Fig 1.
Summary of the reconstruction efforts for iYS854.
(A) Evolution of S. aureus genome scale metabolic reconstructions and their biomass objective function from 2005 to 2018. (B) Graphical representation of the four central objects in the S. aureus GEM; genes, reactions, metabolites, and structures. A representative mapping between all four objects along with relevant metadata are added during the reconstruction process. (C) Percentage of metabolic genes mapped to protein crystal structures and protein homology models, and distribution of metabolic subsystems per category (more details are shown in Table S4 in S2 Appendix).
Fig 2.
Break down of the novel content in iYS854 by metabolic sub-module and COG category.
(A) We compared the gene content in iYS854 to that of the four previous GEMs of S. aureus and categorized them by their metabolic sub-modules. For purposes of clarity we only show a subset of the sub-modules across three COG categories: cofactor and prosthetic group metabolism, transport, and cell wall metabolism. The color scale represents the percentage of novel genes in iYS864 with respect to previous GEMs (columns) by each metabolic sub-module (rows). We highlight the date for the most recent reference that was added in iYS854 for each metabolic sub-module (see S3 Table in S2 Appendix for more details). Genes may have different annotations in previous reconstructions (for example the staphylopine biosynthesis pathway was only uncovered in 2016). Note that “h” represents a metabolic sub-module that was added based on gene homology. (B) We compared the most recently published GEM with iYS854 and highlighted the new additions in reaction content per COG module.
Table 1.
Summary of the modifications made to the starting model.
A single instance is counted towards a metabolite even when it appears in two different subsystems.
Table 2.
Results of growth simulations for iYS864 on seven defined media.
Fig 3.
Comparison of in vivo vs. in silico gene essentiality.
(A) Contingency matrix for the comparison of in silico gene essentiality predictions of iYS854 on rich medium with in vitro observations of tn-seq mutants on TSB. The accuracy is 85.7%, which represents an increase of 10.1% with respect to the most recent model [11]. (B) The genes that fell into the category of false negatives were grouped by the biomass precursor in whose biosynthesis they participate. (C) Predictions of gene knockout on growth phenotype across defined media types. Here, we show a subset of four media types and the subset of conditionally essential genes that are not essential in at least one media type. The full data is available in S3 Table and the full cluster map is available in S1 Appendix.
Fig 4.
Condition-specific GEM validation and evaluation.
(A) Quantitative exo-metabolomics measurements (for both CDM and CDMG) were used to build two condition specific GEMs. Growth was simulated using flux balance analysis and cross-checked against experimentally observed growth rates. See S1 Appendix for comparison of predicted and measured relative growth rate across the two media types. (B) We compared the measured and simulated relative oxygen consumption as well as the relative intracellular ATP concentration between the two condition-specific GEMs (Methods). (C) Growth phenotype predictions for 28 CDM-specific mutant GEMs were compared against experimentally observed transposon mutant growth phenotypes. Two genes (ackA and gudB) were classified as false positives due to the presence of alternative pathways. (D) We simulated single reaction knockouts and compared essential reactions across conditions. The Venn diagram highlights the differences in reaction essentiality between the CDM-specific GEM and the CDMG-specific GEM.
Fig 5.
Comparison of flux distribution across two condition-specific GEMs.
(A) The reactions that were shown to significantly differ in their flux distribution (determined by the Kolmogorov-Smirnoff test) between the CDMG- and CDM-specific GEM are shown. The width of the arrows qualitatively represents the median flux value across 10,000 sampled fluxes. The blue arrows represent the flux simulation results for the CDM-specific GEM and the red arrows represent the flux simulations of the CDMG-specific GEM. The13C labelled intracellular metabolites detected by NMR in both conditions are highlighted in red (metabolic intermediates derived from extracellular glucose in CDMG) and in blue (metabolic intermediates derived from nine extracellular amino acids in CDM) [79]. Metabolites highlighted in grey are present in the extracellular medium (note that D-glucose is only present in CDMG). (B) Differential cycling of ammonium (as well as several cofactors) between the two GEMs highlights the relative contribution to the production and consumption of ammonium for all reactions utilizing or synthesizing ammonium in CDM (blue) and CDMG (red). Several metabolic processes contribute to the ammonium pool in CDM (including the glycine cleavage system) while serine dehydrogenase is the main source of ammonium in CDMG.