Modulation of extracytoplasmic function (ECF) sigma factor promoter selectivity by spacer region sequence.

The ability of bacteria to adapt to stress depends on the conditional expression of specific sets of genes. Bacillus subtilis encodes seven extracytoplasmic function (ECF) sigma (σ) factors that regulate functions important for survival under conditions eliciting cell envelope stress. Of these, four have been studied in detail: σM, σW, σX and σV. These four σ factors recognize overlapping sets of promoters, although the sequences that determine this overlapping recognition are incompletely understood. A major role in promoter selectivity has been ascribed to the core -10 and -35 promoter elements. Here, we demonstrate that a homopolymeric T-tract motif, proximal to the -35 element, functions in combination with the core promoter sequences to determine selectivity for ECF sigma factors. This motif is most critical for promoter activation by σV, and contributes variably to activation by σM, σX and σW. We propose that this motif, which is a feature of the deduced promoter consensus for a subset of ECF σ factors from many species, imparts intrinsic DNA curvature to influence promoter activity. The differential effect of this region among ECF σ factors thereby provides a mechanism to modulate the nature and extent of regulon overlap.

INTRODUCTION

Transcription is the first step in gene expression and can be regulated at multiple levels (1,2). The global transcriptional profile of the cell results from the activity of RNA polymerase (RNAP) as determined by many transcription factors that affect initiation, elongation, and termination of RNA synthesis. In bacteria, a key initial step is the recognition of the promoter by formation of sequence specific DNA–protein interactions between RNAP, and in particular the σ subunit, and promoter DNA (3–5). The widespread σ70 family of initiation factors recognize conserved DNA sequences centered near −35 and −10 bp upstream of the transcription start site that are recognized by σ regions 4 and 2, respectively (3,5,6). Bacteria often contain multiple σ factors, each recognizing distinct promoter sites, thereby allowing the cell to express specific genes in response to changing growth conditions (7,8). The genes controlled by each σ factor are defined as its regulon.

The expression profile of individual genes is often determined by the combinatorial action of multiple regulators (2,9). For many operons, transcription can initiate from more than one promoter element, depending on environmental conditions. Multiple promoter sites are often revealed by the presence of multiple transcription start sites, including some that maybe within genes or within operons, and the resulting transcripts may differ in their stability or the efficiency or site of translation initiation (10). In addition, it is possible for two or more σ factors to recognize the same or largely overlapping −35 and −10 core promoter elements and this may lead to initiation from the same position (11). We refer to this latter phenomenon, where multiple holoenzymes can initiate from the same promoter and from the same start site(s), as overlapping promoter recognition.

One of the best studied cases of overlapping promoter recognition by σ factors occurs in Escherichia coli between the primary housekeeping σ factor, σ70 and the alternative σ factor, σS. These two closely related σ factors recognize very similar core promoter elements, although σS is able to recognize less conserved −35 elements (12), and its activity is stimulated by A/T stretches at positions −29 to −25 and −22 to −18 (11). Moreover, σS is able to better tolerate sub-optimal spacer lengths, whereas σ70 strongly prefers promoters with 17 bp spacing (12). A cytosine at position −13 directly upstream of the −10 element is one characteristic of σS-dependent promoters and, conversely, counter-selected in σ70-dependent promoters (13–16). In addition, these two σ factors differ in their response to the presence and location of upstream stimulatory elements (UP elements) recognized by the RNAP α subunit: a distal UP-element site is beneficial for σS recognition, whereas a proximal UP-element favors σ70 selectivity. Most likely, a combination of several of these factors ultimately determines the σ specificity of each promoter, with many promoter elements being well recognized by both σ factors. Overlapping promoter recognition has also been reported for σ70 and the heat shock σ factor, RpoH (17). In this case, the two σ factors recognize different consensus sequences, but these can be interdigitated in such a manner as to allow initiation from the same start site.

Here, we focus on the phenomenon of overlapping promoter recognition among promoters recognized by extracytoplasmic function (ECF) family σ factors. The ECF family of σ factors is characterized by their generally small size relative to the primary σ factor, corresponding to the presence of only conserved σ regions 2 and 4 (18,19). Multiple ECF σ factor paralogs are present in many organisms with, in extreme cases, 50 or more paralogs in a single species (20). The activation of these alternative σ factors is often associated with signal sensing pathways initiated at the cell surface that inactivate the corresponding membrane-associated anti-σ factor leading to release of active σ (19,21). The resulting stress responses may be largely distinct, or may overlap extensively with those controlled by other ECF σ factors, depending on the extent of overlapping promoter recognition.

