1) Mapping and modeling interpersonal differences in gut microbial biotransformation activity
We integrated high-throughput metabolomics and metagenomics to assess how 265 drugs are biotransformed by 87 gut microbiota communities from humans and preclinical models. Our large-scale, time-resolved screen of over 23,500 drug-microbiota interactions revealed that over 90% of drugs were metabolized (≥25%) by at least one community, underscoring the pervasiveness of microbial drug metabolism. These include novel biotransformation events, supporting the idea that community-level traits emerge beyond isolated strains. This highlights the power of our assay system to uncover hidden metabolic capacities. While chemical similarity explained some patterns, specific functional groups alone were insufficient predictors, revealing the complexity of interactions.
To link interpersonal gut microbiota composition with biotransformation activity, we developed a predictive framework—from linear to non-linear, taxonomically complete models using microbial composition and enzymatic profiles. These showed mid-to-high accuracy for several drugs, validating that microbiota function can be computationally inferred. The models often indicated multi-species contributions rather than single-strain effects, supporting the utility of ensemble approaches.
Despite sample size and modeling limits, our work offers a scalable platform for hypothesis generation and model development. The unified metabolomics workflow and conservative thresholds (25–90% biotransformation) enhance reproducibility amid variability. Our dataset supports AI-driven exploration of microbial drug metabolism and emphasizes the need to consider nuanced chemical and ecological features for accurate predictions.
2) Establishing an in vitro pipeline for the high-throughput quantification of epithelial permeability of gut bacterial metabolites
We developed a scalable workflow to study how gut microbial activity affects epithelial drug and metabolite absorption. This system combines anaerobic bacterial cultures with human Caco2 epithelial monolayers in a transwell setup, integrating untargeted metabolomics to track flux. A novel feature is our marker compound panel for real-time monitoring of epithelial barrier integrity, revealing subtle disruptions missed by TEER.
Using this approach, we assessed the permeability of 482 drugs and 172 microbial metabolites, identifying 33 new bacteria-drug pairs where microbial transformation altered absorption. The workflow recapitulated known biotransformations (e.g. prodrug deacetylation) and revealed new host-microbiota synergies.
We extended the platform to map the permeability of 397 abundant bacterial metabolites from 7240 detected features. Over 79% did not cross the epithelium, reinforcing its selective gatekeeping role. Some metabolites vanished apically without appearing basolaterally, suggesting Caco2 metabolism—supported by 73 newly detected basolateral metabolites.
Notably, bacterial supernatants affected barrier integrity without harming cell viability, implicating microbial metabolites—possibly via Toll-like receptor signaling—in tight junction modulation. The system is compatible with future enhancements (e.g. air-liquid interface, co-cultures), enabling broader gut-organ interaction studies.
This workflow provides a robust, adaptable tool for dissecting microbiota-drug-host interactions, with applications in pharmacokinetics, toxicology, and drug development.
