Catabolite and Oxygen Regulation of Enterohemorrhagic Escherichia coli Virulence

ABSTRACT The biogeography of the gut is diverse in its longitudinal axis, as well as within specific microenvironments. Differential oxygenation and nutrient composition drive the membership of microbial communities in these habitats. Moreover, enteric pathogens can orchestrate further modifications to gain a competitive advantage toward host colonization. These pathogens are versatile and adept when exploiting the human colon. They expertly navigate complex environmental cues and interkingdom signaling to colonize and infect their hosts. Here we demonstrate how enterohemorrhagic Escherichia coli (EHEC) uses three sugar-sensing transcription factors, Cra, KdpE, and FusR, to exquisitely regulate the expression of virulence factors associated with its type III secretion system (T3SS) when exposed to various oxygen concentrations. We also explored the effect of mucin-derived nonpreferred carbon sources on EHEC growth and expression of virulence genes. Taken together, the results show that EHEC represses the expression of its T3SS when oxygen is absent, mimicking the largely anaerobic lumen, and activates its T3SS when oxygen is available through Cra. In addition, when EHEC senses mucin-derived sugars heavily present in the O-linked and N-linked glycans of the large intestine, virulence gene expression is initiated. Sugars derived from pectin, a complex plant polysaccharide digested in the large intestine, also increased virulence gene expression. Not only does EHEC sense host- and microbiota-derived interkingdom signals, it also uses oxygen availability and mucin-derived sugars liberated by the microbiota to stimulate expression of the T3SS. This precision in gene regulation allows EHEC to be an efficient pathogen with an extremely low infectious dose.

to the formation of attaching and effacing (AE) lesions on enterocytes. These AE lesions are responsible for the dynamic remodeling of the host's cytoskeleton to form pedestal-like structures beneath the bacteria (27)(28)(29)(30)(31).
The cross-talk among QseC, QseE, and FusK is important, because they differentially regulate LEE gene expression. Upon QseC phosphorylation of KdpE, KdpE, in conjunction with Cra (a global regulator of genes involved in carbon metabolism [39]), activates the expression of all of the LEE genes by directly binding to the ler promoter under gluconeogenic conditions (34,38,40). Under glycolytic conditions, KdpE and Cra do not activate LEE gene expression (34). Conversely, FusR represses LEE gene transcription upon sensing fucose (33). Sensing of sugar concentrations associated with mucin-derived sugars is essential for EHEC to promote virulence, with fusK, kdpE, and cra mutants being attenuated for mammalian infection (7,33,41). EHEC senses when sugars are more abundant in the lumen or less abundant closer to enterocytes, where the glycophagic microbiota is absent, creating a gluconeogenic environment with more host hormones present. Detecting the alterations in sugar concentrations allows EHEC a way to avoid the premature expression of LEE genes, which should only be expressed at the interface with enterocytes.
In this study, our goal was to determine how much cross-talk exists among Cra, KdpE, and FusR, given their differential roles in the regulation of LEE gene expression. In addition, as oxygen availability is also a key regulator of metabolism, we studied how oxygen availability affects the regulation of LEE gene expression by Cra, KdpE, and FusR. We also evaluated the roles of different mucin-derived sugars in EHEC growth and the secretion of LEEencoded T3SS-secreted EspB under limited-oxygen conditions.

Relationships among Cra, KdpE, and FusR in growth and virulence.