The genome of Bacillus subtilis encodes 7 ECF σ factors: σM, σW, σX, σV, σY, σZ and σylaC (22,23). The regulons for five of these have been defined in detail (reviewed in (23)) and are thought to be determined largely by the sequence determinants in the −35 and −10 elements. σM regulates a large set of genes that include essential functions of cell division and envelope synthesis (24). The σW regulon comprises of at least 60 genes that are expressed in response to membrane active agents and involved in inactivation, sequestering, or eliminating toxic compounds from the cell (25). The σX regulon includes genes which serve to alter cell surface properties to provide protection against antimicrobial peptides (26), whereas σV regulates an overlapping set of ∼30 genes and plays a primary role in lysozyme resistance (27,28). Finally, σY appears to control a small regulon of less than a dozen genes with poorly defined functions (29,30). The regulons and functions of σZ and σylaC have not been clearly defined (31,32), but they are induced in response to some stress conditions (33,34).

Functional redundancy and overlapping promoter recognition among the ECF σ factors of B. subtilis has been thoroughly documented (27,35–38). Indeed, some phenotypes associated with lack of expression of specific genes can only be obtained by mutating two or more of the activating σ factors (31,32). In many cases, this has been ascribed to overlapping promoter recognition, as supported by in vitro transcription experiments (26,35,39) and high resolution start-site mapping (40). Indeed, analysis of the global, condition-dependent transcriptome of B. subtilis revealed a computationally inferred cluster of promoters with the general characteristics of ECF-class promoters (σECF), but the individual regulons could not be discerned under the set of conditions tested (41). In general, these σECF promoters often share a conserved ‘AAC’ motif in the -35 region, and promoters recognized by two or more of σX, σW and σM (collectively defined as σXWM) have a −10 element consensus of ‘CGT’ (32). As shown previously, a subset of σXWM promoters are also recognized σV, thereby defining a σXWMV sub-group (27).

Even though overlapping promoter recognition has been frequently observed among ECF σ factors, the mechanisms that allow some promoters, but not others, to be recognized by more than one σ factor have not been fully resolved. By analogy with the example of E. coli σ70 and σS, overlapping promoter recognition among ECF σ factors is primarily due to similarities in the preferred consensus sequences in the −35 and −10 elements (23), but is strongly modulated by other features within the promoter region. There are clearly some promoters that are highly specific for a single ECF σ factor (there is little if any cross-talk), and others that can be recognized by three or more. For example, sequence comparisons indicate that promoters recognized by σX and σW are distinguished by two key positions in the −10 element (39), and changes to the −10 element were sufficient to switch a promoter from the σX to the σW regulon and vice versa (42). With this particular pair of σ factors, a −10 sequence of CGAC is found in the σX-specific sigX autoregulatory promoter and CGTA confers highly selective recognition of the sigW autoregulatory promoter. However, promoters with a −10 element of CGTC may be recognized by both σ factors (overlapping promoter recognition) (42), and indeed may also be recognized by σM (36) and/or σV (27,28).

Here we identify a −35 proximal homopolymeric T-tract as a novel promoter element involved in promoter discrimination by ECF σ factors in B. subtilis, likely by altering the trajectory of promoter DNA during engagement with RNAP. Based on its conservation in a significant subset of deduced consensus sequences for ECF σ factors, we surmise that this element likely plays a general role in modulating the extent and impact of overlapping promoter recognition among ECF σ factors.

MATERIALS AND METHODS

Strain construction and growth conditions

The B. subtilis 168 wild-type (WT) strain (168) was grown in LB medium. Unless otherwise indicated, liquid media were inoculated from an overnight preculture and incubated at 37°C with shaking at 200 rpm. For selection in B. subtilis, antibiotics were added at the following concentrations: erythromycin (1 μg ml−1) and lincomycin (25 μg ml−1) [for selecting for macrolide-lincosamide-streptogramin B (MLS) resistance], spectinomycin (100 μg ml−1), chloramphenicol (10 μg ml−1), kanamycin (10 μg ml−1) and neomycin (10 μg ml−1). Routine molecular biology procedures were carried out using E. coli DH5α as described (43). Isolation of B. subtilis chromosomal DNA, transformation and specialized SPβ transduction were performed as described (44). Restriction enzymes, DNA ligase and DNA polymerases were used according to the manufacturer's instructions (New England Biolabs).

To construct lacZ transcriptional fusions, the regulatory regions were amplified from B. subtilis 168 genomic DNA by PCR and cloned as a HindIII–BamHI fragment into the integrational plasmid pJPM122 (45). The resulting constructs were linearized with ScaI and transformed into B. subtilis strain ZB307A (46), a strain containing a temperature-sensitive SPβ phage carrying a Tn917-element with a lacZ gene, selecting for neomycin resistance. After plasmid integration adjacent to the lacZ gene, the promoter of choice (P) generates a P-cat-lacZ operon fusion expressing both chloramphicol acetyltransferase (cat) and β-galactosidase (lacZ), as described (45). An SPβ transducing lysate was prepared by heat induction and used to transduce different strain backgrounds as described (45). β-galactosidase activity was assayed in biological triplicate using a modification of the procedure of Miller (47) as described in (48).

