Helicobacter pylori causes gastritis, which has been attributed to the development of H. pylori–specific T cells during infection. However, the mechanism underlying innate immune detection leading to the priming of T cells is not fully understood, as H. pylori evades TLR detection. Here, we report that H. pylori metabolites modified from host cholesterol exacerbate gastritis through the interaction with C-type lectin receptors. Cholesteryl acyl α-glucoside (αCAG) and cholesteryl phosphatidyl α-glucoside (αCPG) were identified as noncanonical ligands for Mincle (Clec4e) and DCAR (Clec4b1). During chronic infection, H. pylori–specific T cell responses and gastritis were ameliorated in Mincle-deficient mice, although bacterial burdens remained unchanged. Furthermore, a mutant H. pylori strain lacking αCAG and αCPG exhibited an impaired ability to cause gastritis. Thus H. pylori–specific modification of host cholesterol plays a pathophysiological role that exacerbates gastric inflammation by triggering C-type lectin receptors.
Helicobacter pylori is a Gram-negative pathogenic bacterium that successfully colonizes the gastric mucosa of half of the world’s population. Persistent H. pylori infection induces chronic gastritis, which is one of most common risk factors for gastric malignancy (NIH Consensus Conference, 1994; Peek and Blaser, 2002). In addition to the virulence proteins secreted from H. pylori (Hatakeyama, 2004), recent studies have underscored the critical role of H. pylori–primed host CD4+ T cells in the development of gastritis (Eaton et al., 2001; Nagai et al., 2007). However, such T cells do not efficiently contribute to the eradication of this pathogen (Adamsson et al., 2017) and are considered deleterious for the host (D’Elios et al., 2005). Hence, insight into mechanisms by which H. pylori primes T cells may help in the prevention of chronic gastritis and malignancy.
H. pylori avoids detection by the canonical pattern recognition receptors, TLRs, by modifying its TLR ligands (Roy and Mocarski, 2007). Instead of having typical pathogen-associated molecular patterns, H. pylori produces a wealth of diverse lipids, in particular assorted cholesterol metabolites (Jan et al., 2016). For example, H. pylori extracts cholesterol from the host and converts it to unique cholesteryl glucosides, such as cholesteryl acyl α-glucosides (αCAGs; Grille et al., 2010; Haque et al., 1996; Hirai et al., 1995; Lebrun et al., 2006; Mayberry and Smith, 1983), which are reported to suppress host immunity (Wunder et al., 2006). On the other hand, their immunostimulatory properties and host pattern recognition receptors have not yet been determined.
Among the germline-encoded innate immune receptors, C-type lectin receptors (CLRs) have recently been demonstrated to recognize various pathogen-derived glycoconjugates (Geijtenbeek and Gringhuis, 2016). Notwithstanding, it is still unclear whether CLR family members recognize steryl glucosides in H. pylori and, moreover, whether CLRs are involved in gastritis induction.
In the present study, we identified host CLRs for H. pylori–derived pathogenic metabolites that cause detrimental inflammation.
H. pylori possesses immunostimulatory lipid components recognized by Mincle
To search for the immunostimulatory lipid components in H. pylori, we isolated a lipophilic extract, separated it into 16 fractions by high-performance TLC (HPTLC; Fig. 1 A), and assessed the ability of each fraction to stimulate bone marrow–derived dendritic cells (BMDCs). Peaked activity corresponding to the distinct spot (Rf = 0.83, chloroform/methanol/water [65:25:4, vol/vol/vol]) was found across fraction 13 to 14 (fraction 13-14) that induced DCs to produce inflammatory cytokines, such as TNF. This activity was independent of the TLR adaptor MyD88 but was completely abolished upon deletion of CARD9 or FcRγ (Fig. 1 B), suggesting that an FcRγ-CARD9–coupled immunoreceptor, including a CLR, is involved in ligand recognition and signaling. We therefore expressed several CLRs in combination with FcRγ in NFAT-GFP reporter cells and found that fraction 13-14 activated cells expressing macrophage-inducible C-type lectin (Mincle; Clec4e) and FcRγ; (Fig. 1, C and D). Weak activity was also detected in cells expressing macrophage C-type lectin (MCL; Clec4d) and FcRγ. MCL forms heterodimers with Mincle and stabilizes its expression on the cell surface (Lobato-Pascual et al., 2013; Miyake et al., 2015; Zhao et al., 2014). We thus created Mincle-CD3ζ and MCL-CD3ζ chimeras and confirmed that Mincle is a receptor that recognizes fraction 13-14 (Fig. 1 E). Consistent with its ability to activate Myd88−/− cells (Fig. 1 B), this fraction did not possess TLR2/4 ligand activity (Fig. S1 A). The production of TNF induced by fraction 13-14 was abrogated in Mincle-deficient BMDCs (Fig. 1, F–H), demonstrating the essential role of this CLR in the recognition of components in fraction 13-14. Fraction 13-14 was visualized by staining with copper acetate-phosphoric acid (lipid stain, Fig. 1 F) and orcinol (carbohydrate stain, Fig. 1 G), suggesting that this fraction contains a glycolipid.
Identification of αCAG as a noncanonical Mincle ligand
To identify its molecular components, fraction 13-14 was repurified by HPTLC (Fig. 1 I), and its dose-dependent activity was verified using reporter cells expressing Mincle (Fig. 1 J) and by binding to Mincle-Fc protein (Fig. 1 K). We then determined its chemical structure using mass spectrometry (MS) and nuclear magnetic resonance (NMR) spectroscopy. High-resolution electrospray ionization (ESI) time-of-flight (TOF) MS showed a molecular-related ion peak at m/z 781.5979 [M + Na]+ (Fig. 1 L), suggesting a cholesteryl acyl-pyranoside such as αCAG (calculated mass, 781.5953) as a likely candidate (Hirai et al., 1995). Methanolysis of the active fraction followed by gas chromatography (GC)–MS allowed for the identification of three saturated fatty acid methyl esters (FAMEs; 14:0, 16:0, 18:0; Fig. S1 B). The presence of a cholesterol moiety was confirmed by the detection of Δ3,5-cholesterol after methanolysis (Fig. S1, B and C).