The transcription factors Cra, KdpE, and FusR directly bind to the ler regulatory region to control LEE gene expression (33,34). To address the interplay among these transcription factors in the regulation of the LEE, we constructed ⌬kdpE, ⌬kdpE ⌬cra, ⌬fusR ⌬cra, ⌬fusR ⌬kdpE, and ⌬fusR ⌬kdpE ⌬cra EHEC deletion strains. EHEC with a single deletion of cra, kdpE, or fusR alone exhibits no growth defects, as previously reported (33,34). However, deleting these genes in combination could impair bacterial growth. To determine whether the double and triple mutant strains were comparable to the wild type (WT) in growth, the constructed deletion strains were grown in either LB or 0.1% glucose with 1 mM pyruvate Dulbecco's modified Eagle's medium (DMEM). None of the deletion strains had growth defects compared to WT EHEC (see Fig. S1 in the supplemental material). Because Cra, KdpE, and FusR affect LEE gene expression, which is necessary for AE lesion formation, we next assessed the roles of these transcription factors in the regulation of AE lesion formation. To form an AE lesion, EHEC uses specialized effectors to intimately attach to mammalian cells and reorganize actin to cup the bacteria, forming a pedestal-like structure. To assay whether deleting cra, kdpE, or fusR either alone or in combination affects EHEC AE lesion formation, HeLa cells were infected and assessed for the amount of AE lesions formed and the amount of bacteria attached to each lesion. Compared to WT bacteria (41.99 Ϯ 5.46), the ⌬fusR (56.74 Ϯ 7.11) and ⌬fusR ⌬kdpE (61.52 Ϯ 13.07) strains had significantly more bacteria attached to the HeLa cells, with 73.68, 70.67, and 77.44% of the attached bacteria able to form AE lesions, respectively ( Fig. 1A to C). Any strain with cra deleted had significantly fewer bacteria attached to HeLa cells or associated with AE lesions. While the ⌬kdpE strain had more attached bacteria (61.56 Ϯ 16.74), with 69.86% forming pedestals compared to the WT, these data were not significant. Overall, these data indicate that Cra is a strong activator of AE lesion formation in EHEC; however, deleting fusR either alone or in combination with kdpE shows that these regulators repress or alter the kinetics of AE lesion formation prior to EHEC attachment to the HeLa cell.
LEE gene expression under aerobic gluconeogenic conditions. AE lesion formation is dependent on LEE gene expression, and previous studies under aerobic gluconeogenic growth conditions showed that Cra and KdpE activate (34) while FusR represses LEE gene transcription (33). However, the potential cross talk among Cra, KdpE, and FusR has not been previously investigated. To assay whether deleting cra, kdpE, or fusR either alone or in combination affects LEE gene transcription, mutants were grown aerobically with low glucose and pyruvate, the same medium used for the assays described above. The expression of genes representative of two LEE operons, eae and espA, was measured. The ⌬cra, ⌬kdpE ⌬cra, ⌬fusR ⌬cra, and ⌬fusR ⌬kdpE ⌬cra strains had significantly reduced LEE gene expression (Fig. 2B). The ⌬kdpE and ⌬fusR ⌬kdpE mutant strains had LEE gene expression levels slightly higher than but similar to those of the WT, while the ⌬fusR strain had significantly enhanced LEE gene expression. To confirm our quantitative reverse transcription (qRT)-PCR data, Western blot assays were used to measure the secretion of the LEE-encoded effector EspB, which is encoded within the LEE4 operon with espA (Fig. 2C). Taken together, these data again indicate that Cra strongly activates expression of the LEE while FusR represses it under these growth conditions. Of the three regulators, Cra has the most dominant effect on LEE gene expression.
As pyruvate is a central key metabolite and Cra activity is dependent on sensing of gluconeogenic conditions, the single, double, and triple mutants were grown aerobically with low glucose only to remove pyruvate as a potentially confounding variable. Independently of pyruvate, Cra still significantly reduced the expression of eae and espA in the ⌬cra, ⌬kdpE ⌬cra, ⌬fusR ⌬cra, and ⌬fusR ⌬kdpE ⌬cra strains (Fig. 3A). The expression of eae was also significantly decreased in the ⌬kdpE strain, while espA expression was unchanged (Fig. 3A). These results are consistent with our previous report (34). EspA is encoded within the LEE4 operon, whose expression is subject to high levels of posttranscriptional regulation, including the RNA binding protein CsrA (42)(43)(44)(45). Consequently, a potential explanation for the differential regulation of espA (LEE4) expression in the kdpE mutant in regard to eae Carlson-Banning and Sperandio ( Fig. 3A) and ler (34) is probably posttranscriptional regulation of the LEE4 operon. In addition, secretion of EspB (which is cotranscribed with espA within the same operon) is significantly reduced in ⌬cra, ⌬kdpE ⌬cra, ⌬fusR ⌬cra, and ⌬fusR ⌬kdpE ⌬cra (Fig. 3B), indicating that Cra is an activator of LEE gene expression independent of pyruvate. Surprisingly, deleting fusR produced eae gene expression levels similar to WT, while espA levels were significantly elevated in the fusR mutant, suggesting that pyruvate somehow affects how FusR represses the LEE. The corresponding Western blot assays for secreted EspB showed that the ⌬fusR strain had levels similar to those of the WT, while the ⌬kdpE and ⌬fusR ⌬kdpE strain levels were slightly higher than those of the WT (Fig. 3B). These data suggest that pyruvate affects the degree of FusR repression of the LEE when oxygen is present. Potentially, EHEC could experience such conditions once the gut epithelium is disrupted by cell death, causing an infusion of blood, rich in both oxygen and pyruvate, to enter the environment. However, at some point in the infection, EHEC may need to repress its T3SS to move on to its next niche environment and FusR may help with such a transition.
