Physiological responses of the scleractinian coral Pocillopora damicornis to bacterial stress from Vibrio coralliilyticus

Coral reef ecosystems are among the most biologically diverse and complex marine ecosystems worldwide. In addition to their biological and ecological importance, coral reefs support major economic and physical functions (e.g. food production, tourism, biotechnology development and coast protection) that are essential for many countries (Lesser, 2004). These ecosystems are principally supported by small, colonial and calcifying organisms, the hermatypic scleractinian corals, which enter into mutualistic symbioses with microalgae of the genus Symbiodinium, dinoflagellates that are also referred to as zooxanthellae (Lesser, 2004).

Because of global climate change and anthropogenic pressure, coral reef ecosystems have been increasingly confronted with severe natural and anthropogenic disturbances over the past three decades (Hughes et al., 2003; Ward and Lafferty, 2004; Donner et al., 2005; Hoegh-Guldberg et al., 2007; Bourne et al., 2009). A recent global assessment of coral reef health showed that approximately 19% of coral reefs were irremediably degraded with no sign of recovery, 15% presented symptoms of an imminent risk of collapse and another 20% were at risk of becoming critically affected in the next few decades (Wilkinson, 2008). Coral diseases are among the major factors in coral reef degradation; their impacts have severely increased in recent decades, in apparent association with global climate change (Weil et al., 2006; Bourne et al., 2009). Indeed, it has been suggested that high temperatures influence the outcome of bacterial infections by lowering the disease resistance of corals and/or increasing pathogen growth, virulence and infectivity (Ward et al., 2007; Rodriguez-Lanetty et al., 2009).

To date, 18 coral diseases have been identified (Sutherland et al., 2004; Willis et al., 2004; Harvell et al., 2007). Among the well-characterized coral infectious diseases, several have been shown to be caused by members of the Vibrionaceae family (Kushmaro et al., 2001; Ben-Haim et al., 2003a; Sussman et al., 2008). Bacteria belonging to the genus Vibrio are ubiquitously distributed in aquatic environments worldwide, from brackish water to deep-sea environments. Kushmaro et al. demonstrated that the bacteria Vibrio shiloi infects and triggers bleaching of the coral Oculinapatagonica (Kushmaro et al., 1996; Kushmaro et al., 1997; Kushmaro et al., 1998; Kushmaro et al., 2001). The same research group showed that the infectivity of this bacteria is temperature dependent and occurs only after an increase in seawater temperature (Kushmaro et al., 1998). A similar temperature-dependent virulence was observed in interactions between Vibrio coralliilyticus and the scleractinian coral Pocillopora damicornis (Ben-Haim et al., 2003b), and the involvement of Vibrio species has been suggested in two other coral diseases: yellow blotch/band disease (Cervino et al., 2004; Cervino et al., 2008) and rapid tissue necrosis (Luna et al., 2007). Among these Vibrio-related diseases, the bleaching disease of O. patagonica caused by the infection of V. shiloi has been characterized to the greatest degree (for a review, see Rosenberg, 2004). In brief, V. shiloi is attracted by chemotaxis and adheres to β-galactoside-containing receptors on coral cells. This adhesion process is specific and dependent on temperature: no adhesion occurs at low temperatures (16–20°C), whereas V. shiloi actively adhere to their target at temperatures between 25 and 30°C. After contact, the bacteria penetrate and multiply within the coral cells, and begin producing two sets of factors: extracellular proline-rich toxins (referred to as toxin P) (Banin et al., 2001) that block photosynthesis, and enzymes that trigger the lysis of zooxanthellae.

However, although Vibrio species have been shown to play crucial roles in several coral diseases, our knowledge of the effects of Vibrio infection on coral physiology remains incomplete. Accordingly, the present study used a global transcriptomic approach to examine the physiological responses of coral colonies when confronted with bacterial stress and/or infection in a realistic ecological context. Specifically, we studied the interaction between the coral P. damicornis and its specific pathogenic bacteria, V. coralliilyticus.

Pocillopora damicornis has a widespread distribution in the Indo-Pacific region (Veron, 2000), is highly sensitive to a wide range of natural and anthropogenic disturbances, and is a good sentinel species for use in monitoring coral reef health (Gates et al., 1992; Ben-Haim and Rosenberg, 2002; Stimson et al., 2002; Hashimoto et al., 2004; Vidal-Dupiol et al., 2009). Vibrio coralliilyticus is a specific pathogen of P. damicornis, and its virulence is temperature dependent; the bacterium has been shown to trigger coral bleaching at moderate temperatures (24–25°C) and coral tissue lysis at higher temperatures (26–29°C) (Ben-Haim Rozenblat and Rosenberg, 2004). Consequently, the P. damicornis–V. coralliilyticus model allows us to compare coral responses to virulent and non-virulent bacteria.

Here, we exposed coral to a consistent supply of bacteria at either 25°C (non-virulent) or under conditions of a gradual temperature increase (from 25 to 32.5°C; induction of virulence) and examined transcriptomic modifications using subtractive hybridization. The differentially regulated genes identified through this approach included several thought to be involved in immune processes. Quantitative reverse-transcription PCR (qRT-PCR) was used to follow the expression levels of a number of selected genes at various time points after the onset of bacterial stress or infection.

