestion and 504 processing of heterotrophic food. The second type of 15 N-hotspot structure, which appeared cup-505 shaped (blue arrows Fig. 2-5), seemed to be more permanent, because it regularly occurred in S. 506 pistillata samples and was also observed in adults and larvae of P. damicornis (49,50). Their electron-507 dense appearance in the TEM indicates either a higher physical density or a high affinity for the heavy 508 metals used in the post-fixation and staining, suggesting the presence of a high degree of negatively 509 charged polar groups. Their close association with lipid bodies and the presence of empty space 510 surrounding these structures might indicate a role in the remobilization of host lipid bodies. Their 511 direct contact with host lipid bodies and their rapid incorporation of heterotrophic N and external 512 ammonia suggest a crucial function in the coral lipid metabolism, potentially related to breakdown of 513 lipid droplets as evidenced by the residual material in the vicinity of the extraction. The third type of 514 hotspot (green arrows in Fig. 2-4) usually showed the highest levels of heterotrophic 13 C and 15 N 515 enrichment (with ca. 40-60% of the brine shrimp 13 C and 15 N enrichment level), but its specific 516 function remains unclear. The high degree of heterotrophic C-and N-enrichment either indicates the 517 lowest degree of degradation and processing of the original food or cellular patches with the fastest 518 turnover that rapidly and extremely efficiently incorporate nutrients into cellular material. Since no 519 clear ultrastructure was distinguishable and such highly enriched gastrodermal hotspots were not 520 found in the autotrophic treatment, we discount however the possibility of dedicated cytoplasmic areas 521 with permanently high C and N turnover. The fact that all these hotspots showed reduced enrichment 522 with labelled brine shrimp material at elevated temperature suggests that all these bodies are involved 523 in the breakdown and storage of heterotrophic nutrients and experienced a lower influx of 524 heterotrophic food. 525 An interesting observation was the strong gradient between host gastrodermis and epidermis with 526 regard to C and N assimilation in both feeding modes. It appears that the thin collagenous mesoglea 527 acted as a border that restricted flow between both tissue layers and limited the distribution and 528 assimilation of nutrients primarily to the side that faces the gastrovascular canal (including shuttling of 529 food vesicles throughout the gastrodermis). The consistent large difference in C and N turnover 530 between both layers indicated either a significant delay in the anabolic utilisation of nutrients 531 originating from the gastrodermis or a fundamentally different anabolic metabolism, with epidermal 532 cells having a much slower turnover. Notably, we detected substantial enrichment from heterotrophic 533 food also in the basal body wall gastrodermis and the adjacent calicodermis, illustrating a direct 534 contribution of heterotrophically derived nutrients to the calicodermal activity that forms the skeleton. 535 536

Assimilation of heterotrophic nutrients by Symbiodinium 537
The labelling of primary and secondary starch deposits in the symbiont with heterotrophic 13 C 538 demonstrated that coral catabolism released brine shrimp C through decarboxylation reactions and the 539 tricarboxylic acid cycle (TCA), which was then directly re-assimilated as CO 2 in symbiont 540 photosynthesis. Indeed, comparing the fate of labelled C from incubation with 1 mM [1-13 C]-pyruvate 541 (label released as CO 2 in the pyruvate dehydrogenase complex) or [3-13 C]-pyruvate (label enters TCA 542 cycle) over 3 h in a separate experiment, confirmed that the initial decarboxylation of pyruvate prior to 543 entering the TCA cycle is a major source for symbiont C recycling (Fig. S5B). In contrast, fewer 13 C 544 was photoassimilated from the CO 2 -release of the TCA cycle and the homogenous host tissue labelling 545 from this form indicated that the label entered primarily pathways of amino acid and fatty acid 546 synthesis via cataplerotic reactions of the TCA cycle and direct Acetyl-CoA use, respectively (Fig  547   S5E). The different labelling patterns from these two incubations confirm that catabolic CO 2 is fixed 548 by the symbiont and that phototrophic carbon translocation is primarily directed towards the host lipid 549 bodies. To what degree host lipid bodies receive phototrophic carbon directly through the use of 550 translocated lipids or indirectly through sugar breakdown and de novo fatty acid synthesis via 551 glycolytically derived Acetyl-and Malonyl-CoA remains however to be answered. 