K-Optforce: Strain Design Using Kinetic Information

Existing computational strain-design approaches relying solely on stoichiometry and on-off regulation ignore the effects of metabolite concentrations, enzyme activity and substrate-level enzyme regulation while identifying metabolic interventions. In this work, we implemented the recently developed k-OptForce procedure, which integrates the available kinetic descriptions of metabolic reactions with stoichiometric models, to sharpen the prediction of intervention strategies for improving the bio-production of a chemical of interest. This procedure seamlessly integrates the mechanistic detail afforded by kinetic models within a constraint-based optimization framework tractable for genome-scale models. Instead of relying on surrogate fitness functions such as biomass maximization or worst-case simulation for predicting flux re-directions, k-OptForce uses kinetic rate expressions to (re)apportion fluxes in the metabolic network. The interventions suggested by k-OptForce are comprised of both direct enzymatic parameter changes (for reactions with available kinetics) and indirect reaction flux changes (for reactions with only stoichiometric information). In some cases, additional modifications are needed to overcome the substrate-level regulations imposed by the representative kinetic model. The mechanism of action of these modifications is often subtle by alleviating substrate inhibition or draining away cofactors from competing pathways. In other cases, kinetic expressions shape flux distributions so as to favor the overproduction of the desired product requiring fewer direct interventions.

k-OptForce requires as input kinetic expressions that accurately capture the substrate-level regulation of metabolic fluxes. To this end, we constructed a kinetic model of E. coli core metabolism that satisfies the fluxomic data for wild-type and seven mutant strains by making use of the recently introduced Ensemble Modeling (EM) concepts. This model consists of 138 reactions, 93 metabolites and 60 substrate-level regulatory interactions and accounts for glycolysis/gluconeogenesis, pentose phosphate pathway, TCA cycle, major pyruvate metabolism, anaplerotic reactions and a number of reactions in other parts of the metabolism. Parameterization of the model was performed using a formal optimization algorithm that minimizes uncertainty-scaled discrepancies between model predictions and flux measurements. The predicted fluxes by the model are within the uncertainty range of experimental flux data for 78% of the reactions (with measured fluxes) for both the reference (wild-type) and seven mutant strains. The remaining flux predictions fall within three standard deviations of measured values. The predicted metabolite concentrations are also within uncertainty ranges of metabolomic data for 68% of the metabolites. Converting the EM-based parameters into a Michaelis-Menten equivalent formalism revealed that 80% of Km and kcat parameters are within one order of magnitude of the literature available values. A leave-one-out cross-validation test performed to evaluate the predictive capability of the model also showed good agreement with test experimental data.

Application of k-OptForce for overproduction of bio-chemicals recapitulated existing intervention strategies, while identifying additional and alternate manipulations for improving the yield of target chemical. k-OptForce identified key regulatory bottlenecks preventing the redirection of flux towards the target chemical, and suggested manipulations, often non-intuitive and distal to the point of regulation to overcome them. This framework paves the way for an integrated analysis of kinetic and stoichiometric models and enables elucidating system-wide metabolic interventions while capturing regulatory and kinetic effects.