Bidirectional promoters in Escherichia coli: regulatory rules and implications for gene expression noise

Abstract

In prokaryotes, bidirectional promoters are pseudo-symmetrical DNA sequences that stimulate divergent transcription. Ubiquitous, and far more likely to drive messenger RNA production than directional promoters, nothing is known about their control. For example, symmetry allows bidirectional promoters to engage RNA polymerase in two possible orientations. As one binding event prevents the other, there is potential for regulation at this step. Here, we show that basal transcription, from all five tested bidirectional promoters, is too low for RNA polymerase competition. Hence, synthesis of one RNA does not impact the divergent pair. Conversely, if transcription in one direction is substantially activated, divergent RNA production can be repressed. Often, this results from RNA polymerase competition alone. Unexpectedly, this also impacts population-level gene expression noise. Specifically, if transcription is constrained, by RNA polymerase interference, cell-to-cell variation is reduced. We anticipate that our findings will help to establish rules for understanding bidirectional promoters, which have hardly been studied, in many bacteria.

Graphical Abstract

Figure 1.

Basal transcription from bidirectional promoters is too low for substantial competition between RNA polymerase molecules. (a) Bidirectional promoter regions used in this study. Schematic representation of the bidirectional promoter regions used in this study. TSS locations are those previously mapped using cappable-seq [32] and are indicated by bent arrows. Distances between divergent TSSs are shown above each schematic. Promoter -10 and -35 elements are shown as boxes and labelled. Filled and open boxes indicate elements used for forward and reverse transcription, respectively. Lighter shading highlights regions of the -10 element shared during forward and reverse transcription. (b) Transcription from bidirectional promoter regions and their derivatives in vitro. The panel shows results of in vitro transcription assays for each bidirectional promoter region and derivatives with -35 element mutations. In each case, the -35 element changes are designed to impede (denoted ‘down’) or enhance (denoted ‘up’) transcription in the indicated direction (see labels above each lane). One unit of RNA polymerase holoenzyme (NEB) was used per reaction. The RNAI transcript is generated from the plasmid DNA replication origin and serves as an internal control. (c) Quantification of forward and reverse transcription from each DNA template. Band intensities were measured and used to determine the amount of each divergent transcript relative to the RNAI control. A two-tailed paired Student’s t-test was used to calculate P, where * <.05, ** <.01, and *** <.001. Error bars show standard deviation from at least three independent replicates.(d) Promoter -35 element scores for forward and reverse transcription. For each DNA template, -35 motif scores are indicated for forward (filled boxes) or reverse (open boxes) transcription. Scores were calculated using a position weight matrix [31]; the consensus -35 element has a score of 3.5. The colour code matches that used in panel (c).

Figure 2.

SoxS induces competition between RNA polymerase molecules at the bidirectional promoter PacrA/acrR. (a) The acrAB/acrR locus. A schematic showing organization of the acrAB operon and divergent acrR gene. Each open reading frame is a block arrow. The expansion shows organization of the bidirectional promoter PacrA/acrR. TSSs are shown as bent arrows. Promoter elements for transcription in the direction of acrR and acrA are filled and open boxes, respectively. The shared portion of the -10 hexamer, used for both forward and reverse transcription, has lighter shading. Extended boxes indicate binding sites for gene regulatory proteins. The closely related factors SoxS and MarA recognise exactly the same sequence and activate transcription in the direction of acrA. The AcrR protein is a repressor. (b) Transcription from PacrA/acrR and derivatives in vitro. The panel shows results of in vitro transcription assays for each bidirectional promoter region and derivatives with -35 element mutations. In each case, the -35 element changes are designed to impede (denoted ‘down’) transcription in the indicated direction (see labels below each lane). The RNAI transcript is generated from plasmid DNA replication origin and serves as an internal control. Two units of RNA polymerase holoenzyme (NEB) were used per reaction. Where present, SoxS and MarA were used at concentrations of 0, 0.25, 0.5, 1, 2, or 3 μM. (c) Transcription from PacrA/acrR and derivatives in vivo. The bar chart shows results of β-galactosidase assays. Experiments were done in triplicate, and error bars show standard deviation. The bidirectional promoter region was cloned upstream of lacZ either in the forward orientation (to monitor acrR transcription, solid bars) or in the opposite orientation (to detect expression of acrA, open bars). Bars labelled ‘empty plasmid’ show values for promoterless lacZ in each strain. SoxS was expressed ectopically, from a low-level constitutive promoter in pJ203-SoxS, as indicated. A two-tailed paired Student’s t-test was used to calculate P, where * <.05 and ** <.01. (d) Models for control of PacrA/acrR by SoxS and AcrR. Proteins are shown as ovals and transcripts as wavy lines. Crosses indicate loss of transcription in a given direction.

