Host-dependent Lewis (Le) antigen expression in Helicobacter pylori cells recovered from Le^b-transgenic mice

Conventionally raised, specific pathogen-free transgenic FVB/N Le^b mice (n = 20) and wild-type littermates (n = 20) were group housed in microisolator cages, maintained on a standard chow diet, and challenged at 6 wk of age (Fig. S2) with H. pylori strain HP1 (98-964), which expressed both Le^x and Le^y but had no detectable Le^a or Le^b expression. At varying times after infection, mice of each genotype were sacrificed, the stomachs were homogenized, and H. pylori was cultured by serial dilution on selective media. H. pylori colonization could only be detected in three out of the five mice in each group after 4 wk but was detected in each mouse at all subsequent time points (8, 16, and 32 wk; n = 5 animals/group/time point). There were no significant differences in bacterial density observed according to mouse genotype at any of the time points surveyed (Fig. 1). In total, we achieved stable (10³ CFU/mouse stomach), long-term (≥8 mo) H. pylori colonization in both wild-type and Le^b-transgenic mice (Fig. 1).

We next addressed whether colonization of the wild-type and Le^b-transgenic mice would select for differing H. pylori Le phenotypes, as determined by Le-specific ELISAs. All bacterial populations recovered expressed Le^x and Le^y, and there were no significant differences in Le^x or Le^y expression between sweeps recovered from wild-type or Le^b-transgenic mice in either the early or late isolates (Fig. S3). None of the isolates recovered from the wild-type or transgenic mice before the 8-mo time point expressed Le^a or Le^b (n = 10 assayed/time point), nor were these epitopes detectable in H. pylori sweeps recovered from wild-type mice 8 mo after challenge (Table I). Although the ratio of Le^x to Le^y in the H. pylori cells recovered from the 8-mo wild-type mice was similar to the inoculum strain, their overall expression was higher. However, in H. pylori sweeps from the 8-mo Le^b-transgenic mice, there was expression of Le^b in addition to Le^x and Le^y in four out of the five mice (Table I), which was significantly (P = 0.001) different from the wild-type mice. In addition, overall Le^x expression in the bacterial populations isolated from the Le^b-transgenic mice was significantly (P = 0.014) lower than in populations from the wild-type mice. There was a trend toward higher colonization levels in the four mice carrying Le^b-positive H. pylori compared with colonization densities of the five wild-type mice (P = 0.054). This result could reflect increased adhesion to host-expressed Le^b through binding by the bacterial Le^b ligand, BabA (Ilver et al., 1998). Although this result suggests that bacterial Le^b expression enhanced colonization and, thus, may have been positively selected, future studies will be required to confirm this observation.

One hypothesis to explain the drop in Le^x expression in the 8-mo Le^b-transgenic mice is that an increase in anti-Le^x antibodies in the transgenic mice selected for cells with reduced Le^x expression, resulting in the emergence of Le^b expression, paralleling phenomena that have been observed previously in other organisms (Bayliss et al., 2001). To test this hypothesis, we determined mouse antibody levels against both H. pylori whole cells and purified H. pylori LPS with distinct Le phenotypes. ELISAs testing serum responses to 98-964 (HP1) whole-cell antigen showed that both wild-type and Le^b-transgenic mice progressively developed anti–H. pylori IgG responses during the course of the challenge (Fig. S4). No IgG was detected in the uninfected (control) mice, but an IgG response appeared during the early infection period (4–8 wk), rising to the highest levels during late infection (4–8 mo). There were no significant differences in IgG levels between the wild-type and Le^b-transgenic mice (n = 10 mice/genotype/time point). These data confirm that the experimental H. pylori challenges induced adaptive humoral responses, as expected. The lack of difference between the groups provides evidence that unlike in previous studies (Guruge et al., 1998), the host genotype did not substantially affect the responses to the heterologous H. pylori antigens.

Next, we tested mouse sera with LPS preparations from wild-type strain J166 (OD values: Le^x = 0 and Le^y = 3.36) with strong Le^y expression, J166 ΔfutC (OD values: Le^x = 2.22 and Le^y = 0) with strong Le^x expression, and J166 ΔfutA/ΔfutB (OD values: Le^x = 0 and Le^y = 0) with neither Le^x nor Le^y expression. Thus, we could examine the serologic response to specific Le epitopes. First, we examined responses to J166 whole-cell antigen, expecting that there would be robust responses, as observed with 98-964 whole-cell antigen (Fig. S4). The high antibody levels in 4- and 8-mo wild-type (OD value = 1.69 ± 1.48) and Le^b-transgenic mice (OD value = 1.89 ± 0.98; Table S1) confirmed that the mice were capable of responding to H. pylori antigens and showed no significant difference in response according to mouse genotype.

