Environmental Viral Genomes Shed New Light on Virus-Host Interactions in the Ocean

Abstract

Metagenomics has revealed the existence of numerous uncharacterized viral lineages, which are referred to as viral "dark matter." However, our knowledge regarding viral genomes is biased toward culturable viruses. In this study, we analyzed 1,600 (1,352 nonredundant) complete double-stranded DNA viral genomes (10 to 211 kb) assembled from 52 marine viromes. Together with 244 previously reported uncultured viral genomes, a genome-wide comparison delineated 617 genus-level operational taxonomic units (OTUs) for these environmental viral genomes (EVGs). Of these, 600 OTUs contained no representatives from known viruses, thus putatively corresponding to novel viral genera. Predicted hosts of the EVGs included major groups of marine prokaryotes, such as marine group II Euryarchaeota and SAR86, from which no viruses have been isolated to date, as well as Flavobacteriaceae and SAR116. Our analysis indicates that marine cyanophages are already well represented in genome databases and that one of the EVGs likely represents a new cyanophage lineage. Several EVGs encode many enzymes that appear to function for an efficient utilization of iron-sulfur clusters or to enhance host survival. This suggests that there is a selection pressure on these marine viruses to accumulate genes for specific viral propagation strategies. Finally, we revealed that EVGs contribute to a 4-fold increase in the recruitment of photic-zone viromes compared with the use of current reference viral genomes. IMPORTANCE Viruses are diverse and play significant ecological roles in marine ecosystems. However, our knowledge of genome-level diversity in viruses is biased toward those isolated from few culturable hosts. Here, we determined 1,352 nonredundant complete viral genomes from marine environments. Lifting the uncertainty that clouds short incomplete sequences, whole-genome-wide analysis suggests that these environmental genomes represent hundreds of putative novel viral genera. Predicted hosts include dominant groups of marine bacteria and archaea with no isolated viruses to date. Some of the viral genomes encode many functionally related enzymes, suggesting a strong selection pressure on these marine viruses to control cellular metabolisms by accumulating genes.

Keywords: genome; marine ecosystem; metabolism; metagenomics; virus.

FIG 1

Proteomic tree. The dendrogram represents proteome-wide similarity relationships among 4,240 prokaryotic dsDNA virus genomes. Branches are colored orange (EVG, environmental viral genome) or black (RVG, reference viral genome), and branch lengths are indicated using a logarithmic scale. TOV, Tara Oceans viromes; OBV, Osaka Bay viromes. The tree is midpoint rooted. Rings outside the dendrogram represent, from inside to outside, sources of genome data, taxonomic groups of known hosts, and viral family classifications.

FIG 2

Genus-level genomic OTU (gOTU) richness. The genome-wide similarity score (S_G) cutoff for clustering was set to 0.15 (i.e., viral-genus-level cutoff). The EVGs and RVGs were clustered together, and subsets of the EVGs and RVGs were then constructed by extracting each member. (A) Rarefaction curves for the number of gOTUs. Rarefaction curves are presented with shading representing 95% confidence intervals obtained from 100 bootstrap replicates using the R package iNEXT (107). Dashed curves represent extrapolations to 5,000 genome sequences. Numbers in parentheses represent the number of genomes and gOTUs. Chao1 richness estimates for the EVGs and RVGs are indicated. (B) Proportions of genus-level gOTU clusters. Colors represent the following cluster categories: EVG-only clusters (red), RVG-only clusters (blue), and shared clusters (gray).

FIG 3

Fifty-eight putative archaeal virus genomes. (A) Part of the proteomic tree with 2 OBV-EVGs (red) and 56 TOV-EVGs (orange), predicted to be derived from euryarchaeal tailed viruses infecting marine group II (MGII) species. Genomes with genes encoding DNA polymerase B (squares) and chaperonin (triangles) are indicated. Clade names and genus-level gOTUs are indicated. Numbers in parentheses represent the number of genomes of each clade or gOTU. The ranges of genome sizes and percent G+C contents for each clade are presented, with the exception that clade 2 includes a long contig (121 kb; asterisk). Branch lengths are logarithmically scaled from the root of the entire proteomic tree in Fig. 1. (B) Genome map of nine archaeal viral genomes that are indicated by stars in panel A. The sequences are circularly permuted and/or reversed. Red arrows indicate the original start position of the sequences. Putative gene functions are indicated. All tBLASTx alignments are represented by colored lines between the two genomes. The color scale represents tBLASTx percent identity.

