Application of Algorithms to Identify and Produce Precursors to Commercial Chemicals

Microbes have been engineered to produce a variety of chemicals, including biofuels, commodity chemicals, and therapeutics. Production can be enhanced by connecting biosynthesis pathways to host metabolism and optimizing the pathway’s expression. However, another important step in engineering microbes to produce desired chemicals is to increase production of their chemical precursors. In this work, we developed optimization-based approaches to trace carbon flux through genome-scale networks. The first method (MapMaker) predicts which substrates and products in a given reaction exchange carbon atoms (with >97% accuracy). MapMaker results were then used by a second algorithm (PathTracer) to identify flux-balanced paths (or cycles) between two metabolites, where each step in a path involves an exchange of carbon atoms between a substrate and product. These methods were used to evaluate what precursors are needed to synthesize a variety of commercial chemicals, and then we designed and engineered strains to produce one of these precursors—pyruvate.

We systematically evaluated what potential non-native products could be produced by Escherichia coli (using both native and heterologous pathways) and then evaluated their distance (in terms of metabolic reactions steps) from central metabolic precursors. Using a genome-scale metabolic model of E. coli and a set of potential heterologous reactions (from the KEGG database), ~1,800 non-native products could potentially be produced in E. coli using heterologous enzymes. Of these 284 have reported commercial applications. The set of 284 chemicals were subsequently analyzed to identify how many heterologous reactions were needed to enable their production and whether they were within five reaction steps of a central metabolic precursor. This analysis identified that of the six central metabolic precursors considered, pyruvate was the closest precursor to the most non-native commercial products.

Since pyruvate has industrial applications and is within 5 reactions steps of many commercial non-native products, we sought to develop a strain of E. coli that could produce pyruvate at high yields. A high-yield pyruvate producing strain could be further engineered to produce a variety of other commerical chemicals. Guided by a genome-scale metabolic model of E. coli, we identified different strategies for enhancing production of pyruvate from glucose. The targeted gene deletions minimize acetyl-CoA production, undesired product (acetate and lactate) formation, and NAD(P)H formation. We constructed a number of strains that achieved yields between ~69% and ~95% of the theoretical yields. These results illustrate how computational models can be used to find precursor-based strategies and identify genetic modifications to enhance precursor production. Current efforts are focused on engineering the pyruvate production strains to produce other native and non-native products, as well as developing other precursor production strains.