The responses to the LPS antigens were heterogeneous (Table S1), with a bimodal distribution, and roughly correlated with those to whole-cell antigen. The responses to Le^y LPS (wild type) and Le⁰ LPS (ΔfutA/ΔfutB) were significantly (P = 0.035 and P = 0.011, respectively) higher in the 8-mo compared with the 4-mo mice, and trended in that direction (P = 0.15) for the Le^x LPS (ΔfutC). Response to LPS also appeared to be independent of LPS Le phenotype; mice with strong anti-LPS responses showed strong responses to LPS of all three Le phenotypes. There were no significant differences in anti-LPS responses between wild-type and Le^b-transgenic mice. Finally, we asked whether there was an association between anti-LPS response and the H. pylori colonization density in the mice. Comparing the log₁₀ CFU (2.43 ± 0.98) of the three mice with the highest anti-Le^x (J166 ΔfutC LPS) responses (3.49 ± 0.7) to the log₁₀ CFU (2.61 ± 0.92) of those with the lowest responses (0.03 ± 0.001) showed no significant differences; results were parallel for the other LPS preparations (unpublished data).

One hypothesis to explain why colonization densities were higher in the Le^b-transgenic mice carrying Le^b+ H. pylori could be the increased bacterial adherence to the gastric epithelia by the H. pylori Le^b ligand, BabA. In a previous challenge of rhesus monkeys (Solnick et al., 2004), a gene conversion event occurred that replaced a portion of babA with babB (as demonstrated by a series of PCRs), which resulted in a loss of Le^b adhesion. To examine this question, we first sought to confirm the previous findings (Solnick et al., 2004) by studying a strain (J166) that successfully colonized rhesus monkeys and comparing it with a J166-derived strain recovered from a monkey 10 mo later (98-169) from our previous monkey challenge studies (Wirth et al., 2006). Sequence analysis of the band produced by primers 834F and AR6 amplified in 98-169 show that the middle region of babA was replaced by babB, indicating that gene replacement had occurred, providing independent confirmation of the previous findings in monkeys (Fig. S7; Solnick et al., 2004). On this basis, we turned to the mouse isolates. When we tested the inoculum strain 98-964 and one isolate each from the wild-type (03-261) and Le^b-transgenic mice (03-270), we found that the babA allele was intact in all cases (Fig. S7), supporting the hypothesis that there was no selection away from babA, and that its maintenance is important for mouse colonization, differentiating the findings in monkeys and mice.

To analyze the diversity in Le expression within H. pylori populations recovered from the wild-type and transgenic mice 8 mo after challenge, single colonies were isolated from sweeps 00-8 (from a wild-type FVB/N mouse) and 00-16 (from an Le^b-transgenic mouse; Fig. S2). The single colonies isolated from sweep 00-8 only showed expression of Le^x and Le^y, with no detectable Le^a or Le^b expression, confirming results obtained from the gastric sweeps (unpublished data). In contrast, 9 out of 11 single colonies isolated from the 00-16 sweep expressed Le^a and/or Le^b in addition to Le^x and Le^y (Fig. 2 A). The phenotypic diversity within the bacterial populations derived from the gastric sweeps of these mouse stomachs is similar to that reported in humans (Wirth et al., 1999; González-Valencia et al., 2008). Expression of Le^a in five of the 00-16 single colonies, in addition to Le^b, indicates diminished futC activity in these isolates (Sanabria-Valentín et al., 2007; Nilsson et al., 2008), preventing substrate Le^a from being fucosylated to form Le^b (Fig. S1). The ratios of Le^a to Le^b and Le^x to Le^y expression in these single colonies are strongly correlated (R = 0.89; P < 0.001; Fig. 2 B), indicating that futC is governing the relative expression levels of the mono- and difucosylated antigens in tandem for the type I and II pathways (Fig. S1). From these studies, three strains were chosen for further genotypic and phenotypic analysis: HP1 (98-964, the inoculum strain from which Le^b was not detectable), 00-8B (03-261, a non-Le^b expressor isolated from a wild-type mouse), and 00-16A (03-270, an Le^b expressor isolated from an Le^b-transgenic mouse; Table I).

Nucleotide sequence analysis was performed to determine the genetic basis for the phenotypic differences in Le antigen expression in the selected representative mouse-derived H. pylori strains. A β-(1,4) galactosyltransferase involved in Le^x and Le^y synthesis is encoded by galT (Fig. S1), which does not contain polynucleotide repeat regions; however, the upstream intergenomic region varies in length between strains (not depicted). Because sequence analysis showed that all three strains share an identical upstream region (unpublished data), their different Le antigen phenotypes were not attributed to galT locus variation.

