Enterohemorrhagic Escherichia coli O157:H7 gal Mutants Are Sensitive to Bacteriophage P1 and Defective in Intestinal Colonization

Abstract

Enterohemorrhagic Escherichia coli (EHEC), especially E. coli O157:H7, is an emerging cause of food-borne illness. Unfortunately, E. coli O157 cannot be genetically manipulated using the generalized transducing phage P1, presumably because its extensive O antigen obscures the P1 receptor, the lipopolysaccharide (LPS) core subunit. The GalE, GalT, GalK, and GalU proteins are necessary for modifying galactose before it can be assembled into the repeating subunit of the O antigen. Here, we constructed E. coli O157:H7 gal mutants which presumably have little or no O antigen. These strains were able to adsorb P1. P1 lysates grown on the gal mutant strains could be used to move chromosomal markers between EHEC strains, thereby facilitating genetic manipulation of E. coli O157:H7. The gal mutants could easily be reverted to a wild-type Gal⁺ strain using P1 transduction. We found that the O157:H7 galETKM::aad-7 deletion strain was 500-fold less able to colonize the infant rabbit intestine than the isogenic Gal⁺ parent, although it displayed no growth defect in vitro. Furthermore, in vivo a Gal⁺ revertant of this mutant outcompeted the galETKM deletion strain to an extent similar to that of the wild type. This suggests that the O157 O antigen is an important intestinal colonization factor. Compared to the wild type, EHEC gal mutants were 100-fold more sensitive to a peptide derived from bactericidal permeability-increasing protein, a bactericidal protein found on the surface of intestinal epithelial cells. Thus, one way in which the O157 O antigen may contribute to EHEC intestinal colonization is to promote resistance to host-derived antimicrobial polypeptides.

Enterohemorrhagic Escherichia coli (EHEC) is an emerging cause of food-borne illness. After ingestion of contaminated food, humans can develop symptoms ranging from mild diarrhea to the severe, and at times life-threatening, condition hemolytic-uremic syndrome. Currently, EHEC is the most common cause of pediatric renal failure in the United States (15). Besides causing significant human disease, EHEC is a major economic burden due to the high cost associated with the recall of contaminated beef (www.fsis.usda.gov/FSIS_Recalls/index.asp). Several EHEC serotypes cause disease, but the O157 serotype is by far the most common cause of EHEC-related disease in North America, Europe, and Japan (6).

Lipopolysaccharide (LPS), found in the outer membrane of gram-negative bacteria, is composed of lipid A, core oligosaccharide, and repeating O-antigen subunits. The O antigen is covalently linked to the outer region of the core oligosaccharide, and it appears to act as a barrier that can protect enteric pathogens against toxic agents encountered in host gastrointestinal tracts (23). For example, galU and galE mutants of Vibrio cholerae, which lacked O antigen, were defective in intestinal colonization, although they had no growth defect in rich medium. These mutants were more sensitive than O-antigen-producing strains to killing by complement and cationic antimicrobial peptides, suggesting that their defect in colonization was attributable to their sensitivity to bactericidal substances elaborated by the host gastrointestinal tract (17).

Like many enteric pathogens, E. coli O157 produces LPS that contains an extensive O antigen. The O157 O-antigen subunit consists of N-acetyl-d-perosamine, l-fucose, d-glucose, and N-acetyl-d-galactose (22). Production of N-acetyl-d-galactose requires that its precursor, galactose, be modified by the enzymes GalE, GalT, GalK, and GalU. Salmonella enterica serovar Typhimurium and E. coli gal mutants do not make O antigen (8, 18, 25).

