Organization of the LEE1 operon regulatory region of enterohaemorrhagic Escherichia coli O157:H7 and activation by GrlA


E-mail; Tel. (+44) 121 414 5439; Fax (+44) 121 414 5925.


Expression of the genes in the locus of enterocyte effacement (LEE) in enterohaemorrhagic Escherichia coli is primarily co-ordinated by expression of the LEE1 operon. GrlA is a LEE-encoded transcription regulator that has been proposed to be involved in the regulation of expression of the LEE1 operon. We describe a simple plasmid-based system to investigate the LEE1 operon regulatory region and to study GrlA-dependent effects. We report that GrlA can activate transcription initiation at the LEE1 P1 promoter by binding to a target located within the 18-base-pair spacer between the promoter −10 and −35 elements, which were defined by mutational analysis. Shortening this spacer to 17 base pairs increases P1 promoter activity and short-circuits GrlA-dependent activation. Hence, at the P1 promoter, the action of GrlA resembles that of many MerR family transcription activators at their target promoters.


Enterohaemorrhagic Escherichia coli (EHEC) serotype O157:H7 causes haemorrhagic colitis and haemolytic uraemic syndrome in humans, exploiting virulence determinants that include a type III secretion system. Effector molecules are translocated through the type III secretion system, and this causes the formation of attaching and effacing lesions in intestinal epithelial cells (Nataro and Kaper, 1998; Bhavsar et al., 2007). The genes responsible for this phenotype are contained in a pathogenicity island termed the locus of enterocyte effacement (LEE) (McDaniel et al., 1995; Perna et al., 1998). The LEE genes are shared by many bacterial strains including enteropathogenic E. coli (Elliott et al., 1998) and Citrobacter rodentium (Schauer and Falkow, 1993). The LEE contains approximately 40 protein-coding sequences that are transcribed in five major polycistronic operons (designated LEE1–5) and several smaller transcriptional units (Mellies et al., 1999). LEE gene expression is controlled by a complex network of signals and transcription factors (reviewed by Mellies et al., 2007; Tree et al., 2009). The ‘master’ regulator of the system is the Ler protein (the LEE-encoded regulator encoded by the ler gene) that acts as the central activator for expression of most of the LEE genes (Mellies et al., 1999; Bustamante et al., 2001; Umanski et al., 2002; Haack et al., 2003; Stoebel et al., 2008).

The ler gene is the first cistron of the LEE1 operon, and the LEE1 regulatory region, located immediately upstream, is crucial in controlling expression of the LEE. In EHEC, this region has been reported to contain two promoters, P1 and P2, respectively, distal and proximal with respect to the ler translation start site (Fig. 1) (Sperandio et al., 2002). Expression from the LEE1 regulatory region is controlled directly or indirectly by a multitude of regulators, including IHF, Fis, H-NS, QseA, BipA GadX, Pch, DksA, Hha, EtrA, EivF and small RNAs together with Ler itself (Mellies et al., 2007; Hansen and Kaper, 2009; Tree et al., 2009; a recent overview summary can be found in the introduction to Kendall et al., 2010). In addition, GrlR and GrlA, encoded by a bicistronic operon grlRA located within the LEE, between the LEE1 and the LEE2 operons (Fig. 1), appear to play a role (Deng et al., 2004; Barba et al., 2005; Jimenez et al., 2010). GrlA (global regulator of LEE activator) is a transcription activator, while GrlR (global regulator of LEE repressor) acts as a repressor by binding to GrlA and preventing its activity (Jobichen et al., 2007). GrlA is homologous to the enterobacterial CaiF proteins (Mellies et al., 2007), which positively regulate expression of the carnitine pathway (Buchet et al., 1999), and a helix–turn–helix DNA-binding motif has been predicted (Deng et al., 2004).

Figure 1.

Schematic representation of the LEE region. The upper line shows the organization of the different LEE transcription units. The lower left part of the figure shows an expanded sketch of the LEE1 regulatory region and illustrates the LEE10–568 fragment and how it was cloned into the pRW224 lac expression vector. The locations of the two LEE1 regulatory region promoters P1 and P2 are indicated, together with the translation initiation region (TIR that contains the Shine–Dalgarno sequence and the translation initiation codon, see Fig. S1). The lower right part of the figure shows an expanded sketch of the grlRA transcription unit and illustrates how it was cloned into vector plasmid pACYC184 to give pSI01, together with two deletion derivatives to give pSI02 and pSI03. E, H, B and S indicate the location of the EcoRI, HindIII, BamHI and SalI sites that were used in cloning (see Experimental procedures).

Many aspects of the action of GrlA at the EHEC LEE1 regulatory region are still open to question. For example, both Huang and Syu (2008) and Jimenez et al. (2010) reported that purified GrlA fusion proteins could bind to the LEE1 regulatory region but did not define its binding target, while Russell et al. (2007) suggested that GrlA may modulate LEE1 expression via the action of another factor. Hence our aim in this work was to improve our understanding of the action of GrlA at the EHEC LEE1 regulatory region. In our previous work with very complex bacterial promoters we had noted that individual regulatory components were often able to function in isolation (Barnard et al., 2004) and, thus, we chose to use a laboratory strain of E. coli K-12 as a ‘test-tube’ to reconstruct GrlA-dependent activation at the LEE1 regulatory region. We have exploited this background to investigate the activity of the P1 and P2 promoters, to find the target for GrlA-dependent activation at the P1 promoter, and to identify functional determinants in GrlA.