Microarray analysis

The WT B. subtilis 168 parental strain and an isogenic strain lacking all seven ECF σ factor genes (designated Δ7) were used for microarray studies (31,49). The WT and Δ7 strains with P were grown in LB to OD600 ∼0.4. Cultures were divided into two parts, and 2 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) was added to one half of the culture and further incubated for 20 min. RNA isolation, cDNA synthesis, microarray hybridization and analysis was done according to (50) using a microarray platform containing 65-mer oligonucleotides (one per annotated open reading frame; Sigma-Genosys). The resulting datasets have been deposited to the NCBI GEO database under accession number GSE95393.

Promoter site directed mutagenesis

The sequence of each promoter region was mutated using overlap extension polymerase chain reaction (PCR) (51). Briefly, for each promoter a set of primers were designed, two universal and flanking, and two pairs of mutagenic overlapping primers to introduce the TTTTT (T5) → AAAAA (A5), T5 → TAAAT or T5 → TTATT changes. Each mutagenic primer was used in pair with a flanking primer to generate two fragments that will have overlapping ends. These two fragments were used in a second PCR joining reaction using only the flanking primers, and the resulting full length product was amplified by PCR.

Protein purification

RNAP was purified from B. subtilis expressing His-tagged β’ subunit as described (52). The sigV gene was PCR amplified from B. subtilis chromosomal DNA and cloned into the pMCSG19 vector (53), which allows the expression of SigV as a His-tagged-maltose binding protein fusion with a site for in vivo cleavage to remove the MBP domain. The resultant plasmid was transformed into BL21/DE3(pLysS) cells harboring the plasmid pRK1037, which expresses the tobacco vein mottling virus protease to cleave the MBP domain. Cells were grown in 1 L of LB medium with ampicillin (100 μg ml−1) and kanamycin (1 μg ml−1) at 37°C to OD600 0.4. IPTG was added to 0.3 mM final concentration and the culture was incubated with shaking overnight at 14°C. Cells were collected by centrifugation and His-tagged σV was purified from the soluble fraction using Prepease Histidine-tagged purification resin (Affymetrix) according to the manufacturer's recommendations. The σM, σW and σX proteins were purified as described (24).

In vitro transcription

Run-off in vitro transcription was performed as previously described (54). Briefly, purified RNAP was incubated with different purified σ factors in a 1:5 molar ratio in transcription buffer (10 mM Tris–HCl pH 8.0, 10 mM MgCl2, 0.5 mM dithiothreitol (DTT), 50 mM KCl, 500 μg/ml acetylated bovine serum albumin, 5% glycerol) and incubated on ice for 15 min. The in vitro transcription reactions contained 10 nM of linear promoter fragment in transcription buffer and 40 nM of RNAP. After 10 min of incubation 37°C, transcription was initiated by adding 0.25 mM (final concentration) of adenosine triphosphate (ATP), guanosine triphosphate (GTP) and cytidine triphosphate (CTP) and 0.025 mM uridine triphosphate (UTP) and 25 μCi of [α-32P]-UTP. After 10 min of incubation, the reaction products were ethanol precipitated in the presence of 2 mM ethylenediaminetetraacetic acid, 0.3 M sodium acetate pH 5.2 and 3 μg glycogen. The RNA pellet was washed with 70% cold ethanol, dried and dissolved in formamide-containing loading buffer and separated on a 6% denaturing polyacrylamide sequencing gel. The gel was then dried and exposed to a phosphorimager screen. The resulting phosphorimage was visualized using a Typhoon FLA 7000 imaging system (GE Healthcare) and analyzed using ImageJ software.

Bioinformatics analysis

Promoter sequences used for alignment and WebLogo analyses for ECF families were obtained from the supplementary data of Staron et al. (20). Promoter sequences for the auto-regulatory region of different sigma factor orthologs from different bacilli were selected based on similarity and Gene context (55). Promoter regions from B. subtilis ECF σ controlled genes were obtained from experimentally determined start sites (23,24,27,35,41,56). Promoter sequence alignment was done using BioEdit and WebLogo 3 (57).

P DNA was modeled onto the crystal structure of the Thermus aquaticus RNAP-fork junction complex (PDB ID:1L9Z) (58) by aligning the non-template strand −35 element of the fork junction DNA and P DNA using the ‘align’ command in PyMol (www.pymol.org). Figure preparation was performed using the UCSF Chimera package (59).

Native PAGE analysis of DNA bending

Annealed oligos encoding the T5dltA promoter (endpoints: −50 to +10) or the G5 variant were electrophoresed on a 15% native polyacrylamide gel at 100 V for 10 h at 4°C in 0.5× tris-acetic acid-ethylenediaminetetraacetic acid (TAE) buffer. The gel was stained for 10 min in EtBr and visualized using a BioRad ChemiDoc XRS.

