Abstract
Xenobiotics, including pharmaceutical drugs, can be metabolized by both host and microbiota, in some cases by homologous enzymes. We conducted a systematic search for all known human proteins with gut microbial homologs. Because gene fusion and fission can obscure homology detection, we built a pipeline to identify not only full-length homologs, but also cases where microbial homologs were split across multiple adjacent genes in the same neighborhood or operon ("split homologs"). We found that human proteins with full-length gut microbial homologs disproportionately participate in xenobiotic metabolism. While this included many different enzyme classes, short-chain and aldo-keto reductases were the most frequently detected, especially in prevalent gut microbes, while cytochrome P450 homologs were largely restricted to lower-prevalence facultative anaerobes. In contrast, human proteins with split homologs tended to play roles in central metabolism, especially of nucleobase-containing compounds. We identify twelve specific drugs that gut microbial split homologs may metabolize; 2 of these, 6-mercaptopurine by xanthine dehydrogenase and 5-fluorouracil by dihydropyrimidine dehydrogenase, have been recently confirmed in mouse models. This work provides a comprehensive map of homology between the human and gut microbial proteomes, indicates which human xenobiotic enzyme classes are most likely to be shared by gut microorganisms, and finally demonstrates that split homology may be an underappreciated explanation for microbial contributions to drug metabolism.