Evidence for a role of protein phosphorylation in the maintenance of the cnidarian–algal symbiosis

Abstract The endosymbiotic relationship between cnidarians and photosynthetic dinoflagellate algae provides the foundation of coral reef ecosystems. This essential interaction is globally threatened by anthropogenic disturbance. As such, it is important to understand the molecular mechanisms underpinning the cnidarian–algal association. Here we investigated phosphorylation‐mediated protein signalling as a mechanism of regulation of the cnidarian–algal interaction, and we report on the generation of the first phosphoproteome for the coral model system Aiptasia. Mass spectrometry‐based phosphoproteomics using data‐independent acquisition allowed consistent quantification of over 3,000 phosphopeptides totalling more than 1,600 phosphoproteins across aposymbiotic (symbiont‐free) and symbiotic anemones. Comparison of the symbiotic states showed distinct phosphoproteomic profiles attributable to the differential phosphorylation of 539 proteins that cover a broad range of functions, from receptors to structural and signal transduction proteins. A subsequent pathway enrichment analysis identified the processes of “protein digestion and absorption,” “carbohydrate metabolism,” and “protein folding, sorting and degradation,” and highlighted differential phosphorylation of the “phospholipase D signalling pathway” and “protein processing in the endoplasmic reticulum.” Targeted phosphorylation of the phospholipase D signalling pathway suggests control of glutamate vesicle trafficking across symbiotic compartments, and phosphorylation of the endoplasmic reticulum machinery suggests recycling of symbiosome‐associated proteins. Our study shows for the first time that changes in the phosphorylation status of proteins between aposymbiotic and symbiotic Aiptasia anemones may play a role in the regulation of the cnidarian–algal symbiosis. This is the first phosphoproteomic study of a cnidarian–algal symbiotic association as well as the first application of quantification by data‐independent acquisition in the coral field.


Abstract
The endosymbiotic relationship between cnidarians and photosynthetic dinoflagellate algae provides the foundation of coral reef ecosystems. This essential interaction is globally threatened by anthropogenic disturbance. As such, it is important to understand the molecular mechanisms underpinning the cnidarian-algal association.
Here we investigated phosphorylation-mediated protein signalling as a mechanism of regulation of the cnidarian-algal interaction, and we report on the generation of the first phosphoproteome for the coral model system Aiptasia. Mass spectrometrybased phosphoproteomics using data-independent acquisition allowed consistent quantification of over 3,000 phosphopeptides totalling more than 1,600 phosphoproteins across aposymbiotic (symbiont-free) and symbiotic anemones. Comparison of the symbiotic states showed distinct phosphoproteomic profiles attributable to the differential phosphorylation of 539 proteins that cover a broad range of functions, from receptors to structural and signal transduction proteins. A subsequent pathway enrichment analysis identified the processes of "protein digestion and absorption," "carbohydrate metabolism," and "protein folding, sorting and degradation," and highlighted differential phosphorylation of the "phospholipase D signalling pathway" and "protein processing in the endoplasmic reticulum." Targeted phosphorylation of the phospholipase D signalling pathway suggests control of glutamate vesicle trafficking across symbiotic compartments, and phosphorylation of the endoplasmic reticulum machinery suggests recycling of symbiosome-associated proteins. Our study shows for the first time that changes in the phosphorylation status of proteins between aposymbiotic and symbiotic Aiptasia anemones may play a role in the regulation of the cnidarian-algal symbiosis. This is the first phosphoproteomic study of a cnidarian-algal symbiotic association as well as the first application of quantification by data-independent acquisition in the coral field.

K E Y W O R D S
cnidarian-algal symbiosis, data-independent acquisition, mass spectrometry, phosphorylation-mediated protein signalling, quantitative phosphoproteomics

