Principles of transcriptional regulation and evolution of the metabolic system in E. coli

Abstract

Organisms must adapt to make optimal use of the metabolic system in response to environmental changes. In the long-term, this involves evolution of the genomic repertoire of enzymes; in the short-term, transcriptional control ensures that appropriate enzymes are expressed in response to transitory extracellular conditions. Unicellular organisms are particularly susceptible to environmental changes; however, genome-scale impact of these modulatory effects has not been explored so far in bacteria. Here, we integrate genome-scale data to investigate the evolutionary trends and transcriptional control of metabolism in Escherichia coli K12. Globally, the regulatory system is organized in a clear hierarchy of general and specific transcription factors (TFs) that control differing ranges of metabolic functions. Further, catabolic, anabolic, and central metabolic pathways are targeted by distinct combinations of these TFs. Locally, enzymes catalyzing sequential reactions in a metabolic pathway are co-regulated by the same TFs. Regulation is more complex at junctions: General TFs control the overall activity of all connecting reactions, whereas specific TFs control individual enzymes. Divergent junctions play a special role in delineating metabolic pathways and decouple the regulation of incoming and outgoing reactions. We find little evidence for differential usage of isozymes, which are generally co-expressed in similar conditions, and thus are likely to reinforce the metabolic system through redundancy. Finally, we show that enzymes controlled by the same TFs have a strong tendency to co-evolve, suggesting a significant constraint to maintain similar regulatory regimes during evolution. Catabolic, anabolic, and central energy pathways evolve differently, emphasizing the role of the environment in shaping the metabolic system. Many of the observations also occur in yeast, and our findings may apply across large evolutionary distances.

Figure 1.

Regulatory targets of general and specific TFs. Schematic representation of transcriptional regulation of E. coli small-molecule metabolism displaying general TFs (red circles), specific TFs (blue), and target enzyme genes (black). TFs are labeled with gene names, and enzymes are grouped according to their COG annotations. Regulatory interactions are shown as lines directed from general TF to functionally enriched target (red), general TF to non-enriched function (gray), and specific TF to target (blue).

Figure 2.

Numbers of TFs controlling metabolic enzymes. (A) Histogram of numbers of TFs regulating all, catabolic, anabolic, and central energy enzymes. (*) Overrepresented groups; (§) underrepresented groups. (B) Box plots of numbers of general and specific TFs targeting different classes of enzymes. (C) Histogram of numbers of enzymes regulated by general TFs only, specific TFs only, and combinations of general and specific TFs.

Figure 3.

Configurations for neighboring enzymatic reactions. Example reactions are given for each configuration.

Figure 4.

Co-regulation and coexpression of metabolic enzyme pairs. (A) Histogram of numbers of pairs of enzymes that are co-regulated by identical sets of TFs in the regulatory network. (B) Box plot of distributions of Pearson correlation coefficients for gene expression profiles of enzyme pairs. The horizontal dashed line displays the median correlation for all pairs of enzymes in the metabolic network.

Figure 5.

Coexpression of enzyme pairs at different levels of separation in the metabolic network. Box plots of Pearson correlation coefficients for gene expression profiles between enzyme pairs separated by different numbers of (A) linear metabolites, (B) convergent junctions, and (C) divergent junctions. There is an upper limit of three intermediate convergent junctions in the metabolic network. The horizontal dashed line displays the median correlation for all pairs of enzymes.

Figure 6.

Conservation and coexpression of metabolic enzymes across 380 bacterial genomes. (A) Box plot of proportion of genomes containing orthologs of catabolic, anabolic, and central energy enzymes. (B) Scatterplot between the Pearson correlation coefficients measuring coexpression and Phi correlation measuring coevolution of enzyme pairs. Data points are shaded according to a normalized proportion of observations in the data set, with darker shades representing higher proportions. (Inset) The density distribution displays the mutual information between the two sets of correlations for the actual data and random expectation from 1000 simulations.