RESULTS

The extent of the σV regulon is increased in the absence of other ECF σ factors

Here, we compared the transcriptional response to induction of σV (using an IPTG-based induction system) in a WT strain (168) and in a strain deleted for all seven ECF σ factor-encoding genes (designated Δ7; (31,49)) (Figure 1). The observed transcriptional response extends and refines that described previously as resulting from induction of σV (27,60).

Figure 1.

Overlap of ECF σ factor regulons. (A) Gene expression as measured by cDNA microarray analysis in WT Bacillus subtilis 168 cells expressing Pspac-sigV showing normalized fluorescence values corresponding to uninduced cells (x-axis) versus those harvested 20 min after the addition of 2 mM IPTG (y-axis). Each gene associated with an ECF σ factor regulon is represented by a symbol as indicated in the inset (e.g. MW indicates genes known to be activated by both σM and σW), with other genes indicated in small circles. (B) Gene expression as measured by cDNA microarray analysis in the Δ7 background expressing PspacsigV showing normalized fluorescence values corresponding to uninduced cells (x-axis) versus those harvested 20 min after the addition of 2 mM IPTG (y-axis). (C) Fold induction of genes in the B. subtilis 168 parental (168) versus Δ7 backgrounds over-expressing sigV. (D) Fold induction of genes in the 168 WT versus Δ7 backgrounds after over-expression of σV sorted by the last position of the −10 region (either CGTA or CGTC) of the cognate promoter. The mean value is shown in circles for each dataset.

The σV regulon includes a single, autoregulated operon (sigVrsiVoatAyrhK) transcribed from a characteristic ECF-class promoter with a −35 element (GCAAAC) and −10 element (CGTC) as defined by primer extension start site mapping (56). This promoter is, for reasons not entirely clear, strongly activated only by σVin vivo (27,28,60). The majority of the σV-induced genes are associated with promoter sites known to be activated by σX, σW and/or σM and associated with promoter sequences previously defined as the σXWMV sub-group (27). Since these ECF σ factors recognize very similar promoter consensus sequences (23), we interpret this as indicative of overlapping promoter recognition. Since these studies were conducted in the Δ7 strain that lacks the other ECF σ factors (including σM and σW), it was unclear whether the ability of σV to activate this full set of promoters is a characteristic of WT cells.

When we compared the transcriptional response to induction of σV in the B. subtilis 168 parental strain and the Δ7 strain we noted a significantly broader and more robust response in the Δ7 background. This effect was most pronounced for a subset of those promoters previously assigned to the σW regulon, which were generally unresponsive to σV induction in the 168 parental strain (Figure 1A), but were strongly induced in the Δ7 background (Figure 1B). This difference is clear when induction under the two conditions is compared directly (Figure 1C). We note that this is the opposite of what we would predict if σV normally functions to activate expression of one or more other ECF σ encoding genes in a cascade fashion, as occurs for example with the σ factors controlling sporulation (61) and with many of the 13 σ factors of Mycobacterium tuberculosis (62).

To gain further insights into the increased breadth and amplitude of gene induction in the Δ7 strain we sorted all of the genes based on the sequence of the −10 region of the promoter, and in particular whether the last base was an A (CGTA) or a C (CGTC) (Supplementary Table S1). Previously, we have shown that the sequence CGTA is associated with recognition by σW, whereas promoters with CGTC are recognized by σM or σX and, in some cases, also by σW (24,42). Promoters with a −10 region CGTA sequence were generally not induced upon σV induction in the WT strain, but many were highly responsive in the Δ7 strain (Figure 1D). In contrast, for those promoters with a −10 sequence of CGTC the average induction in WT was higher, but was generally increased less in the Δ7 strain (an average increase in induction of ∼2-fold versus >10-fold for the CGTA promoters). These observations support the notion that promoters belonging to the σW regulon (and with CGTA in the −10 region) are somehow less responsive to σV induction in WT, perhaps due to promoter occlusion by basal levels of σW holoenzyme. Indeed, the σW regulon has a significant level of basal activity under a wide variety of growth conditions (63).