The 1H-NMR spectrum of fraction 13-14 showed characteristic signals of cholesterol, α-glucopyranoside and saturated fatty acid (Fig. 1, M and N). Furthermore, extensive 2D-NMR analyses, including 1H-1H correlation spectroscopy, total correlation spectroscopy, heteronuclear single-quantum coherence (HSQC), and heteronuclear multiple-bond correlation (HMBC) spectroscopy, gave the main chemical moiety in fraction 13-14 as αCAG (Fig. S1 D). Taken together, these results demonstrate that the major ligand component in H. pylori is cholesteryl 6-O-tetradecanoyl-α-glucopyranoside (αCAG, C14:0; Fig. 1 N). The structure of αCAG is distinct from previously reported Mincle ligands bearing flexible lipid tails (Lu et al., 2018), as it contains a rigid and planar tetracyclic scaffold.
H. pylori extracts cholesterol from the host and adds glucose at the 3′ position using its glucosyltransferase, Hp0421, which is a critical first step for the synthesis of αCAG (Hirai et al., 1995; Lebrun et al., 2006). We extracted lipid components from a mutant H. pylori strain lacking hp0421 gene (H. pyloriΔhp0421). The HPTLC band corresponding to αCAG was absent from H. pyloriΔhp0421, and the equivalent fraction failed to activate Mincle reporter cells (Fig. 1 O) and BMDCs (Fig. 1 P). Thus, αCAG is a dominant H. pylori lipid that signals through Mincle.
Immunostimulatory properties of αCAG
To verify that the activity of fraction 13-14 did not result from minor contaminants, we chemically synthesized αCAG. A synthetic sample of the major αCAG identified in H. pylori, C14:0, potently activated reporter cells expressing Mincle at a level comparable to the authentic ligand trehalose 6,6′-dimycolate (TDM; Fig. 2 A). Cholesteryl α-glucoside (αCG), a biosynthetic precursor of αCAG, did not possess such activity. Direct binding to Mincle was confirmed by probing plate-bound synthetic αCAG with Mincle-Ig fusion protein (Fig. 2 B). In dendritic cells (DCs), synthetic αCAG induced the secretion of inflammatory cytokines (Fig. 2 C), up-regulation of costimulatory molecules (Fig. 2 D), and induction of nitric oxide synthase type 2 (NOS2; Fig. 2 E) through Mincle. Moreover, a comparison with TDM revealed that αCAG is the most potent ligand for human-derived Mincle (Fig. 2 F). Indeed, αCAG efficiently activated human monocyte–derived DCs (hMoDCs) to produce proinflammatory cytokine IL-8 (Fig. 2 G).
We thus analyzed the effect of αCAG on DC-mediated T cell priming. T cells from OVA-specific TCR transgenic mice were cultured with OVA-pulsed BMDCs in the presence or absence of synthetic αCAG. Antigen-specific production of IFN-γ and IL-17 was augmented in the presence of αCAG (Fig. 2 H). To evaluate T cell priming in vivo, mice were immunized with whole OVA protein along with αCAG, and recall response was examined. We detected IFN-γ production after restimulation of T cells from WT mice, whereas levels were lower in Mincle- and FcRγ-deficient mice (Fig. 2 I). Thus, the αCAG–Mincle–FcRγ axis promotes antigen-specific T cell priming through APC activation.
Pathophysiological contribution of Mincle to H. pylori–induced gastritis
To assess the pathophysiological consequences of an αCAG–Mincle interaction, we used a model of chronic H. pylori infection (Nagai et al., 2007). WT and Clec4e−/− mice were infected with H. pylori SS1 strain, and 6 wk after infection, H. pylori–specific recall responses were detected in T cells isolated from gastric LNs and Peyer’s patches. The production of IFN-γ and IL-17 from these T cells was much weaker when Clec4e−/− mice were infected (Fig. 3, A and B). Nevertheless, bacterial numbers in the stomach were comparable between Clec4e−/− and WT mice (Fig. 3 C), suggesting that H. pylori–specific Th1/17 cell responses do not efficiently contribute to the eradication of bacteria. Chronic gastritis was observed in WT mice after infection of H. pylori; however, the severity of gastritis as assessed by histological analysis was ameliorated in Clec4e−/− mice (Fig. 3 D). Increased numbers of neutrophils and macrophages in stomach homogenates from infected mice was suppressed in Clec4e−/− mice (Fig. 3 E and Fig. S2, A and B). Transcriptome analysis of gastric tissue revealed that the expression of inflammatory gene sets was significantly lower in Clec4e−/− mice (Fig. 3 F), further confirming that chronic gastritis was attenuated in the absence of Mincle. This effect is unlikely due to the difference of gastric microbiota in WT and Clec4e−/− mice, as assessed by metagenome analysis (Fig. S2 C). Collectively, these results indicate that Mincle contributes to H. pylori–induced gastritis. In line with these results, antibody (Ab) blockade of Mincle resulted in the suppression of T cell responses (Fig. 3, G–I; and Fig. S2 D) and chronic gastritis (Fig. 3 J and Fig. S2 E) without increasing the number of H. pylori (Fig. 3 K).
