Host-specific viral predation network on coral reefs

Abstract

Viral infections are major modulators of marine microbial community assembly and biogeochemical cycling. In coral reefs, viral lysis controls bacterial overgrowth that is detrimental to coral health. However, methodological limitations have prevented the identification of viral hosts and quantification of their interaction frequencies. Here, we reconstructed an abundance-resolved virus-bacteria interaction network in the oligotrophic coral reef waters of Curaçao by integrating direct microscopy counts with virus-host links obtained from proximity-ligation, prophage integration, and CRISPR spacers. This network of 3013 individual links (97 unique species-level interactions) revealed that the abundance of free viral particles was weakly related to host abundance and viral production, as indicated by the cell-associated virus-to-host ratio (VHR). The viruses with the highest free and cell-associated VHR, interpreted here as highly productive viruses, formed links with intermediate-to-low abundance hosts belonging to Gammaproteobacteria, Bacteroidia, and Planctomycetia. In contrast, low-production viruses interacted with abundant members of Alphaproteobacteria and Gammaproteobacteria enriched in prophages. These findings highlight the decoupling between viral abundance and production and identify potentially active viruses. We propose that differential decay rates and burst sizes may explain the decoupling between free viral abundance and production and that lysogenic infections play an important role in the ecology of high-abundance hosts.

Keywords: MetaHi-C; bacteriophage; infection network; lysogeny.

Figure 1

Viral and host abundances, diversity, and interactions. (A) a summary of the experimental design of the study, including the four main sample types (cell-associated metagenomes, viromes, Hi-C metagenomes, and microscopy counts) and the downstream abundance calculations and network analyses. (B) Sampling sites on the southern coast of Curaçao and associated microscopy counts. The three box columns show microscopy counts for each of the 17 sites (one microscopy sample was lost). The left column shows the virus-to-microbe ratio (VMR), the second column shows the VLP abundance, and the third column shows the bacterial abundance. (C) Genomic diversity of viruses identified in this study with links to prokaryotic hosts. The viral proteomic tree was generated based on the all-versus-all distances calculated from protein sequences encoded by CVDB viral genomes (denoted by circles) and their closest relatives among reference viruses recognized by the International Committee for Taxonomy of Viruses (ICTV). Blue dots indicate the presence of a Hi-C linkage, yellow dots indicate a prophage linkage, and green dots indicate a CRISPR linkage. The absence of a dot indicates a reference virus from ICTV. The outer ring color indicates the host identified through links or known hosts for ICTV viruses. Branch lengths were ignored to highlight the tree topology. Clusters I and II indicate clusters of viruses infecting the same host class (with one exception for cluster I). (D) Relative abundance of bMAGs (bars indicate the median and the standard error across samples). The left side shows the abundance of bMAGs with a viral linkage, whereas the right side shows the abundance of all identified bMAGs in the dataset.

Figure 2

Virus–bacteria interaction network. (A) Bipartite network sorted by modularity (normalized modularity score Q = 0.9099; number of clusters C = 29). The dark grey square indicates a virus-host link, and the red squares indicate modules. (B) Bipartite virus-host network sorted by genome similarity. Virus similarity was calculated from all-versus-all distances between protein-encoding sequences. The bMAG dendrogram was calculated using average nucleotide identity and percent of genome aligned. (C) Visualization of the distribution of link types (line type) across host taxonomy (node color). (D) PN/PS ratios for viral genes identified in the generalist viruses in the top left side of panel a. Only genes with an average PN/PS > 1 are shown. (E) Metabolic gene functions for viruses with different hosts. Magenta bubbles indicate that genes from the same module (pathway) were identified in both the host and virus.

Figure 3

Abundance-resolved network structure. (A) Median rank-abundance curve of hosts (circles) and viruses (triangles) across all samples (curves for each sample are displayed in Fig. S6). Solid lines connect Hi-C linkages, and dashed lines connect prophage linkages. (B) Variance of the viral rank (indicated by its standard deviation across samples) sorted by the median viral rank. (C) Variance of viral abundances (indicated by its standard deviation across samples when accounting for microscopy counts) sorted by the median viral rank. (d) Abundances of predicted temperate and lytic viruses with and without linkages. (E) Same as (d), but temperate is divided into prophage and temperate. (F) Virus-to-host ratio (VHR) of predicted lytic and temperate viruses. Letters above violins indicate groups with significant differences in abundances (Wilcoxon-rank sum test, P value <0.05).

Figure 4

Relationships between viral and host abundances. (A) Relationship between host and viral relative abundances in the cell-associated fraction (> 0.22 μm, light gray symbols, linear regression, R² = 0.21 and P value <0.001;) and free fraction (< 0.45 μm, black symbols, linear regression, R² = 0.02 and P value <0.001). (B) Relationship between host and free viral abundances (< 0.45 μm) in genomes per ml when direct counts from microscopy are incorporated (linear regression, P value = 1.014e-12, R² = 0.03), and informed by the cell-associated VHR (linear regression between viral abundance and cell-associated VHR is shown in Supplementary Fig. 9, P value <2e-16, R² = 0.08). (C) Relative abundances of hosts interacting with viruses in each VHR quartile compared with the relative abundances of all bMAGs in the dataset.

Figure 5

Virus-to-host ratios in free and cell-associated fractions. (A) Relationship between the free and cell-associated VHRs (linear regression, P value <2e-16, R² = 0.52, slope = 0.48). The solid symbols indicate predicted pairs where the bMAG carries a prophage and transparent symbols indicate prophage absence. The red lines represent the upper VHR quartiles, and the blue lines represent the lower quartiles. These quartiles divided the predicted virus–host pairs into four categories, starting from top left: 1) pairs with low cell-associated VHR and high free VHR (n = 18); 2) pairs with high VHR in both cell-associated free fractions (n = 170); 3) pairs with low VHR in both the cell-associated and free fractions (n = 96); and 4) pairs with high cell-associated VHR and low free VHR (n = 6). Host taxonomy in each category is shown in the pie charts. (B) Conceptual figure displaying potential mechanisms driving each category: 1) a low cell-associated VHR with a high VHR in the free-living fraction could be caused by low viral particle turnover rates, where particles are slow to decay in the environment relative to their production rates, leading to a high free VHR relative to the cell-associated. 2) High VHR in both fractions could be caused by highly productive viruses due to large burst sizes, short infection times, or a high proportion of infected host cells. 3) Low VHR in both fractions could be caused by overall low viral production, with the opposite underlying mechanisms compared to the high productivity group just described. 4) High VHR in the cell-associated fraction but low VHR in the free fraction may result from high viral particle turnover or long latent periods. (C) Metabolic modules enriched and depleted in bMAGs of each VHR category, displayed as the difference in abundance compared to the frequency in bMAGs in the middle VHR values.