Marine viruses discovered via metagenomics shed light on viral strategies throughout the oceans

Abstract

Marine viruses are key drivers of host diversity, population dynamics and biogeochemical cycling and contribute to the daily flux of billions of tons of organic matter. Despite recent advancements in metagenomics, much of their biodiversity remains uncharacterized. Here we report a data set of 27,346 marine virome contigs that includes 44 complete genomes. These outnumber all currently known phage genomes in marine habitats and include members of previously uncharacterized lineages. We designed a new method for host prediction based on co-occurrence associations that reveals these viruses infect dominant members of the marine microbiome such as Prochlorococcus and Pelagibacter. A negative association between host abundance and the virus-to-host ratio supports the recently proposed Piggyback-the-Winner model of reduced phage lysis at higher host densities. An analysis of the abundance patterns of viruses throughout the oceans revealed how marine viral communities adapt to various seasonal, temperature and photic regimes according to targeted hosts and the diversity of auxiliary metabolic genes.

Figure 1. Clustering of the MVCs and the reference phage genomes based on the Dice distances.

The MVCs (blue) form novel branches with low similarity to reference phage genomes (red), indicating that they are members of previously unknown lineages of viral diversity. The branch lengths are ignored to better display the clustering topology. Supplementary Data 3 displays a circular version emphasizing exact branch lengths, and Supplementary Data 4 is a circular version that also ignores the branch lengths.

Figure 2. Viral co-occurrence networks.

The large diamonds represent the reference viral genomes colour coded according to the host phylum, and the small grey diamonds represent the MVCs. The line colours follow a gradient according to SparCC score from blue (0.6) to red (0.9). (a) The network displaying the strongest correlations with a SparCC score >+0.6 between reference phage genomes only. (b) The network displaying the strongest correlations with a SparCC score >+0.6 between MVCs and reference phage genomes.

Figure 3. The abundance patterns of the MVCs and the reference viral genomes across 121 marine viromes.

(a) The rank abundance curve of the top 500 most abundant reference viral genomes and MVCs. (b) The x axis shows the prevalence (percentage of samples in which an MVC was detected), while the y axis shows the median relative abundance of such MVCs across the 121 marine virome samples analysed. MVCs are colour coded according to their predicted host phylum. (c) The same as in b but displaying the prevalence and median relative abundance of the reference phage genomes, colour coded according to the phylum of their hosts. Supplementary Data 6 displays the average abundance and prevalence of all MVCs and reference phage genomes across the 121 samples.

Figure 4. Virome nonmetric multidimensional scaling.

The Manhattan distances were calculated based on the viral genome relative abundances and used as the input for a NMDS analysis. (a) POVs from photic (light blue) and aphotic (dark blue) zones. (b) Tara oceans viromes from warm (light green) and cold (dark green) waters. (c) Abrolhos viromes from summer (light red) and winter (dark red) seasons.

Figure 5. Variables displaying significant changes in abundance across sample groupings.

The bar lengths (y axis) are proportional to the number of variables in a given category (x axis) enriched in each of the tested sample groupings (that is, photic, aphotic, warm and cold) as determined by the Mann–Whitney test (corrected P value <0.05). (a) Enriched MVCs grouped according to the predicted host phylum. (b) Enriched reference viral genomes grouped according to the known host genus. Cyanobacteria refers to viruses that infect Prochlorococcus and Synechococcus. (c) Enriched KOs grouped according to the metabolic pathways to which they belong.

Figure 6. Associations between the microbial host abundance and the virus–host ratio.

The x axis displays the abundances of microbial taxa and the y axis displays VHR, calculated based on the relative abundances of microbial taxa and the viruses that infect them in the analysed Tara oceans microbial metagenomes and viromes. (a) Microbial taxa are summed at the taxonomic levels of genus and VHR was calculated using the abundances of reference viral genomes only. (b) Microbial abundances are summed at the taxonomic level of phylum and VHR was calculated using the abundances of both reference viral genomes and the MVCs for which a putative host was identified.

Figure 7. Conceptual model depicting viral strategies for exploiting the marine microbiome.

In the warm waters of the photic zone, Cyanophages would be enriched and display a preference for lysogenic infections. Under these same conditions, Pelagiphages and viruses infecting heterotrophic bacteria would be depleted and prefer lytic infections. In the cold waters of the photic zone, the opposite pattern would occur: Cyanophages depleted and lytic, and Pelagiphages and viruses infecting heterotrophic bacteria would be enriched and lysogenic infections. In the cold waters of the aphotic zone, both Cyanophages and Pelagiphages would be depleted and lytic, while viruses infecting heterotrophic bacteria would be enriched and lysogenic. Throughout these gradients, these viruses carry different types of auxiliary metabolic genes that help them to exploit host metabolism during infection.