Helicobacter pylori (Hp) strains that carry the cag type IV secretion system (cag-T4SS) to inject the cytotoxin-associated antigen A (CagA) into host cells are associated with peptic ulcer disease and gastric adenocarcinoma. CagA translocation by Hp is mediated by β1 integrin interaction of the cag-T4SS. However, other cellular receptors or bacterial outer membrane adhesins essential for this process are unknown. Here, we identify the HopQ protein as a genuine Hp adhesin, exploiting defined members of the carcinoembryonic antigen-related cell adhesion molecule family (CEACAMs) as host cell receptors. HopQ binds the amino-terminal IgV-like domain of human CEACAM1, CEACAM3, CEACAM5 or CEACAM6 proteins, thereby enabling translocation of the major pathogenicity factor CagA into host cells. The HopQ–CEACAM interaction is characterized by a remarkably high affinity (KD from 23 to 268 nM), which is independent of CEACAM glycosylation, identifying CEACAMs as bona fide protein receptors for Hp. Our data suggest that the HopQ–CEACAM interaction contributes to gastric colonization or Hp-induced pathologies, although the precise role and functional consequences of this interaction in vivo remain to be determined.
The highly motile, microaerophilic, Gram-negative bacterium, Heliobacter pylori, infects the gastric mucus layer of about half the human population. This infection is a major cause of gastroduodenal disease, including chronic active gastritis, peptic ulcer disease, mucosa-associated lymphoid tissue (MALT) lymphoma and gastric adenocarcinoma1,2. The 37 kb cag-pathogenicity island (cag-PAI) encodes the cytotoxin-associated gene A (CagA), as well as the cag type IV secretion system (cag-T4SS), which forms a membrane-spanning secretion channel and an extracellular pilus. On contact with target cell receptors such as the α5β1 integrin heterodimer3,4, the effector protein CagA is translocated into host cells5 and secretion of the proinflammatory chemokine interleukin-8 (IL-8) is induced6,7. Injected CagA is tyrosine-phosphorylated on multiple Glu-Pro-Ile-Tyr-Ala (EPIYA) motifs in the C-terminal region of the CagA protein, which enables CagA to interact with a large repertoire of cellular target proteins8,
The carcinoembryonic antigen-related cell adhesion molecules (CEACAMs) are a diverse group of cell surface glycoproteins that are part of the immunoglobulin (Ig) superfamily. CEACAMs are involved in cell–cell recognition and modulate cellular processes ranging from the shaping of tissue architecture and neovascularization to the regulation of insulin homeostasis and T-cell proliferation20,21. A number of human-specific Gram-negative bacteria specifically adhere to CEACAMs22. Here, we report that Hp also exploits CEACAMs to bind to host cells and to inject its major virulence factor, the CagA oncoprotein, into mammalian cells.
Specific binding of Hp to IgV-like domains of human CEACAM1 and CEACAM5
CEACAMs are expressed on the apical surface of epithelial cells, and several human-specific Gram-negative pathogens are known to exploit CEACAMs to infect their host23. To assess the potential interactions between Hp, which also specifically infects humans, and CEACAMs, we incubated soluble CEACAM–green fluorescent protein (GFP) fusion proteins (designated as CEA1–CEA8-N-GFP, Fig. 1a,b) with Hp strain P12 (ref. 24). Centrifugation of intact bacteria pulled down CEA1-N-GFP and CEA5-N-GFP, as detected by immunoblotting with an anti-GFP antibody (Fig. 1b). Flow cytometry measurements verified the binding of CEA1-N-GFP and CEA5-N-GFP (Fig. 1c), but none of the other CEA-N-GFP constructs, as ligands for Hp (Supplementary Fig. 1a). Pulldown of CEA5-N-GFP by P12 was significantly stronger (4.3-fold) than of CEA1-N-GFP (Supplementary Fig. 1b). A similar binding pattern was observed for the CEACAM-binding reference strain Moraxella catarrhalis ATCC25238 (Supplementary Fig. 1c) and for additional Hp strains (Supplementary Fig. 1d). Thus, Hp binds specifically to soluble N-terminal domains of human CEACAM1 and CEACAM5.
From this set of independent Hp strains, three (P12, 26695 and the Mongolian gerbil-adapted strain B8) interacted with CEA1-N-GFP and even more strongly with CEA5-N-GFP, but not with any other CEA-N-GFP protein tested (Fig. 1d–f). In contrast, the mouse-adapted strains SS1 and X47 revealed a moderate to strong binding to CEACAM5, but only weak binding to CEACAM1 (Fig. 1g,h). Interestingly, none of the non-human CEA1-N-GFP (murine, bovine or canine) or CEA5-N-GFP (macaque) fusion proteins reacted with the above set of Hp strains (Supplementary Fig. 2a). These results indicate that the association of Hp with CEACAM1 and CEACAM5 and its intensity varies between strains, but is highly specific for the human receptors.
