Mixed network structure in a coevolving host-parasite system

Genetic interactions between parasites and hosts determine how these antagonists coevolve. Networks depicting infection outcomes between host and parasite genotypes can be analyzed to characterize phenotype-level interactions. However, an open question is whether such network analyses can be used to infer allele-level interaction matrices and draw conclusions about coevolution. We address this question in a system coevolving by Red Queen dynamics and where a matching-allele model has been observed: the virulent bacterium Pasteuria ramosa and its crustacean host, Daphnia magna. We analyze infectivity representing the phenotype under selection. Additionally, because infectivity and resistance are polygenic and depend in part on the attachment site of the parasite spores, we analyze each of 5 attachment sites separately. We find the attachment site-specific matrices vary strongly in network structure. Foregut attachment shows only weak evidence of the modular network structure predicted from matching-allele models. For 3 other attachment sites, we find the more resistant hosts tend to be infected by the more generalist parasites, creating nested network structures. The matrices of the overall phenotype tend toward modular network structures. Overall, our findings suggest network analysis is a weak tool to predict the genetic interaction model underlying coevolution.