Main

Coral reefs are estimated to harbour around one third of all described marine species6, and their productivity supports around one quarter of marine fisheries, but declines in coral abundance and wholesale loss of reef habitats are one of the most pressing environmental issues of our time. The major architects of coral reefs, the scleractinian corals, are anthozoan cnidarians that form obligate endosymbioses with photosynthetic dinoflagellates of the genus Symbiodinium (Fig. 1b). The symbionts confer on the coral holobiont the ability to fix CO2 and to deposit the massive aragonite (a form of calcium carbonate) skeletons that distinguish reef-building corals from other anthozoans such as sea anemones. The association is fragile, however, and collapses under stress. Despite the ecological and economic significance of corals, the molecular mechanisms underlying much of coral biology—including stress responses and disease—remain unclear, but it is clear that corals retain much of the complex gene repertoire of the ancestral metazoan7. To address the lack of molecular data for reef-building corals, we determined the whole-genome sequence of A. digitifera (Fig. 1a–h), a dominant species on Okinawan reefs. Not only are Acropora species the dominant reef-building corals of the Indo-Pacific, but they are also among the most sensitive of corals to increased seawater temperatures8.

Figure 1: The coral Acropora digitifera and an early occurrence of corals on Earth.
figure 1

a, The colony the genome of which was sequenced in the present study. This colony is maintained in aquarium culture at the Sesoko Station, University of the Ryukyus, and is thus available for further investigation of the genome. Scale bar, 10 cm. b, Polyps of the coral showing the presence of symbiotic dinoflagellates (Symbiodinium sp.) (inset, enlargement). c, Natural spawning of the coral. dh, Eggs, embryos, larva and primary polyp of A. digitifera, from which messenger RNA was extracted for transcriptome analyses. Scale bar, 200 μm. d, Fertilized egg; e, blastula at the prawn chip stage; f, gastrula; g, planula larva; and h, primary polyp. i, Molecular phylogeny of corals. 94,200 aligned amino acid positions of proteins encoded by 422 genes were obtained from the sponge Amphimedon queenslandica, from the cnidarians A. digitifera, Nematostella vectensis and Hydra magnipapillata, and from the triploblasts Tribolium castaneum, Drosophila melanogaster, Branchiostoma floridae, Danio rerio and Homo sapiens. The sequences were analysed using maximum likelihood methods, with the plant Arabidopsis thaliacia and the choanoflagellate Monosiga brevicollis serving as outgroups. The scale bar represents 0.1 expected substitutions per site in the aligned regions. The topology was supported by 100% bootstrap value. Approximate divergent times of the occurrence of basal chordates and divergence of vertebrates lineages are shown. This analysis indicates a deeper divergence of Acropora and Nematostella, approximately 500 million years (Myr) ago.

PowerPoint slide

On the basis of flow cytometry, the A. digitifera genome is approximately 420 Mbp (Supplementary Figs 1 and 2) and is therefore similar in size to that of the sea anemone Nematostella9. Sperm from a single colony served as the source of DNA for sequencing using a combination of Roche 454 GS-FLX10 and Illumina Genome Analyser IIx (GAIIx)11 methods. The genome was sequenced to approximately 151-fold coverage (Supplementary Table 1), enabling the generation of an assembly comprising a total of 419 Mbp (Supplementary Tables 2–5; contig N50 = 10.7 kbp and scaffold N50 = 191.5 kbp; Supplementary Fig. 3). The genome is approximately 39% G+C (Supplementary Fig. 4), and contains 23,668 predicted protein-coding loci (Supplementary Table 6). Transposable elements occupy approximately 12.9% of the genome (Supplementary Table 7). The coral gene set is comparable in size and composition with those of Nematostella vectensis9 and Hydra magnipapillata12 (Supplementary Tables 6, 8 and 9). The genome browser is accessible at http://marinegenomics.oist.jp/acropora_digitifera (Supplementary Fig. 5). Approximately 93% of the A. digitifera genes have matches in other metazoans (Supplementary Fig. 6a), and of these, 11% have clear homology only among expressed sequence tag (EST) data from corals13 (Supplementary Fig. 6b), suggesting the presence of a considerable number of coral-specific genes.