The homopolymeric T-stretch is correlated with σV responsiveness

We previously noted that T residues in the −35 proximal region (−26 to −29) seemed to be associated with responsiveness to σV (27). Here, we have extended this analysis by generating sequence alignments and WebLogos of known ECF σ-regulated genes based on their induction in response to σV (Supplementary Figure S1). Overall, those promoters known to be activated by σM, σW and/or σX and also induced (at least 2-fold) upon induction of σV displayed extended conservation in the −35 region (tgaAACntttt; lower case bases are less conserved as shown in Supplementary Figure S1A), whereas those relatively non-responsive to σV induction (<2-fold increase) lacked conservation of these T residues (tgaAAC; Supplementary Figure S1B). Among the σV responsive promoters, conservation of this T-tract is most pronounced for those promoters with a CGTA −10 element (and therefore generally associated with activation by σW), but is also apparent for those with a CGTC −10 element (Supplementary Figure S1C and D). For example, among σW-dependent promoters with a −10 region consensus of CGTA, the −35 region consensus for those induced at least 2-fold by σV (tGAAACnTTTt) is quite distinct from that for those promoter non-responsive to σV induction (TGAAAC) (Supplementary Figure S2).

Effects of mutations affecting the homopolymeric T-stretch in vivo

We used the dltA promoter (P), which can be activated by σX (26), σM (24) and σV (27), as a model system to assess the importance of this homopolymeric T-stretch for expression by holoenzymes bearing different σ factors. The dltABCDE operon encodes proteins involved in D-alanylation of teichoic acids, a cell wall modification pathway that changes the overall net charge (64) and is involved in resistance to cationic antimicrobial peptides and lysozyme (26,27).

We constructed lacZ transcriptional fusions with P and mutant derivatives that altered the −35 proximal homopolymeric T-stretch (Figure 2). In the B. subtilis 168 parental strain disruption of this T-stretch decreased activity substantially, with a 2-fold reduction noted for the single base change converting TTTTT (T5) to TTATT and a 10-fold reduction when T5 was replaced by A5 (Figure 2B). This indicates that this promoter has significant basal activity in B. subtilis 168, and this activity is strongly modulated by this homopolymeric T-stretch.

Figure 2.

A T5-tract modulates gene expression by ECF σ factors. (A) The promoter region of dltA, which is activated by σM, σX or σV showing the −35 and −10 (bold), +1 transcription start site and the T5-tract (underlined). The expression from the P WT and mutant promoters was monitored using P fusions in the indicated background: WT (B), Δ7 with xylose induction of P (C), P (D) and P (E) (values are mean ± SD; n = 3).

To determine how these same mutations might affect recognition by each of the ECF σ factors reported to recognize P, we repeated this analysis in the Δ7 mutant background after induction of σV, σM or σX (Figure 2C–E). Even a single base change within this T-tract (from T5 to TTATT) led to a >3-fold reduction in activity when driven by σX or σV. In contrast, this same change actually improved promoter activity (by nearly 2-fold) when driven by σM (Figure 2D). Thus, mutations that affect the integrity of this homopolymeric T-stretch can differentially affect the activity of ECF σ factors, although in all cases when T5 was replaced by A5 there was a drastic loss of activity.

Generally consistent findings were noted for mutations affecting T-tracts in three other ECF σ factor dependent promoters. Mutation of a −35 proximal T4 tract in the bcrC promoter decreased activity in WT cells and in the Δ7 strain when promoter activity was driven by induction of σV or σM (Supplementary Figure S3). A T4 to TTAT change reduced σV activity, but not that of σM.

Note that this promoter was not strongly activated by induction of σX. Similarly, mutations affecting a −35 proximal T5-tract in the abh and pbpX promoters confirm the general stimulatory effect of this sequence in vivo, and support the notion that this sequence is particularly important when transcription is activated by the σV holoenzyme (Supplementary Figure S4).

Effects of mutations affecting the homopolymeric T-stretch in vitro

We next used in vitro transcription to monitor the effects of these same mutations on activity of purified RNA polymerase reconstituted with various ECF σ factors. Addition of 5-fold molar excess (relative to core RNAP) of either σM, σV, σW or σX resulted in efficient transcription from a known target promoter (Psig, the cognate autoregulatory promoter) for each ECF σ (Figure 3, lane 5). Transcription from these autoregulatory promoters was not detected using the purified RNAP in the absence of added σ factors (Figure 3, lane 6). Each of the reconstituted RNAP holoenzymes were also active with P (Figure 3, lane 1), although to a variable extent. For each holoenzyme this activity was assigned the value of 1.0. Mutation of the T5 region to TTATT had little effect on σM, σX or σW, but led to a 40% reduction in σV activity. In all cases, the T5 to A5 substitution led to a decrease in RNA yield, but this effect was most drastic for σV (a ∼5-fold decrease in activity). Although the agreement is not perfect, these in vitro transcription results corroborate the in vivo expression studies (Figure 2) and indicate that the T5 track is generally stimulatory for ECF σ factors, and this effect is most pronounced for σV.

Figure 3.

