Non-random patterns in viral diversity

Abstract

It is currently unclear whether changes in viral communities will ever be predictable. Here we investigate whether viral communities in wildlife are inherently structured (inferring predictability) by looking at whether communities are assembled through deterministic (often predictable) or stochastic (not predictable) processes. We sample macaque faeces across nine sites in Bangladesh and use consensus PCR and sequencing to discover 184 viruses from 14 viral families. We then use network modelling and statistical null-hypothesis testing to show the presence of non-random deterministic patterns at different scales, between sites and within individuals. We show that the effects of determinism are not absolute however, as stochastic patterns are also observed. In showing that determinism is an important process in viral community assembly we conclude that it should be possible to forecast changes to some portion of a viral community, however there will always be some portion for which prediction will be unlikely.

Figure 1. Distribution of the nine macaque sampling sites.

Number of individuals sampled at each site is indicated. All sites are urban/peri-urban, with known contact between macaques and people, livestock and domestic animals (though frequency of contact is not assessed here). Number of samples collected is not consistent across sites, however sampling effort (number of collection days per site) is the same. Further description of each site is provided in Supplementary Table 3.

Figure 2. Virodiversity of rhesus macaques in Bangladesh.

(a) Diversity of viruses discovered by family. (b) Maximum likelihood phylogeny of picobirnavirus diversity. Although viruses from many families were discovered, we have selected to show the PbV tree because of the substantial diversity observed. One section of the tree is shown as an insert for better visualization. (c) Viral discovery curve to assess saturation and estimate the total richnesss (number of viruses) that exists. Red line is the collector curve. Solid black line is the rarefaction curve. Dotted line indicates Chao2 estimation of asymptotic richness by sample number, and shading indicates 95% confidence intervals. (d) Rank abundance curve of the 184 viruses detected in this study.

Figure 3. Two-mode affiliation network demonstrating the link between viruses and their hosts.

Viruses shaded grey. Size of node indicates abundance. Coloured nodes indicate individual macaques, coloured by site. (a) The overall network shows a non-random pattern of site-associated diversity. (b) Insert showing an area of apparently stochastic distributions. (c) Lower corner (blue shading) shows a pairwise assessment of beta (β) diversity between sites. The Jaccard (incidence based) index was used and demonstrates that little similarity exists between sites. Additional metrics shown in Supplementary Fig. 1. Top corner (green shading) presents results of the null model, and shows that the observed distributions are different from chance (dark green=significant P value). (d) Distribution of pairwise genetic identities for PbVs found in the same host, and those found in different hosts. Results presented are for G1 PbVs, but results consistent for G2. P value (Wilcoxon rank-sum test) indicates that the ‘within same host' values (max=85.5%) are different from chance.

Figure 4. One-mode affiliation network demonstrating the frequency of viral co-occurrence in the same host.

Viruses (coloured by family/genus) are linked if found in the same individual. Frequency of co-occurrence indicated by thickness of the edge connecting each node. Insert shows significant (C-score P<0.05) co-occurrence of AaVs with AdV and HV (significance determined using PAIRS).