Helicobacter pylori Cholesteryl α-Glucosides Contribute to Its Pathogenicity and Immune Response by Natural Killer T Cells

Approximately 10–15% of individuals infected with Helicobacter pylori will develop ulcer disease (gastric or duodenal ulcer), while most people infected with H. pylori will be asymptomatic. The majority of infected individuals remain asymptomatic partly due to the inhibition of synthesis of cholesteryl α-glucosides in H. pylori cell wall by α1,4-GlcNAc-capped mucin O-glycans, which are expressed in the deeper portion of gastric mucosa. However, it has not been determined how cholesteryl α-glucosyltransferase (αCgT), which forms cholesteryl α-glucosides, functions in the pathogenesis of H. pylori infection. Here, we show that the activity of αCgT from H. pylori clinical isolates is highly correlated with the degree of gastric atrophy. We investigated the role of cholesteryl α-glucosides in various aspects of the immune response. Phagocytosis and activation of dendritic cells were observed at similar degrees in the presence of wild-type H. pylori or variants harboring mutant forms of αCgT showing a range of enzymatic activity. However, cholesteryl α-glucosides were recognized by invariant natural killer T (iNKT) cells, eliciting an immune response in vitro and in vivo. Following inoculation of H. pylori harboring highly active αCgT into iNKT cell-deficient (Jα18−/−) or wild-type mice, bacterial recovery significantly increased in Jα18−/− compared to wild-type mice. Moreover, cytokine production characteristic of Th1 and Th2 cells dramatically decreased in Jα18−/− compared to wild-type mice. These findings demonstrate that cholesteryl α-glucosides play critical roles in H. pylori-mediated gastric inflammation and precancerous atrophic gastritis.


Introduction
The gastric pathogen Helicobacter pylori is a bacterium that infects over 50 percent of the world's population [1]. If untreated, this infection leads to chronic gastritis and development of pyloric gland atrophy, peptic ulcer, intestinal metaplasia, gastric carcinoma, and mucosa-associated lymphoid tissue (MALT) lymphoma [2].
The initial host response to H. pylori is strong neutrophilic recruitment, which leads to gastric epithelial damage and is followed by chronic inflammation [3,4]. Such chronic inflammation is associated with infiltration of lymphocytes and plasma cells, forming MALT. In this process, venules in the gastric lamina propria begin to exhibit a high-endothelial venule (HEV)-like phenotype, which likely facilitates immune cell infiltration. Indeed, we have shown that induction of HEV-like vessels is associated with recruitment of mononuclear cells to inflammatory sites, and that eradication of H. pylori with antibiotics and treatment with proton pump inhibitors leads to disappearance of HEV-like vessels and diminished mononuclear cell infiltration [3].
After infection, H. pylori primarily colonizes surface mucosa of the stomach and rarely reaches deeper portions of the gastric mucosa [5,6], although a more invasive and intracellular infection has also been proposed [7]. Gastric mucins are divided into surface and gland mucins [8]. The latter, consists of MUC6, are found in deeper regions of the stomach and are characterized by expression of a1,4-linked N-acetylglucosamine (a1,4-GlcNAc) attached to core 2-branched O-glycans, which is absent in the surface mucin, MUC5AC [6,9]. It is known that MUC6 is exclusively expressed in mucous neck cells and pyloric glands of the gastric mucosa, while MUC5AC is expressed in gastric surface mucous cells in the stomach [10]. These two types of mucins form a surface mucous gel layer exhibiting an alternating laminated array [11]. Since this differential distribution coincides with distribution of H. pylori, we previously examined the antibiotic activity of a1,4-GlcNAc mucin and found that a1,4-GlcNAc-containing mucins inhibit H. pylori growth by blocking synthesis of cholesteryl a-glucosides [12], the major component of H. pylori cell wall lipids [13]. Moreover, mutant mice deficient in a1,4-N-acetylglucosaminyltransferase exhibit adenocarcinoma, indicating that a1,4-GlcNAc-containing mucins function as tumor suppressors [10]. Significantly, H. pylori lacks cholesterol and must incorporate it from surrounding host epithelial cells [14]. Cholesteryl aglucosyltransferase (aCgT) adds an a-glucosyl residue to cholesterol [15], forming cholesteryl a-glucoside (aCGL). aCGL is further derivatized in H. pylori to form cholesteryl acyl aglucoside (aCAG), cholesteryl phosphatidyl a-glucoside (aCPG), and cholesteryl phosphatidyl monoacyl a-glucoside (aCPG (monoacyl)) [13]. We previously cloned aCgT using the shotgun method [16] and showed that its activity is inhibited by core 2 O-glycan capped by a1,4-GlcNAc residues [17]. However, the function of cholesteryl a-glucosides in the pathogenesis of H. pylori infection has not been determined.