To affirm that the deletion strains were not exhibiting polar effects, all strains were complemented and grown under aerobic gluconeogenic conditions. The expression of eae and espA by the complemented strains was measured (Fig. 3A).
Effects of pyruvate on LEE gene transcription and secretion of EspB. Given the effect of pyruvate on FusR and its regulation of LEE gene expression, we investigated further how pyruvate may affect WT EHEC. To test if pyruvate also affects LEE gene expression, WT EHEC was grown aerobically in low-glucose DMEM with or without 1 mM pyruvate. Addition of pyruvate resulted in overexpression of all of the LEE genes tested and increased the secretion of EspB ( Fig. 4A and B). This difference in expression is not due to a growth advantage, as EHEC has similar generation times when grown with or without pyruvate (see Fig. S2 in the supplemental material). While these conditions do not accurately mimic the conditions found in flowing human blood, the results may indicate that once EHEC infections become bloody in the intestines, the influx of oxygen and pyruvate may affect how EHEC regulates LEE gene expression. This is likely a dynamic process, with Cra, KdpE, and FusR shifting whether LEE gene expression is activated or repressed.
Oxygen availability affects LEE gene transcription and secretion of EspB. In addition to the roles of pyruvate and sugar sources in LEE gene expression, oxygen availability in the gut varies from luminal to the epithelial surface (46,47). We next asked if LEE gene transcription would be altered under anaerobic or microaerobic conditions. The intestinal lumen, the first site encountered by EHEC, is predominantly anaerobic (46,47). As fewer bacteria are present near the epithelial apical surfaces, oxygen can diffuse across the enterocytes, creating a microaerobic environment (47). Recently, it was shown that the microbiota can also affect oxygen availability in the gut (48,49). Moreover, Citrobacter rodentium (extensively used as a surrogate EHEC model for murine infections [50]), triggers colonic hyperplasia through the activity of the LEE-encoded T3SS, increasing oxygenation at the mucosal surface, leading to the aerobic expansion of C. rodentium (6). EHEC and C. rodentium intimately attach to enterocytes in an environment that may change from microaerobic to aerobic. The expression of the LEE eae and espA genes is lower under anaerobic conditions (growth in an anaerobic chamber), while it is enhanced under aerobic conditions and reaches even higher levels under aerobic conditions in the presence of pyruvate (Fig. 5).
To investigate the contribution of Cra, FusR, and KdpE regulation under anaerobic conditions, we grew WT EHEC and the single, double, and triple mutants under anaerobic low-glucose conditions to measure the expression of the LEE genes, as well as the secretion of EspB. Unlike the aerobic conditions, KdpE had a strong repressive effect on eae and espA in the ⌬kdpE and ⌬fusR ⌬kdpE strains, and in the ⌬kdpE ⌬cra strain, eae expression was significantly increased (Fig. 6A). Secretion of EspB was also enhanced in the ⌬kdpE, ⌬kdpE ⌬cra, and ⌬fusR ⌬kdpE strains compared to that in the WT (Fig. 6B). FusR had expression levels similar to those of the WT (Fig. 6A). Cra again significantly reduced the expression of eae in the ⌬cra, ⌬fusR ⌬cra, and ⌬fusR ⌬kdpE ⌬cra strains, and the expression of espA in the ⌬cra and ⌬fusR ⌬kdpE ⌬cra strains (Fig. 6A). EspB secretion was significantly reduced in the ⌬cra, ⌬fusR ⌬cra, and ⌬fusR ⌬kdpE ⌬cra strains, again suggesting that Cra is a strong LEE activator (Fig. 6B).
