Epithelial attachment alters the outcome of Helicobacter pylori infection

Abstract

Genetically defined in vivo models are needed to assess the importance of target cell attachment in bacterial pathogenesis. Gastric colonization by Helicobacter pylori in human populations is common and persistent, and has various outcomes including peptic ulcers and cancer. The impact of attachment on the course of infection was examined in transgenic mice expressing a human receptor for H. pylori in their gastric epithelium. Persistent infection by a clinical isolate occurred at comparable microbial densities in transgenic and nontransgenic littermates. However, microbial attachment in transgenic mice resulted in production of autoantibodies to Lewis^x carbohydrate epitopes shared by bacteria and acid-secreting parietal cells, chronic gastritis, and parietal cell loss. This model should help identify bacterial and host genes that produce attachment-related pathology.

Helicobacter pylori colonizes the stomachs of at least half of all humans, surviving largely within the gastric mucus without attaching to host cells (1). The natural history of colonization involves an initial period of rapid bacterial growth associated with gastritis and reduced acid production. In most hosts, the inflammatory response diminishes within several months, leaving an asymptomatic diffuse gastritis with normal acid production that persists for years (2, 3). A subpopulation of colonized individuals develops peptic ulcers (4). In some hosts, inflammation progresses over 3–4 decades, leading to atrophic gastritis that can evolve to adenocarcinoma. H. pylori is estimated to contribute to 30–90% of gastric carcinomas and is a risk factor for mucosal lymphomas (5, 6).

The microbial and host factors that determine the outcome of colonization have been difficult to define, in part because both H. pylori and humans are genetically diverse. One determining factor may be the ability of the microbe to establish physical contact with the gastric epithelium, thereby affecting its ability to influence mucosal epithelial and immune cell populations (7). A number of molecules have been implicated as receptors for H. pylori adhesins (8). Transgenic mice expressing one such receptor, the human blood group antigen Lewis^b (Le^b), have now been used to define the consequences of attachment of clinical isolates to the epithelium. The mouse model is based on the following. In human stomachs, H. pylori is associated with mucus-producing pit cells and their more mature surface mucous cell descendants (9). In vitro binding assays have shown that attachment to these cells can be mediated by epithelial receptors containing Le^b (10). The FVB/N inbred strain of mice expresses lacto-N-fucopentaose I-like oligosaccharides (Fucα1,2Galβ1,3GlcNAcβ) in their pit cell lineage, but not Le^b [Fucα1,2Galβ1,3(Fucα1, 4)GlcNAcβ] (11). Lacto-N-fucopentaose I functions as an acceptor molecule for GDP-β-l-fucose in the synthesis of Le^b by the human Lewis enzyme, which is an α 1,3/4 fucosyltransferase. Production of Le^b glycoproteins was successfully engineered in the pit and surface mucous cells of FVB/N transgenic mice using a fatty acid binding protein gene promoter linked to the α 1,3/4 fucosyltransferase. Le^b-positive pit/surface mucous cells were distributed throughout the glandular epithelium in 1 day to 2 year old transgenic animals. Le^b was absent in all other gastric cell lineages (11). As described below, we have used this model to show that interactions between a bacterial adhesin and a host receptor can alter disease outcome in an animal without affecting colonization levels, thereby demonstrating that the mode of colonization can be crucial for pathogenesis.

MATERIALS AND METHODS

Clinical Isolates.

Gastric biopsies were obtained from adult Peruvians who had clinical and endoscopic evidence of gastritis. The biopsies were used to confirm that they were Le^b-positive (11) and to isolate H. pylori. Mascerated tissue samples were streaked on selective medium (Brain Heart Infusion agar containing 10% calf blood/6 μg/ml vancomycin/5 μg/ml trimethoprim/8 μg/ml amphotericin B). Single colonies were picked after a 3–5 day incubation at 37°C in CampyPakPlus envelopes (Becton Dickinson) and recultured. Isolates were identified as H. pylori by their morphology and their production of urease and catalase (12).

Animals.

Transgenic mice were generated and genotyped as described (11). They were maintained, together with their normal littermates, in a specified pathogen free state and given Pico Lab Rodent Diet 20 (Purina Mills). Gastric epithelial patterns of Lewis epitope production were defined by using a panel of mAbs and immunohistochemical protocols detailed in earlier publication (11, 13).

