UC Berkeley

Directed evolution of E. coli for improved small-molecule production requires a combination of rational design and high-throughput screening technologies. Rational design-based directed evolution schemes use structural analyses and metabolic models to help identify targets for mutagenesis, thus improving the likelihood of identifying the desired phenotype. We used a strictly rational design-based approach to re-engineer cytochrome P450_BM3 for epoxidation of amorphadiene, developing a novel route for production of the anti-malarial compound artemisinin. A model structure of the lowest energy transition state complex for amorphadiene in the P450_BM3 active site was created using ROSETTA-based energy minimization. The resulting enzyme variant produced artemisinic-11S,12-epoxide at titers greater than 250 mg*l^-1. Continued attempts to use ROSETTA and to either improve P450_BM3 epoxidase activity or introduce hydroxylase activity, however, proved unsuccessful. In the absence of a high-throughput screening approach, further improvement of the P450-based production system would be difficult.

As with most small-molecules, there exists no known high-throughput screen for artemisinic-11S-12-epoxide, amorphadiene, or any structurally-related compound. We hypothesized that a generalized method for high-throughput screen or selection design could be based on transcription factor-promoter pairs responding to the target small-molecule. Transcription factors have long been used to construct whole-cell biosensors for the detection of environmental small-molecule pollutants¹, but the work has remained largely un-translated toward screen development. While no known transcription factor binds artemisinic epoxide, a putative transcription factor-promoter pair responsive to 1-butanol, a biofuel molecule of interest in our laboratory, was recently reported². The transcription factor, BmoR, and its cognate promoter, P_BMO, were used to build a short-chain alcohol biosensor for use as a genetic screen or selection. Following optimization of expression temperature, promoter, and reporter 5'-untranslated region, among other parameters, the BmoR-P_BMO system was shown to provide robust detection of 1-butanol in an E. coli host. The biosensor transfer function - relating input alcohol concentration to output fluorescent signal - was derived for 1-butanol and structurally related alcohols using the Hill Equation. The biosensor exhibited a linear response between 100 µM and 40 mM 1-butanol, and a dynamic range of over 8000 GFP/OD600 units. A 700 µM difference in 1-butanol concentration could be detected at 95% confidence. By replacing the GFP reporter with TetA, a tetracycline transporter, a 1-butanol selection was constructed; E. coli harboring the TetA-based biosensor exhibited 1-butanol dependent growth in the presence of tetracycline up to 40 mM exogenously added 1-butanol.

Demonstration of the biosensor in various high-throughput screening and selection applications first required construction of a 1-butanol production host. Studies have reporter 1-butanol production in E. coli through heterologous expression of either the C. acetobutylicum 1-butanol biosynthetic pathway³, or a 2-keto acid-based pathway composed of a L. lactis 2-keto acid decarboxylase, KivD, and the S. cerevisiae alcohol reductase, ADH6⁴. In our hands, the C. acetobutylicum pathway proved non-robust and yielded low titers. In contrast, high-titer production of user-defined 2-keto acid derived alcohols was achieved by introduction of a ΔilvDAYC knockout in E. coli and expression of KivD and ADH6. The engineered strain is auxotrophic for 2-keto acids, and 1-butanol was produced by supplementing the growth medium with 2-oxopentanoate. A liquid culture screen was demonstrated using a 960-member KivD and ADH6 ribosome binding site library. Using the TetA-based biosensor, a strict cut-off between analyte 1-butanol concentration and biosensor output was observed. The assay led to the identification of a variant 2-keto acid-based alcohol production pathway exhibiting an approximately 20% increase in specific 1-butanol productivity.

Attempts to engineer concomitant 1-butanol production and selection in E. coli proved difficult. Both production and detection pathways functioned robustly when individually expressed in engineered E. coli; however, concomitant production and detection resulted in increased plasmid instability and cell death. We conclude by providing an analysis of observed cell stresses, generating negative 1-butanol selective pressures, and outline future strategies that can be used to address these hurdles.