Nontargeting lipidomics reveals inflammatory conversion of H. pylori metabolites
Helical H. pylori bacilli in the stomach transform into dormant coccoid forms under anaerobic conditions, such as in the small intestine and Peyer’s patches (Noach et al., 1994). As previous studies revealed that the coccoid form has more potent immunostimulatory activity, we reproduced this transformation in vitro under anaerobic culture (Fig. 4 A; Nagai et al., 2007). Extracts from helical and coccoid forms were subjected to lipidomic analysis using liquid chromatography coupled with quadrupole/TOF MS (Tsugawa et al., 2020). Nontargeted lipidomics revealed that the lipid composition was markedly altered by the helical/coccoid conversion, particularly for cholesterol-containing lipids (Fig. 4 B), although cholesterol ester species were mostly unchanged (Fig. S3, A and B). TLC analysis also confirmed the alteration of lipid composition, with the most prominent changes being the band shift of αCAG (Fig. 4 C, black arrow) and the appearance of newly generated lipids in coccoid form (Fig. 4 C, gray arrow; designated as Spot-specific for coccoid form [Spot C]). Analysis of the molecular composition of αCAG revealed that longer fatty acids that were trace components in helical form became abundant in the coccoid form (Fig. 4 Dand Fig. S3 C); correspondingly short-chain myristate (C14:0) αCAG decreased in coccoid form. We therefore synthesized αCAG with different fatty acids, myristate (C14:0), palmitate (C16:0), and stearate (C18:0) and found that the activity of αCAG increased as their fatty acyl chains were elongated, as assessed by production of inflammatory cytokines (Fig. 4, E–G) and using reporter cells expressing Mincle (Fig. 4 H).
As described above, the amount of a polar glycolipid, Spot C, that was visualized by orcinol staining was markedly increased in coccoid form (Fig. 4 C, Spot C). When peritoneal macrophages were stimulated with Spot C, IL-6 production was detected in a CARD9-dependent manner (Fig. 4 I). Since Spot C was not recognized by Mincle, we tested several receptors and identified DC immunoactivating receptor (DCAR; Clec4b1; Fig. 4 J), an FcRγ-coupled CLR, as the candidate receptor. DCAR is known to recognize acylated phosphatidylinositol mannosides (AcPIMs) in mycobacteria (Toyonaga et al., 2016). However, as H. pylori does not possess AcPIM species (Tannaes et al., 2000), DCAR must recognize a previously unappreciated ligand in H. pylori. Using ESI-quadrupole Orbitrap MS (ESI-Q-Orbitrap-MS; Fig. 4, K and L; and Fig. S4, A and B), methanolysis followed by GC-MS (Fig. S4, C and D), and NMR spectroscopic analysis (Fig. 4, M and N; and Fig. S4 E), Spot C was identified as cholesteryl phosphatidyl α-glucoside (αCPG; Fig. 4 N). Lipidomics data targeted on αCPG revealed that the relative amount of αCPG was markedly increased in the coccoid form as observed on TLC (Fig. 4 C), whereas in contrast to αCAG, its fatty acid composition was unchanged (Fig. 4 O and FFig. S3 D). αCPG is distinctive in structure from known DCAR ligands AcPIMs (Toyonaga et al., 2016), except that both share a phosphate-containing phosphatidyl group. We thus synthesized αCPG and an αCPG analogue lacking the phosphate moiety, cholesteryl amide-linked α-glucoside (Fig. S4 F). Synthetic αCPG, but not the amide analogue, signaled through DCAR (Fig. 4 P), suggesting that phosphate is a key structural feature for DCAR binding which is absent in αCAG.
Depletion of αCAG/αCPG in H. pylori impairs its virulence
To examine the contribution of αCAG/αCPG to host responses against H. pylori, we investigated the immunostimulatory activity of mutant H. pylori lacking cholesterol glucosyltransferase (H. pyloriΔhp0421), which cannot generate αCAG and αCPG (Lebrun et al., 2006). We confirmed complete loss of these cholesteryl glucosides in both helical and coccoid forms of H. pyloriΔhp0421 (Fig. 5 A). The up-regulation of CD80 on the surface of BMDCs in response to H. pylori was impaired in this mutant strain (Fig. 5 B). Thus, H. pyloriWT and H. pyloriΔhp0421 were further evaluated for their T cell priming potential via co-culture with model antigen-pulsed DCs and OT-II T cells. The antigen-dependent secretion of IFN-γ and IL-17 from OT-II T cells was significantly lower when cultured with H. pyloriΔhp0421 compared with H. pyloriWT (Fig. 5, C and D, left panels). Reduction in cytokine secretion was similar to the suppression detected when we used FcRγ-deficient BMDCs in which Mincle and DCAR are nonfunctional (Fig. 5, C and D, right panels). However, cytokine production was restored following the addition of synthetic αCAG to H. pyloriΔhp0421 (Fig. 5, E and F), suggesting that αCAG is one of the major components of H. pylori responsible for T cell priming.
Finally, we addressed the role of cholesteryl glucosides in H. pylori–induced gastritis. Gastritis was significantly ameliorated in mice infected with H. pyloriΔhp0421 compared with H. pyloriWT, as assessed by inflammatory cell infiltration in the mucosa (Fig. 5, G and H). In contrast, bacterial numbers in the stomach were not significantly altered (Fig. 5 I). Similarly, we detected a substantial titer of anti–H. pylori Ab upon infection with H. pyloriΔhp0421 in WT mice, as well as H. pyloriWT in Clec4e−/− mice (Fig. 5, J and K). These results suggest that the production of cholesteryl glucosides, αCAG and αCPG, by H. pylori is required for its virulence to promote gastritis without affecting humoral immune responses.
