Virulence Meets Metabolism: Cra and KdpE Gene Regulation in Enterohemorrhagic Escherichia coli

ABSTRACT

Gastrointestinal (GI) bacteria sense diverse environmental signals as cues for differential gene regulation and niche adaptation. Pathogens such as enterohemorrhagic Escherichia coli (EHEC), which causes bloody diarrhea, use these signals for the temporal and energy-efficient regulation of their virulence factors. One of the main virulence strategies employed by EHEC is the formation of attaching and effacing (AE) lesions on enterocytes. Most of the genes necessary for the formation of these lesions are grouped within a pathogenicity island, the locus of enterocyte effacement (LEE), whose expression requires the LEE-encoded regulator Ler. Here we show that growth of EHEC in glycolytic environments inhibits the expression of ler and consequently all other LEE genes. Conversely, growth within a gluconeogenic environment activates expression of these genes. This sugar-dependent regulation is achieved through two transcription factors: KdpE and Cra. Both Cra and KdpE directly bind to the ler promoter, and Cra’s affinity to this promoter is catabolite dependent. Moreover, we show that the Cra and KdpE proteins interact in vitro and that KdpE’s ability to bind DNA is enhanced by the presence of Cra. Cra is important for AE lesion formation, and KdpE contributes to this Cra-dependent regulation. The deletion of cra and kdpE resulted in the ablation of AE lesions. One of the many challenges that bacteria face within the GI tract is to successfully compete for carbon sources. Linking carbon metabolism to the precise coordination of virulence expression is a key step in the adaptation of pathogens to the GI environment.

IMPORTANCE

An appropriate and prompt response to environmental cues is crucial for bacterial survival. Cra and KdpE are two proteins found in both nonpathogenic and pathogenic bacteria that regulate genes in response to differences in metabolite concentration. In this work, we show that, in the deadly pathogen enterohemorrhagic Escherichia coli (EHEC) O157:H7, which causes bloody diarrhea, these two proteins influence important virulence traits. We also propose that their control of one or more of these virulence traits is due to the direct interaction of the Cra and KdpE proteins with each other, as well as with their DNA targets. This work shows how EHEC coopts established mechanisms for sensing the metabolites and stress cues in the environment, to induce virulence factors in a temporal and energy-efficient manner, culminating in disease. Understanding how pathogens commandeer nonpathogenic systems can help us develop measures to control them.

Introduction

One of the major challenges faced by bacteria within communities is acquisition of carbon and nitrogen to synthesize primary metabolites. The mammalian gastrointestinal (GI) tract harbors trillions of indigenous bacteria of approximately 1,000 different species (1), whose coexistence relies on the ability of each member to utilize one or a few limiting resources. Invading pathogens have to compete with the microbiota for these resources to establish colonization. These pathogens tend to be aggressive and greedy in their search for a colonization niche and achieve this purpose by precisely coordinating expression of an arsenal of virulence genes.

The GI pathogen enterohemorrhagic Escherichia coli (EHEC) causes hemorrhagic colitis and hemolytic-uremic syndrome (HUS) (2). EHEC employs a type 3 secretion system (T3SS), a needle-like structure through which effectors are translocated into host cells. The T3SS is responsible for the attachment of EHEC to the gut epithelium and the induction of extensive actin rearrangement in the host epithelial cells, culminating in the formation of attaching and effacing (AE) lesions (pedestal-like structures) (2). Most of the genes necessary for their formation are contained within a pathogenicity island (PI) known as the locus of enterocyte effacement (LEE) (2). The majority of the LEE genes are grouped into five major operons, LEE1 to LEE5, and encode the structural components of the T3SS, as well as some of the effectors. The ler gene encodes the master regulator of the LEE genes and is essential for the secretion of LEE and non-LEE-encoded T3SS effectors, pedestal formation, and overall virulence in EHEC (2–5). Expression of ler is regulated by numerous transcription factors, including the response regulator (RR) KdpE (4, 6). KdpE is phosphorylated by its cognate histidine sensor kinase (HK) KdpD in response to potassium and osmotic stress. KdpD and KdpE are encoded within the same operon and regulate the kdpFABC genes important for K⁺ transport and general bacterial homeostasis (7–9). KdpE is also phosphorylated by the noncognate HK QseC in response to the host hormones epinephrine and norepinephrine and a signaling molecule, autoinducer-3 (AI-3), produced by the GI microbiota (6, 10–13, 31, 58). The high level of control of ler expression ensures that, in response to diverse environmental signals, EHEC is able to tightly regulate the expression of the LEE and its virulence.

One important environmental signal that bacteria respond to is carbon nutrients. EHEC’s ability to initiate growth and maintain colonization in vivo depends on whether the carbon source is glycolytic or gluconeogenic (14–17). In vitro studies showed that metabolites can regulate the expression of both metabolism and nonmetabolism genes. The catabolite repressor/activator protein Cra (also known as FruR, a member of the LacI family) is a transcription factor that uses fluctuations in sugar concentrations to activate or inhibit expression of its target genes (18). Cra regulates virulence in Salmonella enterica and Shigella flexneri (19, 20). Cra’s function is cyclic AMP (cAMP) independent but is inhibited by the presence of micromolar concentrations of fructose-1-phosphate (F1P) or millimolar amounts of fructose-1,6-bisphosphate (FBP) (21, 22). These metabolic intermediates bind to the inducer binding domain of Cra, decreasing its binding affinity for target promoters and consequently decreasing its regulatory function.