HopQ is the Hp adhesin binding CEACAM1 and CEACAM5 N-terminal domains
Hp produces a large number of OMPs, including the Helicobacter outer protein (Hop) and Hop-related (Hor) gene families, which are differently present or expressed in independent Hp strains. A phylogenetic tree generated from the Hop and Hor protein families revealed that most of the known Hp adhesins (BabA, BabB, BabC, SabA, HopD and HopZ) clustered together in an isolated branch (Supplementary Fig. 2b). Hp babA, babB, sabB or hopZ mutants did not differ significantly in soluble CEA1-N-GFP and CEA5-N-GFP binding from wild type (wt) P12 (Supplementary Fig. 2c). By generating additional mutant strains within this putative ‘adhesin family’ (Supplementary Table 1), we identified hopQ deletion mutants as defective in the binding of CEA1-N-GFP or CEA5-N-GFP (Fig. 2a). Genetic complementation using the plasmid-based pHel shuttle vector system25 completely restored the CEACAM-binding capacity of this mutant strain (Fig. 2a). Using a peptide-based antiserum (AK298) directed against the HopQ type I-specific protein region, we verified the production of HopQ in the corresponding strains (Fig. 2b). HopQ I proteins are typically associated with Hp strains carrying a functional cag-T4SS, whereas T4SS-negative Hp strains often produce a HopQ II protein26 that differs markedly from HopQ I and is not recognized by AK298 (Fig. 2b and Supplementary Fig. 3). In the genetically complemented P12ΔhopQ::hopQ I strain, HopQ I is overproduced relative to P12 (Fig. 2b), explaining the stronger relative binding to CEACAM1 and CEACAM5 (Fig. 2a). Together, these data show that Hp uses HopQ to bind the CEACAM1 and CEACAM5 N-terminal domains.
HopQ binding of non-glycosylated N-terminal CEACAM domains with high affinity
To determine whether HopQ and CEACAMs interact directly, we produced HopQ as a maltose binding protein (MBP) fusion (Supplementary Table 2). The N-terminal domains of CEACAM1, CEACAM3, CEACAM5 and CEACAM8 (CEA-N) were produced in Escherichia coli (Supplementary Fig. 4a) and their affinity for HopQ was measured by isothermal titration calorimetry (ITC). MBP–HopQ, but not MBP alone (Supplementary Fig. 4b), bound with high affinity to CEA1-N (KD = 23 ± 1 nM) and to CEA5-N (KD = 61 ± 3 nM), with lower affinity to CEA3-N (KD = 268 ± 4 nM), but did not show binding to CEA8-N (Fig. 2c–f) (Supplementary Table 3). The stoichiometry was determined as one MBP–HopQ to half of a CEA-N dimer (1:0.62), similar to other bacterial adhesin–CEACAM5 interactions27. Corroborating our bacterial pulldown assays, these data clearly demonstrate that Hp uses its OMP HopQ as an adhesin to interact directly and with high affinity, with the N-terminal domain of defined CEACAM proteins, even in the absence of CEA-N glycosylation. Thus, the HopQ–CEACAM interaction is the first protein–protein adhesin-receptor interaction described for Hp.
Surface localization of CEACAMs on epithelial cell lines correlates with CagA translocation by Hp
We have previously shown that a P12ΔhopQ mutant strain is severely impaired in its capacity to translocate CagA into gastric epithelial cells18. By flow cytometry analysis, we observed CEACAM1, CEACAM5 and CEACAM6 expression on the surface of AGS, Kato III and MKN45 cells, but not on MKN28, HeLa or HEK293 cells (Fig. 3a). None of this latter set of cells, producing little or no CEACAMs on their surface, was permissive for CagA translocation. In contrast, AGS, Kato III and MKN45 cells, all of which produce significant amounts of these CEACAMs, were competent for translocation of CagA. Notably, MKN45 cells, producing significantly more CEACAM5 and CEACAM6 than the other cell lines, exhibited the strongest capacity for CagA translocation by Hp (Fig. 3a). These data point to a correlation between CEACAM1, CEACAM5 and CEACAM6 cell surface presentation and the ability of Hp P12 to translocate CagA.
HopQ-specific adhesion and CagA translocation in HEK293 cells
To determine whether the absence of cell surface expression of CEACAMs is responsible for a defect in CagA translocation, stably CEACAM1- or CEACAM5-transfected HEK293 cells (Supplementary Table 4) were analysed for proper CEACAM production and surface localization by flow cytometry. Hp infection of these cells resulted in efficient CagA translocation (Fig. 3b,c), demonstrating that CagA translocation can be completely restored by adding a single receptor for the Hp OMP HopQ, either CEACAM1 or CEACAM5.
