Engineering ketoreductase via combined semi-rational and deep learning strategies for production of hydroxypropyl tetrahydropyranetriol

Abstract

Hydroxypropyl tetrahydropyranetriol (HT), an important cosmetic raw material, has attracted wide attention due to its remarkable skincare effects. To overcome the issues of high pollution and purification challenges associated with its chemical synthesis, enzyme-catalyzed biosynthesis has emerged as a favorable alternative. In this study, a ketoreductase cloned from Scheffersomyces stipitis CBS 6045 (SsCR), capable of asymmetrically reducing carbonyl compounds to produce chiral alcohols, was used to synthesize HT. We implemented a multi-combination screening strategy, integrating computer-aided semi-rational design and deep-learning technology, to systematically identify potential mutation sites in ketoreductase, followed by site-directed and combinatorial mutagenesis. Experimental validation showed that the enzyme activity of mutant SsCR-M2 increased by 86.7 % (9.60 U·mg⁻¹), compared to wild-type (WT) SsCR. Molecular docking and dynamics simulations indicated that SsCR-M2 possesses a more stable conformation, facilitating easier entry of β-acetoxyloside into the catalytic center. Notably, SsCR-M2 exhibited higher tolerance to elevated β-acetoxyloside concentrations compared to WT, successfully catalyzing the formation of 18.39 g·L⁻¹ of HT with a conversion rate of 91.93 %. These findings demonstrate the potential of engineered SsCR variants for sustainable HT production.