Small subsets of highly connected taxa predict compositional change in microbial communities

Abstract

For decades, ecological theory has predicted that the complexity of communities should be related to their stability. However, this prediction has rarely been tested empirically, because of both the difficulty of finding suitable systems where the question is tractable and the trouble of defining \"stability\" in real systems. Microbial communities provide the opportunity to investigate a related question: how does community connectivity relate to the rate of compositional turnover? We used a newly developed metric called community \"cohesion\" to test how microbial community connectivity relates to Bray-Curtis dissimilarity through time. In three long-term datasets, we found that stronger connectivity corresponded to lower rates of compositional turnover. Using two case studies of disturbed and reference communities, we found that the predictive power of community connectivity was diminished by external disturbance. Finally, we tested whether the highly connected taxa were disproportionately important in explaining compositional turnover. We found that subsets of highly connected \"keystone\" taxa, generally comprising 1-5% of community richness, explained community turnover better than using all taxa. Our results suggest that stronger biotic interactions within microbial community dynamics are stabilizing to community composition, and that highly connected taxa are good indicators of pending community shifts.