Quantitative contribution of the spacer length in the supercoiling-sensitivity of bacterial promoters

Abstract DNA supercoiling acts as a global transcriptional regulator in bacteria, but the promoter sequence or structural determinants controlling its effect remain unclear. It was previously proposed to modulate the torsional angle between the −10 and −35 hexamers, and thereby regulate the formation of the closed-complex depending on the length of the ‘spacer’ between them. Here, we develop a thermodynamic model of this notion based on DNA elasticity, providing quantitative and parameter-free predictions of the relative activation of promoters containing a short versus long spacer when the DNA supercoiling level is varied. The model is tested through an analysis of in vitro and in vivo expression assays of mutant promoters with variable spacer lengths, confirming its accuracy for spacers ranging from 15 to 19 nucleotides, except those of 16 nucleotides where other regulatory mechanisms likely overcome the effect of this specific step. An analysis at the whole-genome scale in Escherichia coli then demonstrates a significant effect of the spacer length on the genomic expression after transient or inheritable superhelical variations, validating the model’s predictions. Altogether, this study shows an example of mechanical constraints associated to promoter binding by RNA Polymerase underpinning a basal and global regulatory mechanism.


Genome-wide analyses of spacer responses to supercoiling variations in distant bacteria
Transcriptomic data were collected from the literature (Supplementary Tab. S2). Since curated promoter maps were not available, genome-wide TSS maps and associated genes were retrieved from literature (Supplementary Tab. S2), and a scan for promoter motifs upstream of TSSs was conducted with bTSSfinder [1]. Only σ70-dependent promoters were retained for all organisms, except for the cyanobacterium S. elongatus, for which only σA-dependent promoters were kept (primary σ factor), and classified depending on their spacer length and response to the investigated condition. The thresholds for statistical selection procedures are indicated in Supplementary Tab. S2, and were adjusted to generate subsets of act/rep genes of sizes comparable among the different data sets, while having enough statistical power for the analysis. The relation between promoter activation and spacer length was then quantified by a Student's t-test between activated and repressed promoters ( Supplementary Fig. S4), such as in Fig. 4 and 5. All error bars shown are 95% confidence intervals.
To compute the contribution of the spacer sequence in the model (Discussion), we included the sequence-dependence of two parameters of Eq. 1, the twist angle (α 0 ) and the torsional stiffness (k θ ), based on the parameters of the rigid base-pair model inferred from high-resolution crystallographic structures of DNA oligomers [2] and implemented in TwistDNA [3]. Based on these parameters, Eq. 1 was modified to include the total angle between -35/-10 binding sites and the total spacer stiffness, resulting in the following: where σ is the supercoiling level, s is the promoter spacer sequence, k θ (s) is the DNA spacer twist stiffness estimated with ThreaDNA [3], θ P = 543 is the optimal angle between -35 and -10 sites for RNAP binding, as assumed in the main text (with most predictions given in Discussion being independent of its value), θ 0 (s) is the effective angle estimated with ThreaDNA [3] which depends on base composition, α 0 = 34°is the average twist angle between adjacent nucleotides, and k θ is the average (base-pair step) twist stiffness. Then, for each E. coli σ70-dependent promoter, the response to DNA relaxation was predicted starting from a level σ = −0.06, with a relaxation magnitude ∆σ = 0.03 ( Supplementary Fig. S7).

Supplementary figures
Supplementary Figure  In proteobacteria and M. pneumoniae, only σ70 promoters were considered, whereas only σA promoters were considered for S. elongatus. A schematic phylogeny is depicted above.
Supplementary Figure S5: Distribution of spacer lengths at E. coli σ70-dependent promoters, based on the Ecocyc database [7].
Supplementary Figure S7: High-copy-number vector (pUC18 derivative) containing a multiple cloning site upstream of the luc reporter gene, followed by a rrnB terminator and a cat gene conferring chloramphenicol resistance.

Laboratory collection
Supplementary Mycoplasma pneumoniae novobiocin [16] 0.1 0 relaxation [16] Supplementary Table S2: Compilation of investigated species, conditions and references. The thresholds for statistical selection procedures are indicated, and were adjusted to generate subsets of activated/repressed genes of sizes comparable among the different data sets. For a threshold of 0.4 on the log2(fold-change), genes are considered activated for a log2(FC) > 0.4, repressed for a log2(FC) < -0.4, and not significantly affected for a log2(FC) comprised between -0.4 and 0.4. The correlation* condition from S. elongatus corresponds to the phasing of gene expression in the SC circadian oscillation and provides an indirect proxy of gene response to SC relaxation [14].