The inner region of the LPS core oligosaccharide, which is conserved in many enteric bacteria, serves as the receptor for bacteriophage P1. Phage P1 has been a workhorse for genetic manipulation of E. coli K-12 for many decades. P1-mediated generalized transduction enables movement of mutations for generation of isogenic bacterial strains, which is often required for proving the linkage between particular genotypes and phenotypes. In S. enterica serovar Typhimurium, which has an LPS core oligosaccharide similar to that of E. coli K-12, the long O antigen obscures the core oligosaccharide and prevents P1 from adsorbing to the bacteria. O-antigen mutants (Δgal, ΔgalE, and ΔgalU) of S. enterica serovar Typhimurium have been shown to be P1 sensitive (18).

P1-mediated generalized transduction of EHEC has not been described previously. Here, we adopted the strategy that was used to facilitate P1 transduction in Salmonella. Wild-type EHEC O157:H7 strains are resistant to P1, but O157:H7 gal mutants were found to be P1 sensitive and permitted P1-mediated movement of genetic markers between EHEC strains. Furthermore, we developed a method that allows a simple reversion to convert P1-sensitive strains to the wild type. Interestingly, we found that the P1-sensitive galETKM O157:H7 mutant was extremely attenuated in the ability to colonize the infant rabbit intestine and had increased sensitivity to bactericidal permeability-increasing protein (BPI).

MATERIALS AND METHODS

Bacterial strains and growth.

The strains used in this study are shown in Table 1. Unless indicated otherwise, strains were grown in LB broth or on LB agar plates. For antibiotic selection, agar plates were supplemented with ampicillin (80 μg/ml), spectinomycin (100 μg/ml), or tetracycline (6 μg/ml). MacConkey agar plates with 1% galactose or M63 [22 mM KH₂PO₄, 40 mM K₂HPO₄, 15 mM (NH₄)₂SO₄, 0.5 mg/liter FeSO₄] agar plates supplemented with 0.2% galactose and 0.1% Casamino Acids were used to test whether a strain could metabolize galactose.

TABLE 1.

Strains used in this study

Strain	Genotype or description	Reference or source
EDL933	O157:H7	21
TEA007	EDL933 galE::pTHE001	This study
TEA023	EDL933 ΔgalU::aad-7	This study
TEA026	EDL933 ΔgalETKM::aad-7	This study
TEA028	EDL933 ΔgalETKM::tetA	This study
TEA040	EDL933; Gal+ revertant of TEA026	This study
CAG5051	HfrH nadA57::Tn10 thi-1 relA1 spoT supQ80	28
MC4100	araD139 Δ(arg-lac)U169 rpsL150 relA1 flbB5301 fruA25 deoC1 ptsF25	Laboratory stock
EDL933/pKD46	EDL933/pKD46 (oriR101 bla Para-λ gam bet exo)	This study
BW19851/pGP704	RP-4-2-Tc::Mu-1 kan::Tn7 integrant; creB150 hsdR17 endA zbf-5 uidA(ΔMlu1)::pir(wt) recA1 thi/pGP704 (oriR6K mobRP4 Apr)	16
WM3064/pTHE001	thrB1004 pro thi rpsL hsdS lacZΔM15 RP4-1360 Δ(araBAD)567 ΔdapA1341::[erm pir(wt)]/pTHE001 (oriR6K mobRP4 AprgalEEDL933)	This study
DH5αλpir+/pVi36	endA1 hadR17 thi-1 recA1 Δ(lacIZYA-argF)U169 deoR (p80 lacZΔM15) gyrA relA1/pVi36 (oriR6K aad-7)	V. Burrus
BW23473/pAH162	Δ(lacIZYA-argF)U169 rph-1 rpoS396(Am) robA1 creC510 hsdR514 ΔendA95 uidA(ΔMlu1)::pir(wt) recA1/pAH162 (tetA)	11

