A Snapshot of a Coral “Holobiont”: A Transcriptome Assembly of the Scleractinian Coral, Porites, Captures a Wide Variety of Genes from Both the Host and Symbiotic Zooxanthellae

Massive scleractinian corals of the genus Porites are important reef builders in the Indo-Pacific, and they are more resistant to thermal stress than other stony corals, such as the genus Acropora. Because coral health and survival largely depend on the interaction between a coral host and its symbionts, it is important to understand the molecular interactions of an entire “coral holobiont”. We simultaneously sequenced transcriptomes of Porites australiensis and its symbionts using the Illumina Hiseq2000 platform. We obtained 14.3 Gbp of sequencing data and assembled it into 74,997 contigs (average: 1,263 bp, N50 size: 2,037 bp). We successfully distinguished contigs originating from the host (Porites) and the symbiont (Symbiodinium) by aligning nucleotide sequences with the decoded Acropora digitifera and Symbiodinium minutum genomes. In contrast to previous coral transcriptome studies, at least 35% of the sequences were found to have originated from the symbionts, indicating that it is possible to analyze both host and symbiont transcriptomes simultaneously. Conserved protein domain and KEGG analyses showed that the dataset contains broad gene repertoires of both Porites and Symbiodinium. Effective utilization of sequence reads revealed that the polymorphism rate in P. australiensis is 1.0% and identified the major symbiotic Symbiodinium as Type C15. Analyses of amino acid biosynthetic pathways suggested that this Porites holobiont is probably able to synthesize most of the common amino acids and that Symbiodinium is potentially able to provide essential amino acids to its host. We believe this to be the first molecular evidence of complementarity in amino acid metabolism between coral hosts and their symbionts. We successfully assembled genes originating from both the host coral and the symbiotic Symbiodinium to create a snapshot of the coral holobiont transcriptome. This dataset will facilitate a deeper understanding of molecular mechanisms of coral symbioses and stress responses.