The Pocillopora damicornis (Linnaeus 1758) isolate used in this study was harvested in Lombock, Indonesia (CITES no. 06832/VI/SATS/LN/2001) and maintained at the Cap d'Agde Aquarium, France. For experimental procedures, nubbins (10 g; 7×6 cm height×diameter) were propagated by cutting branches from the parent colony and physiologically stabilizing them for 2 months at 25°C.

The coral pathogen Vibrio coralliilyticus strain YB1 (Ben-Haim et al., 2003a) (CIP 107925, Institut Pasteur, Paris, France) was used to challenge or infect P. damicornis (Ben-Haim and Rosenberg, 2002). Vibrio coralliilyticus was cultured in 2216 Marine Broth medium (catalog no. 279110; BD Difco, Le Pont de Claix, France) under aerobic conditions with shaking (150 rpm) at 30°C for routine use, or at the temperature in the coral-containing tank for the Tb and Cb groups (see Experimental infection).

Experiments aimed at determining which cells (host or symbiont) expressed the candidate genes (see Identification of host or symbiont genes) utilized three zooxanthellar isolates corresponding to clonal cultures of zooxanthellae from the B, C and D clades (BURR Culture Collection, University of Buffalo; culture IDs Flap3, Pd44b and MF2.2b, respectively). The representative zooxanthellae were cultured in 250 ml screw-top polycarbonate cell-culture flasks (Corning, Paris, France) in filtered seawater (pore size, 0.2 μm) enriched with 1× Guillard's (F/2) marine water enrichment solution (Sigma-Aldrich, Lyon, France) and supplemented with 1× antibiotic/antimycotic stabilized suspension (Sigma-Aldrich). The zooxanthellae were grown at 25±1°C under an irradiance of 70 μmol photons m^–2 s^–1 provided by daylight fluorescent tubes (Sylvania, Gennevilliers, France) set to a 12 h:12 h light:dark photoperiod. Stock cultures were transferred monthly, and cells were used in the stationary phase.

Bacterial challenge was induced by constantly supplying V. coralliilyticus at a relatively low temperature (25°C; non-virulent), whereas infection was induced by constantly supplying the bacteria and systematically increasing the water temperature from 25 to 32.5°C (thereby activating bacterial virulence).

Bacterial challenge, bacterial infection and their respective controls were established in four independent tanks: (1) constant low temperature (25°C) and bacterial supply (Cb); (2) gradual temperature increase from 25 to 32.5°C and regular bacterial supply (Tb); (3) constant low temperature without bacteria (C); and (4) gradual temperature increase from 25 to 32.5°C in the absence of bacteria (T). For experiments, nubbins of P. damicornis were randomly placed in 120 l tanks (N=27 per tank). After a 14-day acclimatization at 25°C, the C and Cb treatments were maintained at this temperature whereas the temperatures of the T and Tb treatments were gradually increased by 1.5°C every 3 days until the water temperature reached 32.5°C (days 3, 6, 9, 12 and 15). This higher temperature was maintained until the corals bleached or lysed. In the Cb and Tb treatments, bacteria (10³ cells ml^–1 of tank water) were supplied by balneation (Ben-Haim Rozenblat and Rosenberg, 2004). The first bacterial supply was initiated on day 1 and repeated every 3 days (days 1, 4, 7, 10, 13, 16 and 19). The utilized cultures of V. coralliilyticus were prepared at 25°C for the Cb treatment and at the temperature corresponding to the tank temperature for the Tb treatment. For the Tb and T treatments, beginning on day 3, the temperature was increased by +1.5°C every 3 days until it reached 32.5°C.

Three nubbins were randomly sampled from each tank every 3 days (days 0, 3, 9, 12, 15 and 18). Five 0.5-cm apices were cut from each nubbin; three were immediately frozen and stored in liquid nitrogen for later measurement of zooxanthellar density and two were fixed directly in 2.5% glutaraldehyde and subjected to electron microscopy.

The tank temperature was controlled using an aquarium heater (600 W, Schego, Offenbach, Germany) connected to an electronic thermostat (Hobby Biotherm Professional, Koblenz, Germany). Illumination was supplied at an irradiance of 250 μmol photons m^–2 s^–1 (quantum meter; QMSW-SS, Apogee Instruments Inc., Roseville, CA, USA) by metal halide lamps (400 W, Iwasaki 6500 Kelvin, Tokyo, Japan) set to a 12 h:12 h light:dark photoperiod. All other seawater parameters were maintained constant over time and between tanks. A constant water flow was maintained within each tank using a water pump (1300 l h^–1, IDRA, Pozzoleone, Italy). Seawater was continuously recycled at a rate of 10.8 tank volumes per hour by coupling the action of a biological filter and an Aquavie protein skimmer (EPS 600, Montpellier, France), and the water was renewed (2% tank volume per day) with natural filtered Mediterranean seawater heated to 25°C. To avoid the growth of bacterial blooms, the water was continuously treated using a UVC filter (5 W, AquaCristal Series II, JBL, Neuhofen, Germany). During bacterial balneation, all equipment known to remove or kill bacteria (e.g. the protein skimmer and UVC filter) were inactivated for the first 4 h after treatment, allowing the bacteria to adhere to the coral tissues.

