Adult colonies of Stylophora pistillata were collected either from the field or from corals maintained in tanks for at least 20 years within the aquarium system of the Centre Scientifique de Monaco. The corals collected from the field (the Israeli Nature and National Parks Protection Authority approved the collection of corals in this study, permit No. 2011/38353) came from the Gulf of Aqaba in the Red Sea and were transferred after collection to tanks at the Marine Station of Eilat, Israel. The tanks were supplied continuously with seawater from the Red Sea. After an acclimation period of two weeks, colonies of S. pistillata were separated into different tanks that exposed the colonies to different environmental conditions, as described below. The cultured corals were maintained in a 300-liter aquarium supplied with seawater from the Mediterranean Sea (exchange rate 2% h⁻¹) under controlled conditions, as follows: semi-open circuit, temperature of 26.0±0.2°C, salinity of 38.2‰, and light intensity of 175 µmol photons m⁻² s⁻¹ (using fluorescent tubes from Custom Sea Life®) under a 12∶12 h photoperiod. The corals were fed three times weekly with a mixture of Artemia salina nauplii, frozen adults of Artemia salina, and frozen krill.

Colonies from field and cultured sources were grown for two weeks under different conditions (Fig. 2) in order to maximize the expression of the greatest variety of genes. These conditions include different temperatures (18°C, 26°C and 32°C), different light/dark cycles, different pH levels (7 to 8.2), and either fed or not fed. Different field conditions were also included, with colonies being exposed to depths ranging from 5 to 50 meters. After exposure to different treatments, each colony was divided into fragments and snap-frozen in liquid nitrogen.

Total RNA was isolated from each of the fragments from the different treatments described above using TRIzol (Invitrogen) according to manufacturer’s instructions. The quality of all of the RNA was checked using a Bioanalyzer (Agilent) RIN ≥9.5, and pools with the same amount of RNA were then created. The cDNA library was constructed using a Clontech SMARTer PCR cDNA synthesis kit and amplified using the Advantage 2 PCR kit according to the manufacturer's instructions. Subsequently, 2 µg of the amplified cDNA was normalized using the Trimmer kit (Evrogen) following the manufacturer's instructions and purified using the Qiaquick PCR purification kit (Qiagen). The normalized and non-normalized cDNA was sent to Roche for further analysis. Sequencing was performed using a 454 GS-Flx instrument according to the manufacturers' instructions (Roche). In order to obtain the abundant and rare transcripts, the library placed on the 454 plate was divided into two, half containing normalized cDNA, and the other half with non-normalized cDNA. The normalized and non-normalized cDNAs were sheared by sonication to produce short random fragments (300–400 bp) appropriate for 454 sequencing, and oligonucleotide adaptors were then ligated to the fragmented sequences. The 454 GS-Flx running plate was divided for internal comparison of normalized and non-normalized libraries (data not shown).

The sequences were submitted to Newbler assembler version 2.6 for de novo assembly of 454-sequenced EST libraries using the default parameters. The assembled sequences were first automatically annotated with the SwissProt databases and then with several species-specific databases using the BLAST program. The species utilized in this analysis are as follows: the sea anemones Nematostella vectensis, Aiptasia pallida, Metridium senile, and Anemonia viridis; the stony corals Acropora millepora, Acropora palmata, Acropora digitifera, Montastraea faveolata, and Porites astreoides; the hydrozoans Clytia hemisphaerica and Hydra vulgaris; and the protist Symbiodinium sp. The best matches obtained using the following blast parameters: e-value ≤1e-¹⁰, best-hit-overhang  = 0.1, best-hit-score-edge  = 0.1; in order to avoid random short hits. Pathway analysis was later performed by running a pairwise sequence search compared with the KEGG-curated set of human proteins [45].

The alignments of all amino acid sequences were performed with the Multalin server [46] and phylogenetic relationships were investigated using Bayesian techniques as implemented in the computer program MrBayes v3.1.2 [47], starting from a random tree, generating 3,500,000 generations with sampling every 1000 generations, and with four chains in order to obtain the final (consensus) tree and to determine the posterior probabilities at the different nodes.