Here we show that KdpE regulation of ler is glucose dependent and that this dependency is through Cra. We show that Cra and KdpE directly interact with each other to promote ler transcription and AE lesion formation. This convergence of regulation by Cra and KdpE introduces a novel mechanism of regulation that links metabolism to pathogenesis.

RESULTS AND DISCUSSION

Carbon regulation of EHEC pathogenesis.

Given the key role that carbon sources play within the GI tract for niche competition, we investigated the role that carbon sources have in EHEC pathogenesis and their influence on the transcription of ler, the activator of the LEE genes. Using Dulbecco’s modified Eagle’s medium (DMEM) lacking glucose and pyruvate as our base medium, we prepared assay media by adding glucose, glycerol, succinate, or pyruvate (Fig. 1A). Switching to glycolytic conditions by increasing concentrations of glucose (0.1% to 0.4%; 5.56 mM and 25 mM, respectively) or using 0.4% glycerol reduced ler transcription 2-fold, while switching to a gluconeogenic state with 0.4% succinate increased ler mRNA levels 4-fold compared to those with a glucose concentration of 0.4%. The switch from 0.1 to 0.4% glucose that alters virulence gene expression has been shown to be physiologically relevant in humans, where in cholesterol studies the use of 0.4% versus 0.1% glucose has been shown to increase the stimulation of cholesterol absorption in the small intestine (23). EHEC was unable to grow in 0.4% pyruvate as the sole carbon source, but adding it to 0.1% glucose did not vary ler transcription. These data indicate that transcription of the LEE is repressed under glycolytic conditions and activated under gluconeogenic conditions. These findings can be extended to other LEE-containing enteric pathogens, namely, enteropathogenic E. coli (EPEC) and Citrobacter rodentium, in which expression of the LEE genes is also reduced under glycolytic conditions (see Fig. S1 in the supplemental material).

FIG 1.

Carbon regulation of EHEC pathogenesis. (A) qRT-PCR of ler in the following media: no-glucose, no-pyruvate DMEM as the base medium supplemented with low glucose (0.1%), high glucose (0.4%), glycerol (0.4%), succinate (0.4%), or low glucose plus pyruvate (0.1% glucose + 0.4% pyruvate). Results were expressed as fold changes over those of low-glucose DMEM. n/s, not significant. *, P < 0.05. (B) qRT-PCR analysis of ler in WT and ΔkdpE strains complemented with either KdpE or Cra in low- and high-glucose DMEM. (C) Beta-galactosidase assays performed on E. coli K-12 strain BW2511 and its isogenic Δcrp, Δcra, and ΔlacA mutants (from the Keio E. coli knockout library) transformed with plasmid pVS232Z (ler-lacZ, −393 to +86 bp) grown to an OD₆₀₀ of 0.5 in low-glucose DMEM (contains 1 mM pyruvate and 0.1 M NaCl). (D) Schematic representation of the EHEC ler promoter. The transcriptional start sites are indicated with solid arrows. The putative binding site for Cra is depicted with a circle. Probe aa (−450 to −255 bp) was used in subsequent experiments. Underneath is the putative Cra binding sequence on the ler promoter and the Cra binding consensus sequence. (E) qRT-PCR of ler in WT and Δcra strains and the complement in low- and high-glucose DMEM. ler transcript levels were quantified as fold differences normalized to low-glucose WT ler transcript levels. (F) Cra EMSA using probe aa. A radiolabeled kan DNA probe was used as a negative control. (G) Competition EMSA using 70 nM recombinant Cra and increasing amounts of unlabeled ler or kan probes.

To assess whether this carbon source regulation was linked to the QseC/KdpE-dependent LEE regulation, we assessed whether KdpE LEE gene regulation was affected under glycolytic or gluconeogenic conditions. Transcription of ler was decreased in the kdpE mutant compared to that in the wild type (WT) only at 0.1% glucose (gluconeogenic), and was similar to that of the WT at 0.4% glucose (glycolytic), indicating that KdpE activates ler transcription only in gluconeogenic environments (Fig. 1B). These findings were confounding, given that it has been previously reported that under conditions of high glucose availability (glycolytic), IIA^Ntr is dephosphorylated and only in its dephosphorylated form binds to the KdpD HK (the cognate HK for KdpE), increasing its activity and consequently KdpE phosphorylation, leading to higher expression of the KdpE target genes kdpFABC (24). Hence, through this mechanism one would predict that KdpE would activate LEE transcription under glycolytic and not gluconeogenic conditions, which is the opposite of the phenotype that we observed. We then hypothesized that KdpE might regulate the LEE in a glucose-dependent manner through interaction with another transcription factor. It has been well documented that fluctuations in glucose levels lead to different levels of cAMP within bacterial cells, and one of the prominent transcription factors involved in this regulation is cAMP receptor protein (CRP) (25). The CRP binding consensus sequence is very well defined (25), and in silico analysis of the ler regulatory region did not predict any CRP binding sites. Moreover, transcription of a ler::lacZ fusion (−393 to +86 bp) was indistinguishable between WT and Δcrp E. coli K-12 (Fig. 1C), further suggesting that CRP is not involved in ler regulation. However, these analyses identified a putative consensus sequence for Cra (Fig. 1D), a transcription factor that senses changes in metabolite levels as cues to differentially regulate its target genes (22). Cra is a member of the LacI/GalR family that activates genes encoding gluconeogenic enzymes such as FBPase and inhibits genes encoding glycolytic enzymes such as phosphofructokinase (26, 27).

