The sphingosine rheostat is involved in the cnidarian heat stress response but not necessarily in bleaching

Many cnidarians, such as corals and sea anemones, engage in endosymbiotic associations with photosynthetic dinoflagellates in the genus Symbiodinium spp. These partnerships are the centerpiece of coral reef ecosystems and are severely threatened by abiotic stressors associated with climate change, especially elevated temperature (Hoegh-Guldberg et al., 2007). The algal photosynthetic apparatus is sensitive to climate-induced heat and light perturbations, which result in photosystem dysfunction (Warner et al., 1996), and increased reactive oxygen and nitrogen species (ROS and RNS) generation in the algae (Hawkins and Davy, 2012; Lesser, 1996; McGinty et al., 2012) and in turn oxidative stress in the host (Hawkins et al., 2014; Perez and Weis, 2006; Richier et al., 2006). Symbiotic cnidarians have protective mechanisms including heat shock proteins (Hsps), oxygen-scavenging enzymes and fluorescent proteins to reduce the effects of moderate fluctuations in heat and light exposure (Louis et al., 2017; Venn et al., 2008). However, during prolonged heat or light stress, these stress mechanisms can become overloaded, ultimately leading to a massive loss of symbionts, or dysbiosis, from cnidarian hosts, resulting in cnidarian bleaching. Although there have been extensive studies on dysbiosis in a variety of corals and sea anemone model systems (reviewed by Weis, 2008), our understanding of the cellular events preceding symbiont loss is limited. Given that lipids are co-directors of phagocytosis (Heung et al., 2006; Steinberg and Grinstein, 2008), the mode by which Symbiodinium are acquired by hosts, dysbiosis could have significant impacts on lipid signaling, targeting and trafficking events.

Oxidative and heat stress are two cellular events mediated by bioactive sphingolipids in both eubacterial and eukaryotic cells (An et al., 2011; Jenkins et al., 1997; van Brocklyn and Williams, 2012). The lipid messengers sphingosine (Sph) and sphingosine-1-phosphate (S1P) have antagonistic roles in the cell, where Sph initiates programmed cell death pathways while S1P activates cell survival and proliferation (Le Stunff et al., 2004; Olivera and Spiegel, 2001; Spiegel and Milstien, 2003). The enzymatic activities of sphingosine kinase (SPHK) and sphingosine-1-phosphate phosphatase (SGPP) maintain the homeostatic balance of these sphingolipids in this so-called ‘sphingosine rheostat’ (Cuvillier et al., 1996; Mandala et al., 1998). Heat stress studies in mammalian cell lines, yeast and gut microbiome commensals Bacteroides suggest an evolutionarily conserved sphingolipid-mediated heat stress response (HSR) (An et al., 2011; Chang et al., 1995; Jenkins et al., 2002, 1997). The HSR is a defense mechanism controlled by stress sensors and signal transduction pathways to offset the deleterious effects of thermal stress (Hofmann and Todgham, 2010). In yeast and Bacteroides, elevated temperatures decreased cell viability in sphingolipid-deficient strains, and this was rescued with sphingolipid supplementation. Furthermore, in yeast and mammalian cells, synthesis of Sph and its precursor ceramide increased with heat stress (Chang et al., 1995; Jenkins et al., 1997). Elevated Sph levels were reversed in yeast by deletion of the SGPP-like enzyme, which increased the cellular S1P-like metabolites, thereby enhancing survival (Mandala et al., 1998; Mao et al., 1999).

Oxidative stress is also linked to the sphingosine rheostat through modulation of S1P levels in yeast (Lanterman and Saba, 1998), nematodes (Deng et al., 2008), fruit flies (Kawamura et al., 2009), fish (Yabu et al., 2008) and mammals (Ader et al., 2008; Gomez-Brouchet et al., 2007; Pchejetski et al., 2007). The rheostat response is dependent on the severity of oxidative damage, with moderate ROS levels activating SPHK and increasing S1P levels (Ader et al., 2008), while severe ROS levels degrade SPHK, reduce S1P and shift cells toward death (Gomez-Brouchet et al., 2007; Pchejetski et al., 2007). Furthermore, the activation of the Hsp90–endothelial NOS (eNOS) protein complex by S1P causes production of an RNS, cytotoxic nitric oxide (NO), which at low cellular levels can inhibit pro-inflammatory cytokines (De Palma et al., 2006; García-Cardeña et al., 1998) (Fig. 1A). In contrast, immune elicitors and elevated ROS trigger inducible NOS, resulting in high cellular NO and NO-dependent elevations in ceramide through activation of lysosomal acidic sphingomyelinases (A-SMase) (Perrotta et al., 2008) (Fig. 1B). Ceramides then assemble with cholesterol to form ceramide-rich rafts in cellular and organellar membranes, altering membrane fluidity and cell signaling toward cell death (Balogh et al., 2013). These findings suggest an active role for the sphingosine rheostat in the cellular stress response, and led Maceyka and coworkers (2007) to propose SPHK as a central enzyme in the oxidative stress pathway in mammals. Its importance in the invertebrate stress response, however, requires further investigation.