In order to monitor coral health, physiological (e.g. zooxanthellar density measure) and visual (e.g. coral color, tissue lysis appearance, opened or closed polyps) measurements were performed throughout the experiments. Zooxanthellar density was used as a measurement of coral health and the stability/breakdown of symbiosis. Coral tissues were extracted using a water pick (Johannes and Wiebe, 1970) in 50 ml of filtered seawater (0.2 μm pore size), homogenized with a Potter grinder and fixed with 2% formaldehyde. Zooxanthellae were counted under a light microscope using a hemocytometer, and algal densities were standardized per skeletal surface area, as described previously (Stimson and Kinzie, 1991). Every day, two colonies per tank (the same colonies each day) were photographed using a consistent camera setup, and both bleaching and lytic events were thoroughly examined.

In order to follow the infectious process in P. damicornis tissues, the presence of V. coralliilyticus was investigated by transmission electron microscopy (TEM). Two ultrathin slices were prepared from coral fragments fixed in 2.5% glutaraldehyde in filtered seawater (0.2 μm pore size). Sections were prepared as described previously (Ladrière et al., 2008) and observed with a Jeol JEM 100-SX transmission electron microscope (JEOL, Zaventem, Belgium) at an accelerating voltage of 80 kV.

qRT-PCR was used to quantify the amount of V. coralliilyticus present in P. damicornis tissues. Total RNA was extracted from each sample (see Subtractive cDNA library construction) and 100 ng of total RNA were reverse transcribed using hexamer primers and a RevertAid™ kit (Fermentas International Inc., Saint-Rémy-lès-Chevreuse, France). qRT-PCR was performed with 4 μl cDNA (1/100 dilution) in a total volume of 10 μl using a LightCycler 480 System (Roche Diagnostics, Meylan, France), as described below (see Real-time analysis). Oligonucleotide primers were designed based on the aligned sequences from eight different V. coralliilyticus 16S rRNAs (GenBank accession nos EF094886, DQ079633, GQ406805.1, GQ406804.1, GQ406802.1, GQ406801.1, GQ406800.1 and GQ406798.1; National Center for Biotechnology Information database) and the 28S rRNA of P. damicornis (Vidal-Dupiol et al., 2009) (Table 1). RNA extracted from the V. coralliilyticus culture used for infection was taken as the positive control. The qRT-PCR products were cloned and sequenced (see DNA sequencing and sequence analysis).

Visual and physiological monitoring was used to select early bacterially stressed or infected samples, and subtractive hybridization was used to identify genes that were differentially expressed under the non-virulent and virulent conditions. For each set of three nubbins, tissues were extracted with a water pick in 800 ml of filtered seawater (0.2 μm pore size), and both total RNA and mRNA were purified (Vidal-Dupiol et al., 2009). Forward and reverse suppression–subtractive–hybridization (SSH) libraries were constructed using the PCR-Select cDNA Subtraction kit (Clontech, Saint-Germain-en-Laye, France). Tester and driver cDNA were prepared using 2 μg of Poly(A)+ RNA, and enzyme digestion, adaptor ligation, hybridization and PCR amplification were performed according to the manufacturer's instructions (Clontech). The resulting PCR products were cloned into pCR4-TOPO using a TOPO TA cloning kit (Invitrogen, Cergy Pontoise, France) and transformed into One Shot TOP10 chemically competent Escherichia coli cells (Invitrogen).

For each library, 384 clones were randomly selected and single-pass sequenced by GATC Biotech, using an ABI 3730xl apparatus. Vector and adaptor sequences were trimmed from all sequences using Sequencher™ software (Gene Codes Corp., Ann Arbor, MI, USA). High-quality expressed sequence tags (ESTs) (>100 bp) were assembled into clusters or identified as unique sequences, and used for database searches with the BLASTX and BLASTN programs. Specific domain searches were performed using the InterProscan software. ESTs displaying significant similarities with a gene of known function were assigned to the appropriate functional class. The various EST sequences have been submitted to the dbEST database; in an effort to minimize redundancy, sequences displaying 100% identity with one another were submitted as a single sequence.

qRT-PCR was used to analyze the expression profiles of selected genes. Total RNA was extracted, and this total RNA (2.5 μg) was reverse transcribed using oligo dT primers and the Superscript II enzyme (Invitrogen). The resulting cDNA products were purified using a Nucleotrap Gel Extraction Trial kit (Clontech), and qRT-PCR was performed with 2.5 μl of purified cDNA (diluted 1:50 in water) in a total volume of 10 μl containing 1× LightCycler® 480 SYBR Green I Master Mix (Roche Diagnostics) and 70 nmol l^–1 of each primer. The primers, which were designed using Light Cycler Probe Design Software version 1.0 (Roche Diagnostics), are presented in Table 1. Amplification was performed using a LightCycler 480 System (Roche Diagnostics) and the following reaction conditions: activation of the Thermo-Start® DNA polymerase at 95°C for 15 min, followed by 45 cycles of denaturation at 95°C for 30 s and annealing/extension at 60°C for 1 min. Melting curve profiles (95 to 70°C decreasing by 0.5°C every 10 s) were assayed to ensure that we were looking at a single product. Each run included a positive cDNA control that was sampled at the beginning of the experiment (day 0) and also from each amplification plate; this positive control was also used as the calibrator sample, and blank controls (water) were included for each primer pair. The PCR products were resolved by electrophoresis, the bands were isolated directly from agarose gels and DNA was extracted using the Gel Extraction PCR Purification System V (Promega, Charbonnières-les-Bains, France). The resulting qRT-PCR products were single-pass sequenced as described above.

