(629c) Improving De Novo Enzyme Design through a Preorganization-Oriented Computational Strategy: Kemp Elimination As a Case Study

Computational de novo enzyme design is a technique used to create artificial enzymes capable of catalyzing non-naturally occurring chemical reactions. Despite its potential, the current energy functions and conformational sampling methods have limitations in achieving sub-angstrom accuracy, which can result in low activities of de novo enzymes. In addition, the active site models used in de novo enzyme design often fail to fully characterize the complex catalytic geometrical constraints between the transition state and catalytic residues. This can result in a mismatched electrostatic environment of the scaffold, which further limits the efficiency and accuracy of the designed enzyme. To tackle the issues of the inaccuracy and complexity of active site models in de novo enzyme design, a preorganization oriented computational strategy based on computational enzyme design tool PRODA was developed and demonstrated by the creation of Kemp elimination enzymes. Quantum mechanics calculations were used to propose a preorganized active site model consisting of six catalytic residues that facilitate proton transfer from a carbon with a low energy barrier. Using the ProdaMatch algorithm, the complicated active site model was anchored into the TIM barrel scaffold 3AOF, the endoglucanase from Thermotoga maritima. The scaffold was selected from a backbone library consisting of 17,434 protein structures sourced from the Protein Data Bank. To stabilize the catalytic productive geometry, the low-energy amino acid sequences near the binding pocket were generated computationally using iterative enzyme redesign and molecular dynamics simulation. After three rounds of design selection and optimization, and experimental assays on only four mutants, the optimal designed variant (3AOF-KE03), which contained 17 mutation sites, was confirmed to have catalytic activity (k_cat/K_m = 14.04 M^-1s^-1) towards Kemp elimination, with measured rate (k_cat = 0.033 s^-1) enhancement of up to 104-fold. The computational strategy used in this study has the potential to be used for the creation of a diverse range of artificial enzymes capable of catalyzing reactions of industrial significance.