In symbiotic cnidarians, the molecular crosstalk between the host and symbionts during dysbiosis is still only partially understood. In stable associations, symbionts are housed in host-derived vesicles, or symbiosomes, that resist phagosomal maturation (Chen et al., 2004, 2003, 2005). Transcriptional studies on symbiotic cnidarians exposed to elevated temperature or high irradiance stress indicate that symbiont removal mechanisms such as host apoptosis, innate immunity and exocytosis are upregulated (DeSalvo et al., 2008; Richier et al., 2006; Starcevic et al., 2010). Modulation of sphingolipids could underlie these changes through activation of phagosomal maturation (Heung et al., 2006) and cell death (Maceyka et al., 2002). The cnidarian sphingosine rheostat has been implicated in symbiosis maintenance (Hemond et al., 2014; Rodriguez-Lanetty et al., 2006), immunity and the HSR (Detournay and Weis, 2011). In the sea anemone Aiptasia pallida (also known as Exaiptasia pallida; Grajales and Rodriguez, 2014), exposure to an elevated temperature with the addition of exogenous Sph increased algal loss by 40%, whereas incubation in exogenous S1P reduced bleaching by nearly 50% (Detournay and Weis, 2011). Moreover, elevated Sph led to increased host caspase activity, an indicator of programmed cell death. However, cytotoxic NO production, implicated in symbiosis breakdown (Hawkins et al., 2013; Perez and Weis, 2006; Ross, 2014) and functionally coupled to sphingolipid metabolism in other systems (Perrotta et al., 2008), was unaffected by Sph and S1P (Detournay and Weis, 2011).

Given that exogenous sphingolipids altered the amount of symbiont loss and host cell death in A. pallida during heat stress, we predicted that endogenous transcriptional regulation of rheostat enzymes (AP-SPHK and AP-SGPP) and concomitant changes in sphingolipid metabolites contribute to the initiation of dysbiosis and a host HSR. The sea anemone Aiptasia sp. is a widely used model system for the study of coral symbiosis physiology and cell biology (Goldstein and King, 2016; Weis et al., 2008). Its ease of culture, rapid growth rate and well-developed genomic resources (Baumgarten et al., 2015; Lehnert et al., 2012) make it a valuable resource for empirical studies of symbiosis that cannot be carried out in corals. In this study, we experimentally examined the response of the host sphingosine rheostat to elevated temperature through time by quantification of AP-SPHK, AP-SGPP and HSR biomarker Hsp90 gene expression, sphingolipid concentrations and symbiont loss over a prolonged heat stress lasting 7 days.

Symbiotic A. pallida (Agassiz in Verrill 1864) from the Weis Lab population (VWA; Grajales and Rodríguez, 2016) were maintained in artificial seawater (ASW) at room temperature (RT, ∼25°C) under 40 µmol quanta m⁻² s⁻¹ intensity on a 12 h light:12 h dark photoperiod to promote symbiont photosynthesis and proliferation. Anemones were fed Artemia sp. nauplii twice a week up to 2 weeks prior to the commencement of experiments, after which time they were starved. The symbionts, Symbiodinium minutum strain CCMP830 (type B1, Bigelow Laboratory for Ocean Science, East Boothbay, ME, USA) previously isolated from A. pallida (Lesser, 1996), were grown in sterile f/2 media (Guillard and Ryther, 1962) at 25°C at an irradiance of 40 µmol quanta m⁻² s⁻¹ on a 12 h light:12 h dark cycle.

Anemones were randomly placed into 6-well plates (for qPCR, three anemones per well; for lipid analyses, one anemone per well) with 10 ml of 0.45 µm-filtered artificial seawater (FASW) 1 week prior to treatment. During the 2 weeks before and throughout the week of temperature exposure, the anemones were starved, given daily FASW changes and any expelled symbionts were removed. All experiments were carried out in temperature-controlled incubators. For the qPCR assay, anemones were exposed to a range of temperatures – 25 (ambient), 27, 30 and 33°C – for 1 week. Anemones were sampled at 0.5, 1, 2, 4 and 7 days (n=3 per time point). Additional samples were collected at 3 and 6 h for ambient and 33°C to investigate early expression differences at these experimental temperatures. The intermediate elevated temperatures (27 and 30°C) did not result in detectable gene expression differences (see Results); therefore, only ambient and 33°C-treated anemones were used for quantification of sphingolipids. Anemones were sampled over two experiments, at 1, 4 and 7 days in experiment 1 (n=4) and 6 h and 7 days in experiment 2 (n=10) based on gene expression results. At designated time points, anemones were rinsed in FASW before being frozen in liquid nitrogen and stored at −80°C until processed for qPCR or sphingolipid extraction.

To quantify symbiont loss from host tissues during exposure to elevated temperature, symbiont chlorophyll autofluorescence was measured from the set of anemones that were destined for lipid analysis. Measurements were taken from each individual until they were killed at their designated time point. First, each anemone was imaged under white light on the Zeiss Stemi 2000-c stereo-microscope (Zeiss, Oberkochen, Germany) with standardized camera settings and 1 s exposure to determine well location and measure oral disc size. Autofluorescence was then detected using a long-pass emission filter (>665 nm) and digitally captured with a 25 s camera exposure after excitation by 470 nm LED blue light at 130 µmol quanta m⁻² s⁻¹ intensity. Mean fluorescence intensity for each individual at each time point was measured from 10 random areas of the oral disc or base of the tentacles using ImageJ v1.47 software (Rasband, 1997-2015). Autofluorescence was normalized to the mean fluorescence intensity from five areas adjacent to the anemones considered to be background levels. The random effect of repeated measures of autofluorescence from the same individual was significant in the generalized linear mixed-effects model (GLMM; Chi-square goodness of fit test, P<0.001), but not the well position in the plate. Therefore, normalized values were fitted with a GLMM model with fixed effects of time (continuous) and temperature (factor, two levels), and random effect of sample. Anemone size was estimated by their amount of total protein, a common proxy for A. pallida size in hyperthermal experiments (Bieri et al., 2016; Goulet et al., 2005; Sunagawa et al., 2008). For each sample, total protein was quantified using a Bradford assay from a 100 µl aliquot of the homogenized tissue from the lipid extraction described below. To determine whether host size affected the loss of symbionts, the percentage symbiont loss was calculated as: 1−(6 h mean fluorescence/0 h mean fluorescence)×100, and correlated to total protein using a Pearson's correlation test.

