Statistical design of a synthetic bacterial community that clears a multi-drug resistant gut pathogen

Communities of microbes perform critical functions across many environments on Earth. Elucidating principles of their design is difficult despite our abilities to interrogate communities and their constituent bacteria at unprecedented resolutions. We present a process for engineering defined microbial communities de novo that is composition and mechanism agnostic. Our approach is based on a modified Design-Build-Test-Learn approach (DBTL+) coupled with statistical inference that considers only strain presence or absence of designed communities. Using clearance of multi-drug resistant (MDR) Klebsiella pneumoniae as a target function, we converged on a generative statistical model of K. pneumoniae clearance in just a single round of DBTL+. Statistical analysis of this model defined 15 strains that were key for community function. Combining these 15 strains into a community (SynCom15) cleared K. pneumoniae across various in vitro environments and was more efficacious than a whole stool transplant in a pre-clinically relevant mouse model of infection. Considering metabolic profiles instead of strain presence/absence resulted in a poor model for generating functional communities, demonstrating the utility of strain presence/absence for deriving principles of community design. Our work introduces the concept of statistical design for engineering synthetic bacterial communities, opening the possibility of synthetic ecology more broadly.