In Vitro Bacterial Binding Assays.

Formalin-fixed normal human stomachs were obtained from the surgical pathology files of Barnes Hospital (St. Louis, MO). Stomachs from transgenic mice and their normal littermates were fixed in 10% formalin/PBS or frozen and fixed in methanol. Bacteria were cultured in selective medium at 37°C, harvested 1, 3, and 6 days later, labeled with fluorescein isothiocyanate, and applied to tissue sections according to ref 11. Parallel experiments used unlabeled bacteria that were detected, after incubation with the sections, using rabbit antibodies to H. pylori surface proteins (Dako). Blocking controls were performed by pre-incubating labeled or unlabeled bacteria with purified Lewis-human serum albumin (HSA) neoglycoconjugates (11, 13). Neoglycoconjugates were obtained from IsoSep (Tullinge, Sweden) and typically contained 20–30 oligosaccharides covalently linked to each molecule of HSA.

Infections.

Clinical isolates were plated on selective medium and incubated for 3 days at 37°C. 10⁷ colony-forming units (cfu) (A₆₀₀ = 0.1) of each isolate (n = 8) were mixed together in 200 μl tryptic soy broth. Organisms were introduced into stomachs by gavage after mice had been placed in metabolic cages for 4 h to allow emptying of their gastric contents. Four hours to 16 weeks later, animals were placed in metabolic cages for 4 h and killed. One half of each stomach was used for histochemical analysis. The other half was homogenized in 0.5 ml tryptic soy broth and serial dilutions of the suspension were plated on selective medium supplemented with polymyxin B (3.3 μg/ml), bacitracin (200 μg/ml), and nalidixic acid (10 μg/ml). Following a 3-day incubation, cfu were defined. Survival rates were the same between transgenic and normal mice after 8 and 16 weeks of infection (88–92%).

PCR-Based DNA Fingerprinting.

Fingerprinting was performed (14) by using genomic DNA prepared from three individual colonies, and from a full plate of colonies (≈10⁵ cfu), cultured from each stomach. Results were compared with the PCR fingerprints of each input isolate.

ELISA Assays.

Protocols described in ref. 15 were used to define lipopolysaccharide-associated Lewis epitopes in H. pylori isolates and to measure circulating antibodies to Lewis epitopes in mouse sera.

Analysis of Host Immune Responses.

Stomachs were sectioned parallel to their cephalocaudal axis (≥100 serial 5 μm sections per stomach). Hematoxylin and eosin stained sections were scored for inflammatory changes by three observers in a single blinded fashion. Frozen gastric sections were stained with rat or hamster monoclonal antibodies to mouse CD4, CD8, αβT cell receptor, γδT cell receptor, and CD45R/B220 (PharMingen). Antigen-antibody complexes were detected with tyramide signal amplification (16).

Immune cells were recovered from gastric mucosa pooled from nine transgenic and six normal mice and analyzed by fluorescence-activated cell sorter. Spleens were removed from the infected mice, pooled from each group, and single cell suspensions prepared. Bacteria were cultured in selective medium for 3 days, heat-inactivated, and 10²–10⁵ organisms were incubated with 5 × 10⁶ splenocytes/ml in a T cell proliferation assay (17). To detect parietal cell autoantibodies, sera from infected normal and transgenic mice were obtained at the time of death, diluted 1:50 in PBS/blocking buffer (11) and applied to formalin-fixed sections of normal uninfected FVB/N stomachs. Bound antibodies were detected with Cy3-conjugated sheep anti-mouse Ig. Control experiments established that the secondary antibody did not react with any gastric epithelial lineage. Blocking controls were performed by incubating an aliquot of each serum (100 μl of a 1:50 dilution in PBS/blocking buffer) with Le-HSA neoglycoconjugates (10 μg), or with 10⁷ Hp1 cells, overnight at 4°C prior to application to sections.

Statistical Analysis.

The statistical significance of observed differences between transgenic and nontransgenic mice were analyzed by Student’s t test.

RESULTS

Le^b-Dependent Binding of Clinical Isolates of H. pylori to Human Pit and Surface Mucous Cells in Vitro Is Reproduced in Transgenic Mice.

