Metagenomic analysis indicates that stressors induce production of herpes-like viruses in the coral Porites compressa

Results and Discussion

Experimental Design and Rationale.

To characterize coral-associated viruses, specimens of the finger coral, Porites compressa, were collected in the Hawai'i Marine Laboratory Refuge. This species of coral was chosen because it is a common and endemic species of coral in the Hawaiian archipelago and is the dominant species of coral across reefs in Kane'ohe Bay where Coconut Island and the Hawai'i Institute of Marine Biology are located. Numerous studies have been conducted on this coral and there is a legacy of information on the oceanography and ecology of these reefs, providing pertinent data for insights into the viral ecology of corals across this region.

Coral specimens were exposed to thermal (5° C above ambient), dissolved organic carbon (25 mg·mL⁻¹ glucose), nutrient (10 μM nitrate, nitrite, phosphate, and ammonium), and pH (1.7 units below ambient) stress for 64 hours [supporting information (SI) Fig. S1]. In brief, 3 corals were broken into fingers and placed in aquaria containing a given treatment. Coral samples or “nubbins” were collected just before the experiment (“time zero”) to measure shifts in the viral consortia that resulted from isolation from the reef, and to provide several unadulterated samples for comparison. Subsequent samples were taken after 1, 4, 16, and 64 h of stressor exposure. Six viromes were generated by pooling 1 μg of DNA from each time point: time zero, “aquarium control,” “temperature,” “DOC,” “pH,” and “nutrient” (Fig. S1). Pyrosequencing was conducted at 454 Life Sciences on a GS20. Sequence reads were parsed and analyzed using bioinformatics.

Library Statistics and Bioinformatic Analysis.

The sequenced viromes contained between 30,000 and 51,000 reads with an average length of 105 bases, a dataset of >25 million bases. Several bioinformatic techniques identified changes in the eukaryotic viral consortia across each library. Sequence reads were first compared using BLASTn to the nonredundant database at the National Center for Biotechnology Information (Table S1). A sequence was assigned as “known” if it had a similarity (e ≤ 10⁻⁴) to the nonredundant database. These low stringency parameters were chosen because viral DNA often has little similarity to known sequences in the non-redundant database (35). “Knowns” were then taxonomically sorted (viral, bacterial, archaeal, and eukaryotic) based on their highest similarity. In agreement with previous findings, these metagenomes contained few similarities to known viral sequences (Table S1) (36). A majority of sequences were found to be similar to bacterial and eukaryotic annotations. Analyses of 16S and 18S rDNA PCR and similarities to the human genome (see Methods) demonstrated that eukaryotic and microbial DNA contamination did not contribute to these sequences. This high percentage of known sequences similar to Bacteria and Eukaryota is likely the result of (i) protein conservation across viral, microbial, and eukaryotic genomes; (ii) the larger size and larger number of microbial genomes at the National Center for Biotechnology Information database; (iii) prophage and retrotransposon-like sequences within microbial and eukaryotic genomes; and (iv) the genetic mosaicism of viruses. The entire library was also compared with 4 other marine viromes (Table S2) to determine whether the coral viromes resembled seawater viruses. Few similarities were found between the coral viromes and water samples, suggesting that seawater contamination was not an issue.

Coral-Associated Viral Consortia Change with Experimental Manipulation.

One goal of this study was to determine the types of eukaryotic viruses infecting the coral holobiont. Phage data will be presented elsewhere. As few sequences (Table S1) were found to be similar to any viral sequences at the National Center for Biotechnology Information, reanalysis of the metagenomic reads was conducted using tBLASTx (e ≤10⁻⁶) to a boutique database of the fully sequenced eukaryotic viral genomes from the National Center for Biotechnology Information database.

Metagenomic sequences with significant similarities (e ≤10⁻⁶) were assigned as “knowns” based on the sequence's best similarity (i.e., the top hit) to each reference viral genome. To compare and contrast the overall differences between each library, sequences were first sorted at the family level, and the percentage of known similarities to each viral family was calculated.