In the present study, we provide the first example of an innate immune recognition of self-lipid–derived virulence factor generated by bacterial pathogens. These immune responses did not, however, efficiently contribute to H. pylori clearance (Adamsson et al., 2017).
In addition to the established role of Th1 (Eaton et al., 2001), recent studies have underscored the contribution of the Th17 population for the induction and development of gastritis during H. pylori infection (Ericksen et al., 2014; Gray et al., 2013). In the present study, H. pylori lipids augmented both Th1 and Th17 responses, which is consistent with the reported characteristics of CLR signaling (Geijtenbeek and Gringhuis, 2016). Another T cell subset, invariant natural killer T (iNKT) cells, are reported to be activated by cholesteryl glucosides (Chang et al., 2011; Ito et al., 2013; Shimamura, 2012), although we did not observe detectable iNKT cell activation by αCAG and αCPG (Fig. S2 F). Alternatively, αCAG and αCPG might activate iNKT cells through macrophage/DC-derived IL-12 in a TCR-independent manner (Cohen et al., 2011). However, there was no significant difference in the phenotypes of WT mice and Jα18-deficient (Traj18−/−) mice lacking iNKT cells after H. pylori infection (Fig. S2, G and H), suggesting that iNKT cells play a limited role in immune responses against H. pylori in the present SS1 model.
Some bacteria and fungi produce nonsteryl aliphatic ligands bearing flexible tails that signal through Mincle (Lu et al., 2018). H. pylori–derived Mincle ligand is atypical in its structure (Fig. 1 N) and recognition mode (Fig. S1 E), and the induction of gastritis is a pathology specific to H. pylori. Possibly, the continuous priming of APCs by rigid steroid-based ligands during chronic infection in gastric lymphoid tissues may efficiently induce pathogenic T cells. Most likely, these events take place in a less acidic environment, as the immune-active coccoid form of H. pylori is efficiently taken up by DCs in the duodenum or small intestine, particularly in secondary lymphoid organs (Nagai et al., 2007). Indeed, these DCs are activated in H. pylori–infected individuals and induce Th1 response (Bimczok et al., 2010).
αCAG is the most potent Mincle ligand in human. Although humans lack a DCAR orthologue (Flornes et al., 2004), αCPG potently activated human DCs (Fig. S4 G), indicating that αCPG may also exert their effect in humans via an unidentified receptor. Our recent report of murine DCAR–ligand complex structure (Omahdi et al., 2020) will help with the structure-based identification of human counterpart. Collectively, these results suggest that blockade of Hp0421 may prevent gastritis in humans by reducing the level of αCAG and αCPG. Although an H. pyloriΔhp0421 mutant did not show an apparent growth disadvantage in our experimental setting, inactivation of this enzyme in other clinical strains of H. pylori led to impaired growth (Kawakubo et al., 2004; McGee et al., 2011), which may further provide therapeutic benefit.
αCAG and αCPG are unique to Helicobacter spp. (Grille et al., 2010; Haque et al., 1996; Hirai et al., 1995; Mayberry and Smith, 1983). Although the physiological advantage of cholesterol glycosylation in H. pylori remains to be fully understood, several beneficial roles of αCAG for the bacteria have been reported (Grille et al., 2010; McGee et al., 2011; Morey et al., 2018; Wunder et al., 2006). The function of αCPG has not been demonstrated in both bacteria and host. It is also proposed that sterol glycosylation and subsequent acylation by H. pylori may detoxify environmental sterols that may be harmful to bacteria (Shimomura et al., 2013).
On the other hand, the conservation of apparently disadvantageous recognition of these metabolites by the host implies that these interactions may confer other advantages to the host, although we could not observe it in the context of H. pylori infection. Given that Mincle plays a protective role against mycobacterial infection (Kabuye et al., 2019; Lu et al., 2018), it may have conferred a survival advantage to the host. Such a trade-off between “chronic inflammation” and “protective immunity” may provide selection pressures that could potentially alter the CLR family lineup during evolution. Indeed, some CLR members have been lost/pseudogenized or arisen by gene duplication in mammalian species (Flornes et al., 2004).
Nontargeted lipidomics allowed for the identification of uncharacterized cholesteryl lipid species in H. pylori. In addition to the above-mentioned “diacyl” αCPG, we detected lyso-type CPG C19c:0 (lyso-αCPG) in the coccoid form, yet its biological function remains unclear. We also identified cholesteryl glycero-phosphate glycoside and cholesteryl ethanolamine-phosphate glycoside (Fig. S5), which have not been previously reported. These are potential candidates for as-yet-uncharacterized immunomodulatory metabolites.
Clinical studies have demonstrated that a low vitamin D concentration is correlated with severe gastritis and that vitamin D administration attenuated H. pylori infection and decreased gastritis; however, the molecular basis remains unclear (Antico et al., 2012; Hosoda et al., 2015; Kawamura et al., 2006). Vitamin D3 is a 3β-OH steroid that is efficiently assimilated by H. pylori (Shimomura et al., 2013), implying that one of the molecular mechanisms of the effect of vitamin D is through its action as a competitive inhibitor of Hp0421. Development of vitamin D derivatives that inhibit this enzyme might provide a harmless regimen for the prevention of H. pylori–induced gastritis and subsequent malignancy.