Introduction
Coral reefs are estimated to harbor roughly one-third of all described marine species, and their productivity supports approximately one-quarter of marine fisheries.The major architects of coral reefs, the scleractinian corals, are anthozoan cnidarians that form obligate endosymbioses with photosynthetic dinoflagellates of the genus Symbiodinium.The unicellular symbionts are harbored in the host coral's gastrodermal (endodermal) tissue in intracellular vacuoles known as symbiosomes [1], which are thought to originate from the plasma membrane of host cells during the initial acquisition of symbionts by a phagocytic process [2].Although many of the details of the interaction between the host and Symbiodinium remain to be explored, this association enables the massive rates of calcification that distinguish reef-building corals from other anthozoans, such as sea anemones and zoanthids.
Coral reefs face a range of environmental changes, including ocean acidification, seawater temperature increases, and declines in coral abundance.Extensive loss of reef habitats is one of the most pressing environmental issues of our time [3][4][5].Recently, increasing instances of ''coral bleaching'' have been observed.In most cases, coral bleaching is a breakdown of the mutualism between the coral and the photosynthetic dinoflagellate, resulting from a stress response to environmental perturbation.The integrity of the coral holobiont -a complex symbiosis between the coral animal, its endosymbiotic zooxanthellae, and an associated community of microorganisms is essential for maintaining coral health [6].The molecular mechanisms underlying the collapse of symbiosis and the bleaching response are complex and are still little understood.Bleaching is currently viewed as a host response to a compromised symbiont that is analogous to the innate immune responses that occur in other host-microbe interactions.Expulsion or elimination of the symbiont from host tissues is thought to involve a variety of mechanisms, including exocytosis, host cell detachment, and host cell apoptosis [7].
Genomic information for cnidarians has been accumulating; the genomes of two non-symbiotic cnidarians, the anemone Nematostella vectensis (Anthozoa) [8], and a hydra, Hydra magnipapillata (Hydrozoa) [9], have been sequenced.The genome of a scleractinian coral, Acropora digitifera, was decoded using nextgeneration sequencing (NGS) technology [10].The latter genome was estimated to comprise 420 Mbp and contains about 23,700 predicted protein-coding genes.Recently NGS-based transcriptome datasets have become available for several anthozoan cnidarians including several coral species: Acropora millepora [11,12], A. palmata [13] and Pocillopora damicornis [14], and a sea anemone Aiptasia pallida [15].In addition, transcriptomes of two cultured strains of Symbiodinium have been reported [16] and the nuclear genome of S. minutum was decoded [17].In order to better understand the molecular interactions, it is worthwhile to simultaneously capture the molecular states of all organisms comprising a coral holobiont.
The massive stony corals of the genus Porites are common, important reef builders in the Indo-Pacific Ocean.More than 80 named species and numerous unclassified forms have been identified [18].In contrast to Acropora species, Porites corals generally transmit their symbionts directly from parents to offspring (vertical transmission) rather than acquiring them anew from the environment (horizontal transmission) in each generation [19].Porites species have thicker tissues and appear more robust to thermal stress than other corals, such as Acropora, that have thinner tissues [20].Thus Porites corals have been used for comparative analyses of stress responses [21].Laboratory studies indicate that acidified seawater reduces calcification of Porites corals [22][23][24].Geochemical tracers, such as oxygen isotope ratios, strontiumcalcium ratios, and heavy metal concentrations in growth rings of the CaCO 3 Porites skeleton have been used to monitor changes in sea surface temperature, salinity, and/or marine pollutants [25][26][27][28].By such means, Porites taxa have been used to reconstruct past environmental changes so as to better understand tropical climate systems and to predict climate change [29,30].Nevertheless, molecular information about Porites corals is still limited.To address this deficiency and to improve the utility of Porites as an environmental indicator, we sequenced the transcriptome of Porites australiensis and its symbiotic algae (Figureo ˆ1) using next generation technology (Illumina HiSeq2000) and constructed a transcriptome dataset that contains a large proportion of a symbiotic alga.

RNA isolation, Transcriptome sequencing and assembly
The coral sample used in this study was collected at Sesoko island, Okinawa, Japan, under the Okinawa prefecture permission (Number: 20-69).A small Porites australiensis colony that has been maintained in the Sesoko research station at the University of the Ryukyus for five years (Figureo ˆ1) was used in this study.A fragment (2 cm diameter) from the coral was snap frozen in liquid nitrogen and pulverized with an iron mortar and pestle.Total RNA was isolated using an RNeasy RNA extraction kit (QIAGEN).Total RNA was then fragmented into about 200 bp lengths and an RNA-seq library was prepared using a TruSeq RNA Sample Prep Kit v2 (Illumina).cDNA normalization using a duplex-specific nuclease was also performed.The library was sequenced (100 bp paired-end reads) using the Illumina HiSeq 2000 platform.Library preparation, cDNA normalization, and sequencing were outsourced to Hokkaido System Science Corporation, Japan.Low quality bases (Phred quality value, QV$20) were trimmed from the raw data and read pairs of at least 80 bp were retained using SolexaQA [31].Possible PCR duplicates that originated during library preparation were removed with Con-DeTri [32].Contamination with TruSeq adapter sequences was removed by ea-utils (http://code.google.com/p/ea-utils).Subsequently high-quality paired end reads were assembled with Velvet/Oases software [33,34] using different K-mer sizes [35,45,55] and merged by K-mer 27 using Oases [34].For each word size, the longest isoform of each locus was selected.In addition, in order to recover low-expression genes and to produce longer assemblies, different coverage cut-off values (5, 8, 10 and 12) were also applied.Contigs originating from different coverage cutoff values were concatenated and redundant sequences and transcript-variants were removed by CDHIT-EST [35] using default parameters.Contigs over 200 bp length were retained.Additionally, contigs containing translated Open Reading Frames (ORFs) in which 95% of the amino acids were identical were considered duplicates.In such cases, one of the duplicates was removed using CDHIT [35].