We confirmed that ler transcription was decreased in both Δcra EHEC and Δcra E. coli K-12 grown in 0.1% (low) glucose and that this phenotype could be rescued by expressing Cra in trans (Fig. 1D and E). Transcription levels of ler were similar between WT and Δcra strains in 0.4% (high) glucose (Fig. 1B and E), suggesting that Cra-dependent activation of ler also occurs only in low glucose. To confirm the predicted Cra binding site around −350 bp (Fig. 1D), we performed electrophoretic mobility shift assays (EMSAs) using a ler probe (−450 to −255 bp) (Fig. 1D and F). A Cra concentration of 7 nM was sufficient to shift the ler probe, while the negative-control probe kan did not shift with up to 10 µM Cra (Fig. 1F). To confirm specificity, competition EMSAs showed that Cra binding to ler could be competed by an unlabeled ler probe with a ratio as low as 1:1 but not by the nonspecific unlabeled kan probe (Fig. 1G). Using DNase protection assays, we verified the specific nucleotides in the ler promoter recognized by Cra (Fig. 2) and confirmed that Cra binds to its in silico predicted binding site.

FIG 2.

Footprinting of Cra on ler. (A) To identify the actual nucleotides of the ler promoter involved in binding with Cra, a DNase I footprint assay was carried out using increasing amounts of Cra. The protected region is indicated within the parentheses. (B) DNA sequence of the ler promoter region showing the −35 and −10 positions of both the proximal and distal promoters. The arrow shows the position of probe aa (containing the Cra binding region used in the EMSA studies), and the Cra binding site is indicated in bold. (C) Alignment of the actual binding site with the consensus binding site sequence of Cra.

Cra binding to its targets can be displaced by micromolar amounts of F1P or millimolar amounts of FBP (21, 56, 57). F1P and FBP are intermediates of the glycolysis metabolic pathway (Fig. 3A) (28). FBP is produced either through the glucose phosphorylation metabolic pathway or by the phosphorylation of F1P. To assess the role that glucose and/or its catabolites play in the binding of Cra to the ler promoter, EMSAs were performed with 100 µM F1P and 10 mM and 50 mM FBP. Fructose-6-phosphate (F6P) and glucose-6-phosphate (G6P) were used as negative controls. At a concentration of 350 nM, Cra completely shifted the ler probe (Fig. 3B). Addition of 100 µM F1P significantly reduced this shift, bringing the amount of free DNA in the reaction mixture to about 25% of the original protein free reaction mixture in lane 1 of Fig. 3B. For FBP, a 50 mM concentration was sufficient to decrease binding, bringing the amount of free DNA to approximately 50% (Fig. 3B). The ability of the glucose catabolite FBP to inhibit binding of Cra to ler in vitro may mirror FBP’s role in vivo as a negative inducer of the Cra-ler complex formation. These results support the idea that Cra directly and specifically binds to the ler promoter region and that this binding is inhibited by metabolites such as F1P and FBP that accumulate under glycolytic conditions. Increasing the glucose concentration in the medium pushes the cell toward glycolytic metabolism that increases the amount of FBP in the cell. This would favor ler inhibition through the reduction of Cra binding. The switch to gluconeogenic metabolism using succinate decreases the amount of FBP available in the cell. This would promote Cra binding and increase ler transcription (Fig. 1A).

FIG 3.

Catabolite regulation of Cra binding to DNA. (A) Schematic representation of glucose and fructose metabolism. The catabolites known to be inducers of Cra are boxed. (B) Inducer-supplemented EMSA. Indicated concentrations of intermediates in the fructose and glucose metabolism cascade were added to 2 ng (400 pM) radiolabeled ler probe (bp −450 to −255) and 70 nM Cra. G6P and F6P were used as negative controls.

EHEC is a GI pathogen, Bacteroides thetaiotaomicron is a predominant species within the GI microbiota (29), and EHEC is likely to encounter a large population of this organism in the intestine. Hence, we investigated whether this sugar-dependent regulation of the LEE genes through Cra also occurs when EHEC is cocultured with B. thetaiotaomicron (Fig. 4), conditions that would more closely mimic the multibacterial species environment of the GI tract. Expression of the LEE gene espA is also higher in low glucose than in high glucose in the presence of B. thetaiotaomicron, and Cra is still necessary for activation of its expression (Fig. 4). We have previously reported that the human microbiota produces the AI-3 signal (13) and proposed that EHEC, given its low infection dose, would not sense self-produced AI-3 in the initial stages of infection but sense the AI-3 produced by the intestinal microbiota (30). Indeed, coculture with B. thetaiotaomicron increased LEE gene expression in EHEC (Fig. 4), further substantiating our hypothesis that flora-produced AI-3 activates LEE expression.

FIG 4.

qRT-PCR of espA in high and low glucose for mixed populations of EHEC and B. thetaiotaomicron (1:9). EHEC WT and Δcra strains were cocultured with B. thetaiotaomicron, and the transcription of espA was evaluated. rpoA mRNA levels were used as an internal control to normalize the output C_T values to take into account variation in bacterial numbers. *, P < 0.05; n/s, not significant.

Cra and KdpE interplay in LEE regulation.