We next developed a flow cytometry-based quantitative binding assay using the GFP-labelled Hp P12 strain and stably transfected HEK293 cells producing the N-terminal (CEACAM5ΔIgC) or constant domains of CEACAM5 (CEACAM5ΔIgV) (Fig. 4a). HEK293 cells, producing CEACAM5ΔIgC, showed a significant HopQ-dependent adherence (Fig. 4b) and CagA translocation (Fig. 4c), whereas CEACAM5ΔIgV-producing cells did not. Thus, the presence of the N-terminal domain of CEACAM5 on host cells is necessary and sufficient to support both Hp adherence and CagA translocation into HEK293 cells in vitro.
The amino-acid sequences of the N-terminal domains of human CEACAMs are well conserved between CEACAM1, CEACAM3, CEACAM5 or CEACAM6, whereas CEACAM4 and CEACAM8 diverge substantially (Supplementary Fig. 4c). Transient transfection of HEK293 cells with expression plasmids encoding these additional CEACAMs resulted in CEACAM surface localization (Fig. 4e, bottom). Notably, for CEACAM3- and CEACAM6-, but not for CEACAM4- or CEACAM8-producing cells, we observed specific binding of Hp (Fig. 4d), as well as CagA translocation (Fig. 4e, top). The binding of Hp to cell surface-exposed CEACAM3 and CEACAM6 was in contrast to our results obtained with soluble CEA3-N-GFP and CEA6-N-GFP, which did not show binding (Fig. 1c), but is supported by the ITC data (Fig. 2c–f). Thus, membrane localization and potential clustering of these receptors may provide better accessibility of binding epitopes for HopQ as compared to the constructs.
A specific siRNA knockdown of CEACAM1, CEACAM5 and CEACAM6 in AGS cells was up to 100% for CEACAM1, 70% for CEACAM5 and 73% for CEACAM6 (Supplementary Fig. 5a). Knockdown of individual CEACAMs did not show a significant inhibition of CagA translocation, suggesting a functional redundancy. However, the simultaneous knockdown of all three CEACAMs reduced CagA translocation to a level comparable to that observed for the Hp P12ΔhopQ mutant on AGS cells (Supplementary Fig. 5a,b). These observations suggest that AGS cells express at least one additional unknown receptor recognized by an unknown Hp adhesin, which supports CagA translocation and seems to be absent in HEK293 cells (Supplementary Fig. 5a,b).
Correlation of Hp adherence, HopQ-mediated recruitment of CEACAM5 and CagA injection
To study the functional relevance of the HopQ–CEACAM interaction for CagA translocation, we performed confocal laser scanning microscopy (CLSM) using different human gastric epithelial cell lines, AGS, MKN45 and Kato III. Infection resulted in a strong adherence and co-localization of Hp P12 WT and the genetically complemented hopQ mutant (P12ΔhopQ::hopQ I) with CEACAM5 on the cell surface (Supplementary Fig. 6a–c, yellow arrows). In the absence of HopQ (P12ΔhopQ), co-localization of Hp with CEACAM5 was not observed on any of the tested cells (Supplementary Fig. 6, arrowheads). However, in the P12ΔhopQ strain, Hp binding to AGS cells was only slightly reduced (statistically not significant), but almost abolished for MKN45 cells (Supplementary Fig. 6a,b,d,e). In contrast, Kato III cell binding of the mutant strain did not show any significant differences from the WT strain, whereas the complemented strain showed enhanced binding (Supplementary Fig. 6c,f). Our data suggest that on AGS and Kato III cells additional receptors besides CEACAMs are present, whereas in MKN45 cells CEACAMs might represent the major receptor for Hp. Generally, HopQ seems to efficiently recruit and cluster CEACAM5 on the surface of the host cells and even cover Hp partially or completely. These data are corroborated by the CagA translocation results, showing a correlation between CEACAM adhesion capability of Hp and CagA translocation (Supplementary Fig. 5b–e), suggesting that adherence via outer membrane proteins is a prerequisite for CagA translocation by Hp.
Pathology and stabilization of the cag-T4SS in the CEACAM5-humanized mouse model
To further study the role of a putative HopQ-mediated adhesion in the stomach mucosa, we used the Mongolian gerbil and mouse models. The gerbil-adapted Hp strain P322 and an isogenic hopQ knockout strain revealed a similar colonization rate (Supplementary Fig. 7a), no significant differences in pathology scores28 (Supplementary Fig. 7b) and a HopQ-independent binding to gerbil gastric tissue sections (Supplementary Fig. 7c–e). This suggests that no compatible CEACAM receptor(s) for HopQ-mediated binding of Hp is/are present on gerbil antrum or corpus tissue, although we cannot rule out a denaturation of CEACAM IgV protein domains in fixed gastric sections.