| INTRODUC TI ON
Scleractinian corals are reef-building architects that provide the foundation of one of the most diverse marine ecosystems, coral reefs (Reaka-Kudla, Wilson, & Wilson, 1997). Corals are cnidarian animals that rely upon a functional symbiosis with their intracellular, photosynthetic dinoflagellate algae in the family Symbiodiniaceae (Davy, Allemand, & Weis, 2012;LaJeunesse et al., 2018). The endosymbiotic algae supply the coral with metabolites by translocating substantial amounts of photosynthetically fixed carbon to the host (Davy, Lucas, & Turner, 1996;Kopp et al., 2015;. In return, the coral provides lipids, amino acids and substrates for photosynthesis (Imbs, 2013;Wang & Douglas, 1998), as well as a light-rich and sheltered environment to the microalgae (Enríquez, Méndez, & Prieto, 2005;. This symbiotic interaction is essential for the coral holobiont, which encompasses a broad suite of coral-associated microbes (e.g., bacteria, archaea, viruses and fungi), allowing it to thrive and flourish in oligotrophic waters as well as to support highly biodiverse ecosystems (Jaspers et al., 2019;Rohwer, Seguritan, Azam, & Knowlton, 2002).
Coral reefs are globally under threat due to environmental and anthropogenic pressure. The causes of coral decline, among others, are increasing sea surface temperatures, ocean acidification and water eutrophication (Hoegh-Guldberg et al., 2007;Rädecker, Pogoreutz, Voolstra, Wiedenmann, & Wild, 2015), which can lead to the breakdown of the coral-algal association and subsequent loss of the microalgae, a process known as coral bleaching. It is therefore crucial to better understand the molecular mechanisms and regulation of the coral-algal association. Notably, studying the cnidarian-dinoflagellate symbiosis in corals is difficult for several reasons, including the challenges with maintaining corals in aquarium settings, long reproductive cycles and the obligate dependency on their symbiotic association, that is, the absence of a "control" aposymbiotic (symbiont-free) state (Voolstra, 2013).
Aiptasia is easily reared in laboratory tanks, has a short asexual reproductive cycle and can be kept in an aposymbiotic state, enabling direct comparison between symbiotic and aposymbiotic conditions. In corals, this type of comparison is only possible in larvae, which are commonly aposymbiotic and acquire their symbionts horizontally (Dubinsky & Stambler, 2010;Schnitzler & Weis, 2010;Voolstra et al., 2009).
Comparison of aposymbiotic and fully symbiotic Aiptasia at the transcriptomic, proteomic and metabolic level revealed regulation of host biological processes that include metabolism, nutrient transport, energy storage and vesicle trafficking (Baumgarten et al., 2015;Ganot et al., 2011;Lehnert et al., 2014;Matthews et al., 2017;Oakley et al., 2016;Sproles et al., 2019). The marked overlap in the patterns at different molecular levels suggests that these biological processes are likely to be critical in defining the symbiotic phenotype. Because proteins are the functional unit of the cell and more directly relate to the prevalent phenotype than expressed genes, we investigated the role of protein signalling in the cnidarian-algal interaction. We focused on protein phosphorylation, a post-translational modification (PTM) with a prevalent role in the control of protein activity and signal transduction (Aivaliotis et al., 2009;Humphrey, James, & Mann, 2015).
In eukaryotes, at least one-third of the total proteome is phosphorylated mainly on hydroxyl amino acids (serine, threonine, tyrosine), and highly conserved cellular processes are thereby regulated (Jünger & Aebersold, 2014;Ochoa et al., 2016). Among others, protein phosphorylation is commonly employed to accommodate reversible adjustments in metabolism, sugar transport, oxidative stress responses, control of heat-shock and adaptation to light (Cozzone, 2005). Phosphorylation has also been described in pathogenic symbioses, where changes in the host cellular structure and function occur as a consequence of intracellular pathogens surviving within the invaded cell (Cozzone, 2005;Schmutz et al., 2013). In this regard, more recent work in corals alludes to "selfish algal symbionts" (Aranda et al., 2016;Nielsen, Petrou, & Gates, 2018;Rädecker et al., 2018;Sproles et al., 2019). Thus, protein phosphorylation is an attractive target to explore in order to better understand the mechanistic underpinnings of the cnidarian-algal interaction.
In this study, we used mass spectrometry (MS)-based phosphoproteomics to unravel the phosphorylation dynamics in the cnidarian-algal association. To do this, we first generated a reference phosphoproteome spectral library using data-dependent acquisition (DDA or shotgun proteomics), which provided comprehensive phosphoproteome coverage across aposymbiotic and symbiotic conditions in Aiptasia polyps. The spectral library served to retrieve the phosphopeptide identity during quantification using DIA (data-independent acquisition)-based SWATH-MS (sequential windowed acquisition of all theoretical fragment ion mass spectra), known to enable highly accurate and reproducible quantification of protein phosphorylation. To allow for discrimination between translational regulation (de novo protein synthesis) and PTM (addition/removal of phosphate groups to existing proteins), we additionally generated a total proteome data set used for normalization of phosphorylation against protein amount. In doing so, we found characteristic phosphoproteomic profiles distinguishing aposymbiotic and symbiotic states and differential phosphorylation targeting biological processes that have not been previously described in the context of the cnidarian-algal symbiosis, namely "phospholipase D signalling pathway" and "protein processing in the endoplasmic reticulum." We suggest that changes in the phosphorylation status of these signalling pathways have a putative role in the control of an established cnidarian-algal association. More broadly, the finding of extensive regulation at the level of the phosphoproteome may argue for dynamic control of the symbiotic state.

