Computational Strategies for Engineering Non-Heme and Heme Iron-Dependent Enzymes

Enzymes are macromolecules that function as biological catalysts, facilitating over 5,000 biochemical reactions through enzymatic catalysis. These reactions occur under relatively mild conditions and exhibit high efficiency and selectivity. Iron, the most abundant transition metal in biological systems, serves as a cofactor for many enzymes. Based on the structure of their active sites, iron-dependent enzymes can be classified into different types. In this study, we conducted an in-silico investigation of two non-heme iron-dependent dioxygenases, Novosphingobium aromaticivoran (NOV1) and Caulobacter segnis Dioxygenase (CsO2), both of which have similar and relatively narrow substrate scopes. Our aim is to perform computational approaches to identify the potential mutants capable of processing substrates not recognized by wild-type NOV1 and CsO2. Firstly, we utilized the deep learning-based protein design tool, proteinMPNN, to redesign the enzymes sequences, enhancing their ligand binding affinity and catalytic activity. Secondly, we employed the Rosetta protein design suite, a physics-based approach, to redesign NOV1 and CsO2 active sites for stabilizing reaction transition states. This process involved: i) identifying reaction transition states through quantum mechanics calculations, ii) constructing a theozyme based on the transition states and key residues, and iii) subsequently using RosettaMatch algorithm to refine the enzyme active site. CYP153A6, a heme iron dependent monooxygenase that catalyzes allylic hydroxylation with highly regioselectivity. However, due to the inherent structural flexibility of cytochrome P450 enzymes and the presence of multiple substrate and water tunnels, we applied an ensemble docking strategy to study this enzyme’s molecular recognition. We then identified and resigned key hotspot residues within these tunnels. Through tunnel engineering, CYP153A6 can be optimized to improve catalytic efficiency and broaden its substrate scope. We hope our computational strategies can provide potential applicability to general enzymatic systems for enhancing catalytic activity and expanding substrate specificity.