Sequence analysis of the α-(1,3/4) fucosyltransferase gene (futA; Fig. S1), which affects both the type I and II Le synthesis pathways, showed that all three strains had identical in-frame futA alleles. Sequence analysis of futB (Fig. S1) revealed extensive intrastrain poly-C tract length variation among the three strains (Fig. S5), but overall there were no significant differences between the three strain populations. Thus, differential activity of the α-(1,3/4) fucosyltransferases in these strains does not correlate with the Le^b phenotypic differences observed but is consistent with the overall phenotypic variation in Le expression within the populations of cells, as reflected in the Le phenotypes of single-colony isolates (Fig. 2 A).

The poly-C tract of futC was cloned and sequenced for each strain (10 clones/strain) and, as expected, revealed extensive intrastrain variation (Sanabria-Valentín et al., 2007). The primary sequence data suggest that only 2 (61C6 and 70C7) out of 30 intrastrain variants tested would produce a full-length α-(1,2) fucosyltransferase (Fig. S6 B). Although multiple mechanisms can be present (Wang et al., 1999), some α-(1,2) fucosyltransferase activity must exist in strains 98-964, 03-261, and 03-270, because all three produce Le^y and 03-270 also expresses Le^b. The distribution of poly-C tract lengths in Le^b-expressing 03-270–derived clones was significantly (P < 0.01 and P < 0.001) different from that in clones derived from 98-964 and 03-261, respectively, consistent with enhanced selection for Le^b expression in the Le^b-transgenic mice.

Sequence analysis of β-(1,3)galT showed that the non-Le^b–expressing strains 98-964 and 03-261 were identical in sequence (Fig. 3 A). Both contain C₁₄ poly-C tracts, whereas Le^b-expressing strain 03-270 has a C₁₆ tract (Fig. 3 B). Based on the predicted (ATG) start codon of β-(1,3)galT in sequenced strain J99 (Fig. 3 A; Alm et al., 1999) as the putative translation start site, all three strains appear to have an out-of-frame β-(1,3)galT (Fig. 3 D). Thus, primary sequence data would not be sufficient to explain differences in Le antigen expression between these strains. However, if the alternate translation initiation codon (TTG), active in ∼9% of H. pylori genes (Alm et al., 1999), is used in these strains at the same position as the GTG start codon (∼8% of H. pylori genes) of β-(1,3)galT in reference strain 26695 (Fig. 3 A; Tomb et al., 1997), then frame status is determined by poly-C tract length. Strains 98-964 and 03-261 encode seven β-(1,3)galT heptad repeats versus four in strain 03-270 (Fig. 3 C); similar changes were observed in a previous mouse gastric challenge (Salaün et al., 2005).

Because of these uncertainties, allelic replacement mutagenesis was performed to determine whether β-(1,3)galT is essential for H. pylori type I Le antigen expression (Fig. S8 B). A plasmid containing a copy of β-(1,3)galT interrupted by a nonpolar kanamycin resistance (aphA-3) cassette (p98B13k) was introduced into test H. pylori strains via natural transformation (Fig. S8 B), and insertion of the plasmid-encoded sequences was confirmed by PCR analysis of chromosomal DNA from transformed cells. Strains 99-8 and JP26 were included as positive controls for Le^a and Le^b phenotypic expression, respectively, and strain JP26 was transformed with pCTB8 (Cover et al., 1994) as a positive control for transformation. Introduction of p98B13k resulted in loss of Le^b expression in strains JP26 and 03-270 (Fig. 4 A), and loss of Le^a expression in strain 99-8 (not depicted). Also as expected, introduction of control plasmid pCTB8 into JP26 had no effect on Le antigen phenotype (unpublished data). In total, these studies provide evidence that open reading frame (ORF) jhp0563 encoding the presumed β-(1,3)galT is essential for synthesis of type I Le antigens.

To determine whether the loss of type I Le antigen expression was specifically caused by β-(1,3)galT inactivation and not an adventitious event, complementation studies were performed. To accomplish this, jhp0563 from strain 03-270 was introduced in trans (p70B13comp) into H. pylori strains in which the native ORF had been interrupted (Fig. S8 C), and its placement was confirmed by PCR. Introduction of β-(1,3)galT at the ureA locus resulted in restoration of type I Le antigen expression in the JP26 and 03-270 β-(1,3)galT null mutants (Fig. 4 A). Complementation of strain 03-261 led to a low level but new expression of Le^b, which had not been observed in either the wild-type or mutant strains (Fig. 4 A). However, there was no complementation of Le^a expression in the 99-8 β-(1,3)galT null mutant and no de novo synthesis of Le^a or Le^b in strain 98-964 (unpublished data). Collectively, these results provide evidence that β-(1,3)galT is necessary, but not sufficient, for type I antigen expression.