To determine which organism (host or symbiont) expressed the candidate genes, cross-PCR experiments were performed on DNA and RNA extracted from the holobiont (host plus symbiont) and from pure cultured zooxanthellae, as previously described (Vidal-Dupiol et al., 2009). Briefly, PCR amplifications were performed with oligonucleotides amplifying the targets, the gene encoding the small ribosomal subunit RNA Symbiodinium spp. (Rowan and Powers, 1991) and the cDNA encoding the major basic nuclear protein (MBNP) of Symbiodinium spp. (Table 1).

Because conditions of normality and homoscedasticity were not satisfied, within-set variations in zooxanthellar density were analyzed with the non-parametric Kruskal–Wallis (KW) test using GraphPad InStat 3 (GraphPad Software, La Jolla, CA, USA). Variations in gene expression over time were analyzed with a maximum normed residual test (Grubbs, 1969; Stefansky, 1972) using JMP (SAS Institute, Inc., Cary, NC, USA). Analyses were performed on qRT-PCR results obtained for a double-passed pool of three nubbins from the same treatment at each sampling date. Differences were considered statistically significant at P<0.05.

In order to investigate the effect of bacterial challenge and/or infection, we designed two sets of experiments. The first set examined the effects of non-virulent bacterial challenge by V. coralliilyticus. A constant temperature of 25°C was used, as exposure of P. damicornis to V. coralliilyticus at this temperature was previously shown to trigger bleaching without tissue lysis (Ben-Haim Rozenblat and Rosenberg, 2004). Tanks were maintained at 25°C with (Cb) and without (C) bacteria (10³ bacteria ml^–1). In the second set of experiments, which examined the effects of infection by virulent bacteria, a similar setup was used, except that the tanks with (Tb) and without (T) bacteria were subjected to a gradual temperature increase from 25 to 32.5°C. At each time point, the bacteria added to Tb were cultured at the tank temperature.

Coral health and the stability/breakdown of symbiosis were examined by analysis of zooxanthellar density and visual monitoring. For the C and Cb treatments, the zooxanthellar densities (cells cm^–2 of skeleton) remained stable throughout the experiment (mean±s.d.: C=1.07×10⁶±2.35×10⁵ and Cb=5.18×10⁵±2.89×10⁵ cells cm^–2). There was no significant difference in zooxanthellar density between C and Cb (KW, P>0.05), indicating that bleaching did not occur under these experimental conditions. For the T and Tb treatments, the zooxanthellar densities remained stable over the first 15 days (mean±s.d.: T=6.21×10⁵±1.61×10⁵ and Tb=4.36×10⁵±1.69×10⁵ cells cm^–2), but dropped to 6.26×10³±5.05×10³ and 1.72×10⁵±6.23×10⁴ cells cm^–2, respectively, by day 18. This decrease was only significant for the Tb treatment (KW; P=0.0535 and P=0.0147 for T and Tb, respectively).

Visual monitoring was used to follow the occurrence of symptoms such as tissue lysis and/or bleaching. In the C and Cb treatments, the nubbins appeared healthy throughout the duration of the experiment (Fig. 1). In the tank subjected to only increased temperature (T treatment), polyps were closed at day 15, bleaching became visible in the upper portion of nubbins at day 16 and bleaching was complete by day 18 (Fig. 1). These bleached corals were pale but still alive; when the tank temperature was returned to 25°C, the polyps reopened (Fig. 2). In the Tb treatment, however, tissue lysis was first observed on day 12; lytic plaques were observed in the coral tissues and calcified skeletons (seen at the beginning as small white spots) became evident on the nubbins (Fig. 2). This effect continued until lysis was complete (Fig. 1; day 21). In order to confirm that corals subjected to bacterial infection were dead, the tank temperature was returned to 25°C. After several days at this temperature, polyps did not reopen and the skeleton began to be colonized by algae.

In order to substantiate that V. coralliilyticus was responsible for the observed symptoms, three randomly chosen samples were harvested from each tank at each time point, and the presence/amount of bacteria was investigated using electron microscopy and qRT-PCR. Tb samples showed no bacteria during the first 6 days. On day 9, bacterial aggregates (bacteriocyte-like structures) were first detected inside the ectodermal host cells (Fig. 3A). Similar aggregates were reported by Ben-Haim et al. (Ben-Haim et al., 2003b) when sections of P. damicornis infected by V. coralliilyticus YB1 were analyzed using the same methodology. This suggests that the bacteria multiply inside the ectoderm before tissue lysis becomes visible to the naked eye. Three days later (day 12, 29.5°C, Fig. 3B–D), the tissues surrounding the bacteriocyte-like structures were disorganized and most cellular organelles were lysed (Fig. 3B,C). In contrast, identically treated and processed samples from the other treatments (C, Cb and T) failed to show any bacterial aggregates or lytic events during the experimental period.