Antibiotic eradication of H. pylori is an established therapy to prevent H. pylori–induced gastritis. Recently, however, such methods are limited due to the emergence of drug resistance and microbial dysbiosis (Labenz, 2001; Wu et al., 2012). Furthermore, incomplete treatment with antibiotics increases the risk of accelerating gastritis through induction of the coccoid form (Poursina et al., 2013). Therefore, targeting Hp0421 or conversely the innate immune receptors for its products may provide a complementary approach to current antibiotic treatments. Helicobacter-specific T helper cells are also implicated in the pathophysiology of hematopoietic diseases such as mucosa-associated lymphoid tissue lymphoma and idiopathic thrombocytopenia purpura (D’Elios et al., 2005; Frydman et al., 2015), which can be addressed in mice using clinical isolates of Helicobacter spp. (Matsui et al., 2014). Thus, the identification of host lipid-derived bacterial adjuvants and their host receptors that drive T cell activation is potentially of broad clinical significance.
Materials and methods
Mincle-deficient mice (Yamasaki et al., 2009) were backcrossed for at least 16 generations with C57BL/6. FcRγ-deficient mice (Park et al., 1998) were provided by T. Saito (RIKEN, Yokohama, Japan). CARD9-deficient mice (Hara et al., 2007) were provided by H. Hara (Kagoshima University, Kagoshima, Japan). MyD88-deficient mice (Adachi et al., 1998) were purchased from Oriental Yeast. Jα18-deficient mice (Cui et al., 1997) were provided by M. Taniguchi (RIKEN, Yokohama, Japan). OVA-specific TCR OT-II transgenic mice (Barnden et al., 1998) were used on a C57BL/6 background. All mice were maintained in a filtered-air laminar-flow enclosure and given standard laboratory food and water ad libitum. All animal protocols were approved by the committee of Ethics on Animal Experiment, Faculty of Medical Sciences, Kyushu University and Research Institute for Microbial Diseases, Osaka University.
TDM and αCG were purchased from Sigma-Aldrich. Synthetic αCAG, αCG, αCPG, and αCPG analogue were synthesized as described below. Anti-CD11b (M1/70), Ly6G (1A8), B220 (RA3-6B2), and CD40 (3/23) mAbs were from BD Biosciences. Anti-CD80 (16-10A1), TCR-β (H57), and CD69 (H1.2F3) mAbs were from BioLegend. Anti-F4/80 (BM8) and NOS2 (CXNFT) mAbs were from eBioscience. Anti-Mincle (1B6) mAb was from MBL. For histological analysis, anti-CD3 (A0452) Ab and F4/80 (Cl:A3-1) mAb were from Dako and AbD Serotec, respectively. The ELISA kits for TNF, IL-6, and IFN-γ were from BD Biosciences. The ELISA kits for MIP-2, IL-17, and human IL-8 were from R&D Systems and Invitrogen, respectively.
2B4-NFAT-GFP reporter cells expressing murine CLRs and human Mincle were prepared as previously described (Kiyotake et al., 2015; Miyake et al., 2013; Toyonaga et al., 2016; Yamasaki et al., 2008; Yonekawa et al., 2014). HEK293-derived reporter cells that stably express an NF-κB–inducible secreted embryonic alkaline phosphatase (SEAP) reporter gene and TLR2 or TLR4 gene were obtained from InvivoGen. BMDCs and hMoDCs were prepared as previously described (Kiyotake et al., 2015; Nagata et al., 2017). DN32.D3 (Lantz and Bendelac, 1994) cells were provided by Y. Kinjo (Jikei University School of Medicine, Tokyo, Japan).
The H. pylori strain SS1, a mouse-adapted human isolate, was used for all the experiments. The Hp0421-deficient SS1 strain (H. pyloriΔhp0421) was described previously (Chang et al., 2011). Glycerol stocks of H. pylori SS1 were first grown on 5% sheep blood agar plates (BBL: 251239) under microaerobic conditions for 2 d. After the plate culture, H. pylori SS1 was grown in Brucella broth (BD Biosciences) with 5% FCS for 14–16 h at 37°C under microaerobic conditions with gentle agitation. For the preparation of the coccoid form, SS1 was incubated in Brucella broth with 5% FCS for 2–3 d under anaerobic conditions after the microaerobic liquid culture or on 5% sheep blood agar plates for 7 d under anaerobic conditions after the first growth on the agar plates for 3 d.
Lipid extraction and purification
H. pylori was washed with PBS and then treated with chloroform/methanol (2:1, vol/vol) for 1 d. This mixture was partitioned by centrifugation and the lower organic phase was used as the lipid extract. To isolate αCAG or αCPG, the lipid extracts were separated by HPTLC (Merck) and visualized by copper(II) acetate-phosphoric acid (180°C, 15 min) or orcinol (120°C, 5–15 min) staining. The fractions containing these lipids were scraped from the plate, and chloroform/methanol (2:1, vol/vol) was used to elute the lipids. The purified lipids were filtered using Millex-LG (0.2 μm; Merck) to remove silica gel contamination.