Separation of host or symbiont-originated sequences
To distinguish whether contig sequences originated from host coral or symbiont Symbiodinium, nucleotide sequences of the assembled contigs were aligned to the genome sequences of both coral A. digitifera [10] and dinoflagellate S. minutum [17] by BLASTN, and contigs that show nucleotide similarities with A. digitifera or S. minutum were identified.Contigs aligned with the both genomes were not annotated in this study.Contigs that aligned only to the A. digitifera genome were annotated as ''Porites contigs'' while those that aligned only to the S. minutum genome were annotated as ''Symbiodinium contigs'', respectively.Different evalue cutoffs (1e 21 , 1e 22 , 1e 23 , 1e 24 , 1e 25 and 1e 210 ) were examined and a cutoff that maximized the number of Porites or Symbiodinium contigs was adopted (Table S3).We also applied the same analyses for reported coral transcriptome data from A. millepora [12], A. palmata [13], and P. damicornis [14], in order to investigate proportions of Symbiodinium sequences in each dataset.

Annotation of transcriptome assembly
Assembled transcriptome data were annotated as follows: 1) by BLAST homology searches against public protein databases: NCBI non-redundant protein sequences (NR) and Swiss-Prot [36], 2) by assignment of Gene Ontology (GO) terms [37], 3) by mapping to pathways using the KEGG annotation service KAAS [38], and 4) conserved protein domain searches with Pfam [39].BLASTX homology searches were conducted against the Swiss-Prot and NCBI NR protein databases and an e-value cutoff of 1e 25 was applied.For GO annotation, GO IDs were assigned based on the UniProt IDs of the best matches in the Swiss-Prot database [36].Then Generic GO slim terms were assigned using map2slim.plin go-perl (http://search.cpan.org/,cmungall/goperl/).For KAAS pathway annotation and analysis, we used the bi-directional best-hit (BBH) method to query the set of organisms representative for 'eukaryotes' as suggested on the KAAS website, using default settings.To screen and identify conserved protein domains, we used the Pfam database (Pfam-A.hmm, release 24.0; http://pfam.sanger.ac.uk) [39].Translated amino acid entries matching conserved domains were identified using HMMER searches (hmmer3) [40].In order to avoid eliminating Symbiodiniumor coral-specific rapidly evolving domains, we used an evalue cutoff of 1e 23 , as proposed by Kawashima et al. [41].For assessing completeness of the Porites contigs, mutual best-hit blast analyses (TBLAST, BLASTX, 1e 25 ) were performed against the sea anemone Nematostella vectensis proteome [8] in order to identify possible orthologs between Porites and Nematostella.We used Nematostella as a reference proteome in order to compare number of orthologous pairs between A. digitifera [10] and Nematostella with those of Porites and Nematostella.Alignment coverage of each Porites contig across full-length amino acid sequence of its possible Nematostella ortholog were investigated based on the BLAST search.For Symbiodinium contigs, S. minutum proteome dataset [17] was used for identifying orthologous pairs.

Estimation of polymorphism in Porites
For estimating the polymorphism rate of P. australiensis, highquality, trimmed Illumina reads (see above) were re-mapped against possible host-originated contigs (Porites contigs) using the Burrows-Wheeler Aligner (BWA) [42] and SNPs (single nucleotide polymorphisms) in each contigs were detected.Because BWA allows gaps in alignments, we set the maximum indel size to 5 bp (bwa aln -n 0.05; since the maximum read length is 100 bp) in order to detect small indels.SNPs and small indels were called using SAMtools software packages [43].To ensure reliable and high quality variant calling, SNPs and small indels were called only for positions with a minimal mapping quality (-Q) of 25, a coverage value (-d) of 10, and a maximum read depth (-D) of 200 using the varFilter command in the SAMtools package.