To confirm that the observed bacteria were not opportunistic bacteria of another species, we used qRT-PCR to quantify the levels of the V. coralliilyticus 16S rRNA in the tissue samples. Vibrio coralliilyticus-specific primers were designed and used for amplification of cDNA, and the levels of V. coralliilyticus 16S rRNA were compared to that of the P. damicornis 28S rRNA. V. coralliilyticus 16S rRNA corresponding amplicons were cloned and 40 randomly picked clones were sequenced. All sequences obtained corresponded to the V. coralliilyticus 16S rRNA. Consistent with the above results, qRT-PCR showed that the relative amount of V. coralliilyticus 16S rRNA increased significantly by day 6 and remained higher than the control until the end of the protocol (Fig. 4). By comparison, no such increase was observed in Cb group (Fig. 4). This strongly suggests that coral tissue invasion occurs upon increasing water temperature only.

Two SSH experiments were performed to monitor the coral response to bacterial challenge or infection. The first was conducted using pooled samples (days 12 and 15) from the Cb treatment versus the C treatment, whereas the second was conducted using pooled samples (days 9–15) from the Tb treatment versus the T treatment. This yielded four cDNA libraries: the bacteria-induced (BI) and bacteria-repressed (BR) libraries, which corresponded to the subtractive hybridization of Cb versus C, and the thermo-bacteria-induced (TBI) and thermo-bacteria-repressed (TBR) libraries, which corresponded to the subtractive hybridization of Tb versus T. The ‘induced’ libraries contained transcripts putatively induced by bacterial stress or infection whereas the ‘repressed’ libraries contained gene putatively downregulated under the experimental conditions.

The general features of each library are given in Table 2. Three hundred and eighty-four clones were sequenced from each library, yielding 279, 276, 275 and 229 high-quality cDNA sequences (>100 bp) from the BI, BR, TBI and TBR libraries, respectively (Table 2). The ESTs from the BI library coalesced into 63 contigs and 123 singletons, suggesting that the overall redundancy of the library was approximately 56%. The BR, TBI and TBR ESTs distributed to 59, 38 and 23 contigs and 149, 184 and 166 singletons, for overall redundancies of 46, 33 and 28%, respectively (Table 2).

Singleton and consensus clusters were subjected to BLASTN and BLASTX searches and classified into three groups: unknown genes, genes of unknown function, and genes of known function (Fig. 5). Most of the sequences of known gene function presented similarities to predicted proteins from the genomes of Nematostella vectensis (an aposymbiotic cnidarian) and Branchiostoma floridae (Cephalochordata), both of which have recently been sequenced (Putnam et al., 2007; Putnam et al., 2008). The putative functions of these sequences were deduced by domain analysis using the InterProscan program. The identified genes clustered into the following 13 functional categories: apoptosis, cellular homeostasis, cellular metabolism, chromatin remodeling, cytoskeletal organization, immune responses, intracellular protein transport, photosynthetic processes, protein metabolism, stress responses, RNA metabolism, signal transduction and system development (Fig. 5, supplementary material Table S1).

To investigate the response of coral to virulent and non-virulent bacterial challenges, five candidate genes were selected from the SSH libraries, and their transcription levels at various post-challenge time points were measured by qRT-PCR (Fig. 6Aii–vi). Three were selected because of their putative involvements in antibacterial responses; these comprised a ferritin-like protein from the TBI library (similar to Periserrula leucophryna ferritin; E=2×10^–66), a selenium-binding-protein-like protein from the BI library (similar to Mus musculus selenium binding protein; E=2×10^–25) and a cystatin B-like protein from the BR library (similar to Danio rerio cystatin B; E=4×10^–7). The other two selected genes, both lectin-like molecules, were selected based on the importance of lectins in invertebrate immunity. They included a P-selectin-like protein from the TBR library (similar to Homo sapiens P-selectin; E=6×10^–28) and a concanavalin A lectin/glucanase-like protein from the TBI library (similar to Synechococcus sp. thrombospondin N-terminal-like domain-containing protein; E=2×10^–9). These candidates were classified as lectins because the InterProscan analysis of their translated amino-acid sequences revealed the presence of a concanavalin A-like lectin/glucanase domain (InterProscan accession no. SSF49899) and a P-selectin-like domain (InterProscan accession no. PR00343), both of which contain carbohydrate-binding sites. Following our recent discovery in a previous study of another lectin from P. damicornis, PdC-lectin (Vidal-Dupiol et al., 2009), we further investigated the transcript variation of this third lectin under the same experimental conditions (Fig. 6Ai). To verify that the observed variations were linked to the targeted physiological effects and not to profound bacteria-induced disturbances, we also examined the expression levels of three other genes implicated in general cellular processes (P. damicornis: 60S ribosomal protein L22, 60S ribosomal protein L40A and 60S acidic ribosomal phosphoprotein P0; GenBank accession nos HO112261, HO112283 and HO112666, respectively) for comparison (Fig. 6Bi–iii).

To verify that all of the selected candidates (including the reference genes) were expressed by the coral cells rather than their symbionts, we performed cross-PCR experiments using DNA and cDNA extracted from the holobiont (host plus symbiont) and from pure cultured representatives of Symbiodinium spp. clades B, C and D (Fig. 7). The utilized primers included candidate-gene-specific oligonucleotides; primers ss5Z and ss3Z, which amplified the Symbiodinium spp. small ribosomal subunit (Rowan and Powers, 1991); and primers specific for the cDNA of the MBNP of Symbiodinium spp. The results revealed that the candidate genes could be amplified only from DNA and cDNA extracted from holobionts (Fig. 7), whereas the ss5Z/ss3Z and MBNP primers amplified all tested DNA and cDNA samples, respectively. This result shows that all of the selected candidate genes were specifically expressed by coral cells. In addition, results of this experiment for the genes encoding ferritin-like protein and the selenium-binding-protein-like protein show higher amplicon size when the matrix used was genomic DNA (Fig. 7). This is probably due to the presence of intronic sequences in both genes.