Figure 3.

Competition between RNA polymerase molecules at semi-synthetic PacrA/acrR derivatives dependent on CRP. (a) Transcription from CRP regulated PacrA/acrR and derivatives in vitro. The panel shows results of in vitro transcription assays for each bidirectional promoter region and a derivative with -35 element mutations. The -35 element changes are designed to impede (denoted ‘down’) transcription in the indicated direction (see labels below each lane). The RNAI transcript is generated from the plasmid DNA replication origin and serves as an internal control. Two units of RNA polymerase holoenzyme (NEB) were used per reaction. Where present, CRP was used at concentrations of 0, 0.5, 1, 2, and 4 μM. Schematic representations of each promoter region derivative are shown below the gel images and are labelled as in prior figures. (b) CRP regulation of bidirectional promoters in vivo. Promoter sequences, with CRP binding sites in different positions, were fused to lacZ in each possible orientation to measure acrA or acrR transcription. Derivatives of constructs with -35 element mutations that reduce transcription in the acrA direction are labelled ‘acrA down’. The effect of CRP is seen by comparing expression levels in wild type (open bars) and Δcrp (speckled bars) for acrA. For acrR, the equivalent bars are solid and striped, respectively. Data are the mean of three replicates and error bars show standard deviation. Schematics show organization of the wild-type DNA fragments. The P-value was calculated as described in Fig. 2, where *** <.001.

Figure 4.

The MerR family activator SoxR directly represses divergent RNA production at the bidirectional promoter PsoxS/soxR. (a) Organization of PsoxS/soxR. The schematic shows organization of the bidirectional promoter PsoxS/soxR. Genes (not to scale) are shown as block arrows and TSSs as bent arrows. Promoter elements for forward and reverse transcription are shown as filled and open boxes, respectively. Lighter shading indicates -10 element DNA used for both forward and reverse transcription. Binding sites for transcription factors are shown as elongated boxes and are labelled. (b) Binding of purified SoxR protein to PsoxS/soxR. The panel shows results of electrophoretic mobility shift assays with either the wild-type promoter region or derivatives carrying mutations in two putative SoxR sites. The complete SoxR site is labelled in panel (a) and mutation abolishes SoxR binding. The partial SoxR site, consisting of only one half site, is located between the shared -10 hexamer and the -35 element for transcription in the direction of soxR. Mutation of this partial binding site does not alter SoxR binding. Concentrations of SoxR used were 0, 1, 2, or 4 μM. (c) SoxR activates soxS transcription, and directly represses soxR transcription, from PsoxS/soxR. Transcripts generated from the soxRS promoter by RNA polymerase holoenzyme in the presence or absence of purified SoxS or SoxR protein. Two units of RNA polymerase holoenzyme (NEB) were used per reaction. Concentrations of SoxR were 0, 0.05, 0.1, 0.2, 0.4 μM. Concentrations of SoxS were 0, 0.375, 0.75, 1.5, 3 μM. The RNAI transcript is derived from the plasmid replication origin and serves as an internal control.(d) Model for simultaneous activation and repression by SoxR. (e) Structural model of SoxR bound to PsoxS/soxR. The -10 element, used for bidirectional transcription, and the soxR TSS are highlighted.

Figure 5.

Competition between RNA polymerase molecules at PacrA/acrR impacts population-level gene expression noise. (a) Schematic representation of the acrAB/acrR locus in different strains. The E. coli genome was modified to encode PAmCherry between acrA and acrB (top). A derivative of this strain was made lacking acrR (middle) and this strain was then altered further by improving the -35 element for transcription in the direction of acrR. The cross indicates the position of the genetic scar left after deleting acrR. (b) Representative confocal microscopy images of each bacterial strain. The smaller field of view is the boxed area from the larger field of view for each strain. Cells marked with a blue or white arrow highlight the extremes of PAmCherry expression levels observed for the ΔacrR strain.(c) Distribution of fluorescence in cells measured by confocal microscopy. The ‘x’ symbol marks the mean fluorescence value per strain. Boxes indicate the lower and upper quartiles, separated by a line indicating the median. Whiskers indicate the largest and smallest data points within 1.5 times the interquartile range. Outliers are shown as individual data points. A two-tailed Student’s t-test (assuming unequal variance) was used to calculate P. (d) Fluorescence intensity as a measure of PAmCherry expression in each strain, shown as a histogram.