A unique approach to monitor stress in coral exposed to emerging pollutants

Metabolomic profiling of the hexacoral Pocillopora damicornis exposed to solar filters revealed a metabolomic signature of stress in this coral. It was demonstrated that the concentration of the known steroid (3β, 5α, 8α) -5, 8-epidioxy- ergosta- 6, 24(28) - dien- 3- ol (14) increased in response to octocrylene (OC) and ethylhexyl salicylate (ES) at 50 µg/L. Based on the overall coral response, we hypothesize that steroid 14 mediates coral response to stress. OC also specifically altered mitochondrial function at this concentration and above, while ES triggered a stress/inflammatory response at 300 µg/L and above as witnessed by the significant increases in the concentrations of polyunsaturated fatty acids, lysophosphatidylcholines and lysophosphatidylethanolamines. Benzophenone-3 increased the concentration of compound 14 at 2 mg/L, while the concentration of stress marker remained unchanged upon exposition to the other solar filters tested. Also, our results seemed to refute earlier suggestions that platelet-activating factor is involved in the coral inflammatory response.


Compared effect of OC and ES on coral metabolomes.
In the current study, P. damicornis nubbins were exposed to ES for 7 days at concentrations of 5, 50, 300 and 1000 µg/L. As with OC previously, the extracts prepared from the coral nubbins exposed to ES were analyzed by UHPLC-ESI + -HRMS 2 and compared with control experiment 7 .
First, unlike what happened with OC, ES or ES-analogs do not seem to accumulate in the coral, although both compounds possess a 2-ethylhexyl side chain. Our hypothesis is that the ester group in OC is more stable than in ES. ES would then be degraded by carboxylesterase-mediated ester hydrolysis 17 or via the bacterial ortho degradation pathway 18,19 while OC is degraded by hydroxylation of the 2-ethylhexyl chain and subsequent grafting of fatty acids. As a result, lipophilic OC analogs accumulate in coral tissues, which might lead to further accumulation by the trophic chain.
The metabolomic profiles were compared with those of control corals treated with DMSO only (0.25% v/v). Eighteen compounds are significantly upregulated at 1000 µg/L ES, and in some cases are also upregulated at lower ES concentrations ( Table 2, Fig. 1).
Compounds 1-3 are the polyunsaturated fatty acids eicosapentaenoic, docosahexaenoic and arachidonic acid (AA, Fig. 1). The structures were determined based on the molecular formulas and examination of fragmentation spectra with MS-Finder. Identification was confirmed by comparison of retention times and MS/MS spectra with those of commercially available standards. These compounds were significantly upregulated at 1 mg/L ES and were not affected at lower concentrations (Fig. 1). Until recently, ω3 long-chain polyunsaturated fatty acids (PUFAs) would have been considered as Symbiodiniaceae 20 metabolites. Vertebrates lack the ωx desaturases necessary for PUFA synthesis and it was largely accepted that all animals should lack it too. However, it is now well established that many marine invertebrates, including corals (P. damicornis among them), have acquired ωx desaturases by horizontal gene transfer [21][22][23] .
In mammals, ω3 PUFAs, and in particular AA, that come from the diet are stored and are involved in the inflammatory cascade after cleavage from phosphatidylcholine by phospholipase A2 (PLA2) and transformation into several derived signaling molecules such as leukotrienes and prostaglandins. In corals, increased production of eicosanoids, linked to an increase in the expression of an allene oxide synthase-lipoxygenase (AOS-LOXa) was shown for the soft (octo)coral Capnella imbricata in response to mechanical injury and thermal stress 22,24 . In stony (hexa)corals, eicosanoids such as 8-hydroxyeicosatetraenoic acid (8-HETE), (S,5Z,11Z,14Z)-8-hydroxy-9-oxoicosa-5,11,14-trienoic acid and (5Z,12a,14Z)-9-oxo-prosta-5,10,14-trien-1-oicacid were also observed, both in extracts of whole coral tissue, and after incubation of 14 C-labeled AA with tissue homogenates of Acropora millepora, A. cervicornis and Galaxea fascicularis 23 . These results, combined with previous transcriptomic analysis showing upregulation of enzymes linked to eicosanoid production in corals undergoing black disease or thermal stress 25,26 , led to the hypothesis that the production of eicosanoids could also be associated with stress in stony corals 23 . Finally, experiments with host switching using the sea anemone Exaiptasia pallida also show evidences towards increases of eicoisanoids after colonization with a heterologous Symbiodiniaceae 27 . Our blastp analysis of the P. damicornis genome identified several enzymes linked to the production of eicosanoids, including cytosolic and secreted PLA2, AOS-LOXs, 5-lypooxygenases and a leukotriene A4 hydrolase, confirming previous results in the literature (Supplementary Table S1) 23,26 . However, other than PUFAs, we could not detect any of the eicosanoids, leukotrienes or prostaglandins in P. damicornis extracts, nor could we identify a leukotriene receptor based on blastp searches using the mouse LTB4R and CYSLTR1 as queries. The closest seven-transmembrane G-protein coupled receptors in P. damicornis were related to receptors in the alpha and beta groups of rhodopsin family (defined by Fredriksson et al.) 28 and not the gamma and delta groups as other known leukotriene receptors.