Isolates of H. pylori were obtained from patients with gastritis living in Lima, Peru. H. pylori is endemic in this region and the Le^b phenotype is predominant in Peruvians (18). Therefore, these patients should harbor bacteria that have undergone in vivo selection (15) for their ability to bind Le^b. H. pylori isolates were recovered from gastric biopsies obtained from 11 patients who expressed Le^b in their pit and surface mucous cells. Vacuolating cytotoxin (VacA) induces acidic vacuoles in epithelial cells and is a virulence factor (19). Cytotoxin-associated protein (CagA) is encoded by a gene located at the end of a pathogenicity island (cag) and is considered to be a marker of virulence although its own role is unknown (19, 20). PCR studies revealed that all 11 isolates were vacA⁺ (toxigenic type MI allele) and cagA⁺.

Each of the 11 isolates was cultured for 1, 3, and 6 days, and then applied to fresh frozen or formalin-fixed sections of transgenic and normal mouse stomachs (n = 10 animals/group) and formalin-fixed sections of 10 normal Le^b-positive human stomachs. All isolates bound to both human and transgenic mouse stomachs. Binding was limited to Le^b-positive pit and surface mucous cells and was blocked by pre-treatment of the isolates with Le^b-HSA neoglycoconjugate but not by Le^a-HSA, Le^x-HSA, Le^y-HSA, or HSA alone (Fig. 1 A–C). Binding was not affected by bacterial growth phase, or by the presence or absence of flagella. None of the isolates bound to nontransgenic Le^b-negative stomachs.

Figure 1.

In vitro and in vivo binding of H. pylori to Le^b-producing pit and surface mucous cells. (A) Section of a transgenic mouse stomach incubated with a Le^b-specific mAb (detected with Cy3-conjugated donkey anti-mouse Ig). Le^b epitopes (red) are confined to the pit/surface mucous cell lineage. (B) Multilabel analysis of the same section as in A incubated with fluorescein isothiocyanate-conjugated isolate Hp1 and a nuclear stain (bis-benzimide; blue). Bacteria (yellow) are bound to Le^b-positive cells (red). (C) Pre-treatment of Hp1 with Le^b-HSA blocks its ability to bind to an adjacent section prepared from the mouse in (A, B). (D) There are no statistically significant differences in cfu/stomach between transgenic (black bars) and normal (gray bars) mice at any time during persistent infection. Mean values ± 1 SD are plotted. (E and F) Frozen sections of an unperfused transgenic mouse stomach following an 8-week infection. (E) Double exposure after staining with Le^b mouse mAb, fluorescein isothiocyanate-conjugated donkey anti-mouse Ig, rabbit antisera against H. pylori surface proteins, and Cy3-conjugated sheep anti-rabbit Ig. (F) Triple exposure of another section, incubated with the reagents in E plus the nuclear stain bis-benzimide (blue). Bacteria (yellow-orange) are attached to Le^b-positive pit and surface mucous cells (green; e.g., solid arrows in E), and are also affiliated with the gastric mucus (e.g., open arrow in E). (G and H) Frozen sections from an 8-week infected nontransgenic mouse, processed as in E and F, respectively. In the Le^b-negative stomach, bacteria (red) are only evident in the luminal mucus (open arrow in G) and not associated with the epithelium (H). (Bars = 25 μm.)

A Comparable High-Density Infection Is Established in Normal and Transgenic Mice but with Sustainable Differences in Bacterial Distribution.

Eight of the isolates were pooled (10⁷ cfu per isolate) and administered in a single gavage to 3–4-month-old transgenic and normal mice. Animals were killed 4 h, 1 day, and 1, 4, 8, or 16 weeks later (9–11 mice/group/time point/experiment; 2–3 experiments/time point). Gastric contents were cultured on selective medium. Four hours after gavage, the number of viable organisms was 7 orders of magnitude less than in the inoculum. There was a marked increase in the number of cfu recovered from the stomachs of mice killed over the course of the next 4 weeks. There was no statistically significant difference in the percentage of transgenic and normal animals infected 8 or 16 weeks after inoculation (89 vs. 80%, respectively, at 8 weeks; 75 vs. 73% at 16 weeks), nor was there a significant difference in the total number of viable organisms in their stomachs (Fig. 1D). In contrast, there was a distinct difference in the distribution of the organisms: after 8 and 16 weeks, bacteria were only seen in the mucus layer in normal animals while in transgenic mice, bacteria were associated with both the mucus layer and pit/surface mucous cells (Fig. 1 E–H). PCR fingerprinting of H. pylori recovered after 8 and 16 weeks of infection revealed that only one of the eight input isolates (Hp1) was present in transgenic and normal stomachs (10 animals/group).