| Aiptasia rearing
Aposymbiotic and symbiotic Aiptasia anemones of the clonal strain CC7 (Baumgarten et al., 2015) were reared in incubators at 25°C, in 0.5-1-L tanks filled with freshly collected, autoclaved, 0.22-μmfiltered (Cat. GSWP04700; MF-Millipore) seawater (AFSW) from the Red Sea with salinity adjusted to 37 PSU, and fed Artemia twice weekly. Aposymbiotic Aiptasia were generated by repetitive cold-shock of 4-hr cycles in AFSW at 4°C and treatment with 50 μm of the photosynthetic inhibitor diuron (Cat. D2425; Sigma-Aldrich). Aposymbiotic anemones were kept in a dark incubator for more than 1 year and regularly examined for residual symbiont recolonization/replication under a fluorescence microscope (Leica DMI3000 B) ( Figure S1). Part of the aposymbiotic population was recolonized with the Breviolum minutum strain SSB01 (clade B) (LaJeunesse et al., 2018;Xiang, Hambleton, DeNofrio, Pringle, & Grossman, 2013). We additionally assessed symbiont presence/ absence in Aiptasia batches by quantitative polymerase chain reaction (qPCR) using ITS2 primers that discriminate between Symbiodinium spp. (clade A) and B. minutum (clade B). We tested for these two clades because the former is the native symbiont of CC7, while the latter was used for recolonization. In both instances, the qPCR did not produce amplicons with 30 cycles, thus confirming that animals were aposymbiotic (Table S1). Full recolonization occurred in about 30 days at 12 hr of light exposure. To avoid Artemia contaminations, the food supply was ceased 5 days before Aiptasia sampling.

| Tissue lysis, protein extraction and digestion
The detailed protocol describing sample preparation and data acquisition for both phosphopeptide spectral library generation and phosphopeptide quantification in Aiptasia is publicly available (Simona, Zhang, & Voolstra, 2018). Briefly, for the spectral library, a pool of 20 small-sized (2-mm basal disc) anemones per experimental condition (aposymbiotic and symbiotic) were collected and separately processed throughout the entire protocol. For quantification, each biological replicate per condition (five aposymbiotic, five symbiotic) was constituted by a pool of 10 small-sized anemones. Samples were lysed in 8 M urea buffer supplemented with protease (Cat. 4693159001; Roche Applied Science) and F I G U R E 1 (a) Aiptasia sample preparation workflow for phosphoproteomics. Proteins were extracted from aposymbiotic and symbiotic Aiptasia polyps and digested in-column. For spectral library generation, peptides were HPLC-fractionated before phosphopeptide enrichment, and MS acquisition was performed in DDA mode. For phosphopeptide quantification no HPLC fractionation was performed and acquisition was performed in DIA/SWATH-MS. (b) Data processing and analysis workflow. The DDA raw files were imported into MaxQuant for phosphopeptide search and used for spectral library generation in Spectronaut X. The quantitative DIA data were imported into Spectronaut X for retrieval of phosphopeptide identity by spectral library match. Biological and statistical data analysis was conducted in Perseus

| Peptide fractionation
Deep coverage of the phosphopeptide spectral library was obtained by high pH reversed-phase fractionation according to the protocol of Batth & Olsen (2016). Briefly, an XBridge Peptide BEH C18 column (Cat. 186003570; Waters Corp.) was connected to a Surveyor Plus high-performance liquid chromatography (HPLC) system (Thermo Fisher Scientific) and a 120-min gradient of buffer B (90% ACN, 5 mM ammonium hydroxide) in buffer A (5 mM ammonium hydroxide) was designed as follows: from 0%-5% to 25% of buffer B within 60-90 min, from 25% to 40% of buffer B within 5-10 min, ramp up to 60%-70% of buffer B within 5 min, maintain at 70% for 5 min before ramping down to 0%-5% of buffer B. A total of 110 fractions were collected and reconcatenated orthogonally (by mixing different parts of the gradient) into 25 fractions, and their volume was reduced to ~50 µl with a SpeedVac system (Thermo Fisher Scientific).

| Phosphopeptide enrichment
TiO 2 bead-based phosphopeptide enrichment was performed according to the protocol of Engholm-Keller et al. (2012). First, desalted and/or fractionated peptides were brought to a volume of 1 ml with 80% ACN, 5% TFA and 1 M of freshly added glycolic acid. Then, peptides and TiO 2 beads (Cat. 502075000; GL Sciences) were incubated at a 1:6 ratio (w/w), shaking for 10 min in a vortex mixer. After supernatant removal, the TiO 2 -bound phosphopeptides were washed in 80% ACN/1% TFA followed by 20% ACN/0.5% TFA. After complete evaporation of the solvent, the phosphopeptides were eluted from the dry TiO 2 beads by incubation in 4% ammonia solution in water, pH 11 (32% ammonia stock solution, Cat. 21192.323; VWR), rocking for 10 min. The phosphopeptide-rich supernatant was acidified with formic acid (FA) (14% by vol.) and TFA (4% by vol.) and transferred on top of a homemade C18 STAGE tip assembled and preconditioned according to the protocol developed by Rappsilber, Mann, and Ishihama (2007).
After two washes with 0.1% TFA in water, phosphopeptides were eluted twice in 75% ACN/0.1% TFA and completely dried via a SpeedVac.

