Toxigenic Helicobacter pylori Infection Precedes Gastric Hypochlorhydria in Cancer Relatives, and H. pylori Virulence Evolves in These Families

Helicobacter pylori infection is a major risk factor for the development of distal gastric adenocarcinoma (1), but who develops this condition depends on host genetic susceptibility, bacterial virulence, and environmental cofactors. The best characterized host genetic factors are polymorphisms leading to increased interleukin 1β (IL-1β) production in response to infection (2, 3). These polymorphisms are even more strongly associated with gastric hypochlorhydria and atrophy (2). This supports the contention that hypochlorhydria and atrophy are precursor conditions of gastric cancer (4) and suggests that cytokine polymorphisms predispose to gastric cancer through predisposing to these precursor conditions.

H. pylori virulence factors associated with gastric cancer risk include toxic s1/m1 forms of the vacuolating cytotoxin (VacA; refs. 5–7), possession of the cytotoxin-associated gene (cagA; refs. 8, 9), and possession of a cagA allele encoding four or more tyrosine phosphorylation motifs (TPM; refs. 10–12). The vacA gene is polymorphic; it encodes VacA that can be either type s1 or s2 in the signal region and either type m1 or m2 in the midregion (13). Each strain has one vacA allele, and this can be type s1/m1, s1/m2, s2/m1 (rarely), or s2/m2 (13). Type s2 alleles encode a nontoxic form of VacA, and s1/m1 alleles encode a form which binds to and vacuolates a wider range of epithelial cells than s1/m2 (14). The cag pathogenicity island (PaI) encodes a type IV secretion system (15) that can penetrate epithelial cells and deliver the effector molecule CagA into the cytosol where it becomes tyrosine phosphorylated (16, 17). CagA phosphorylation is dependent upon the number of TPMs within the COOH terminal variable region (VR) of the protein (18). CagA with greater numbers of TPMs becomes more phosphorylated and increases the extent of SHP-2–dependent (19) cytoskeletal alterations and cellular elongations (18, 20, 21). H. pylori strains that possess CagA with greater numbers of TPMs are more likely to be associated with gastric cancer than those possessing CagA with fewer TPMs (10–12).

We hypothesized that if VacA and CagA increased cancer risk through predisposing to hypochlorhydria and atrophy, they would be even more strongly associated with these conditions than they are with cancer itself, as cancer does not inevitably arise in the atrophic stomach. We therefore studied these associations in a group of first-degree relatives of patients with gastric cancer that had previously been used to study the association of IL-1β polymorphisms with gastric hypochlorhydria and atrophy (2). Studying family groups of adults who had not cohabited since childhood also gave us the opportunity to assess whether family members had an identical strain of H. pylori, implying childhood acquisition and so that infection predated hypochlorhydria/atrophy. Whilst doing this work, we noticed that virulence factors very frequently evolved in members of these families. This phenomenon is potentially important for the gastric cancer risk of individuals, and so we characterized and quantified this finding.

H. pylori are genetically diverse, and this diversity has arisen largely through intergenomic and intragenomic recombination (22, 23). Microevolution describes minor genetic change within a strain that leads to a stable change in phenotype. Recently, microevolution of H. pylori virulence genes vacA (24, 25) and cagA (26) occurring in real time have been characterized on a case report basis. These changes lead to a changed virulence phenotype (24, 26). Kersulyte et al. showed an example of a strain that deleted the whole cag PaI through recombination (27). Other studies have used PCR to show that strains with similar genetic fingerprints may have different vacA or cagA types (28–32) or to show deletion of cagA or of other genes in the cag PaI (33–35). However, these studies do not show whether phenotype (and so potentially virulence) is changed.

The collection of H. pylori strains gave us the opportunity to perform, for the first time, a detailed quantitative study of microevolution of vacA and cagA. We show that microevolution of virulence is common and extensive within these cancer relatives, not only within individual stomachs, but also between family members, occurring in about half of strain transfer events within families. It occurs through recombination and duplication/deletion and alters the phenotype of the strains.

