COS 83-3
Using hierarchical, multi-response models to disentangle coexposure and parasite interactions across multiple scales in wetland amphibian communities

Wednesday, August 12, 2015: 2:10 PM
322, Baltimore Convention Center
William E. Stutz, Ecology and Evolutionary Biology, University of Colorado, Boulder, Boulder, CO
Cheryl J. Briggs, Dept. of Ecology, Evolution & Marine Biology, University of California, Santa Barbara, Santa Barbara, CA
Andrew R. Blaustein, Department of Integrative Biology, Oregon State University, Corvallis, OR
Jason R. Rohr, Integrative Biology, University of South Florida, Tampa, FL
Jason T. Hoverman, Department of Forestry and Natural Resources, Purdue University, West Lafayette, IN
Pieter T.J. Johnson, Ecology and Evolutionary Biology, University of Colorado, Boulder, CO
Background/Question/Methods: Interactions between coinfecting pathogens have the potential to affect hosts (pathology and fitness), parasites (virulence and transmission) and disease dynamics in natural systems. An ongoing challenge is infer the presence and strength of such interactions using field-data collected across multiple spatial (e.g. sites or regions) and biological (host species) scales. Specifically, correlations in parasite load or the probability of infection at the individual level can indicate pathogen interactions, but such associations can also arise through correlated exposure at other scales even in the absence of direct parasite interactions. Both spatial- and species-level heterogeneity in parasite occurrence/abundance can generate patterns of co-exposure, leading to potentially spurious inferences about the presence or strength of pathogen interactions among individuals. We measured abundances of three potentially interacting species of pathogen – Batrachochytrium dendrobatidis, Ranaviruses, and the larval trematode Ribeiroia ondatrae – from a sample of 2152 amphibian hosts (5 species) collected from 93 California wetlands in a single year. We show how a multi-response, hierarchical, generalized linear model framework can be used to draw inferences about the strength and potential sources of correlated infection at across sites and species, thus enabling stronger inference about potentially interacting parasites.

Results/Conclusions: Each pathogen varied substantially in prevalence and mean abundance among sites, host species, and individuals. At the site scale, Ranavirus and R. ondatrae mean abundances correlated negatively (r = - 0.38, credible interval [-0.62,-0.11]), suggesting the presence of an unmeasured site-level variable driving opposing patterns of exposure for these two parasites. However, after accounting for the (co)variation among sites and species, we found positive mean correlations for two of three pathogen pairs at the individual host level. Ranavirus was positively correlated with both R. ondatrae (r = 0.23, CI[0.03,0.42]) and Bd abundance (r = 0.21, CI[0.001, 0.41]), suggesting that these pathogens may interact to facilitate infection, for example by increasing host susceptibility or reducing hosts ability to clear infections. We found no strong correlations in mean abundance among the five host species. Finally we illustrate how including additional predictors such as individual size or the density of intermediate hosts can indicate and account for sources of co-exposure at different scales of analysis. Overall, these results illustrate how appropriately formulated and parametrized hierarchical models can be used to assess patterns of coinfection across scales and to enable stronger inference about potentially interacting pathogens.