Carboxylic acid reductases (CARs) are biocatalysts of industrial importance. Their properties, especially their poor stability, render them sub-optimal for use in a bioindustrial pipeline. Here, we employed ancestral sequence reconstruction (ASR) – a burgeoning engineering tool that can identify stabilizing but enzymatically neutral mutations throughout a protein. We used a three-algorithm approach to reconstruct functional ancestors of the Mycobacterial and Nocardial CAR1 orthologues. Ancestral CARs (AncCARs) were confirmed to be CAR enzymes with a preference for aromatic carboxylic acids. Ancestors also showed varied tolerances to solvents, pH and in vivo-like salt concentrations. Compared to well-studied extant CARs, AncCARs had a Tm up to 35 °C higher, with half-lives up to nine times longer than the greatest previously observed. Using ancestral reconstruction we have expanded the existing CAR toolbox with three new thermostable CAR enzymes, providing access to the high temperature biosynthesis of aldehydes to drive new applications in biocatalysis.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Multi‐step enzyme reactions offer considerable cost and productivity benefits. Process models offer a route to understanding the complexity of these reactions, and allow for their optimization. Despite the increasing prevalence of multi‐step biotransformations, there are few examples of process models for enzyme reactions. From a toolbox of characterized enzyme parts, we demonstrate the construction of a process model for a seven enzyme, three step biotransformation using isolated enzymes. Enzymes for cofactor regeneration were employed to make this in vitro reaction economical. Good modelling practice was critical in evaluating the impact of approximations and experimental error. We show that the use and validation of process models was instrumental in realizing and removing process bottlenecks, identifying divergent behavior, and for the optimization of the entire reaction using a genetic algorithm. We validated the optimized reaction to demonstrate that complex multi‐step reactions with cofactor recycling involving at least seven enzymes can be reliably modelled and optimized.

Multi-step enzyme reactions offer considerable cost and productivity benefits. Process models offer a route to understanding the complexity of these reactions, and allow for their optimization. Despite the increasing prevalence of multi-step biotransformations, there are few examples of process models for enzyme reactions. From a toolbox of characterized enzyme parts, we demonstrate the construction of a process model for a seven enzyme, three step biotransformation using isolated enzymes. Enzymes for cofactor regeneration were employed to make this in vitro reaction economical. Good modelling practice was critical in evaluating the impact of approximations and experimental error. We show that the use and validation of process models was instrumental in realizing and removing process bottlenecks, identifying divergent behavior, and for the optimization of the entire reaction using a genetic algorithm. We validated the optimized reaction to demonstrate that complex multi-step reactions with cofactor recycling involving at least seven enzymes can be reliably modelled and optimized. Significance statementThis study examines the challenge of modeling and optimizing multi-enzyme cascades. We detail the development, testing and optimization of a deterministic model of a three enzyme cascade with four cofactor regeneration enzymes. Significantly, the model could be easily used to predict the optimal concentrations of each enzyme in order to get maximum flux through the cascade. This prediction was strongly validated experimentally. The success of our model demonstrates that robust models of systems of at least seven enzymes are † Present address: Manchester readily achievable. We highlight the importance of following good modeling practice to evaluate model quality and limitations. Examining deviations from expected behavior provided additional insight into the model and enzymes. This work provides a template for developing larger deterministic models of enzyme cascades. phosphotransferase (PAP), which allows the regeneration of ADP from AMP (24). The combination of PAP and PPK therefore allowed complete regeneration of ATP from AMP (25). In an alternative approach, adenylate kinase (which catalyzes the reversible phosphorylation of AMP by ATP) was used in place of PPK. Coupling of this enzyme with PAP pushes the equilibrium towards ATP regeneration ( Figure 1B) (26). The application of process modellingSynthetic biology has long promised the use of well-characterized parts for the rational design of new metabolic pathways (27)(28)(29). However the use of modelling to optimize these in vivo reactions is yet to be fully realized (30). In contrast, modelling of enzymes in vitro can be fairly robust and offers solutions for reaction engineering (31). This could allow the combination of enzymes, for which validated mathematical models exist, to be rapidly engineered and tested in silico (2).