Because colonies of H. pylori represent a mixture of cells of varying genotypes (Appelmelk et al., 1998; Kuipers et al., 2000; Sanabria-Valentín et al., 2007), especially in reference to metastable loci such as homopolymeric tracts (Salaün et al., 2004; Sanabria-Valentín et al., 2007), we analyzed nucleotide sequences from the complemented strains. Analysis of the 5′ 500 nucleotides of the complemented copy of β-(1,3)galT revealed sequences identical to 03-270 wild-type β-(1,3)galT in all five complemented mutants tested. These results indicate that sequence variation in the 5′ region is not responsible for the phenotypic variation observed. However, as determined by direct PCR sequencing, poly-C tract lengths varied among the complemented mutant strains. Therefore, Le antigen phenotypes were determined by ELISA for 10 single colonies isolated from strain 03-270::p98B13k::p70B13comp. The Le^b expression in these strains ranged from the level of 03-270 wild type to no detectable expression (Fig. 4 B). The β-(1,3)galT heptad repeat region was identical in length in these single-colony isolates, indicating that variation in Le^b expression is not caused by variation in this region (unpublished data). To identify the basis of this variation, we examined genomic DNA from isolate 70C1A with high Le^b expression (Le^b = 0.74), from isolate 70C1E with an intermediate Le^b phenotype (Le^b = 0.25), and from isolate 70C1F with no detectable Le^b expression to PCR amplify the poly-C tract regions (OD values are shown). After cloning the products, sequence analysis of the recombinant pGEM-T Easy plasmids revealed extensive β-(1,3)galT poly-C tract length variation (Fig. 4 C). 4 out of 10 70C1A-based sequences contained C₁₆ poly-C tracts, identical to 03-270 wild type, versus 2 out of 10 from the 70C1E sequences and 0 out 10 from the 70C1F sequences. The number of clones with a β-(1,3)galT C₁₆ tract, identical to the 03-270 poly-C tract, correlated with the level of Le^b expression in each isolate. The 70C1A poly-C tract lengths (16.2 ± 1.48 nucleotides) and the 70C1E poly-C tract lengths (15.8 ± 2.04 nucleotides) were significantly (P < 0.001) longer than those in 70C1F (11.8 ± 0.79 nucleotides). Thus, within populations of H. pylori cells, β-(1,3)galT frame status as well as homopolymeric tract length vary extensively, critically affecting Le^b expression.

To better understand the distribution of Le^b phenotypic expression within a population of H. pylori cells, wild-type strains and their β-(1,3)galT mutants were examined by flow cytometry. Cells were incubated with an anti-Le^b monoclonal antibody and anti–H. pylori polyclonal antibodies as a positive control. Cells of all three H. pylori strains (98-964, 03-261, and 03-270) were detectable with anti–H. pylori, essentially to the same extent (Fig. 5 B). As expected, when stained with antibodies directed against Le^b, the 03-270 wild-type strain showed the greatest levels of fluorescence, with a bimodal distribution, indicating that most but not all cells were Le^b positive (Fig. 5 C). In contrast, also as expected, the 03-270 β-(1,3)galT mutant showed fluorescence levels similar to the background level of cells stained with secondary antibodies alone (Fig. 5 C). The 03-270 β-(1,3)galT–complemented mutant was positive for Le^b, but the percentage of Le^b-positive cells was lower than in the wild type (Fig. 5, C and F), indicating that the complementation was partial. The net mean fluorescence intensity (MFI; region M1) for the Le^b-positive cell populations in the wild-type and complemented strains was similar, indicating that a proportion of the cells had been fully complemented (Fig. 5 G). Similar flow cytometry experiments performed on JP26 and 03-261 indicated that the complemented mutant populations were a mixture of Le^b+ and Le^b− cells (Fig. 5, D and E). The Le^b positivity in the JP26-complemented cells was significantly (P < 0.001) lower than in the wild-type strains, confirming that complementation was partial and less efficient than in strain 03-270 (Fig. 5 F). However, as in strain 03-270, net MFI levels showed full complementation in a portion of the population (Fig. 5 G). Collectively, these results paralleled phenotypes determined by ELISA (Fig. 4 A) and define the range of Le^b phenotypes exhibited by wild-type H. pylori strains and their β-(1,3)galT mutants.