KdpE and Cra follow similar patterns toward sugar-dependent regulation of LEE expression, suggesting that these two transcription factors work together to integrate LEE regulation with signaling and metabolism. Through the genetic, bioinformatic, and biochemical analyses depicted in Fig. 1 and 2, we identified the Cra binding region within the ler promoter. Unlike Cra, which has a very well defined consensus sequence, KdpE tends to bind primarily to AT-rich DNA and does not have a very well defined consensus. Hence, to address the mechanism of KdpE-dependent ler regulation, we performed nested deletion analyses of the ler regulatory region (Fig. 5A). This deletion analysis narrowed the region of the ler promoter necessary for KdpE-dependent activation to between −173 and −42 bp (Fig. 5A and B). KdpE activates transcription of ler by directly binding to the ler regulatory region (Fig. 5C and D), and this interaction is specific, given that in a competition EMSA (Fig. 5E), unlabeled ler probe was able to compete with the labeled ler probe for KdpE binding but unlabeled kan probe (negative control) was unable to compete. Interestingly, the unphosphorylated KdpE showed higher binding affinity to the ler promoter than did the phosphorylated KdpE (Fig. 5F). This is in contrast to the KdpE regulation of the kdpFABC genes, to which the phosphorylated KdpE has higher binding affinity (24). During glycolytic growth, there is high glucose availability, IIA^Ntr is dephosphorylated and binds to the KdpD HK to increase KdpE phosphorylation (24), and as a result there is increased kdpFABC transcription and decreased KdpE-dependent LEE expression (Fig. 1B), given that the phosphorylated form of KdpE has lower affinity to the ler promoter (Fig. 5F). These results are in agreement with the observation that KdpE activates ler transcription only under gluconeogenic (low-glucose) conditions (Fig. 1B). Here, we defined that KdpE activates LEE transcription by binding within the bp −173 and −42 region, while Cra binds upstream to the bp −393 and −255 region (Fig. 1 to 5), and that under glycolytic conditions binding of both proteins to the ler promoter is diminished.

FIG 5.

KdpE regulation of the ler promoter. (A) Cartoon representation of plasmids used for nested deletion analysis. Fragments of the ler regulatory region encompass the distal promoter (−173 to −42 bp, pYN01), proximal promoter (−42 to +86 bp, pYN02), and both promoters (−173 to +86 bp, pVS224). (B) Nested deletion analysis in WT and ΔkdpE strains and the complement. The beta-galactosidase assays were performed on samples grown to an OD₆₀₀ of 0.5 in low-glucose DMEM (containing 1 mM pyruvate and 0.1 M NaCl). (C) KdpE EMSA of the ler promoter region using 2 ng (300 pM) probe bb (−255 to −5 bp). Increasing amounts of His-purified recombinant KdpE were used to shift the radiolabeled ler DNA probe. A radiolabeled kan DNA probe was used as a negative control. (D) Cartoon depicting the Cra and KdpE binding regions on ler and probes aa and bb used for EMSAs. (E) Competition EMSA using 5 µM recombinant KdpE and probe bb. A ratio of hot probe to cold probe of 1:10 decreased the shift due to 5 µM KdpE. Unlabeled kan DNA probe was used as a negative control. (F) EMSAs of KdpE and ler in the absence and presence of acetyl phosphate. *, P < 0.05.

How these two transcription factors act in concert to activate ler expression remains undefined. Protein-protein interactions are an important mechanism for molecular processes in the cell. Different members of the LacI family have been shown to form homomultimers and to also interact with other proteins and metabolites as part of their regulatory mechanism (18, 56, 57). Since both KdpE and Cra activate ler transcription in a glucose-dependent manner, by directly binding the ler promoter, we investigated the possibility that Cra and KdpE interact with each other using far-Western blotting (FWB). His-tagged Cra, KdpE, and, as a negative control, QseB were run on gels, transferred to membranes, and then probed with whole-cell lysate (wcl) of a ΔkdpE Δcra mutant expressing either Flag-tagged Cra or Flag-tagged KdpE. Duplicate membranes were then washed and probed with either anti-His or anti-Flag antibody. As expected, all three pure proteins were detected using anti-His-tag antibody (see Fig. S2 in the supplemental material). However, when wcl overexpressing Flag-tagged Cra was used to probe the membranes, only the lanes containing His-tagged Cra and His-tagged KdpE were detected with anti-Flag antibody, indicating that Flag-tagged Cra interacts with itself and KdpE but not with the negative control QseB (Fig. 6A). To further confirm this interaction, we reversed the bait-prey proteins (Fig. 6B). The wcl overexpressing Flag-tagged KdpE interacted with His-tagged KdpE as well as with Cra but not the control QseB. As additional negative controls, we either left replicate membranes unprobed by wcl or probed replicate membranes with the double mutant wcl only before probing with anti-Flag (see Fig. S2). These findings suggest that the two ler-activating proteins Cra and KdpE interact in vitro.

FIG 6.

Cra and KdpE proteins interact in vitro. (A and B) Far-Western blotting of the interaction between Cra and KdpE in vitro. Recombinant His-tagged Cra, KdpE, and QseB (negative control) on a membrane were probed first with whole-cell lysate (wcl) overexpressing Flag-tagged Cra or Flag-tagged KdpE and then with anti-Flag antibodies. Cra is 37 kDa; KdpE and QseB are both 25 kDa. Bands indicate interaction between the membrane-bound His-tagged protein (bait) and the probing Flag-tagged protein (prey). Flag-Cra interacted with His-Cra and His-KdpE but not His-QseB (A). Flag-KdpE interacted with His-Cra and His-KdpE but not His-QseB (B). (C) Cartoon depicting the Cra and KdpE binding regions on ler and probes aa and bb used for EMSAs. (D and E) Mixed protein competition EMSAs were performed using probe bb (−255 to −5 bp). The EMSAs were performed with a constant concentration of KdpE and increasing concentrations of Cra (D) or with a constant concentration of Cra and increasing concentrations of KdpE (E). (F) Beta-galactosidase measurements of ler-lacZ fusion pVS224 (lacking the Cra binding site) in WT, Δcra, and ΔkdpE strains and complement and ΔkdpE Δcra strains. *, P < 0.05. (G) EMSAs of the ler probes aa and bb with Cra. (H) Schematic representation of EHEC ler promoter region indicating the Cra and KdpE binding sites and position of the probe cb (−392 to −5 bp) used for the CD assay. (I) CD spectra were recorded from 190 nm to 290 nm in 1-nm steps using a 1-mm-path-length cell. Samples in 50 mM phosphate buffer (pH 8, 25°C) were scanned three times and averaged for DNA only (red), DNA plus Cra (green), DNA plus KdpE (black), and DNA plus Cra plus KdpE (purple). The changes observed between 240 and 280 nm indicate DNA conformational changes due to the addition of protein.