At most E. coli promoters, the key sequences required for activity are the −10 and −35 hexamer elements, which are recognized by different domains of the RNA polymerase σ factor, and the optimal length of the spacer region between these elements is 17 base pairs (McClure, 1985). Most transcription activators function by binding to targets located upstream from or overlapping with the −35 element (Browning and Busby, 2004). Here we report that the functional −10 and −35 elements at the LEE1 regulatory region P1 promoter are separated by a non-optimal 18-base-pair spacer and that GrlA activates expression from P1 by binding to a target within this spacer.


Activity of the LEE1 promoters and nested deletion analysis

The regulatory region of the E. coli O157:H7 LEE1 operon is located upstream from the ler gene. Thus, the start point of this work was to create a construct in which lacZ expression is controlled by the LEE10–568 fragment, which carries the base sequence from position −568 to position −19 upstream of the functional ATG start codon of the ler gene (Fig. 1). Figure 2A illustrates a series of eight nested deletions of this fragment, showing the location of the upstream end of each fragment with respect to the two proposed LEE1 promoters, P1 and P2. Each fragment was cloned into the lac expression vector, pRW224, and the resulting recombinants were transformed into the Δlac E. coli K-12 laboratory strain, M182, and into two EHEC serotype O157:H7 strains (Sakai and EDL933). Measurements of β-galactosidase expression illustrated in Fig. 2B show the effects of the deletions on the activity of the LEE1 promoters. The overall patterns of activity in all three E. coli strains resemble each other in both LB medium and DMEM medium (see Fig. S2). Expression increases approximately two- to fourfold as upstream sequences are removed and optimal expression is found with the LEE10-275 and LEE10–215 fragments, which carry DNA sequence up to positions −275 and −215 upstream from the ler start codon. Longer deletions to positions −203, −195 and −155 result in a large decrease in expression. Since the region from positions −215 to −155 contains the proposed sequence elements for the P1 promoter (Sperandio et al., 2002), we conclude that, in our conditions, P1 is the major functional promoter for the LEE1 operon. With the LEE10–155 and LEE10–115 fragments, which lack all P1 elements, expression is 10–15% of the maximum. This expression is most likely due to the P2 promoter, since the longer deletion in the LEE10–75 fragment, which cuts into P2, causes a reduction in expression to basal levels.

Figure 2.

Nested deletions of the LEE1 regulatory region.
A. Schematic representation of a set of EcoRI–HindIII DNA fragments covering the LEE1 regulatory region. The co-ordinates of the upstream and downstream end of each fragment refer to the number of base pairs upstream from the functional ATG start codon of the ler gene reported by Yerushalmi et al. (2008). The P1 and P2 promoters are indicated by bent arrows and the shaded black boxes represent the cognate −10 and −35 hexamer elements.
B. Bar chart to illustrate measured β-galactosidase activities in E. coli K-12, EHEC Sakai and EHEC EDL933 cells carrying pRW224 with each of the different fragments. Vector refers to empty pRW224 with a short linker between the EcoRI and HindIII sites. Activities were measured after growing the cells in LB medium to an optical density at 650 nm of ∼0.5 at 37°C. The values are the average of three independent assays. Standard errors are shown with error bars.

In a complementary experiment, the LEE10–568, LEE10–315, LEE10–275 and LEE10–203 fragments were shortened to move the downstream HindIII site to position −158, rather than position −19, thereby deleting the downstream P2 promoter, and generating the LEE20–568, LEE20–315, LEE20–275 and LEE20–203 fragments (Fig. 2A). After cloning these fragments into pRW224, the observed patterns of β-galactosidase expression in both the O157:H7 and M182 strains were similar to with the LEE10 fragment series (Fig. 2B). The data are consistent with the conclusion that the P2 promoter makes but a minor contribution to the activity measured here. Note that the increased expression seen with the LEE20 fragments is probably due to the shorter untranslated leader sequence upstream of the reporter lacZ gene.

Identification of functional elements at the LEE1 regulatory region P1 promoter

In the experiment illustrated in Fig. 2, the shortest fragment that shows maximum promoter activity is the LEE20–275 fragment. This fragment carries the DNA sequence from positions −275 to −158 upstream of the ler start codon and includes the P1 promoter. The complete sequence of this fragment is shown in Fig. 3, where the sequence has been renumbered 1–118, starting immediately downstream of the EcoRI linker. We adopted this simpler numbering system to describe the ensuing mutational analysis of the LEE20–275 fragment and the shorter LEE20–203 fragment.

Figure 3.