These studies showed that experimentally challenged wild-type and Le^b-transgenic FVB/N mice can stably maintain H. pylori gastric colonization for up to 8 mo, regardless of Le antigen expression, consistent with studies in other mouse genetic backgrounds (Takata et al., 2002; Lozniewski et al., 2003). That H. pylori populations expressing Le^b were somewhat more efficient than non-Le^b expressors in colonizing Le^b-transgenic mice (Table I) may reflect greater host tolerance to self-antigens, although future studies are needed to confirm this observation. This phenomenon also could be attributed to enhanced adhesion of bacterial cells to the gastric epithelia via BabA binding to host Le^b, because unlike in monkeys (Solnick et al., 2004), babA recovered from the mice 8 mo after challenge showed intact copies of babA and no evidence of gene conversion (Fig. S7). The host Le^b–bacterial BabA interaction also could be a target of selection. Enhanced bacterial Le^b expression could result in BabA-mediated bacterial aggregation, contributing to the higher colonization densities observed in Le^b-transgenic mice; further experimentation is required to test this hypothesis.

Selection for Le^b expression among H. pylori cells persistently colonizing Le^b-positive transgenic but not Le^b-negative wild-type mice is consistent with increased fitness of H. pylori variants that match the Le phenotype of their host, extending previous studies in rhesus monkeys (Wirth et al., 2006) and in some (Wirth et al., 1997) but not all (Taylor et al., 1998; Heneghan et al., 2000) human studies. One hypothesis to explain the appearance of Le^b expression in H. pylori recovered from the transgenic mice 8 mo after challenge is the development of an anti-Le^x response in these mice, which is consistent with the lower expression of Le^x in the H. pylori cells recovered from the Le^b-transgenic mice after 8 mo. Such a response would provide selection for the expansion of Le expression variants, a phenomenon that has been observed previously in Neisseria meningitidis (Bayliss et al., 2001). To test this hypothesis, we determined levels of anti-Le antibodies by ELISA using purified LPS with well-defined Le antigen phenotypes. The results showed a wide range in responses among the mice, with no correlation between mouse genotype and antibody response. Further, antibody responses were independent of both LPS Le phenotype and colonization density of the mice. These results provide evidence that serum antibody pressure was not the driving force behind the emergence of bacterial Le^b expression. However, our observations are limited by a small sample size and an inability to directly test for anti-Le^a and anti-Le^b antibody levels. Future studies are needed to further investigate this hypothesis.

Sequence analysis showed that selection of Le^b-positive H. pylori variants in the transgenic mice was mediated by phase variation of β-(1,3)galT. Solnick et al. (2004) provided evidence that phase variation of babA occurred in vivo, supporting the hypothesis that this is a gene regulation mechanism used by H. pylori to adapt to changing host environments.

Experimental challenge of wild-type mice and gerbils for up to 5 mo with H. pylori isolates led to no substantial Le antigen expression diversification (Wirth et al., 1999). Similarly, when wild-type FVB/N and Le^b-transgenic mice were experimentally challenged for 8 or 16 wk, no change in Le antigen expression was detected (Guruge et al., 1998). One explanation for those results is that there was no phenotypic selection, or alternatively, that the H. pylori founding populations were too small and/or colonization periods too short for detection of differential fitness (Webb and Blaser, 2002). That no changes were observed before 8 mo in the present studies may reflect both the relatively low level of H. pylori colonization of conventionally raised mice compared with humans (Atherton et al., 1996) and/or relatively small differential fitness (Webb and Blaser, 2002) because of bacterial Le^b expression in the Le^b-transgenic mice.

Sequence analysis of β-(1,3)galT in HP1 and its descendants revealed a 5′ 5-bp deletion compared with H. pylori reference strain J99, suggesting either a truncated ORF or that the annotated 5′ ATG is not the initiation codon in these strains. Because Le^b is expressed in strain 03-270, a more downstream start codon may be used for translation, translational frameshifting is occurring, or β-(1,3)galT is undergoing recombination with its upstream homologue, jhp0562, in some H. pylori cells within the population. TTG present in some strains at the same position as the 26695 initiation GTG (Fig. 3 A) may be an alternative initiation codon (Alm et al., 1999), representing another translational control locus of Le synthesis gene expression.

Our findings indicate that as with H. pylori futA, futB, and futC, poly-C tract length likely regulates β-(1,3)galT translation, with C₁₆ in 03-270 critical for Le^b expression. The three strains tested also differed at the β-(1,3)galT heptad repeat region, paralleling observations in another mouse challenge experiment (Salaün et al., 2005), potentially explaining β-(1,3)galT enzymatic activity variation, paralleling that shown in futA and futB (Ge et al., 1997; Lin et al., 2006; Ma et al., 2006; Nilsson et al., 2006).