Invariant natural killer T (iNKT) cells are recognized as immune cells that react with glycolipids. iNKT cells express the T cell receptor (TCR) encoded by Va24-Ja18 and Va14-Ja18 rearrangements in human and mouse, respectively [18,19]. These TCRs recognize glycolipid antigen presented by CDld, a nonclassical MHC class I-like antigen distinct from c-type lectins that mediate leukocyte and lymphocyte adhesion [20,21]. iNKT cells exhibit unique aspects of both innate and adaptive immunity, distinguishing them from innate immune natural killer (NK) cells [22][23][24]. Both activated iNKT and NK cells can rapidly produce large amounts of various cytokines such as interleukin (IL)-4, interferon-c (IFN-c), tumor necrosis factor-a (TNF-a) and IL-17, which likely stimulate different immune cell populations with diverse functions [19]. The potent TCR antigen of iNKT cells is a-galactosylceramide but also includes galactosyl diacyl glycerol present in Lyme disease-causing Borrelia burgdorferi [25,26], agalacturonic ceramide from Sphingomonas spp [27], and aglucosyldiacylglycerol from Streptoccocus pneumonia [28]. However, how iNKT cells respond to these bacteria during the course of infection in human patients is not known.
Here, to characterize aCGL function in the innate immune response, we first isolated the aCgT gene from H. pylori retrieved from stomach tissues of H. pylori-infected patients. We found that the activity of cloned aCgT from clinical isolates positively correlates with the atrophy score of stomach tissue. We then constructed recombinant H. pylori harboring aCgT from different clinical isolates and found that aCgT activity is positively correlated with susceptibility to iNKT cells. Moreover, H. pylori containing highly active aCgT were recovered from iNKT celldeficient mice at levels dramatically higher than from wild-type (WT) mice. In vitro and in vivo analysis identified aCPG (monoacyl) is the most potent antigen for iNKT cells among H. pylori cell components. These findings demonstrate that cholesteryl aglucosides induce an immune response by iNKT cells, thus causing stomach inflammation due to H. pylori infection.

Results
H. pylori aCgTs isolated from Japanese patients show varying levels of activity relative to aCgT from H. pylori 26695 To determine the role of cholesteryl a-glucosides in H. pylori pathogenesis in the stomach, aCgT genomic DNA was isolated from clinical H. pylori isolates from the stomachs of 24 H. pyloriinfected Japanese patients. Amino acid sequences deduced from various aCgT genomic sequences showed at least 20 different amino acid substitutions compared to aCgT from control WT H. pylori 26695, whose whole genome has been sequenced [29] ( Figure 1A). DNA encoding aCgT H. pylori 26695 WT was mutated by site-directed mutagenesis to create sequences corresponding to clinical isolates, and mutant proteins were expressed in a bacterial expression vector [30] and their activities measured. Some enzymes showed activity higher than WT aCgT from H. pylori 26695, while others showed decreased activity ( Figure 1B), as indicated in yellow and blue, respectively, in Figure 1A.
The amino acid sequence of aCgTs derived from clinical isolates of 18 European and 5 Indian patients was also determined (data not shown). A tyrosine substitution for WT histidine at position 41, which is an activating mutation, is observed in all Japanese isolates; that mutation was only occasionally seen in isolates of European and Indian origin (data not shown), indicating that protein sequences from Japanese patients are more uniform than those isolated from Indian and European individuals. Moreover, all H. pylori isolates from Japanese patients harbored genes encoding the most toxic form of cagA and vacA (cagA-positive and vacA s1/m1, data not shown) [31][32][33]. However, more than half of the Indian and European clinical isolates harbored the much less toxic vacA s1/m1 or non-toxic vacA s2/m2, and about a quarter of the European H. pylori specimens lacked cagA (data not shown). Due to this diversity, for the remainder of the experiments reported here, we analyzed H. pylori from Japanese patients only.
To determine the effect of amino acid substitutions seen in different H. pylori clones, the entire aCgT sequence in H. pylori 26695 was replaced with sequences present in the 16 different patterns of substitutions of aCgT amino acids and expressed in Escherichia coli, and the mutant aCgT proteins were purified. The activity of those recombinant proteins showed significant variation among clinical isolates, and more than half of the aCgT variants showed increased activity relative to WT H. pylori ( Figure 1C).