Total RNA was extracted from each anemone using an extraction protocol described by Poole et al. (2016), with TRIzol Reagent (Invitrogen, Carlsbad, CA, USA) and an RNeasy Mini kit (Qiagen, Valencia, CA, USA). The homogenization step in this extraction method does not specifically target S. minutum RNA; however, it may have been extracted with the host RNA.

To assess primer specificity for AP-SPHK and AP-SGPP with the qPCR assay, we also tested amplification of S. minutum cDNA. Cultured S. minutum RNA was extracted using a different combination of protocols from Rosic and Hoegh-Guldberg (2010) and an RNeasy Mini kit (Qiagen). Briefly, 15 ml of CCMP830 cells in exponential growth phase at a concentration of 10⁶ml⁻¹ was pelleted by centrifugation at 13,000 g for 5 min. The pellet was resuspended in 1.5 ml of f/2 media, transferred to a FastPrep screw-top tube (MP Biomedical, Santa Ana, CA, USA), and repelleted at 14,000 g for 30 s. The supernatant was removed and the pellet frozen in liquid nitrogen. Acid-washed 0.45–0.6 mm glass beads (Sigma-Aldrich, St Louis, MO, USA) and 0.5 ml TRIzol were added to the pellet, and placed in the FastPrep FP120 cell disruptor (MP Biomedical) at the low speed (4.0 s) for two 20 s pulses. The homogenate was then processed as described above for the anemone samples using a combination of the TRIzol protocol through phase separation followed by the RNeasy Mini kit (Qiagen).

RNA quality and quantity was assessed by gel electrophoresis and the NanoDrop ND-1000 (NanoDrop Products, Wilmington, DE, USA), respectively. Samples with a NanoDrop 260 nm/230 nm ratio less than 1.5 underwent the RNA purification protocol described by Poole et al. (2016). Each sample was then treated with the DNA-free kit (Invitrogen) or TURBO DNA-free kit (Invitrogen) to remove residual genomic DNA contamination. cDNA was synthesized using SuperScript III First-Strand Synthesis System (Invitrogen) with 500–750 ng of RNA and 50 μmol l⁻¹ oligo (dT)₂₀ primer. Each cDNA sample was diluted to a final concentration of 300 ng μl⁻¹ before use in the qPCR reactions. Aiptasia pallida (n=6) and CCMP830 (n=1) cDNA was tested for amplification of the symbiont-specific peridinin chlorophyll protein using previously published primers (Amar et al., 2008) following the GoTaq Flexi polymerase protocol (Promega, Madison, WI, USA).

Partial transcripts of AP-SPHK and AP-SGPP were identified from the A. pallida transcriptome (Lehnert et al., 2012) using tBLASTn with an expect value (E-value) ≤e−20 in Geneious v8.0.3 (Altschul et al., 1990; Kearse et al., 2012). The query protein sequences used were Rattus norvegicus SPHK1 (NCBI: ABF30968.1) and Homo sapiens SGPP (NCBI: CAC17772.1). To generate full-length cDNA sequences, we performed rapid amplification of the cDNA ends (RACE) using a FirstChoice RLM-RACE kit (Invitrogen) and RNA template from symbiotic A. pallida. Nested gene-specific primers (Table 1) were designed by hand for the 3′ and 5′ ends following the manufacturer's guidelines. Successful PCR products detected by gel electrophoresis were ligated into plasmids with the pGEM-T easy vector system (Promega) overnight at 4°C and cloned into MAX Efficiency DH5α competent cells (Invitrogen), following the manufacturer's protocol. Plasmids were purified using the QIAprep Spin Miniprep kit (Qiagen), product size verified with FastDigest EcoRI (Thermo Scientific, Waltham, MA, USA) or Promega EcoRI (Promega), and Sanger sequenced on the ABI 3730 capillary machine at the Center for Genome Research and Biocomputing at Oregon State University. Sequence products from RACE were assembled with the transcriptome fragments using Geneious v8.0.3 (Kearse et al., 2012). We searched for conserved protein domains of SPHK and SGPP (Kohama et al., 1998; Mandala et al., 2000), nuclear localization sequence of SPHK2 (Igarashi et al., 2003) and transmembrane segments using the Consensus Constrained TOPology prediction web server (Dobson et al., 2015).