Although the inoculum strain HP1 expressed Le^x and Le^y in relatively equal proportions, the gastric sweeps and single colonies varied greatly in their Le antigen phenotypes. That several single colonies isolated from within the Le^b-expressing bacterial populations also expressed Le^a, which was not detected in the parental strain, indicates futC phase variation among the strains, consistent with the extensive sequence variation observed (Fig. S6). Diversity in futC sequences in differing strain backgrounds (Alm et al., 1999; Wang et al., 1999; Salaün et al., 2004; Salaün et al., 2005; Sanabria-Valentín et al., 2007) may reflect strain-specific adaptations that have evolved to facilitate Le expression control, particularly the tension between mono- and difucosylated forms (Wang et al., 1999; Rasko et al., 2000; Kang and Blaser, 2006a).

Loss of type I Le antigen expression in a β-(1,3)galT mutant confirms a previous report (Appelmelk et al., 2000) that in trans complementation of β-(1,3)galT restored the type I pathway phenotype lost in null mutants, and introduced Le^b expression into a strain previously negative for type I antigens expands on these previous studies. The lower type I phenotypes in the complemented strains compared with wild type probably reflects lower β-(1,3)galT expression from the heterologous ureA promoter than from the wild-type promoter. Alternatively, the complemented strains are cell populations mixed in β-(1,3)galT frame status because of poly-C tract phase variation, as was indicated by sequence analysis.

Flow cytometric analysis of Le^b phenotypes in wild-type and mutant H. pylori strains confirmed results determined by ELISA, providing an alternative phenotyping method. Although ELISA determines the overall phenotype of a cell population, flow cytometry determines the phenotype of each individual H. pylori cell, providing an indication of the population structure, which is highly advantageous for studying genetically diverse organisms. The results indicate that the descendants of the complemented strain vary in their Le^b phenotype, and as indicated by the sequence data, generated by phase variation of the metastable poly-C tracts of β-(1,3)galT, futA, futB, and futC.

In conclusion, we have developed a tractable animal model to examine the hypothesis that changes in Le phenotypes are a mechanism used by H. pylori to adapt to specific host milieus. Our studies provide evidence that H. pylori can change Le phenotype to the type I pathway in vivo to match the Le phenotype of its host, extending previous studies in humans and rhesus monkeys (Wirth et al., 1997; Wirth et al., 2006). That the pathway involves phase-variable Le antigen synthesis genes provides a mechanistic explanation for the observed phenotypic changes, and can be harnessed to allow quantitative analysis of the force of selection.

Wild-type and Le^b-transgenic FVB/N mice breeding pairs were provided by the Washington University Animal Facility, and were bred and maintained under specific pathogen-free conditions in microisolators in a barrier facility, as previously described (Falk et al., 1995; Guruge et al., 1998). All experiments using mice were performed using protocols approved by the Animal Studies Committee of Washington University. DNA was extracted from the mouse tails using the QIAamp Tissue Kit (Promega) and used as a template in two PCR reactions to determine the mouse genotype: one with primers specific for α-actin (control; gene amplified in all mice), and one with primers specific for hGH (only present in transgenic animals; Table S2).

Mice were challenged with H. pylori strain HP1, isolated from a Peruvian patient with gastritis (Guruge et al., 1998). Before inoculation, H. pylori strain HP1 was grown for 24 h and harvested in Brucella broth. Cell concentration was adjusted to an OD of 2 (∼6 × 10⁷ cells) at 550 nm, and 0.4 ml of cell suspension was used to inoculate, via orogastric gavage, 20 wild-type FVB/N and 20 Le^b-transgenic FVB/N mice. Orogastric challenge was repeated for three consecutive days. At 4 wk, 8 wk, 4 mo, and 8 mo, five mice of each genotype were sacrificed and their stomachs were removed. One third of each stomach was homogenized in 1× PBS (pH 7.4), and undiluted and 1:10 diluted suspensions were plated on Skirrow's medium agar plates (BBL Microbiology Systems) and incubated for ∼72 h at 37°C and 5% CO₂. CFUs were determined for each gastric sweep and cell populations were harvested; 10–11 single colonies were picked from each gastric sweep and expanded for further analysis.

H. pylori strains used in this study were routinely grown on Trypticase soy agar/5% sheep blood plates (BBL Microbiology Systems) or Brucella agar (BA) with 10% newborn calf serum (NCS) supplemented with the appropriate antibiotic (e.g., vancomycin, kanamycin, or chloramphenicol). H. pylori strains JP26, a wild-type Le^b-positive strain isolated in Japan, and 99-8, an Le^a-positive strain, were included as controls in mutagenesis and complementation experiments.

H. pylori Le antigen phenotypes were determined by ELISA using monoclonal antibodies to Le^a, Le^b, Le^x, or Le^y (Signet Laboratories, Inc.) and protocols described previously (Wirth et al., 1996). ODs at 410 nm were determined on a microplate reader (MRX; Dynatech Laboratories Inc.). Corrected OD values were determined by averaging the OD values of two or three wells per sample and subtracting the blank (Escherichia coli strain HB101).