To determine potential effects of amino acid substitutions on aCgT structure, we firstly attempted to determine the aCgT crystal structure. However, since we could not accomplish this task due to aCgT hydrophobicity, we searched databases for enzymes of similar structure [34]. Our search clearly identified aCgT as a member of the GT-4 family. GT-4 proteins exhibit two Rossmann-fold domains with the active site in a cleft between the two domains. The best hit with a known 3-dimensional structure was phosphatidylinositol mannosyltransferase from Mycobacterium smegmatis (PDB code 2GEJ) [35]. In this case, sequence identity was only 17%, but a ''Z-score'' of 265.2 indicated high structural similarity (a value of 29.5 indicates 97% confidence) [34]. There was only one major gap in the primary sequence of aCgT: a 12-residue insertion in a loop far from the GDP-mannose binding site. We next built a 3-dimensional model using 2GEJ as template ( Figure 2). The model was of high quality as judged by the distribution of hydrophobic residues in the protein core and hydrophilic residues on the surface. Moreover, analysis of the UDP-glucose binding pocket revealed that most of the critical binding residues were identical, including residues implicated in catalysis [35]. Notably, all of the aCgT amino acid substitutions from Japanese patients are located in the surface region of the aCgT sterical structure and are absent in the UDP-Glc binding pocket ( Figure 2). We next used homologous recombination to replace the H. pylori 26695 aCgT sequence with sequences from H. pylori harboring aCgT of higher (aCgT high , strain #10) and lower (aCgT low , strain #17) activity or to create H. pylori lacking the aCgT gene altogether (aCgT D ) in order to compare these variants with parental H. pylori 26695 (aCgT cont ) ( Figure S1 and Table S1 in File S1). As anticipated, H. pylori aCgT high synthesized greater amounts of cholesteryl a-glucosides than did H. pylori aCgT low , while H. pylori lacking aCgT synthesized no cholesteryl aglucosides ( Figure 3A). Among different cholesteryl a-glucosides, aCGL was the most abundant in products of H. pylori aCgT low , most likely because aCGL was not converted to aCAG or aCPG.
Significantly, the aCgT D H. pylori strain grew much more slowly and entered plateau phase earlier than did the parental H. pylori 26695 (WT H. pylori), and the aCgT high H. pylori clone grew faster than the WT form in liquid culture ( Figure 3B). Electron microscopic analysis showed that some aCgT D H. pylori exhibited an aberrant coccoid form [36,37] (Figure 3C). These results indicate that cholesteryl a-glucosides are critical for H. pylori growth and normal morphology.

H. pylori aCgT activity is highly correlated with stomach atrophy
We then asked if aCgT levels, and thus those of cholesteryl aglucosides, were correlated with pathogenesis of H. pylori infection. Histological grading of 24 human biopsy samples of the gastric mucosa collated together with the 24 Japanese H. pylori strains were judged by five different criteria using the updated Sydney The aCgT amino acid sequence from 24 clinical isolates was compared to that of H. pylori 26695, whose whole genome sequence has been reported. Only variant residues are shown. Residues in yellow and blue represent substitutions that yield higher and lower aCgT activity, respectively, relative to aCgT from H. pylori 26695. Proteins, of which amino acid residues shown in white boxes were substituted, were not soluble as a recombinant protein in a bacterial expression system and therefore enzyme activity was not assayed. (B) cDNA encoding the amino acid sequence of aCgT from H. pylori 26695 in an expression vector was mutated by site-directed mutagenesis to reproduce residues seen in H. pylori clinical isolates. Bacterially expressed aCgT was assayed using [ 3 H]UDP-glucose and cholesterol as described in File S1. (C) The entire aCgT sequence in the expression vector was replaced with sequences from H. pylori clinical isolates and activity of expressed aCgT was assayed. The assay was performed in triplicate and repeated twice in both (A) and (B). Representative results are shown. doi:10.1371/journal.pone.0078191.g001 System [38]: H. pylori infection load, recruitment of neutrophils, infiltration of mononuclear cells, glandular atrophy (antrum and corpus), and intestinal metaplasia, as illustrated in Figure S2A and B. These criteria were evaluated each as four grades scored from 0 to 3: normal (score 0), mild (1), moderate (2), and marked (3). In general, more advanced atrophy was characterized by a decrease in the number of pyloric glands in the antrum and fundic glands in the corpus (data not shown). More advanced atrophy was also associated with recruitment of mononuclear cells and formation of intestinal metaplasia (data not shown). Histological assessment of the gastric mucosa was undertaken based on pathology reports, which were reviewed by one senior pathologist (J.N., Figure S2B). When aCgT activity of all strains was plotted against these parameters, activity was highly correlated with the total atrophy score (antrum and corpus) ( Figure 4A). Replotting of those data, as indicated by the dotted rectangle in Figure 4A, showed that the total atrophy score is positively correlated with mononuclear cell recruitment ( Figure 4B). Moreover, the number of H. pylori colonies was inversely correlated with aCgT activity ( Figure 4C). These results strongly suggest that cholesteryl a-glucosides induce an immune response causing increased inflammation, yet that response decreases the number of surviving H. pylori. Furthermore, the atrophy score was correlated with intestinal metaplasia ( Figure 4D), a precancerous phenotype [39], indicating that a high atrophy score predicts progression to gastric carcinoma.