| Phosphopeptide preparation for LC-MS/MS
Prior to liquid chromatography coupled to tandem mass spectrom-

| LC-MS/MS acquisition in DDA
Prior to DDA MS, phosphopeptides were separated with an UltiMate 3000 RSLCnano UHPLC system (Thermo Fisher Scientific) on a 25-cm Acclaim PepMap 100 C18 column (Cat. 164261; Thermo Fisher Scientific) at a flow rate of 300 nl/min, and over the following 75-min gradient of buffer B (0.1% FA in 80% ACN) in buffer A (0.1% FA, 0.1% ACN, in LC-MS grade water): from 5% to 40% of buffer B within 55 min, ramp up to 90% of buffer B within 5 min and maintain at 90% for 5 min, ramp down to 2% of buffer B and maintain for another 10 min. The nano-LC (nLC) system was connected to a Q-Exactive HF mass spectrometer (Thermo Fisher Scientific). The phosphopeptides were injected in the spectrometer with a Nonospray Flex ion source with electrospray potential of 1.9 kV and the temperature of the ion transfer tube set at 275°C. A full MS scan (350-1,400 m/z) was acquired at a resolution of 60,000 (at 200 m/z) in profile mode and with a target value of 3 × e 6 . The maximum ion accumulation time was set at 100 ms, and charge state screening for precursor ions was activated. The 10 most intense ions (threshold > 2 × e 4 ) carrying multiple charges were selected for fragmentation by higher energy collision dissociation (HCD). The dd-MS2 resolution was set at 15,000 and dynamic exclusion for HCD fragmentation at 20 s.
Other settings for fragment ions included a maximum ion accumulation time of 100 ms, target value of 1 × e 5 , normalized collision energy at 28%, and isolation width of 1.8.

| LC-MS/MS acquisition in DIA/SWATH-MS
Prior to DIA/SWATH-MS, phosphopeptides were separated on a similar nLC system as for DDA, but connected to a 50-cm EASY- Trypsin/P was selected as the digestion enzyme, allowing a maximum of two missed cleavages. Phosphopeptide identity retrieval was done by sequence alignment against the Aiptasia genomic gene set database version 1.0 (Baumgarten et al., 2015), containing 29,269 entries. We also queried the phosphopeptides against the B. minutum genomic gene set (Aranda et al., 2016), which confirmed the presence of some B. minutum protein groups in symbiotic Aiptasia, the majority of those being notably absent in aposymbiotic samples ( Figure S5, Dataset S5). The MaxQuant search output was imported into Spectronaut X (Cat. Sw-3001; Biognosys) for spectral library generation using default settings: minimum fragment ions length of three amino acids, fragment ions m/z range of 300-1,800, minimum relative intensity of 5%, and three to six fragment ions per precursor peptide. The DDA raw data and the spectral library are available via ProteomeXchange with identifier PXD014076.

| DIA data processing
DIA raw data files were imported in Spectronaut X and the generated spectral library was assigned for identity retrieval of the quantified phosphopeptides, using default settings. Extraction and scoring of MS1 and MS2 mass tolerance as well as the retention time (RT) for the extracted ion chromatogram (XIC) were set to dynamic and correction factors of 1 were applied. Calibration of iRT standards was done automatically by local (non-linear) regression. Mutated decoys (one to three random amino acids swap) were generated at a library size fraction of 0.1, a q-value cutoff for precursor and protein identification was set to 0.01, and single protein hits were defined by stripped peptide sequences. Interference correction was applied so that the least interfering three fragment ions and two precursor ions were kept. All fragment ions not removed during interference correction were used for quantification, calculated from the area under the curve between the XIC boundaries of each targeted ion.
Label-free normalization was applied on the whole data set to minimize the effect of the potential variability generated by the sample preparation and the LC-MS performance. Spectronaut X uses an algorithm based on the Local Regression Normalization described by Callister et al. (2006). This normalization approach assumes that the majority of the precursors in an experiment are not regulated (stable background), and when regulated, there is no preference for upor down-regulation. The q-value sparse defines the normalization strategy; all precursors that passed the q-value threshold at least in one sample across the experiment were used for normalization.