When we grew WT EHEC and all of the mutants under microaerobic conditions (the oxygen concentration was measured under microaerobic and aerobic conditions [see Fig. S5 in the supplemental material]), we saw the strong repressor and activator phenotypes become ameliorated, except in the ⌬fusR ⌬kdpE ⌬cra strain (Fig. 7A). Secretion of EspB was not significantly different from that of the WT either (Fig. 7B). Overall, these data indicate that the presence or lack of oxygen significantly alters how EHEC regulates LEE gene expression through Cra, KdpE, and FusR.
EHEC growth with different sugars found in the gut. Because the gut has different oxygen concentrations, especially near the epithelial cell surface, and EHEC alters the expression of its virulence armamentarium dependent on the available oxygen, we next asked how limited oxygen availability affects the growth of EHEC on sugars available in the gut. EHEC and commensal E. coli predominantly consume monosaccharides (51)(52)(53). The mucinderived sugars available for E. coli to utilize in the human gut include glucose, fucose, galactose, sialic acid, N-acetylgalactosamine, N-acetylglucosamine, fructose, xylose, and mannose (12,54). However, the preference for which monosaccharides to metabolize differs between EHEC and commensal E. coli (52). EHEC strains defective in the following metabolism pathways have colonization defects compared to WT strains: fucose (⌬fucAO), ribose (⌬rbsK), mannose (⌬manA), arabinose (⌬araBAD), N-acetylglucosamine (⌬nagE), and galactose (⌬galK) (52). The ability to switch to alternative, less abundant carbon sources is therefore key for EHEC to colonize the gut.
Monosaccharides that feed into the Embden-Meyerhof-Parnas (EMP) pathway are a preferred carbon source for E. coli. However, pathogens often exploit nonpreferred carbon sources to gain a niche advantage. To measure the effects of different mucinderived sugars from the large intestine, we grew WT EHEC and each of our deletion strains with individual sugars in DMEM and measured the overall generation times under oxygen-limited mi-croaerobic conditions, an environmental condition most likely experienced as EHEC moves closer to the epithelial surface.
None of the strains harboring mutations in kdpE, cra, or fusR grew significantly better than WT EHEC. We were surprised that   Expression of representative LEE genes by WT EHEC grown anaerobically, microaerobically, or aerobically in low-glucose DMEM with or without 1 mM pyruvate. Significance was assessed with Student's t test.
the ⌬fusR mutant did not grow faster than the WT when grown in fucose only, as was previously shown for aerobically grown cultures (33). This difference may result from growing this mutant under microaerobic conditions, further indicating the sensitivity of these transcription factors to the available oxygen. However, for six of the sugars, strains with mutations had generation times significantly worse than those of the WT grown with the same carbon source (Table 1). For pyruvate, it should be noted that strains harboring a ⌬cra mutation failed to grow, as has been previously demonstrated (55).
EHEC changes the secretion of a virulence protein when grown with different sugars. Given the differences in the generation times of WT EHEC with different mucin-derived sugars, we next asked whether these different sugars affect the expression of the LEE-encoded protein EspB. The WT EHEC and ⌬espB mutant strains were grown to early stationary phase under microaerobic conditions before growth was halted. With all of the sugars tested, the ⌬espB mutant strain grew similarly to WT EHEC. As expected, WT EHEC grown under high-glucose conditions expressed low concentrations of EspB (Fig. 8). Low EspB production was also observed when galactose, fructose, rhamnose, fucose, ribose, xylose, or arabinose was the sole carbon source. Secretion of EspB increased substantially when galacturonic acid, gluconic acid, glucuronic acid, pyruvate, sialic acid, or mannose was the sole carbon source. A modest increase in EspB secretion was observed when N-acetylglucosamine and N-acetylgalactosamine were used. These data demonstrate that the utilization of at least eight alternative sugars by EHEC can promote the expression and secretion of virulence genes.

DISCUSSION
The GI tract is composed of varied environments and niches that are colonized by different microbial communities. These differences in the gut biogeography of the microbiota are largely dictated by differences in nutrient availability and oxygen tension (1). It is fundamentally important for enteric pathogens to inform themselves of their location and the available resources within the gut and translate this information to regulate the expression of their virulence repertoire.