Altogether, these results suggest that ω3 PUFAs concentration increased likely in the context of a stress process triggered by ES. This effect was only noticeable at the highest exposition concentration, but the downstream signaling molecules or putative receptors were absent or not measurable in our study.  Supplementary Fig. S14), the 5:1 peak intensity ratio of product ions at m/z 104 and 147 indicated that 5 was an sn-1-LPC regioisomer 29 . The identification of compounds 5-7 was finally unambiguously established by comparing the retention times and MS/MS spectra with those of commercially available standards.
As for ω3 PUFAs, these compounds were significantly upregulated only at 1 mg/L ES (Fig. 1). LPCs are widespread in the animal kingdom 30 . In mammals, LPCs are proinflammatory lipids upregulated in an event of inflammation, inducing pro-inflammatory cytokine secretion and activating B lymphocytes [31][32][33] . LPCs stimulate time-and concentration-dependent release of arachidonic acid (compound 3) in human coronary artery smooth muscle cells 33 . Interestingly, it has been proposed that the coral immune system is similar to that of higher organisms, including mammals [34][35][36][37] . In Porites sp., platelet-activating factor (PAF) concentration has been reported to increase during interactions with Acropora cervicornis 30 as has the expression of a gene coding for a putative LPC acetyltransferase, the protein responsible for converting LPC to PAF, in response to stress and inflammation in mammals 38 . PAF was not detected in our study, even though the genome of P. damicornis codes for an acyl transferase with high homology to the mammalian LPC acyltransferase, containing the motif HXXXXD responsible for acyltransferase activity, and a putative acyl-acceptor binding pocket domain (Supplementary Table S1). Remarkably, compound 7 is an isomer of PAF and could easily be mistaken for PAF. In our study, its structure was confirmed by comparison with standards of 7 and PAF (Supplementary Figs. S2 and S3). The fragmentation product at m/z 104 from [7 + H] + further confirms the identification of 7 as the 1-O-octadecanoyl-sn-glycero-3-phosphocholine 39,40 . In addition, unlike what was reported for Acropora digitifera, we did not identify a gene coding PAF receptor based on blastp searches using the human PTAFR receptor as a query (Supplementary Table S1). The closest seven-transmembrane G-protein-coupled receptors in P. damicornis were related to cholecystokinin receptors in beta groups of the rhodopsin family (defined by Fredriksson et al.) 28 and not in the delta group as the PTAFR receptor. Finally, our results show that the expression of LPCs fluctuates widely between  www.nature.com/scientificreports www.nature.com/scientificreports/ replicates ( Fig. 1) and, therefore, cannot be considered as good stress indicators. The signal was weakly significant (p < 0.05 for all 4 LPCs) at 1000 µg/L ES and was not significant at lower concentrations.  The X-axis is in Log scale. For each ion, the leftmost point represents the minimal value, and the rightmost point the maximal one. The rectangular box represents the 25% quantile to 75% quantile ranges. The dark line shows the average of the distribution. Significance levels relative to DMSO were determined by an ANOVA followed by a Tukey HSD test. The differences were not significant unless otherwise stated. ***p < 0.001, **p < 0.01, *p < 0.05. (2020) 10:9601 | https://doi.org/10.1038/s41598-020-66117-3 www.nature.com/scientificreports www.nature.com/scientificreports/ MS 2 fragmentation patterns, in which the migration of the enol ether chain on the amino group leads to the major fragmentation products (Fig. 3, Supplementary Scheme S3). The identification of compound 10 was confirmed by comparison with a commercially available standard. Compounds 10 and 11 were significantly upregulated at 1000 µg/L ES.
