Application of a robust dynamic flux balance analysis framework to a wine fermentation for understanding and steering aroma formation

Various commercial S. cerevisiae yeast strains are available for use in industrial wine fermentations, each with its owncombination of growth characteristics and organoleptic (sensory) profiles. However, the metabolism underlying thesedifferences is complex and not well understood, making manipulation of these strains difficult. One tool for elucidatingstrain-to-strain differences and identifying key metabolic pathways related to flavor production is the use of genome-scale metabolic models (GSMM). Despite progress in the field, most current models either focus on aerobic systems,contain models that focus on a carbon-limited medium, and/or lack the detailed coverage of Ehrlich and amino aciddegradation metabolic pathways that have been shown experimentally to be highly correlated with aroma formation.One way to capture the power of these models is to use dynamic flux balance analysis (dFBA) to predict the fluxdistribution of all the metabolites within the cell over the course of an entire fermentation. Using this approach, it ispossible to test the predictive capability of these models by comparing predictions with experimental fermentation data.Once the models fit dynamic data, they can be used to understand differences between commercial strains andsuggest genetic modification strategies towards steering certain aroma formation in strains of interest. Here, in thiswork, we applied a robust dFBA framework to an expanded GSMM of S. cerevisiae over the course of a winefermentation. Recently, we showed that ester production during alcohol fermentation is regulated by a newlydiscovered family of alcohol acyl transferases (AATs) in S. cerevisiae. However, mysteries still persist regarding thewhether other metabolic mechanisms are responsible for ester formation. By applying our expanded GSMM of yeast,which has the most comprehensive representation aroma forming pathways, we can more accurately predict metabolicfluxes for various yeast strains. Furthermore, this model can be calibrated with experimental data to reasonablypropose genetic and process engineering strategies to improve the enological performance S. cerevisiae strains ofinterest.