EHEC has a very small infectious dose, estimated to be 50 CFU (14). Consequently, it is of paramount importance for EHEC to correctly determine its location and the resources available to regulate the expression of virulence genes. EHEC regulation of the LEE pathogenicity island is complex and involves the sensing of both microbiota-and host-derived signals and metabolites. The transcription factors Cra, KdpE, and FusR play an important role in the integration and sensing of these environmental cues, leading to optimal regulation of the LEE genes (7,32,35,38,56). EHEC colonizes the colon, where the major source of carbon is the mucus, which is decorated with sugars such as fucose. These sugars can be harvested by saccharolytic members of the microbiota, such as Bacteroides thetaiotaomicron, and made available to other bacterial species that lack this capability (10). B. thetaiotaomicron causes a remarkable modification of the metabolic landscape in the gut, leading to enhanced concentrations of succinate and metabolites that characterize a gluconeogenic signature (34). These are sensed by Cra, which acts in concert with KdpE to directly activate LEE gene expression (7,34). There is also an additional interplay among Cra, KdpE, and FusR, whose cognate HK (FusK) senses fucose cleaved from the mucus by B. thetaiotaomicron (33). B. thetaiotaomicron is also important for liberating many other sugars from the mucosal glycoconjugates, including sialic acid, N-acetylneuraminic acid, and others (57). Here we examined how specific mucin-derived sugars liberated by the microbiota affect EHEC generation times and the secretion of virulence proteins. EHEC has the shortest generation time in the ED pathway sugar  galacturonic acid, which was twice as short as the generation time in glucose. In addition, the most abundant secretion of EspB was observed with the tested ED pathway sugars ( Fig. 8; see Fig. S3 and S4 in the supplemental material). Utilization of the ED pathway may help EHEC compete among anaerobes that rely on the EMP pathway (classical glycolysis) for ATP production (58).
ED pathway sugars are common in the large intestine from the metabolism of pectin, which is present in fruit and vegetable cell walls. Pectin is constructed of long chains of ␣-1,4-glycosidelinked D-galacturonic acid that are decorated with other terminal sugars such as rhamnose, D-xylose, L-fucose, D-glucuronic acid, and others (59). Importantly, pectin is degraded only in the large intestine, where B. thetaiotaomicron and other microbiota species encode the enzymes to break the long or branched side chains (60,61). E. coli also degrades smaller pectin substrates when B. thetaiotaomicron is present in the large intestine (62). Thus, the amount of galacturonic acid can remain high in diets steady in plant fiber (61). The short generation time of EHEC when galacturonic acid is the sole carbon source and under limited-oxygen conditions and the increased secretion of LEE-encoded EspB suggest that the ED pathway may help initiate EHEC infection and that diet may affect the outcome of EHEC infections.
Oxygen availability dictates the utilization of different metabolic pathways and is important in enteric infections (6,(63)(64)(65)(66). As EHEC moves through the large intestinal lumen, it encounters primarily anaerobic conditions (46,47,61,64,67). This radial oxygen gradient is dependent on atmospheric pressure, the host's ability to sequester oxygen, and the aerotolerant members of the microbiota consuming oxygen in the outer mucosal layer (46,48,68). When EHEC moves toward the gut epithelial cells, oxygen concentrations increase because of diffusion across the microvillus capillary network, thus creating a microaerobic environment (46,47,66,68,69). Indeed, an in vitro model of polarized human epithelial cells indicates that the expression of LEE-encoded EspA, EspB, and Tir increases and promotes adherence under microaerobic conditions (66). Moreover, in murine infections that promote host inflammation, aerotolerant species such as C. rodentium and Campylobacter jejuni blossom when the host microbiota is eliminated or reduced, resulting in higher oxygen concentrations (70). Low butyrate concentrations, which would be associated with a disrupted microbiota, also promote the expression of the T3SS in EHEC (71). Recently, the expression of the LEE-encoded T3SS in C. rodentium (an EHEC surrogate murine infection model extensively employed in the field) was directly tied to its ability to promote an increase in host oxygen availability, which allowed the bacteria to bloom and exploit nutrient niches occupied by the microbiota (6). This increase in oxygen could explain the expansion of aerobic Proteobacteria during a C. rodentium infection (70), which ultimately, through competition for nutrient sources, limits the C. rodentium infection (51). Together, these data suggest that variations in the concentration of oxygen should be sensed by EHEC to successfully control virulence gene expression.