| Data analysis
Biological and statistical analysis of DDA and DIA data was performed in the Perseus environment (version 1.6.2.1) (Tyanova et al., 2016). The MaxQuant search output of the 44 DDA files used to generate the phosphoproteome spectral library was imported in Perseus to retrieve the number of detected phosphosites, the ratio of phosphorylated amino acid residues, the phosphosite localization probabilities and the kinase motifs assignments. The phosphosite localization probabilities were assigned according to Olsen et al. (2006). That is, class I phosphosites were confidently assigned with probability above 0.75. Classes II and III had localization probabilities between 0.25 and 0.75, but to the former a matching kinase motif was assigned. No filtering based on phosphosite probability was made when generating the spectral library, because confident phosphosite assignment in the spectral library does not imply high probability assignment in the DIA data. As currently there is no control of correct phosphosite assignment for DIA in Spectronaut X, we relied on retention time shift of different phosphopeptides as well as manual validation of the spectra of interest.
The quantitative DIA data matrix covered 75% of the phosphopeptides compiled in the spectral library. The matrix was quality-filtered so that only phosphopeptides with at least 70% of valid values across samples were considered; this reduced the phosphopeptide coverage to 62%. Prior to biological and statistical analysis,

| The Aiptasia phosphoproteome
To quantify protein phosphorylation in Aiptasia by DIA/SWATH-MS, we first generated a phosphopeptide spectral library of highly frac-  (Figure 2b). When the phosphopeptide identity could not be confidently assigned due to the high sequence homology among phosphoproteins, these phosphoproteins clustered in the same PG. Phosphoproteins belonging to the same PG usually shared function, family, or presented redundant annotation. Phosphorylation occurred on the hydroxyl amino acid residues serine (Ser), threonine (Thr) and tyrosine (Tyr) at 85.7%, 13.5% and 0.8%, respectively (Figure 2c). Phosphorylation was confidently assigned to 61.7% of the amino acid residues (class I phosphosite, p > .75), whereas 33.4% of the phosphosites were assigned with a lower probability (.25 < p < .75) (Figure 2d). Nevertheless, for most of the low-probability assignments a matching kinase binding motif was retrieved (class II phosphosites). Only in 1.4% of the assignments no phosphorylating kinase motif was identified (class III phosphosites).

| Quantification of protein phosphorylation differences across symbiotic states
We measured protein phosphorylation in aposymbiotic and symbiotic Aiptasia by DIA/SWATH-MS and quantified 3,779 phosphopeptides, namely 75% of the phosphopeptides compiled in the spectral library ( Figure 3a). For downstream analysis, we only retained phosphopeptides that were quantified in at least 70% of the experimental samples (i.e., present in at least 7 of the 10 samples). This reduced the matrix to 3,105 phosphopeptides (62% of the spectral library) and 1,641 phosphoproteins clustering into 1,206 PGs (Dataset S1).
We quantified protein phosphorylation in aposymbiotic and symbiotic anemones based on phosphopeptide abundance and visualized sample clustering by PCA (Figure 3b control." We next investigated which proteins and signalling pathways were targeted by differential phosphorylation.

| Differential protein phosphorylation across symbiotic states group into distinct biological processes
Differential phosphorylation of proteins was statistically inferred by two-sided t-test (FDR 0.01, S0 0.1). The 644 differentially abundant phosphopeptides totalling 539 phosphoproteins (clustering into 390 PGs) were represented on a Volcano plot to illustrate highly abundant phosphopeptides in the symbiotic or aposymbiotic condition ( Figure 4a and Dataset S1). To distinguish cases of differential phosphorylation from changes in phosphoprotein expression, we generated a total proteome data set from the same aposymbiotic and symbiotic replicates used for phosphoproteomic analysis (Figures S1 and S2, Dataset S3) and we normalized phosphopeptide amount against total protein abundance. This was done for all the differentially phosphorylated proteins highlighted by the KEGG enrichment analysis (Figure 4b). According to the KEGG enrichment analysis, among the biological processes and signalling pathways that were differentially phosphorylated (p < .05), we found "carbohydrate metabolism," "protein digestion and absorption (ko04974)," "phospholipase D signalling pathway (ko04072)," "protein folding, sorting and degradation" and "protein processing in endoplasmic reticulum (ER) (ko04141)" (Dataset S2). Nevertheless, the total proteome normalization revealed that differential phosphorylation of proteins belonging to the KEGG terms "carbohydrate metabolism" and "protein digestion and absorption" co-occurred with changes in the total amount of these respective proteins. Examples were key metabolic enzymes (e.g., ATP-citrate synthase, AIPGENE6154; acetyl-CoA carboxylase, AIPGENE11605; long-chain-fatty-acid-CoA ligase 1, AIPGENE7314) and amino acid transporters (e.g., b(0,+)-type amino acid transporter 1, AIPGENE1716 and AIPGENE15008; rBAT, AIPGENE 7824). Because discrimination between translational and post-translational changes is critical for the inference of biological regulation, we focused on the KEGG terms that were target of differential phosphorylation at constant protein abundance.

| The phospholipase D signalling pathway
In the KEGG enrichment analysis, the "phospholipase D signalling pathway (ko04072)" was highly enriched (enrichment factor 3.12), and thus targeted by differential phosphorylation in symbiosis. The   Phosphorylation occurred at Ser849-p and Ser854-p, in the predicted cytosolic terminus following the 7th transmembrane domain.