As Cra has previously been shown to enhance CRP binding to its targets (27), we examined whether Cra could have a similar effect on KdpE binding to ler. Using probe bb (−255 to −5 bp), which lacks the identified Cra binding site (Fig. 6C), we conducted mixed EMSAs where the KdpE concentration was kept constant and the Cra concentration was varied (Fig. 6D). We also repeated this assay, keeping the Cra concentration constant and varying the concentration of KdpE (Fig. 6E). When the concentration of KdpE was kept constant (Fig. 6D, lanes 4 to 7), we observed an increase in the amount of DNA shifted with increasing concentrations of Cra. The maximum amount of Cra added (1.5 µM, Fig. 6D, lane 3) was not sufficient to shift this probe on its own, but when supplemented with 2.5 µM KdpE, it significantly altered the shifting pattern compared to a reaction mixture with 2.5 µM KdpE only (lane 4). When this experiment was repeated keeping Cra constant and adding increasing amounts of KdpE, we again observed a supershift (Fig. 6E). These results indicate that the two proteins Cra and KdpE interact with each other to promote ler transcription. To further test whether the effect of Cra enhancement of KdpE binding could enhance KdpE-dependent ler transcription, we utilized a ler-lacZ fusion, pVS224 (Fig. 5A), containing only the bp −173 to +86 region (which lacks the Cra binding region, Fig. 6C and G) to monitor KdpE-dependent ler transcription in the absence or presence of Cra. As predicted, transcription of this ler-lacZ fusion was unaffected in the cra mutant compared to that in WT (Fig. 6F), given that Cra does not interact with this region of the ler promoter (Fig. 6D and G). In agreement with our nested deletion analyses (Fig. 5), transcription of this fusion was decreased in the kdpE mutant and decreased even further in the double kdpE cra mutant (Fig. 6F), suggesting that interaction between Cra and KdpE has an additive effect on the expression of ler. However, these proteins bind to the ler regulatory region at sites that are distant from one another (Fig. 1 to 5), suggesting that in order for them to interact there has to be DNA bending and looping. Using circular dichroism (CD), we observed that binding of KdpE to the ler promoter slightly changes the DNA structure, while binding of Cra causes a much bigger change, and this change is exacerbated when the two proteins bind together (Fig. 6H and I). The LEE is a PI horizontally acquired by EHEC and has a very low GC content (34%) compared to the GC content of the E. coli backbone genome (50%) (32). It has also been extensively reported that because of this low-GC-content feature, the regulatory region of ler is prone to DNA bending (33), and ler transcription is subject to regulation by several architectural proteins that promote DNA bending such as H-NS, Fis, and Ihf (4). Hence, it is feasible that through DNA bending, Cra and KdpE interact to optimally activate ler transcription.

Cra and KdpE in AE lesion formation.

Ler is the master activator of the LEE genes (Fig. 7A) (4). In low glucose, the decreased expression levels of the LEE2, LEE3, and LEE5 operons in the cra and kdpE mutants were comparable (Fig. 7B), in agreement with the role of these two transcription factors in activating transcription of ler (Fig. 1 to 6). However, the mRNA level of LEE4 (measured by espA, which encodes the T3SS translocon and which is itself secreted through the T3SS) was significantly decreased in the Δcra strain but not in the ΔkdpE strain (Fig. 7B). It is worth noting that expression of the LEE4 operon is also subject to high levels of posttranscriptional regulation (34) and that the RNA binding protein CsrA (involved in posttranscription carbon metabolism regulation [35–37]) differentially affects expression of LEE4 (38). Hence, a potential explanation for the differential LEE4 regulation between KdpE and Cra may be that in addition to modulating ler transcription, these proteins also differentially affect expression of posttranscriptional regulatory systems that exclusively act on LEE4. Transcription of ler is decreased in high glucose compared to that in low glucose, and this phenotype is mediated through both Cra and KdpE (Fig. 1 to 6). Switching to high glucose reduced espA transcription in the WT but did not affect the mRNA levels in the Δcra and ΔkdpE mutants (Fig. 7C). Both the expression (Fig. 7D) and the secretion (Fig. 7E) of EspA were decreased in WT grown in high glucose. Although the ΔkdpE strain had levels of expression and secretion similar to those of WT in low glucose, these levels were unaffected by switching to high glucose. No EspA expression/secretion was observed in the Δcra strain. Altogether, these results indicate that carbon sources influence LEE expression not only transcriptionally but also posttranscriptionally and that KdpE and Cra act in concert in the transcriptional regulation but differ in the posttranscriptional regulation.

FIG 7.

LEE regulation by Cra and KdpE. (A) Schematic representation of the LEE pathogenicity island. (B) qRT-PCR of the other LEE genes in low-glucose DMEM. The mutant mRNA levels were expressed as fold changes over WT mRNA levels. (C) qRT-PCR of espA/LEE4 in WT, Δcra, and ΔkdpE strains in low and high glucose. For all the samples, rpoA mRNA levels were used as an internal control to normalize the output C_T values in order to take into account variation in bacterial numbers. (D and E) Western blots of wcl (D) and SP (E) of WT, Δcra, and ΔkdpE strains grown in low or high glucose were probed with antisera against EspA. RpoA and BSA were used as the loading controls for the wcl and SP blots, respectively. L, low glucose; H, high glucose. P < 0.05; n/s, nonsignificant.