Population.H. pylori strains were isolated as sweeps (see below) from relatives (siblings or offspring) of gastric cancer patients from western Scotland (2, 36). We now studied 51 of these relatives who had H. pylori isolates available for further analysis. Twenty-seven (53%) of the relatives were female, and the median age was 42 y (range 29-71 y). Gastric acid secretion data were available (37), and severe hypochlorhydria was defined as acid secretion at ≤10 mmol/h. Inflammation in the antrum and corpus was classified by the modified Sydney system (36). Twenty (39%) of the relatives had histologically defined gastric atrophy. H. pylori sweeps (combination of all colonies of H. pylori grown from a single biopsy sample on a blood-agar plate at 37°C in a microaerobic environment) were initially genotyped for vacA and cagA VR. Sweeps that exhibited different vacA types or size variations in cagA were plated to single colonies, and single colonies of each type were isolated and used for further genotypic and phenotypic characterization.

PCR typing of H. pylori strains. Strains were cultured on horse blood agar plates (Oxoid) for a maximum of five passages. H. pylori genomic DNA was extracted as described (38). Random-amplified polymorphic DNA (RAPD)–PCR was carried out using one or more of primers 1254, 1281, 1283, or 1290 (39). The primers used for PCR amplification and nucleotide sequencing are listed in the Supplementary Table S1. Nucleotide sequencing of the entire cagA gene was carried out after PCR amplification of overlapping fragments using cag primer pairs A21F-A22R, A22F-A23R, A23F-A24R, A24F-A28R, and A28F-A27R and cloning into pGemT-Easy (Promega).

Vacuolation assay. VacA-induced vacuolation of AGS and RK13 epithelial cells was determined microscopically after coculture with H. pylori strains for 16 to 24 h, as described (40).

CagA phosphorylation assay and IL-8 ELISA. The extent of CagA phosphorylation was determined by coculturing H. pylori strains with AGS cells for 6 h as described (20). Determination of CagA-induced cytoskeletal change (16) and IL-8 secretion by ELISA was carried out as described (20).

Statistical analysis. We used Mann-Whitney U tests, χ² tests, and Fisher exact tests, as appropriate.

Association of H. pylori virulence factors with gastric hypochlorhydria and atrophy. Of the 51 gastric cancer relatives we studied, 20 had gastric atrophy and 11 of these had severe hypochlorhydria (acid secretion, ≤10 mmol/h). Initial typing of vacA and cagA from 51 sweeps then single colonies (see Materials and Methods) showed that 4 of 51 subjects had H. pylori isolates of more than one vacA type and 9 of 51 had H. pylori isolates of more than one cagA size; these subjects were omitted from the following analyses.

First, we studied associations between the most toxigenic form of vacA (type s1/m1) and precancerous conditions. We found a significant association between vacA s1/m1 type and low gastric acid secretion. Of the nine people with severe hypochlorhydria, eight (89%) were infected with H. pylori with s1/m1 vacA, whereas only 34% of the remaining 38 subjects were colonized by an s1/m1 vacA strain (P = 0.004; Fisher exact test). Gastric atrophy was associated with infection with s1/m1 vacA strains, but this did not reach formal significance (P = 0.07; χ² test). However, people with gastric atrophy had significantly less stomach acid secretion if they were infected with s1/m1 vacA strains than s1/m2 vacA strains (P = 0.009; Mann-Whitney U test). This suggests that toxigenic H. pylori may increase cancer risk through increasing the risk of these precursor conditions, particularly gastric hypochlorhydria.

Next, we turned our attention to the cag PaI and cagA. Most (44 of 51) people in this Scottish population were infected with H. pylori strains possessing the cag PaI—the other seven had strains confirmed as cag PaI negative genotypically and phenotypically. Thus, our study was not powered to show any association of cag PaI with hypochlorhydria or atrophy, and indeed, no such associations were identified. More surprisingly, we also found no association between the number of CagA TPMs and acid secretion: Of the eight people with severe hypochlorhydria, six were infected with H. pylori possessing CagA with three TPMs (75%), one lacked the cag PaI, and only one possessed CagA with four TPMs. Of the remaining 34 people, 22 (65%) had CagA with three TPMs or fewer and 6 (18%) had strains lacking the cag PaI (P > 0.25). This raises the possibility that the increased cancer risk associated with higher numbers of CagA TPMs (10–12) may not be through increased risk of hypochlorhydria and atrophy, although the subject numbers studied here were fairly small so a type 2 error is possible.