Oxygen availability affects LEE gene expression in WT EHEC and also how Cra, KdpE, and FusR alter expression of the LEE (Fig. 2 to 7). The LEE is poorly expressed under anaerobic conditions and optimally expressed under aerobic conditions, especially in the presence of pyruvate (Fig. 5). Under anaerobic conditions, KdpE and FusR are repressors of LEE gene expression, while Cra activates LEE gene expression (Fig. 6). EHEC encounters an anaerobic environment near the lumen or the outer mucosal layer-an unfavorable location at which to express a T3SS. Microaerobic conditions support LEE gene expression for all of our mutant strains (Fig. 7). These data suggest that the limited amount of oxygen available is most similar to moving toward or having close contact with intestinal epithelial cells (46,47,66). This site is where EHEC should optimally express its T3SS. Of note, oxygen availability is higher at this site and increases with the infection (6). When oxygen is readily available under aerobic conditions, Cra is a strong activator of the LEE (Fig. 2 and 3) and this activation may be enhanced by its physical interaction with KdpE (34). These conditions would mimic disruption of the intestinal epithelial barrier and infusion of blood into the gut. Surprisingly, FusR is more repressive when pyruvate is present. Pyruvate may signal that the expression of a T3SS is no longer warranted and perhaps EHEC should move on to another environmental niche. In summary, oxygen availability and utilization of nonpreferred mucin-derived carbon sources affect the ability of EHEC to express its T3SS. EHEC is versatile in sensing a plethora of environmental cues to accurately time the deployment of the energetically costly T3SS. The savviness of EHEC in highjacking ancient metabolic systems to deploy its virulence armamentarium highlights the deftness of enteric pathogens in causing disease and their ability to precisely sense and adapt to microniches within the intestine.

MATERIALS AND METHODS
Strains and plasmids. The strains and plasmids used in this study are listed in Table S1 in the supplemental material. Standard molecular biology methods were used (72). The primers used are listed in Table S2 in the supplemental material. Nonpolar mutants were constructed by using the Red protocol so that all would be in the same genetic background (73). To construct strains harboring a ⌬cra mutation, a PCR product was amplified with NEB Phusion polymerase and primers JcraredF and JcraredR (see Table S2 in the supplemental material). For a ⌬kdpE mutation, a PCR product was amplified with primers kdpERed-F and kdpERed-R (see Table S2 in the supplemental material). For a ⌬fusR mutation, a PCR product was amplified with primers Z0463lambdaredP1 and Z0463lambdaredP2. All deletion PCR products used pKD4 as the template and were gel purified (Qiagen). The PCR product was then electroporated into the prepared cells and recovered in S.O.C. medium (2% tryptone, 0.5% yeast extract, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl 2 , 10 mM MgSO 4 , 20 mM glucose) for 3 h at 30°C and plated on LB containing kanamycin overnight at 37°C. Colonies were then screened for ampicillin sensitivity and kanamycin resistance and PCR verified with primers for cra (KCB topo cra FW/KCB topo cra RV), kdpE (KCB topo kdpE FW/KCB topo kdpE RV), or fusR (Z0463for/Z0463rev) for the absence of the gene (see Table S2 in the supplemental material). To create nonpolar mutants, the kanamycin resistance cassette was resolved with 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.
To construct the complementation plasmids, each gene was cloned individually into the Invitrogen pCR-Blunt II-TOPO plasmid in accordance with the manufacturer's instructions. PCR products to clone into the TOPO plasmid were amplified with NEB Phusion polymerase with EHEC 86-24 as the template. The primers used to amplify the PCR products of cra (KCB topo cra FW/KCB topo cra RV), kdpE (KCB topo kdpE FW/KCB topo kdpE RV), and fusR (Z0463for/Z0463rev) are described in Table S2 in the supplemental material. To construct complemented strains that express each gene equally, NEBuilder cloning was used to place each gene under the control of the P cat promoter. Briefly, puc19 was linearized with NEB restriction enzyme Eco53kI. With the appropriate TOPO vector (KCB01, KCB02, or KCB03) as a template, each gene was PCR amplified with NEB Phusion polymerase as follows (from Table S2 in the supplemental material): cra, Pcat to cra FW/Puc19 to cra RV; kdpE, pcat to kdpE FW/puc19 to kdpE RV; fusR, pcat to fusR FW/puc19 to fusR RV. With pACYC184 as the template, the P cat promoter for each gene was amplified as follows (from Table S2 in the supplemental material): cra, puc19 to pcat FW/cra to pcat RV; kdpE, puc19 to pcat FW/kdpE to pcat RV; fusR, puc19 to pcat FW/fusR to pcat RV. Linearized puc19 and the PCR products for each gene and the corresponding P cat promoter PCR products were gel purified (Qiagen) prior to incubation at 50°C for 20 min with the NEBuilder HiFi DNA Assembly master mix in accordance with the manufacturer's instructions. Ligated products were then diluted and electroporated into DH5␣ cells. Cells recovered at 37°C for 1 h. Transformants were selected for ampicillin resistance and confirmed by sequencing with M13 RV as a primer. These plasmids are pKCB04, pKCB05, and pKCB06.