Mutational analysis of the LEE20–275 fragment. The figure shows the nucleotide sequence of the upper strand of LEE20–275 fragment. The base sequences are numbered 1–118, starting with the first cloned LEE1 regulatory region base (i.e. 275 base pairs upstream from the ler translation start codon). The positions of randomly generated single mutations that reduced expression from this fragment are illustrated by showing the substituted base and, in each case, the adjacent number records the number of times that the particular substitution was obtained. The locations of the P1 promoter −35 and −10 hexamer elements, deduced from this study, are shown by grey shading and the transcript start, at position 107A, determined from data in Fig. 4B is indicated. The upstream end of the smaller LEE20–203 fragment is indicated by a bent solid/dotted line.

To identify elements essential for promoter activity, error-prone PCR was used to generate 10 independent preparations of the LEE20–275 fragment carrying random point mutations. The fragments were then cloned into the pRW224 lac reporter plasmid, the resulting recombinant plasmids were transformed into E. coli strain M182, and transformants were grown on MacConkey indicator plates. As expected, the majority of colonies scored as Lac+ (red), but, after screening over 100 000 transformants, we identified ∼100 Lac- (pale pink) colonies. Sequence analysis showed that 43 of these carried single mutations in the LEE20–275 fragment cloned in pRW224.

The locations of the different single point mutations that reduced lac expression from the LEE20–275 promoter fragment are illustrated in Fig. 3. Strikingly, 25 of the 43 point mutations fall in the TTGACA motif at positions 73–77 that was predicted by Sperandio et al. (2002) to be the LEE1 P1 promoter −35 hexamer, and which corresponds exactly to the consensus hexamer −35 element for E. coli promoters. The effects of the different substitutions in this element were quantified and the data are shown in Table 1. The results are consistent with this being the functional −35 element controlling expression from the LEE20–275 fragment. Another of the point mutations (71A) falls just upstream of the −35 element, and its effects are consistent with the lower promoter activity of the LEE20–203 fragment compared with the LEE20–275 fragment (Fig. 2B).

Table 1.  Mutational analysis of the LEE1 P1 promoter. Thumbnail image of

A second cluster accounts for 15 of the 43 point mutations and appears to identify the promoter −10 hexamer (Fig. 3). Recall that base pairs at positions 1 and 2 of promoter −10 hexamers in E. coli are most crucial for promoter activity (McClure, 1985), so this suggests that the motif TACACA at positions 97–102 is the likely functional −10 hexamer element. Since the consensus −10 element for E. coli promoters is TATAAT, we used site-directed mutagenesis to create complementary mutations to check this suggestion. Data presented in Table 1 show that the 97C, 98C, 98G, 98T and 100G substitutions cause > 90% reductions in expression, while the 99A, 99G, 101G, 102C and 102G substitutions have lesser effects. These results are consistent with our assignation of TACACA as the −10 element. However, this would mean that the spacer between the proposed −35 and −10 hexamer elements is 18 base pairs rather than the optimal 17 base pairs. We therefore constructed a LEE20–275 fragment derivative in which a T in the spacer was deleted (Δ94T), thus shortening the spacer to the optimal 17 base pairs. Data presented in Table 1 show that the measured promoter activity is doubled by the Δ94T mutation. Note that one of the Lac- mutants, selected after random mutagenesis, carried the LEE20–275 fragment with an extra T between position 93 and 94. This insertion extended the spacer to 19 base pairs, and caused a sharp reduction in lac expression (Table 1), suggesting that P1 promoter activity tails off as the spacer length increases from the optimal 17 base pairs.

To confirm the location of the functional promoter in the LEE20–275 fragment, we studied complexes between purified DNA fragments and purified E. coli RNA polymerase holoenzyme. In these experiments, we compared the starting ‘wild-type’ DNA fragment with a corresponding fragment carrying the 98C substitution at position 2 of the −10 hexamer element. An experiment in which binary RNA polymerase–DNA complexes were probed with potassium permanganate, a reagent that modifies T residues in the single-stranded ‘bubble’ produced after local unwinding of promoter DNA around the transcription start, is illustrated in Fig. 4 (Savery et al., 1996; Browning et al., 2009). The experiment reveals extensive unwinding just downstream of the promoter −10 hexamer and, crucially, this unwinding is greatly reduced with the 98C mutant. We then analysed the labelled transcripts formed after labelled nucleoside triphosphates were added to the binary complexes. In this experiment, RNA polymerase runs to a terminator in the DNA fragment synthesizing labelled RNA molecules that are detected by gel electrophoresis (Fig. 4B). The major transcript starting in the LEE20–275 fragment is ∼102 bases and this corresponds to a transcript starting at position 107A, five bases downstream from the proposed −10 element (Fig. 3). This RNA species is completely absent when the DNA fragment carrying the 98C substitution is used (Fig. 4B).

Figure 4.