Because mice generally do not have a CEACAM5 gene29 and murine CEACAM1 is not recognized by Hp (Supplementary Fig. 2a), a transgenic C57BL/6 mouse expressing the human CEACAM5 gene (CEACAM5-humanized mouse, here designated ‘CEA mouse’)30 was used to further study Hp CEACAM interactions. The production of human CEACAM5 in the mouse gastric mucosa was verified by immunostaining (Supplementary Fig. 8a,b).
Infection experiments by the mouse-colonizing Hp strain PMSS1 revealed a similar rate of colonization for C57BL/6 WT and CEA mice after 6 weeks, 3 months or a long-term infection of 12 months (Supplementary Fig. 8c–e). The histopathological evaluation of mouse stomach tissue did not show significant differences in activity or chronicity of CEA- versus WT-infected animals up to 3 months of infection (Supplementary Fig. 9a), but the long-term infected animals showed a significant difference in the activity of gastritis after 12 months (P < 0.05) (Supplementary Fig. 9b). Re-isolated Hp from CEA versus WT mice after 6 weeks, but not after 3 months or 12 months of infection showed a significantly higher rate of bacteria with a functional cag-T4SS, as measured by their capability for CagA translocation into AGS cells (Supplementary Fig. 9c). After 3 and 12 months of infection, the capacity for IL-8 induction by Hp was significantly reduced in CEA as compared to WT-infecting re-isolates (Supplementary Fig. 9f–h).
Collectively, these data suggest that in the early phase of infection (6 weeks) a stronger adhesion of Hp in the CEA as compared to the WT mouse gastric mucosa might increase gastric inflammation and the host immune response. Subsequently, according to the rheostat model31, the stimulated inflammation would provoke Hp in CEA mice to more quickly switch off the cag-T4SS to avoid its elimination, supporting a role for CEACAM5 as a functional receptor in this mouse model in vivo.
Expression of CEACAM5 in the infected human stomach is cag-T4SS-dependent
To further evaluate the role of CEACAMs during Hp infection in the human stomach, we characterized gastric biopsy isolates from Hp-infected patients for virulence traits (for example, functional cag-T4SS, CagA tyrosine phosphorylation in AGS cells, IL-8 induction). We therefore generated thin sections from formalin-fixed paraffin blocks for microscopy studies to determine the pathology, and classified the gastric biopsy material into three groups: (1) non-Hp-infected (by urea breath test (UBT) and culture, six biopsies), (2) infected with CagA-positive and translocation-competent Hp (CagA+P+; 11 biopsies) and (3) infected with CagA translocation-negative (CagA+P−) or CagA-negative (CagA−P−) strains (four biopsies). Strikingly, CEACAM5 expression was significantly higher in CagA+P+ strain infections relative to all others (Figs 5 and 6a). Despite a strong surface localization of CEACAM5 in CagA+P+ infected patients, an AGS luciferase reporter cell line did not sense a direct transcriptional upregulation of CEACAM5 promoter-driven luciferase expression by the Hp cag-T4SS, or phosphorylated injected CagA under in vitro infection conditions for up to 3 h (Supplementary Fig. 10). These data suggest that not Hp per se, but the inflammation, or the specific cytokine milieu induced by CagA+P+ Hp infection in the gastric mucosa, might be driving CEACAM5 expression in infected tissue.
We have discovered here, for the first time, distinct members of the CEACAM protein family as novel protein receptors for Hp. In addition, the outer membrane protein HopQ, a member of the comprehensive Hp Hop protein family, could unambiguously be identified as a novel bacterial adhesin binding CEACAM receptors. We show that the IgV-like N-terminal domains of CEACAM1, CEACAM3, CEACAM5 and CEACAM6 are sufficient for Hp binding. The HopQ–CEACAM binding is essential for CagA injection into host cells carrying no additional receptors for Hp binding (for example, HEK293 cells transfected with human CEACAMs, MKN45 cells). Thus, CagA injection correlates with the binding capacity of Hp to the corresponding host cells. These results also indicate that a direct binding of components of the cag-T4SS (CagL, CagA, CagY and CagI) to β1 integrin receptors3,4 is not sufficient for CagA translocation by Hp. Instead, apart from integrin binding, interaction of at least one outer membrane adhesin with a distinct cellular receptor (for example, HopQ–CEACAM, or another adhesin-receptor pair) is required.
