Computer-Aided Directed Evolution Achieves Balanced Activity, Thermal Stability, and Selectivity in Stereoselective Carbonyl Reductase

Achieving simultaneous improvements in activity, substrate specificity, stereoselectivity, and thermal stability remains a central challenge in laboratory enzyme evolution. Enzymes with industrial properties require synergistic development of various functions. This study developed a computational pipeline integrating sequence-structure information to regulate the activity, thermal stability, and stereoselectivity of carbonyl reductases. Our framework employs an unsupervised epistasis model to map residue interdependencies across the entire protein structure, combined with ΔΔ G fold calculations and conservation analysis, enabling global evolutionary engineering. This strategy dramatically reduced 6,720 potential mutations to 27 prioritized candidates. Greedy combinatorial strategy generated optimized mutants with enhanced activity (up to 28-fold), high stereoselectivity toward 22 structurally diverse substrates, and improved thermal stability (Δ T m up to 5.8 °C). For the substrate 2-acetylpyridine (H1), I51L/Y61F/D147E (M3) is the dehydrogenase with the highest activity reported so far. Systematic analysis of crystal structures and molecular dynamics simulations revealed that distal mutations reorganized interdomain communication networks, increasing the active population of prereaction state conformations. The introduction of distal mutations balanced overall protein fluctuations by redistributing flexibility across different regions, contributing to the simultaneous improvement of catalytic activity and thermal stability. This work demonstrates the efficiency of a computer-aided protein design approach for synergistically enhancing multifunctional compatibility, offering a transformative strategy for advancing biomanufacturing of high-value chiral compounds.