To construct the complement vectors for each single gene in pA-CYC184, the plasmid was digested with NEB enzymes EcoRV and SalI. pKCB04 was digested with SfoI and SmaI, pKCB05 was digested with SfoI and SalI, and pKCB06 was digested with SfoI and SalI (NEB). The digested template and respective genes were gel purified (Qiagen) before ligation with NEB T4 ligase overnight at 16°C. Ligation products were electroporated into DH5␣ cells that were allowed to recover at 37°C for 1 h. Transformants were selected for chloramphenicol resistance and tetracycline sensitivity prior to confirmation by sequencing with the appropriate primers used as described above. These plasmids were then transformed into the corresponding single-deletion EHEC strains to construct KCB07, KCB08, and KCB09.
To construct the remaining complement vectors in low-copy-number plasmid pACYC184, the NEBuilder method was again used to make the double-and triple-gene complements. pACYC184 was linearized with NEB EcoRV and gel purified (Qiagen). To make the kdpE cra complement (pKCB10), PCR amplification of the pKCB05 template was done with primers p184kdpE to cra FW and p184 to cra RV and pKCB06 was PCR amplified with primers p184 to kdpE FW and p184 cra to kdpE RV. To make the fusR cra complement (pKCB11), PCR amplification of the pKCB04 template was done with primers p184 to fusR FW and p184cra to fusR RV and the pKCB05 template was PCR amplified with primers p184fusR to cra FW and p184 to cra RV. To make the fusR kdpE complement (pKCB12), PCR amplification of the pKCB06 template was done with primers p184 to kdpE FW and p184fusR to kdpE RV and the pKCB04 template was PCR amplified with primers p184kdpE to fusR FW and p184 to fusR RV. To make the fusR kdpE cra complement (pKCB13), PCR amplification of the pKCB06 template was done with primers p184 to kdpE FW and p184fusR to kdpE RV, the pKCB04 template was PCR amplified with primers p184kdpE to fusR FW and p184cra to fusR RV, and the pKCB05 template was PCR amplified with primers p184fusR to cra FW and p184 to cra RV. Each PCR product was gel purified (Qiagen) prior to incubation with linearized pACYC184 and NEBuilder HiFi DNA Assembly master mix at 50°C for 1 h. Ligated products were then diluted and electroporated into DH5␣ cells. Cells recovered at 37°C for 1 h. Transformants were selected for chloramphenicol resistance and tetracycline sensitivity prior to confirmation by sequencing with the appropriate primers used as described above. These plasmids were then transformed into the corresponding EHEC deletion strains to construct KCB09, KCB10, KCB11, and KCB12.
RNA extraction and qRT-PCR. For EHEC strains grown aerobically and shaking, cultures were grown in 0.1% glucose with or without 1 mM pyruvate DMEM to an optical density at 600 nm (OD 600 ) of 1.0. Complemented EHEC strains were grown aerobically with shaking in 0.1% glucose DMEM to an OD 600 of 1.0. For EHEC strains grown statically (microaerobic) or in an anaerobic chamber, cultures were grown for 6 h to Carlson-Banning and Sperandio early stationary phase in 0.1% glucose DMEM to an OD 600 of 0.6. RNA from three replicates was extracted with the RiboPure bacterial isolation kit in accordance with the manufacturer's protocols (Ambion). qRT-PCR was performed as described previously (38). Briefly, diluted extracted RNA was mixed with validated primers (see Table S2 in the supplemental material), RNase inhibitor, and reverse transcriptase (AB). The mixture was used in a one-step reaction utilizing an ABI 7500 sequence detection system. Data were collected with ABI Sequence detection 1.2 software, normalized to endogenous rpoA levels, and analyzed by the comparative critical threshold (C T ) method. Analyzed data are presented as fold changes over WT levels. The Student unpaired t test was used to determine statistical significance. A P value of Ͻ0.05 was considered significant.