While this project was in process, the A. pallida genome (v1.0) was sequenced (Baumgarten et al., 2015). The predicted gene models from the genome were searched using the full-length transcripts of AP-SPHK and AP-SGPP (generated as described above) as BLASTx queries to identify homologs at Reef Genomics database (reefgenomics.org). Moreover, given that SPHK and SGPP proteins are highly conserved across eukaryotes, A. pallida sequences from this study were also compared with the S. minutum genome (OIST: symb_aug_v1.120123) using BLASTx with an E-value ≤e−10. DNA sequences of AP-SPHK and AP-SGPP from this study were aligned with BLAST hits from the A. pallida genome using MUSCLE (Edgar, 2004) or converted to predicted protein in Geneious v5.4.3 (Kearse et al., 2012) and aligned with representative proteins from A. pallida, S. minutum and mammalian model R. norvegicus (SPHK1: NCBI ABF30968.1; SPHK2: NCBI AAH79120.1; SGPP1: UniProt Q99P55.2; SGPP2: UniProt F1LZ44) using MUSCLE (Edgar, 2004).

Before developing the qPCR assay, the full-length transcripts from RACE amplification were surveyed for single nucleotide polymorphisms (SNPs), which have been shown to interfere with amplification and quantification of PCR products if they occur within primer annealing sites (Taris et al., 2008). SNP primers for the genes of interest AP-SPHK and AP-SGPP were designed with Primer3 software v 2.3.4 (Table 1) (Rozen and Skaletsky, 1999) and verified with cDNA from six anemones using cloning and sequencing methods described above (data not shown). The Hsp90 gene sequence (AIPGENE3199) was recovered from the A. pallida genome (reefgenomics.org). For all genes, qPCR primers were designed to exclude SNPs and amplify a product ranging in size from 100 to 200 bp (Table 1). PCR products were verified through cloning and sequencing as described above.

Genes of interest were normalized to reference genes selected for stable expression during experimental treatments. Candidate reference genes included poly-A binding protein (PABP), ribosomal large subunit 10 and 12 (L10 and L12), eukaryotic translation elongation factor 1 alpha 1 (EEF1A1) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH); they were tested on 25 and 33°C-incubated samples at days 1 and 4 as described previously (Poole et al., 2016). Based on the stability values from GeNorm and NormFinder (Table S1), the combination of PABP, L10 and GAPDH genes was selected as the reference genes for elevated temperature-treated anemones. After reference gene selection, cDNA from each sample was run in triplicate 20 µl reactions of 10 µl Power SYBR^® Green PCR master mix (Applied Biosystems, Foster City, CA, USA), 5 µmol forward and reverse primer pair, 0.5 µl cDNA and RNase-free H₂O on the ABI Prism 7500 Real-Time PCR machine (Applied Biosystems) with standard settings and an additional melt curve. No-template, no-reverse transcriptase and no-primer controls were tested, as well as one interplate calibrator sample per plate. C_t values calculated by ABI 7500 software v2.0.6 (Applied Biosystems) with a set baseline threshold of 0.2 were imported into GenEx v5.3.7.332 (MultiD Analyses AB, Göteborg, Sweden) to adjust interplate variation and normalize expression to reference genes. The ΔΔC_t method (Livak and Schmittgen, 2001) was used to calculate relative quantities for each treatment group (27, 30 and 33°C) to the mean ΔC_t of 25°C samples at each time point (n=3). Relative quantities are presented on the log₂ scale.

The best statistical model for AP-SGPP and AP-SPHK relative quantities was determined through Chi-square goodness of fit tests for the fixed effects of time (seven levels) and temperature (four levels: 25, 27, 30 and 33°C), and random effect of well placement. Relative quantities of AP-SGPP were fitted to a GLMM with the fixed effect of temperature and random effect of well placement, whereas AP-SPHK relative quantities were fitted to an intercept model. Post hoc two-tailed Student's t-tests were run for treated samples at each time point (n=3 anemones) compared with the time-matched non-heat-stressed control. P-values were adjusted for multiple-test comparisons using a Bonferroni correction. There was no evidence for the inclusion of the random effect of well placement in a GLMM model of Hsp90 relative quantities (Chi-square goodness of fit, P>0.05). Therefore, the effect of temperature by time interaction on Hsp90 expression was tested using a two-way ANOVA test. Differences between temperature and time points were found using Tukey's post hoc test. The significant threshold for all statistical tests was P≤0.05.

Two experimental trials were run for the lipid analysis. In both experiments, anemones were homogenized in 0.5 ml of 1× phosphate buffered saline (PBS) and spun at 2500 rpm for 3 min to gently pellet cellular debris and symbionts. For each sample, the supernatant was divided, with 100 µl set aside for protein quantification using a Bradford assay. In experiment 1, the remainder was transferred to a 13 mm borosilicate glass tube for lipid extraction using established protocols for lipid extraction for A. pallida (Bligh and Dyer, 1959; Garrett et al., 2013). An internal standard mixture was added to the lipid portion consisting of 25 ng ml⁻¹ of C17-base-D-erythro-sphingosine-1-phosphate (C17-S1P) and 25 ng ml⁻¹ of C17-base-D-erythro-sphingosine C17 (C17-Sph; Avanti Polar Lipids, Alabaster, AL, USA). A 1:2 chloroform:methanol mixture was added to each sample, agitated by vortexing for 20 s, and incubated at RT for 5 min. Then, equal volumes of chloroform and PBS were added to the mixture. After centrifugation at 4°C for 5 min at 3000 rpm, a two-phase separation was visible. The lower organic phase was collected in a clean glass tube and dried under gentle nitrogen stream. In experiment 2, the remaining homogenized tissue was placed in a low-protein binding tube (Eppendorf, Hamburg, Germany) and stored at −80°C until shipped to Lipidomics Core Facility at Wayne State University (Detroit, MI, USA), where lipids were extracted. The same internal standard mix was added to the samples at a concentration of 25 ng ml⁻¹. A solvent mixture of ethyl acetate−isopropanol−water (60:30:10 v/v) was added to each sample followed by centrifugation at 6000 g to extract the lipids. The extraction was repeated twice and the organic supernatant from all three extractions was evaporated with nitrogen. In both experiments, lipid residue was dissolved in LC-MS grade methanol:water:formic acid:ammonium formate (60:40:0.2:2 mmol l⁻¹ v/v), and stored at −20°C.

