Pathway Engineering Of Isobutanol-producing Bacillus Subtilis Based On Genome-scale Metabolic Network

Isobutanol-producing Bacillus subtilis was engineered via synthetic biology in thepresent work. A rational strain improvement approach was established by taking the―model construction-target prediction-experiment guidance‖art of systems metabolicengineering into consideration, which improved the isobutanol biosynthetic capacityof the engineered strains.In order to obtain the stable isobutanol-producing B. subtilis, an efficientheterologous Ehrlich pathway was constructed by putting kivd of Lactococcus lacticsand adh2of Saccharomyces cerevisiae under the control of the strong promoter P43ofB. subtilis, and integrated into the chromosome of wild-type B. subtilis168. Under themicroaerobic conditions (20%work volume,240rpm and37°C in screw-cap shakeflasks), the maximal isobutanol titer of the recombined strain BSUL02-1was up to0.93±0.12g/L. For higher isobutanol productivity, strain BSUL03was obtained byintroducing an efficient heterologous α-ketoisovalerate biosynthetic pathway, which iscomposed of P43, alsS of B. subtilis, ilvC and ilvD of Corynebacterial glutamicum,into BSUL02-1. By using20g/L glucose as carbon substrate and supplementing theglucose feeding solution to the original concentration at18h, the production of strainBSUL03reached2.62±0.19g/L in fed-batch fermentations under microaerobicconditions.For the sake of efficient strain improvement, a rational model-guided strainengineering approach was established. The genome-scale metabolic network of theisobutanol-producing B. subtilis BSUL03was for the first time constructed andanalyzed by elementary mode analysis. A total of11,342elementary modes wereobtained, and239of which were the qualified isobutanol biosynthetic modes. On thisbasis,12target genes were predicted by flux correlation and flux flexibility analysis.To verify the correction of model and prediction algorithm, targets ldh and pdhC werechosen for in silico flux distribution analysis of the corresponding suppositionalmutants. In addition, the mutants were further experimentally constructed forconfirmation. In two-stage fed-batch fermentations (aerobic-microaerobic,40%workvolume, initial carbon substrate10g/L glucose,1.6mL500g/L glucose feedingsolution was added when glucose concentration was below1g/L), the double mutant BSUL05produced5.5±0.3g/L isobutanol, which is about70%more than that of theparental strain BSUL03. Meanwhile, isobutanol yield increased by83%to0.31±0.02C-mol isobutanol/C-mol glucose (C-mol/C-mol), reaching53%of the maximaltheoretical value (0.59C-mol/C-mol). The consistency between model prediction andexperimental results demonstrates the rationality and accuracy of this model-basedapproach for target identification as well as strain optimization.Guided by model prediction, the present isobutanol-producing B. subtilisBSUL05was rationally engineered for improved productivity. By disruptingglucose-6-phosphate isomerase encoding gene pgi, overexpressing the endogenousglucose-6-phosphate dehydrogenase encoding gene zwf and Escherichia colitranshydrogenase encoding gene udhA, the concentration of intracellular redoxNADPH of the resulted strain BSUL08was increased to2.6fold compared to thecontrol BSUL05. Moreover, the redox equilibrium was well balanced. Under thetwo-stage fed-batch fermentations, isobutanol production of BSUL08was11%higherthan that of BSUL05, reaching6.12±0.53g/L. Meanwhile, isobutanol yield wasenhanced by19%to0.37C-mol isobutanol/C-mol glucose, which is63%of thetheoretical value. These results demonstrate the significance of model-based targetprediction to guide the rational metabolic engineering of the current strains. Moreover,it emphasized the importance of intracellular redox for isobutanol biosynthesis.