Model-Driven Design and Analysis of Synthetic Auxotrophs and Their Predicted Metabolite Cross-Feeding in Co-Culture

Synthetic microbial communities have shown promise as a system for improving the yield of chemical compounds of interest. Furthermore, studies have shown that nascent communities can be constructed by pairing microbial knockout strains that exhibit synthetic auxotrophy. Here, we used constraints-based reconstruction and analysis (COBRA) techniques and the most recent genome-scale metabolic model (M-model) of E. coli-K12 MG1655 (iJO1366) to design co-cultures of synthetic auxotrophs that will require high strain production of specific cross-feeding metabolites in order for the community to grow. This was first done by designing and implementing OptAux, an algorithm that solves for reaction knockouts which will maximize the minimum uptake requirement of the targeted metabolite in order for the strain to grow. For the 166 exchange metabolites optimized in simulated aerobic, glucose minimal media conditions, 73 single, double, and triple knockout strain designs were found capable of producing auxotrophs. These auxotrophs were categorized as a major subsystem elimination (MSE) knockouts, representing 42 designs,  if >5 metabolites were capable of restoring growth. The algorithm accomplished these design by inhibiting a major biosynthetic pathway such as the citric acid cycle, nucleotide synthesis, or amino acid synthesis pathways and generally showed a trade off between number of metabolites capable of restoring growth and metabolite uptake requirement. The remaining 31 designs were catagorized as essential biomass component (EBC) elimination knockout meaning <5 metabolites were capable of restoring growth. The 31 EBC designs were further categorized into 16 specific auxotroph designs (only 1 metabolite can restore growth) and 26 non-specific auxtroph designs (2-5 metabolites can restore growth). The EBC designs showed high restorative metabolite specificity but low uptake requirement due to the models inhibility to synthesize a single biomass component. By simulating all possible auxotroph strain pairings using community M-models as well as a model community model of metabolism and expression (ME-model), there were 42 MSE vs specific EBC pairings and 39 MSE vs MSE pairings predicted to cross-feed high fluxes of only one metabolite to each strain (0.704-7.673 and 0.062-0.295 mmol*gDW^-1*hr^-1 of metabolite cross-fed to the MSE and specific EBC strain, respectively at growth rate of ~0.65 hr^-1) . These results suggest that by optimizing these strain pairs via adaptive laboratory evolution, the system could be evolved to cross-feed L-Lactate, Succinate, L-Alanine, L-Glutamate, 2-Oxoglutarate, Glycoxylate, or Uridine at high fluxes. Therefore, this method could be implemented as a possible means to create to design production strains for metabolites that are not easily growth-coupled.