Host microbiomes are ubiquitous and play an important ecological role in preventing establishment of microbial pathogens in a variety of host systems. However, few studies simultaneously test multiple alternative hypotheses to distinguish among various possible mechanisms through which host-associated microbiota are able to mediate pathogen success. Here, we use an invasion-community ecology framework in the well-studied amphibian system to ask whether five microbial community traits—(1) microbial community richness, (2) richness of Bd-inhibitory bacteria, (3) proportion of Bd-inhibitory bacteria, (4) compositional difference between individuals (dispersion) and (5) compositional differences over time (stability)—are predictors of infection by the emerging fungal pathogen Batrachochytrium dendrobaditis (Bd). Amphibian skin communities were sampled through time, before and after Bd exposure, in five amphibian species that range from highly susceptible to Bd-tolerant. Data from unexposed individuals was used to create Bayesian estimates of baseline microbial community traits for each species. Then, microbiome measurements from exposed amphibians were compared against baseline trait distributions to assess whether deviations in richness, composition, or stability were predictors for Bd infection. Finally, random forest models were used to assess how accurately infection risk or intensity could be predicted using community-level traits only, in comparison to ASV (bacterial species) data.
Results/Conclusions
Richness of Bd-inhibitory bacteria was negatively correlated with infection risk, while proportion of Bd-inhibitory bacteria was negatively correlated with infection intensity. This suggests that establishment and spread of Bd within a host is controlled by different mechanisms. Random forest models predicting infection risk and intensity did not perform better on validation test sets if individual bacterial ASV data were included, compared to a minimal model including only microbial community traits. This demonstrates that community-level traits have the potential to be just as informative as bacterial identity when quantifying microbial community-mediated pathogen resistance. In conclusion, our data shows that pathogen success can be predicted by community-level traits of the host-associated microbial community, and that resistance to Bd infection risk and intensity are controlled by distinct community traits.