552 Analogously to heterotrophic C assimilation by the symbiont, a substantial enrichment of the symbiont 553 with heterotrophic N and the occurrence of occasional N hotspots illustrated a rapid assimilation of 554 brine shrimp N. Our data indicated that ~14-24 % of the assimilated heterotrophic C and N pool was 555 found in Symbiodinium after 6 h (based on cross-sectional areal budgets, Fig. 7), roughly matching 556 earlier observations in Oculina arbuscula and O. diffusa (10-20% of 15 N over 4-28 h (32, 51)). Since 557 the measurements here were in hospite, previous concerns about incomplete homogenization and host 558 tissue contamination of the symbiont pellet as experimental artefacts were avoided. An interesting 559 aspect with regard to the transfer of nutrients from host to symbiont is the question of how fast the 560 symbiont population can access heterotrophic N. Piniak and Lipschultz (51) argued that the occurrence 561 of heterotrophic N in the symbiont within 4 h was insufficient time for a complete recycling of 15 N 562 from host metabolism ("i.e. host digestion, synthesis into host macromolecules, catabolism, excretion, 563 uptake by zooxanthellae") and that the symbionts must have assimilated prey N directly. Our data 564 seem to support this notion, considering that the relative amount of replaced N in the symbiont 565 population under ambient temperatures in two of the three colonies was only 19% and 20% lower than 566 the corresponding host gastrodermal values after 6 h (the third one was 52% lower). Considering their 567 intracellular location and the consistent observation of assimilation of organic substrates (and amino 568 acids in particular) by Symbiodinium (52-55), these symbionts likely directly assimilate amino acid 569 monomers as well as released ammonium as they appear in the tissue in the process of the breakdown 570 of the brine shrimp material (assuming transport across the symbiosome membrane). Indeed, the 571 NanoSIMS images from auto-and heterotrophically 15 N-enriched symbiont cells were qualitatively 572 similar, notwithstanding the difference in absolute level of labelling (compare Figs. 1C,F and S3B). 573 Thus, in our interpretation, the small 15 N hotspots in the symbionts represent the immediate 574 concentration and storage of nitrate and ammonium ions into crystalline uric acid (this study, 37), 575 whereas the more or less homogenous labelling of all subcellular symbiont structures represents the 576 incorporation of more complex N-containing compounds from the brine shrimp into the general 577 cellular infrastructure. Despite the additional influx of prey N, internal N recycling by the symbionts is 578 a key mechanism for the coral and previous studies have estimated that up to ~80-90 % of the 579 symbiont N is derived from this process (16, 33). Here, heterotrophic N considerably supplemented 580 the symbiont's N status (exceeding its own nitrate fixation by a factor of 2.4; Fig. 6A) and this supply 581 in combination with the influx of other limiting elements such as phosphorus likely lifted the general 582 nutrient limitation (56-58). 583

Modulation of symbiont autotrophy by holobiont feeding 584
The motivation to investigate symbiont autotrophy efficiency under different feeding regimes came 585 from the observation that regularly fed corals tend to display higher host tissue thickness and increased 586 symbiont densities that alter the light microenvironment for individual cells (reviewed in 58). While 587 such tendencies were also observed here (Table S4), the NanoSIMS data showed that the symbionts in 588 fed corals assimilated on average 10% more C, but not more N (Fig. 6A, B). Assuming that the 589 relative activity of host anabolism and catabolism was not fundamentally different between unfed and 590 regularly fed corals, this extra amount of photosynthetically assimilated C was primarily utilized by 591 the symbionts themselves similar to the observation in another feeding experiment (12). These 592 observations thus confirm that the enhanced N supply from regular feeding increased individual 593 symbiont C investment into cellular structure (as reflected by the NanoSIMS data and the tendency of 594 increased symbiont size and density in fed corals). Thus, the uncoupling of symbiont photosynthesis 595 and growth under nutrient limitation is eased under regular host feeding, leading to the expected effect 596 that the symbionts retain a larger share of their fixed C for growth (reviewed in 59). We attribute the 597 unchanged state in nitrate fixation to the isometric scaling of increased host tissue biomass and 598 symbiont biomass (symbiont density per host biomass was not significantly affected; Table S4). Thus, 599 the tendency for increased host tissue biomass and elevated ammonia production that might have 600 shifted the N prevalence of the in hospite symbiont from nitrate to the preferred ammonia (60-65), as 601 is the case in anemones (45), was counterbalanced by increases in symbiont numbers (p=0.053, Table  602 S4). For the effect of elevated temperature, we can conclude that the feeding state of the holobiont 603 made no difference for the generally reduced autotrophic acquisition of C or N. 604

Coral autotrophy vs. heterotrophy and the impact of elevated temperature 605
