Model-based media selection to minimize the cost of metabolic cooperation in microbial ecosystems

Motivation: Simple forms of mutualism between microorganisms are widespread in nature. Nevertheless, the role played by the environmental nutrient composition in mediating cross-feeding in microbial ecosystems is still poorly understood. Results: Here, we use mixed-integer bilevel linear programming to investigate the cost of sharing metabolic resources in microbial communities. The algorithm infers an optimal combination of nutrients that can selectively sustain synergistic growth for a pair of species and guarantees minimum cost of cross-fed metabolites. To test model-based predictions, we selected a pair of Escherichia coli single gene knockouts auxotrophic, respectively, for arginine and leucine: Delta argB and Delta leuB and we experimentally verified that model-predicted medium composition significantly favors mutualism. Moreover, mass spectrometry profiling of exchanged metabolites confirmed the predicted cross-fed metabolites, supporting our constraint based modeling approach as a promising tool for engineering microbial consortia.