Binding experiments with soluble CEACAM IgV constructs, or recombinant HopQ and CEACAM domains in ITC experiments, showed that binding of Hp is highly specific, strong and independent of glycosylation of CEACAMs. HopQ does not recognize murine, canine or bovine CEACAM receptors, which clearly underpins the very high species adaptation during a long co-evolution of Hp with humans32. We also did not detect a specific binding of the HopQ adhesin to the antrum or corpus of Mongolian gerbil tissue, suggesting that no compatible CEACAM receptor(s) for HopQ-mediated binding of Hp is/are present on gerbil antrum or corpus tissue. In the humanized CEACAM5 mouse, we observed a CEACAM5-dependent effect concerning gastritis activity after 12 months of infection, indicating that CEACAM receptors play a role for Hp infection and might have long-term consequences in vivo, but further work will be necessary to determine the role and functional consequences of the HopQ–CEACAM interaction in vivo.
In gastric sections of Hp-infected patients we observed a significantly higher CEACAM5 expression in the corpus versus the antrum of the patient gastric mucosa (Fig. 6a), suggesting that CagA+P+ Hp strains might exploit the inflammatory conditions to more efficiently adhere to the gastric corpus. Although our in vitro CagA translocation data show a correlation between the amount of CEACAM expressed on the cell surface and CagA translocation efficiency (Fig. 3), it remains at present unclear whether upregulated CEACAM5 expression in the gastric corpus enhances CagA delivery in vivo. CagA-P− or CagA + P− Hp are seen on the surface of gastric sections (Fig. 5b,c). However, these bacteria do not co-localize with CEACAM5 (Fig. 5b,c), which is in contrast to the CagA+P+ Hp strains, which often show intimate contact with CEACAM5, suggesting direct adhesion. Thus, our data support the general finding that CagA+P− and CagA–P− Hp strains are mainly located in the mucus layer, or adherent to surface-associated glycoprotein receptors, such as Lewisb or other Lewis antigens, rather than on the epithelial cell surface or in the tissue.
Hp might also encounter other human CEACAMs in the gastric mucosa, such as CEACAM1, or the granulocyte-specific member of the CEACAM family, CEACAM3, which was identified as the major receptor responsible for opsonin-independent phagocytosis and bacterial elimination33. By contacting these receptors, Hp might induce intracellular signalling events via stimulation of immunoreceptor tyrosine-based inhibition motifs (ITIM) or immunoreceptor tyrosine-based activation motifs (ITAM) in the cytoplasmic tail of CEACAM1 or CEACAM3, respectively. A better understanding of the molecular details of these interactions could help to understand the strategies of Hp for chronic infection and consequently to devise better treatment schemes against Hp infections.
Bacterial strains and cell culture
Hp strains were cultured at 37 °C under microaerophilic conditions on GC agar plates (Oxoid) supplemented with vitamin mix (1%), horse serum (8%), trimethoprim (5 μg ml–1) and nystatin (1 μg ml–1) (serum plates). Chloramphenicol (6 μg ml–1), kanamycin (8 μg ml–1) or streptomycin (250 μg ml–1) were added to select for transformants or screen colonies for resistance to either drug, when necessary. Plates were incubated for 24–48 h under microaerobic conditions (85% N2, 10% CO2, 5% O2) at 37 °C. Moraxella catarrhalis strains 11994 and 9143 were incubated at 37 °C using Columbia blood agar plates (Oxoid). The gerbil-colonizing strain P322 (cag-PAI positive, vacAs1m1, hopQ I), a patient isolate, was passaged several times through the gerbil stomach. All bacteria were passaged two to three times before the first experiments were performed. The agar plates contained GC (Oxoid) supplemented with 8% horse serum (PAA) and 1% vitamin mix (100 g l–1 d-glucose, 10 g l–1l-glutamine, 26 g l–1l-cystein, 0.1 g l–1 co-carboxylase, 20 g l–1 Fe(III) nitrate, 3 g l–1 thiamin, 13 mg l–1 p-aminobenzoic acid, 250 mg l–1 nicotinamide-adenine dinucleotide, 10 mg l–1 guanine, 0.15 g l–1 l-arginine and 0.5% uracil). All cells were grown at 37 °C and 5% CO2 and subcultured every 3–4 days. AGS and HeLa cells were grown in RPMI/10% FCS, MKN45 cells in RPMI/20% FCS and Kato III cells in DMEM/20% FCS. For MKN28, RPMI medium was supplemented with 10% FCS, 10 mM HEPES, 1 mM sodium pyruvate and 0.1 mM non-essential amino acids. HEK293 cells were cultured in DMEM/15% FCS and supplemented with 1 mg ml–1 G418 when they were transfected with CEACAM constructs. All cell lines used in this study are routinely tested for mycoplasma contamination (once a year) and proved to be negative (4′,6-diamidino-2-phenylindole (DAPI) staining; LookOut Mycoplasma PCR Detection Kit, Sigma Aldrich, catalogue no. MP0035). Cells are routinely checked for their morphology (size, granularity), growth behaviour and species identity (human cell lines). All cell lines used have been verified by a cell line authentication test (Eurofins Medigenomics). AGS cells (CRL-1739) and Kato III cells (HTB-103) were purchased from ATCC, HEK293 cells from Tularik (taken over by Amgen) and MKN45 cells (ACC 409) from DSMZ.