Two of the three selected lectins (concanavalin A-like protein and PdC-lectin) displayed similar transcriptional profiles (Fig. 6Ai,ii). Exposure to non-virulent bacteria triggered significant downregulation of concanavalin A-like protein and PdC-lectin at day 9 (more than fourfold and eightfold for PdC-lectin and concanavalin A-like protein, respectively), followed by upregulation at day 18 (only significant for PdC-lectin; 3.6-fold, P<0.05; Fig. 6Ai). In contrast, samples exposed to the virulent conditions showed significant upregulation of both genes from day 6 (2.9- and 3.3-fold for PdC-lectin and concanavalin A-like protein, respectively) to day 15, with a dramatic decrease seen on day 18. The maximum increase was observed on day 12 (7.7- and 9.3-fold for PdC-lectin and concanavalin A-like protein, respectively). The remaining lectin, P-selectin-like protein, differed from the other two; its expression decreased at all tested time points by more than 3.2-fold during the two experiments (Fig. 6Aiii).

The transcription patterns for ferritin-like protein and the selenium-binding-protein-like protein were very similar. In samples exposed to the non-virulent treatment, significant upregulation was observed across most of the tested time points (Fig. 6Aiv,v). In samples exposed to the virulent treatment, expression of ferritin-like protein was strongly downregulated at day 6 (10.2-fold), and then significantly upregulated at the end of the experiment (by more than 16.1-fold on day 18; Fig. 6Aiv). The gene encoding the selenium-binding-protein-like protein displayed a very similar profile (Fig. 6Av). Finally, cystatin B-like protein was downregulated at all tested time points in response to the non-virulent treatment (more than 3.1-fold; Fig. 6Avi), whereas under the virulent treatment it was significantly upregulated at day 15 and then downregulated at day 18 (more than 3.5-fold; Fig. 6Avi).

In our system, no significant expressional variation was observed among the general cellular process corresponding to reference genes (Fig. 6Bi–iii).

This work represents the first transcriptomic study investigating the physiological response of a scleractinian coral (P. damicornis) to its specific and temperature-sensitive pathogen, V. coralliilyticus YB1. More specifically, subtractive hybridization of mRNA from the holobiont subjected to balneation with bacteria at a constant (25°C) or increasing (25 to 32.5°C) temperature was used to identify immunity-related genes that appeared to be associated with bacterial stress or infection.

Although a previous study showed that bleaching was triggered by balneation of P. damicornis with V. coralliilyticus YB1 at 25°C (Ben-Haim Rozenblat and Rosenberg, 2004), we surprisingly failed to observe bleaching under these conditions. However, this difference may be due to between-study differences in the utilized holobiont populations, which are known to display genetic variability (Pinzon and LaJeunesse, 2010). Eugene Rosenberg's group used an isolate of P. damicornis from Eilat in the Gulf of Aquaba (Red Sea, 29.30°N), where seawater temperatures range from 20.5 to 27.3°C (Ben-Haim Rozenblat and Rosenberg, 2004; Loya, 2004). In contrast, we used clones of an isolate of P. damicornis from Lombock Island (Indonesia, 8°S), where seawater temperatures range from 26.8 to 29.4°C (IGOSS: Integrated Global Ocean Services System; http://iridl.ldeo.columbia.edu/SOURCES/.IGOSS/). Thus these holobiont populations may have adapted to their local thermal conditions, as has been well documented in other species, including Drosophila melanogaster, Podosphaera plantaginis, Scottolana canadensis, Daphnia magna and various crustaceans (Lonsdale and Levinton, 1985; Whiteley et al., 1997; Mitchell and Lampert, 2000; Ayrinhac et al., 2004; Laine, 2008). In addition, differential responses under similar experimental conditions may arise through the actions of the mucus-associated bacterial community, as proposed in the hologenome theory of evolution (Rosenberg et al., 2007b). Although still under some debate, this theory contends that changes in the mucus-associated bacterial population can cause the coral holobiont to adapt to changes in the environment, including the presence of pathogens (Leggat et al., 2007; Rosenberg et al., 2007a; Rosenberg et al., 2007b).

In the context of temperature-dependent virulence, no lytic symptoms were observed at the low temperature (as expected). The first lytic events that could be observed by visual monitoring were noted on day 12 (29.5°C) in the Tb treatment. Microscopic analysis revealed that these visible lytic events were preceded by the development of bacterial aggregates inside ectodermal cells (first observed on day 9; Fig. 3A). In addition, the tissues surrounding the bacteria aggregates were disorganized and the cellular organelles were mostly lysed (Fig. 3C). Similar aggregates were observed in a previous study by Ben-Haim et al. (Ben-Haim et al., 2003b), in which sections of P. damicornis infected by V. coralliilyticus YB1 were analyzed using the same microscopic approach. Although Vibrio species are usually considered to be extracellular pathogens, several studies have reported intracellular and viable Vibrio. This was evidenced for another coral pathogen, Vibrio shiloi, invading Oculina patagonica epithelial cells (Banin et al., 2000), and for Vibrio cholerae, which can colonize amoebas (Abd et al., 2007) and phagocytes (Ma and Mekalanos, 2010). These results suggest that the applied temperature increase allowed infection by V. coralliilyticus, and that this infection was responsible for the lytic events leading to coral death.

