Genome-scale reconstruction of the sigma factor network in Escherichia coli: topology and functional states

Abstract

Background: At the beginning of the transcription process, the RNA polymerase (RNAP) core enzyme requires a σ-factor to recognize the genomic location at which the process initiates. Although the crucial role of σ-factors has long been appreciated and characterized for many individual promoters, we do not yet have a genome-scale assessment of their function.

Results: Using multiple genome-scale measurements, we elucidated the network of σ-factor and promoter interactions in Escherichia coli. The reconstructed network includes 4,724 σ-factor-specific promoters corresponding to transcription units (TUs), representing an increase of more than 300% over what has been previously reported. The reconstructed network was used to investigate competition between alternative σ-factors (the σ70 and σ38 regulons), confirming the competition model of σ substitution and negative regulation by alternative σ-factors. Comparison with σ-factor binding in Klebsiella pneumoniae showed that transcriptional regulation of conserved genes in closely related species is unexpectedly divergent.

Conclusions: The reconstructed network reveals the regulatory complexity of the promoter architecture in prokaryotic genomes, and opens a path to the direct determination of the systems biology of their transcriptional regulatory networks.

Figure 1

Molecular basis of transcription and a reconstruction of σ-factor-transcription unit gene (σ-TUG) network from multi-omic experimental datasets. (a) Diagram shows bacterial transcription process by an RNA polymerase (RNAP) core enzyme and an associated σ-factor. (b) Four-step process of multi-omic data integration to reconstruct the σ-TUG network. First, we identified RNAP-binding regions (RNAP map) and σ-factor binding regions (σ map) from RpoB and σ-factor chromatin immunoprecipitation and microarray (ChIP-chip) data (the missing σ²⁴ binding information was taken from a public database [6]), resulting in the genome-wide holoenzyme binding map (Eσ map). The Eσ map was then combined with experimental transcription start site (TSS) information (TSS map), resulting in he strand-specific promoter map (P-map), which was integrated with previously reported TU information [7], resulting in the σ-network. With this σ-network, we then performed further analysis, such as network reprogramming, motif analysis, promoter overlapping, and alternative TSS usage. Subfigure I: IOPR, intensively overlapped promoter region; OPR, overlapped promoter region; SPR, single promoter region; Orphan, orphan promoter region. Subfigure III and IV: green and brown circles represent σ⁷⁰ and σ³⁸, yellow circles represent TUs, and red dots represent genes. Edges show regulatory interactions between elements. (c) Datasets used for σ-TUG network reconstruction: ChIP-chip dataset with RNAP and six σ-factors, and the TSS dataset. The TSS dataset for exponential phase was taken from a previous study [9].TSS subpanel: exp, exponential phase; stat, stationary phase; heat, heat shock; gln, alternative nitrogen source with glutamine. (d) Magnified examples of rpoD (left panel, genomic region ranging from 3,196 to 3,214 kbp), fecI and fecRAB (right panel, genomic region ranging from 4,494 to 4,517 kbp).

Figure 2

Properties of the reconstructed σ-factor network in Escherichia coli. (a) Extensive overlapping between σ-factor binding sites. For each σ-factor, σ⁷⁰, σ³⁸, σ⁵⁴, σ³², σ²⁸, σ²⁴, and σ¹⁹, we identified 1,643, 903, 180, 312, 65, 51, and 7 binding regions, respectively. The number of binding regions overlapping between any two σ-factors is shown. For instance, 805 binding regions that were bound by both σ⁷⁰ and σ³⁸ were identified. (b) Number of promoters bound by multiple σ-factors showed a complex overlap between different σ-factors, indicating complicated alternative σ-factor usage. (c) A regulatory network between σ-factors in E. coli, in which σ⁷⁰ and σ³⁸ regulate expression of most of the seven σ-factors; σ⁷⁰ and σ²⁴ auto-regulate themselves. (d) Reconstruction of a three-layered network of σ-factors, transcription units (TUs), and genes. This network shows that many transcription start sites (TSSs) are shared by multiple σ-factors, suggesting possible competition between σ-factors for promoter binding. (e) Examples of thrLABC and hypBCDE-fhlA transcription units that are differently regulated by multiple σ-factors, and result in different TUs containing different sets of genes. For instance, TU001 is regulated by σ⁷⁰ and contains four genes, thrLABC, while TU0005 is regulated by σ³⁸ and had only two genes, thrB and thrC.

Figure 3

Competition between σ⁷⁰ and σ³⁸ in overlapping promoter regions. (a) Recruitment of RNA polymerase (RNAP) core enzyme to promoters upstream of 1,139 σ³⁸-specific genes was recovered when rpoS was knocked out. RNAP binding intensity on the y-axis was the chromatin immunoprecipitation and microarray (ChIP-chip), intensity; the three red lines represent the first, second, and third quantiles. (b) Comparison of transcriptional expression of genes in wild type (WT) and ΔrpoS strains. Of 1,139 genes with σ³⁸-specific promoters,178 had up-regulated transcription (red background) and 291 had down-regulated transcription (blue background). (c) Expression level of σ⁷⁰ and σ³⁸ was measured at both th transcriptional and translational levels. The amount of σ⁷⁰ was abundant in exponential and stationary phase, and so it was absent in rpoS. (d) After rpoS knock-out, up-regulated genes were more strongly bound by σ⁷⁰ than down-regulated genes.

Figure 4

Conservation and divergence in transcriptional regulation by σ-factors. (a) Clustering σ-factor binding patterns revealed conserved and divergent transcriptional regulation of 2,876 orthologous genes. (b)crp is regulated by σ⁷⁰ and σ³⁸ in both species, showing regulation conservation. (c) In Esherichia coli, cutA is a part of the dcuA-cutA-dipZ transcription unit (TU) and is regulated by σ⁷⁰ and σ³⁸, while cutA in Klebsiella pneumoniae is the first gene in its TU, and is directly bound by σ⁷⁰. (d) In K. pneumoniae, panD is a part of the panBCD TU, which is regulated by σ⁷⁰. However, in E. coli, panD is separated from panBC by yadD, making another distinct TU. These two TUs are both regulated by σ⁷⁰. (e) A genomic region containing ydeA and marC in both species was inverted, and this genomic inversion was accompanied by a transcription regulation switch between σ⁷⁰ and σ³⁸.