Coinfection Organizes Epidemiological Networks of Viruses and Hosts and Reveals Hubs of Transmission.

Multiple virus infections affect the competence of host plants to transmit disease. The effects of coinfection on transmission are expected to produce ecologically complex pathogen and host-pathogen interactions. However, the prediction of disease risk will rely on untangling nonrandom from random patterns of infection to identify underlying processes that drive these interactions. Are the spatial distributions of infections in complex multispecies systems random or not? For the first time, we use an empirical evaluation of this basic but nontrivial question to test the hypothesis that coinfection contributes to (i) nonrandom ecological interactions between hosts and viruses and (ii) structuring infection distributions. We use a novel approach that decomposed the ecological interactions of 11 generalist viruses in 47 host species in four habitats of an agroecosystem into single-infection and coinfection "modes." Then, we relate ecological structuring in infection networks to the distribution of infection using generalized regression models. The network analyses of coinfection showed that virus-host interactions occurred more often than expected at random in one of the four habitats, Edge. A pattern of specific interactions was shared between Edge and the ecosystem, indicating scale invariance. The regression modeling also showed that the plant community characteristics of Edge were unique in explaining infection distributions. The results showed that the spatial distribution of infection at the ecosystem level was not only a species-specific phenomenon but also, strongly structured by specific virus-virus and host-virus interactions. The evidence of scale invariance and the special role of Edge as a reservoir suggest that ecological interactions were less strongly structured by community differences among habitats than by wider-scale processes and traits underlying the interactions. Addressing whether reservoir communities significantly contribute to epidemiological processes at the ecosystem scale is a promising avenue for future research.