Identification of the symbiotic Symbiodinium type
Illumina sequence data that originated from the Symbiodinium internal transcribed spacer 2 (ITS-2) region of the nuclear ribosomal array were identified using DNA sequences of Symbiodinium-specific ITS-2 primers used for genotyping: itsD and ITS-2rev2 [44].Sequences reads containing these primer sequences were detected using BLAST (BLASTN) and 100% matched sequences were retrieved.Then these were subsequently assembled into a contig using Phrap [45] with default parameters.Sequence variations within the assembled ITS-2 sequence were checked with BWA [42] and SAMtools [43] as mentioned above.

Sequencing and assembly of the transcriptome
Total RNA was isolated from a single colony of P. australiensis (Figureo ˆ1A), and then a normalized cDNA library was prepared and sequenced with the Illumina HiSeq2000 system.We obtained 71 million paired-end sequences (14.3 Gbp, Table S1).Raw sequence data were submitted to the DDBJ Sequence Read Archive (DRA) under accession number DRA000906 (BioProject ID: PRJDB731).Subsequently, 45 million quality-trimmed read pairs (QV$20, both paired read $80 bp and PCR duplicates and TruSeq adapter sequences removed, approximately 9 Gbp, Table S1) were assembled using Velvet/Oases software (Figureo ˆ2, Tableo ˆ1).Contigs of ,200 bp were discarded because the RNA was fragmented into approximately 200-bp lengths during library preparation (see Material and Methods) and these short contigs were most likely to be truncated.After exclusion of redundant sequences and transcript variants, we obtained 74,997 contigs without ambiguous sequence gaps, totaling 94.73 Mbp (Figureo ˆ2, Tableo ˆ2).Although the size distribution was weighted toward smaller contigs, there were 31,199 contigs with lengths .1,000bp, 1,952 contigs .5,000bp, and 198 contigs .10,000bp, respectively (Figureo ˆ3A).The maximum contig length reached 54,796 bp; the average was 1,263 bp, and the N50 size was 2,037 bp (Tableo ˆ2), indicating that the assembly is of comparable or better quality than extant anthozoan transcriptome assemblies using NGS platforms (Table S2).Assembled sequences have been submitted to the DDBJ/EMBL-Bank/GenBank Transcriptome Shotgun Assembly (TSA) Database under accession numbers FX435232-FX505330 and FX799345-FX804242.

Dissecting the ''holobiont'' transcriptome
Because total RNA was isolated from an adult P. australiensis colony (Figureo ˆ1A), the transcriptome assembly contains genes from both corals and their zooxanthellae symbionts (Figureo ˆ1B).When the GC content distribution of all contigs in the assembly was analyzed, two clear peaks of approximately 40% and 53% were detected (Figureo ˆ3B), suggesting that these peaks possibly originated from the host (P.australiensis) and the symbiont (Symbiodinium sp.), respectively.The GC % of A. digitifera exons is about 40% [10] and that of a sea anemone, Nematostella, is similar [8], suggesting that a GC content of about 40% may be typical for anthozoan cnidarians.Therefore the 40% peak probably reflects P. australiensis genes.Taking into consideration the mean GC % of exons in the Symbiodinium minutum genome is about 50% [17], the secound peak might corresponds to the contigs from Symbiodinium.
Next we tried to identify contigs in the assembly that originated with Porites or Symbiodinium.We found that nucleotide alignment to both the Acropora and Symbiodinium genomes by BLASTN effectively separated them.In this study, we adopted an e-value cutoff 1e 24 (Table S3).30,446 sequences hit against the A. digitifera genome and 30,415 sequences against S. minutum genome (3,788 sequences sequences overlapped) respectively (Table S3).We subsequently removed overlapping sequences, and annotated 26,658 sequences as ''Porites contigs'' and 26,627 sequences as ''Symbiodinium contigs''.Annotated contigs comprised about 70% of all contigs and GC distributions of Porites and Symbiodinium contigs clearly matched the two peaks detected in the whole assembly (Figureo ˆ3B), indicating that we effectively separated  Porites and Symbiodinium genes in the assembly.These annotated contigs were used for further analyses (Tableo ˆ2, Figureo ˆ3, 4).
In most transcriptome analyses of anthozoans [12,13,15], aposymbiotic materials were used for sequencing in order to minimize Symbiodinium contamination.Thus, we suspect that Symbiodinium sequences in those transcriptome assemblies have been under-detected.It has also been reported that RNA of adult P. damicornis colonies containing Symbiodinium was extracted, sequenced on a 454 platform and assembled, but that few Symbiodinium sequences had been isolated [14].In contrast with previous studies, a large proportion of Symbiodinium sequences (at least 26,627 contigs, 35% of total assembled sequences) were in the transcriptome assembly (Figureo ˆ3C).We confirmed that coral transcriptome data reported to date contained very few Symbiodinium-originated sequences (Figureo ˆ3C).Therefore this study represents the first simultaneous transcriptome profile of both counterparts of coral holobiont.