Shared strains between noncohabiting adults implies childhood acquisition. The association between s1/m1 vacA strains and gastric hypochlorhydria could be due to the strains predisposing to these conditions or to them preferentially colonizing the hypochlorhydric stomach. One of our reasons for conducting this study in families of adult siblings was to address this issue. H. pylori strains from different individuals only have the same genetic fingerprints when they are passed between these individuals or acquired from a common source. Genetic fingerprinting methods have been used to show that strains are usually acquired from within the family in childhood (41–44). We argued that if noncohabiting adult siblings shared the same strain, this would be good evidence that they had acquired it in childhood. Of the gastric cancer relatives used in this study, 42 came from 12 families that provided from two–six members to the study; the other nine had no other family members in the study and are excluded from the following analysis. Family members F2-B and F12-D were infected with two distinct H. pylori strains by RAPD-PCR analysis, so there were 44 strains present within the 12 families. RAPD-PCR analysis (Fig. 1) showed that 29 of 44 (66%) H. pylori strains were shared with at least one other family member in 9 of 12 families (Table 1). Fifteen of 22 subjects with vacA s1/m1 strains shared that strain with at least one other family member, indicating that at least 68% of s1/m1 strains were likely acquired in childhood. Similarly, 66% of all strains were shared, implying childhood acquisition. Thus, acquisition of s1/m1 vacA strains likely preceded gastric hypochlorhydria.

H. pylori microevolution is common within individual stomachs. PCR-based genotyping of sweeps from gastric biopsies from our 51 subjects had shown that cagA of more than one size was amplified in nine subjects, and in two subjects, both the m1 and m2 vacA midregions were amplified. The most obvious explanation of this would be mixed strain infection. However, recent publications by us and others have shown occasional individuals to be infected by separate H. pylori isolates with different cagA or vacA genotypes, but have shown these isolates to be the same strain by various genetic fingerprinting techniques, most usually RAPD-PCR (24, 26, 31). This phenomenon has been termed microevolution. Because this phenomenon seemed fairly common among these cancer relatives, and because it has important implications for cancer risk, we explored it further. We plated out H. pylori samples with multiple vacA or cagA types to single isolates which had, as expected, only one type of cagA or vacA, and showed that all isolates from these 11 individuals had similar RAPD profiles (Table 1), indicating microevolution within these genes. To confirm this and to study the mechanism, we PCR amplified and did nucleotide sequencing on all or part of cagA or vacA from paired isolates, which were similar by RAPD-PCR and did phenotypic assays (VacA-induced vacuolation of AGS and RK13 cells, CagA phosphorylation, and CagA-induced hummingbird formation of AGS cells) to assess whether the isolates had changed their virulence.

Microevolution in cagA within individuals. Microevolution of cagA seemed to have occurred in 9 of 51 (18%) individuals (Table 1). Nucleotide sequencing of the cagA VRs showed that variation was always due to duplication or deletion of a 102-bp sequence containing the coding region for TPM type C, and in the case of strain 9, variation was due to single and double duplication/deletion of this region. The sequence surrounding the repeat region was identical between the paired isolates of seven strains and differed by only 1 and 3 bp, respectively, in the other two. Comparisons of this region of cagA between unrelated strains in this population showed variation in 18 to 38 bp. This confirmed the recent clonal origin of the paired isolates and suggested that microevolution had occurred by intragenomic recombination—either deletion or duplication of this region. To determine whether differences in the number of CagA TPMs altered the phenotypes of the strains, AGS cells were cocultured with the two isolates from strain 10 (3 and 4 TPMs). As expected, the CagA with four TPMs became more phosphorylated and induced longer cellular protrusions than the CagA with only three TPMs (Fig. 2). Further examples are discussed later.

Microevolution in vacA within individuals. Among the 42 subjects in the 12 families, only two had single colony isolates of H. pylori with different vacA types (s1/m1 and s1/m2). These subjects belonged to the same family and were siblings (F7-C and F7-D) who, unlike other family members in our study, lived with each other as adults. All four isolates were found to be the same strain (strain 14) by RAPD-PCR analysis. PCR amplification of vacA from single colony isolates of strain 14 from both family members showed that both individuals had isolates with both s1/m1 and s1/m2 vacA. Nucleotide sequencing of the entire vacA gene of both isolates from both siblings revealed that, whereas the s1/m1 alleles in isolates from both siblings were identical, the s1/m2 forms were not (Fig. 3). Both s1/m2 forms of vacA were found to possess stop codons close to the 5′ end that prevented VacA expression (as confirmed by Western blot). As expected, isolates possessing these forms of vacA did not induce vacuolation of RK13 epithelial cells, whereas isolates possessing s1/m1 vacA did induce vacuolation, confirming that microevolution lead to a changed phenotype.