Our data clearly demonstrated that the influx of heterotrophic C and N to host tissues substantially 606 exceeded any photosynthate contribution, in particular for N from nitrate fixation (Fig. 8). When 607 extrapolated from these numbers, the single feeding event immediately (within 6 h) provided the host 608 gastrodermis with the C and N equivalent of ~14 and ~283 daylight hours worth of translocated 609 photosynthates. This gradient was even stronger for the epidermis, with the C and N received from the 610 single feeding event equating to ~97 and ~903 daylight hours of photosynthetic input, respectively. 611 Although an earlier model calculation estimated that approximately only one third of the daily N 612 demand is covered through nitrate assimilation in Stylophora pistillata, heterotrophic N influx would 613 still considerably exceed the symbiont's inorganic N fixation (NO 3 and NH 4 + make up 75% of the 614 daily budget) (66). Thus, while symbiont C is mainly derived from seawater DIC, internal recycling 615 and heterotrophic input are the main source for symbiont N (Fig. 8). Our NanoSIMS data 616 demonstrated that whereas photoautotrophic C is mainly directed towards storage in the host lipids, 617 heterotrophic C is primarily assimilated into the host tissue (Fig. 8). We suggest that this was likely 618 due to particular proteinaceous nature of the ingested zooplankton, with a limited deamination of 619 amino acids to generate keto acids and ketone bodies that would be used in glycolytic or TCA-linked 620 pathways to form Acetyl-CoA for lipogenesis. 621

628
As previously reported, negative effects of elevated temperature (11.2 DHW) in this Red Sea 629 Stylophora pistillata were largely restricted to lowered autotrophic nutrient turnover and not reflected 630 in most other physiological variables (39). This observation also applies to the heterotrophic feeding 631 mode. Although still providing a net influx of nutrients, turnover on the tissue-level scale showed even 632 stronger temperature-induced reductions for N (all compartments) and C assimilation (both host tissue 633 layers) from the heterotrophic feeding mode compared to the autotrophic mode. The simplest 634 explanation for the reduced assimilation could be a reduced feeding rate at elevated temperature as a 635 previous study on a Red Sea Stylophora pistillata has shown a decline in polyp ingestion rates by ca. 636 67% at 31°C (67). This matches the observed average drop in C and N assimilation in our experiment 637 (57-66% reduction for the host tissue layers) and also explains the reduced enrichment in the coral 638 tissue structures that degrade and process food material (i.e. the heterotrophic hotspots). The consistent 639 shift in the internal partitioning of nutrients in fed corals for both feeding modes, benefiting symbiont 640 and epidermis, suggests a modulating impact of temperature specific to the gastrodermis as central hub 641 of autotrophic and heterotrophic nutrient assimilation and distribution. This is furthermore supported 642 by the fact that the ~1.5-ratio between epidermal and gastrodermal area was unaltered by temperature 643 in fed corals. Thus, the change in nutrient partitioning was truly related to nutrient allocation and 644 assimilation efficiency and not just due to size changes of the compartments as e.g. observed in 645 Pocillopora damicornis at elevated temperature (68). The specific mechanism behind this shift 646 remains unclear. Notably, the symbiont gains more of this increased C and N share under elevated 647 temperature than the epidermis and the NanoSIMS images indicate that this is related to a stronger 648 acquisition of respiratory CO 2 and organic and inorganic N derived from the food breakdown. Thus, it 649 might be reasonable to assume a stronger diversion of heterotrophic food for catabolic activity in the 650 gastrodermis at elevated temperature, which provides additional energy under elevated temperature for 651 the host on the one hand, but also supports symbiont nutrient fixation and internal recycling. This 652 suggestion of enhanced gastrodermal catabolic activity at elevated temperature should however be 653 treated carefully, since to date it has not been possible to verify this through other proxies for 654 respiratory activity (i.e. metabolic CO 2 release) that can be measured in a tissue-specific manner. 655 Increased catabolization of heterotrophic C is nevertheless consistent with observations in some corals 656 recovering from bleaching (69, 70). The increased symbiont share of heterotrophic C and N at elevated 657 temperature provides some support for the proposed role of heterotrophic feeding in alleviating 658 symbiont nutrient and CO 2 -limitation, which under bleaching conditions might contribute to 659 preventing sink limitation of photosynthesis and maintaining autotrophic capacity (71). 660 Contrasting the abundance and distribution of autotrophic and heterotrophic nutrients in the coral 661 symbiosis using NanoSIMS represents a new tissue-and cellular-level view of the two main 662 fundamental modes of coral nutrition and shows the tight connection between auto-and heterotrophy 663 in a symbiotic coral. This study nevertheless represents a 6 h snapshot of the tissue in a healthy coral 664 known to have a comparably high temperature threshold in its location (39). Thus, the observed role of 665 autotrophy and heterotrophy has to be considered within this context. A number of studies have shown 666 the significant role that heterotrophy plays in the survival and recovery of corals after severe stress 667 events that left their autotrophic capacity impaired (69,(72)(73)(74)(75)(76). Thus, while the preservation of 668 symbiont functioning and initial host energy reserves are key elements of coral resistance to thermal 669 stress (77), it is also the complementing interplay of auto-and heterotrophy that defines the nutritional 670 state and ability of corals to recover after a severe symbiotic disturbance in a warmer ocean (78). 671