Bacterial genome architecture shapes global transcriptional regulation by DNA supercoiling

Abstract

DNA supercoiling acts as a global transcriptional regulator in bacteria, that plays an important role in adapting their expression programme to environmental changes, but for which no quantitative or even qualitative regulatory model is available. Here, we focus on spatial supercoiling heterogeneities caused by the transcription process itself, which strongly contribute to this regulation mode. We propose a new mechanistic modeling of the transcription-supercoiling dynamical coupling along a genome, which allows simulating and quantitatively reproducing in vitro and in vivo transcription assays, and highlights the role of genes' local orientation in their supercoiling sensitivity. Consistently with predictions, we show that chromosomal relaxation artificially induced by gyrase inhibitors selectively activates convergent genes in several enterobacteria, while conversely, an increase in DNA supercoiling naturally selected in a long-term evolution experiment with Escherichia coli favours divergent genes. Simulations show that these global expression responses to changes in DNA supercoiling result from fundamental mechanical constraints imposed by transcription, independently from more specific regulation of each promoter. These constraints underpin a significant and predictable contribution to the complex rules by which bacteria use DNA supercoiling as a global but fine-tuned transcriptional regulator.

Figure 1.

Illustration and main components of the transcription-supercoiling coupling model. (A) Snapshot of the simulation of the stochastic binding (green arrows; the basal initiation rate k_on of each promoter is shown), elongation, and dissociation (red arrows) of a set of RNAPs along a 1D genome (here a 5-kb plasmid). (B) The SC profile is updated at each timestep, and is affected by elongating RNAPs as well as by topoisomerase activity. This level is constant between topological barriers, i.e., either elongating RNAPs (blue) or fixed proteic barriers (black). (C) The local SC level affects each promoter through an activation curve derived from thermodynamics of open complex formation, which modulates its specific strength (basal initiation rate). (D) Topoisomerases bind in a deterministic but heterogeneous way, according to the local SC level (see text).

Figure 2.

Calibration of the model on in vitro transcription experiments with plasmids. (A) The promoter activation curve (Figure 1C) is calibrated from pelE expression levels measured on purified plasmids prepared at different SC levels (4). Due to the absence of topological barriers in the plasmid, transcription-induced supercoils do not accumulate and SC levels remain constant. In this assay, this promoter from Dickeya dadantii is activated around 20-fold by negative SC. The weakly expressed promoter ampR was kept inactive in the simulations. (B) Topoisomerase activity constants are calibrated from a SC accumulation assay (22). The instantaneous initiation rate (top) was measured over time, and is reproduced by our simulations where the average plasmid SC level can be inferred (bottom). Positive supercoils first accumulate in the absence of DNA gyrase, resulting in a progressive repression of the promoter, and are then released in an exponential timecurve (red curve) reflecting gyrase activity.

Figure 3.

In vivo TSC-induced co-regulation of adjacent genes on the Escherichia coli chromosome. (A) The expression of the inducible tetP promoter (by increasing tet inducer concentration or absence of tet repressor) progressively activates a divergent promoter (tufB) (14). This effect is reproduced semi-quantitatively in our simulations by only changing the tetP basal initiation rate; small discrepancies with the data may reflect a different SC-sensitivity of the two promoters (see text). (B) The gyrA promoter is activated by the expression of the upstream isodirectional gene uidA, almost independently of the distance between the two genes (from 0 to 6 kb) (16). This observation is fully consistent with our hypothesis of fast diffusion of SC over kilobase distances. A reversed promoter activation curve was used here to account for the unusual properties of the gyrA promoter (see text and Supplementary Figure S2B). (C) Simulations (without insert) show that the gyrA activation factor is strongly dependent on the expression strength of the uidA gene.

Figure 4.

(A) ChIP-Seq data show that topoisomerase I and the DNA gyrase preferentially bind in divergent and convergent regions, respectively (43). (B) The observed distribution naturally emerges from our modeling of TSC in physiologically relevant conditions (moderate gene expression and 0.25 μM gyrase concentration, see Materials and Methods). (C) Simulations were carried on a model genome with three distinct topological domains, each carrying two active genes (yellow) in different orientations, and a central inactive gene (gray) as a regulatory probe. The heterogeneous SC levels are shown before (blue) and after (orange) relaxation by gyrase inhibition, and they give rise to the heterogeneous recruitment of topoisomerases observed in B. (D) Transcriptomics data in Escherichia coli from (9) show that genes’ response to chromosomal relaxation is tightly related to their local orientation. (E) The same is observed in new RNA-Seq data from Dickeya dadantii, in exponential as well as at transition to stationary (Supplementary Figure S4B) phase, confirming that this feature is not organism- or condition-specific. (F) Simulating the action of the antibiotics exhibits a different effect on the promoters’ initiation rates, depending on their local orientation. (G) As a result, relaxation favours convergent genes versus divergent ones, in agreement with genome-wide experimental data. (H) The average SC level (upper panel) and relative convergent/divergent foldchange due to relaxation (lower panel) was computed for a range of initial DNA gyrase concentrations and expression strengths (of all active genes), corresponding to effective waiting times of 1, 2.5 and 10 min between transcription events at each promoter, respectively. Data shown in B, C, F and G correspond to the central datapoint with moderate expression.

Figure 5.

Gene expression profiles in clones from the long-term evolution experiment with E. coli (29,30) exhibits the signature of TSC-mediated regulation. (A) 2K and 20K clones from the Ara-1 population evolved increased fitness relative to their ancestor after 2000 and 20,000 generations, respectively (data from (7)). (B) The SC level sequentially increased, owing to the selection of two SC-modifying mutations (data from (7)), first in topA (among six mutations in 2K), then in fis (among 45 mutations in 20K, including the topA mutation) (32). (C) In 2K, divergent genes are more activated than convergent ones with respect to the ancestor (P = 1.6 × 10⁻³), as predicted by TSC. Differentially expressed genes were selected with a loose threshold (see text): we indicate the number of convergent+divergent responding genes. (D) Same for the 20K clone (P = 0.016). (E) We simulated the mutation in topA by reducing the activity of topoisomerase I by 2-fold: as expected and observed in vivo, divergent genes are favored by the resulting increase in negative SC. (F) Conversely, a 2-fold increase in topoisomerase I activity favours convergent genes, in the same manner as gyrase inhibition (Figure 4G).