To address the possibility that the detected aggregates could represent another bacterial species present in coral tissues, we quantified the amount of V. coralliilyticus in tissue samples from the Cb and Tb treatments at all time points. Our results revealed that the amount of V. coralliilyticus 16S rRNA in the P. damicornis tissues increased significantly by day 6 and remained high at the end of the experimental period. Collectively, these results suggest that V. coralliilyticus penetrates coral tissues, multiplies inside the ectodermal cells and triggers the lysis of organelles, cells and coral tissues.

In terms of transcriptomics, our SSH and EST analyses identified several interesting genes related to innate immunity as potentially being involved in the bacterial challenge. Six of them were chosen and used to monitor the response of coral to bacterial stress or infection.

Three of the chosen candidates corresponded to lectin-like molecules, which are known to play key roles in the pathogen-recognition processes of various invertebrates (Iwanaga and Lee, 2005; Schulenburg et al., 2008; Dunn, 2009). The first selected lectin-like molecule was similar to a mammalian P-selectin (Ley, 2003); the second displays similarities for concanavalin A-like (Rudenko et al., 1999; Grishkovskaya et al., 2000; Tisi et al., 2000; Buschiazzo et al., 2002); and the third, PdC-lectin, is likely to play a crucial role in the acquisition of zooxanthellae by coral cells (Vidal-Dupiol et al., 2009). Indeed, PdC-lectin is highly similar to another mannose binding lectin, the Millectin of Acropora millepora, which was shown to bind to both zooxanthellae and V. coralliilyticus (Kvennefors et al., 2008).

We found that the PdC-lectin and concanavalin A-like transcripts displayed similar expression patterns in non-virulent and virulent experiments. Their levels were upregulated in the virulent experiment from day 6 to 12, plateaued on day 15 (when the tank temperature reached 32.5°C) and then returned to the control level by day 18 (Fig. 6Ai,ii). This transcriptional induction is not surprising, as lectin-like genes have been found to be induced in a large set of studies on experimental infections of invertebrate species (O'Rourke et al., 2006; Huang et al., 2007; Gowda et al., 2008; Soonthornchai et al., 2010; Zhang et al., 2010). The day 18 decrease is more surprising, but many coral cells were undergoing necrosis at this stage (Fig. 1). Consequently, we can hypothesize that this day 18 decrease illustrates the physiological collapse of the coral, which could lead to a general decrease in the transcription levels of the genes responsible for a variety of functions. A general transcriptional decrease under extreme physiological conditions (lethal stress) has been observed previously in other species (Wang et al., 2006; Su et al., 2009).

In coral subjected to the non-virulent treatment, we observed a drastic transient decrease in the transcriptional expression levels of PdC-lectin and concanavalin A-like protein (Fig. 6Ai,ii). This transient decrease peaked on day 9 and the transcripts had returned to control levels by day 12. This observation is curious because no symptoms of infection were observed in this group. However, a similar downregulation of PdC-lectin was previously observed in P. damicornis nubbins confronted with thermal stress (Vidal-Dupiol et al., 2009), and occurred just prior to the dissociation of symbiosis and the onset of bleaching. Based on the putative role of PdC-lectin in the process of zooxanthellar acquisition, we previously hypothesized that this downregulation could lead to bleaching (Vidal-Dupiol et al., 2009). As corals have been reported to bleach under a wide variety of stress conditions, including high and low seawater temperature, increased heavy metal concentrations and decreased salinity (Brown, 1997; Douglas, 2003), we herein hypothesize that downregulation of PdC-lectin and concanavalin A-like protein could be induced by various stresses that lead to bleaching. The transient downregulation of these transcripts observed in the Cb treatment could be the result of stress induced by bacterial balneation. Although we did not observe any decrease in zooxanthellar density among these samples, it is possible that there may have been a temporary ‘bleaching-like effect’ without the appearance of visible symptoms. The further observation that the transcription levels of these genes returned to control levels allows us to hypothesize that the corals may have responded to the bacterial stress by reallocating their resources to prepare for an antibacterial response (e.g. the induction of ferritin expression; see below). Then, when the bacterial threat was gone, their allocations returned to ‘business as usual’.

The third lectin, the P-selectin–like protein, was strongly downregulated after the initial addition of bacteria in both treatment groups, and this decrease was maintained for the duration of the experiments. The P-selectins have been well studied in vertebrate models, where they are known to be involved in various immune processes. However, their function in invertebrates is not yet known, making it somewhat difficult to interpret our present results. Interestingly, however, the P-selectins are known to be involved in adhesion processes (Rudenko et al., 1999; Grishkovskaya et al., 2000; Tisi et al., 2000; Buschiazzo et al., 2002), and studies have shown that some virulence factors present in the excretion–secretion products of marine Vibrio spp. reduced the adhesive properties of invertebrate hemocytes (Choquet et al., 2003; Allam and Ford, 2006; Labreuche et al., 2006a; Labreuche et al., 2006b; Mateo et al., 2009). Hemocyte-like cells have not yet been identified in scleractinians, but endodermal and circulating granular and agranular cells may have phagocytic properties (Mullen et al., 2004). Consequently, it may be useful in the future to explore the potential effect of V. coralliilyticus on the adhesive properties of P. damicornis phagocytes.