Of the ≈100 viral families described thus far (37), 19 were found in the combined coral-associated viromes (Fig. 1A). For example, the time zero samples contained 23% Baculoviridae-like and 30% Polydnaviridae-like sequences. Baculoviridae-like sequences also accounted for 23% of the similarities in the pH (blue diamonds) sample. The aquarium control (filled circles) contained 30% and the carbon-loaded DOC samples (orange hexagons) 18% sequences similar to the Nanoviridae. A large proportion of the best sequence similarities in the stressed samples were related to the Herpesviridae family, with the nutrient, pH, and temperature-perturbed samples containing 64%, 53%, and 70% herpes-like sequences, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

A) Percentages of best similarities (tBLASTx, e ≤10⁻⁶) to 19 families of viruses in each individual coral-associated viral metagenome: Time Zero (open circles), Aquarium Control (filled circles), Temperature (red triangles), pH (blue diamonds), Nutrient (green arrows), and DOC (orange hexagons). (B) Predicated host ranges across each metagenome, with libraries containing vertebrate (light gray) and invertebrate (black bars) associated viral sequences and plant (white bars) and algal (dark-gray bars) viruses. Blue bars represent similarity to viral genomes that are associated with Protists.

Coral-Associated Viruses Infect a Range of Hosts.

A discriminating characteristic in viral taxonomy is host range (38). Corals are a collective of cnidarian tissue, microbial symbionts, fungi, and other microeukaryotes. To determine the potential hosts of the coral-associated viruses, sequences were categorized according to a predicted generalized host range defined by the known host of the similar reference genome. These host ranges were modified from the Universal Virus Database from the International Committee on the Taxonomy of Viruses (37). Using these eukaryotic viral assignments, the predicted host range of the coral-associated viruses changed with each stressor treatment (Fig. 1B). In each of the viromes, >38% of the similarities were associated with vertebrate hosts, including representatives of most of the herpesviruses, poxviruses, adenoviruses, papillomaviruses, circoviruses, and iridoviruses (light gray bars). The Aquarium Control and DOC samples contained a large fraction of sequences similar to ssDNA viruses from plants (white bars) such as geminiviruses and nanoviruses (39). Invertebrate infecting viruses (black bars) were also abundant (>60%) within the time zero library and were mostly similar to the baculoviruses and polydnaviruses, 2 viral families that predominantly infect arthropods (40–42).

However, a caveat to this analysis is that the predicted host range is biased by the larger number of completed vertebrate-associated viral genomes. This analysis is also based on the sequences' best similarity, but almost certainly represents distant relatives of the reference viruses. These data suggest that coral-associated viruses possibly have many target hosts, including metazoans, protists, fungi, and plants.

Stressors Increase the Abundance of Herpes-Like Viral Sequences.

Each virome sequence may have multiple similarities due to gene conservation. Therefore, each library was also analyzed for the total number of sequence similarities (as opposed to the best similarity in Fig. 1) to each of the eukaryotic viral genomes at the National Center for Biotechnology Information database, using tBLASTx (e ≤10⁻⁶). Individual viromes had between 300 and 11,000 predicted protein level sequence similarities to all eukaryotic viral genomes (Table 1). However, the number of these sequences that had homology to the Herpesviridae was different between the stressors. The percentage of all of the viral similarities to the Herpesviridae was similar in the time zero (≈19%) and the control (≈19%) but was elevated to 30%, 32%, and 32%, in the temperature, nutrient, and pH samples, respectively. The DOC sample had few sequences (< 2%) related to Herpesviridae.

The large abundance of Herpesviridae similarities in the temperature, pH, and nutrient samples indicated that these stressors induced production of viruses with distant relation to the herpesviruses. A caveat to this analysis, however, is that the short read length combined with low sequence homology limits taxonomic resolution (43). To verify if herpes-like viruses were present in the samples, sequences from all of the libraries were combined, compared using tBLASTx (e ≤10⁻⁶) to the eukaryotic viral boutique database, and then plotted along the corresponding reference genomes. The combination library contained both broad and deep coverage to some herpesvirus (44–48)genomes (Fig. 2 A–C and Table S3), but low coverage to other marine and cnidarian-associated viral genomes (Fig. 2 E–H and Table S3), such as a panaeid shrimp virus, a zoochlorella infecting phycodnavirus (2D), a phytoplankton virus (2E), and a fish iridovirus (2F) genome (49). From these analyses, we hypothesized that herpes-like viruses (HLVs) are present in corals and become more abundant as a result of stress events.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Coverage of selected Herpesviridae genomes and other marine associated viral genomes. Libraries were combined and sequences analyzed for similarity (tBLASTx e ≤10⁻⁶) to a boutique database of 2,020 fully sequenced eukaryotic viral genomes collected from the National Center for Biotechnology Information. Genomes were separated into 2,500 base pair bins. (A–C) Broad and high fold coverage was found for many mammalian associated Herpesviridae genomes, such as a suid herpesvirus (A) and 2 cercopithecine/macaque herpesviruses (B and C). (D–F) Low and uneven coverage was found for the symbiont associated chlorella phycodnavirus (D), a coccolithophorid virus (E), and a fish iridovirus (F).