ESI-TOF-MS and GC-MS analysis of αCAG and αCPG
ESI-TOF-MS was measured with a Bruker micro-TOF mass spectrometer in the positive ESI mode (Bruker Daltonics). For conversion to FAMEs for GC-MS analysis, αCAG from the helical form of H. pylori was subjected to methanolysis by heating with 10% HCl/methanol in a sealed tube at 80°C for 3 h. The reaction mixture was diluted with methanol and extracted with n-hexane. The n-hexane extract was concentrated in vacuo to give a mixture of FAMEs. The FAMEs were dissolved in acetone and subjected to GC-MS with GC-MS-QP2010 SE (Shimadzu). The conditions of GC-MS were as follows: column, INERTCAP 5MS/SIL (0.25 mm i.d. × 30 m; GL Science); column temperature, 100–280°C; rate of temperature increase, 18°C/min. Three major FAMEs were detected and the retention time (tR), and the proportions (%) of each FAME of αCAG were as follows: methyl tetradecanoate (14:0), tR [min] = 12.3, 35.4%; methyl palmitate (16:0), tR [min] = 14.5, 6.7%; methyl stearate (18:0), tR [min] = 16.6, 5.8%. Dehydro cholesterol was also detected: Δ3,5-cholesterol, 41.4%, tR [min] = 24.4, m/z = 368 (M+). FAMEs of αCPG: methyl tetradecanoate (14:0), tR [min] = 12.3, 20.4%; methyl palmitate (16:0), tR [min] = 14.5, 3.0%; methyl octadecenoate (18:1), tR [min] = 16.4, 3.7%; methyl stearate (18:0), tR [min] = 16.6, 11.1%; methyl nonadecanoate, tR [min] = 18.1, 14.3%; methyl 11,12-methyleneoctadecanoate (19c:0), tR [min] = 19.0, 4.9%; methyl nonadecanoate (19:1), tR [min] = 19.4, 1.9%; dehydrocholesterol was also detected: Δ3,5-cholesterol, tR [min] = 24.4, 29.2%.
ESI-Q-Orbitrap-MS analysis of αCPG
Lipid fractions were dissolved in 5 mM ammonium acetate (wt/vol) in methanol/water (95:5, vol/vol; 50 μg/ml). Flow injection analysis was performed using a Dionex Ultimate 3000 RSLC system (Thermo Fisher Scientific) coupled with a Q Exactive, a high-performance benchtop Q-Orbitrap-MS, fitted with an ESI ion source (Thermo Fisher Scientific). The flow injection conditions in the positive and negative ionization modes were as described previously (Nagata et al., 2017).
NMR analysis of αCAG and αCPG
1H- and 13C-NMR spectra were recorded on an Agilent INOVA 600 spectrometer. The conditions for αCAG were as follows: 1H: 600 MHz, CDCl3/CD3OD (1:1, vol/vol), 298 K and 13C: 150 MHz, CDCl3/CD3OD (1:1, vol/vol), 298 K. The conditions for αCPG were as follows: 1H: 600 MHz, CDCl3/CD3OD/D2O (65:35:5, vol/vol/vol), 298 K and 13C: 150 MHz CDCl3/CD3OD/D2O (65:35:5, vol/vol/vol), 298 K. 1H chemical shifts were assigned by 1H-1H correlation spectroscopy and total correlation spectroscopy experiments and 13C chemical shifts were assigned by HSQC and HMBC experiments.
CG and CAGs were synthesized in a multistep route as follows. Glycosylation of cholesterol using 3,4,6-tri-O-acetyl-2-O-benzyl-α-d-glucopyranosyl N-phenyltrifluoroacetimidate afforded cholesteryl 3,4,6-tri-O-acetyl-2-O-benzyl-α-d-glucopyranoside. Transfer hydrogenolysis of the benzyl groups (cyclohexene, Pd(OH)2/C, ethanol) and then deacetylation (NaOMe, methanol/THF) afforded αCG. Trimethylsilylation (TMSCl, Et3N, CH2Cl2) of αCG yielded cholesteryl 2,3,4,6-tetra-O-trimethylsilyl-α-d-glucopyranoside. Removal of the primary trimethylsilyl group (NH4OAc, CH2Cl2/methanol) and acylation with myristoyl, palmitoyl, or stearoyl chloride (DMAP, pyr/CH2Cl2) afforded the cholesteryl 6-O-acyl-2,3,4-tri-O-trimethylsilyl-α-d-glucopyranosides. Finally, treatment with Dowex 50 (H+ form) resin in methanol/CH2Cl2 provided αCAG-C14:0, αCAG-C16:0, and αCAG-C18:0. Natural αCPG was chemically synthesized as below. First, a common αCG derivative having a single hydroxyl group at the C6 position of glucose residue was prepared based on an in situ anomerization method to form an α-glucoside linkage. The conversion of penta-O-trimethylsilyl-d-glucose into the corresponding glucosyl iodide followed by the coupling with cholesterol afforded the desired cholesteryl glucoside as a mixture of diastereoisomers. Subsequent four-step reaction sequence, including the removal of TMS groups, C6-protection by the trityl group, the introduction of the Fmoc group to other hydroxyl groups, and the removal of the trityl group, gave the αCG derivative in pure form. Glycerol moiety found in natural αCPG was prepared starting from commercially available l-(−)-1,2-isopropylideneglycerol. The coupling of the glycerol derivative and myristic acid in the presence of N,N′-dicyclohexylcarbodiimide and 4-dimethylaminopyridine followed by the removal of the isopropylidene group afforded the diol product. The primary alcohol was then protected as a trityl ether. The introduction of phytomonic acid to the remaining secondary alcohol and subsequent removal of the trityl group by hydrogenolysis provided the desired glycerol unit used for natural αCPG in good yields. The prepared glycerol unit was treated with 2-cyanoethyl N,N,N′,N′-tetraisopropylphosphoroamidite and 1H-tetrazole to form the corresponding phosphoroamidite derivative, which was then subjected to the phosphoroamidite coupling reaction with the above-mentioned αCG derivative. The coupling successfully yielded the corresponding fully protected target framework. Finally, global deprotection delivered natural αCPG. On the other hand, the amide-linked αCPG analogue was chemically prepared via 15 reaction steps starting from the commercially available d-glucose pentaacetate. First, d-glucose pentaacetate was converted into phenyl 4,6-O-di-tert-butylsilylene-2,3-di-O-p-methoxybenzyl-1-thio-β-d-glucopyranoside as the glycosyl donor in six steps. The coupling of the glucosyl donor and cholesterol in the presence of dimethyl(methylthio)sulfonium triflate and 2,4,6-tri-tert-butylpyrimidine in dichloromethane at room temperature followed by the removal of the DTBS group of the glucose moiety afforded the 4,6-dihydroxy α-glucosyl cholesterol derivative. Subsequent Mitsunobu reaction with phthalimide, 1,2-bis(diphenylphosphino)ethane, and bis(2-methoxyethyl) azodicarboxylate installed an imide functionality at the C6 position of the glucose. The 6-phthalimide derivative obtained was then transformed into the 6-NH2 glucosyl cholesterol having tert-butyldimethylsilyl groups at O2, O3, and O4 positions of the glucose. The prepared 6-NH2 glucosyl cholesterol derivative was condensed with the same glycerol unit as that used for natural αCPG in the presence of N,N′-disuccinimidyl carbonate and triethylamine, giving the fully protected amide-linked αCPG framework in good yield. Finally, the removal of tert-butyldimethylsilyl groups by the action of tetra-n-butylammonium fluoride furnished the amide-linked αCPG analogue.
