Network models for the computational design of proficient enzymes

Billions of years of evolution have made enzymes superb catalysts capable of accelerating reactions by several orders of magnitude. The underlying physical principles of their extraordinary catalytic power still remains highly debated, which makes the alteration of natural enzyme activities towards synthetically useful targets a tremendous challenge for modern chemical biology. The routine design of enzymes will, however, have large socio-economic benefits, as because of the enzymatic advantages the production costs of many drugs will be reduced and will allow industries to use environmentally friendly alternatives. The goal of this project is to make the routine design of proficient enzymes possible. Current computational and experimental approaches are able to confer natural enzymes new functionalities but are economically unviable and the catalytic efficiencies lag far behind their natural counterparts. The groundbreaking nature of NetMoDEzyme relies on the application of network models to reduce the complexity of the enzyme design paradigm and completely reformulate previous computational design approaches. The new protocol proposed accurately characterizes the enzyme conformational dynamics and customizes the included mutations by exploiting the correlated movement of the enzyme active site residues with distal regions. The guidelines for mutation are withdrawn from the costly directed evolution experimental technique, and the most proficient enzymes are easily identified via chemoinformatic models. The new strategy will be applied to develop proficient enzymes for the synthesis of enantiomerically pure β-blocker drugs for treating cardiovascular problems at a reduced cost. The experimental assays of our computational predictions will finally elucidate the potential of this genuinely new approach for mimicking Nature’s rules of evolution.

NetMoDEzyme is organized in four subprojects. First, Markov State network models are applied to characterize the enzyme structure and dynamics thus identifying the enzyme’s most populated conformational states (S1). Second, network community analysis is performed on these states to determine the enzyme regions correlated to active site residues and substrate binding (S2). Third, the rules of operation of the laboratory-based DE are analyzed in terms of occurrence frequencies and evolutionary conservation, and a Sequence-Activity relationship are developed (S3). Finally, the new protocol has to be established (from the outcome of S1-3), applied, and validated in a superfamily of enzymes relevant for the potential applications in the synthesis of enantiomerically pure β-blockers (S4).

We have successfully generated Markov State Models (MSM) for most of the training enzymes (S1), have developed new unprecedented tools based on correlation-based measures and graph theory (S2) that allow, for the first time, the prediction of distal active site mutations that lead to enhanced enzymatic activity (ACS Catal. 2017, 7, 8524; ACS Catal. 2021, 11, 13733), (S3) have performed some experimental validations within the host institution and also via external collaborations for identifying the key parameters for activity, and (S4) have applied the developed methodologies for the design of new enzymes for the synthesis of beta-blocker drugs.

The obtained results have been presented in multiple international and national conferences: plenary lectures at Gordon Research Conference on Biocatalysis 2016, EuChemS 2022, and TheBio2017; invited lectures at 252nd ACS meeting (Washington), 256th ACS meeting (Boston), 262th ACS meeting (Atlanta), Biotransformations 2017, Protein Engineering Canada 2018, 7th EuChemS 2018, Rideal conference (2018), CECAM workshop (Stuttgart 2018), AmineBioCat 2020, Biotrans 2021, IUPAC Canadian Chemistry conference (2021), FEBS practical course (2021), among others.

We have also established some collaboration agreements with two companies for applying the computational pipelines for enzyme design: one NDA signed with the chemical company BASF, one collaboration project with the French company SEQENS for the 2022-2023 period.

Our studies included in S1 and S2 have provided evidence that Molecular Dynamics simulations, coupled to correlation-based tools similar to those used to investigate processes such as allosteric regulation and molecular recognition, can be successfully applied in the enzyme (re)design field. We have developed the Shortest Path Map (SPM), which analyzes the different conformational sub-states sampled along the MD trajectory and identifies which residues are important for the sub-state inter-conversion. Therefore, if catalytically competent states are sampled in the MD simulation, the new tool facilitates identification of residues that contribute to the inactive-to-active inter-conversion. SPM thus identifies both active site but also distal residues that could lead to a population shift toward the catalytically competent conformation for novel activity. This is totally unprecedented, and thus opens the door to new computational paradigms that are not restricted to active site alterations. The development of subprojects S3 and S4 have involved the experimental characterisation of many of the computational designs to reveal the key descriptors for enhanced catalytic activity. In this line, we have computationally designed many different enzyme variants based on alpha,beta-hydrolase scaffolds and also based on Halohydrin Dehalogenases for the synthesis of enantiomerically pure beta-blocker drugs. We have also found that the combination of SPM with ancestral sequence reconstruction is a powerful strategy to be exploited for computational enzyme design.

Periodic Reporting for period 5 - NetMoDEzyme (Network models for the computational design of proficient enzymes)

Partager cette page

Télécharger