S1P and Sph were resolved using high-performance liquid chromatography (HPLC) on the Prominence XR system (Shimadzu, Kyoto, Japan) using a Targa C8 (5 µm, 2.1×20 mm, Higgins Analytical, Mountain View, CA, USA) column by the Lipidomics Core Facility at Wayne State University. The mobile phase consisted of a gradient between (i): methanol–water–ammonium formate–formic acid (5:95:1 mmol l⁻¹:0.2 v/v) and (ii): methanol–ammonium formate–formic acid (100:2 mmol l⁻¹:0.2 v/v) with a flow rate of 0.5 ml min⁻¹. The HPLC eluate was introduced to the ESI source of a QTRAP5500 mass analyzer (AB SCIEX, Framingham, MA, USA) in the positive ion mode and molecular ion–daughter ion transitions for each sphingolipid were detected using the multiple reaction monitoring (MRM) method. The MRM transition chromatograms were captured by Analyst v1.6.2 software (AB SCIEX) and quantified by MultiQuant software (AB SCIEX). Sample analytes that passed signal to noise thresholds and quality measures were quantified against internal standards and normalized to total protein (mg). The fold-change of each analyte was calculated as the 33°C-treated anemones referenced to the mean of the concentrations of control anemones at each respective time point. Differences in fold-changes and sphingolipid concentrations were tested with a non-parametric Kruskal–Wallis test followed by Dunn's post hoc test. In addition, each sample used for lipid extraction had recorded autofluorescence over the duration of the temperature treatment; therefore, the lipid concentrations could be correlated to final mean fluorescence using a Pearson's correlation test.

The rheostat enzymes SPHK and SGPP are highly conserved across eukaryotes, occurring in both cnidarian and dinoflagellate genomes. One AP-SPHK and one AP-SGPP gene were recovered from the A. pallida transcriptome (Lehnert et al., 2012) and matched BLAST queries with 43% (E-value=3e−43 to R. norvegicus SPHK1, NCBI: ABF30968.1) and 83% (E-value=9e−63 to Homo sapiens SGPP, NCBI: CAC17772.1) coverage, respectively. These partial transcripts were RACE-amplified and then searched against the A. pallida and S. minutum genomes that became available after this study began.

From the A. pallida genome, two putative SPHK (AIPGENE12769 and AIPGENE12827) and one SGPP (AIPGENE2150) were identified (Baumgarten et al., 2015). In mammals, slime mold, yeast and fruit flies, there are two SPHK isoenzymes, SPHK1 and SPHK2, which differ in cellular location, leading to some overlapping and opposing functions (reviewed by Maceyka et al., 2005; Alemany et al., 2007; Chan et al., 2013). It is unclear whether the two SPHK enzymes arose from a common ancestor or are the result of lineage-specific gene duplication and diversification. To determine whether the predicted gene models recovered from the genome were two distinct SPHK genes, the AP-SPHK sequence was compared with the gene models using multiple sequence alignments. A nucleotide pairwise comparison of the RACE-amplified AP-SPHK sequence with the genome sequences resulted in 97.9% identity to AIPGENE12769 and 58.5% identity to AIPGENE12827 (Fig. S1). For the A. pallida genome assembly, annotation was automated based on translated sequence similarity to the Swiss-Prot database. The transcript AIPGENE12769 that had higher similarity to the RACE-amplified sequence was annotated based on similarity to the mouse SPHK2 in Swiss-Prot (Q9JIA7). The other transcript (AIPGENE12827) was annotated by similarity to Arabidopsis SPHK1 in Swiss-Prot (Q8L7L1) but only matches with 32.6% identity to the Swiss-Prot sequence, mainly at the C-terminal end. Furthermore, the two genome sequences were found adjacent to one another on the same scaffold (data not shown). This arrangement can arise from segmental duplications during genome draft assembly of diploid organisms with heterozygous allele copies, especially those in highly divergent regions (Kelley and Salzberg, 2010). Given that only one transcript has been identified in the transcriptome, RACE amplification and other cnidarian resources (Table S2), we suggest that the two gene models identified in the genome originate from sequence mis-assembly rather than from gene duplication. Therefore, the results presented in this study are based on the full-length RACE-amplified AP-SPHK transcript available at NCBI, accession no. KY038070. To determine whether AP-SPHK was more similar to the cytosolic SPHK1 or membrane-bound SPHK2 identified in mammals, we compared the protein translation of AP-SPHK with R. norvegicus SPHK1 and SPHK2 using a multiple protein sequence alignment (Fig. S2). AP-SPHK did not share the hallmark features of membrane-bound SPHK2, which includes a 200 amino acid proline-rich insert between domains 4 and 5 (Liu et al., 2000), four transmembrane segments (Liu et al., 2000) and a nuclear localization sequence at the N-terminus (Igarashi et al., 2003) (hydropathy data not shown).

