Computational analysis of phenotypic space in heterologous polyketide biosynthesis--applications to Escherichia coli, Bacillus subtilis, and Saccharomyces cerevisiae.

Polyketides represent a class of natural product small molecules with an impressive range of medicinal activities. In order to improve access to therapeutic polyketide compounds, heterologous metabolic engineering has been applied to transfer polyketide genetic pathways from often fastidious native hosts to more industrially-amenable heterologous hosts such as Escherichia coli, Saccharomyces cerevisiae, or Streptomyces coelicolor. Efforts thus far have resulted in titers either inferior to the native host and significantly below the theoretical yield, emphasizing the need to computationally investigate and engineer the interaction between native and heterologous metabolism for the improved production of heterologous polyketide compounds. In this work, we applied flux balance analysis on genome-scale models to simulate cellular metabolism and 6-deoxyerythronolide B (the cyclized polyketide precursor to erythromycin) production in three common heterologous hosts (E. coli, Bacillus subtilis, and S. cerevisiae) under a variety of carbon-source and medium compositions. We then undertook minimization of metabolic adjustment optimization to identify single and double gene-knockouts that resulted in increased polyketide production while maintaining cellular growth. For the production of 6-deoxyerythronolide B, the results suggest B. subtilis and E. coli are better heterologous hosts when compared to S. cerevisiae and that several single and multiple gene-knockout mutants are computationally predicted to improve specific production, in some cases, over 25-fold.