As deletion of either cra or kdpE affects expression of the LEE that affects AE lesion formation, we next investigated whether deletion of these transcription factors would impact pedestal formation. HeLa cells were infected with wild type or the mutant strains, actin was stained with fluorescein isothiocyanate (FITC)-phalloidin (green), and HeLa nuclei and bacteria were stained with propidium iodide (red). Pedestals were visualized as brilliant green patches underneath red bacteria. Although the ΔkdpE mutant formed slightly fewer pedestals than did the WT, the Δcra strain had significantly reduced pedestal formation, and this could be complemented by introduction of a plasmid carrying cra (Fig. 8). The observation that, in addition to decreased LEE1 to LEE3 and LEE5 expression, the Δcra mutant also has a severe decrease in EspA expression (Fig. 7B to E) while the ΔkdpE mutant does not may explain the disparity in pedestal formation between these two mutants. Deletion of both kdpE and cra led to an inability to form pedestals (Fig. 8). Expression of Cra alone could partially complement this defect, while expression of KdpE alone could not. Full complementation was achieved only when both proteins were expressed in the double mutant (Fig. 8B). These data further advocate a synergistic role for these two transcription factors in virulence regulation.

FIG 8.

Cra and KdpE regulation in AE lesion formation. (A) Pedestals are green (actin) cups beneath red bacteria. (B) Quantification of pedestal formation. These were quantified (examining at least 50 HeLa cells per slide, 3 slides each) as percentages of pedestals per attached bacterium. The standard deviation is indicated in parentheses. *, P < 0.05; n/s, nonsignificant.

Conclusions.

The GI microbiota resides in the loose mucus layer and is not in close contact with the host epithelium (39). Growth within the GI tract is determined by the available concentration of nutrients. Consequently, for two species that compete for the same nutrients in the mucus layer and are not attached to the epithelial cells, the one that utilizes these nutrients more efficiently will eliminate the other strain (17). In the mammalian GI tract, EHEC has to compete with the gammaproteobacteria for nutrients, because they have similar preferences for carbon source utilization. However, commensal E. coli is more proficient than EHEC in the utilization of these carbon sources. EHEC uses glycolytic substrates for initial growth but is unable to effectively compete for these carbon sources beyond the first few days and begins to utilize gluconeogenic substrates to stay within the intestine (17). A second strategy used by EHEC to establish colonization of the GI tract is the expression of the LEE-encoded T3SS to closely attach to the host enterocytes, leading to AE lesion formation (2). Activation of LEE expression relies on the sensing of the microbiota and host-derived signaling molecules AI-3 and epinephrine/norepinephrine through the QseC HK (6, 10, 13, 31). Upon sensing these signals, QseC initiates a complex signaling cascade, which through the phosphorylation of the KdpE RR leads to activation of the LEE genes (6). Here we show that optimal direct activation of ler, the master regulator of the LEE genes, by KdpE occurs in conjunction with Cra and that this regulation exclusively occurs under gluconeogenic conditions, showing that activation of the T3SS for epithelial attachment is coordinated with the gluconeogenic shift that EHEC undergoes during intestinal colonization of mammals (17).

Bacteria share common evolutionary progenitors. EHEC diverged from its nonpathogenic relatives about 4.5 million years ago (40), obtaining virulence traits, such as the LEE, through the insertion of mobile genetic elements (2, 41). Here we also show that through convergent evolution, two proteins that were originally designed to regulate nonpathogenic functions have been coopted by a pathogen to regulate virulence factors encoded within a horizontally acquired PI. This regulation also responds to differences in metabolite concentrations and the phosphorylation state of transcription factors, which can be modulated by the availability of carbon sources. This nutrient-based modulation of virulence expression is also intrinsically intertwined with interkingdom chemical signaling. In the GI tract environment, where about 1,000 different bacterial species coexist, the exquisite integration of different cues to regulate virulence gene expression is essential for an invading pathogen to successfully establish itself within a host.

MATERIALS AND METHODS

Strains and plasmids.

Strains and plasmids are listed in Table S1 in the supplemental material. Standard molecular biology methods were used (42). Primers are listed in Table S2 in the supplemental material. Nonpolar mutants were constructed using λ-red (43). Briefly, for the construction of Δcra and ΔkdpE Δcra mutants (see Table S1), respectively, 86-24 (WT) and ΔkdpE cells containing pKD46 were prepared for electroporation. A cra PCR product was amplified using JcraredF and JcraredR primers (see Table S2) with pKD4 as the template and then gel purified (Qiagen). The PCR product was then electroporated into the prepared cells and recovered in SOC medium for 6 h at 30°C and plated on LB containing kanamycin overnight at 42°C. Colonies were then screened for ampicillin sensitivity and kanamycin resistance and PCR verified using primers JcraexF and JcraexR for the absence of the gene. In order to create nonpolar mutants, the kanamycin cassette was resolved using the resolvase plasmid pCP20. The mutants were electroporated with pCP20, and resultant colonies were patched for kanamycin sensitivity. Final verification of proper deletion was performed by sequencing. The construction of the ΔkdpE mutant has been previously published (6).

