Abstract
With hundreds of species interacting with each other as well as with specific proteins and cells in our body, the gut microbiome is a complex ecosystem embedded within a complex organism. Microbiome impacts on host health can shape key aspects of fitness, such as development1, fecundity2, and lifespan3,4, while the host in turn can shape the microbiome5. However, complex interactions between microbes can make impacts unpredictable, such as when toxin-producing Clostridium species cause pathogenesis after antibiotics reduce gut diversity6. A pressing need exists to deconstruct the effects of gut diversity on host health, and new mathematical frameworks are needed to quantify the high dimensionality of this problem. Central to the microbiome-host relationship are questions of how bacterial diversity is maintained in the gut7 and how this diversity impacts host fitness8. Here we show that interactions between bacteria are major determinants of host physiology and the maintenance of bacterial diversity. We performed a complete combinatorial dissection of the naturally low-diversity Drosophila gut microbiome using germ free flies colonized with each possible combination of the five core species of bacteria, forming a discrete 5-dimensional cube in ecological state space. For each species combination, we then measured the resulting bacterial community abundances and fly fitness traits including (i) development, (ii) reproduction, and (iii) lifespan. Notably, we found that the fly gut environment promotes bacterial diversity, which in turn accelerates development, reproduction, and aging. From these measurements we calculated the impact of bacterial interactions on fly fitness by adapting the combinatorial geometry approach of Beerenwinkel-Pachter-Sturmfels9 (BPS), originally built to measure genetic interactions, to the microbiome10. We found that host phenotypes (e.g. lifespan) from single associated bacterial species are not predictive of host phenotypes when in diverse communities. Furthermore, we found evidence that high-order interactions (involving 3, 4 and 5 species) are widely prevalent and impact both host physiology and the maintenance of bacterial diversity, which ecologists have recently predicted11. In regard to evolution, the impacts of bacterial interactions on community composition parallel the impacts on host fitness traits, providing a feedback, which, propagated over time, may poise a population for emergence of co-evolving microbiome-host units.