Cholesteryl a-glucosides are responsible for an iNKT cell immune response To determine which immune cells play a critical role in H. pylori infection, we first assayed phagocytosis by macrophage-like differentiated THP-1 cells ( Figure S3A) and antigen presentation by dendritic cells (DCs) ( Figure S3B). Phagocytosis was observed in all aCgT clones, regardless of aCgT activity ( Figure S3A). Activation of DCs was measured by the expression of 3 markers, CD86, HLA-DR, and CD40. CD86 and HLA-DR are known antigen-presentation markers and CD40 is a differentiation marker of mature DCs. None of aCgT clones exhibited a significant difference in terms of recognizing H. pylori expressing aCgT high , aCgT low , or aCgT D , in antigen presentation by DCs ( Figure S3B). This result suggests that generally all aCgT clones analyzed are similarly recognized by macrophages and DCs.
We then tested the possibility that iNKT cells exert a differential response to H. pylori expressing aCgT high or aCgT D . Upon recognition of WT H. pylori, a mouse hybridoma iNKT cell line produced significant amounts of IL-2, an indicator of iNKT cell activation. Those levels were roughly equivalent to stimulation by a-galactosylceramide, a bona fide antigen for iNKT cells. By contrast, H. pylori lacking aCgT were barely recognized by the same iNKT hybridoma based on failure to elicit an IL-2 response ( Figure 5A). These results indicate that iNKT cell immune responses are largely due to recognition of cholesteryl a-glucosides.
Cholesteryl a-glucosides constitute 25% of total H. pylori lipids and comprise three major forms and one minor form ( Figure S4) [13]. All four forms were synthesized and their structure confirmed by NMR. When we evaluated the three major forms of synthetic cholesteryl a-glucosides in vitro, cholesteryl phosphatidyl a-glucoside (aCPG) elicited the highest iNKT cell response when glycolipids were initially dissolved in DMSO ( Figure S5A), although the response toward aCPG was significantly less robust than to a-galactosylceramide. This observation is consistent with a generally weak interaction of the glycolipid with CD1d ( Figure  S5B) and/or the T-cell receptor. Cholesteryl b-glucoside (bCGL) was not recognized by iNKT cells ( Figure S5A). In other experiments, DCs were isolated from bone marrow and differentiated using GM-CSF. These immature DCs were then incubated with the four synthetic forms of cholesteryl a-glucosides and injected intraperitoneally into WT mice. The presence of iNKT cells in liver, where they are more abundant than in other tissues, was evaluated 16 hours later. Interestingly, the monoacylated form of aCPG, which is reportedly a minor component of cholesteryl a-glucosides [40], was the most potent antigen in the in vivo assay ( Figure 5B). Consistently, an isoelectrofocusing assay showed that aCPG (monoacyl) was the only lipid that interacted with CD1d ( Figure S5B). These results suggest that fatty side chain(s) in cholesteryl a-glucosides are potentially important for CD1d recognition.