Transcriptome annotation
For transcriptome annotation, BLAST homology searches against protein databases (NCBI NR and Swiss-Prot) were performed.Among 74,997 contigs, 33,935 (45.2%) have significant similarities (BLASTX, 1e 25 ) with NCBI NR and 25,893 (34.5%) with Swiss-Prot database entries, respectively.As mentioned above, 30,446 sequences showed similarity to the A. digitifera genome and 30,415 sequences to the S. minutum genome, respectively (Tableo ˆ2).Taken together, we were able to annotate 60,516 (80.7%) sequences using public protein databases (NCBI NR and Swiss-Prot) and A. digitifera [10] and S. minutum [17] genomic data.The remaining 14,481 sequences (total 7.12 Mbp) could not be annotated by BLAST and their average length (491.7 bp) was much shorter than that of the total assembly.These may be fragmented regions of other contigs, assembly artifacts, or bacterial contaminants.Some show similarities to coral microsatellite sequences deposited in the NCBI nucleotide (NT) database (data not shown).
Based on a BLAST search against the Swiss-Prot database, we were able to assign at least one GO term to the 11,048 Porites and 9,915 Symbiodinium contigs, respectively.These 20,963 contigs were grouped into 101 categories in the second hierarchy of the Generic GO slim.Figures S1, S2, S3 show GO annotation results for three GO sub-ontologies (cellular component, biological process, and molecular function).Since Porites corals are multicellular animals, cell death, neurological system process, cell proliferation, cell adhesion, and cell-cell signaling terms are more heavily represented than in Symbiodinium (26 more, Figure S1).DNA binding, enzyme binding, and transcription factor binding activity also seem to be higher in Porites (Figure S2).These may reflect the fact that 464 Porites contigs possess transcription factor domains (Table S4), whereas only 28 Symbiodinium contigs do (data not shown).On the other hand, transcripts of genes related to photosynthesis, transport, and generation of energy are more abundant among Symbiodinium contigs (Figure S1).