Corals are morphologically very similar to sea anemones, but their evolutionary origins are obscure. Reef-building Scleractinia first appeared in the fossil record in the mid-Triassic (approximately 240 million years ago)5, but were already highly diversified, suggesting much earlier origins. The availability of fully sequenced genomes for three cnidarians—Acropora (the present study), Nematostella9 and Hydra12— allowed the estimation of the depth of the divergence between corals and other metazoans. Molecular phylogenetic analyses based on an alignment of 94,200 amino acid positions suggest a divergence time of Acropora and Nematostella between 520 to 490 million years ago (the late Cambrian or early Ordovician) (Fig. 1i). The implied earlier origin of Scleractinia indicates that corals have persisted through previous periods of major environmental change, including the mass extinction event at the Permian/Triassic boundary, when global CO2 and temperature were much higher than at present. However, whereas the Scleractinia as a lineage has persisted on evolutionary time scales, whether modern coral reefs can adapt to rapid environmental change on ecological time scales is a very different question.

The obligate endosymbiosis of corals dates at least from the mid-Triassic, and the longevity of this association might therefore be expected to have resulted in changes within the coral genome. We were unable to find any Symbiodinium DNA sequences in the coral genome, hence there is as yet no evidence for horizontal gene transfer from symbiont to host (Supplementary Fig. 6). However, comparative analyses indicated that, in the case of Acropora, the coral host might be metabolically dependent on the symbiont. Using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database14, the metabolic repertoire of Acropora was compared to that of its non-symbiotic relative, the sea anemone Nematostella (Supplementary Table 10), leading to the identification of an apparent metabolic deficiency in Acropora. The biosynthesis of cysteine from homocysteine and/or serine requires the activities of two enzymes, cystathionine β-synthase (Cbs) and cystathionase (cystathionine γ-lyase; Cth) (Table 1). Whereas we were able to identify genes encoding the latter in both A. digitifera and Nematostella, the former could not be identified in Acropora despite a clear match being present in Nematostella (Supplementary Fig. 7). Although extensive transcriptomic data are available for various Acropora spp13, we could find no evidence for a Cbs transcript in any of these. Moreover, whereas a polymerase chain reaction (PCR) strategy confirmed the presence of Cbs in some other corals, Galaxea fascicularis, Favites chinenis, Favia lizardensis and Ctenactis echinata, no amplification products could be obtained for two different Acropora species (Table 1 and Supplementary Fig. 8). Although the analyses presented here do not rigorously exclude the presence of Cbs activity in Acropora, they raise the intriguing possibility of a metabolic basis for the obligate nature of symbiosis in Acropora; differences in dependency could potentially explain not only the phenomenon of symbiont selectivity, but also the high sensitivity of Acropora to environmental challenges.

Table 1 The presence or absence of a gene encoding Cbs for l-cysteine biosynthesis

Reef-building corals typically inhabit shallow and relatively clear tropical waters and are therefore constantly exposed to high levels of ultraviolet irradiation. As corals are particularly susceptible to bleaching when exposed to both raised temperatures and high solar radiation2,4, one intriguing question is how corals protect themselves against ultraviolet damage. Photo-protective compounds, such as the mycosporine amino acids (MAAs), have been isolated from corals15,16 but, because similar compounds have been identified in algae, the sources of these compounds were unknown. Recently a short (four-step) pathway encoded by a gene cluster (DHQS-like, O-MT, ATP-grasp and NRPS-like) (Fig. 2 and Supplementary Figs 9–12) has been demonstrated to be both necessary and sufficient in the cyanobacterium Anabaena variabilis to convert pentose-phosphate metabolites to shinorine, a photo-protective MAA17. Scanning the available whole-genome data allowed us to identify clear homologues of all four members of the cyanobacterial shinorine gene cluster in both A. digitifera and N. vectensis (Fig. 2), indicating that both the coral and the sea anemone have the ability to carry out de novo synthesis of ultraviolet-protective compounds. Hence, MAA synthesis in corals and other cnidarians is not symbiont dependent.

Figure 2: The genes required for the biosynthesis of shinorine are present in anthozoan cnidarians.
figure 2

Top, the organization of the gene cluster involved in the biosynthetic pathway of the photo-protective molecule shinorine, a mycosporine-like amino acid, in the cyanobacterium Anabaena variabilis. Bottom, the presence of corresponding genes in various organisms is indicated (+). The Acropora and Nematostella genomes contain homologues of each of the four genes, in which DHQS-like and O-MT are fused with each other.