To obtain further evidence that we were dealing with a single parental strain, we did nucleotide sequencing of the housekeeping genes mutY and yphC of both s1/m1 and both s1/m2 isolates. mutY was identical between all four isolates, as was the region 5′ to yphC confirming parental strain identity. However, the yphC allele in the s1/m1 vacA isolate from subject C differed extensively to that of the other three isolates of strain 14, suggesting that, like vacA, it had been acquired horizontally from another strain. Taken together, these data suggest that the s1/m2 vacA isolates (which had developed minor genetic differences in vacA) were passed between these individuals many years previously, perhaps in childhood. In contrast, the s1/m1 vacA allele was likely acquired horizontally by the strain in one individual, and this isolate was passed to the other individual much later in life (because there were no nucleotide substitutions in s1/m1 vacA alleles between subjects). Other genes were also likely acquired horizontally by this strain, as analysis of two randomly selected housekeeping genes showed that one had undergone partial allelic replacement in one isolate.

Strains circulating within families have frequently microevolved to change virulence. Our analysis of strains passed between family members showed that, whereas family members had the same strains (by RAPD-PCR analysis), the isolates from different family members frequently varied in either vacA type or cagA type. We now confirmed and quantified this novel and potentially important phenomenon of virulence factor microevolution within families of gastric cancer patients. Among the family members we studied, RAPD-PCR analysis showed that strains were passed from one person to another in 18 cases, of which eight (44%) displayed microevolution; this occurred in the cagA VR in six cases and the vacA midregion in two cases (Table 1).

Turning first to the cagA microevolution, nucleotide sequence analysis of the cagA VR from the six cases showed that cagA had evolved by single or double duplication of a 102-bp sequence encoding a TPM. Phenotyping confirmed that this resulted in a changed phenotype (Fig. 4). Nucleotide sequencing of mutY and yphC confirmed the RAPD-PCR results that microevolved isolates were the same strain.

For vacA, nucleotide sequencing of s1/m1 and s1/m2 vacA isolates from strain 4 in members of family F2 showed that the vacA alleles differed between these isolates. There were, excluding the regions of no homology (comprising 429 nucleotides), 403 nucleotide substitutions (10.9% of the gene) scattered throughout vacA. VacA was expressed by both the s1/m1 isolate from subject F2-A and the s1/m2 isolate from subject F2-B. However, we could not show a changed phenotype in this case, as both isolates induced vacuolation in RK13 cells during coculture experiments but neither isolate caused vacuolation of AGS cells. Nucleotide sequencing of mutY and yphC from different isolates of strain 4 from subjects F2-A and F2-B showed that the mutY fragments and the majority of the yphC gene were identical, confirming that these isolates were the same “strain.” However, we identified another horizontally acquired DNA fragment of 600 bp which comprised the 5′ 210 nucleotides of yphC, the upstream intergenic region (82 nucleotides), and the first 308 nucleotides of the neighboring gene hp0833; 25 nucleotides in this region differed between the two isolates. This showed that yphC, as well as vacA, had been a target for partial allelic replacement by homologous recombination.

Virulence genes may be donated to unrelated strains through homologous recombination. As virulence genes can be acquired horizontally from unrelated strains, we next asked whether we could show donation of virulence genes to unrelated strains. In family F3, subjects B, C, and D had H. pylori isolates which were the same strain (strain 8) by RAPD-PCR, whereas subject A carried a different strain (by RAPD), strain 7. Strain 8 isolates from subjects C and D possessed cagA encoding 3 and 4 TPMs, respectively, whereas isolates of strain 8 from subject B (which showed microevolution within an individual) encoded cagA with either four or five TPMs. Unrelated strain 7 possessed cagA with four TPMs that was almost identical (one base change) to the cagA VR of strain 8 from subject D, suggesting that strain 7 may have acquired cagA from strain 8 in subject D. Phenotypic analysis revealed that CagA from strain 7 was phosphorylated to a similar extent after coculture with AGS cells to CagA from subjects B and D (Fig. 4).