Estimating transcriptome completeness
About 44% of Porites contigs contain conserved protein domains and 4,528 unique Pfam domains were identified (Tableo ˆ2).In the gene models of the A. digitifera genome, unique 3,684 Pfam domains were detected [10], indicating that the Porites contigs Table 2. Summary of annotations of the transcriptome assembly.probably contain a comparable variety of protein domains.In addition, we performed KEGG annotation and 3,633 unique KO IDs were identified among the Porites contigs (Tableo ˆ2).We detected 3,684 unique KO IDs in the A. digitifera proteome using the KAAS BBH method.A comparable number of KO IDs might indicate that, although RNA-seq was performed using a single RNA library, the majority of P. australiensis genes were successfully recovered in the transcriptome assembly.However it could be possible that those numbers in Porites contigs are overestimated due to incomplete clustering of transcript variants in the assembly.An increasing number of studies have recently identified genes that encode transcription factors and signal transduction molecules in corals and have determined their roles in cnidarian evolution [46][47][48][49][50]. Therefore, for further assessing the completeness of the Porites transcriptome, the repertoire of transcription factors and signal transduction molecules was compared with those of other cnidarians whose genomes have been decoded (Table S4, 5).We detected comparable numbers of transcription factors and signaling molecules among the Porites contigs to what we previously found in the A. digitifera genome [10], e.g.Porites: 89 homeobox, 24 HMG box and 15 wnt genes; A. digitifera: 97 homeobox, 26 HMG box and 15 wnt genes (Table S4, 5).Next we identified orthologous gene pairs between Porites contigs and Nematostella by mutual-best-hit blast analysis; 7,825 pairs were detected.The similarities of the A. digitifera and Nematostella genomes resulted in 8,416 orthologous pairs, suggesting that Porites contigs represent the majority of the gene repertoire for Porites australiensis.6,324 orthologous genes out of 7,825 have more than 80% alignment coverage across Nematostella orthologs, suggesting that most of the Porites contigs recover full-length open reading frames.
On the other hand, 3,569 Pfam domains and 1,352 KO IDs were detected among Symbiodinium contigs (Tableo ˆ2), while 4,035 Pfam domains and 2,622 KO IDs were detected in the S. minutum gene model [17], respectively.We also identified 28 Symbiodinium contigs containing transcription factor domains (data not shown).It has been reported that in transcriptome assemblies from two cultured strains of Symbiodinium, 156 and 87 sequences, respectively, contained at least one known transcription factor domain [16].The S. minutum genome was found to contain 32 transcription factor genes [17].We detected 15,914 orthologous pairs between the Symbiodinium contigs and S. minutum genome.In S. minutum genome, 23,487 orthologous groups were identified by compared with 150 genomes [17].This might reflect that not all genes were expressed at the sampling point.

Polymorphism rate in Porites
Since total RNA was extracted from a single Porites colony, it is possible to estimate the polymorphism rate of P. australiensis within an individual.High quality Illumina reads were re-mapped against the reference Porites contigs using BWA software [42], and high quality SNPs and small indels (,5 bp) within the Porites colony were identified using SAMtools [43].Among the 44,569,031 nucleotides of the 26,658 Porites contigs, 425,728 SNPs and 11,788 small indels (total 20,213 bp) were detected.Taken together, the polymorphism rate between haplotypes, including small indels (less than 5 bp) and SNPs in the exons of Porites transcripts is estimated to be 1.0%, which is slightly higher than that observed in other cnidarian genomes (A.digitifera [0.4%, Shinzato and Mungpakdee, unpublished data], Nematostella [0.65%, 8], and Hydra [0.69%, 9].

Symbiont type
The genus Symbiodinium displays tremendous taxonomic diversity with nine divergent lineages.Clades A-I have been described in Symbiodinium based on nuclear ribosomal DNA (rDNA) and chloroplast 23S rDNA [51].Each clade contains multiple genetic types, often resolved using the internal transcribed spacer 2 (ITS-2) regions [44,52].We failed to detect assembled sequences of nuclear rDNA or chloroplast 23S rDNA in our dataset.Accordingly, the ITS2 region-containing sequence reads were identified and assembled, and then we managed to reconstruct an ITS-2 region (341 bp) of the symbiotic Symbiodinium.BLAST homology searches against the NCBI NT database revealed that this sequence shows high similarity (340/341 bp identity) with clade C Symbiodinium sp.C15 (GenBank: JN558044).We also checked the sequence variation within the ITS2 region by BWA mapping and SAMtools, but no variation was detected (data not shown), indicating that the C15 type, a type common in Porites species across the Pacific [53], is the major Symbiodinium type in the P. australiensis colony used in this study.