There are also two SGPP genes with three conserved domains (Fig. S3) (Spiegel and Milstien, 2003). We converted the RACE-amplified AP-SGPP transcript to protein in silico and compared it with the A. pallida genome match AIPGENE2150, R. norvegicus SGPP1 and SGPP2 using a multiple protein sequence alignment (Fig. S3). Pairwise identity was 96.7%, 33.2% and 31.6%, respectively. The low sequence similarity observed to the R. norvegicus SGPPs is comparable to the predicted homolog in oyster Crassostrea gigas (Timmins-Schiffman and Roberts, 2012).

In the S. minutum genome, one SPHK homolog was recovered (symbB1.v1.2.011198.t1, E-value=7e−41), but none were recovered for SGPP with the rat or human sequences as queries. However, using the plant Arabidopsis thaliana SGPP protein sequence NP_191408 as a query, two homologs from S. minutum (symbB1.v1.2.002394.t1, E-value=4e−32; and symbB1.v1.2.027512.t1, E-value=3e−18) were identified that contained a phosphatidic acid phosphatase type 2 domain found in SGPP and a larger group of lipid phosphate phosphohydrolyases. Based on protein sequence similarity, AP-SPHK and S. minutum SPHK share just 25.3% pairwise identity (Fig. S2). Therefore, both partners have the molecular machinery necessary to cycle S1P and Sph, but the enzymes are highly divergent at the protein level.

Given that homologs of SPHK and SGPP were identified in A. pallida and S. minutum genomes, we developed a host-specific qPCR assay that excluded amplification of the symbiont enzymes. During RNA extraction, anemones had varying amounts of accompanying symbiont RNA. This contamination was confirmed with amplification of a Symbiodinium-specific protein, peridinin chlorophyll protein, from sample cDNA (data not shown). Therefore, we tested the specificity of the A. pallida qPCR primers for amplification of S. minutum cDNA. AP-SPHK qPCR primers failed to amplify a product from symbiont cDNA but AP-SGPP qPCR primers did amplify a product at a high mean C_t cycle of 34.208±0.35. Although a product was detected with the AP-SGPP qPCR primers, this amplification was shifted by 7.2 cycles from A. pallida cDNA expression (mean C_t=27.0.13±0.19) and therefore considered to be a non-specific PCR product. These results indicated that the qPCR primers captured host-specific expression.

To test the hypothesis that the sphingosine rheostat is part of the immediate HSR in cnidarians, relative expression of rheostat genes AP-SPHK and AP-SGPP was measured in symbiotic anemones incubated in ambient (25°C) and hyperthermal (27, 30 and 33°C) temperatures at several time points. A range of temperatures was tested over 1 week to determine thermal and temporal activation of rheostat transcription. Relative quantities of AP-SGPP decreased 0.8-fold in 33°C-incubated anemones, but not with more moderate elevated temperatures (GLMM conditional R²=0.716, P=0.003; Fig. 2A). However, AP-SGPP expression was not downregulated over the entire time course. It significantly declined between 3 h and 1 day (post hoc t-test, 3 h: P=0.008, 6 h: P=0.003, 12 h: P=0.012, 1 day: P=0.031), began to increase by the second day and eventually exceeded control expression at 1 week (post hoc t-test, 7 days: P=0.011). In contrast, relative quantities of AP-SPHK did not differ between any of the four temperature treatments (linear regression R²=0.2037, P=0.096; Fig. 2B). The post hoc test, however, revealed significantly decreased AP-SPHK expression at 12 h in 30°C and 1 day in 33°C conditions compared with ambient temperature (post hoc t-test, 12 h: P=0.002, 1 day: P=0.031). So, although there was no difference between hyperthermal temperatures in terms of AP-SPHK expression over time, downregulation was observed for specific time points. By the end of 1 week, an opposing expression pattern of the rheostat genes emerged, but it was not significant for AP-SPHK (post hoc t-test, P=0.112; Fig. 2). Overall, the relative quantities of both rheostat genes were initially downregulated with acute hyperthermal stress and then there was a shift to upregulation of AP-SGPP by the end of the experiment.

From the range of temperatures tested, the most extreme 33°C hyperthermal treatment initiated a greater transcriptional response of the sphingosine rheostat than the intermediate hyperthermal treatments (Fig. 2). Therefore, we only compared animals incubated at ambient and 33°C in subsequent analyses. We were next interested to see how the timeline of symbiont loss and onset of a HSR of the host corresponded to changes in sphingosine rheostat expression during incubation at elevated temperature. We hypothesized that if the rheostat is driving bleaching, then changes in rheostat gene expression would occur before evidence of bleaching. Relative symbiont densities of anemones were quantified by measuring mean chlorophyll autofluorescence (Fig. 3A). At elevated temperature, there was evidence for a loss in symbiont autofluorescence over time (GLMM, conditional R²=0.68). For each additional day, mean autofluorescence was 2.36 times lower in heat-stressed anemones compared with the controls (GLMM, P<0.001; Fig. 3B). By 1 and 4 days, the mean autofluorescence in the heat-stressed animals was 28.78±4.97% and 50±4.93% lower in the hyperthermal treatment compared with the initial mean fluorescence (Fig. 3B). At 7 days, very little additional loss (1.37%) was observed (Fig. 3B). Relative symbiont loss was variable under hyperthermal stress. Therefore, we compared the percentage symbiont loss at 6 h with total protein to see whether anemone size affected symbiont loss. No correlation was detected between symbiont loss and anemone size in either temperature treatment (Pearson's correlation, 25°C correlation=−0.06 and P=0.65, and 33°C correlation=0.18 and P=0.27; Fig. S4).