In family F7 described previously, nucleotide sequencing of the entire vacA gene from strain 13 from family member F7-B revealed that this sequence was nearly identical to the s1/m2 vacA of strain 14 in subject D: only 7 of 3,915 nucleotides were different, but one of these substitutions meant that there was no 5′ stop codon in the vacA allele in strain 13 (Fig. 3). As expected, VacA was expressed in strain 13 and induced vacuolation in RK13 cells. Thus, it seems that vacA from strain 13 in subject B was acquired by strain 14 in subject D, where it was mutationally inactivated, and this isolate was passed to subject C. This likely occurred in childhood, as a small number of nucleotide substitutions are now present between vacA alleles in these three isolates.

A third example is discussed in more detail below and shown in Fig. 5.

Studying microevolution of virulence factors allows passage of strains within families to be mapped (Fig. 5). In family F12, H. pylori strain 25 was isolated from all four members of the family, and these isolates were shown to be the same strain by RAPD-PCR and by sequencing of mutY and yphC (Table 1). Strain 26 from subject D (who was also infected with strain 25) was unrelated (Table 1). Single-colony isolates of H. pylori strain 25 from subjects C and D both showed intrastrain evolution of cagA by duplication/deletion within the cagA VR and had identical nucleotide sequences between subjects; it is possible that the strain had microevolved in one family member, and then both isolates had been passed to the other. The cagA VR in strain 25 from subject A had 19 nucleotide substitutions, implying that it was acquired horizontally from another strain. Comparison with the cagA VR of strain 26 showed that this strain was the likely donor, in that the nucleotide sequence exhibited only 2-bp differences, although the strain 25 isolate in subject A had also undergone a 102-bp duplication of the third TPM of strain 26. Nucleotide sequencing of the entire cagA alleles from H. pylori strain 25 from subject A and strain 26 from subject D, and the shorter cagA allele (encoding four TPMs) of strain 25 from subject D showed that there had been extensive recombination throughout cagA. The cagA alleles from these three isolates varied in 119 positions with 56 of these changes occurring within 966 bp of the 3′ end of the gene, which includes the VR. Taken together this suggests that strains 25 and 26 had recombined in subject D, such that cagA from strain 26 had recombined extensively with cagA from strain 25, and this included donation of its VR. This recombined strain 25 would then have been passed to subject A where duplication of the third TPM may have occurred (Fig. 5). The strain 25 H. pylori isolate from subject B was found to lack the entire cag PaI. This isolate may have evolved from any of the otherwise identical strains from the other members of the family (Fig. 5). Phenotypic analysis revealed the isolate of strain 25 from subject B, lacking the cag PaI, failed to induce IL-8 from cells, whereas cag PaI–positive isolates of strain 25 and strain 26 from other members of the family did induce IL-8 secretion, as expected, and had CagA translocated into cells and phosphorylated to an extent proportional to the number of CagA TPMs. Thus, this family showed genetic and phenotypic evolution of virulence factors in a strain which was passed between family members, and this allowed tracking of the passage of this strain within the family.

H. pylori strains possessing type s1/m1 vacA were strongly associated with gastric hypochlorhydria in this population of first-degree relatives of gastric cancer patients in Scotland. Intestinal-type distal gastric cancer arises in a stepwise manner with simple H. pylori–associated gastritis developing to atrophy and hypochlorhydria, metaplasia, dysplasia, and cancer (4). Human genetic polymorphisms leading to high-level IL-1β secretion increase gastric cancer risk, but increase the risk of gastric hypochlorhydria even more markedly (2), implying that they convey their cancer risk through increasing the risk of this precursor condition. In this study, we show that vacA type s1/m1 strains of H. pylori, which several studies have shown are associated with increased gastric cancer risk (5–7), also increase the risk of hypochlorhydria, suggesting that they too increase cancer risk by increasing the risk of this precursor condition.

Studying H. pylori strains within families showed that most strains (66%) were shared between family members. As siblings rarely cohabit as adults and strains are rarely passed between spouses (43, 44), this suggests that strains were acquired during childhood, as has been shown for other populations (41, 42). This therefore implies that H. pylori strains possessing type s1/m1 vacA were acquired before the development of gastric atrophy and hypochlorhydria and suggests that the association between toxigenic strains and these conditions is likely to be causal.