Plasmids for bacteria and eukaryotic cell manipulation
Hp strain P12 was used for the generation of marker-free outer membrane protein mutants. OMP genes babA, babB, sabB, hopZ and hopQ were deleted using the contra-selectable streptomycin susceptibility method, as described in ref. 34. The up- and downstream regions of the gene of interest were cloned in a pBluescript II SK(+) vector (Agilent Technologies) flanking a resistance cassette. Primers used for amplification of DNA fragments adjacent to the genes encoding outer membrane proteins are listed in Supplementary Table 2. In strain P12, two identical gene copies of babB (in the babB and babC loci) and sabB (in the sabB and sabA loci) had to be deleted to obtain marker-free mutants. By homologous recombination of the adjacent gene regions, the resistance marker was first inserted in the locus of the gene of interest, creating the distinct outer membrane protein gene mutant strain. In the second step, the resistance cassette was completely removed, resulting in a marker-free mutant strain according to the contraselection method34. P12ΔcagI was generated using the method described above and served as negative control for CagA translocation competence in all in vitro infection experiments35. For genetic complementation of the hopQ mutant, the P12 gene was cloned in an E. coli–H. pylori shuttle vector based on the pHel3 plasmid of Hp, as described in ref. 18.
Mammalian expression plasmids encoding the GFP-tagged amino-terminal IgV-like domains of human CEACAM1 (hCEA1-N), murine CEACAM1 (mCEA1-N), bovine CEACAM1 isoforms a or b (bCEA1a-N or bCEA1b-N, respectively), as well as canine CEACAM1 (cCEA1-N) have been described previously36. The CEACAM5 orthologue from Macaca mulatta was constructed by using macaque cDNA (reverse transcribed from mRNA derived from a colon tissue biopsy and generously provided by C. Roos, German Primate Center, Göttingen, Germany) and amplifying the coding sequence for the amino-terminal IgV-like domain of macaque CEACAM5 with primers macCEA-IF-sense′ and CEACAM5-NT-IF-anti (Supplementary Table 2). The resulting PCR fragment was cloned into pDNR-Dual using the In-Fusion PCR Cloning Kit (Clontech) and transferred by Cre-mediated recombination into pLPS-3′ EGFP (Clontech) resulting in GFP fused to the carboxy-terminus of the expressed protein.
For the cloning of HopQ and CEACAM N-terminal domains, HopQ (residues 25–460, Hp strain P12) was cloned into a NotI/XhoI modified pET21d vector containing an N-terminal maltose binding protein-tag (MBP, containing several mutations: D82A/K83A/E172A/N173A/K239A) and a C-terminal hexa-histidine tag. Human CEACAM5 (residues 34–141, CEA-N) was cloned into a NcoI/XhoI cut pET21d vector with a stop codon before the hexa-histidine tag. CEACAM1, CEACAM3 and CEACAM8 were synthesized and codon optimized as GeneArt Strings (Life Technologies). GeneArt Strings were cut and cloned into an NcoI/XhoI cut pET21d vector without a purification tag.
Mammalian expression plasmid pRc/CMV-CEA has been described previously37. CEACAM5 truncations were constructed by PCR using inverted phosphorylated primers and subsequent DpnI digestion followed by ligation. Deletion of IgC-like domains was performed with primers LH90/LH91, resulting in plasmid pLH45 and deletion of IgV-like domains with primers LH79/LH89 (Supplementary Table 2) leading to pLH46.
Bacterial pulldown assay
The bacterial pulldown was performed as described in ref. 24. Bacteria of an overnight culture were resuspended in 1 ml PBS and the optical density at 550 nm (OD550) was measured. A total of 4 × 106 bacteria, calculated by the OD500 measurement, were incubated with cell culture supernatants containing CEA-N-GFP. After an incubation time of 30 min at room temperature with head-over-head rotation, the bacteria were washed twice with PBS containing 0.9 mM CaCl2 and 0.5 mM MgCl2 and resuspended in SDS loading buffer. After heat denaturation, the samples were analysed by western blotting as described previously24. Alternatively, bacteria were resuspended in PBS/2% FCS and analysed by flow cytometry (FACS Canto II, BD Biosciences).
