AI-FRET-guided engineering of subtilisin for green aqueous-phase amide bond formation with enhanced catalytic activity

Amide bond formation is a pivotal reaction in biochemistry. Chemical synthesis often relies on stoichiometric coupling reagents or non-green catalytic systems, which typically generate large amounts of waste, pose potential threats to human health, and cause substantial environmental risks. Here, we report that wild-type subtilisin can directly catalyse the condensation of free carboxylic acids and amines in purely aqueous media, thereby eliminating the need for hazardous and environmentally detrimental chemicals. We demonstrated that wild-type subtilisin successfully catalyzed the coupling of peptide fragments Suc-Ala-Ala-Pro-Phe and Gly-Tyr-NH₂ to form the amide product Suc-Ala-Ala-Pro-Phe-Gly-Tyr-NH₂ in PBS buffer (pH 8.0) at 40 °C. To address the inefficiency of conventional enzyme optimization requiring large mutant libraries and lengthy validation, we integrated an artificial intelligence (AI)-guided prediction platform with a Förster resonance energy transfer (FRET)-based rapid detection platform. Binding free energy analysis of three critical catalytic residues (N155, N218, and M222) yielded 7999 in silico candidates, which were narrowed to five prioritized mutants for experimental validation. The optimal mutant, N155G/N218D/M222Q, was validated within 2 min and exhibited a 753 ± 14 % increase in relative catalytic activity compared to the wild-type enzyme. This engineered mutant successfully catalyzed the coupling of biotin with the peptide fragment DFELL-NH₂ under identical aqueous conditions, expanding substrate scope beyond conventional peptide-peptide ligations to include non-peptidic substrates. This study establishes a green catalytic route for amide bond formation in purely aqueous media catalyzed by wild-type subtilisin, and demonstrates an integrated strategy combining AI-guided prediction with FRET-based rapid detection to accelerate enzyme engineering.