In vitro stimulation assay
To stimulate the cells, each lipid was diluted in isopropanol and the 20 μl dilutions were added to each well of the 96-well plate followed by evaporation of the solvent. LPS (10 ng/ml) was used as a positive control. The reporter activity of 2B4-NFAT-GFP cells was analyzed by flow cytometry and SEAP secretion from HEK293-based NF-κB cells were detected using an alkaline phosphatase detection reagent (QUANTI-Blue; InvivoGen).
In vitro Mincle binding assay
The Mincle-Ig protein, composed of the C-terminus of the extracellular domain of Mincle fused to the N-terminus of the human IgG1 Fc region (hIgG1), was prepared as described previously (Miyake et al., 2013). For the in vitro binding assay, 3 μg/ml hIgG1-Fc (Ig) and Mincle-Ig diluted in binding buffer (20 mM Tris-HCl, 150 mM NaCl, 1 mM CaCl2, and 2 mM MgCl2, pH 7.0) were incubated with plate-coated glycolipids. The bound proteins were detected with anti-hIgG-HRP followed by the addition of a colorimetric substrate. Peroxidase activity was measured using the UV/Vis spectrophotometer.
Co-culture of OT-II CD4+ T cells and BMDCs
BMDCs were stimulated with H. pylori, plate-coated αCAG, or both during this assay. OT-II CD4+ T cells were sorted by CD4 microbeads (Miltenyi Biotec) from single-cell suspensions of spleen and inguinal LNs from OT-II mice. Sorted OT-II CD4+ T cells were co-cultured with BMDCs at the ratio of 10:1 for 3 d in the presence of OVA323–339 peptide.
Mice were sensitized by a subcutaneous injection with 200 μg OVA in oil-in-water emulsion (mineral oil/Tween-80/PBS [9:1:90, vol/vol/vol]) containing 100 μg αCAG. At day 7, mice were challenged by an injection of 200 μg heat-aggregated OVA (70°C, 1 h) in 20 μl PBS into both footpads. For the in vitro restimulation analysis, at 7 d after the challenge, B cell–depleted inguinal LN cells were stimulated with the OVA protein for the indicated periods.
H. pylori SS1 were prepared from plate and liquid cultures. Mice were orally administered 3 × 108 CFUs of H. pylori SS1 in 400 μl of 5% FCS/Brucella broth three times in a week. For the anti-Mincle (1B6) mAb administration, mice were intravenously injected with 100 μg anti-Mincle mAb or control rat IgG twice a week for 4 wk. After the indicated time period, mice were sacrificed, and gastric LNs were collected before harvesting other tissues. Half of the stomachs were fixed with 10% formalin for 24 h and embedded in paraffin. Specimens were stained with H&E or taken for immunohistochemical analysis with anti-CD3 and F4/80 Abs. The remainder was used for the preparation of gastric mononuclear cells (MNCs) and RNA and the calculation of bacterial CFUs. The CFU was determined by plating serial dilutions of the stomach homogenates onto H. pylori selective agar plates (NISSUI) and counting colonies.
Preparation of gastric MNCs
Mouse stomachs were minced into small pieces and treated with 3 mM EDTA. After treatment, small pieces of stomach were digested with 250 μg/ml collagenase II (Sigma-Aldrich) and 100 μg/ml DNaseI (Roche) for 30 min at 37°C. Gastric MNCs were purified by Percoll density gradient centrifugation.
Recall responses after infection
Gastric LNs, Peyer’s patches, mesenteric LNs, and spleens were collected from H. pylori–infected mice. These tissues were homogenized, and then single-cell suspensions were stimulated with H. pylori lysate in the presence or absence of WT BMDCs (tissue cells/BMDCs = 10:1). CD4+ T cells were sorted by CD4 microbeads from splenocytes.
Preparation of H. pylori lysate
H. pylori SS1 was grown on 5% sheep blood agar plates under microaerobic conditions for 3 d and was sequentially incubated on 5% sheep blood agar plates under anaerobic condition for 7 d. After cultures, H. pylori SS1 were collected and sonicated in PBS. The supernatants after centrifugation were collected and used as the whole lysate of H. pylori. The protein concentration of the filtered whole lysate was determined using the Protein Assay Bicinchoninate Kit (Nacalai Tesque).