To detect the onset of the HSR in A. pallida, we measured the gene expression of a characterized biomarker of stress in cnidarians, molecular chaperone Hsp90 (see review by Louis et al., 2017) . In A. pallida, there was a significant interaction between temperature treatment and time on the relative quantities of Hsp90 (two-way ANOVA, P<0.001; Fig. 4). At 3 and 6 h, the relative quantities of Hsp90 were upregulated 4.22- and 3.45-fold, respectively, in the hyperthermal treatment compared with the time-matched controls (Tukey's post hoc, 3 and 6 h: P<0.001). The expression of Hsp90 remained elevated in anemones in the hyperthermal condition between 12 h and 1 day, but this did not differ from the baseline expression at 2 and 7 days in this treatment group or the control group (Fig. 4). There was evidence for a slight downregulation of Hsp90 expression at 4 days in the hyperthermal treatment that statistically differed from the upregulation observed between 3 h and 1 day (Fig. 4).

Although transcriptional differences between rheostat genes were not observed until several days after the onset of heat stress, post-translational activation of AP-SPHK and AP-SGPP proteins could be part of the immediate HSR. We indirectly measured the enzymatic activity of rheostat enzymes by quantifying their products, S1P and Sph. Between 6 h and 4 days, Sph concentrations were lower in the heat-stressed anemones than in their corresponding controls at ambient temperature, resulting in lower lipid concentrations and fold-changes below 1 (Fig. 5A; Fig. S5A). However, the median fold-change of Sph increased by 0.77 between 6 h and 7 days (Fig. 5A; Dunn's post hoc test, P=0.021). Median S1P fold-change remained low throughout the experiment (Fig. 5B), but S1P concentrations at 6 h differed from those at 4 and 7 days in anemones in the hyperthermal condition (Fig. S5B; Dunn's post hoc test, 4 days: P=0.023 and 7 days: P=0.031).

Given that loss of symbionts varied across individuals, the severity of the hyperthermal stress on symbiosis breakdown was not uniform. More thermotolerant anemones (those with less symbiont loss) might therefore undergo regulation of sphingolipid metabolism that contributes to resistance to bleaching. To examine this, we correlated the mean fluorescence of ambient and 33°C-treated anemones to their recovered sphingolipid concentrations. At 33°C, there was no evidence that Sph or S1P concentrations were correlated with hyperthermal exposure and symbiont loss (Pearson's correlation, Sph correlation=0.23 and P=0.19, and S1P correlation=0.07 and P=0.68; Fig. 6).

We examined the activation of the sphingosine rheostat with heat stress as part of the signaling cascades involved in the cnidarian HSR and symbiosis breakdown. Our findings, summarized in Fig. 7, suggest that although the rheostat is involved in the HSR, symbiosis breakdown precedes changes in the gene expression and activity of the rheostat. Therefore, there is no evidence in this study that the rheostat directly initiates cnidarian bleaching.

The HSR dynamically alters almost all cellular processes in a highly orchestrated fashion and occurs in two response phases: acute (phase I) and chronic (phase II) (Balogh et al., 2013; Chen et al., 2013). In phase I, those processes required for immediate action including post-translational modifications of existing proteins, and rapid elevations in transcription and synthesis of Hsps and other stress-responsive proteins, coincide with the downregulation of most non-essential cellular processes (Graner et al., 2007; Morimoto, 1993). Therefore, the expression of Hsps is an important biomarker for phase I of the HSR in cnidarians. After 1 day of thermal stress exposure, Hsp90 was upregulated in the corals Orbicella faveolata (DeSalvo et al., 2008), Acropora millepora larvae (Rodriguez-Lanetty et al., 2009), Acropora palmata (DeSalvo et al., 2010), Acropora aspera (Leggat et al., 2011), Acropora tenuis (Yuyama et al., 2012) and Porites asteroides (Kenkel et al., 2011). However, under chronic stress conditions, Hsp90 was downregulated in P. asteroides (Kenkel et al., 2013). Following phase I, processes essential for slower, longer term acclimatization are regulated by changes in gene expression, metabolism and membrane organization as part of phase II (Balogh et al., 2013; de Nadal et al., 2011). The timing of phase II is dependent on the severity and duration of the stress and the thermal tolerance of the partners, all of which contribute to the cnidarian HSR (Hofmann and Todgham, 2010; Tchernov et al., 2004).

