Final Report Summary - DIREVENZYME (Computational Exploration of Directed Evolution rules for tuning enzymatic activities)
Biocatalysis is based on the application of natural catalysts for new purposes, for which the enzymes were not designed. Although the first examples of biocatalysis were reported more than a century ago, biocatalysis was revolutionized after the discovery of an in vitro version of Darwinian evolution called Directed Evolution (DE). Despite the recent advances in the field, major challenges remain to be addressed. Up to date, the best experimental approach consists of creating multiple mutations simultaneously but limit the choices using statistical methods. Still, tens of thousands of variants need to be tested experimentally. In addition to that, little information is available as to how these mutations lead to enhanced enzyme proficiency. Significant advances in computational tools have enabled the de novo design of enzymes catalyzing unnatural reactions making use of the so-called inside-out computational protocol developed by the groups of Prof. Baker and Prof. Houk. Despite the initial computational successes, the most active computationally designed enzymes still perform quite poorly in comparison with the natural and DE-engineered enzymes. This project aims to computationally unveil the molecular basis of improved catalysis achieved by Directed Evolution. In particular, quantum mechanics (QM), and Molecular Dynamics (MD) simulations have been used to study: a) the reversal of enantioselectivity of natural enzymes, b) the conversion of natural enzymes into other biocatalysts, c) the expansion of the substrate scope of natural enzymes, and finally d) the proposal of new directions and improvements for the computational protocol.
MD simulations and QM calculations have been carried out to investigate the reversal of enantioselectivity achieved either by single active site mutations or remote aminoacid substitutions in an alcohol dehydrogenase (ADH) and D-sialic acid aldolase to L-KDO aldolase enzymes. Additionally, the molecular basis for converting natural enzymes into other biocatalysts has been elucidated for the case of esterases and epoxide hydrolases. The simulations indicate that just by mutating the obvious catalytic residues is not enough for conferring the enzyme new catalytic activity, as neither the catalytic triad, nor the binding residues required for stabilizing the transition states of the reaction are well positioned in catalytically competent arrangements. The application of long timescale MD simulations to epoxide hydrolases has also revealed a key conformational state, not previously observed by means of X-ray crystallography and short MD simulations, that presents the loop containing one of the catalytic residues in a wide-open conformation, which is likely involved in the binding of the epoxide substrate. Thus, the identification of such conformational state is key for the engineering of the enzyme active site and conformational dynamics to alter its substrate scope and allow the acceptance of bulkier pharmacologically-relevant targets.
The computational study of the laboratory-engineered enzyme variants has elucidated the strengths and weaknesses of existing computational protocols. Based on the results obtained in this project, the conformational dynamics of the enzyme is a key feature that needs to be carefully analyzed for computationally predicting mutations to engineer the enzyme activity, selectivity, and substrate scope. The development of more robust computational methods to predict amino-acid changes needed for activity is of the utmost importance as the need for experimentally probing randomized sequences would be greatly reduced, rendering the route to novel biocatalysts much more efficient. This might represent a cheap and environmentally friendly alternative for industries to produce active catalysts for any desired target.
MD simulations and QM calculations have been carried out to investigate the reversal of enantioselectivity achieved either by single active site mutations or remote aminoacid substitutions in an alcohol dehydrogenase (ADH) and D-sialic acid aldolase to L-KDO aldolase enzymes. Additionally, the molecular basis for converting natural enzymes into other biocatalysts has been elucidated for the case of esterases and epoxide hydrolases. The simulations indicate that just by mutating the obvious catalytic residues is not enough for conferring the enzyme new catalytic activity, as neither the catalytic triad, nor the binding residues required for stabilizing the transition states of the reaction are well positioned in catalytically competent arrangements. The application of long timescale MD simulations to epoxide hydrolases has also revealed a key conformational state, not previously observed by means of X-ray crystallography and short MD simulations, that presents the loop containing one of the catalytic residues in a wide-open conformation, which is likely involved in the binding of the epoxide substrate. Thus, the identification of such conformational state is key for the engineering of the enzyme active site and conformational dynamics to alter its substrate scope and allow the acceptance of bulkier pharmacologically-relevant targets.
The computational study of the laboratory-engineered enzyme variants has elucidated the strengths and weaknesses of existing computational protocols. Based on the results obtained in this project, the conformational dynamics of the enzyme is a key feature that needs to be carefully analyzed for computationally predicting mutations to engineer the enzyme activity, selectivity, and substrate scope. The development of more robust computational methods to predict amino-acid changes needed for activity is of the utmost importance as the need for experimentally probing randomized sequences would be greatly reduced, rendering the route to novel biocatalysts much more efficient. This might represent a cheap and environmentally friendly alternative for industries to produce active catalysts for any desired target.