Amino acid biosynthesis pathways in Porites holobiont
Bacteria, plants, and many fungi are able to synthesize all of the 20 amino acids commonly found in proteins, whereas all animals studied to date either lack the ability to synthesize one or more of these amino acids, or else they are unable to synthesize quantities sufficient to meet their metabolic needs.These amino acids are termed ''essential'' and must be obtained from the diet.For vertebrates, eight or more amino acids are essential; threonine, valine, methionine, leucine, isoleucine, phenylalanine, lysine, and tryptophan are required by all vertebrates, while arginine and/or histidine are also essential in some cases [54].Although amino acid biosynthesis occurs via conserved pathways and is relatively wellcharacterized [55], amino acid exchange in cnidarian photoautotrophic symbioses is poorly understood.The issue of amino acid requirements is complicated by the presence of symbiotic algae [56,57].For example, examination of the amino acid biosynthetic capacity of the sea anemone Aiptasia pulchella -a facultative host of Symbiodinium sp.-suggested that seven amino acids (histidine, isoleucine, leucine, lysine, phenylalanine, tyrosine, and valine) are synthesized by the symbiotic algae and translocated to the sea anemone, and that methionine and threonine are likely to be synthesized by A. pulchella itself [57].A genome-wide survey of amino acid biosynthetic pathway components in A. digitifera revealed that Acropora corals may be able to synthesize ten nonessential amino acids, but not cysteine [10], however amino acid metabolism in coral holobionts is still unclear (Figureo ˆ4A).We reconstructed the amino acid biosynthetic pathways in the Porites holobiont based on KEGG IDs (Figureo ˆ4B, C).In this study we took amino acid biosynthetic pathways missing one or two enzymes into consideration.Most of enzymes involved in amino acid biosynthetic pathways were detected in the transcriptomes of Porites and Symbiodinium, while four enzymes cannot be detected in the lysine anabolic pathway (Figureo ˆ3C).
Interestingly many enzymes involved in essential amino acid biosynthesis are detected only in Symbiodinium contigs (Figureo ˆ3B, Figureo ˆ3C left side).Overrepresentation of ''transport'' GO terms in Symbiodinium contigs (Figure S1) implies that Symbiodinium transport essential amino acids to host cells.In contrast, enzymes for non-essential amino acid pathways were detected in both Symbiodinium and Porites (Figureo ˆ3B, Figureo ˆ3C, right side).Glutamic acid, glutamine, aspartic acid, and alanine are probably produced in both Porites and Symbiodinium.According our prediction, asparagine is synthesized in Symbiodinium while cysteine is synthesized in Porites (Figureo ˆ3).We have shown previously that Acropora corals lack an essential enzyme for cysteine biosynthesis, cystatione ß-synthase [10], however, Porites seems to posses it (Figureo ˆ3C), indicating that it does not depend upon Symbiodinium for cysteine biosynthesis and might account for its greater resilience to environmental stresses.An interesting example is the methionine biosynthesis pathway.Half of the enzymes reside in the Porites and Symbiodinium contigs, respectively, suggesting that methionine might be produced by intimate cooperation between host and symbiont in coral holobionts.While transcriptome sequencing was based upon a single RNA-seq library in this study, we were able to recover a surprisingly large gene repertoire for amino acid biosynthesis.Our data appear to provide the first direct molecular evidence of complementarity and syntrophy between coral hosts and their symbionts in amino acid metabolism.