Real-Time PCR on Original Temporal Samples Confirms the Production of HLVs.

Viral production, or the generation of new virions by the host, has been effectively quantified using real-time PCR to genes of various RNA and DNA viruses (50). To determine temporal changes in viral abundance across the manipulation experiments, metagenomic sequences were assembled, annotated, and given a taxonomic classification according to the sequence's best protein similarity using PSI-BLAST to the non-redundant database (see SI Methods for details). Sequences with significant homology (e ≤10⁻⁶) to herpesviruses were chosen, and primers were generated for both conventional and real-time PCR. One sequence was 62% identical (e <4 × 10⁻¹¹) to a single copy thymidylate synthase gene from Saimiriine herpesvirus 2 (Fig. S2). This sequence was amplified, and the resulting PCR product cloned and Sanger sequenced to verify the assembly (GenBank accession no. EU660955). No differences were found between the original assembly and the cloned nucleotide sequence (data not shown).

These primers were used for real-time PCR, and it was found that the copy number of this putative herpes-like viral thymidylate synthase sequence increased during the original stressor experiment (Fig. 3A). The target sequence was below the detection level in all of the zero hour time points with the exception of the Nutrient (green triangles). The target sequence was also elevated in all of the 1-, 4-, 16-, and 64-h treated corals. The abundance of this target sequence typically peaked at 4 h after treatment exposure, and the maximum increase was ≈6 orders of magnitude. For example, the pH (blue diamonds) and Temperature (red triangles) stressors increased the abundance of this viral sequence from below the detection limit (≈10 copies) to >10⁵ and 10⁷ copies in only 4 h. Viral abundance was then reduced to 10³ for the remainder of those experiments.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Herpes-like viral sequence abundance across two 64-h experiments. (A and B) Real-time PCR on a putative herpes-like virus thymidylate synthetase gene was conducted on all temporal samples collected during the original metagenomic experiment (A) and samples from a second experiment conducted 1 year later (B). In the original experiment, coral specimens were exposed to thermal (5° C above ambient), dissolved organic carbon (25 mg·mL⁻¹ glucose), nutrient (10 μM nitrate, nitrite, phosphate, and ammonium), and pH (1.7 units below ambient) stress for 64 hours. (B) In the second 2007 experiment, corals were acclimated to the tanks for 2 weeks before treatment with stressor, thermal stress was an elevation of seawater temperature 3°C, acidity stress was a 0.2 unit reduction of ambient pH, and samples were collected at 0, 4, and 64 h.

In the Aquarium Control sample (Fig. 3, black circles) large numbers (≈10⁷) of this HLV specific target gene were detected after 16 h and demonstrated that even in the absence of applied exogenous abiotic stress, physical manipulation of the corals led to viral production. The viral sequence was detected in only 1 of the 5 DOC (orange hexagons) time points, corroborating the low abundance of sequences in the DOC metagenome. Both this low copy number in the DOC samples and the large copy number in the nutrient zero time point indicate that external herpesvirus contamination was not a confounding factor.

Second Experiment Conducted One Year Later Confirms that HLVs Are Commonly Associated with Stressed Corals.

The possibility that the viruses discovered in the corals were unique to the original metagenomic experiment and/or the result of unnatural stress conditions was tested by conducting a second experiment 1 year later. This second repeat experiment used the same sampling scheme (Fig. S1) with minor modifications to reflect more likely environmental conditions: (i) corals were acclimated to the tanks for 2 weeks before treatment with stressor, (ii) thermal stress was an elevation of seawater temperature 3°C, (iii) acidity stress was a 0.2 unit reduction of ambient pH, and 4) nubbins were only collected at 0, 4, and 64 h. Identical primer sets were used to amplify the HLV thymidylate synthase sequence.

These analyses showed an increase in the herpes-like virus target sequence when stressors were applied to the corals (Fig. 3B). The target sequence was not detected at the zero hour time point in any of the tissue samples, but was found in the temperature (red squares), pH (blue diamonds), and nutrient (green triangles) samples at 4 and 64 h. The DOC exposed and aquarium control corals contained no detectable target sequence until 64 h.

