Comparisons between experimental data and model predictions for gene essentiality under minimal-medium conditions. In silico gene deletion studies were carried out by evaluating the effect of deleting a gene from the model. If the model was unable to generate all of the biomass components due to a gene deletion, as determined by flux balance analysis, then the mutant was considered lethal. Experiments were also conducted by testing for growth in glycerol-supplemented minimal medium using the 3,888 mutants that were viable when grown in rich medium. In total, 819 out of the 896 mutants (91%) showed growth behaviors in glycerol minimal medium in agreement with computational predictions. KO, knockout.

Model-based evaluation of experimentally essential genes involved in cofactor biosynthesis. The map shows the metabolic pathways involved in vitamin and cofactor metabolism, where biosynthesis proceeds in the downward direction and the end products are indicated below each pathway. Red and blue arrows are used to indicate reactions associated with experimentally identified essential genes in glycerol minimal medium and rich medium (1), respectively. Gray lines indicate cases in which the gene(s) associated with the corresponding reaction was experimentally nonessential. Only the experimentally essential genes are listed next to the corresponding reactions. “&” indicates that the genes form a complex. For pathways containing essential genes, a “No Gene” designation is placed next to reactions that lack a corresponding gene assignment. Gene names listed in black indicate discrepancies between the model predictions and experimental data (i.e., the model predicts a nonlethal phenotype for growth on glycerol minimal medium but the gene deletion is lethal experimentally), and genes in gray indicate agreement (both model and data predict essentiality). Genes in black next to gray reactions (coaA, coaE, aldA, and pabC) indicate that the model predicts that a gene deletion is essential whereas the corresponding mutant strain is viable. Note that unlabeled gray reactions also represent agreement of the model prediction and experimental data, as both find the associated gene(s) to be nonessential.

Glycerol- versus glucose-supplemented minimal medium essential genes. The overlap between the data set presented in this work and the recently published data for conditionally essential genes for growth on glucose-supplemented minimal medium (1) is shown. A large majority of conditionally essential genes are consistent between the glycerol and glucose data sets, with a relatively small fraction of conditionally essential genes apparently specific to one growth condition or the other. Brief descriptions of the condition-specific essential genes are provided on the left (glycerol specific) and right (glucose specific) of the Venn diagram, and a more detailed treatment is provided in the text.

Essential genes involved in arginine and proline biosynthesis. The map shows the biosynthesis pathway for the conversion of glutamate (upper left) into arginine and proline. Red arrows are used to indicate reactions associated with an experimentally identified essential gene in glycerol-supplemented minimal medium, while gray arrows indicate reactions in which the associated gene(s) is experimentally nonessential. Associated genes are listed next to their corresponding reactions (“or” indicates that the gene encodes isozymes). Gene names listed in black indicate that the model makes an incorrect prediction regarding essentiality, while those in gray indicate the model and data are in agreement. The genes proB and proA are both essential experimentally during growth on minimal medium, but the model predicts that argA, argB, argC, and argE can be used instead to produce glutamate 5-semialdehyde for proline biosynthesis.

Essential genes involved in glycerol metabolism. The pathway map in the center shows the metabolic pathways involved in the first steps of glycerol metabolism. Red and blue arrows are used to indicate reactions associated with experimentally identified essential genes in glycerol minimal medium and rich medium (1), respectively. Gray lines indicate cases in which the gene(s) associated with the corresponding reaction was experimentally nonessential. Associated genes are listed next to their corresponding reactions (“or” indicates that the genes encode isozymes, and “&” indicates that the genes form a complex). Gene names listed in black indicate that the model makes an incorrect prediction regarding essentiality, while those in gray indicate that the model and data are in agreement. The phosphorylation of dihydroxyacetone is carried out by DhaKLM coupled with PtsHI; of these five genes, only ptsI is essential during growth on minimal medium. Three enzymes can convert glycerol 3-phosphate into dihydroxyacetone phosphate, but the data indicate that glpD is the only essential gene. The model predicts that the only gene that is essential on this map is glpF, the glycerol transporter, since the conversion of glycerol into dihydroxyacetone phosphate can proceed through either of two routes. The four plots show results from quantitative RT-PCR performed to assess the expression levels of genes that encode enzymes associated with reactions on the map. Gene expression levels during growth on glucose-supplemented and glycerol-supplemented M9 minimal medium are reported relative to levels of the control gene acpP (mean ± standard error). The gene expression data indicate that the GlpK-GlpD branch is dominantly active during growth on glycerol-supplemented minimal medium, in agreement with the conditional-essentiality data.

Phylogenetic distribution of essential genes by subsystem. The phylogenetic distribution of essential subsystems (see Materials and Methods for further details) across a diverse set of bacterial species is shown. Each row of the heat map corresponds to one of the 18 essential subsystems that were considered; next to each subsystem name, in parentheses, is the number of essential genes (both experiment and model assessed) found in the subsystem in E. coli over the total number of genes found in the subsystem in E. coli. Each column represents one of the 31 representative organisms used in this analysis. The dark-gray elements indicate the presence of the subsystem within the organism, whereas light gray indicates its absence. The dendrograms display the result of hierarchical clustering using the Hamming distance metric and average linkage for subsystems (left) and organisms (top), respectively. Organism abbreviations are as follows: ATU, Agrobacterium tumefaciens; AQA, Aquifex aeolicus; BAS, B. subtilis; BOB, B. burgdorferi; BRJ, Bradyrhizobium japonicum; BRM, Brucella melitensis; BUA, Buchnera aphidicola; CAJ, Campylobacter jejuni; CAC, Campylobacter crescentus; CHT, C. trachomatis; CLA, Clostridium acetobutylicum; COG, Corynebacterium glutamicum; DER, Deinococcus radiodurans; FUN, Fusobacterium nucleatum; HAI, Haemophilus influenzae; HEP, Helicobacter pylori; LIM, Listeria monocytogenes; MYT, Mycobacterium tuberculosis; MYP, M. pneumoniae; NEG, Neisseria gonorrhoeae; NOS, Nostoc; PSA, Pseudomonas aeruginosa; RAE, Ralstonia eutropha; RIP, R. prowazekii; SHO, Shewanella oneidensis; STA, Staphylococcus aureus; STP, Streptococcus pneumoniae; SYN, Synechocystis; THM, Thermotoga maritima; TRP, T. pallidum; XYF, Xylella fastidiosa.