H. pylori growth increases in Va14 iNKT cell-deficient mice
To determine the role of iNKT cells in the in vivo immune response to cholesteryl a-glucosides, H. pylori clones expressing different aCgTs were inoculated into the stomach of WT or Va14 iNKT-cell knockout (Ja18 2/2 ) mice, which were generated by genetic deletion of a T cell receptor (Va14) that recognizes CD1d-bound glycolipids and is unique to iNKT cells [41]. Ten days after the 3rd inoculation, mice were sacrificed and the stomach was excised. Previous reports indicate that macrophage and neutrophil recruitment subsides by 10 days after H. pylori inoculation, while T lymphocyte recruitment is initiated 10 days after inoculation [42]. Indeed, histochemical analysis showed that surface mucosa from H. pylori-infected Ja18 2/2 and WT mice was indistinguishable and only a few mononuclear cells, neutrophils or macrophages had been recruited by the 10 day time point (Figure S6A and B). Under these same conditions, aCgT high H. pylori were recovered at lower levels from the stomach of WT mice than from aCgT cont H. pylori-infected WT mice. Furthermore, recovery of aCgT high H. pylori was five times greater in Ja18 2/2 than in WT mice ( Figure 6A). Such a substantial increase relative to WT mice was not observed when aCgT D H. pylori were inoculated into Ja18 2/2 mice. Significantly, increased mRNA expression of iNKT cells (Table S2 in File S1) were expressed in the stomach and/or underwent proliferation upon inoculation of aCgT high H. pylori and WT H. pylori relative to controls ( Figure 6B). Significantly, a greater number of transcripts expressed in iNKT cells were present in stomach tissue derived from patients infected with H. pylori than in control samples, based on quantitative real-time PCR analysis of polyA + RNA isolated from those tissue specimens ( Figure S7). Overall, these results suggest that H. pylori containing cholesteryl a-glucosides induce proliferation and/or recruitment of iNKT cells to the stomach, where they attack H. pylori in infected tissue.
To evaluate the consequences of an iNKT response upon H. pylori infection, we examined levels of mRNAs encoding cytokines. Following infection of mice with aCgT high bacteria, IFN-c, IL-12p40, and IL-4 transcript levels were significantly decreased in Ja18 2/2 relative to WT mice ( Figure 6C, D, and E). Similarly, expression of IL-2, IL-5, IL-10, and lymphotoxin (LT)-b was reduced in Ja18 knockout mice ( Figure S8A, B, D, and F). These results indicate that iNKT cell activation by H. pylori cholesteryl aglucosides promotes proliferation and/or recruitment of cells producing Th1 cytokines (IFN-c, IL-2, and IL-12p40) [43], Th2 cytokines (IL-4, IL-5), and a regulatory cytokine (IL-10) [44]. Th17 cells, which are marked by RORct expression [45], are reportedly associated with an autoimmune responses [46]. We also observed increased RORct expression as well as that of IL-10, which encodes an inducer of regulatory T cells (Tregs), in stomach tissue from WT infected mice, both indicators of iNKT cell activation ( Figure S8E and G). However, Foxp3 [47] (Table S2 in File S1) and IL-22 transcripts were hardly detectable, even following analysis using different sets of RT-PCR primers ( Figure 6F and Figure S8H), indicating that Tregs were not activated. These results indicate that recognition of cholesteryl aglucosides by iNKT cells stimulates immune cell responses in various cell lineages, including Th1, Th2, and Th17 cells, and that these responses are associated with decreased recovery of H. pylori.

Discussion
Infection of the stomach with H. pylori induces an acute immune response mediated predominantly by neutrophil infiltration, but the subsequent innate response promotes chronic inflammation mediated by various immune cells. This long-term phase apparently induces HEV-like vessels in gastric mucosa, facilitating T and B lymphocyte recruitment to inflammatory sites in the stomach. We previously found that the presence of H. pylori in the stomach is necessary to facilitate lymphocyte recruitment, and that H. pylori eradication abrogated HEV-like vessels and dramatically decreased the number of mononuclear cells [3]. Similarly, we found recently that gastric MALT lymphoma is associated with the appearance of HEV-like vessels that express MECA-79 negative sialyl Lewis X [48], suggesting that non-sulfated sialyl  Lewis X recruits lymphocytes during progression of gastric MALT-lymphoma.
Here we show that cholesteryl a-glucosides in H. pylori play a critical role in the early phase of inflammation in H. pylori-infected mice. Moreover, we observed highly significant diversity in aCgT amino acid sequences depending on clinical isolates and found that aCgT activity was highly correlated with progression of gastric mucosal atrophy. aCgT amino acid substitutions were not seen at the UDP-Glc binding site, which is located at an inner hydrophobic pocket of aCgT. This observation supports our conclusion that aCgT is essential for H. pylori growth and that mutations that alter UDP-Glc binding abolish H. pylori viability. Amino acid substitutions that we observed in aCgT from the clinical isolates likely result in aCgT conformational changes, resulting in increased or decreased aCgT activity.
