Variance in C. elegans gut bacterial load suggests complex host-microbe dynamics

Variation in bacterial composition inside a host is a result of complex dynamics of microbial community assembly, but little is known about these dynamics. To deconstruct the factors that contribute to this variation, we used a combination of experimental and modeling approaches. We found that demographic stochasticity and stationary heterogeneity in the host carrying capacity or bacterial growth rate are insufficient to explain quantitatively the variation observed in our empirical data. Instead, we found that the data can be understood if the host-bacteria system can be viewed as stochastically switching between high and low growth rates phenotypes. This suggests the dynamics significantly more complex than logistic growth used in canonical models of microbiome assembly. We develop mathematical models of this process that can explain various aspects of our data. We highlight the limitations of snapshot data in describing variation in host-associated communities and the importance of using time-series data along with mathematical models to understand microbial dynamics within a host.Author summary: Bacterial population density is known to vary across individual hosts. What drives this variation is unclear. In this study, we use C. elegans as a easily controllable host, controlling for the age of the host and genetic heterogeneity to address this question. We quantify populations of individual bacteria species and their interactions with C. elegans. We found that bacteria behaved differently when grown in a host compared to the standard logistic growth observed in vitro. When bacteria grew within the host C. elegans, they exhibited density-dependent growth and the emergence of two distinct subpopulations of worms, one with high bacterial density and the other with low bacterial density. We also observed that hosts can switch between high and low population densities of bacteria. To describe this behavior, we developed a phenomenological model and a switching model for the bacterial dynamics inside the host that we simulated and compared with experimental data.