Western blot assays for secreted proteins. From cultures grown in 0.1% glucose with or without 1 mM pyruvate DMEM, secreted proteins were isolated as previously described (18). Twenty micrograms of bovine serum albumin was added to secreted protein samples as a loading control. Proteins were separated by 12% SDS-PAGE, transferred to a polyvinylidene fluoride membrane, and blocked with 10% milk in phosphatebuffered saline (PBS) containing 0.05% Tween (PBST). Membranes were probed with an anti-EspB primary antibody, washed, and then incubated with a secondary antibody conjugated to streptavidin-horseradish peroxidase. GE enhanced-chemiluminescence reagent was added, and the membranes were exposed either to film or with the Bio-Rad ChemiDoc Touch Imaging System (software 1.0.0.15) with Image Lab 5.2.1 software for image display. Each growth condition was replicated a minimum of three times.
Fluorescein actin staining assays. Fluorescein actin staining assays were performed as previously described (74). Briefly, HeLa cells were grown overnight to about 80% confluence at 37°C in 5% CO 2 on coverslips in wells containing DMEM supplemented with 10% fetal bovine serum and a 1% penicillin-streptomycin-glutamine antibiotic mixture. Prior to infection, fresh medium lacking antibiotics replaced the overnight medium. To infect HeLa cells, overnight static bacterial cultures were diluted 100:1 (bacteria to DMEM). After 3 h of infection, the wells were again replaced with fresh medium lacking antibiotics. After 6 h of infection, the coverslips were washed, fixed, and permeabilized. The samples were treated with fluorescein isothiocyanate-labeled phalloidin to visualize actin accumulation and propidium iodide to visualize bacterial DNA and HeLa cell nuclei, respectively. The coverslips were then mounted on slides and imaged with a Zeiss Axiovert microscope. Pedestal formation was quantified as the percentage of pedestals formed per attached bacterium. Replicate coverslips from multiple experiments were quantified, and statistical analyses were performed with Student's unpaired t test. Serially diluted samples of the original bacterial cultures were also plated to confirm that similar CFU ratios were used for infection.
Protein purification and ELISA. pET21-based plasmids expressing EspB were induced with isopropyl-␤-D-thiogalactopyranoside (IPTG), and the proteins were purified with nickel columns (Qiagen). His-EspB was buffered exchanged to PBS with Amicon filters before being diluted for standard control. For the enzyme-linked immunosorbent assay (ELISA), overnight WT and ⌬espB mutant EHEC cultures were diluted 1:100 into 96-well plates in a 0.4% concentration of the sugar indicated in DMEM. The OD 600 was measured at early stationary growth phase with a BMG Labtech FLUOstar Optima microplate reader to confirm that the two strains were comparable in growth. Bacterial growth was quenched with 4ϫ STOP solution (0.92 M sodium azide and 100 l of Sigma protease cocktail inhibitor in PBS). Quenched reaction mixtures were diluted 1:2 and incubated in Dynatech Laboratories Microtiter ELISA plates. Wells were blocked with 5% milk in PBST. Samples were washed in PBST prior to incubation with an anti-EspB primary antibody and a secondary antibody conjugated to streptavidin-horseradish peroxidase. Plates were developed with Sigma 3,3=,5,5=-tetramethylbenzidine and stopped with 2N HCl. The OD 450 was measured with a BMG Labtech FLUOstar Optima microplate reader, and EspB concentrations were calculated on the basis of known control protein standard curves. The Student unpaired t test was used to determine statistical significance. A P value of Ͻ0.05 was considered significant.

ACKNOWLEDGMENTS
We thank Sebastian Winter and Maria Winter for help and use of an anaerobic chamber. This work was supported by NIH grants AI053067 and AI05135.

FUNDING INFORMATION
This work, including the efforts of Vanessa Sperandio, was funded by HHS | NIH | National Institute of Allergy and Infectious Diseases (NIAID) (AI053067). This work, including the efforts of Vanessa Sperandio, was funded by HHS | NIH | National Institute of Allergy and Infectious Diseases (NIAID) (AI05135).