The three other candidates chosen to monitor the effect of V. coralliilyticus on the physiology of P. damicornis were cystatin B-like, ferritin-like and selenium-binding-protein-like. Cystatin B is a cytoplasmic protein that reversibly inhibits the cysteine proteases to affect protein turnover, and also plays a role in antibacterial defense (Zavasnik-Bergant, 2008). In vertebrates, cystatin B is upregulated in monocytes exposed to lipopolysaccharide challenge and treatment with immune enhancers (e.g. nitric oxide) (see Zavasnik-Bergant, 2008). In invertebrates, cystatin B is reportedly upregulated in response to bacterial challenge (Lefebvre et al., 2004) and parasitic infection (Kang et al., 2006). Our present results are in agreement with the previous findings, in that the transcriptional expression of the cystatin B-like protein increased when corals were infected by bacteria (virulent treatment). This induction peaked at day 15 and then sharply decreased by day 18. This drastic decrease could also be explained by the physiological collapse described above. Downregulation of cystatin B-like protein transcript levels was also observed in P. damicornis exposed to non-virulent bacteria, perhaps because of the existence of a trade-off mechanism allowing the coral to maximize its energy re-allocation to external defenses (i.e. the production of antibacterial compounds for release into the mucus). This phenomenon has been described for several invertebrate species in response to stress (Agell et al., 2004; Brulle et al., 2007; Lourenco et al., 2009).

The other putative immune effectors considered in the present study are two metal-binding proteins, a ferritin-like protein and a selenium-binding-protein-like protein. Ferritins, which can sequester iron, play dual roles in detoxification and iron storage (Arosio et al., 2009; Arosio and Levi, 2010), and the selenium-binding protein covalently binds selenium (Bansal et al., 1989; Jeong et al., 2009). Interestingly, iron appears to be essential for bacterial growth (Doherty, 2007), especially for Vibrio species (Wright et al., 1981; Tolmasky and Crosa, 1991; Wyckoff et al., 2007), and selenium is a trace element crucial for the survival of all living organisms through the formation of selenocysteine, a modified amino acid largely implicated in antioxidant defense (e.g. glutathione peroxidase, thioredoxin reductase) (Burk et al., 2003; Stadtman, 1996). Upregulation of ferritin and selenium-binding-protein by infected host cells can trigger iron and selenium sequestration, thereby reducing the amount of these essential trace elements necessary for bacterial growth. Infection-induced iron and selenium sequestration by ferritin and selenium-binding-protein have been observed in several other invertebrate–bacteria interactions (Beck et al., 2002; Song et al., 2006; Altincicek et al., 2008; Li et al., 2008; Kong et al., 2010; Simão et al., 2010), and the crucial role of ferritin in antibacterial defense was recently demonstrated in a study showing that the injection of recombinant ferritin into Vibrio harveyi-infected shrimp (Peneaus monodon) increased the host survival rate (Maiti et al., 2010).

Under non-virulent conditions, both ferritin-like and selenium-binding-protein-like were upregulated, perhaps reflecting that bacteria have been detected and the coral is preparing its defenses by sequestering iron and selenium. Under our virulent experimental conditions, the expression of ferritin-like and selenium-binding-protein-like corresponding genes was significantly downregulated during the early stage of infection, but upregulated near the end of the infection. This could illustrate that V. coralliilyticus is able to inhibit the intracellular immune response during the initial stages of infection. A similar phenomenon was observed in abalone following infection by V. harveyi (Travers et al., 2009). The upregulation of ferritin-like and selenium-binding-protein-like genes later in the process of infection could limit bacterial infection by sequestering iron and selenium. In addition, zooxanthellae under thermal stress have been shown to suffer photo-inhibition leading to ROS over-production (Weis, 2008). Thus, the overexpression of ferritin and selenium-binding-protein might also participate in ROS detoxification. Nevertheless, this late increase seems insufficient to overcome the Vibrio infection process, at least under our experimental conditions.

In conclusion, the present study represents the first molecular examination of the response evoked by a scleractinian coral when confronted by its pathogenic Vibrio (V. coralliilyticus). Several genes that appeared to be involved in the coral immune response were identified, and their expression patterns were studied in corals exposed to virulent and non-virulent bacteria. The results revealed clear differences for some of the candidate genes, which could prove useful as functional biomarkers for measuring coral health and immunity in monitoring programs and public aquariums. Future studies will be required to fully examine the molecular and cellular mechanisms involved in this process, but our study also suggests several new hypotheses concerning the effect of V. coralliilyticus on the physiology of P. damicornis.

The authors are indebted to Marc Manetti for his help with the experimental procedures, Céline Cosseau for many helpful discussions and Andrew Carroll for English proofreading and helpful discussions. We thank Alain Pigno and Boris Rota from the Cap d'Agde Public Aquarium for their help with the project, and Jérôme Bossier for helping with the statistical analyses. We also thank the Department of Environmental Sciences and Management (Prof. J. P. Thome) and the Dean of the Faculty of Sciences (Prof. J. M. Bouquegneau), both of the University of Liege, for financial support. Finally, the authors thank Mary-Alice Coffroth for allowing us to use cultures from the BURR Culture Collection.