Biologically-derived hydrocarbons are considered to have great potential as next-generation biofuels owing to the similarity of their chemical properties to contemporary diesel and jet fuels. However， the low yield of these hydrocarbons in biotechnological production is a major obstacle for commercialization. Several genetic and process engineering approaches have been adopted to increase the yield of hydrocarbon， but a model driven approach has not been implemented so far. Here， we applied a constraint-based metabolic modeling approach in which a variable demand for alkane biosynthesis was imposed， and co-varying reactions were considered as potential targets for further engineering of an E. coli strai... More
Biologically-derived hydrocarbons are considered to have great potential as next-generation biofuels owing to the similarity of their chemical properties to contemporary diesel and jet fuels. However， the low yield of these hydrocarbons in biotechnological production is a major obstacle for commercialization. Several genetic and process engineering approaches have been adopted to increase the yield of hydrocarbon， but a model driven approach has not been implemented so far. Here， we applied a constraint-based metabolic modeling approach in which a variable demand for alkane biosynthesis was imposed， and co-varying reactions were considered as potential targets for further engineering of an E. coli strain already expressing cyanobacterial enzymes towards higher chain alkane production. The reactions that co-varied with the imposed alkane production were found to be mainly associated with the pentose phosphate pathway (PPP) and the lower half of glycolysis. An optimal modeling solution was achieved by imposing increased flux through the reaction catalyzed by glucose-6-phosphate dehydrogenase (zwf) and iteratively removing 7 reactions from the network， leading to an alkane yield of 94.2% of the theoretical maximum conversion determined by in silico analysis at a given biomass rate. To validate the in silico findings， we first performed pathway optimization of the cyanobacterial enzymes in E. coli via different dosages of genes， promoting substrate channelling through protein fusion and inducing substantial equivalent protein expression， which led to a 36-fold increase in alka(e)ne production from 2.8 mg/L to 102 mg/L. Further， engineering of E. coli based on in silico findings， including biomass constraint， led to an increase in the alka(e)ne titer to 425 mg/L (major components being 249 mg/L pentadecane and 160 mg/L heptadecene)， a 148.6-fold improvement over the initial strain， respectively; with a yield of 34.2% of the theoretical maximum. The impact of model-assisted engineering was also tested for the production of long chain fatty alcohol， another commercially important molecule sharing the same pathway while differing only at the terminal reaction， and a titer of 1506 mg/L was achieved with a yield of 86.4% of the theoretical maximum. Moreover， the model assisted engineered strains had produced 2.54 g/L and 12.5 g/L of long chain alkane and fatty alcohol， respectively， in the bioreactor under fed-batch cultivation condition. Our study demonstrated successful implementation of a combined in silico modeling approach along with the pathway and process optimization in achieving the highest reported titers of long chain hydrocarbons in E. coli.