Plasmids encoding recombinant proteins were constructed by amplifying the coding regions from the EHEC strain 86-24 using Phusion polymerase (NEB), digesting them with appropriate restriction enzymes (NEB), and ligating them into plasmids as summarized in Tables S1 and S2 in the supplemental material. Briefly, for the pBAD33-based plasmids pJN49 and pJN57, the primer pairs JkdpE33_2F/JkdpE33_2R and Jcra33F/Jcra33R were used to amplify the kdpE and cra genes, respectively, with strain 86-24 as a template. The resulting PCR products were cloned into the XbaI and HindIII cloning site of vector pBAD33 (44). To construct Flag-tagged versions of the above-described plasmids, the reverse primers were replaced with Jkdpe33flagR and Jcra33flagR and the cloning process was repeated to create pJN45 and pJN46, respectively. Plasmid pJN56 was constructed by amplifying the cra gene from strain 86-24 using primers JcramycF and JcramycR and cloning the PCR product into the KpnI and EcoRI cloning site of vector pBADMycHisA (Invitrogen). The protein expression plasmid pJN55 was constructed by amplifying the cra gene using primers Jcra21F and Jcra21R and cloning the resulting PCR product into the BamHI and NotI cloning site of vector pET21 (EMBD Biosciences). Proper cloning of the plasmids was confirmed by sequencing.

RNA extraction and qRT-PCR.

Cultures were grown in DMEM to an optical density at 600 nm (OD₆₀₀) of 1.0. RNA from 3 replicates was extracted using the RiboPure bacterial isolation kit according to the manufacturer’s protocols (Ambion). Quantitative reverse transcription-PCR (qRT-PCR) was performed as described previously (6). Briefly, diluted extracted RNA was mixed with validated primers (see Table S2 in the supplemental material), RNase inhibitor, and reverse transcriptase (AB). The mix was used in a one-step reaction utilizing an ABI 7500 sequence detection system. Data were collected using ABI Sequence Detection 1.2 software, normalized to endogenous rpoA levels, and analyzed using the comparative critical threshold (C_T) method. Analyzed data were presented as fold changes over WT levels. The Student unpaired t test was used to determine statistical significance. A P value of ≤0.05 was considered significant.

Nested deletion analysis.

Reporter plasmids were constructed as previously described (47). For the construction of pYN01 and pYN02, the ler promoter region was amplified from the 86-24 strain, using the primer pairs Y2/R1 and Y1/R2, respectively. The resulting PCR products were then cloned into the BamHI and EcoRI cloning site of pRS551 (46). Construction of pVS224 has been previously published (47). The beta-galactosidase assays were performed as described previously (47). Briefly, appropriate strains containing different lacZ fusion-expressing plasmids (see Table S1 in the supplemental material) were grown overnight aerobically at 37°C in LB. Dilutions of 1:100 were grown in triplicate in clear DMEM (low glucose, 0.1 M salt, 0.001 M pyruvate) and appropriate antibiotics to mid-exponential phase (OD₆₀₀ of 0.5). Cells were diluted in Z buffer and lysed with chloroform and 0.1% SDS. After addition of o-nitrophenyl-β-d-galactopyranoside (ONPG), the reaction was timed and stopped using 1 M Na₂CO₃. The OD₄₂₀ was measured and used to calculate the Miller units as previously described (48). The Student unpaired t test was used to determine statistical significance. A P value of ≤0.05 was considered significant.

Protein purification, Western blotting, and FWB.

pET21-based plasmids were induced with isopropyl-β-d-galactopyranoside (IPTG), and the proteins were purified using nickel columns (Qiagen). A modified protocol from the work of Wu et al. was used to perform far-Western blotting (FWB) assays (49). Briefly, equimolar amounts of purified His-tagged protein were separated on a 12% SDS gel, transferred, and blocked with 10% milk in Tris-buffered saline containing 0.05% Tween (TBST). Replicate membranes were then probed with whole-cell lysates (wcl) of the ΔkdpE Δcra double mutant (negative control) or the double mutant overexpressing either Flag-tagged KdpE or Cra. As a further (negative) control, a replicate membrane was left unprobed by the wcl. All membranes were then probed with either anti-His or anti-Flag primary antibodies and then incubated with a secondary antibody conjugated to streptavidin-horseradish peroxidase (HRP). Enhanced chemiluminescence (ECL) reagent (GE) was added, and membranes were exposed to film to detect interacting proteins.

For Western blotting, wcl and secreted proteins (SP) were isolated as previously described (50). One hundred micrograms of bovine serum albumin (BSA) was added to SP for loading control.

EMSAs.

EMSAs were performed as previously described (51). Briefly, defined regions of the promoter (see Text S1 and Table S2 in the supplemental material) were amplified by PCR, purified, quantified, and end labeled using radiolabeled [γ-³²P]ATP (PerkinElmer) and T4 polynucleotide kinase (NEB) according to the manufacturer’s instructions. The radiolabeled probes were then repurified to remove unincorporated ATP. EMSAs were performed by adding increasing amounts of purified recombinant protein to 2 ng labeled probe in binding buffer [60 nM HEPES, pH 7.5, 5 mM EDTA, 3 mM dithiothreitol (DTT), 300 mM KCl, 25 mM MgCl₂, 50 ng poly(dI-dC), 500 µg/ml BSA (NEB)] (52). In relevant experiments, metabolites were added to the indicated final concentrations. The reaction mixtures were incubated for 20 min at room temperature and then loaded on a 6% polyacrylamide gel after addition of a 5% Ficoll DNA loading buffer. The gel was run at 180 V for 6 h or 50 V overnight, dried, and exposed on a phosphorimager.

DNase I protection (footprinting) assay.