It has been reported that H. pylori is one of the most diverse and variable bacterial species studied. Genetic variation can be generated in a bacterial population by mutation and/or recombination between different strains. As a result, H. pylori exhibits extensive genetic variation, so that almost every infected individual carries their own H. pylori clinical strain. [49]. To investigate the effect of genomic diversity on aCgT activity, we analyzed 24 clinical isolates from Japanese patients, as well as samples from 5 Indian and 18 Europe individuals. Genotyping analysis of H. pylori cagA and vacA, which encode toxic factors [31][32][33], showed that isolates from all Japanese patients contained toxic forms of these genes (cagA-positive, vacA; s1/m1), while some H. pylori isolates from European and Indian patients showed cagA-negative or weakly toxic or non-toxic vacA subtypes. Moreover, substitution of the histidine residue at position 41 with tyrosine activated aCgT relative to wild-type H. pylori. That substitution was observed in all clinical isolates from Japanese patients, while only a fraction of European and Indian patients harbored that substitution. This finding, together with the presence of CagA and the toxin known as VacA, may account for the high prevalence of gastric cancer in Japan. Additional large-scale examination of H. pylori isolates from patients in different countries such as European nations is necessary to support this hypothesis.
The present study revealed that aCgT activity is higher in H. pylori isolated from patient clinical isolates. As biospecimens were isolated upon diagnosis, the specimens used in this study came from patients who had been infected for varying periods of time. It is noteworthy that aCgT activity is highly correlated with glandular atrophy, regardless of infection history. The degree of atrophy is also correlated with intestinal metaplasia, a putative precancerous condition, supporting the idea that the inflammatory response leads to gastric cancer [50]. Moreover, unlike normal gastric mucosa, which shows stable expression of a1,4-GlcNAc residues in gastric glands, it is reported that expression of a1,4-GlcNAc residues containing O-glycans in H. pylori-associated intestinal metaplasia is significantly reduced [51]. This fact suggests that disappearance of core 2 O-glycans capped by a1,4-GlcNAc residues may function in the process of intestinal metaplasia. Although previous studies demonstrate a role for iNKT cells in various chronic infection-inflammation states, our work demonstrates that iNKT cell-mediated chronic inflammation is directly correlated to disease progression in human patients.
Here, we generated recombinant H. pylori in which only the aCgT gene was replaced with forms seen in clinical isolates. We found that H. pylori harboring aCgT high grows more efficiently than does H. pylori expressing aCgT low . Moreover, some H. pylori lacking aCgT exhibited a coccoid morphology. These results demonstrate that aCgT is critical for H. pylori growth. However, forms of H. pylori that express different amounts of cholesteryl aglucosides did not induce a differential response toward macrophages and DCs, supporting the idea that cholesteryl a-glucosides are antigens recognized by iNKT cells. Notably five-fold greater levels of H. pylori aCgT high were recovered from Ja18 2/2 than from WT mice. This difference was much greater than differences observed in WT and Ja18 2/2 mice infected with either aCgT cont or aCgT D H. pylori. These results clearly show that excess cholesteryl a-glucosides are recognized by iNKT, reducing H. pylori infection. Our findings differ from a previous report showing that cholesteryl a-glucosides protect H. pylori from immune cell attack [14]. That work relied on only one H. pylori mutant lacking cholesteryl a-glucosides. Thus, anomalies in that mutant H. pylori may have perturbed the immune response.
Among three major H. pylori cholesteryl a-glucosides, aCPG was identified as the most potent antigen for iNKT cells based on an in vitro assay, even though the response toward aCPG was less potent than toward a-galactosylceramide. It has been shown that CD1d has two pockets at the a-galactosylceramide binding site and that the two acyl chains of the latter fit into these pockets [52]. It is possible that two acyl chains of aCPG similarly fit into these pockets. When these different cholesteryl a-glucosides are presented from DCs, aCPG (monoacyl) is the best antigen in vivo. Therefore, in vivo one acyl chain and cholesterol with a side chain may be the optimal antigen for iNKT interaction with the CD1d pocket.
iNKT cells reportedly produce cytokines that stimulate different immune cells [19]. Under the experimental conditions used, an increase in mononuclear cells was not observed 10 days after infection [42]. iNKT cells constitute less than 3% of all T lymphocytes in many tissues [53]. Because of their low abundance, an increase in iNKT cells was not observed in either WT or Ja18 2/2 mice. It has been reported that H. pylori infection leads to Th1 cell activation [43]. However, Th2 cell activation has not been well described [54]. The present work shows that 10 days after H. pylori inoculation, the Th2 cell response is as robust as the Th1 cell response, and that those responses largely depend on iNKT cells. Since Th2 cells play a role in the tolerogenic response [44], Th2 cell activation by H. pylori may promote long-lasting attenuation of the immune response, which might underlie the chronic nature of H. pylori infections. In our present study, we did not detect an increase in Tregs ten days after infection, while a recent study reports that Tregs increase during influenza virus infection in the presence of H. pylori or aCAG [55]. Our findings suggest that the increase in Th2 cells induced by excess cholesteryl a-glucosides may promote a tolerogenic effect following H. pylori infection through iNKT cell activation.