To generate the deletion-insertion mutations in the gal genes, we used a one-step gene inactivation method adapted from the method of Datsenko and Wanner (5). In this method, a temperature-sensitive plasmid (pKD46) encoding lambda red recombinase was transformed into EDL933. To make the ΔgalU::aad-7 mutant TEA023 and the ΔgalETKM::aad-7 mutant TEA026, the spectinomycin resistance gene (aad-7) was amplified from the pVi36 plasmid (provided by V. Burrus, University of Sherbrooke) template using primers TE139 (5′-ATGGCTGCCATTAATACGAAAGTCAAAAAAGCC) and TE140 (5′-TTACTTCTTAATGCCCATCTCTTCTTCAAGCCA) and primers TE137 (5′-ATGCTATGGTTATTTCATACCATAAGCCTAATGGAGCCCGGCGGATTTGTCCTACTC) and TE138 (5′-TTACTCAGCAATAAACTGATATTCCGTCAGGCTCTAAGCACTTGTCTCCTGTTTA), respectively. To obtain ΔgalETKM::tetA mutant TEA028, the tetracycline resistance gene (tetA) was amplified from the pAH162 plasmid (11) template using primers TE141 (5′-ATGCTATGGTTATTTCATACCATAAGCCTAATGGAGGATGCCTGGCAGTTCCCTACTC) and TE142 (5′-TTACTCAGCAATAAACTGATATTCCGTCAGGCTTTAGGTGGCGGTACTTGGGTCGA). After electroporation of the PCR products, cells were incubated in SOB (Invitrogen) containing 0.2% l-arabinose for 2 h and then plated on selective media at 37°C. For the ΔgalU::aad-7 mutation, the spectinomycin resistance gene replaced all of the galU gene except the first 33 bp and the last 30 bp of the galU open reading frame. For the ΔgalETKM::aad-7 and ΔgalETKM::tetA mutations, the antibiotic resistance gene replaced all of the galETKM operon except the first 36 bp of the galE gene and the last 30 bp of the galM gene.

We constructed the pTHE001 plasmid to generate an insertion mutation in galE. First, a 460-bp internal fragment of the galE gene was amplified by PCR using primers TE013 (5′-GCAAGGATCCGACGTTTGTTGAAGGCGATA) and TE014 (5′-GGCATAAGGGAATTCGGAATGCCTTGCGGA). The PCR product was digested with BamHI and then cloned into the BglII site of the conditional plasmid pGP704 (16). The resulting plasmid, pTHE001, was mobilized using the RP4⁺ helper strain WM3064 (provided by W. Metcalf, University of Illinois, Urbana-Champaign) into EDL933.

To generate the Gal⁺ revertant TEA040, we first moved the galE::pTHE001 mutation from TEA007 into the ΔgalETKM::aad-7 strain (TEA026) by P1 transduction, selecting for ampicillin resistance. The resulting strain was then used as a recipient for P1 transduction of the galETKM⁺ allele from TEA023. Gal⁺ transductants were selected on M63 agar plates supplemented with 0.2% galactose and 0.1% Casamino Acids.

P1 adsorption and sensitivity assays.

P1 adsorption assays were performed using a P1 lysate grown on E. coli K-12 strain MC4100. Approximately 100 μl of an overnight culture was pelleted by centrifugation and resuspended in 100 μl of LB broth. Then 100 μl of the P1 lysate was added to the cells. After 15 min of incubation at 37°C, cells were pelleted by centrifugation at 6,000 rpm in an Eppendorf centrifuge at 4°C for 2 min. The P1 titers of the supernatants were then determined by plaquing, using MC4100 as an indicator strain. To plaque P1, we spotted phage lysates on top agar (LB broth containing 2 mM MgSO₄, 10 mM CaCl₂, and 0.7% agar) lawns of MC4100.

To test for sensitivity of strains to P1 lysis, each strain was cross-streaked against P1. A single line of P1 (100 μl; ∼10⁹ PFU/ml) was allowed to dry on an LB agar plate. For each bacterial strain, a single streak was then drawn perpendicular to a line of phage P1. Strains resistant to P1 grew on both sides of the line of P1; susceptible strains were partially lysed following an encounter with P1.

P1 lysate production.

