Main

Tropical reef-building corals form a multipartite symbiosis with the photosynthetic dinoflagellate Symbiodinium and a diverse microbial community that includes bacteria, archaea, fungi, protists and viruses; collectively termed the coral holobiont (Rohwer et al., 2002). The Symbiodinium–coral symbiosis can be disrupted by heat stress, which results in the loss of Symbiodinium cells from coral tissues, i.e., coral bleaching (Hoegh-Guldberg, 1999). Heat stress has been alleged to promote lytic viral infections of Symbiodinium, as virus-like particles have been found in heat-stressed Symbiodinium from the temperate sea anemone Anemonia viridis (Wilson et al., 2001) and from the corals Pavona danai, Acropora formosa and Stylophora pistillata (Wilson et al., 2005; Davy et al., 2006). Additionally, following heat shock at 31 °C, the expression of a protein with homology to a eukaryotic viral protein increased >100-fold in a Symbiodinium-enriched fraction of Stylophora pistillata tissue (Weston et al., 2012).

Using transcriptome data generated by Levin et al. (2016) for two Symbiodinium type C1 populations cultured at 27 °C and 32 °C (n=4), we explore the effect of heat stress on viruses associated with Symbiodinium and Symbiodinium anti-viral responses. The Symbiodinium populations were originally isolated from the coral Acropora tenuis at South Molle Island (SM) and Magnetic Island (MI) (Great Barrier Reef, Australia). The thermosensitive SM population was found to suffer physiological damage in culture and to bleach in hospite at 32 °C, whereas the thermotolerant MI population was unaffected (Howells et al., 2012; Levin et al., 2016).

As many viral RNAs are polyadenylated (Wilson et al., 2000; Priet et al., 2015), they were retained in the poly(A)+ purified Symbiodinium RNA samples used for RNA-Seq (Levin et al., 2016). Viral transcripts in each de novo transcriptome were identified through a robust BLASTx bit score approach adapted from Boschetti et al. (2012), which calculates the difference between the highest viral and the highest non-viral bit score to determine if a transcript is from a virus or the host, followed by GC content, trinucleotide frequency and codon usage analyses (Supplementary Materials and Methods, Supplementary Figure 1 and Supplementary Datasets 1 and 2). Symbiodinium anti-viral transcripts were identified by searching the transcriptomes for transcripts encoding anti-viral gene types found in corals and for transcripts with anti-viral Gene Ontology (Supplementary Materials and Methods).

The thermosensitive SM and thermotolerant MI transcriptomes contain 306 and 238 viral transcripts (Supplementary Discussion), as well as 62 and 65 anti-viral transcripts, respectively (NCBI GEO accession: GSE77911). In both transcriptomes, viral transcripts show homology to genes from NCLDVs Mimiviridae and Phycodnaviridae and the major capsid protein (MCP) gene from the dinoflagellate-specific +ssRNAV Alvernaviridae (Dinornavirus), which is in agreement with previous sequencing and transmission electron microscopy findings (Wilson et al., 2005; Correa et al., 2013, 2016). The +ssRNAV transcripts in the thermosensitive SM transcriptome share 94% nucleotide (nt) identity (TR74740|c13_g1_i1, 5202 nt; TR74740|c13_g1_i2, 2154 nt). The +ssRNAV transcript in the thermotolerant MI transcriptome is much shorter (TR97578|c0_g1_i1, 475 nt) but shares 100% nt identity with TR74740|c13_g1_i1. Successful PCR amplification of the MCP genes from complementary DNA reverse-transcribed from RNA—but not from genomic DNA—of both Symbiodinium populations supports that they are from +ssRNAVs (Supplementary Materials and Methods and Supplementary Figure 2). However, phylogenetic analysis of the translated MCP gene sequences revealed that they are highly divergent from the Dinornavirus MCP gene and previously identified partial-length Dinornavirus-like MCP genes (Supplementary Discussion, Supplementary Table 1 and Supplementary Figure 3).

Both +ssRNAV transcripts in the thermosensitive SM transcriptome contain a putative viral internal ribosomal entry site (Supplementary Figure 4), which is related to the internal ribosomal entry site of the +ssRNA cricket paralysis virus, directly upstream from the MCP gene. The longer +ssRNAV transcript, TR74740|c13_g1_i1, also encodes an unannotated open reading frame (ORF) determined to be a +ssRNAV RNA replicase (RNA-dependent RNA-polymerase) polyprotein based on protein structure modeling (Supplementary Figure 5). The RNA replicase gene precedes the internal ribosomal entry site and MCP gene, giving the full transcript a markedly similar arrangement to the Dinornavirus and cricket paralysis virus complete genomes (Wilson et al., 2000; Nagasaki et al., 2005). Thus, we conclude TR74740|c13_g1_i1 to be the RNA genome of a novel +ssRNAV, making this the first discovered genome of any virus infecting Symbiodinium (Figure 1a). Surprisingly, the conserved dinoflagellate spliced leader, which is present on >95% of Symbiodinium mRNAs (Zhang et al., 2013), is at the 5′ end of TR74740|c13_g1_i1. Although, if the +ssRNAV is dinoflagellate-specific like Dinornavirus, incorporation of the dinoflagellate spliced leader in the viral RNA genome is likely a case of molecular mimicry, a well-documented viral strategy to evade host immune responses that detect foreign nucleic acids (Elde and Malik, 2009), or for efficient cap-dependent translation by host polysomes (Zeiner et al., 2003).