PowerPoint slide

Surveys of Acropora for genes associated with innate immunity18, apoptosis19 and autophagy19 indicate not only the complexity of these systems in Acropora (Supplementary Figs 13–23), but also that the coral innate immune repertoire is more sophisticated than that of Nematostella. For example, whereas a single canonical Toll/TLR protein is present in N. vectensis18, the Acropora genome encodes at least four such molecules, as well as five IL-1R-related proteins and a number of TIR-only proteins (Fig. 3). Likewise, the Acropora repertoire of NACHT/NB-ARC domains, which are characteristic of primary intracellular pattern receptors20, is again highly complex: an order of magnitude more NACHT/NB-ARC domains are present in coral than in other animals (Supplementary Table 11), and some of these cnidarian proteins have novel domain structures (Supplementary Fig. 23b). In terms of the apparent expansion and divergence of NACHT-encoding genes, the coral resembles amphioxus21, the sea urchin22 and angiosperms23. The greater complexity of the coral innate immunity network may in part reflect adaptations associated with the symbiotic state and coloniality.

Figure 3: Repertoires of TIR-domain-containing proteins of three cnidarians.
figure 3

The schematic representation of the domain structures of TIR-domain-containing proteins identified in A. digitifera, alongside the corresponding complements from Nematostella vectensis and Hydra magnipapillata. The repertoire of Toll/TLR, IL-1R-like and TIR-only proteins is significantly more complex in the case of A. digitifera than in N. vectensis or H. magnipapillata. TIR, TIR domain. DEATH, DEATH domain. IG and IGc2, Ig domain. LRR, LRY-TRY, LRR-CT and LRR-NT, leucine-rich repeats.

PowerPoint slide

The coral repertoire of genes with predicted roles in skeleton deposition is of particular interest given the likely impact of ocean acidification resulting from rising atmospheric CO2 on coral calcification. Surveys of the Acropora genome for specific groups of proteins associated with calcification, including the eukaryotic-type carbonic anhydrases24 are given in Supplementary Table 12. In general, the soluble fraction of the organic matrix in scleractinian corals is very rich in acidic amino acids, and has a particularly high aspartic acid composition25. A number of candidate organic-matrix proteins were identified in Acropora (Supplementary Fig. 24). For several of these, orthologues could be identified in A. millepora and/or A. palmata but only one of these (Adi-SAP6) was found in other coral species (Supplementary Table 13). Galaxins, first purified from the coral Galaxea fascicularis, are unique to corals and are the only coral skeletal matrix protein for which the complete primary structure has been determined26. However, galaxin possesses neither acidic regions (the fraction of Asp+Asn in the galaxin is 9.7%) nor obvious Ca2+-binding domains26. Four genes encoding galaxin-related proteins were identified in the A. digitifera genome (Supplementary Fig. 25), including two likely A. digitifera homologues of Gfa-galaxin.

Here we decoded the 420-Mbp genome of the reef-building coral Acropora digitifera, with the aim of providing a platform for understanding the molecular basis of symbiosis and responses to environmental change. Some of the main findings are: (1) a relatively deep divergence of the lineage leading to the reef-building corals; (2) although we could find no evidence for horizontal gene transfer from symbiont to coral despite the long evolutionary history of the association, Acropora may have lost a gene essential for cysteine biosynthesis and thus be metabolically dependent on its symbionts; (3) the coral host has the ability to independently carry out de novo synthesis of the MAA family of photoprotective compounds; (4) the innate immune repertoire of coral is highly complex in comparison with the non-symbiotic and solitary sea anemone Nematostella; and (5) a number of coral-specific gene families are likely to have evolved in the context of calcification. These data also provide a basis for systems biology approaches to understanding the establishment, function and collapse of coral symbioses. If and when a whole-genome sequence becomes available for the dinoflagellate symbiont of corals Symbiodinium sp. (zooxanthellae), these resources will together provide additional perspectives on the symbiosis and a powerful resource for understanding the response of the holobiont to environmental stresses such as raised seawater temperatures or ocean acidification.

Methods Summary

Sperm DNA obtained from a single colony of the coral Acropora digitifera was used for genome sequencing by Roche 454 GS-FLX10 and Illumina Genome Analyser IIx (GAIIx)11. The 454 shotgun and paired-end reads were assembled de novo by GS De novo Assembler version 2.3 (Newbler, Roche)10, and subsequent scaffolding was performed by SOPRA27 and SSPACE28 using the Illumina mate-pair information. Transcriptome analysis was also performed. A set of gene model predictions (the A. digitifera Gene Model v. 1) was generated mainly by AUGUSTUS29, and a genome browser has been established using the Generic Genome Browser (GBrowser) 2.17. The annotation and identification of Acropora genes were performed by three approaches, individual methods or combinations of the methods: reciprocal BLAST analyses, screening the gene models against the Pfam database30 and phylogenetic analyses.