F I G U R E 3
In silico kinase motif assignment indicated that Ser849-p but not Ser854-p on Aiptasia mGluR4 is a substrate for PKC-mediated phosphorylation. Downstream of mGluR4, the PLC subunit β1 (PLCβ1, AIPGENE26754) was also highly phosphorylated at Ser851-p in symbiosis.
The PLD pathway cross-regulates multiple signalling cascades horizontally and vertically. We found differential phosphorylation of effector molecules belonging to various PLD-interacting pathways. Tuberin (TSC2, AIPGENE9407) is a regulator of the mTOR pathway that also has a role in microtubule-mediated protein transport and in the regulation of the GTPase activity of the Ras-related protein Rab5. TSC2 was highly phosphorylated at Ser1223-p in symbiosis. Dynamin-1 (DYN1, AIPGENE26789), a microtubule-associated phosphoprotein involved in vesicular trafficking, was highly phosphorylated in symbiosis at Ser778-p, which is a conserved phosphosite in proximity to the mammalian regulatory phosphosite Ser774-p (not present in Aiptasia). Additionally, our proteome data set showed enrichment of the process "nervous system" ( Figure S4b and

| Protein processing in the endoplasmic reticulum
Another highly enriched pathway (enrichment factor 2.50) in the KEGG analysis was "protein processing in ER (ko04141)," a pathway that belongs to the higher order process "folding, sorting and degradation," which was also slightly enriched (enrichment factor 1.78). Interestingly, the process "folding, sorting and degradation"   subunit δ (SSRD, AIPGENE5652) and of the translocation protein SEC62 (AIPGENE12061, +0.49-fold). Higher phosphorylation was also observed at Thr367-p on the lectin-type chaperone calnexin (cnx, AIPGENE23380), a Ca 2+ -binding protein with a major function in the quality control of protein folding. Properly folded proteins are sorted to the Golgi (secretory pathway), whereas misfolded proteins undergo ER-associated degradation (ERAD). Across symbiotic states, we found substantial differential phosphorylation of the ERAD apparatus. For instance, we found that dephosphorylation of the E3 ubiquitin-protein ligase AMFR (AMFR, AIPGENE14825) at Ser503-p. AMFR is part of the VCP/p97-AMFR/gp78 complex, which is involved in the final step of ERAD. The NSFL1 cofactor p47 (NSF1C, AIPGENE8167), a regulator of the ATPase activity of the above-mentioned VCP (transitional endoplasmic reticulum ATPase), was highly phosphorylated on Ser203-p, whereas the translocating chain-associated membrane protein 1-like 1 (TR1L1, AIPGENE17645) and the heat shock protein HSP 90-alpha 1 (H90A1, AIPGENE3199) were dephosphorylated on Ser360-p and Ser230-p, respectively.

