Purpose: The objective of this project was to construct GEMs for five Salmonella strains and to analyze differentiating metabolic capabilities through model predictions.
Methods: Salmonella Abaetetuba str. ATCC 35640, Salmonella Enteritidis str. P125109, S. 4,[5],12:i:- str. CVM23701, Salmonella Typhimurium str. LT2 and str. UK1 were chosen for analysis. The semi-automated resource KBase was used to generate draft GEMs. In silico nutrient utilization predictions for individual carbon sources under aerobic and anaerobic conditions were conducted, and compared to in vitro data, leading to refinement and improvement in the GEMs predictive accuracy.
Results: Each GEM contained >1,250 metabolites participating in >1,330 reactions, and 1252 reactions are shared in common for all strains. There are 67 reactions unique to the GEM of str. LT2, 13 unique to the GEM of str. CVM23701 and 2 unique to the GEM of str. UK1 compared to the rest. Experimental data of str. LT2, str. UK1 and str. ATCC 35640 showed that from the 95 carbon sources tested in vitro, 72 were present in all three GEMs thus allowing validation of the models predictive accuracy, and the agreements between in silico and in vitro results were >90% for carbon utilization thus permitting simulation of multi-nutrient host niches for all Salmonella strains.
Significance: Although there are multiple GEMs built for foodborne pathogens, prior to this study there was only one GEM for a Salmonella strain. The high predictive accuracy for these Salmonella GEMs represents new tools for detection, control and prevention of Salmonella spp. in food and in hosts.