Constructing a High Quality Genome-Scale Metabolic Network for Streptomyces Lividans TK24

Streptomyces lividans is increasingly attracting attention as a potential industrial host for heterologous production of proteins and secondary metabolites. The current interest in S. lividans, combined with the increasing relevance of constraint-based modeling for strain improvement, urges for the creation of a high-quality genome-scale metabolic network (GSMN).

The most recent GSMN of model organism S. coelicolor— with high genomic similarity to S. lividans—is taken as a starting point for model construction. Comparative genomics is used for the conversion into S. lividans gene-reaction relationships. Reactions which lack S. lividans enzymes are removed when this did not result in dead-end reactions with expressed (RNA-seq) enzymes or loss of growth on known growth substrates did not occur. Furthermore, KEGG reaction identifiers are added to the model, and KEGG metabolite identifiers and Enzyme Commission numbers are re-evaluated.

The model is extended using genomic, metabolomic and phenomic data. From the genome, metabolic reactions catalyzed by predicted enzymes not yet included in the model are added, as well as their substrates if they are not yet included as metabolites. Furthermore, additional metabolites, detected through LC-MS, are linked to the model by addition of metabolic reactions whenever this is reasonable with respect to the in silico phenotype and literature. Biolog Phenotype MicroArrays are used for testing respiration of S. lividans TK24 on 348 substrates and subsequent gap-filling of the model. To maintain high quality, the library for gap-filling is limited to 13 selected published bacterial GSMNs. Gap-filling was performed by searching for minimal sets of reactions that restore growth on a given substrate through mixed integer linear programming, and selecting the most appropriate solution. [Research funded by EU FP7-KBBE-2013-7 StrepSynth (grant n°613877).]