In A. pallida, S. minutum loss is a rapid process and is temporally associated with phase I of the HSR (Fig. 7). We observed 40% loss after 1 day at 33°C, a value within the range of other studies that noted significant symbiont pigment and density loss after acute heat stress (Detournay and Weis, 2011; Dunn et al., 2004; Gates et al., 1992; Hawkins et al., 2013; Paxton et al., 2013; Perez and Weis, 2006; Sawyer and Muscatine, 2001; Tolleter et al., 2013). This dysbiosis coincides with early upregulation of Hsp90 expression (Fig. 7) observed in this study and elevated caspase activity, NO production and ultimately host cell death in anemones exposed to elevated temperatures from 31.5 to 33.5°C reported in other studies (Black et al., 1995; Detournay and Weis, 2011; Dunn et al., 2007; Hawkins and Davy, 2012; Paxton et al., 2013). However, not all host–symbiont combinations of A. pallida bleach as quickly. The A. pallida strain CC7 with S. linucheae (subclade A4) did not have significant symbiont loss until 7 days of hyperthermal treatment, but caspase activity was elevated as early as 2 h (Bieri et al., 2016). These differences in bleaching response could result from different thermal sensitivities of the A. pallida genotype, symbiont species or both.

The activity of the sphingosine rheostat, a regulator of host cell death and NO production, was not part of phase I, thereby refuting our original hypothesis. During phase I, both rheostat genes and lipid concentrations were downregulated (Fig. 7). Based on a transcriptomic study, symbiotic anemones have lower constitutive AP-SGPP gene expression than aposymbiotic anemones (Rodriguez-Lanetty et al., 2006). Therefore, we predict that the symbiotic A. pallida could have a surplus of AP-SPHK transcripts from elevated AP-SPHK expression that could be recruited to mitigate the heat-induced cellular damage before initiating a transcriptional response of the rheostat. The high sample variation in S1P concentrations at early time points suggests that some anemones had elevated AP-SPHK activity. Expression of the gene encoding the pro-apoptotic enzyme AP-SGPP and subsequent elevations in Sph concentrations were induced by longer term incubations at the highest temperature of 33°C in A. pallida, but not by moderately high temperatures of 27 and 30°C. In a previous study, the slow heating of A. pallida revealed two peaks in proteolytic activity of caspase 3 and 9, two cysteine proteases essential in apoptosis, at day 1 and 5 at the maximal temperature (33°C) (Hawkins et al., 2013).The increase in Sph between days 4 and 7 observed in this study could trigger pro-apoptotic mechanisms as part of the chronic HSR. Our results suggest that initiation of dysbiosis is part of the acute response (phase I) and sphingosine rheostat activation is part of the chronic response (phase II) in symbiotic A. pallida (Fig. 7).

Despite the fact that additions of exogenous S1P and Sph altered bleaching in A. pallida (Detournay and Weis, 2011), the endogenous rheostat examined in this study did not change during rapid bleaching in phase I. This suggests that symbiont loss is independent of sphingosine rheostat activation. Thus, the cellular programs in the host controlling initial bleaching do not appear to include the rheostat. We measured the response of the whole animal to thermal stress, but it is important to consider that the microenvironment of gastrodermal tissue where Symbiodinium spp. reside might be significantly different from the animal as a whole. Further investigation into the tissue-specific effects of the rheostat using single-cell expression profiling or fluorescently tagged sphingolipid localization will help reveal the involvement of sphingolipids in dysbiosis.

Other components of the sphingolipid signaling and their downstream effectors may contribute to symbiont removal and increased apoptosis or NO production during the acute HSR. An important step in the mammalian HSR is the rapid accumulation of ceramide (see Fig. 1) that precedes apoptosis in mammalian cell lines (Goldkorn et al., 1991; Jenkins et al., 2002; Nagai et al., 2011). In addition to its role in cell death, ceramide can activate eNOS leading to high NO levels (Florio et al., 2003). NO production regulates ceramide levels, synthesis and trafficking, and determines cell fate (see Fig. 1). Furthermore, Hsp90 is an allosteric enhancer of eNOS (García-Cardeña et al., 1998) and inducible NOS (Yoshida and Xia, 2003). Immunoprecipitation studies suggest the presence of similar allosteric interactions in the soft coral Eunicea fusca subjected to a heat stress (Ross, 2014). Pathways for ceramide production in A. pallida are unexplored. Future studies aimed at investigating ceramide levels and enzymes involved in accumulating ceramide will enhance our understanding of the cnidarian HSR and HSR evolution in eukaryotes.

In summary, we have shown that transcriptional activation and activity of the cnidarian sphingosine rheostat are not associated with the acute phase HSR or initiation of dysbiosis but are instead part of the longer term chronic HSR. The sphingosine rheostat could therefore be participating in longer term mechanisms of host survival under heat stress, but not a mediator of the complex mechanisms underlying symbiont removal and dysbiosis. To isolate the physiological changes of heat stress from those associated with symbiosis breakdown, other non-symbiotic model cnidarians such as the sea anemone Nematostella vectensis or hydrozoan Hydra vulgaris are good candidates for examining the role of the sphingosine rheostat in the HSR. Nevertheless, this study demonstrates a shift in the sphingosine rheostat in a symbiotic cnidarian under prolonged thermal stress toward cell death and increases our understanding of its role in an evolutionarily conserved stress response.

The authors would like to thank Dr Robert Mason, Dr Camille Paxton, Dr Angela Poole, Trevor Tivey, Jessica Flesher, Kinsey Matthews and Darian Thompson for their assistance in the execution of the study. In addition, the authors would like to thank Drs Dee Denver, John Fowler, Nathan Kirk, Eli Meyer and Barbara Taylor for their thoughtful feedback on the data analysis and manuscript.