FIG 4

Gene phylogenetic trees of DNA polymerase B and chaperonin. (A) Maximum likelihood tree of DNA polymerase B. The tree is rooted by four distant bacterial sequences (not shown) and includes 348 sequences. (B) Maximum likelihood tree of chaperonin. The tree is midpoint rooted and includes 381 sequences. In panels A and B, numbers in parentheses represent the number of sequences in each collapsed node. Colors represent taxonomies. Asterisks indicate collapsed nodes that include MGII (*) and MGI (**) sequences. The scale bar refers to the estimated number of amino acid substitutions per site. Numbers near the nodes represent bootstrap percentages of >50%. MGIIA and MGIIB indicate sequences from reported genomes ( and , respectively).

FIG 5

Genomes with Fe-S cluster assembly-related genes. (A) Four parts of the proteomic tree with genomes carrying Fe-S cluster assembly genes (i.e., ATC [■] and IscU [▲] genes). Branch lengths are logarithmically scaled as described for Fig. 3A. Genus-level gOTUs and genome identifiers (IDs), lengths, and percent G+C compositions are indicated. (B) Maximum likelihood tree of IscU genes. The tree contains protein sequences encoded in OBV_N00005 (red), five TOV-EVGs (orange), and 21 Proteobacteria and cyanobacterial genomes (black). The scale bar refers to the estimated number of amino acid substitutions per site. Numbers close to the nodes represent bootstrap percentages of >50%. The tree is rooted by the cyanobacterial Nostoc species sequence. (C) Genome map of OBV_N00085 and Pelagibacter phage HTVC008M. The HTVC008M sequence is circularly permuted at 97,000 bp and reversed. A red arrow indicates the original start position of the HTVC008M sequence. Putative gene functions of OBV_N00005 and HTVC008M (described in reference 10) are indicated. All tBLASTx alignments are represented by colored lines between the two genomes. The color scale represents tBLASTx percent identity. FAD, flavin adenine dinucleotide; NAD, nicotinamide adenine dinucleotide.

FIG 6

Two parts of the proteomic tree with EVGs of putative Flavobacteriaceae phages. Branch lengths are logarithmically scaled as described for Fig. 3A. Genus-level gOTUs are indicated. Numbers in parentheses represent the number of genomes in each gOTU. (A) Group 1 distributed in 29 gOTUs, including two Persicivirga phages (black), 5 OBV-EVGs (red), 147 TOV-EVGs (orange), and 7 other EVGs (blue). (B) Group 2 distributed in 25 genus-level gOTUs, including two Cellulophaga phages (phi40:1 and phi38:1; black; G508), IAS virus (black; G520), 3 OBV-EVGs (red), 75 TOV-EVGs (orange), and 2 other EVGs (blue). Genomes encoding chaperonins are indicated by a triangle.

FIG 7

Genomic alignment between the whole sequence of OBV_N00085 and a genomic region (385,000 to 425,000 bp) of “Candidatus Puniceispirillum marinum” IMCC1322. The OBV_N00085 sequence is circularly permuted at 15,000 bp for clarity, and a red arrow indicates the original start position of the sequence. Putative gene functions and function categories of OBV_N00085 are indicated by texts and colors. All tBLASTx alignments are presented. The color scale represents tBLASTx percent identity.

FIG 8

Recruitment of photic POV sequences to RVGs (blue) and to a pool of EVGs and RVGs (red). Mappings were performed with tBLASTn (proteins) and BLASTn (reads). In both mappings, the initial filtering of hits involved an E value of <1e−3, and an additional filtering was based on ≥60% identity and ≥80% alignment length of the query sequence. (A) Recruitment of proteins. (B) Recruitment of reads.