P1 lysates of various EHEC strains were generated by growing 1:100 dilutions of overnight cultures of each strain in 2.5 ml of LB broth containing 2 mM MgSO₄ and 10 mM CaCl₂ and incubating the preparations for 1 h at 37°C with agitation. Then 100 μl of a P1 lysate grown on MC4100 was added to each culture. After 2 to 3 h of incubation at 37°C with agitation, lysis of the cultures was observed. Tubes were transferred to ice, and any remaining intact bacteria were lysed with 0.5 ml chloroform. Lysates were then centrifuged at 13,000 rpm for 1 min, diluted in phosphate-buffered saline (PBS), and spotted on top agar lawns of MC4100 for titration. Lysates were stored at 4°C in the dark with 0.5 ml chloroform.

P1 transduction.

Overnight cultures (0.5 ml) of recipient bacteria grown in LB broth were pelleted and resuspended in 100 μl MC (5 mM MgSO₄, 50 mM CaCl₂). About 50 μl of P1 lysate was added to the cells, which were then incubated at 37°C for 15 to 30 min. LB broth with 10 mM sodium citrate (0.5 to 1 ml) was added to each tube and incubated for 1 h at room temperature. Each tube was centrifuged at 6,000 rpm for 2 min, and the pellets were resuspended in 100 μl 1 M sodium citrate and plated on selective media.

In vitro competition assays.

A 1:1 mixture of a mutant (TEA026) and wild-type strain EDL933 or TEA040 initially containing 5 × 10⁷ bacteria/ml was incubated in LB broth at 37°C with agitation overnight. Each assay mixture was then diluted and plated on LB agar. After overnight growth, bacteria were replica plated on selective media to determine the numbers of mutant and wild-type bacteria. Each assay was performed at least three times.

Competition assays with infant rabbits.

We used the infant rabbit model to test the colonization ability of EHEC strains (27). Three-day-old New Zealand White rabbits were orally inoculated with a 1:1 mixture of TEA026 and wild-type strain EDL933 or TEA040 containing 2.5 × 10⁸ bacteria which were washed one time and resuspended in PBS. Seven days after inoculation, rabbits were sacrificed, and their gastrointestinal tracts were removed. Portions of the ileum, midcolon, and cecum were then homogenized, diluted, and plated on MacConkey agar containing sorbitol. After overnight growth, bacteria were replica plated on LB agar containing spectinomycin (100 μg/ml) to determine the number of TEA026 bacteria.

BPI sensitivity assays.

We used the BPI-derived peptide P2, which has the antibacterial activity of whole protein (10, 13), to assess EHEC sensitivity. For these assays, bacteria grown overnight in LB broth were washed once in PBS and resuspended in PBS (pH 6.2). Then 5 × 10⁷ bacteria with or without 30 μg P2 (Tufts University Core Facility) were incubated in 0.5 ml PBS (pH 6.2) for 45 min at 37°C. After incubation, each assay mixture was placed on ice, diluted, and plated on LB agar. Each assay was performed in triplicate and repeated in three independent experiments.

RESULTS AND DISCUSSION

O157:H7 gal mutants can adsorb P1 and are sensitive to P1 lysis.

In many enteric bacteria, extensive LPS O antigens obscure the LPS core oligosaccharide and thereby prevent adsorption of and lysis by phage P1. Since synthesis of the O157:H7 O antigen requires modification of galactose by the galE, galT, galK, and galU gene products, we hypothesized that EHEC strains with mutations in these genes would produce no O antigen and therefore would be far more susceptible to P1 infection. To test this hypothesis, we compared phage infection of and adsorption by the sequenced O157:H7 isolate EDL933 to those of four gal derivatives: the ΔgalETKM::aad-7 (TEA026), ΔgalETKM::tetA (TEA028), and ΔgalU::aad-7 (TEA023) deletion mutants and the galE::pTHE001 (TEA007) insertion mutant.

To assess whether gal loci influence infection of EHEC by phage P1, we performed cross-streak experiments. A streak of each bacterial strain was drawn perpendicular to a line of P1, and the consequences of encountering phage were assessed. There was no change in the growth of wild-type strain EDL933 in response to P1, indicating that this strain is resistant to phage infection. In contrast, all four gal mutants were lysed by phage P1 (data not shown).

