Systematic discovery of complex in vivo dynamics of the microbiota

Longitudinal studies of host microbial ecosystems are important for determining how the microbiota affect human health and disease. However, the complexity of these dynamic systems necessitates robust modeling approaches to elucidate alterations in microbial communities relevant to the state of the host. We developed an integrated experimental and computational framework to systematically analyze time-dependent changes in gut microbial ecosystems after a defined perturbation. Employing a murine model of infectious colitis, we generated 2 month long time-series of 16S rRNA gene signatures at multiple intestinal sites from mice infected with an enteric pathogen, or from uninfected controls. We introduce a computational model that infers temporal trajectories for individual microbial taxa from high-throughput sequencing data, and groups trajectories with similar dynamics to identify members of the ecosystem that share common behaviors. Application to our 16S dataset revealed alterations in dynamics of the microbiota that correlate with introduction of the pathogen and phases of the host response. These changes differ across anatomic sites at multiple levels of resolution, from systems-level metrics to effects among individual species. Computational inference of site specific functional groupings of taxa reveal cascades of coordinate behavioral changes, uncovering sub-communities that may interact synergistically or antagonistically with the pathogen or host. Using quantitative culture data, we validate predictions for predominant organisms, demonstrating that our model accurately reconstructs dynamic microbial signatures to the species level of resolution. This integrated framework provides new insights into the behavior of complex host ecosystems and enables focused hypothesis generation to direct further experimental studies.