Effect of T5-tract on in vitro transcription of dltA promoter. RNAP without addition of any σ factors was unable to transcribe the cognate autoregulatory promoter region for each ECF σ (lane 6), indicating the absence of significant amounts of the cognate σ factor in the purified RNAP. Addition of σM, σV, σW or σX to RNAP at 5:1 molar ratio (in lanes 1–5) enables the resulting RNAP–σ complex to transcribe the autoregulatory promoter of each σ, indicating that all purified proteins are functional (lane 5). Transcription activity of WT (lane 1) and mutant (lanes 2–4) dltA promoter regions was determined using different RNAP-σ holoenzymes.

Addition of a homopolymeric T-tract alters promoter specificity

We next sought to determine if addition of a T5-tract to a promoter lacking one would increase promoter activity. For this, we chose the intragenic promoter within the murG gene (Figure 4A) that is normally dependent on σM (24). This promoter (P) was very weakly activated after induction of σV in either the 168 or Δ7 strains (Supplementary Table S1).

Figure 4.

Addition of a T5-tract enhances promoter recognition by ECF σ factors. (A) Sequence of the P promoter that lacks a T5-tract showing the −35 and −10 (bold), +1 transcription start site. (B) Fold change in P expression of mutant (CCGAG to T5) to WT promoter region in WT, Δ7 or Δ7 over-expressing σM, σV or σX (values are mean ± SD; n = 3). Actual mean values of Miller units for the WT promoter were 4.95 (168), 0.34 (Δ7), 4.45 (induction of σM), 0.4 (induction of σV) and 0.32 (induction of σX). (C) In vitro transcription of the WT P promoter and its mutant (CCGAG to T5) by RNAP associated with σM, σV, σW or σX.

When the P −30 to −26 region was mutated from CCGAG to T5 the level of activity in WT cells increased 6-fold (Figure 4B). This is consistent with the general stimulatory effect of the T5 sequence, as noted above. To determine how individual ECF σ factors might respond to the introduction of the T5 sequence in this promoter context, we used the Δ7 strain expressing σM, σX or σV under xylose induction. For both σM and σX, induction led to activation of P, but the level of expression was not strongly impacted (∼2-fold) by the introduction of the T5 sequence. In contrast, the ability of σV to initiate transcription from this promoter was increased ∼10-fold by addition of the T5 sequence (Figure 4B).

These results were corroborated using in vitro transcription assays with P as template. With the WT promoter sequence, transcription was most active with the σM holoenzyme, as expected, with little to no transcription detected in the presence of σV, σW and σX (Figure 4C). Introduction of the T5-tract results in a promoter that can be transcribed by all four ECF σ factors, although activity with σM is still the highest (Figure 4C). Consistent with the in vivo results, the stimulatory effect of the T5-tract was highest with the σV holoenzyme (∼8-fold). These results support the notion that a homopolymeric T-tract in an ECF σ factor regulated promoter is, in general, stimulatory and can extend the range of σ factors that can potentially recognize the promoter. Conversely, the absence of a stimulatory T-tract may be beneficial when seeking to restrict the activation of a promoter to a single ECF σ factor. Thus, −35 region proximal T-tracts provide a tool for modulating the extent of regulon overlap.

Homopolymeric T-tract sequences in the −35 proximal region of ECF σ-dependent promoters

ECF σ factors are the smallest alternative σ factors, as well as the most abundant and diverse group (20,23). A previous bioinformatic analysis of 2708 ECF σ factors (those containing conserved sigma regions 2 and 4, while lacking region 3) revealed that 2/3 of these could be organized into 43 phylogenetically distinct clusters with at least 10 members each (designated ECF01–ECF43), with the remainder forming smaller clusters (20). Since most ECF σ factors regulate their own synthesis, inferences regarding the likely promoter selectivity of each ECF σ group were developed from a comparison of known and conserved promoter elements upstream of each σ factor operon. Promoters recognized by ECF σ factors typically have a conserved AAC motif in their −35 regions with group-specific -10 region consensus sequences (20).

In B. subtilis, σW, σV and σY belong to groups ECF01, ECF30 and ECF31, respectively, whereas σM and σX did not fall into any of the major groups (20). As reported previously (20), comparison of the deduced consensus sequences for the ECF01, ECF30, and ECF31 families (Figure 5A) indicates a weak conservation of T residues proximal to the −35 element for ECF01(σW) and a more pronounced conservation for ECF30(σV) and ECF31(σY). It is notable that the consensus for σW derived from the regulon defined in B. subtilis includes a T-rich sequence (23,35), whereas that derived from only autoregulatory promoters does not. One interpretation of this observation is that some σW targets belong to more than one regulon, whereas autoregulatory promoters, at least in B. subtilis, tend to be highly selective for their cognate ECF σ factor (23,39,60,65).

Figure 5.

Conservation of a T5-tract in ECF σ factor families. Alignment of ECF σ factor-dependent promoter regions was analyzed using WebLogo 3 (57). (A) WebLogo analysis of ECF01, ECF30 and ECF31 families. Promoter sequences data were extracted from the Supplementary Tables in (20). (B) WebLogo analysis of the auto-regulatory region of σM, σX, σV and σW orthologs from different bacilli. ECF σ factors were selected based on similarity and Gene context (55).

