A variant of flux balance analysis shows consistency with proteomic and transcriptomic data. Parsimonious enzyme usage FBA (pFBA) is used to label all metabolic genes based on simulation results. (A) pFBA classifies each gene based on its ability to contribute to the optimal growth rate predictions and its flux level. These classes include (1) essential genes; (2) pFBA optima, which includes genes that are predicted to be used for optimal growth in silico; (3) ELE, which includes genes that will increase cellular metabolic flux if used; (4) MLE, which includes genes predicted to decrease the growth rate if used; and (5) pFBA no-flux, which includes genes that cannot be used in the given growth conditions. (B) The omic data show good coverage of essential genes and the pFBA optima, and low coverage of the genes that are predicted to be non-functional. In addition, in laboratory evolution experiments, genes within these optimal states are upregulated, whereas non-functional genes are downregulated. These results support predicted optimal growth states, and suggest that laboratory-evolved strains further enhance these optimal growth states.

pFBA classes are consistent with omic data. Simulations for each growth condition were used to classify each gene according to the efficiency of its associated reaction(s). The coverage of (A) expressed genes above a statistical cutoff and (B) identified proteins were determined. More genes and proteins in the essential and pFBA optima classes were expressed than genes and proteins in the less efficient (ELE and MLE) and conditionally non-functional (pFBA no-flux) pathways. Almost half of the missing essential and pFBA optima proteins are membrane associated, hence its lower coverage. (C) In all, 99 and 98% of the active reactions associated with the essential genes and pFBA optima, respectively, are supported by the union of expressed genes and proteins. Gene expression levels are also consistent with the pFBA classes, as shown in density plots for growth on (D) glycerol and (E) lactate. (F) Pairwise comparisons of classes show the significant ordering of expression levels as Essential>pFBA optima>ELE>MLE>pFBA no-flux for growth on glycerol (upper triangle) and lactate (lower triangle) (one-sided Wilcoxon test).

The subnetwork providing optimal growth emerges in adaptively evolved strains. (A) For each growth condition, pFBA was used to classify all genes and their associated reactions. (B) Subsequently, these classifications were compared with differentially expressed genes or proteins for each growth condition (a representative portion of the metabolic network with glycerol strain proteomic data is shown). (C) A quantitative assessment was performed, in which the sum of fold-change values from all downregulated genes in each class (x axis) was plotted against the fold change sum of all upregulated genes (y axis), and then scaled by the variance of the sum of randomly selected differentially expressed genes. The cloud represents the normalized distribution of the summed upregulated and downregulated genes or proteins of randomly chosen differentially expressed genes. This analysis shows that for all data sets, genes and proteins within the essential set and the pFBA optima show more upregulation and much less downregulation than expected from randomly selected differentially expressed metabolic genes and proteins. This emergence of the optimal pathways is enhanced by the lack of upregulation and the significant downregulation within the less efficient ELE and MLE pathways and the conditionally inactive pFBA no-flux pathways. ELE, enzymatically less efficient; MLE, metabolically less efficient.

In adaptive evolution through serial passaging, E. coli evolves to a higher growth rate and biomass yield at exponential growth. Growth rates and substrate uptake rates were acquired for each strain before and after adaptive evolution, as reported earlier (Fong et al, 2005; Charusanti et al, submitted for publication). For growth on (A) glycerol, (B) lactate, and for (C) the Δpgi strain grown on glucose, the evolved strains (red) all improved their growth rate and biomass yield at exponential growth, compared with the unevolved parent strains (blue).

Proteins and genes associated with metabolic processes dominate differential expression in evolved strains. Adaptively evolved strains have hundreds of altered gene and protein expression levels covering a broad range of COGs. However, the relatively large coverage of genes and proteins in COGs associated with metabolism (red brackets) clearly suggests the importance of changes in the metabolic network. The most significantly enriched metabolic COG is ‘carbohydrate metabolism and transport.' In addition, the proteomic data suggest that the transcription and translation machinery may also have an important function in the physiological changes witnessed in the adaptively evolved strains.