The inflammatory response toward H. pylori infection exemplifies an inflammatory response that leads to cancer [50,56]. The novel function of H. pylori glycans that we report here significantly extends our previous understanding of the roles of glycosylation in pathogenesis [57]. Future studies should determine whether aCgT inhibition constitutes an alternative treatment for H. pylori-induced inflammation and cancer.

H. pylori strain and bacterial culture
The standard H. pylori strain 26695 (ATCC700392) was purchased from American Type Culture Collection (ATCC, Manassas, VA), and routinely grown on Tripticase Soy agar with 5% sheep blood (TSA II) (Becton Dickinson, Franklin Lakes, NJ) for 2 to 3 days at 35uC in 12% CO 2 . Bacteria were precultured in Brucella broth (Becton Dickinson) supplemented with 10% fetal bovine serum (FBS) (HyClone, Logan, UT). Subsequently, bacteria was diluted to 4610 7 cells/ml and cultured in brain heart infusion (Becton Dickinson) supplemented with 0.2% yeast extract (Becton Dickinson) and 10% FBS (BHI/YE/FBS). Bacteria cultured in BHI/YE/FBS were used for all experiments except for generation of recombinant aCgT mutants. For targeting of the aCgT gene, bacteria on TSAII plates were used directly without liquid culture.
Mice C57BL/6 mice were purchased from the Jackson Laboratory. Ja18 2/2 mice on a C57BL/6 background were generated by Dr. Masaru Taniguchi (Riken, Yokohama, Japan) [41]. All mice were housed in specific pathogen-free conditions. Animals were treated according to the guidelines of the National Institute of Health and the experiments were approved by the Institutional Animal Care and Use Committee of the Sanford-Burnham Medical Research Institute (PHS-Assurance number; A3053-01).

Ethics statement
The experimental protocol and use of all human pathology specimens for research were approved by the Ethical Committee of Shinshu University School of Medicine (Matsumoto, Japan). The Ethical Committee also granted a waiver of informed consent to use H. pylori clinical isolates and the formalin-fixed and paraffinembedded biopsy specimens retrieved from the pathology file of the Shinshu University Hospital, because the diagnostic use of the samples was completed before the study. Thus no risk to the patients involved was predicted. Samples were also coded to protect patient anonymity.

Clinical isolates
Twenty-four gastric biopsy specimens from Japanese patients were obtained by endoscopic examination at Shinshu University Hospital, Matsumoto, Japan. The patients consisted of 7 male and 17 female (ranging in age from 12 to 83 years; average 49.7 years), and only H. pylori-positive patients were evaluated. For histological assessment of chronic gastritis, at least 5 biopsies were taken; biopsy specimens were fixed in phosphate-buffered 10% formalin (WAKO, Osaka, Japan), embedded in paraffin, and cut into 5-mmthick sections. Pathological diagnosis was evaluated based on the updated Sydney System.
Clinical bacterial stocks from each specimen were stored at 280uC. H. pylori were cultivated at 35uC under microaerophilic conditions. H. pylori-selective agar plates were utilized (Eiken, Tokyo, Japan), and single colonies were incubated on TSAII plates (Becton Dickinson) for isolation of genomic DNA. Similarly, genomic DNA was isolated from 5 clinical isolates from India, 10 from Sweden, 3 from Germany, and 5 from Spain, all of which were stored in Umeå University, Umeå, Sweden [58]. Genotyping for cagA and vacA was conducted using published PCR primers [59]. The experimental protocol was approved by the Ethics Committee of Shinshu University School of Medicine and Umeå University.