In vitro permanganate footprinting and transcription assays.
A. The figure shows an autoradiogram that identifies the potassium permanganate-sensitive sites in complexes of holo RNA polymerase with a DNA fragment that includes the LEE20–275 sequence (WT) or a derivative carrying the 98C mutation (98C). Lanes 1 and 3 show the result after control incubations without RNA polymerase, while lanes 2 and 4 show the analysis of samples with 50 nM RNA polymerase. The gel was calibrated using a Maxam-Gilbert sequence reaction (GA) and relevant positions are indicated. The location of the LEE1 P1 promoter −10 and −35 elements are indicated by black boxes and the asterisks identify six consecutive residues just downstream of the −10 element that display RNA polymerase-dependent reactivity to permanganate.
B. Autoradiogram of an analysis by gel electrophoresis of 32P-labelled RNA transcripts made by RNA polymerase holoenzyme from PstI–BamHI fragments carrying wild-type LEE20–275 sequences (lanes 1–3) and the 98C derivative (lanes 4–6). The RNA polymerase concentration was: lanes 1 and 4, no enzyme; lanes 2 and 5, 200 nM; lane 3 and 6, 400 nM. The gel was calibrated with the pSR plasmid-encoded 108-base RNA-I transcript (lane 7) and a Maxam-Gilbert sequence reaction (GA). The proposed LEE1 P1 transcript is indicated by an asterisk.

Activation by GrlA

A major aim of this study was to investigate the action of GrlA at the LEE1 operon regulatory region. To do this, we used a series of pACYC184 derivatives carrying the grlRA operon (pSI01), grlA alone (pSI02), grlR alone (pSI03) or empty vector (Fig. 1), pRW224 derivatives carrying different LEE1 regulatory region fragments (Fig. 2), and the E. coli K-12 M182Δlac strain as a ‘test tube’. In a preliminary experiment, M182 was co-transformed with pRW224 carrying the LEE10–568 fragment and the different pACYC184 derivatives, and β-galactosidase expression was measured. The results show approximately sixfold activation in the presence of GrlA (Fig. 5). As expected (Jobichen et al., 2007), the activation found with GrlA alone (pSI02) was suppressed by the presence of GrlR (pSI01).

Figure 5.

Activation of expression from the LEE1 regulatory region by GrlA. The bar chart illustrates measured β-galactosidase expression in M182 cells containing pRW224 with the LEE10–568 promoter fragment together with plasmid pACYC184ΔHN (R−A−), pSI01 (R+A+), pSI03 (R+A−) or pSI02 (R−A+). Measurements were made after growing the cells in LB medium at 37°C to an optical density of ∼0.5 at 650 nm. Each bar represents the mean of three independent experiments together with the standard error.

GrlA-dependent activation at the LEE1 regulatory region carried by fragments with the two series of nested deletions described earlier (Fig. 2) was measured. Activation is observed with the LEE20 series of fragments that lack the P2 promoter, but is lost with the LEE10–195, LEE10–155, LEE10–115 and LEE10–75 fragments in which segments of the P1 promoter are deleted (Fig. 6). From this, we deduce that the LEE1 P1 promoter is the target for activation by GrlA, and this can be measured even with the 52-base-pair LEE20–203 fragment. Note that GrlA-dependent activation at the LEE1 P1 promoter carried by LEE20–203 fragment can also be measured in the Sakai EHEC serotype O157:H7 strain (see Fig. S3).

Figure 6.

GrlA-dependent activation at different LEE1 regulatory region fragments. The left part of the figure illustrates different fragments covering the LEE1 regulatory region that were cloned into pRW224, using the same conventions as in Fig. 2. The right part of the figure presents β-galactosidase activities measured in M182 cells carrying pRW224 containing each of the fragments. Cells also contained either pSI02 (+GrlA) or empty vector, pACYC184ΔHN (−GrlA). Measurements were made after growth in LB at 37°C to an optical density of ∼0.5 at 650 nm. Standard deviations were obtained from at least three independent experiments. The fold activation by GrlA for each fragment is shown in the third column.

Jimenez et al. (2010) recently reported that GrlA primarily activates LEE1 expression by counteracting repression by the global repressor, H-NS, but also reported GrlA-dependent activation of LEE1 in the absence of H-NS. To investigate this, we measured expression from the LEE1 regulatory region and activation by GrlA in E. coli K-12 strain MG1655 and its isogenic Δhns derivative. With the full-length LEE10–568 fragment, LEE1 promoter activity is clearly repressed by H-NS, and, as expected, the repression is overcome by GrlA which activates to higher levels in both the hns+ and Δhns backgrounds (Fig. 7). In contrast, with the short LEE20–203 fragment, LEE1 promoter activity is not repressed by H-NS and a similar level of GrlA-dependent activation is observed in both genetic backgrounds. Hence, GrlA can activate the LEE1 P1 promoter independently of H-NS.

Figure 7.

Repression by H-NS and activation by GrlA at the LEE1 regulatory region.
A. Schematic representation of the LEE10–568 and LEE20–203 fragments of LEE1 operon regulatory region.
B. The bar chart illustrates measured β-galactosidase activities in E. coli K-12 strain MG1655 (WT) and a Δhns derivative carrying pRW224 with either the LEE10–568 or LEE20–203 fragments. Cells also contain either empty vector pACYC184ΔHN (−GrlA: open bars) or pSI02 (+GrlA: shaded bars). Measurements were made after growth the cells in LB at 37°C to an optical density of ∼0.5 at 650 nm. Values are the average of at least three independent assays, and standard errors are shown with error bars.