Bacterial Attachment Affects the Cellular and Humoral Immune Responses of the Host.

After 8 weeks of infection, a patchy chronic active gastritis was evident throughout the glandular epithelium of normal and transgenic mice (n = 28 animals/group). Gastritis was not present in the stomachs of controls that had been gavaged with medium alone (n = 20/group). Surveys of hematoxylin and eosin-stained serial sections revealed that the severity of the inflammation was worse in transgenic animals (Fig. 2 A, B), and that the gastritis was distributed over a greater area of the corpus of the stomach. Immunohistochemical and fluorescence-activated cell sorter analyses disclosed that ≈50% of the lymphocytic infiltrate in both groups was composed of αβ T cell receptor-positive T cells and 10% of B cells. However, there was a 2–3-fold greater fractional representation of macrophages and NK1.1-positive natural killer cells in infected transgenic hosts. In vitro stimulation of splenic T cells by Hp1 produced a significantly greater proliferative response when cells were obtained from infected transgenic vs. normal animals (Fig. 2C).

Figure 2.

Bacterial binding to the gastric epithelium of transgenic mice is associated with enhanced recognition of bacterial antigens and chronic gastritis. (A and B) Hematoxylin and eosin-stained sections from transgenic mice prior to (A) and 8 weeks after infection (B). (A) Parietal and zymogenic cells are indicated by arrows and arrowheads. (B) Infected mice develop gastritis with numerous granulocytes (e.g., arrows) and lymphocytes (arrowhead), loss of parietal and zymogenic cells, plus epithelial cell proliferation (e.g., open arrow) and nuclear atypia. (C) After 8 weeks of infection, a significantly greater (P < 0.05) proliferative response is seen upon exposure of transgenic vs. nontransgenic splenic T cells (black and gray bars, respectively) to varying amounts of heat-inactivated Hp1. Mean values ± 1 SD are shown. (D–G) Parietal cell autoantibodies in a transgenic mouse infected for 8 weeks. Serum from the animal was incubated with sections of stomach from a normal uninfected FVB/N mouse. (D) Section stained with fluorescein isothiocyanate-Dolicos biflorus agglutinin (DBA), a lectin that only binds to parietal cells. (E) Same section incubated with transgenic mouse serum and Cy3 donkey anti-mouse Ig. (F) Double exposure showing that the autoantibodies stain the cytoplasm of most parietal cells (orange). (G) Binding of autoantibodies to parietal cells is blocked by preincubation of the serum with Le^x-HSA. The section was processed as in (E). (H) Section prepared from an uninfected normal mouse stomach. The section was incubated with fluorescein isothiocyanate-conjugated Hp1, Le^x mAb (detected with Cy3-sheep anti-mouse Ig), and biotinylated DBA (detected with aminomethylcoumarin acetate-streptavidin). Bacteria (green) do not bind to parietal cells (blue) that express Le^x (magenta) but selectively adhere to pit and surface mucous cells. (I and J) H&E stained sections of a transgenic mouse stomach after 16 weeks of infection showing chronic gastritis with loss of parietal cells (I) and MALT (J). (Bars = 25 μm.)

Parietal cell autoantibodies have been correlated with the occurance of atrophic gastritis in H. pylori infected humans (21, 22). To screen for autoantibodies, sera from infected transgenic and normal mice were incubated with sections prepared from uninfected normal mouse stomachs. After 4 weeks of infection, neither group had detectable autoantibodies to any gastric epithelial lineage. After 8 weeks, 21 of 28 transgenic animals (75%) had autoantibodies that reacted with parietal cells (Fig. 2 D–F) compared with only 3 of 23 (13%) nontransgenic mice (P < 0.05). After 16 weeks, autoantibodies were detected in 87% of transgenic and 26% of nontransgenic animals (P < 0.05; n = 25/group). Uninfected normal and transgenic mice gavaged 8 or 16 weeks earlier with medium alone did not produce autoantibodies (10–15 animals/group/time point).

Parietal Cell Autoantibodies: Shared Le^x Immunodeterminants between Bacteria and Host Cells.