Generation of aCgT mutants by homologous recombination
The strategy used to disrupt the aCgT gene and replace mutated aCgT from clinical isolates is shown in Figure S2 and File S1. Briefly, aCgT D were bacteria deficient in aCgT, and aCgT high and aCgT low harbored high and low activities, respectively, of aCgT. aCgT cont was a control clone. In vitro response to iNKT hybridoma cells Lysates from recombinant aCgT D or WT H. pylori (at 1610 7 or 2610 6 CFU/well, respectively), or 1 mg/well of synthetic compounds were incubated for 24 hours in 96-well microplates coated with 10 mg/ml mouse CD1d-tetramer, according to published methods [60]. For controls, 6 ng/well of a-galactosylceramide and/or 1610 7 CFU/well S. yanoikuyae lysate were used [27]. After washing wells with PBS, 1610 5 of mouse iNKT hybridoma cells (clone 1.2 or 1.4) were cultured for 16 hours, and then IL-2 secreted into supernatants was measured by a sandwich ELISA (BD Pharmingen, La Jolla, CA). Synthetic glycolipids were initially dissolved in dimethyl sulfoxide (DMSO) and then prepared using a series of 10-fold dilutions with assay medium prior to the assay. In parallel, the glycolipids were incorporated in liposomes as described previously [61] and assayed after 20-fold dilution.

In vivo iNKT cell activation
Mouse dendritic cells were prepared by culturing bone marrow cells in media containing 10 ng/ml mouse recombinant GM-CSF (Kyowa-Hakko-Kirin, Tokyo, Japan) for 7 days. 1610 6 DCs were then incubated with 50 mg/ml of synthetic cholesteryl aglucosides, 100 ng/ml of a-galactosylceramide, or 500 ng/ml of BbGL-IIc (a glycolipid derived from B. Burgdorferi) for 24 hours. After washing cells with PBS, 5610 5 of glycolipid-pulsed DCs were intravenously injected into C57BL/6 WT mice. Liver mononuclear cells were collected 16 hours later and analyzed directly for IFN-c and TNF-a levels by FACS Calibur (BD Bioscience) in iNKT cells. Intracellular cytokine staining of agalactosylceramide-CD1d tetramer-positive cells was carried out according to a published protocol [60].
Short-term H. pylori infection assay C57BL/6 WT or Ja18 2/2 mice were fasted overnight and orogastrically inoculated with 3610 8 CFU H. pylori in BHI/YE/ FBS by a gastric intubation tube three times at one-day intervals. Mice were maintained on a fasting regime for an additional 4 hours after each infection. Control mice were administered media only.
Mice were sacrificed at post-infection day 10 after the third bacterial infection, and the stomach was cut into along the greater curvature, washed with diethylpyrocarbonate (DEPC)-treated PBS, and divided into two pieces along the lesser curvature. For each stomach, one piece was used for measuring wet weight and then placed in 1 ml PBS for quantification of bacterial colonies. The other piece was immediately frozen on dry ice and stored at 280uC for total RNA isolation. RNA extraction and RT-PCR analysis was performed as described in File S1.
Quantification of H. pylori colonies from mouse stomach Stomach tissue prepared as described above was homogenized in 1 ml of PBS three times. Homogenates at 1:5, 1:10, 1:20, and 1:40 dilutions were prepared in PBS, and then 50 ml of each dilution was put onto an H. pylori selective-agar plate containing 30 mg/ml kanamycin in duplicate. Plates were incubated at 35uC for 5 days under microaerophilic conditions, and colonies were evaluated as CFU/gram of stomach tissue.

Statistical analysis
Statistical analysis was carried out using Prism 5 (GraphPad Software, Inc., La Jolla, CA) and evaluated by an unpaired t-test. P values of ,0.05 were considered statistically significant. Correlation coefficients as described by r values were analyzed by calculating Spearman's rank correlation coefficient.
Other experimental procedures are described in File S1. . Va24Ja18 mRNA expression was not detected in specimens from the remaining 9 patients. Total RNA derived from normal human stomach (purchased from Clontech) served as a reference control. The transcript of aCgT was not detected by RT-PCR. Anomalous levels seen in patient #16 sample could be due to decreased H. pylori, which was eradicated by antibiotic treatment before biopsies were taken.

Supporting Information
(TIF) Figure S8 Expression of cytokines and immune cell markers in H. pylori-infected stomach tissues 10 days after infection, related to Figure 6. Transcript levels of IL-2 (a Th1 cytokine, A), IL-5 and IL-6, (Th2 cytokines, B and C), IL-10 (a regulatory cytokine, D) IL-17A and IL-22 (Th17 cytokines, E and H), RORct (a Th17 cell marker, G), and LTb (F) were determined by RT-PCR. For each experiment, four WT or Ja18 2/2 mice were infected with H. pylori. Stomach samples shown in Figure 6 in the text were used. Statistical significance was evaluated using an unpaired t-test (*, P,0.05). Mean 6 S. D. are shown. (TIF) File S1 Supporting Information. Table S1. Primers used to amplify fragments of the 59 arm, the KanR gene, AECgT and the 39 arm in targeting vectors.