The footprinting assays were performed as previously described (52). Briefly, radiolabeled probes were made as described for the EMSAs. The binding reactions were also performed as described for the EMSAs. After the 20-min incubation, a 1:5 dilution of DNase I (Invitrogen) was added and allowed to digest unprotected DNA at room temperature for a set amount of time. The digestion was then stopped by adding 100 µl of stop buffer (200 mM NaCl, 2 mM EDTA, and 1% SDS). Protein was then removed using isoamyl-phenol-chloroform, and the DNA was precipitated using 3 M potassium acetate, 100% ethanol, and 1 µl glycogen. The concentrated samples and a sequencing reaction mixture (Epicentre) were then run on an 8% polyacrylamide gel, dried, and exposed on a phosphorimager. To generate the sequencing reaction, the initial PCR products were used as the template and amplified with end-labeled reverse primers according to the manufacturer’s instructions.

Fluorescein actin staining assays.

Assays were performed as described by Knutton et al. (54). Briefly, HeLa cells were grown on coverslips in wells containing DMEM supplemented with 10% fetal bovine serum (FBS) and 1% PSG antibiotic mix at 37°C, 5% CO₂, overnight to about 80% confluence. The wells were then thoroughly washed with phosphate-buffered saline (PBS) and replaced with fresh medium supplemented with arabinose (0.2% final concentration) and lacking antibiotics. Overnight static cultures of bacteria were then used to infect cells at a dilution of 100:1 (bacteria to DMEM). After a 6-h infection at 37°C, 5% CO₂, the coverslips were washed, fixed, and permeabilized. The samples were then treated with fluorescein

Isothiocyanate (FITC)-labeled phalloidin and propidium iodide were used to visualize actin accumulation and bacteria, respectively. Propidium iodide also stained HeLa nuclei red. The coverslips were then mounted on slides and visualized with a Zeiss Axiovert microscope. Pedestal formation was quantified as percentage of pedestals formed per attached bacterium. Replicate coverslips from multiple experiments were quantified, and statistical analyses were performed using the Student t test. Serially diluted samples of the original bacterial cultures were also plated to confirm that similar CFU ratios were used for infection.

CD.

Circular dichroism (CD) experiments were performed as previously described (55). CD curves were recorded on an Aviv model 62DS spectropolarimeter using a 1-mm-path-length cell. CD spectra were recorded from 190 nm to 290 nm in 1-nm steps in 50 mM phosphate buffer (pH 8, 25°C) on samples of the proteins Cra and KdpE and the ler promoter DNA fragment cb, which encompasses both the Cra and the KdpE binding site (−392 to −5 bp). Three scans were taken and averaged. The effect of proteins on DNA conformation was tracked between 240 nm and 280 nm, as most proteins show no significant CD spectrum in this wavelength range.

Coculture conditions.

B. thetaiotaomicron VPI-5482 (ATCC 29148) was grown anaerobically overnight at 37°C in liquid TYG medium (59) plus 200 µg ml⁻¹ gentamicin. Wild-type EHEC O157:H7 strain 86-24 and the isogenic Δcra mutant were grown anaerobically overnight at 37°C in liquid LB plus 100 µg ml⁻¹ gentamicin. The overnight cultures were pelleted and resuspended in either low-glucose DMEM or high-glucose DMEM (Invitrogen). B. thetaiotaomicron was plated in a 10-fold excess over EHEC O157:H7 to represent the composition of the intestinal microbiota. The bacteria were grown anaerobically in 12-well tissue culture plates in 2 ml of either low-glucose or high-glucose DMEM for 6 h in a GasPak EZ anaerobe container (Becton, Dickinson). RNA was extracted from three biological replicates using a RiboPure bacterial RNA isolation kit (Ambion) according to the manufacturer’s guidelines.

SUPPLEMENTAL MATERIAL

Text S1

Supplemental materials and methods. Download Text S1, DOCX file, 0.1 MB.

Figure S1

Sugar-dependent regulation of virulence genes in other enteric bacteria. qRT-PCR of LEE genes and non-LEE-encoded effector gene nleA in low and high glucose for EPEC (A) and C. rodentium (B). For all the samples, rpoA mRNA levels were used as an internal control to normalize the output C_T values in order to take into account variation in bacterial numbers. *, P < 0.05. Download Figure S1, DOCX file, 0.1 MB.

Figure S2

Cra and KdpE proteins interact in vitro. (A) Recombinant His-tagged Cra, KdpE, and QseB (negative control) were purified, and their size and purity were confirmed using a Coomassie blue gel. Cra is 37 kDa; Kdpe and QseB are both 25 kDa. Far-Western blotting was used to visualize the interaction between Cra and KdpE in vitro. The three His-tagged proteins were run on a gel, transferred, and blocked with 10% milk. Next, the membranes were left unprobed by wcl or probed with wcl alone or wcl overexpressing Flag-tagged Cra or Flag-tagged KdpE. (B) The membranes were probed with anti-His antibodies. (C) The control membranes, left unprobed or probed with wcl alone, did not test positive for Flag (panels 1 and 2, respectively). Download Figure S2, DOCX file, 0.1 MB.

Figure S3

Involvement of Cra in Salmonella enterica serovar Typhimurium invasion of HeLa cells and intramacrophage survival. (A) Invasion of HeLa epithelial cells by WT, Δcra, and complemented strains. (B) Intramacrophage survival (J774 macrophages) of WT, Δcra, and complemented strains. *, P < 0.05. Download Figure S3, DOCX file, 0.1 MB.

Table S1

Strains and plasmids used in this study.

Table S2

Oligonucleotide primers used in this study.