Levels of anti–H. pylori antibodies in mouse sera were determined by ELISA, essentially as previously described (Wirth et al., 1997). Mouse sera were tested against H. pylori strains 98-964 and J166 whole-cell antigens, as well as the following LPS antigens: wild-type strain J166 (Le^x = 0; Le^y = 3.36), J166 ΔfutA/ΔfutB (Le^x = 0; Le^y = 0), and J166 ΔfutC (Le^x = 2.22; Le^y = 0; a gift of E. Sanabria-Valentín, New York University School of Medicine, New York, NY).

Three H. pylori isolates were chosen for sequence analysis: HP1 (98-964; inoculum strain), 00-8B (03-261; non-Le^b expressor, isolated from a wild-type mouse), and 00-16A (03-270; Le^b expressor, isolated from an Le^b-transgenic mouse). Isolates were grown for 48 h, and harvested in 1ml sterile PBS. Cells were pelleted for 5 min at ∼4,300 g, and the extracts were prepared for genomic analysis with the Wizard Genomic DNA Purification Kit (Promega).

Genomic DNA recovered from these strains was used as a template for PCR amplification of the Le antigen synthesis genes using primers specific for each known gene (Table S2). PCR amplification was verified with agarose gel electrophoresis, and products were purified using a PCR purification kit (QIAGEN), quantified, and subjected to sequence analysis (SeqWright, Inc.). Because of the inherent difficulties of analyzing long homonucleotide repeat regions and regions with a high GC content (Stirling, 2003), the poly-C tract regions of β-(1,3)galT, futB, and futC were amplified with primers specific to these regions (Table S2). PCR products underwent direct sequence analysis or were cloned into pGEM-T Easy (Promega), and recombinant plasmids were analyzed by standard sequencing methods (SeqWright, Inc.) and by using a protocol for GC-rich DNA (GeneWiz).

To further characterize the intrastrain variation in the homonucleotide regions of futB and futC, the pGEM-T Easy plasmids containing PCR fragments of these regions were transformed into competent E. coli DH5α (Invitrogen) and plated on Luria-Bertani agar plates with 50 µm/ml X-gal (Thermo Fisher Scientific) and 100 µm/ml ampicillin. For each strain, 5–10 transformants were selected, and the recombinant plasmids were purified and subjected to sequence analysis with the universal primers T7F and SP6R (Promega). To determine the nucleotide sequence of the 5′ region of the complemented copy of β-(1,3)galT, PCR amplification was performed using primers A17476 and jhp0563(+482)R (Table S2), and products were purified and sequenced as described. To determine the length of the poly-C tract in the complemented mutants, 10 single colonies were isolated from 03-270::p98B13k::p70B13comp and their Le antigen phenotypes were determined. Subsequently, the isolates with the highest, lowest, and intermediate Le^b expression provided template genomic DNA for PCR amplification of the β-(1,3)galT poly-C tract region using primers jhp0563(+351)F and Gal(1,3)R(+770). PCR products were cloned into pGEM-T Easy, recombinant plasmids were transformed into E. coli as described, recombinant plasmids were purified from 10 single transformants per strain, and DNA sequence analysis was performed using primers jhp0563(+351)F and Gal(1,3)R(+770).

To disrupt β-(1,3)galT, a knockout plasmid, p98B13k, was constructed. The β-(1,3)galT ORF was amplified from strain 98-964 (HP1) with primers Jhp0562(+630)F and Jhp0564(−10)R (Table S2), and the 1.8-kb product was cloned into pGEM-T Easy. A nonpolar kanamycin resistance cassette (aphA-3) was amplified from pUCK18K2 (a gift from H. de Reuse, Institut Pasteur, Paris, France; Ménard et al., 1993) and inserted into a unique HindIII site to interrupt β-(1,3)galT. Orientation of the insertion was confirmed by restriction digestion using EcoRI and EcoRV. Transformation of H. pylori strains with p98B13k was performed essentially as previously described (Israel et al., 2000), and transformation with pCTB8 (Cover et al., 1994) was included as a positive control. Transformants were selected on BA plus 10% NCS containing 10 µg/ml vancomycin and 25 µg/ml kanamycin. Confirmation of insertion of the plasmid into the chromosome was confirmed by PCR using primers Jhp0562(+630)F and Jhp0563(+482)R (Table S2). The Le antigen phenotypes of all mutants were determined by ELISA as described.