Together, these results show that, although the temporal changes in HLV abundance were analogous in both experiments, the magnitude of this change was different. Compared with the 2006 metagenomic experiment, the total abundance of HLVs was lower in the second 2007 experiment. This may have resulted from the more environmentally relevant levels of stressors applied in the repeat experiment. However, environmental conditions were also milder during the second 2007 experiment with no extended rainy or windy periods before or during the experiment. In contrast, during the original 2006 experiment, precipitation along with significant runoff occurred. As previously mentioned, the abundance of the HLVs was found to be high in the original experiment's nutrient zero hour sample. This indicates that production of virus had already begun on the reef, perhaps as a result of degraded environmental conditions.

A two-way ANOVA conducted on the real-time data indicated significant effects of both treatment (F = 9.80, P < 0.01) and time (F = 4.82, P < 0.01), and a significant interaction effect (F = 4.54, P < 0.01). The combination of the two main effects and the interaction term explains >70% of the variance. These results indicated that the different treatments applied to corals stimulated herpes-like viral production and that the level of production varied significantly over the time course of the experiment. A significant interaction term suggested that the effects of both treatment and time on viral production were amplified multiplicatively. This implies that just studying each treatment individually over time or all treatments simultaneously over 1 time point would not have shown the same differences in production.

Additional Evidence for Herpes-Like Viruses in Corals and All Cnidarians.

In addition to our datasets, transmission electron micrographs (TEM) and sequences from the Nematostella genome and hydra EST project provide further evidence that Cnidarians, not just corals, are hosts for herpes-like viruses. Corals contain large enveloped icosahedral virions morphologically similar to the Herpesviridae (28, 29). Herpesvirus essential genes are also present in Cnidarian genomes (28, 51, 52). Within the Nematostella draft genome, alkaline exonucleases, herpes major outer envelope proteins, herpes DNA primases, herpes ubiquitin-specific proteases as well a herpes surface glycoprotein and transcription activation factor were identified (Table S4). Also, several copies of a latent membrane protein were found suggesting the presence of endogenous or latent herpes-like viruses in Nematostella. Additionally, expressed sequence tags from Hydra magnipapillata contain herpes specific tegument proteins and herpes-associated proteases, suggesting that herpes-like viral transcripts were produced.

Herpes-like viral sequences were also recently identified in metagenomic libraries from P. astreoides and D. strigosa corals. In the P. astreoides metagenome, most of the sequences with similarity to eukaryotic viruses were related to fish iridoviruses, but the second most abundant viral similarities were to Herpesviridae (30). In addition, >50% of the all viral sequence similarities (including phages and Archaeal viruses) from linker amplified shotgun libraries of healthy and bleached D. strigosa were to Herpesviridae, indicating that herpesviruses are commonly associated with corals and not a result of our methodology.

Caveats.

The initial metagenomic experiment was designed both to address questions about the viral consortia present in corals and to determine whether this high-throughput metagenomic approach is amenable for empirical experiments. Therefore, treatments were designed to be harsh and as a result did not reflect realistic stressor levels. However, the second 2007 experiment was redesigned to represent more environmentally relevant levels of stress and used to confirm trends from the metagenomes.

In addition, equal amounts of DNA were combined from each replicate coral at each time point to generate individual stressor libraries that contained between 36 and 45 separately extracted viromes. This pooling approach provided a reduction in the variance, or normalizing shifts in the viral consortia over-time. The metagenomes therefore represent an underestimation of the true changes in the viral consortia. The individual temporal samples were then used for real-time PCR to verify identified trends.

The method used to collect the coral subsamples or nubbins required removal of the corals from the aquaria and repeated tissue damage to the coral finger. Corals from the no treatment aquarium control had elevated numbers of herpes-like viral sequences in both experiments. It was apparent from the real-time PCR that this method was disruptive enough to induce herpes-like viral particle production albeit at a reduced level compared with the other stressors. However, naturally occurring damage to corals results from predation, weathering events like hurricanes and earthquakes, and human activities, such as the aquarium coral and ornamental fish trade, recreational diving, and research (53). Our experimental design may have inadvertently mimicked fragmentation and small scale predation on corals. It is therefore likely that these activities and events are stressful enough to induce production of herpes-like viruses on the reef and may result in short or long term reductions in coral health that have until now gone unnoticed.

Last, the metagenomic DNA was amplified using multiple displacement amplification with Phi29 polymerase before sequencing, which could artificially inflate the occurrence of rare sequences, ssDNA, and small circular genomes, thus biasing the analysis (54, 55). However, multiple displacement amplification generally provides an even representation of genomes except at the ends (56) and, as evidenced by the elevation of ssDNA viruses in only the DOC sample, did not appear to bias these data.