The normalized and non-normalized cDNA libraries constructed from S. pistillata holobiont RNA starting materials were based on an RNA pool collected from different environmental conditions to maximize the diversity of rarely expressed genes. Normalization of the library decreased the amounts of abundant transcripts and maximized the chances of finding new genes. We divided the 454 plate to run normalized and non-normalized cDNA libraries to obtain both abundant and rare transcripts. The two datasets were merged before assembly to generate a database of 523,533 sequenced reads. Assembly of these reads produced in 15,052 contigs with a mean length of 1,078 bp and N50 1,256 bp. These results are available from the NCBI and from (http://data.centrescientifique.mc/). Using BLAST searches against SwissProt database we were able to annotate 51% of the obtained sequences.

S. pistillata transcript similarity searches were first performed against proteome libraries using blastX (see methods for specific parameters) (Table 1). At this level, 10,050 out of the 15,052 contigs (∼67%) exhibited a positive match to proteins from one or more species. As expected, most of the S. pistillata transcripts matched sequences from the coral A. digitifera [7] and N. vectensis. The human proteome ranked third in terms of the number of hits, followed by the third diploblast Hydra magnipapillata. The two protostomes (the fruit fly and nematode) presented the lowest numbers of hits and were ranked last. The general tendency of the homolog presence across lineages is even more apparent if we group S. pistillata homologs across taxonomic groups, as shown in Fig 3. Most of the coral transcripts (5,381) were found in all three groups followed by transcripts found specifically in at least one of the species belonging to the diploblast group, (2,977). Moreover, a large set of diploblast homologs was found among the deuterostome representatives (1,579), supporting the findings of previous work in cnidarians [3], [4], [6], [10] that has shown an unexpectedly high homology of diploblasts with deuterostomes in comparison to protostomes [48].

To study the differences between our EST library and those of the Cnidaria in general, we compared it to the available cnidarian EST libraries (see methods). Cross matches of Stylophora transcript homologs across Cnidaria (Table 2 and Fig. 4) showed 13,216 hits with stony corals, 8,787 with anemones and 6,246 with hydrozoans. Out of these, 5,988 were common to all three families.

After the evolutionary comparisons of homology between related and non-related lineages were complete, we successfully classified 7,764 S. pistillata transcripts using the STRING database v.6.3 [49] and a stringent assignment process. The graphical distribution of functional classes is illustrated in Figure 5. When comparing functional data between S. pistillata transcripts and the human proteome, we observed the presence of equivalent functional elements. These elements are sufficient to reconstitute key processes in signaling and cell adhesion, which we chose to further explore by focusing on Wnt and BMP pathways (Fig. 6). The wingless (Wnt) and bone morphogenetic protein (BMP) pathways are essentially independent signaling mechanisms, although they often regulate similar biological processes. Using DAVID bioinformatics resources [50], we revealed the presence of equivalent functional elements in the Wnt and TGF-β signaling pathways (Fig. 6).

The transcriptome reported in this study comprises 15,052 sequences, which is less than the estimated number of genes in the cnidarian genomes described to date (32,338, 27,273 and 23,668 for Hydra magnipapillata, Nematostella vectensis, and Acropora digitifera, respectively). The raw reads were also assembled by MIRA 3.02 software [54] which produced a significantly higher number of contigs (48,075 ), however, a comparison between the two assemblies demonstrated that Newbler assembly was much less fragmented (data not shown), and hence was used for downstream analysis. We assume that the low number of contigs assembled is due to limited coverage, and that Newbler software produced an uncompleted albeit un-fragmented EST library.

