A phylogenetic model for the recruitment of species into microbial communities and application to studies of the human microbiome

Abstract

Understanding when and why new species are recruited into microbial communities is a formidable problem. Much theory in microbial temporal dynamics is focused on how phylogenetic relationships between microbes impact the order in which those microbes are recruited; for example species that are closely related may exclude each other due to high niche overlap. However, several recent human microbiome studies have instead found that close phylogenetic relatives are often detected in microbial communities in short succession, suggesting factors such as shared adaptation to similar environments play a stronger role than competition. To address this, we developed a mathematical model that describes the probabilities of different species being detected in time-series microbiome data, within a phylogenetic framework. We use our model to test three hypothetical assembly modes: underdispersion (species are more likely to be detected if a close relative was previously detected), overdispersion (likelihood of detection is higher if a close relative has not been previously detected), and the neutral model (likelihood of detection is not related to phylogenetic relationships among species). We applied our model to longitudinal high-throughput sequencing data from the human microbiome, and found that for the individuals we analyzed, the human microbiome generally follows an assembly pattern characterized by phylogenetic underdispersion (i.e. nepotism). Exceptions were oral communities, which were not significantly different from the neutral model in either of two individuals analyzed, and the fecal communities of two infants that had undergone heavy antibiotic treatment. None of the datasets we analyzed showed statistically significant phylogenetic overdispersion.