Evidence that GrlA binds to the spacer region at the LEE1 P1 promoter

Most bacterial transcription activators function by binding to specific sites upstream of the −35 hexamer at target promoters (Browning and Busby, 2004). However, this cannot be the case for GrlA at the LEE1 P1 promoter since GrlA-dependent activation was observed with both the LEE10–203 and LEE20–203 fragments, which begin immediately upstream of the −35 TTGACA hexamer element (Fig. 8A). To identify essential promoter determinants for activation, we constructed a series of mutations throughout the LEE20–203 fragment and measured GrlA-dependent activation. The effects of 23 point mutations throughout this fragment, which carries only 46 base pairs of the LEE1 regulatory region sequence, were measured. The results (Fig. 8) show that GrlA-dependent activation is completely suppressed by the 89G, 90C, 91C, 92A and 98C substitutions (note that the numbering is as in Fig. 3 and Table 1). As noted, the 98C substitution falls at the second position of the −10 hexamer and completely knocks out P1 activity (Fig. 4). In contrast, positions 89–92 fall within the 18-base-pair promoter spacer region. A simple explanation for the effects of changes at these positions would be that they fall within the binding site for GrlA and hence suppress GrlA binding and ensuing activation.

Figure 8.

Mutational analysis of the LEE20–203 fragment.
A. Base sequence of the LEE20–203 fragment, numbered and annotated as in Fig. 3. The asterisks and shading indicate bases where substitutions suppress GrlA-dependent activation.
B. The figure illustrates β-galactosidase activities measured in M182 cells carrying pRW224 containing the starting LEE20–203 fragment (WT) or derivatives with different single-base substitutions, indicated on the x-axis. Measurements were made in cells containing either pSI02 (+GrlA: shaded bars) or empty vector, pACYC184ΔHN (−GrlA: open bars).

In a complementary set of experiments, effects of altering the length of the spacer length on GrlA-dependent activation at the LEE1 P1 promoter were measured, using derivatives of the LEE20–203 fragment carrying insertions or deletions in the spacer either upstream or downstream of, but not within positions 89–92. The results, presented in Table 2, show that GrlA-independent expression from the LEE20–203 fragment is increased by single-base deletions at positions 79 or 94, but decreased by two-base deletions at positions 79 and 80, or by single-base insertions between positions 78 and 79 or between positions 93 and 94. Hence, as expected, in the absence of GrlA, optimal LEE1 P1 activity is found with a 17-base-pair spacer. However, GrlA-dependent activation of the LEE1 P1 promoter is found only with the starting 18-base-pair non-optimal spacer, and it is suppressed by all of the deletions and insertions that were tested. Note that, in this experiment, we selected locations for insertions and deletions that did not appear to be involved in GrlA binding.

Table 2.  Effects of spacer length on activation by GrlA at the LEE1 P1 promoter. Thumbnail image of

We sought to use the well-established in vitro band shift and footprinting techniques (Minchin and Busby, 2009) to identify the binding target for GrlA at the LEE1 P1 promoter. However our efforts were unsuccessful as we failed to purify soluble functional GrlA. Thus, we turned to genetic approaches and, first, used suppression genetics. Error-prone PCR was used to create random mutations in the grlA-coding sequence within pSI02. Mutated pSI02 DNA was then transformed into E. coli strain M182 containing pRW224 carrying the LEE20–203 fragment with the 92A substitution that reduces lac expression (Fig. 8), resulting in Lac- colonies on MacConkey indicator plates. After screening over 75 000 transformants, from 10 independent preparations of mutated pSI02, we identified 11 Lac+ colonies containing pSI02 with single-base substitutions leading to the RK53, IT59 or IV59 substitutions in GrlA. These substitutions fall in the predicted helix–turn–helix motif of GrlA (Fig. 9A). We found that the RK53, IT59 and IV59 substitutions partially restore GrlA-dependent activation with the LEE20–203 fragment carrying the 92A mutation but not with other mutations (Fig. 9B). This implies that, when bound at the LEE1 P1 promoter, residues 53 and 59 may be sufficiently close to make contact with the base pair at position 92 and suggests that the predicted helix–turn–helix in GrlA is functional in binding the target operator site. To investigate this further, a set of pSI02 derivatives encoding GrlA with alanine substitutions at sequential residues from L52 to S63 was constructed and GrlA-dependent activation of expression from the starting and mutant LEE20–203 fragments was measured. We found that the RA54 substitution causes the largest activation defect (Fig. 10A). Taken together with the suppression genetics experiment, this suggests that the R54 side-chain makes a contact with the GrlA binding target somewhere in its target from positions 89 to 92, and that the RK53 substitution in the neighbouring side-chain makes a contact that compensates for the 92A substitution.

Figure 9.