Online Methods

Biological specimen

Under permits from the Aquaculture Agency of Okinawa Prefecture (the number 20–27), part of an A. digitifera colony was collected and has subsequently been maintained in an aquarium at the Sesoko Station, Tropical Biosphere Research Center, University of the Ryukyus.

The number of chromosomes, diploidy and genome size of Acropora digitifera

The number of chromosomes was determined by their preparation from nuclei of embryonic cells. The diploidy of the genome was examined by fluorescent in situ hybridization (FISH) of BAC clones31, which were constructed in pKS146 (ref. 32). The genome size was estimated by flow cytometry33 using sperm nuclei from the same colony that was used to sequence the genome.

Genome sequencing and assembly

The sperm was obtained from the single colony and sperm DNA was used for genome sequencing and BAC library construction. Genome sequence data were obtained using single read, paired-end and mate-pair protocols on the Roche 454 GS-FLX10 and Illumina GAIIx11 instruments. The genomic DNA was fragmented, libraries prepared and sequencing conducted according to the manufacturer’s protocols. The 454 shotgun and paired-end reads were assembled de novo by GS De novo Assembler version 2.3 (Newbler, Roche)10 in heterozygotic mode with adjusted algorisms to reflect an increase in the expected variability in sequence identity. Possible PCR duplicates in Illumina mate-pair reads were removed by MarkDuplicates in Picard tools (http://picard.sourceforge.net), and then subsequent scaffolding of the 29,765 Newbler output was performed by SOPRA27 and SSPACE28 using the Illumina mate-pair information. Gaps inside the scaffolds were closed with Illumina paired-end data using GapCloser34. To overcome potential assembly errors arising from tandem repeats, sequences that were aligned to another sequence over 50% of the length by BLASTN (1 × 10−50) were removed from the assembly35.

Transcriptome analyses

RNA was isolated from eggs, gastrulae, planulae, polyps and adults. Total RNA was extracted following the manufacturer’s instructions (Invitrogen) and purified using DNase and an RNeasy micro kit (QIAGEN). Transcriptome libraries for 454 GS-FLX were prepared36 and sequenced as per manufacturer’s instructions. In addition, Illumina 50-bp paired-end RNA-seq sequencing was performed. All high-quality sequences (quality value ≥15) were assembled by a Velvet/Oases assembler37 with hash length 27.

Gene prediction

A set of gene model predictions (the A. digitifera Gene Model v. 1) was generated using AUGUSTUS29. AUGUSTUS 2.0.4 was trained on the 877 EST assemblies recommended by PASA38 for this purpose. The gene models were created by running AUGUSTUS on a repeat-masked genome produced by RepeatMasker39, and improved by PASA38. A genome browser has been established using the assembled genome sequences using the Generic Genome Browser (GBrowser) 2.17 (ref. 40).

Identification of Acropora genes involved in the response to environmental change

Three approaches, individual methods or combinations of the methods, were used to annotate the protein-coding genes in the A. digitifera genome. A primary approach to the identification of putative orthologues of A. digitifera genes was reciprocal BLAST analysis. This was carried out on the basis of mutual best hit in BLAST analyses for human, mouse, or Drosophila genes against the A. digitifera gene models (BLASTP) or the assembly (BLASTN). A second approach used in the case of genes encoding proteins with one or more specific protein domains, was to screen the merged models against the Pfam database (Pfam-A.hmm, release 24.0; http://pfam.sanger.ac.uk)30, which contains 11,912 conserved domains using HMMER (hmmer3)41. In the case of complex multigene families, a third annotation method was used; sets of related sequences were subjected to phylogenetic analyses to determine more precisely orthology relationships. For these purposes, amino acid sequences were aligned using ClustalW42 or ClustalX42 under the default options. Gaps and ambiguous areas were excluded using Gblocks 0.91b43 with the default parameters and then checked manually. On the basis of the alignment data sets, phylogenetic trees were constructed by neighbour joining and/or maximum likelihood. Calculations and tree construction were performed in SeaView44. The KEGG pathway database14 was used to examine the metabolic repertoire of Acropora in comparison to that of the sea anemone Nematostella.