Designing γ-Carbonic Anhydrase as a Broad-Scope Metalloreductase with Ultrathermostability and Organic-Solvent Tolerance

Enzymes are essential for industrial catalysis but are often limited by poor thermostability and low activity in organic solvents. In this study, we utilized ProteinMPNN, a deep learning-based protein sequence design algorithm combined with evolutionary information, to design γ-carbonic anhydrase (γ-CA) as a broad-scope metalloreductase. Compared with the wild-type enzyme, the designed variants exhibited significantly enhanced catalytic activity. Subsequent optimization through directed evolution further improved the catalytic activity and superior enantioselectivity toward the asymmetric reduction of ketones and alkenes. Remarkably, the engineered γ-CA demonstrated ultrathermostability with a Tm of 96 °C, and resistance to organic solvents, even including 50% polar organic solvents, addressing key challenges in industrial biocatalysis. Mechanistic study through protein crystallography, MD simulations, and QM/MM calculations revealed the advantage of the identified L83A mutant, along with the reaction profile through the terminal zinc hydride intermediate. This study showcases the potential of combining deep learning-based protein design with traditional engineering methods to create robust and efficient biocatalysts. The results not only establish a framework for optimizing γ-CA but also provide a generalizable strategy for engineering enzymes with enhanced stability and activity under extreme conditions, paving the way for potential industrial applications.