Analysis of GrlA-dependent activation using suppression genetics.
A. The figure shows the GrlA amino acid sequence. The helix–turn–helix DNA-binding motif predicted by Deng et al. (2004), and confirmed by the Dodd & Egan, GYM2.0 and Jpred prediction tools (see Experimental procedures), is indicated by grey shading. Substitutions at R53 and I59, identified after random mutagenesis and selection for suppressors of the 92A mutation in the LEE1 P1 promoter, are indicated. The number adjacent to each substitution is number of times that it was isolated.
B. The figures show the effect of GrlA substitutions on the activity of LEE1 P1 promoter mutants. The bar charts illustrate β-galactosidase activities in M182 cells carrying pRW224 containing the LEE20–203 fragment with single-base substitutions, indicated on the x-axis. Measurements were made in cells containing either pSI02 encoding wild-type GrlA (WT, open bars) or GrlA with the RK53, IT59 or IV59 substitutions, as indicated (grey bars).

Figure 10.

Epistasis analysis of GrlA interactions. The figure shows bar charts that illustrate measured β-galactosidase activities in M182 cells carrying pRW224 containing the LEE20–203 fragment with the starting LEE1 P1 promoter sequence (A), and with the 89G (B), 90C (C) or 91C (D) mutations. Measurements were made as in Fig. 5 with cells carrying pSI02 encoding wild-type GrlA (WT) or the different alanine substitutions indicated on the x-axis. For each promoter, the data are expressed as a percentage of the activity with wild type, the values are the average of three independent assays, and standard errors are shown with error bars.

In a complementary experiment, we investigated whether any of the alanine substitutions could relieve the ‘down’ effects of specific base substitutions in the putative GrlA operator target in the LEE20–203 promoter fragment. Results in Fig. 10B and C show that the FA57 substitution partially reverses the effects of the 89G and 90C substitutions. The simplest explanation is that residue F57 clashes with the base pairs at positions 89–90 in the mutated LEE20–203 89G or 90C fragments, probably because residue F57 is close to positions 89 and 90 in the GrlA–target DNA complex. A similar relationship was found between residue R53 and position 91 (Fig. 10D).


In pathogenic E. coli and related bacteria, the LEE plays a key role during infection and the expression of the different LEE genes must be carefully orchestrated in time and space. This is likely to account for the complexity of LEE1 regulation. Here, we have focused on one single aspect, namely activation by GrlA, and studied it in the context of a simplified system, using a non-pathogenic E. coli K-12 strain as a host.

The first issue that we addressed was the relative activity of the two promoters, P1 and P2. Our conclusion that P1 is the major promoter is in agreement with Porter et al. (2005). Other groups have reported a bigger role for P2 (e.g. Sperandio et al., 2002; Sharp and Sperandio, 2007) but differences are likely due to the precise fusions, genetic backgrounds and conditions used for the measurements. These differences underscore the point that any set of results applies only to the conditions in which the experiments are performed. Our experimental set-up allowed us to make a detailed mutational analysis of the key P1 elements (Fig. 3 and Table 1) and this identified TTGACA and TACACA as the functional −35 and −10 hexamer elements respectively. The −35 element, which was predicted by Sperandio et al. (2002) and other groups, corresponds exactly to the established consensus for E. coli promoters (McClure, 1985). The −10 element was originally identified by Sperandio et al. (2002) as TACACA but, due to uncertainty in the location of the P1 transcript start, was later reassigned to a sequence two base pairs upstream, TTTACA (Porter et al., 2005; Russell et al., 2007; Sharp and Sperandio, 2007). Our mutational analysis argues unambiguously for TACACA, as originally assigned, and a key consequence of this is that the spacer region between the −10 and −35 elements is 18 base pairs, one above the optimal 17 base pairs for promoter activity (McClure, 1985). An important point to note is that there is considerable flexibility in the location of the 5′ end of transcripts with respect to −10 hexamer regions at bacterial promoters and the structural basis for this is now understood (Darst, 2009). Indeed, the AT-rich nature downstream of the −10 element may confer unusual flexibility at the LEE1 P1 promoter and this may be reflected in the greater than usual unwinding (> 20 base pairs) seen in the open complex probed by permanganate (Fig. 4A; compare with Browning et al., 2009). The consequence of this is that the different start points reported in different experiments could all be correct. For many bacterial promoters, the upstream end of the −10 hexamer provides the major anchoring point for RNA polymerase and the location of the first templated base is not fixed (Sclavi, 2009). For this reason, rather than referencing LEE1 P1 promoter locations with the transcript start as +1, we used an arbitrary numbering system from an upstream fixed point to describe our experiments (Fig. 3).

Most summaries of transcriptional regulation in the LEE (e.g. Kendall et al., 2010) show GrlA as a positive regulator of expression from the LEE1 regulatory region and yet the current literature is not completely unambiguous. We were able to show significant activation at LEE1 by GrlA in our K-12 system (Fig. 5) and this activation was suppressed by GrlR, which suggests that our set-up is reporting physiologically relevant effects. Our most striking observations were GrlA-dependent activation from a 52-base-pair promoter fragment (LEE20–203) that lacks LEE1 sequence upstream of the P1 −35 hexamer, and the localization of the DNA target for GrlA to the spacer region between the P1 −10 and −35 hexamer elements. In agreement with Jimenez et al. (2010), our results (Fig. 7) confirm that GrlA plays a dual role at the LEE1 regulatory region by counteracting repression by the global repressor, H-NS, and activating the P1 promoter. Our observation that expression from the 52-base-pair LEE20–203 promoter fragment is activated by GrlA, but not subject to repression by H-NS, argues that GrlA can function as a true transcription activator, most likely by directly accelerating one of the transactions between RNA polymerase holoenzyme and the LEE1 P1 promoter.