| D ISCUSS I ON
Our work represents the first phosphoproteome generated for a cnidarian-algal symbiotic association as well as the first application of label-free phosphoproteomic quantification by DIA/SWATH-MS in the coral field. As such, we focused on the generation of a comprehensive phosphopeptide spectral library for Aiptasia, namely a collection of phosphopeptide spectra that serves as a reference database for phosphopeptide identification during quantification in DIA/ SWATH-MS. The generation of a spectral library that equally represents the aposymbiotic and symbiotic states (Figure 2a) was relevant to avoid quantification biases deriving from heterogeneous phosphopeptide representation. The ratio of phosphorylated amino acid residues (Figure 2c) was comparable to that found in other organisms (Olsen et al., 2006;Robles, Humphrey, & Mann, 2017), and in most of the cases, phosphosites were confidently assigned or identified as a substrate of known protein kinases (Figure 2d). Quantification in DIA/SWATH-MS revealed that the phosphoproteomic profile of the samples reflected their symbiotic state (Figure 3b,c). Of the F I G U R E 5 Model of a symbiotic cnidarian cell representing putative phosphorylation-mediated regulation of the KEGG terms "PLD signalling pathway" and "protein processing in ER." For each differentially phosphorylated protein, the phosphosite is represented in a circle as phosphorylated (yellow) or dephosphorylated (grey) in symbiotic versus aposymbiotic Aiptasia. For the sake of completeness, the model includes additional proteins that belong to the two pathways. Bold font indicates proteins that were measured but not differentially abundant. The colored font represents upregulation (red) and downregulation (green) of protein abundance in symbiosis  (Figure 4a), the enrichment analysis showed that only 84 clustered into KEGG terms and therefore contributed to the enrichment. This suggests that differential phosphorylation occurs across a broad range of proteins rather than pre-dominantely targeting selected biological processes or signalling pathways. Here we focused on the differentially phosphorylated proteins highlighted by the enrichment analysis in order to describe potential signalling pathways associated with coral-algal symbiosis.
The differential abundance of a measured phosphopeptide can result either from changes in phosphorylation at a specific phosphosite or from variation in the translated protein (at constant phosphosite occupancy or stoichiometry), or a combination of the two events (Wu et al., 2011). Among the phospho-regulated proteins, we promptly recognized some for which regulation at the level of protein expression during symbiosis was previously described (Oakley et al., 2016). In such cases, differential phosphopeptide abundance probably reflected an increase in total phosphoprotein amount (by de novo protein synthesis) rather than differential phosphorylation. Although such cases did not allow for sensu stricto discrimination between translational and post-translational change, the presence of phosphoproteins belonging to the KEGG terms "carbohydrate metabolism" and "protein digestion and absorption" represented a "sanity check" of our data set. This is because genes/ proteins associated with "carbohydrate metabolism" and "protein digestion and absorption" (or amino acid transport) have been characterized at the transcriptional/translational level in symbiotic cnidarians and showed adjustment to the availability of photosynthetically fixed carbon translocated by the endosymbiotic algae as well as increased amino acid synthesis and transport (possibly to the algae for primary production) (Kopp et al., 2015;Lehnert et al., 2014;Oakley et al., 2016;Rädecker et al., 2015). Here we suggest that proteins associated with these processes are targeted by phosphorylation in cnidarians. Indeed, some phosphosites on these proteins are conserved and have known regulatory function in higher vertebrates (Berg, Tymoczko, Stryer, & Clarke, 2002;Frahm, Li, Grevengoed, & Coleman, 2011;Ray, Suau, Vincent, & Dalla Venezia, 2009), suggesting their possible relevance also in Aiptasia.
Among the KEGG pathways that were differentially phosphorylated (at constant protein abundance) in symbiosis, we found the "PLD signalling pathway". To our knowledge, this pathway has not been previously described in the context of the coral-algal symbiosis.
The PLD pathway can be activated by the metabotropic glutamate receptor mGluR4 that self-phosphorylates upon glutamate binding. In symbiosis, we found higher phosphorylation of mGluR4 at phosphosite Ser849-p, a potential substrate for PKC-mediated phosphorylation and a binding site for Ca 2+ /calmodulin (CaM) ( Figure 5). Based on sequence alignment, this phosphosite resides in a sequence stretch that is not present in the mGluR4 of higher vertebrates. However, mGluR4 is closely related to mGluR7, the most conserved mGluR subtype in mammals (Nakajima, Yamamoto, Nakayama, & Nakanishi, 1999). Interestingly, not only is Ser849-p conserved between Aiptasia mGluR4 and the mammalian mGluR7, but also phosphorylation at this phosphosite in mGluR7 is mediated by PKC and inhibited by the binding of Ca 2+ /CaM (Nakajima et al., 1999). This suggests that similarly to mGluR7 in higher vertebrates, mGluR4 in Aiptasia may regulate the PLD signalling transduction cascade in a PKC/Ca 2+ /CaM/PLC-dependent manner. The phospholipase PLCβ1 was indeed also highly phosphorylated in symbiosis. In higher vertebrates, PLCβ1 can be phosphorylated at multiple sites, although only phosphorylation at Ser982-p has yet been shown to positively regulate its enzymatic activity, and thus the formation of the second messenger molecules diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP3) (Xu et al., 2001). Interestingly, PLC and its reaction products, DAG and IP3, are also involved in the "phosphatidylinositol signalling system," process that was previously described in the regulation of endocytosis in other symbiotic systems and implied in the modulation of the coral-algal interaction (Peleg-Grossman, Volpin, & Levine, 2007;Rosic et al., 2014).
In the symbiotic state, we also found phosphorylation of effector proteins belonging to interacting pathways downstream of PLD.
Intriguingly, our total proteome data set revealed upregulation of both endophilin and VGLUT1, as well as differential expression of several components of the large SNARE protein complex that has a major function in vesicular trafficking and glutamate release.