To gain insights into whether conservation of this T-rich region extends to σM and σX family proteins (which were not part of the 43 major groups defined in (20)), we generated sequence alignments using an analogous process in which the auto-regulatory regions were recovered from orthologous σ factor genes (restricting the search however to Bacillus spp.) and used to generate promoter consensus sequences. In this case, orthology was judged by conservation of both σ factor sequences and genomic context, rather than relying on σ factor sequence alone. For comparison with (20), we included in this analysis those ECF σ factors orthologous to σW (ECF01) and σV (ECF30). For each σ, the −35 proximal T-tract was conserved (Figure 5B). The σV consensus derived from Bacillus strains (Figure 5B), is generally consistent with that derived for the broader ECF30 cluster (Figure 5A), and also with the previously reported σV consensus derived from comparison of promoters activated by overproduction of σV in B. subtilis (27). The σX autoregulatory consensus differs in the −10 region (CGAC) from that of the regulon as a whole (CGwC; w = A or T). This is consistent with the previous observations that the autoregulatory promoter is highly specific for σX (due to the CGAC −10 element), whereas many promoters in the broader σX regulon are also recognized by other ECF σ factors (and have a more widely recognized CGTC −10 element) (42).

Collectively, the results above indicate that the ability of a −35 proximal T-tract sequence to stimulate transcription varies significantly among the ECF σ factors tested. The largest stimulatory effect was noted for σV, with the least effect for σW. The presence of short oligo-dT (alternatively known as oligo-dA) tracts are known to induce intrinsic DNA bends in B-form DNA, with maximal bending noted for phased T5–6 sequences (66). We therefore postulated that the stimulatory effect of this −35 proximal oligo-dT sequence was likely due to changes in the trajectory of DNA during binding of the promoter region with RNAP.

To assess the effect of the T5-tract in the dltA promoter we first modeled the promoter region as B-form DNA with and without a T-tract sequence. As expected, the T5-tract is predicted to impart an intrinsic DNA bend (Supplementary Figure S5A). We also used native polyacrylamide gel electrophoresis to compare the mobilities of WT P DNA and P DNA in which the T5-tract was replaced with a G5-tract (Supplementary Figure S5B). Even though the two DNA fragments are the same length (60 bp), WT P migrated more slowly than the G5 variant, consistent with the increased bend angle of WT P observed in silico. The effect of this change in DNA trajectory on the interaction with RNAP was modeled by docking both forms of the P promoter (with and without the T-tract) onto the crystal structure of the T. aquaticus RNAP holoenzyme in complex with fork junction DNA (PDB ID: 1L9Z) (58). While crystal structures of complete RNAP open complexes are available (67), the fork junction complex represents an earlier intermediate (68,69) and may more closely emulate initial binding of RNAP to promoter DNA. Modeling indicates that this motif would likely bend the DNA toward RNAP, leading to a near superposition on the intermediate state represented by the fork-junction DNA (Figure 6). If RNAP binds initially at the −35 consensus element, such bending is predicted to facilitate the subsequent engagement of σ region 2 with the −10 consensus element. We hypothesize that this facilitating mechanism might be most important for those ECF σ factors that are relatively constrained in their ability to engage simultaneously with both the −35 and −10 consensus element due to small linker region between the corresponding protein recognition domains, domain 4 and domain 2, respectively. One factor that can restrict RNA interactions with promoter DNA is the limited flexibility of σ domain 4, which is constrained by interactions with the β flap tip helix (Figure 6) (67). In general, we note that there is a correlation between short interdomain linker regions (which potentially restrict σ factor flexibility) and the stimulatory activity of the −35 proximal T-rich sequence (Supplementary Figure S6).

Figure 6.

Effect of T5 tract on DNA trajectory in during RNAP–DNA complex formation. The RNAP holoenzyme (PDB ID:1L9Z) (58) is shown as a surface representation (α2, β, β’, σA are indicated) and the fork junction DNA is shown as a ribbon (FJ). B-form WT (dltA T) and variant (dltA G) dltA promoter DNA was generated and modeled on to the Thermus aquaticus RNAP-fork junction DNA crystal structure. The ‘Align’ command in PyMol (www.pymol.org) was used to align the resulting B-form models to the non-template strand of the fork junction −35 element and the figure was generated using UCSF Chimera (59).

DISCUSSION

RNAP holoenzyme contacts a large region of DNA during the processes of promoter engagement, melting, and clearance (5). It is therefore not surprising that sequences throughout this extended region can impact activity. These include the −35 and −10 (core) promoter elements (18), upstream AT-rich UP elements (70), and the region surrounding the transcription start site (including the discriminator element) (71). Here, we postulate that the presence of a spacer-region T-tract, proximal to the −35 recognition element, changes the trajectory of the DNA (Figure 6) and facilitates productive engagement of the holoenzyme with the promoter region.

