Abstract
Background: Despite the availability of numerous complete genome sequences from E. coli strains, published genome-scale metabolic models exist only for two commensal E. coli strains. These models have proven useful for many applications, such as engineering strains for desired product formation, and we sought to explore how constructing and evaluating additional metabolic models for E. coli strains could enhance these efforts.Results: We used the genomic information from 16 E. coli strains to generate an E. coli pangenome metabolic network by evaluating their collective 76,990 ORFs. Each of these ORFs was assigned to one of 17,647 ortholog groups including ORFs associated with reactions in the most recent metabolic model for E. coli K-12. For orthologous groups that contain an ORF already represented in the MG1655 model, the gene to protein to reaction associations represented in this model could then be easily propagated to other E. coli strain models. All remaining orthologous groups were evaluated to see if new metabolic reactions could be added to generate a pangenome-scale metabolic model (iEco1712_pan). The pangenome model included reactions from a metabolic model update for E. coli K-12 MG1655 (iEco1339_MG1655) and enabled development of five additional strain-specific genome-scale metabolic models. These additional models include a second K-12 strain (iEco1335_W3110) and four pathogenic strains (two enterohemorrhagic E. coli O157:H7 and two uropathogens). When compared to the E. coli K-12 models, the metabolic models for the enterohemorrhagic (iEco1344_EDL933 and iEco1345_Sakai) and uropathogenic strains (iEco1288_CFT073 and iEco1301_UTI89) contained numerous lineage-specific gene and reaction differences. All six E. coli models were evaluated by comparing model predictions to carbon source utilization measurements under aerobic and anaerobic conditions, and to batch growth profiles in minimal media with 0.2% (w/v) glucose. An ancestral genome-scale metabolic model based on conserved ortholog groups in all 16 E. coli genomes was also constructed, reflecting the conserved ancestral core of E. coli metabolism (iEco1053_core). Comparative analysis of all six strain-specific E. coli models revealed that some of the pathogenic E. coli strains possess reactions in their metabolic networks enabling higher biomass yields on glucose. Finally the lineage-specific metabolic traits were compared to the ancestral core model predictions to derive new insight into the evolution of metabolism within this species.Conclusion: Our findings demonstrate that a pangenome-scale metabolic model can be used to rapidly construct additional E. coli strain-specific models, and that quantitative models of different strains of E. coli can accurately predict strain-specific phenotypes. Such pangenome and strain-specific models can be further used to engineer metabolic phenotypes of interest, such as designing new industrial E. coli strains.
Original language | English (US) |
---|---|
Article number | 182 |
Journal | BMC Systems Biology |
Volume | 5 |
DOIs | |
State | Published - Nov 1 2011 |
Bibliographical note
Funding Information:We are grateful and would like to acknowledge the following for provision of bacterial strains used in this study: Prof. Scott Hultgren (Dept of Molecular Microbiology, Washington University School of Medicine, St. Louis) for provision of E. coli strain UTI89, Prof. Patricia Kiley (Dept. of Medical Microbiology, University of Wisconsin-Madison) for provision of E. coli K-12 strain MG1655, and Prof. Diana Downs (Dept. of Microbiology, University of Wisconsin-Madison) for provision of Salmonella LT2. We would also like to thank the Great Lakes Bioenergy Research Center for use of the UPLC instrument for analysis. This work was funded by the US Department of Energy Genomics:GTL and SciDAC Programs (DE-FG02-04ER25627) for the BACTER Institute Post-Doctoral Fellowship (D.J.B) and also the National Library of Medicine, National Institutes of Health, Grant No. 5T15LM007359 to the Computation and Informatics in Biology and Medicine Training Program for a Post-Doctoral Traineeship (D.J.B) and in part by NIH grant #NIH/GNM GM62994-02. Finally, we would like to thank Dr. Eric Cabot and Christopher Tervo for assistance in generation of the SBML model files.