To date, it has proven impossible to reconstitute GrlA-dependent transcription activation at the LEE1 regulatory region in vitro. Indeed, we were unable to purify functional GrlA, in order to reproduce the in vitro binding reported by Huang and Syu (2008) and Jimenez et al. (2010). However, we were able to exploit different genetic approaches to confirm the proposed DNA-binding helix–turn–helix motif in GrlA, and to identify its target in the spacer region. Our conclusion that GrlA binds to the LEE1 P1 promoter spacer region was surprising, since most bacterial transcription activators bind sites that are upstream of or overlapping the −35 element of the target promoter (Browning and Busby, 2004). However several members of the MerR family of transcription factors provide interesting exceptions to this rule since they also bind to sites located between the −10 and −35 hexamers at target promoters (reviewed by Brown et al., 2003). Furthermore, they activate transcription at target promoters where the spacing between the −10 and −35 hexamer elements is greater than the optimum 17 base pairs. Current models for activation by MerR family members suggest that their binding in the spacer bends or twists the target DNA, resulting in juxtaposition of the target promoter −10 and −35 elements that facilitates RNA polymerase binding (reviewed by Brown et al., 2003). The existence of an 18-base-pair spacer in the LEE1 P1 promoter, which, when changed to 17 base pairs, increases promoter activity and stops GrlA-dependent activation, argues that a similar activation mechanism may be occurring here (Fig. 11), even though there is no significant sequence similarity between GrlA and any member of the MerR family. Note that our conclusions are consistent with Jimenez et al. (2010) who reported that the DNA site for GrlA is downstream of position −54 at its target promoter in the LEE1 regulatory region of enteropathogenic E. coli.

Figure 11.

Model for activation of LEE1 P1 promoter by GrlA.
A. Weak recognition of the promoter by RNA polymerase due to suboptimal spacer length.
B. Efficient recognition of the promoter by RNA polymerase due to optimized spacer length.
C. GrlA interacts with the spacer sequence and facilitates RNA polymerase activity.

In conclusion, it is clear that the interaction between GrlA and the LEE1 P1 promoter represents an intriguing case study in the complexity of bacterial gene regulation, of interest, not just to researchers studying bacterial pathogenesis, but also to those interested in promoter organization. Further insights into the action of GrlA may emerge from studying its other targets and its interactions with other proteins. To date, the full extent of the GrlA regulon is unknown, although it has been reported to activate the expression of the Ehx enterohaemolysin (Saitoh et al., 2008) and repress genes responsible for the formation of flagella (Iyoda et al., 2006).

Experimental procedures

Strains, plasmids and oligonucleotides

In this work we used E. coli K-12 strain M182 (Casadaban, 1976), MG1655 and a Δhns derivative (Grainger et al., 2008), and two EHEC O157:H7 strains lacking the stx toxin, Sakai 813, obtained from Chihiro Sasakawa and EDL933, given by Arthur Donohue-Rolfe. The vector plasmids used were pRW224 (Lodge et al., 1992), pSR (Kolb et al., 1995) and pACYC184 (Chang and Cohen, 1978). Plasmid pRW224 is an RK2-based low-copy-number lac expression vector, encoding resistance to tetracycline, designed to facilitate the cloning of EcoRI–HindIII fragments carrying a promoter directed towards the HindIII end of the fragment but without a translation start. Thus, a translation initiation signal for lacZ is located in the vector immediately adjacent to the HindIII site as described by Bingham and Busby (1987) and in Fig. S1. Plasmid pSR is a colE1-based general cloning vector, encoding resistance to ampicillin. Promoters that are cloned into pSR on EcoRI–HindIII fragments run into the bacteriophage λoop terminator downstream of the HindIII site (Browning et al., 2009). Plasmid pACYC184 is a P15A-based general cloning vector, encoding resistance to tetracycline and chloroamphenicol. We routinely cloned DNA segments so as to disrupt resistance to tetracycline. Hence, for use as an empty vector, we constructed the pACYC184ΔHN derivative by deleting intervening sequences between the HindIII and NruI sites. This derivative encodes resistance to chloramphenicol but not to tetracycline. Standard techniques for recombinant DNA manipulations were used throughout this work using PCR with synthetic oligodeoxyribonucleotide primers made by Alta Biosciences (University of Birmingham) that are listed in Supporting information (Table S1). All cloned sequences were checked using the University of Birmingham Functional Genomics Facility (