Protein expression of MBP–HopQ and MBP–CEACAM5 in E. coli and purification
HopQ, MBP and CEA-N were expressed in E. coli BL21(DE3) pLysS cells. Cells were grown at 37 °C until an absorbance at 600 nm (A600) of 0.6 and were induced with 1 mM isopropyl-β-d-thiogalactopyranoside and grown for 4 h at 37 °C (CEACAMs) or 18 °C (HopQ or MBP). HopQ and MBP cells were resuspended in ice-cold 20 mM Tris, 500 mM NaCl, pH 7.5 (NiBB) and lysed by sonication. HopQ and MBP were purified by nickel affinity chromatography (HisPur Ni-NTA Resin, Thermo Scientific) by passing the soluble fraction over the beads and removing unbound protein with 20 ml of NiBB. Beads were further washed with NiBB, 50 mM imidazole, pH 7.5 (15 ml) before elution with NiBB, 400 nM imidazole, pH 7.5 (10 ml). The elution and wash steps were combined and diluted twofold with 20 mM Tris, 150 mM NaCl, 1 mM EDTA, pH 7.5 (amylose binding buffer) and loaded onto amylose resin (NEB). Resin was washed with 30 ml of amylose binding buffer and protein eluted with 15 ml of amylose binding buffer containing 40 mM maltose. HopQ and MBP were further purified by size exclusion chromatography using a Superdex 200 column (GE Healthcare) equilibrated with 20 mM Tris, 150 mM NaCl, 40 mM maltose, pH 7.5. CEACAM cells were resuspended in ice-cold NiBB, 10 mM EDTA, 1% vol/vol Trition-X-100 and lysed by sonication. The soluble fraction was discarded and inclusion bodies resuspended in the same buffer and further disrupted by sonication. Inclusion bodies were then washed with 50 mM Tris, 1 M NaCl, pH 8.0 and then NiBB. Inclusion bodies were finally dissolved in 30 mM Tris, 150 mM NaCl, 8 M urea, pH 8.3 and CEACAMs refolded at 4 °C by diluting the redissolved inclusion bodies slowly (100 µl min−1) 1:20-fold into 50 mM N-cyclohexyl-2-aminoethanesulfonic acid (CHES), 500 mM l-arginine, pH 9.2, overnight. Refolded CEACAMs were dialysed against 50 mM Tris, pH 8.0 and concentrated by MonoQ (GE Healthcare). CEACAMs were further purified by size exclusion chromatography using a Superdex 200 column equilibrated with 30 mM Tris, 150 mM NaCl, pH 8.3. See Supplementary Table 2 for oligonucleotides and Supplementary Table 4 for plasmid constructs.
Isothermal titration calorimetry (ITC)
All proteins were dialysed against 30 mM Tris, 150 mM NaCl, 5 mM maltose, pH 7.5. ITC experiments were performed using an iTC200 instrument (GE Healthcare). A typical experiment consisted of loading the syringe with CEACAM (240 µM) and MBP or HopQ placed in the cell (24 µM). Titrations were performed at 25 °C with 17 injections of 2.46 µl aliquots, with 210–240 s intervals between injections. For CEACAM8, the number of injections was decreased but larger volumes were injected. Heats of dilutions were also measured and subtracted from each data set. CEACAM5 binding to MBP alone was also tested. All data were analysed using Origin 7.0 software.
Infection experiments of eukaryotic cell lines
Before infections, cells were overnight serum-starved using cell culture medium without FCS. The next morning, cells were grown for 2 h in PBS containing 10% FCS and then infected with bacteria. In all experiments, a MOI (multiplicity of infection) of 60 was used and the infection was stopped after 4 h on ice. Cells were collected in PBS containing 1 mM sodium vanadate, 1 µM leupeptin, 1 µM pepstatin and 1 mM phenylmethylsulfonyl fluoride and resuspended in SDS loading buffer. Equal amounts of samples were loaded on a gel and analysed by western blotting.
Antibodies used for flow cytometry or western blotting
For the detection of CEACAM-N-GFP constructs in the bacterial pulldown, monoclonal GFP antibody (clone JL-8, Clontech) was used. The densitometry of GFP signals was evaluated by Image Lab 4.1 software, and the band intensity was normalized against rabbit polyclonal AK263 (α-RecA) antibody signals38. In the in vitro infection model, polyclonal rabbit anti-CagA antibody (AK257)5 was used to detect CagA, monoclonal anti-phosphotyrosine antibody (clone 4G10, Millipore) was used to monitor CagA tyrosine phosphorylation, AK268 (ref. 39) for the detection of CagA, and anti-tubulin antibody (Abcam) for normalization of cell lysates.
The antibody AK298 (this study) was generated against a conserved peptide sequence of HopQ (type I) of Hp strain P12 (SKGEKLEAHVTTSKYQ) by immunization of rabbits (Genosphere Biotech). The anti-peptide antibody was verified to be specific for HopQ type I by using a P12ΔhopQ strain and was used for the detection of HopQ.
Cells were stained with specific antibodies (Aldevron) against human CEACAM1 (clone GM-8G5), CEACAM5 (clone 6/3/13), CEACAM6 (clone 9A6), CEACAM8 (clone LT-3AG-B2) or pan CEACAM (clone D14HD11) according to the manufacturer's instructions and analysed in a flow cytometer (FACS Canto II, BD Biosciences).