The outcome of the cross matches from the Venn diagram shown in Figure 3 illustrates that the highest number of unique hits exists between S. pistillata and the stony corals, as expected. Actiniaria are also hexacorallia but are solitary animals and do not produce a skeleton. The number of unique cross matches between these two groups is lower, suggesting that biomineralization involves an important number of specific genes, as suggested by Shinzato et al. [7]. Finally, the lowest numbers of cross matches are obtained among the hydrozoans, which are more divergent (see the phylogenic tree in Fig. 4 obtained from Technau and Steele 2011), solitary and freshwater animals, which presumably accounts for part of the greater difference in the transcriptome.

It should be noted here that the cDNA library was constructed from holobiont tissue; hence, both zooxanthellae and host sequences are present in our library. Most coral libraries are constructed from larvae, which do not contain symbionts, with P. damicornis and P. astreoides being exceptions. However, a blastX comparison with Symbiodinium EST libraries [51], [52] downloaded from NCBI shows that few sequences mapped exclusively to the Symbiodinium genome and not to the coral A. digitifera (375 out of 15,052, i.e., ∼ 2.5% of sequences) illustrating that most of the contigs belong to the coral itself.

Among Stylophora ESTs, there are sequences coding for a specific organic matrix protein that provides the skeleton shape of the coral. The only protein to have been fully characterized from the calcifying matrix of scleractinian corals is galaxin, which was originally identified from the coral Galaxea fascicularis [53]. Based on similarity searches performed in an EST library from Acropora millepora [3], [54], Reyes-Bermudez et al. [55] reported the sequences of three genes encoding galaxin-related molecules and their expression patterns during settlement and metamorphosis. A fourth galaxin protein, galaxin 2, has been characterized by Shinzato et al. [7], based on the sequencing of Acropora digitifera. blastX and blastn searches of the galaxin sequences against the S. pistillata EST library identified two ESTs with strong similarity. A phylogenetic tree generated by MrBayes program classified these ESTs to galaxin 2 and galaxin-like 1 (Figure 7). No homolog for Amgalaxin-like 2 was found, probably since its expression is restricted to post-settlement polyps and is not observed in adults [56]. The presence of an Amgalaxin-like 1 homolog in an adult expression library is surprising, as Reyes-Bermudez et al. [55] showed that its expression remains restricted to the settlement and metamorphosis phases. Though, a species-specificity in galaxin-like diversity cannot be excluded. The availability of new coral genomes, from both robust and complex clades, is then a key feature in understanding the role and the impact of matrix proteins in coral calcification.

Concerning the functional classification, half the transcripts show similarity to sequences in the KEGG database (51%). The distribution of functional classes is consistent with those in both the human and fruit fly complete proteomes [56], with the most abundant categories being RNA processing and modification (A) signal transduction (T), protein turnover (O), and translation (J), indicating that the coverage of the EST library presented here was adequate (Figure 5). We decide to focus on the Wnt and BMP signal transduction pathways because of their importance in coral development [4]. These two pathways do not share any major components and can function independently of each other, which is accomplished by means of different ligands, different receptors and different cytoplasmic and nuclear signal transducers. Most of the proteins found in humans are also present in corals. However, some of them are missing in the EST data obtained in this study due to 1) the technical methodology applied (for example, the chordin gene in the BMP pathway, absent from our EST library, is present in other cnidarian species [57], [58] and was found in a new library constructed using genome sequencing (in prep.) and/or 2) the absence of these genes in Cnidaria in cases where the genes appeared after the coelomata split, as observed for the secreted cytokines IFN gamma [personal data].

Wnt and BMP play roles in vertebrate biomineralization in addition to their roles in development. Studies have shown that both BMP and Wnt pathways are involved in osteogenesis [59], [60]. Furthermore, in adult coral, BMP expression is restricted to the skeletogenic ectoderm [42], and it has been suggested that BMP might be an organic matrix protein [61]. Although biomineralization and the nature of the calcifying organic matrix have been extensively studied in other taxa, such as mollusks [62], there have been relatively few corresponding studies of these elements in stony corals. We show here that some signaling pathways are conserved in S. pistillata providing new tools to understand the mechanisms of biomineralization. For instance, the study of BMP receptors and their location in polyps would help us to better understand the role of BMPs in adult corals.