Together, these findings suggest that in the cnidarian-algal symbiosis the PLD signalling pathway may be activated by the binding of glutamate to its receptor mGluR4, triggering a downstream phosphorylation cascade that regulates vesicular trafficking across symbiotic compartments. It was shown that mGlur4 could localize on cytoplasmic vesicles (Iacovelli et al., 2004), perhaps sensing host cytosolic glutamate concentrations and activating its transport to the symbiosome. This would imply that glutamate regulates its transport inside VGLUT1 vesicles across symbiotic compartments, thus reconfirming its importance in the recycling of nitrogen and modulation of the cnidarian-algal symbiotic interaction, especially given the nitrogen-poor state of reef waters (Macdonald, Lin, Russell, Thomas, & Douglas, 2012).
Another pathway that was differentially phosphorylated (at constant protein abundance) in symbiosis was "protein processing in ER." Here we found higher phosphorylation of some components (rrbp1, SSRA) of the translocon, the protein complex responsible for the import of newly synthesized proteins from cytosolic ribosomes into the ER ( Figure 5). Phosphorylation occurred together with the upregulation of some other translocon components (SSRD, Sec62), as well as phosphorylation of the lectin-type chaperone calnexin, a key player in the control of protein fate (Bergeron, Brenner, Thomas, & Williams, 1994). Also, several components of ERAD (AMFR, NSF1C, TR1L1, H90A1) were differentially phosphorylated in symbiosis. Notably, Ser503-p of AMFR is conserved in higher vertebrates (corresponding to Ser522-p). In human cell lines, AMFR phosphorylation at a nearby site (Ser516-p) triggers its ubiquitination and degradation (Wang et al., 2018). Indeed, while properly folded proteins leave the ER via the secretory pathway (i.e., proteins are packed into vesicles and delivered to the Golgi for further distribution to target compartments), misfolded proteins are tagged for ubiquitination in the ER and retrotranslocated to cytosolic proteasomes for degradation via the ERAD apparatus (Fumagalli et al., 2016). The coordinated phosphorylation and upregulation of different components of the ER machinery during symbiosis may indicate sustained import of newly synthesized proteins in the ER for folding and delivery to other compartments via the secretory pathway, as well as high protein turnover via degradation.
Differential regulation of the ER machinery in the cnidarian-algal symbiosis has not been extensively described before, in agreement with the notion that we detected enrichment at the post-translational but not at the translational level. Oakley et al. (2017) showed proteostasis disruption and ER stress in Aiptasia under thermal-shock, suggesting that damage of this compartment in cnidarians may be involved in symbiosis breakdown. Additionally, none of the identified phosphosites has been functionally characterized in other organisms, and therefore no direction of regulation could be inferred in our system. We found some literature on the relevance of the protein secretory pathway (ER, Golgi and vesicle trafficking) in symbiotic metazoans and plants (Wang & Dong, 2011;Wang et al., 2010). This pathway is highly conserved among eukaryotes, and sessile organisms such as plants rely on the secretory pathway to adjust to changing environmental conditions, avoid pathogen infection, and achieve symbiosis (Wang & Dong, 2011). Intriguingly, in plant-root microbial symbioses the secretory pathway is used by the host to deliver proteins to the symbiosome (or periarbuscolar) membrane. Although not strictly secreted, these proteins would still be processed in the ER and Golgi, packed into vesicles and directed to the symbiosome via cytoskeleton and associated motor proteins (Wang & Dong, 2011). In concert, regulation of multiple components of the secretory pathway suggests its essential function in maintaining the complex mosaic of proteins that decorate the symbiosome (Maunoury et al., 2010;Pumplin & Harrison, 2009;Wang et al., 2010). In the cnidarian-algal symbiosis, the differential phosphorylation of various components of the ER machinery and its implication in protein translocation, folding and degradation could suggest a potentially important regulatory function of this compartment in the recycling and degradation of the large suite of proteins necessary to maintain the symbiosome as an arrested phagosome.
Overall our findings highlight changes in the phosphorylation status of PLD signalling and of protein processing in the ER, suggesting that phosphorylation may be a relevant regulatory mechanism of these pathways during an established cnidarian-algal interaction.
This supports the concept of "dynamic stability" of the cnidarianalgal association (Davy et al., 2012). This dynamism may reflect a system in continuous readjustment with possible implications for the stability of the association, which can be disrupted at every stage if no longer fulfilling the requirements of the symbiotic partners.
Clearly, all our findings need further functional validation.
Because protein phosphorylation is a reversible modification,

| CON CLUS IONS
This is the first phosphoproteomic study of a cnidarian-algal symbiotic association as well as the first application of label-free quantification by DIA/SWATH-MS in the field. The phosphopeptide spectral library is a publicly available resource for cross-laboratory quantification studies and the quantitative data set can be used to explore regulation of more than 1,000 phosphoproteins in the cnidarian-algal symbiosis. Quantification of protein phosphorylation in aposymbiotic and symbiotic Aiptasia revealed changes in the phosphorylation status of the PLD signalling pathway and protein processing in the ER, revealing for the first time a putative layer of post-translational regulation in a cnidarian-algal symbiotic interaction. We anticipate that the establishment of the DIA/SWATH-MS method in the field will pave the way to further studies and more sophisticated experimental designs and will help addressing new biological questions in the context of the cnidarian-algal symbiosis.

ACK N OWLED G EM ENTS
The research reported in this publication was supported by the King Abdullah University of Science and Technology (KAUST). We thank the Core Labs at KAUST for collaborating on the project, Rúben Costa and Hagen Gegner for reviewing the manuscript, and Hagen Gegner for providing the Aiptasia pictures for Figure 1.