Conclusions
We sequenced a Porites australiensis holobiont and demonstrated that decoded genomic data of an Acropora coral [10] and a Symbiodinium minutum [17] greatly assisted the characterization of assembled contigs from a mixture of RNAs from different organisms.In this study we were able distinguish about 70% of contigs as host or symbiont.The assembled sequences contain a wide variety of genetic information from both the coral and its symbiont, Symbiodinium, including genes for most enzymes in all amino acid biosynthetic pathways.When genome sequences of P. australiensis and its symbiotic Symbiodinium are available, the annotation could be further improved.The Porites holobiont transcriptome dataset allows us to utilize Porites australiensis, an abundant coral species from the Indo-Pacific, to reveal molecular mechanisms of coral symbiosis and coral stress responses.Our approach provides an opportunity to simultaneously analyze coralsymbiont interactions on transcriptomic level.Furthermore, analyzing the molecular bases of Porites calcification may enable researchers to improve the accuracy of future climate change prediction.Table S3 Results of nucleotide sequence alignment of the assembled sequences with different e-value settings of BLASTN against Acropora digitifera and Symbiodinium minutum genome sequences.We selected e-value cutoff as 1e 24 in this study as the number of Porites contigs is the largest.(PDF)

Figure 3 .
Figure 3. Analyses of the Porites australiensis ''holobiont'' transcriptome assembly.(A) Length distribution of transcripts in the transcriptome assembly.Yellow boxes indicate all contigs.Blue and brown lines indicate Porites contigs and Symbiodinium contigs, respectively.(B) Distribution of GC percentages of the assembled contigs.Black line: all contigs, blue dot line: Porites contigs, brown dot line: Symbiodinium contigs.(C) Comparison of proportions of Symbiodinium sequences between the scleractinian coral transcriptome assemblies.Each sequence was aligned to A. digitifera and S. minutum genome sequences by BLASTN (1e 25 ).Note that the high percentage of A. digitifera sequences shared with the A. millepora and A. palmata datasets occur because these three corals are congeneric.doi:10.1371/journal.pone.0085182.g003

Figure 4 .
Figure 4. Amino acid metabolism in Porites holobiont.(A) Schematic drawing of amino acid exchange between a coral polyp and symbiotic Symbiodinium.(B) Summary of amino acid biosynthesic capabilities of Porites (host) and Symbiodinium (symbiont) inferred from KEGG annotation.Numbers indicate how many amino acids can be produced by Porites or Symbiodinium, and numbers in blankets show numbers of essential amino acids.Blue circle indicates Porites and brown circle indicates Symbiodinium, respectively.(C) Detail annotations of amino acid biosynthetic pathways in the Porites holobiont.Arrows indicate enzymes involved in each amino acid biosynthetic pathway.Enzymes that are not annotated as Porites or Symbiodinium are shown with a black line, and those not detected in the transcriptome dataset are shown with dotted lines.Blue indicates enzymes identified in Porites contigs.Brown indicates enzymes detected in Symbiodinium contigs.Brown arrows boxed in blue lines indicate enzymes detected in both Porites and Symbiodinium.Amino acids that can be produced by Symbiodinium are boxed in brown and those that can be produced by Porites are boxed in blue, respectively.Amino acids boxed in dotted lines indicate that biosynthetic capabilities are not clear.doi:10.1371/journal.pone.0085182.g004

Figure
Figure S1 Analysis of GO term enrichment of the ''biological process'' category for Porites and Symbiodinium contigs.GO terms containing at least 100 sequences are shown.The Y-axis represents proportions of contigs in each category of GO-assigned contigs (Porites: 9806, Symbiodinium: 9147).Blue bars indicate Porites contigs.Red bars indicate Symbiodinium contigs.(PDF)

Table S4
Comparison of the number of genes with transcription factor-related domains of Porites australiensis, Acropora digitifera, Nematostella vectensis and Hydra magnipapillata.(PDF) Table S5 Comparison of the number of genes with signaling molecule-related domains of Porites australiensis, Acropora digitifera, Nematostella vectensis and Hydra magnipapillata.(PDF) File S1 List of Porites contigs (tab-separated text; contig names and NCBI accessions).File S2 List of Symbiodinium contigs (tab-separated text; contig names and NCBI accessions).File S3 Annotation files for all contigs (tab-separated text; contig name, NCBI accession, Swiss-Prot blast result, Pfam ID, KEGG ID).(ZIP)