H. pylori contains two fucosyltransferase genes, one of which is an α 1,3 fucosyltransferase that produces Le^x (20, 23, 24). Five of the input isolates, including Hp1, expressed Le^x epitopes ex vivo. All eight isolates expressed Le^y. Production of these bacterial Lewis epitopes persisted after 8 and 16 weeks of infection in transgenic and normal mice (7–11 animals/group/time point). Bacterial Le^b and Le^a immunodeterminants were not detected before or after infection.

Binding of parietal cell autoantibodies was blocked by preincubation of sera with Le^x- and Le^y-positive Hp1 cells, or with Le^x-HSA (Fig. 2G). Binding was not blocked by Le^y-HSA, Le^b-HSA, Le^a-HSA, or HSA alone. At least 50% of parietal cells in uninfected normal and transgenic FVB/N mice have detectable Le^x immunodeterminants. Le^x is limited to parietal cells in both groups of mice, whether or not they are infected. None of the clinical isolates bound to Le^x-positive parietal cells in vitro (Fig. 2H). ELISA assays revealed a significantly higher titer of circulating antibodies to purified H. pylori lipopolysaccharide-associated Le^x and to Le^x-HSA in transgenic mice with parietal cell autoantibodies, compared with infected transgenic or normal mice without detectable autoantibodies (P < 0.05; 7–8 animals/group at the 8-week time point).

Gastritis, Parietal Cell Loss, and Mucosa-Associated Lymphoid Tissue (MALT) in Transgenic Animals.

Mice inoculated with hybridoma cells secreting antibodies to an H. pylori lipopolysaccharide-Lewis epitope that mimics parietal cell Lewis epitopes develop gastritis (25–27). In our model, development of parietal cell autoantibodies was directly correlated with more extensive and pronounced reactive atypia in the gastric epithelium, and more pronounced parietal cell loss. Areas of reactive atypia contain elongated and architecturally distorted gastric units (glands) with a general increase in proliferating epithelial cell populations, moderate to severe nuclear atypia, and altered nuclear to cytoplasmic ratios (Fig. 2B). A panel of antibodies and lectins (28) were used to establish that the foci of reactive atypia were associated with loss of parietal cells and a block in terminal differentiation of pepsinogen-producing zymogenic (chief) cells. These pathologic changes were not seen in uninfected transgenic mice.

The granulocytic component of the gastritis was lost and its lymphocytic component diminished in both groups of mice after 16 weeks of infection. Areas of mild reactive atypia lacking parietal and zymogenic cells were present in 33% of transgenic mice (Fig. 2J), but not in any normal littermates.

The normal mouse stomach lacks organized MALT. Persistent H. pylori infection in humans is associated with formation of MALT (29). MALT was not evident after 8 weeks of infection in transgenic or normal mouse stomachs. After 16 weeks, MALT, composed predominately of B220-positive B cells, was present in 8 of 15 (53%) transgenic animals (Fig. 2I), while smaller and less numerous mucosal lymphoid aggregates were present in only 3 of 15 (20%) normal mice (P < 0.05).

Clinical Isolate Hp1 Is Sufficient to Produce the Spectrum of Host Responses to Infection.

To determine whether infection with Hp1 alone could produce gastric pathology, or whether early collaborative interactions between several isolates are necessary, transgenic and normal mice were gavaged with 10⁷ cfu of Hp1 and killed 8 and 16 weeks later (10 mice/group/time point/experiment; n = 2 independent experiments). The percentage of mice in each group that were infected, the density and distribution of bacteria within their stomachs, the incidence of autoantibodies that bound to parietal cells in a Le^x-dependent manner, and the inflammatory responses were similar to that observed in mice gavaged with the mixture of eight isolates.

We have not yet defined the time course of predominance of Hp1, nor do we know why it becomes the predominant organism. Hp1 was obtained from a patient with active ulcer disease. Some of the other isolates had growth rates in vitro that were similar to Hp1. Nonetheless, Hp1 had one distinguishing phenotypic characteristic: when cultured, it was able to retain a helical (spiral) form to a greater degree than any of the other isolates. When initially isolated from patients, H. pylori is present in the helical form, which is considered to be the actively growing form of the organism (30, 31). Conversion to a coccoid form has been reported to occur ex vivo during the course of nutrient deprivation, or in vivo after treatment with antibiotics or inhibitors of parietal cell H⁺/K⁺ ATPase (32–35).