To explore why the gal mutants are more susceptible to P1 infection, we examined whether the mutants showed enhanced adsorption of P1. Phage adsorption to the host bacterium is an essential step in phage infection. In these assays, each gal mutant was incubated with P1 and then centrifuged to pellet any phage adsorbed to the bacteria. Supernatants were then assayed to determine the number of unadsorbed P1 particles. As shown in Table 2, the wild-type strain adsorbed ∼30% of the phage, suggesting that there may be some nonspecific interactions between EHEC and P1, since the wild-type strain is resistant to P1 infection and therefore is presumed to have inaccessible core oligosaccharides. However, all four gal mutants adsorbed ∼95% of the P1, which is consistent with the hypothesis that access to P1's receptor is less impeded in these strains.

TABLE 2.

P1 adsorption and P2 sensitivity of EHEC strains

Strain	% Adsorbed P1a	% Survival after P2 challengeb
EDL933 (wild type)	27.4	38
TEA007 (galE::pTHE001)	93.5	0.054
TEA023 (ΔgalU::aad-7)	96.6	0.29
TEA026 (ΔgalETKM::aad-7)	95.2	0.014
TEA028 (ΔgalETKM::tetA)	98.4	0.35
TEA040 (Gal+ revertant)	15.6	32

^aThe percentage of adsorbed P1 was determined as follows: (number of P1 plaques obtained after incubation with bacteria/number of P1 plaques obtained after incubation without bacteria) × 100%. Each experiment was performed in triplicate, and the data are means. The P value for a comparison of each of the gal mutants and the wild type was less than 0.006. The P value for a comparison of TEA040 and the wild type was 0.65.

^bThe percentage of survival after P2 challenge was determined as follows: (number of bacteria after incubation with P2/number of bacteria after incubation without P2) × 100%. Each experiment was performed in triplicate, and the data are means. The P value for a comparison of each of the gal mutants and the wild type was less than 0.03. The P value for a comparison of TEA040 and the wild type was 0.10.

Antibiotic resistance markers can be transduced between EHEC O157:H7 gal mutants by P1.

We next assessed whether EHEC gal mutants could serve as recipients or donors in P1 transduction. As expected, a nadA::Tn10 marker from an E. coli K-12 strain (CAG5051) could not be transduced into the wild-type EDL933 strain by P1. In contrast, this marker could be transduced into the ΔgalU::aad-7 (TEA023), ΔgalETKM::aad-7 (TEA026), and galE::pTHE001 (TEA007) strains. To determine if chromosomal markers could be transduced from the EHEC gal mutants, each of the mutants was used to generate a P1 lysate. The ΔgalU::aad-7, ΔgalETKM::aad-7, and galE::pTHE001 mutations could be transferred readily into the O157:H7 ΔgalETKM::tetA strain or into E. coli K-12 strain MC4100. For each of these EHEC P1 transduction experiments, between 10 and 100 transductants were obtained. Clearly, O157:H7 gal mutants can be transduced by P1 and can be used to generate P1 lysates. However, the frequencies of transduction between the O157:H7 gal strains were ∼100-fold lower than the frequency of P1 transduction between K-12 strains. The lower O157:H7 transduction efficiency could be due to the abundance of prophage genes in O157:H7 that may affect replication or production of P1.

Reversion of the gal mutant using P1 transduction.