Previous studies have begun to provide insights into the basis of promoter recognition by ECF σ factors (18,72,73). On average, promoters recognized by ECF σ factors adhere closely to consensus, suggestive of strong initial binding (22,23). Promoters activated by ECF σ factors may be specific for only one σ, or may display overlapping promoter recognition. Promoter selectivity is likely due to differences in the −35 and −10 consensus elements and in optimal spacer lengths, which range from 14 to 17 bp (74). Suboptimal spacer lengths alter both the separation and relative rotational positioning of the −35 and −10 elements (75–78), and this presumably requires conformational adaptation on the part of the σ factor.

Here, we describe the novel role of a T-tract in the spacer region between −30 and −26 as an additional element affecting promoter selectivity by ECF σ factors in B. subtilis. Previous studies with E. coli σ70 have demonstrated that spacer region sequences can influence promoter strength by affecting either RNAP–DNA contacts or DNA conformation. For example, E. coli σ70 region 3 contacts the extended −10 region (consensus TGTGn) (3), the linker between σ70 region 2 and 3 interacts with the DNA backbone at −18 (79,80), and the β' subunit zipper region interacts with the ‘Z-element’ near position −21 (81). The spacer region tolerates large scale substitutions of DNA sequences and these modulate, but generally do not eliminate, promoter activity (75,82–84). Previously, T-tract sequences were found to reduce activity of an already consensus σ70 promoter (83,84), but the role of sequence-directed DNA-bending in the −35 proximal region described here has not been previously reported.

Bioinformatic studies suggest that T-tract sequences may be a common feature in promoter recognition by ECF σ factors. Among 29 of the major ECF σ groups (20), three showed a relatively high conservation of T residues in this part of the spacer region. These groups include ECF02 (including E. coli RpoE), ECF30 (including B. subtilis σV) and ECF31 (74). However, these consensus sequences were all derived from presumed autoregulatory promoters, which by analogy with B. subtilis (23,60), may have been selected to have a restricted level of crosstalk. The prevalence of T-tracts in the overall regulons for most ECF σ factors is unknown, and the contribution of the T-tract to promoter recognition may have been systematically underestimated in this prior analysis.

The role of this −35 proximal T-tract in promoter recognition is not yet clear, but we speculate that the key parameter is introduction of an intrinsic DNA-bend (Figure 6). The ability of short 4 to 5 nt T-tracts to bend DNA is well documented (85), and the observation that disrupting this sequence (e.g. changing T5 to TTATT) reduces the stimulatory activity significantly argues in favor of this interpretation. The possible role of intrinsic DNA-bending within the spacer region has not been systematically investigated, but bending is known to accompany promoter engagement (5). Perhaps the closest analogy to the results reported here is the introduction of DNA-bending and twisting by MerR-family regulatory proteins. For example, E. coli CueR binds to spacer region DNA and, in a Cu(I)-responsive manner, twists and bends the DNA to enable RNAP to effectively engage with both the −35 and −10 elements (86). In this case, protein-induced changes in DNA conformation are required to compensate for the non-optimal 19 bp spacer region (87). By analogy, the presence of a T-tract directed DNA-bend in some ECF σ factor dependent promoters may compensate for conformational restrictions imposed by a relatively short amino acid linker between the −10 and −35 recognition domains of σ (Supplementary Figure S5B). Indeed, there is an inverse correlation between the stimulatory activity of a T-tract and the length of the amino acid linker separating the promoter recognition domains in σ.

Our results provide further insights into the mechanisms that contribute to the nature and extent of functional overlap among the ECF σ factor regulons in B. subtilis. Spacer-region T-tracts modulate promoter responsiveness and thereby fine-tune the extent of activation of each promoter by the various σ factors, and therefore in response to various stresses. A set of ECF σ factors (selected to represent diverse groups) was shown previously to recognize their cognate promoters with little if any apparent cross-talk (orthogonality) (74). Since the −35 and −10 promoter elements are recognized by two distinct protein domains, hybrid promoters can be constructed by matching −35 and −10 elements that can then be recognized by domain-swapped σ factors (74). Consideration of the possible effects of −35 proximal T-tract sequences will likely be important in efforts to use ECF σ factors in systems biology, and consideration of linker length when constructing chimeric ECF σ factors may also be advisable. Ultimately, a fuller understanding of all of these components will be essential for efforts to predict the regulons controlled by ECF σ factors and for efforts to co-opt these regulatory systems for use as tools for bioengineering of new pathways and regulatory circuits.

AVAILABILITY

Microarray datasets have been deposited to the NCBI GEO database under accession number GSE95393.

Click here for additional data file.