Decoding Substrate Specificity in a Promiscuous Biocatalyst by Enzyme Proximity Sequencing

Abstract Substrate specificity is a defining feature of enzyme function, but its molecular underpinnings remain difficult to decode and engineer. Here, we leveraged enzyme proximity sequencing (EP-Seq) to systematically map how single-point and combinatorial mutations reshape the substrate preferences of D-amino acid oxidase (DAOx) from Rhodotorula gracilis, a model promiscuous enzyme. We generated ∼40,000 sequence–phenotype pairs, enabling us to profile the activities of ∼6,500 unique DAOx variants against five D-amino acid substrates with distinct physicochemical properties. Our analysis revealed that substrate-specific mutations are distributed throughout the enzyme structure. Mutations near the active site drive strong specificity shifts but also incur catalytic penalties, while distal mutations subtly rewire intramolecular contacts in order to modulate specificity with minimal loss of activity. We identified and validated positional hotspots that act allosterically to influence specificity, and characterized key variants that acquired exclusive substrate specificity or exhibited up to 230-fold changes in substrate preference. Combining mutations with complementary effects further sharpened substrate discrimination, enabling rational design of highly selective biocatalysts. This work provides a powerful framework for decoding enzyme specificity and provides unique foundational datasets to advance AI-guided enzyme engineering.