It has been well established that LPEs and LPCs are produced by hydrolysis of phosphatidylethanolamines and phosphatidylcholines, releasing arachidonic acid or other fatty acids linked to the sn-2 position. This reaction is usually mediated by a phospholipase A2. An increased phospholipase A2 activity in coral exposed to ES would account for the observed concomitant increase of ω3 PUFAs (1-3), LPCs (4-7) and LPEs (8)(9)(10)(11). As discussed above P. damicornis genome mining revealed that the genome of this species codes for a number of secretory (sPLA2) and cytosolic PLA2 (cPLA2) presumably involved in this mechanism. Similar to in mammals 42,43 , sPLA2 upregulation in coral is visibly associated with the activity of the innate immune system and/or a stress response 26,38 . Increases in PLA2 expression and arachidonic acid was also shown after host switching with a heterologous Symbiodiniaceae with sea anemone Exaiptasia pallida even though that study did not see differences in LPEs and LPCs 27 . Overall, the metabolomic response of coral after exposition to ES resembles the signature of a stress/inflammatory process triggered by ES.
To the exception of compound 14, compounds 12-17 ( Fig. 1) were detected with somewhat small response peaks, and it was not always possible to obtain MS/MS spectra. Compounds 12-17 seemed related based on their molecular formulas and on their ionization and fragmentation patterns. Interestingly, the differential analysis  www.nature.com/scientificreports www.nature.com/scientificreports/ showed that all these compounds were significantly upregulated when the coral was exposed to ES at concentrations ranging from 50 to 1000 µg/L. It turned out that the concentration of 14 also increased upon exposition to OC 7 , and at this point, it was clearly necessary to determine unambiguously the structure of compound 14. Extensive fractionation allowed for the isolation of 14 in its pure form, which was ultimately identified by NMR as the known steroid (3β,5α,8α)-5, 8-epidioxy-ergosta-6,24(28)-dien-3-ol. Compound 14 along with many epidioxy sterol analogs have been described essentially in marine invertebrates, including several cnidarians [44][45][46][47] . The role and fate of this compound in invertebrates remains an unresolved question. However, steroids are present in the whole animal kingdom. They regulate life cycles, mating, and development. In Cnidarians, bioregulatory pathways and hormonal-like signaling remain largely uncharacterized [48][49][50] . Nonetheless, a family of nuclear receptors has been found to bind the ancient hormone paraestrol A 50,51 . According to Khalturin et al., Cnidarian steroids can be transported in the digestive tract and through the mesoglea, and may be involved in intercellular communication. Here, compound 14 concentration increased in coral exposed to pollutants, and these pollutants eventually triggered a stress response witnessed by the increased production of specific lipids. We hypothesize that the increased concentration of this class of steroids -and in particular the major compound 14 -is a signal triggering a coral inflammatory-like response. Owing to the very small standard deviation within replicates (Fig. 2), 14 could be considered as a good marker of stressed corals. In our experiments, the concentration of 14 significantly increased with both ES and OC at 50 µg/L, indicating that both solar filters had a negative impact on coral. However, ES triggered a stress/inflammatory response while OC also specifically altered mitochondrial function 7 .
Last, it should be mentioned that the relative concentration of five metabolites decreased when the coral was exposed to ES at 1 mg/L (Supplementary Figs. S1, S51-S61). These metabolites have been identified as four monogalactosyl diacylglycerols (MGDGs 19-22) and one cerebroside (23). In MS 2 , MGDG fragmentation pattern shows the length and number of unsaturation of each acyl chain. The relative position of the acyl chains can be assigned based on the relative peak intensities of the sodium adduct fragmentation products. The peak intensity of the product ion resulting from the loss of the sn-1 fatty acid is always higher than the one resulting from the loss of the sn-2 fatty acid 52 . Although cerebrosides are widely distributed in the eukaryotes, galactolipids including MGDGs are the main components of plant chloroplast membrane lipids 53,54 . Decreased MGDGs concentration might point towards an effect of the UV filter on the coral symbiont. It is possible that the coral had begun to bleach although bleaching was not visible to the naked eye. However, specific toxicity mechanism unbalancing Symbiodiniaceae metabolism or any process disrupting algal symbiont metabolite translocation cannot be ruled out.

Effect of BEMT, BM, BP3, DBT, DHHB, ET, HS, and MBBT. The effect of 8 other solar filters was
examined based on the comparison of global metabolomic profiles between exposed and unexposed coral, but it was also examined in light of the putative increased concentration of compound 14 in exposed corals. In general, the first step consisted of exposing coral at 1000 µg/L of each solar filter and testing lower concentrations when an effect was detected. BP3 was tested at 2000 µg/L as well because very high environmental concentrations of this compound have been occasionally detected 3 . The relative variations in compound 14 concentration are reported in Fig. 4.