Cloning of LEE1 regulatory region fragments and mutational analysis

PCR was used to amplify the LEE10–568 fragment using the D61221 and D61222 oligos and genomic DNA from the O157:H7 Sakai 813 strain. The fragment was designed with flanking EcoRI and HindIII sites, respectively, located upstream and downstream of the LEE1 regulatory region, with the HindIII site positioned such that the fragment excluded the ler gene Shine–Dalgarno sequence (Fig. 1). The resulting product was restricted with EcoRI and HindIII and cloned into pRW224. The different fragments in the LEE10 and LEE20 sets of nested deletions, illustrated in Fig. 2, were made using primers listed in Supporting information (Table S1). Each fragment is described by the location of its upstream end, in terms of the number of bases upstream from the functional ler ATG codon defined by Yerushalmi et al. (2008). Different mutations were introduced into the LEE20–275 and LEE20–203 fragments cloned in pRW224 by using error prone PCR (Leung et al., 1989) with flanking primers D10520 and D53463, or by using megaprimer PCR (Perrin and Gilliland, 1990) as in Chismon et al. (2010). The different bases at the LEE1 P1 promoter in the LEE20–275 fragment are numbered 1–118, as shown in Fig. 3. We used this numbering system to describe different P1 promoter mutations, in both the LEE20–275 and LEE20–203 fragments. Note that we did not adopt the customary convention of numbering bases with respect to the transcript start point because of uncertainty about its location.

Cloning of grlRA and mutational analysis

PCR was used to amplify a HindIII–SalI fragment carrying the grlRA operon using the D62895 and D62897 oligos and genomic DNA from the O157:H7 Sakai 813 strain. The resulting product was restricted with HindIII and SalI and cloned into pACYC184 to give pSI01. To construct pSI02, which is a derivative of pSI01 carrying a large in-frame deletion in grlR, we used PCR with primers D63209 and D63210 and pSI01 as a template. The resulting product was cut with BamHI and circularized by ligation to give pSI02. To construct pSI03, which is a derivative of pSI01 deleted for grlA, we used primers D62895 and D63698 and pSI01 as a template. The resulting product was restricted with HindIII and SalI and cloned into pACYC184 to give pSI03. Different mutations were introduced into pSI02 by using error prone PCR (Leung et al., 1989) with flanking primers D63048 and D63049, or by using megaprimer PCR (Perrin and Gilliland, 1990).

β-Galactosidase assays

To assay promoter activity in the LEE1 regulatory region we used the different LEE10 and LEE20 fragments (Fig. 2) cloned in pRW224. E. coli strains carrying different pRW224 recombinants were grown aerobically with shaking at 37°C in LB medium or Dulbecco's modified Eagle's medium (DMEM) containing 35 µg ml−1 tetracycline. In experiments to measure the effects of grlR and grlA, encoded by pACYC184 derivatives, 35 µg ml−1 chloramphenicol was also included. β-Galactosidase levels were measured by the Miller (1972) method. Recorded activities are shown in Miller units and are the average of at least three independent experiments.

Permanganate footprinting and run-off transcription assays

Purified PstI–BamHI DNA fragments were derived from caesium chloride preparations of plasmid pSR carrying the LEE20–275 fragment. Fragments were labelled at the BamHI end using [γ-32P]-ATP and polynucleotide kinase. Permanganate footprints were performed following the protocol of Savery et al. (1996) using holo E. coli RNA polymerase, which was purchased from Epicenter (Madison). Each reaction contained approximately 3 nM labelled DNA fragment in 20 mM HEPES, pH 8.0, 5 mM MgCl2, 50 mM potassium glutamate, 1 mM DTT and 0.5 mg ml−1 BSA and 50 nM holo RNA polymerase as required. Products of footprinting reactions were analysed on denaturing 6% polyacrylamide sequencing gels, calibrated with Maxam Gilbert ‘G+A’ sequencing reactions.

The in vitro transcription experiments were performed as described by Browning et al. (2009) using PstI–BamHI DNA fragments purified from caesium chloride preparations of pSR carrying the LEE20–275 fragment. These fragments served as a template for multiple round in vitro transcription assays in which 20 ng of fragment was incubated in transcription buffer containing 40 mM Tris pH 7.9, 10 mM MgCl2, 1 mM DTT, 100 mM KCl, 100 µg ml−1 bovine serum albumin, 200 µM GTP, 200 µM ATP, 200 µM CTP, 10 µM UTP and 5 µCi [α-32P]-UTP. Reactions were started by adding holo E. coli RNA polymerase purchased from Epicenter (Madison). RNA products were analysed on a denaturing 5.5% polyacrylamide gel and visualized using a Fuji phosphor screen and Bio-Rad Molecular Imager FX. Gels were calibrated with Maxam Gilbert ‘G+A’ sequencing reactions.

Helix–turn–helix prediction for GrlA

The DNA-binding helix–turn–helix of GrlA, suggested by Deng et al. (2004), was confirmed computationally by the Dodd and Egan (1990) method,, the GYM2.0 method, (Narasimhan et al., 2002), and the Jpred structure prediction tool, (Cole et al., 2008).


This work was supported by a BBSRC project grant and by a Commonwealth PhD studentship to M.S.I. We are grateful to Kerry Hollands for help with the footprinting and transcription experiments.