Designing Consensus Catalytic Geometries for De Novo Multistep Enzymes

Abstract

Natural enzymes that catalyze multistep reactions possess highly organized active‑site geometries that require minimal atomic rearrangement during catalysis (<1.5 Å inter‑state average RMSD). While de novo enzymes for multistep reactions have previously been designed, none have explicitly leveraged this principle - a limitation likely contributing to their poor catalytic performance. In this thesis, I introduce a computational framework for the design of multistep de novo enzymes based on pre-optimized consensus geometries, using a Michael addition as a model system. Using density functional theory, theozymes were constructed for all seven transition states of a proposed mechanism mediated by a Lys/Tyr/His catalytic triad. Optimizing for a shared consensus geometry yielded a theoretical pathway with an average inter-state RMSD of 1.1 Å, successfully mirroring the organization of natural multistep enzymes. To physically realize this theoretical model, the consensus theozymes were embedded into a structural ensemble derived from a TIM-barrel scaffold. However, experimental characterization revealed that the designed sequences failed to adopt the target fold, largely reflected by inclusion body formation. Consequently, the viability of the consensus theozyme hypothesis remains experimentally untested. Nevertheless, this negative result stress-tests the design hypothesis and isolates limitations in the protein backbone redesign workflow. By diagnosing these failure points, this work establishes a foundation for deriving multistep consensus theozymes and provides an actionable roadmap for their future implementation using de novo generative scaffolding models.