We devised a simple strategy to change Gal⁻ strains back to Gal⁺ strains in order to enable study of Gal⁺ EHEC strains that have been engineered using P1. We were unable to revert the ΔgalETKM::aad-7 mutant (TEA026) by transducing the galETKM⁺ allele from TEA023, perhaps because the frequency of transduction was too low to obtain a revertant. Therefore, we utilized a two-step reversion method. We first moved the galE::pTHE001 mutation from TEA007 into the galETKM strain (TEA026). This strain was then used as a recipient for P1 transduction of the galETKM⁺ allele from a donor lysate grown on the ΔgalU strain (TEA023). Gal⁺ transductants were selected on agar plates containing M63 minimal medium supplemented with 0.2% galactose and 0.1% Casamino Acids and were verified to be Gal⁺ strains on MacConkey agar containing 1% galactose. This transduction was as efficient as transfer of other EHEC chromosomal markers. We chose one Gal⁺ revertant strain (TEA040) to test for P1 sensitivity. Like the isogenic wild-type strain, this Gal⁺ revertant had a reduced capacity to adsorb P1 phage compared to the gal mutants (Table 2), was resistant to P1 lysis in a P1 cross-streak experiment, and was not able to serve as a recipient for P1 transduction (data not shown).

O157:H7 galETKM mutant is dramatically impaired in colonization of the infant rabbit intestine.

Previous studies have shown that galE mutants of S. enterica serovars are defective in intestinal colonization (9, 12). We were interested in determining whether the EHEC galETKM deletion resulted in a similar defect in intestinal colonization. To do this, we tested the galETKM::aad-7 deletion mutant (TEA026) in a competition assay with the isogenic wild-type strain (EDL933) using the EHEC-infant rabbit model. We found that the galETKM deletion mutant (TEA026) was ∼500-fold less able to colonize the infant rabbit ileum, cecum, and midcolon (Fig. 1A). To demonstrate that this dramatic defect in intestinal colonization was due to the galETKM deletion, we performed a competition experiment with the Gal⁺ reverted strain and its ΔgalETKM::aad-7 parent strain. The Gal⁺ revertant (TEA040) outcompeted the galETKM mutant (TEA023) to an extent similar to the extent observed for the wild type (Fig. 1B), strongly suggesting that the galETKM deletion accounts for the colonization defect of TEA026. In vitro competition assays in which the galETKM strain and the wild type or the Gal⁺ revertant were grown in LB broth at 37°C revealed that the galETKM mutation resulted in a slight (∼2-fold) but statistically significant growth defect in rich medium (Fig. 1). This minor in vitro growth defect could not account for the drastic colonization defect of the gal mutant. Overall, these findings suggest that the O157 O antigen is critical for EHEC intestinal colonization.

FIG. 1.

In vivo and in vitro competition assays with an EHEC galETKM mutant (TEA026) and wild-type strain EDL933 or with TEA026 and a Gal⁺ revertant of TEA026 (TEA040). Equal numbers of TEA026 and EDL933 (A) or TEA026 and TEA040 (B) were coinoculated into infant rabbits or into LB broth. After 7 days of infection, sections of the ileum, midcolon, and cecum and samples of stools were processed to determine competitive indices. The in vitro competition assays in LB broth were carried out for 16 h. The competitive index was determined as follows: [(number of output mutant bacteria)/(number of output wild-type or revertant bacteria)]/[(number of input mutant bacteria)/(number of input wild-type or revertant bacteria)]. Each of the competitive indices shown is statistically different from 1 (P < 0.00001, as determined by Student's t test).

O157:H7 gal mutants are more sensitive to BPI killing.

Host organisms often produce antimicrobial peptides, such as cryptidins, as effectors of innate immunity. BPI is a 55- to 60-kDa protein that is found in the blood and on surfaces of epithelial cells throughout the gastrointestinal tract (2). BPI binds the lipid A component of LPS and has potent bactericidal activity against gram-negative bacteria (7). Given that O157 gal mutants likely lack O antigen, we hypothesized that these mutants would be more sensitive to BPI than the wild-type strain, since the lipid A portion of the gal strains' LPS would be more accessible to BPI binding. We used P2, a peptide that contains BPI residues 86 to 104 and that has the same cytotoxic activity as the entire protein, to test if the O157:H7 gal mutants had increased susceptibility to BPI. Bacteria were incubated in the presence or absence of P2, and then the number of bacteria was determined. Each of the gal mutants was far more sensitive to P2 than the wild-type EHEC strain or the Gal⁺ revertant (Table 2). These data suggest that O157 O antigen is important for EHEC resistance to BPI.