DISCUSSION

There are several reasons why H. pylori infection provides an excellent opportunity to study the role of attachment in bacterial pathogenesis. Colonization occurs frequently but with varied outcomes. Although bacteria can exist within the gastric mucus without host cell attachment, a variety of potential epithelial receptors for bacterial adhesins have been described. Genetic diversity among hosts and H. pylori could generate various combinations of adhesins and receptors within and between infected populations that could determine whether the organism attaches to the epithelium of a colonized individual or remains in the mucus.

Our study illustrates how it is possible to test the biological consequences of attachment. A putative receptor for H. pylori adhesins (Le^b) was produced in a clinically relevant target cell lineage in a genetically defined transgenic host. Nontransgenic FVB/N mice provided an ideal control: they were colonized with similar efficiency and microbial density for equivalent periods as their isogenic transgenic littermates, indicating that the receptor was not required to establish and maintain an infection. Distinct, sustainable differences in microbial compartmentalization occurred between normal and transgenic mice, establishing that the genetically engineered Le^b receptor was biologically active. In addition, both uninfected and infected normal and transgenic FVB/N mice constitutively express Lewis^x epitopes in their acid-producing parietal cells. Because these epitopes are often present in the lipopolysaccharide of clinical H. pylori isolates, they provide an opportunity to determine if bacterial attachment affects the host response to shared carbohydrate structures. This system differs from other reported models of H. pylori infection because microbial attachment to the host can be genetically manipulated while keeping the bacterial strain “constant” (36).

A microbial basis for autoimmunity in chronic inflammatory diseases is widely postulated but remains largely unproven. The effect of H. pylori attachment was to promote development of autoantibodies to acid-producing parietal cells, chronic gastritis, and loss of parietal cells. The mechanisms underlying autoantibody production likely include molecular mimickry between constitutively expressed bacterial lipopolysaccharide- and parietal cell-associated Le^x epitopes as well as facilitated presentation of the bacterial epitopes to a recruited gastric mucosal immune population. While we cannot rule out subtle attachment-induced alterations in the structure of Hp1 or host Le^x epitopes, comparisons of Hp1 recovered from normal and transgenic mice indicated that the presence or absence of Le^b in the host did not induce a change in the classes of Lewis immunodeterminants produced by the microbe. Moreover, the presence or absence of Hp1 did not elicit a change in Lewis immunodeterminants produced by transgenic or normal mice.

The results of a diphtheria toxin-induced ablation of parietal cells in another FVB/N transgenic mouse model (28) underscore the potential contributions of parietal cell destruction to H. pylori pathogenesis. The parietal cell is the only principal epithelial cell type that completes its differentiation within the stem cell zone (isthmus) of gastric units (glands) (37). The lineage ablation experiment disclosed that parietal cells are required to complete differentiation of zymogenic cells. Ablation of parietal cells produced augmented proliferation of stem cells in the absence of inflammation (28). Amplified stem cells and their immediate committed daughters expanded beyond their normal niche. Cells with the morphologic appearance of gastric stem cells and their immediate committed daughters eventually formed lesions with adenomatous features and extended through the muscularis mucosa into the submucosa (Q. Li, A. Syder, R. G. Lorenz, S. M. Karam and J. I. Gordon, unpublished manuscript). These findings, together with the results described in this report, suggest that if a host is colonized by a strain that expresses adhesins that promote attachment to epithelial receptors, and if that strain also expresses surface antigens that mimic host (parietal) cell structures, then the course of infection may be skewed toward autoimmunity, atrophic gastritis with parietal cell loss, and perhaps neoplasia.

Engineering carbohydrate receptors for bacteria in transgenic animals may be a generally useful approach for examining whether host cell attachment simply provides a more effective means for delivering constitutively expressed bacterial gene products, or whether intimate association with cellular targets serves to modify patterns of bacterial or host gene expression. The sequence of the H. pylori genome is now known (20). Further definition of the factors that mediate attachment-related H. pylori pathology should come from comparisons of gene expression and genome structure in isolate Hp1 recovered before and after infection of the Le^b transgenic mice and their normal littermates. Because Hp1 is genetically manipulatable, using these mice to assess the effects of disrupting H. pylori genes should also be informative.

ABBREVIATIONS

Le

Lewis blood group antigen

HSA

human serum albumin

MALT

mucosa-associated lymphoid tissue

cfu

colony forming units