Figure 1
figure 1

Genome of the novel +ssRNAV and its expression in Symbiodinium transcriptomes. (a) Genome model of the +ssRNAV infecting Symbiodinium (TR74740|c13_g1_i1, GenBank accession: KX538960) and its similarities to the RNA genomes of the dinoflagellate-specific Dinornavirus (Heterocapsa circularisquama RNA virus strain 34, NCBI accession: AB218608) and cricket paralysis virus (NCBI accession: AF218039). (b) TMM-normalized FPKM at 27 °C (averaged across replicates for each transcript on day −1, n=8) for the +ssRNAV transcripts (TR74740|c13_g1_i1, 5757 FPKM; TR74740|c13_g1_i2, 1996 FPKM; TR97578|c0_g1_i1, 1 FPKM), NCLDV transcripts and non-viral transcripts in the thermosensitive SM and thermotolerant MI transcriptomes. Black lines mark the mean FPKM+s.d. for each subset of transcripts.

Full size image

Viral transcripts had similar expression levels to many non-viral transcripts at 27 °C, with the exception of the two +ssRNAV transcripts in the thermosensitive SM transcriptome (Figure 1b). Both +ssRNAV transcripts in the thermosensitive SM transcriptome maintained average expression levels >1300 fragments per kb of transcript per million mapped reads (FPKM) on all sampling time points at 27 °C, whereas the +ssRNAV transcript in the thermotolerant MI transcriptome had an average expression level <2 FPKM on all sampling time points. The vastly dissimilar expression levels of +ssRNAV transcripts between the transcriptomes suggest that the thermosensitive SM Symbiodinium population was experiencing a severe viral infection.

Differential expression analysis (Supplementary Materials and Methods) confirmed that no viral or anti-viral transcripts were differentially expressed between experimental groups of either population on day −1 (preheating; all samples acclimated at 27 °C) (Figures 2a–e). In both transcriptomes, upregulation of NCLDV transcripts was induced at 32 °C and increased from day 9 to day 13 (Figures 2b and d, and Supplementary Tables 2–9). However, the upregulated NCLDV transcripts in the thermosensitive SM transcriptome encoded for a greater diversity of genes (F-box and FNIP repeat-containing proteins, resolvase, transposase, ankyrin repeat protein) compared to those in the thermotolerant MI transcriptome (only F-box and FNIP repeat-containing proteins). Viral F-box and FNIP repeat-containing proteins and ankyrin repeat proteins have possible roles in degrading proteins through protein–protein interactions, countering host defences and exploiting the host's ubiquitin-proteasome system to create an appropriate cellular environment for viral replication (Suhre, 2005; Sonnberg et al., 2008; Fischer et al., 2010), whereas resolvases and transposases directly participate in DNA manipulation (Iyer et al., 2001; Schroeder et al., 2009).

Figure 2
figure 2

Viral infections and anti-viral responses of Symbiodinium under heat stress. Differential expression of (a) thermosensitive SM transcriptome +ssRNAV transcripts, (b) thermosensitive SM transcriptome NCLDV transcripts, (c) thermosensitive SM transcriptome anti-viral transcripts, (d) thermotolerant MI transcriptome NCLDV transcripts and (e) thermotolerant MI transcriptome anti-viral transcripts at 32 °C. Only transcripts with ⩾2-fold expression between 27 °C and 32 °C treatments on at least one sampling time point were analyzed. Transcripts were considered to have no differential expression on a sampling time point where the false discovery rate (FDR) was >0.001. Upregulated transcripts at 32 °C are shown in yellow. Downregulated transcripts at 32 °C are shown in purple. The gray regions represent the preheating sampling time point on day −1 when all replicates (n=8) were still at 27 °C. On day 0, replicates were ramped to 32 °C (n=4) or maintained at 27 °C (n=4) for the duration of the study (Supplementary Materials and Methods). Differential expression and transcript annotation results are detailed in Supplementary Tables 2–9.

Full size image

In the thermosensitive SM transcriptome, heat stress resulted in downregulation of the highly expressed +ssRNAV transcripts and upregulation of anti-viral transcripts (Figures 2a and c and Supplementary Tables 2–3 and 6–7). Conversely, the thermotolerant MI transcriptome showed no differential expression of its lowly expressed +ssRNAV transcript and downregulation of anti-viral transcripts at 32 °C (Figure 2e, Supplementary Tables 8–9). Symbiodinium anti-viral responses may therefore become activated from increased DNA manipulation by NCLDVs or initial upregulation of highly expressed +ssRNAV transcripts at 32 °C that was mitigated by the sampling time points on days 9 and 13.

Our study exemplifies how RNA-Seq data can be used to gain valuable insight into resident viruses. Our results indicate that only the thermosensitive SM Symbiodinium population experienced an extreme +ssRNAV infection and thermally induced NCLDV DNA manipulation. Thus, viral infections may factor into Symbiodinium thermal sensitivity, and consequently, coral bleaching.