Total RNA was extracted from the stomachs with Sepasol-RNA I Super G (Nacalai Tesque) according to the manufacturer’s instructions. Library preparation was performed using a TruSeq stranded mRNA sample prep kit (Illumina) according to the manufacturer’s instructions. Whole-transcriptome sequencing was applied to the RNA samples by using an Illumina HiSeq 2500 platform in a 75-base single-end mode. The Illumina Casava ver.1.8.2 software was used for base calling. Sequenced reads were mapped to the mouse reference genome sequences (mm10) using TopHat ver. 2.0.13 in combination with Bowtie2 ver. 2.2.3 and SAMtools ver. 0.1.19. The number of fragments per kilobase of exon per million mapped fragments was calculated using Cufflinks ver. 2.2.1. The transcriptome RNA-sequencing datasets have been deposited to the National Center for Biotechnology Information Gene Expression Omnibus database under the accession number GSE136203.
16S rRNA gene sequencing of gastric microbiota
Gastric mucosal swabs were collected from uninfected WT mice, H. pylori–infected WT mice, and Clec4e−/− mice for 12 wk. Bacterial DNA was extracted from swab samples using a DNeasy PowerSoil Pro kit (Qiagen). Each library was prepared according to the Illumina 16S Metagenomic Sequencing Library Preparation Guide with primer set 27Fmod: 5ʹ-AGRGTTTGATCMTGGCTCAG-3ʹ and 338R: 5ʹ-TGCTGCCTCCCGTAGGAGT-3ʹ targeting the V1–V2 region of 16S rRNA genes; 301-bp paired end sequencing of the amplicons was performed on a MiSeq system (Illumina) using a MiSeq Reagent v3 600 cycle kit. The paired end sequences obtained were merged, filtered, and denoised using DADA2. Taxonomic assignment was performed using QIIME2 feature-classifier plugin with the Greengenes 13_8 database. The QIIME2 pipeline, version 2020.2, was used as the bioinformatics environment for the processing of all relevant raw sequencing data. Metagenomic datasets are available at the DNA Data Bank of Japan Sequence Read Archive (DRA010478).
Whole lipids of cultured H. pylori SS1 were prepared by single-phase extraction as described by Tsugawa et al. (2020). Briefly, 1.0 × 109 CFUs of dried bacterial cells were suspended in 100 μl chloroform. After 1 h incubation, 195 μl methanol and 5 μl EquiSPLASH (Avanti Polar Lipids) were mixed and incubated for another 2 h at room temperature. Thereafter, 20 μl of water was added, and the samples were incubated for 10 min. After extraction, samples were centrifuged at 2,000 ×g for 10 min, and the supernatants were collected. The extracted lipids were measured using the ACQUITY UPLC I class system (Waters) coupled with a TripleTOF 6600 (AB Sciex; Tsugawa et al., 2020). Liquid chromatography separation was performed using a reverse-phase column (ACQUITY UPLC BEH C18 column [2.1 mm i.d. × 50 mm, particle size 1.7 μm; Waters]), and data acquisition was performed by data-dependent MS/MS in the negative and positive ion modes. The detected lipid species were annotated using MS-DIAL (Tsugawa et al., 2015) and Peak View (AB Sciex).
For the assessment of gastric histopathology, two H&E-stained sections per mouse were analyzed for lymphocytic inflammation using a previously described method (Eaton et al., 2007). Briefly, lymphocytic inflammation was defined as inflammatory cell infiltration that displaced the gastric glands. Positive field numbers were divided by the total number of fields and multiplied by 100% to calculate the percentage of the affected fields.
An unpaired two-tailed Student’s t test was used for all the statistical analyses. Asterisks denote level of statistical significance (*, P < 0.05; **, P < 0.01).
The transcriptome RNA-sequencing and metagenomic datasets generated during this study are available at the National Center for Biotechnology Information Gene Expression Omnibus database (GSE136203) and the DNA Data Bank of Japan Sequence Read Archive (DRA010478), respectively.
Online supplemental material
Fig. S1 shows structural analysis of immunostimulatory components in H. pylori. Fig. S2 shows cellular and humoral immune responses during H. pylori infection. Fig. S3 shows nontargeted lipidomics of helical and coccoid form of H. pylori. Fig. S4 shows structural analysis of Spot C. Fig. S5 shows MS spectra of uncharacterized cholesteryl lipid species in H. pylori.
We thank S. Iwai, S. Torigoe, Y. Hosono, and J. Maaskant for technical support; H. Mimuro, E. Kuroda, T. Watanabe, and M. Ito for discussion; M. Tanaka, Y. Baba, K. Kaseda, and M. Ikawa for embryonic engineering; D. Motooka and D. Okuzaki for bioinformatics analysis; and the Cooperative Research Project Program of the Medical Institute of Bioregulation, Kyushu University, and Genome Information Research Center, Research Institute for Microbial Diseases, Osaka University, for technical support.
This research was supported by the Japan Agency for Medical Research and Development (JP19gm0910010, JP19ak0101070, and JP19fk0108075), Japan Society for the Promotion of Science KAKENHI (JP17H04087 and JP15H05897), the Australian Research Council (DP160100597), and the Takeda Science Foundation.
Author contributions: M. Nagata, E. Ishikawa, B.J. Appelmelk, and S. Yamasaki conceptualized research; M. Nagata, K. Toyonaga, E. Ishikawa, S. Haji, N. Okahashi, M. Takahashi, and T. Miyamoto performed investigations; A. Imamura, K. Takato, H. Ishida, S. Nagai, P. Illarionov, B.L. Stocker, M.S.M. Timmer, D.G.M. Smith, S.J. Williams, and B.J. Appelmelk provided resources; N. Okahashi, M. Takahashi, Y. Izumi, T. Bamba, T. Miyamoto, and M. Arita performed data curation; S. Yamasaki supervised the research; M. Nagata, K. Toyonaga, E. Ishikawa, and S. Yamasaki wrote the manuscript.
Disclosures: The authors declare no competing interests exist.