We observed that BEMT, DBT, DHHB, ET and MBBT do not seem to affect the overall coral metabolome at the highest concentration tested. Specific measurement of the concentration of compound 14 in treated versus untreated coral confirmed the apparent innocuousness of these five molecules, as the concentration of 14 remained stable. It was found that the concentration of 14 increased at 2 mg/L BP3, indicating that BP3 most certainly affects wild coral at the highest published environmental concentration 3 . Coral exposure to BM does not induce an increase in the concentration of compound 14. Nevertheless, the presence of undetermined ions in the global coral metabolome will require further investigation. Last, HS does not alter either the concentration of compound 14 or the overall metabolome of P. damicornis. However, polyps of coral exposed to HS at 1 mg/L were closed at the end of the assay, while those of control corals were not, as if the coral reacted although its metabolome was not significantly altered ( Supplementary Fig. S61).

conclusion
This work establishes a metabolomic signature of stressed P. damicornis. In this coral, the known steroid (3β, 5α,8α)-5,8-epidioxy-ergosta-6,24(28)-dien-3-ol (14) is viewed as a steroid hormone that could trigger coral response to pollutants. OC was the most toxic of the tested UV filters. It induced coral stress response, while triggering mitochondrial dysfunction at 50 µg/L. Of concern was also the previously reported accumulation potential of possibly toxic coral-modified OC derivatives. ES comes second in terms of toxicity. ES triggered coral stress response at 50 µg/L, inducing a significant increase in the concentration of compound 14. At 300 µg/L ES and above, the relative concentration of several PUFAs, LPCs and LPAs also increased. ES may also have induced partial coral bleaching at 1 mg/L although this remains to be firmly established. BP3 was also toxic at 2 mg/L in our assay. BEMT, BM, DBT, DHHB, ET, HS and MBBT did not affect coral metabolism at any of the concentrations tested, although in the case of HS and BM, further investigations are needed to evaluate their potential effects. More investigations are also needed to clarify the role of compound 14 in coral hormonal response to pollutants.

Methods
General experimental procedures. Nuclear Magnetic Resonance (NMR) spectra were recorded on a JEOL ECZ500R spectrometer equipped with a 5-mm inverse detection FG/RO Digital Autotune Probe. Chemical shifts (δ) are reported as ppm downfield from tetramethylsilane, and the coupling constants (J) are reported in Hertz. High-resolution MS/MS analyses were conducted with a Thermo UHPLC-HRMS system 7  www.nature.com/scientificreports www.nature.com/scientificreports/ FullMS-data dependent MS 2 mode. In FullMS, the resolution was set to 70,000, and the AGC target was 3.10 6 . In MS 2 , the resolution was 17,500, AGC target 10 5 , isolation window 0.4 Da, and stepped normalized collision energy 15/30/45 was used, with 15 s dynamic exclusion. The lock mass option was set for an ion at m/z 144.98215, corresponding to Cu(CH 3 CN) 2 + . For coral profiling and comparison of coral profiles with standards, the UHPLC column was a Phenomenex Luna Omega polar C-18 150 × 2.1 mm, 1.6 µm. The column temperature was set to 42 °C, and the flow rate was 0.5 mL.min −1 . The solvent system was a mixture of water (solution A) with increasing proportions of acetonitrile (solution B), and both solvents were modified with 0.1% formic acid. The gradient was as follows: 2% B 3 min before injection; then from 1 to 13 min, there was a shark fin gradient increase of B up to 100% (curve 2), followed by 100% B for 5 min. The flow was diverted (not injected into the mass spectrometer) before injection, up to 1 min after injection. For fast analysis of the fractions of coral extract, the UHPLC column was a Thermo Scientific Accucore Vanquish C18 + 50 × 2.1 mm, 1.5 µm. The column temperature was set to 42 °C, and the flow rate was 0.5 mL.min −1 . The gradient was as follows: 2% B 1 min before injection; then from 0 to 4 min, there was a shark fin gradient increase of B up to 100% (curve 2), followed by 100% B for 1 min. The flash chromatography was performed with a Teledyne ISCO CombiFlash Companion equipped with a Büchi FlashPure Ecoflex C18 50 µm 40 g column. The column was equilibrated with water:CH 3 CN 4:6. The flow rate was 40 mL/ min. The injection was performed in the solid phase, with fraction F3 impregnated on C18 silica (1 