Within the population studied here, we found extensive microevolution in cagA and significant microevolution in vacA in strains within individuals and in strains passed between family members. The finding in individuals is not novel: our studies follow previous case reports of microevolution in vacA and cagA leading to changes in phenotype (23, 24), although we show that this occurs more commonly than suspected, at least in these gastric cancer relatives. Our finding of H. pylori microevolution in these cancer families and its high frequency is novel and has important biological and clinical implications.

Turning first to H. pylori cagA and vacA microevolution within individuals, we found this in 11 of 51 (22%) subjects, which is likely to be a vast underestimate of the true frequency. Inevitably, our sampling represents a snapshot in time and also involves only a tiny area of the stomach—that contained in a single biopsy specimen. Also our study gives no clue as to how long cocolonization of phenotypically different isolates in the same stomach has persisted and will persist. This will depend on whether the phenotypic change confers a selective advantage and the extent of any such advantage. Given these limitations, it is interesting that our study gives similar frequencies to those recently reported in an Irish population. Like us, these workers screened for differences in the size of the cagA VR by PCR in strains from individual stomachs, although they did not check results or study likely mechanisms by performing nucleotide sequence analysis and nor did they check for phenotype (31). Nevertheless, their PCR analysis suggested that 3 of 19 (16%) individuals examined may have had isolates of the same strain with different cagA sizes coexisting in their stomach. This proportion is similar to that we report.

The 12 family groups in our population allowed us to study the transmission of strains between relatives of cancer patients and for the first time to show and quantify microevolution of vacA and cagA in strains passed between members of the same family. Besides its biological and clinical relevance, determining microevolution within families is a much better indicator of the extent of microevolution than looking in individuals. The main advantage is that, in effect, it is the result of sampling over a longer time period; family studies will pick up microevolution which has occurred at any stage because the parental strain was passed between two individuals, most usually in childhood. Sampling isolates from separate stomachs also avoids the potential criticism that microevolution may be an in vitro phenomenon due to recombination between subclones during initial culture; in the family scenario, these subclones are kept entirely apart. The extent of microevolution we describe in strains shared between family members (44%) is both important and entirely unexpected. Owen and Xerry (30) studied 16 individuals in five families, and although they did not focus on cagA or vacA, they described one example of a strain shared between husband and wife which differed for vacA midregion by PCR typing, indicating that microevolution may have taken place in that strain. Carroll et al. studied one sibling pair from a family and found different vacA signal region types, providing an indication that this part of vacA may have microevolved (31).

The frequency with which microevolution of virulence factors occurs within families has important implications for bacterial persistence. That it occurs at all implies that there must be selective advantages to H. pylori of having different vacA types or different numbers of cagA TPMs in different hosts. The environmental drivers of this are unknown. For vacA, they are not likely to be differences in acid production or presence or absence of atrophy, as our results suggest that these changes follow rather than precede colonization by vacA s1/m1 strains. Equally, for cagA, atrophic hypochlorhydric subjects in our study were not preferentially colonized by strains with higher or lower numbers of CagA TPMs. We are now performing further studies in vitro and in vivo to elucidate the selective advantages different virulence factor polymorphisms offer to individual strains under different conditions.

Microevolution in vacA and cagA has important implications for H. pylori pathogenesis. Changing vacA or cagA type may provide a partial explanation of why studies examining the relationship between bacterial virulence and disease do not show tighter associations. Indeed, if H. pylori microevolution occurs throughout the life of a human host, close associations of current virulence factor genotypes with disease would be unexpected. We already know that multiple strain infection is common, especially in developing countries (45–47), and knowing also that patients can be infected by the same strain with different virulence levels further complicates this.

Finally, and perhaps most importantly, our work has important clinical implications for families of gastric cancer patients. Although we did not have strains from cancer patients themselves available for study, our investigation of their families shows that strains are usually shared. However, this does not mean that different individuals have H. pylori with the same virulence; indeed, in about half the cases where strains are shared, this is not the case. Thus, if our clinical practice changes (as we believe it should), such that we screen young relatives of cancer patients for H. pylori, the safest option would be to treat all who are positive. We still do not know for sure whether this would change their cancer risk, but subgroup analysis of a large randomized treatment trial suggests that it might do so in patients who have not yet developed gastric atrophy (48).