To restore β-(1,3)galT activity in H. pylori, a complementation plasmid, p70B13comp, was introduced into the β-(1,3)galT::aphA-3 strains via natural transformation (Israel et al., 2000). The complementation plasmid was created using pADC::uvrD (pHPuD), a plasmid containing the H. pylori ureA/ureB genes interrupted by a chloramphenicol resistance cassette (Kang and Blaser, 2006b). A copy of β-(1,3)galT, amplified from strain 03-270 (Le^b positive) using primers B(1,3)compF1 and B(1,3)compR2 (Table S2), was inserted just downstream of the ureA promoter. Transformants were selected based on chloramphenicol resistance and were rapidly screened for disruption of the ureA/ureB locus by the urease test. Insertion of β-(1,3)galT into the ureA/ureB locus was confirmed by PCR using chromosomal DNA from the transformed strains that had urease-negative, chloramphenicol-resistant phenotypes, and the Le antigen phenotypes were determined by ELISA. Transformants were not recovered for any strains tested when p70B13comp was used as donor DNA, with the exception of JP26. Thus, to increase the efficiency of transformation, first a streptomycin-resistant derivative of 98-964 transformed with p98B13k and p70B13comp was selected, and after verification of genotype, chromosomal DNA from this mutant was used as donor DNA in a subsequent transformation with the remaining β-(1,3)galT null mutants.

For each H. pylori strain examined by flow cytometry, one plate of cells was harvested and resuspended in 1 ml of sterile saline and pelleted at 8,000 krpm for 5 min, washed, pelleted, and resuspended in 1.2 ml saline. The cell suspension was then passed through a 40-µm cell strainer (Falcon; BD), and 50-µl aliquots were pelleted and resuspended in 100 µl 1% Tween 20 in PBS buffer (pH 7.4) to prevent cell aggregation. Cells were incubated for 30 min with 10 µl of 1:25 diluted anti-Le^b monoclonal antibody (mouse IgM) as described, washed in Tween-PBS buffer to remove unbound antibody, and resuspended in buffer. As a positive control, cells were also incubated for 30 min with 10 µl of 1:25 diluted high titer serum from an H. pylori–positive patient (Blaser et al., 1995). Cells were washed again, resuspended, and stained for 20 min simultaneously with 10 µl of 20 µg/ml Alexa Fluor 488 goat anti–mouse IgM (μ chain specific) and Alexa Fluor 647 goat anti–human IgG (human specific; Invitrogen). The cells were washed again and resuspended in 100 µl of buffer. Then, 400 µl of 1.25% formaldehyde in saline was added, followed by 5 µl of 0.1 mg/ml propidium iodide (PI). Cells were allowed to fix overnight at 4°C. As controls, cells were reacted with the secondary antibodies alone. Flow cytometric analyses were performed on a FACSCalibur (BD). PI fluorescence intensity was used to discriminate cells from inorganic particulates. Forward scatter and 90° angle scattered laser light intensities were used to distinguish intact bacteria (R1) from cellular debris (Fig. 5 A). For each sample, 20,000 R1-gated events were acquired in listmode and subsequently analyzed using CellQuest Pro software (BD).

The Welch two-sample t test and the Mann-Whitney test were used where appropriate, with P < 0.05 considered significant. Ratios of Le^a to Le^b expression of 0 for 00-16–derived isolates were adjusted to 0.01 to calculate log₁₀ values in Fig. 2 B.

Fig. S1 outlines H. pylori Le antigen synthesis pathways. Fig. S2 is a schematic of the experimental challenge of wild-type and Le^b-transgenic mice with H. pylori. Fig. S3 shows the Le^x and Le^y phenotypes of H. pylori populations recovered 4 wk to 8 mo after challenge of wild-type and Le^b-transgenic mice. Fig. S4 shows serum levels of anti–H. pylori IgG in mice after challenge with strain HP1 (98-964). Table S1 shows serum levels of anti-LPS antibodies after challenge with strain HP1 (98-964). Fig. S5 shows the variation in futB in clones derived from H. pylori strain 98-964 and progeny. Fig. S6 shows the variation of futC in clones derived from H. pylori strain 98-964 and progeny. Fig. S7 shows PCR amplification of the babA locus in monkey- and mouse-derived H. pylori isolates. Fig. S8 shows the complementation strategy for β-(1,3)galT. Table S1 shows serum levels of anti-Le antigen IgG in mice 4 or 8 mo after challenge with H. pylori strain HP1. Table S2 lists the oligonucleotide primers used in this study.

The authors thank H. de Reuse for pUC18K2 and E. Sanabria-Valentín for purified LPS preparations.

This work was supported in part by grants RO1GM63270 and 2T32 AI007180 from the National Institutes of Health, and by the Medical Research Service of the Department of Veterans Affairs.

The authors have no conflicting financial interests related to this work.