mL). The gradient was 60% acetonitrile until 1 min after injection; then, from 1 to 20 min, there was a linear gradient increase of acetonitrile up to 100%, followed by 100% acetonitrile for 10 min. Effluents were collected in 30 s fractions. The preparative HPLC was conducted with a hybrid system mounted with a Reodyne manual injector, 2 Varian PrepStar 218 HPLC pumps, a Dionex Ultimate 3000 variable wavelength detector, and a Dionex Ultimate 3000 fraction collector. The column was a Phenomenex Luna C18 250 × 21.20 mm, 5 µm. The gradient was as follows: Relative integration values for compound 14 ion peak in exposed coral compared to control animals. The vertical dashed line illustrates the 1:1 ratio. The rectangular box represents the min to max ranges. The dark line shows the average of the distribution. Significance levels relative to DMSO were determined by an ANOVA followed by a Tukey HSD test. The differences were not significant unless otherwise stated. ***p < 0.001, **p < 0.01, *p < 0.05. The multiple data points for same concentrations are for repeat experiments. pocillopora damicornis. Fragments of the coral P. damicornis were collected in Oman in 2014 (CITES permit 37/2014). This procedure had no impact on the wild population as 1-5% of few coral colonies were collected. The coral was acclimated in tanks at the Banyuls Oceanological Observatory. New colonies were obtained in the laboratory from the fragments and used for our experiments. The corals were maintained in artificial sea water (ASW) prepared with reverse osmosis purified water and Reef Salt SeaChem salts. Salinity was adjusted to 36 g/L, pH = 8, and the temperature was set at 24 °C. All experiments were conducted with the same ASW.
Exposition of coral to solar filters, extraction and metabolomic analyses. The coral exposition protocol, the extraction, the metabolomic profiling and the statistical analyses were conducted as described before for OC 7 . They are also reported in Supplementary Information. Five replicates were used for each condition, and in some cases, the conditions were repeated. The standard compounds were diluted in MeOH (≈ 10 µg/ mL each, 1 µL injected) and were analyzed in the same conditions as the coral extracts for comparison of retention times, MS and MS/MS spectra. All standards were identical to the coral metabolites in retention times and MS/ MS spectra (see supplementary information). Interpretations of the metabolomic data and MS/MS spectra were conducted with the help of Compound Discoverer 2.

Isolation and characterization of compound 14.
Approximately 200 2-5 cm-long coral nubbins were cut from the branch tips of mother colonies. The nubbins were placed in a large Erlenmeyer flask and covered with MeOH:CH 3 CN 1:1. After 24 h at room temperature, the flask was sonicated for 20 min, the coral pieces were removed by filtration, and the solvent was evaporated to give the crude coral extract (2.5 g). The crude extract was dissolved in the smallest amount DMSO possible and was purified by SPE with a phenomenex Strata C-18 150 mL column. The column was equilibrated successively with CH 3 CN and water. Elution was performed with water (300 mL, F1), with water:CH 3 CN 1:1 (300 mL, F2) and finally with MeOH:CH 2 Cl 2 8:2 (300 mL, F3). Fractions were evaporated and diluted in MeOH at 1 mg/mL for LC/MS analyses (1 µL injected). The target compound at t R ≈ 11.02 min and m/z 395.3308 was detected in fraction F3 (0.29 g). Fraction F3 was purified by flash chromatography, providing 80 fractions. The analysis of the fractions demonstrated that the target compound was detected in fractions 45 to 63. These fractions were gathered and evaporated. For injection, fraction F3.45-63 (49.95 mg) was diluted in water:CH 3 CN 1:9 (2 mL). Preparative HPLC was performed twice with 1 mL injected. The targeted compound was concentrated in fractions 40 to 45, which were gathered and evaporated. The resulting fraction F3.45-63.40-45 (5.91 mg) was diluted in water:CH 3 CN 15:85 (1.5 mL) and was purified by small-scale preparative HPLC by portions of 250 µL (6 injections). The desired compound 14 was isolated in pure form from fractions 27-29 (0.9 mg).

P. damicornis genome mining.
To interpret the possible roles of the identified stress biomarkers, we queried a number of possible putative enzymes and receptors leading to the synthesis, modification and binding of different phospholipids, polyunsaturated fatty acids and oxylipins against proteins coded by the P. damicornis