Conclusions.

EHEC strains, especially E. coli O157 strains, are significant causes of disease in many developed nations, yet there are no methods for transducing markers in this important group of pathogenic E. coli. It has been known for a long time that S. enterica serovar Typhimurium gal mutants have little or no O antigen, which makes these strains sensitive to P1 (18). Our findings strongly suggest that the O157 O antigen obscures the EHEC core oligosaccharide, the P1 receptor, in a similar fashion, since we found that O157:H7 gal mutants are sensitive to P1 infection. We also developed a relatively simple P1 transduction-based means to revert gal mutants to a Gal⁺ phenotype. Although we describe a method to genetically manipulate O157:H7 with P1, this technique can likely be adapted for use with other EHEC serotypes and probably with other pathogenic E. coli strains (such as enteropathogenic E. coli and uropathogenic E. coli) as well. Figure 2 outlines a general scheme for using P1 transduction for genetic manipulation of EHEC. This figure shows the method that our lab routinely uses to generate isogenic EHEC strains. This ability to move chromosomal markers enables generation of isogenic strains containing single or multiple mutations. Given the relative ease of the P1 transduction technique outlined here, it should be possible to backcross mutations generated with the lambda red system into genetically identical backgrounds.

FIG. 2.

General scheme for P1-facilitated genetic engineering of EHEC. (A) The desired mutation in the ΔgalETKM::tetA or ΔgalETKM::aad-7 background is generated (e.g., using the lambda red recombination system) so that the mutation can be moved by P1 transduction. (B) The desired mutation is moved into the galE::pTHE001 background using P1 transduction, selecting for the antibiotic resistance gene to generate an isogenic backcrossed mutant. (C) A second P1 transduction into the backcrossed strain is performed, selecting for growth on galactose minimal plates. Abbreviations: x, gene of interest; ab^R, antibiotic resistance gene; tet^R, tetracycline resistance gene.

Our observations revealed that the O157 O antigen is an important EHEC colonization factor. Tissue culture experiments have suggested that the O157 O antigen is not likely to be involved in EHEC adherence to intestinal epithelial cells and may reduce EHEC type 3 secretion of effectors that promote adherence (1, 4, 19). The sensitivity of the O157 gal mutants to BPI supports the idea that at least part of the colonization defect of the galETKM mutant is explained by its reduced resistance to host antimicrobial factors. Although O157 strains are most commonly associated with EHEC-related illness in North America, Europe, and Japan, other EHEC serotypes (O26, O91, O103, and O111) can cause significant human disease (20). It will be interesting to assess whether O antigens from different EHEC serotypes provide equal resistance to host killing and act as effectively as the O157 O antigen in intestinal colonization. In fact, differences in the LPSs of the O1, O7, and O8 serotypes of K1 E. coli correlate with differences in the virulence of these serotypes (24). The P1 transduction method described here should facilitate the construction of isogenic strains which vary only in their O antigens, thus enabling exploration of the role of particular O antigens in EHEC virulence.

O antigens likely act as important barriers against extracellular assaults mounted by the host immune system. O-antigen mutants of enteric bacteria like V. cholerae, Klebsiella serotype O1:K20, and Shigella flexneri are more susceptible to antimicrobial peptides and complement killing (14, 17, 29). S. enterica serovar Typhi, Brucella melitensis, and V. cholerae O-antigen mutants have been shown to be impaired in intestinal colonization as well (3, 9, 17, 26). In fact, the only oral live attenuated vaccine against S. enterica serovar Typhi used in the United States is a galE mutant. Given the pronounced attenuation of the O157:H7 ΔgalETKM::aad-7 mutant, perhaps EHEC gal mutants could also be developed into useful vaccines.