HEK293 adherence assay
Cells were infected with GFP expressing Hp using an MOI of 60 for 1 h at 37 °C and 5% CO2. After three washing steps with PBS containing 2% FCS, the cells were collected and analysed in a flow cytometer (FACS CantoII, BD Biosciences). The mean fluorescence intensity from non-infected cells was subtracted from infected samples. At least three experiments were performed and analysed with Student's t-test.
Transfection of HEK293 cells
HEK293 cells were transfected in 10 cm2 culture dishes with Lipofectamine 2000 (Life Technologies) reagent following the manufacturer's recommendations. After 24 h, the cells were washed and 1 mg ml–1 geneticin (Life Technologies) was added to select for stably transfected cells. After 3–4 weeks of subculturing, cells were stained with monoclonal CEACAM antibodies (pan CEACAM clone D14HD11 or CEACAM5 clone 26/3/13, Aldevron, 1.2 µg per 106 cells) and subsequently with secondary Alexa488-coupled anti-mouse IgG antibody before FACS (FACS Aria-II, BD Biosciences). After sorting, the cells were reanalysed for CEACAM expression with a Canto-II (BD Biosciences) flow cytometer. The same protocol was used for transient transfections. Expression analysis was performed 24 h after transfection.
HEK293 cells were found to produce no CEACAM1, 5 and 6 on their surface. Furthermore, they are easily and efficiently transfected and proved to be free of mycoplasma. We therefore chose these cells for the stable CEACAM transfection experiments. In the International Cell Line Authentication Committee (ICLAC) database they are described as cross-contaminated with HeLa cells. Because we also tested HeLa cells as negative for CEACAM production, this cross-contamination did not interfere with the results of our experiments (for example, CagA translocation). MKN28 cells are also listed in the ICLAC database as cross-contaminated with human gastric MKN74 cells. MKN28 cells were identified as CEACAM-deficient and for this work only used to support the correlation between CEACAM surface expression and CagA translocation capability of Hp.
CEACAM5-luciferase reporter assay
For construction of the CEACAM5 promoter plasmids pGL4.21-CEACAM5_P0.4Kb and pGL4.21-CEACAM5_P3.1Kb, a 430 bp or a 3,109 bp region upstream of the CEACAM5 gene was amplified by PCR using a plasmid template pCEA3300 with primers ZQ113 and ZQ112 or ZQ111 and ZQ112, respectively, and cloned into the EcoRV and HindIII sites of pGL4.21(Promega). AGS cells were co-transfected with pGL4.21-CEACAM5 promoter plasmids and the internal control pGL 4.74 (Promega) in the ratio 1:1, except the promoter-less control comprised AGS cells co-transfected with empty vector pGL4.21 and the internal control pGL4.74 (n = 3). Transiently transfected AGS cells were infected with H. pylori strains at an MOI of 60 for 3 h at 37 °C, 5% CO2. Firefly luciferase and Renilla luciferase activities were measured with a Dual-Glo Luciferase Assay system (Promega). Transfection control was normalized as 1.
CEACAM knockdown in AGS cells
AGS cells were transfected with Hs_CEACAM1_11 siRNA (CTGCACAGTACTCCTGGCTTA) targeting CEACAM1, CEACAM5 and CEACAM6, Hs_CEACAM5_6 (CGCAGTATTCTTGGCGTATCA), Hs_CEACAM6_7 (AAGAAG ATAGATCCAATTAAA) (Qiagen) and AllStars Negative Control siRNA (Qiagen) by Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions. After 48 h, the siRNA effect on expression of CEACAM1, 5 and 6 was evaluated against AGS control cells by flow cytometry analysis (FACS Canto II, BD Biosciences). AGS control cells and AGS/CEACAM-siRNA with a downregulation of CEACAM1, CEACAM5 and CEACAM6 were infected with Hp strains and tested for CagA translocation.
Immunofluorescence of gastric epithelial cell lines AGS, Kato III and MKN45
AGS and Kato III cell lines were cultured in RPMI medium supplemented with 10% fetal calf serum (Gibco–Life Technologies). For cell line MKN45, RPMI medium supplemented with 20% fetal calf serum (Gibco–Life Technologies) was used. All cell lines were cultured at 37 °C and 5% CO2. For immunofluorescence, cells were seeded on sterile cover slides in a 24-well plate (Costar) at a cell density of 2 × 105 cells per cm2 and cultivated further. After 2 days, cells were infected with Hp WT or isogenic mutant strains with an MOI of 10 for 4 h at 37 °C and 5% CO2. For immunostaining, cells were fixed with 4% paraformaldehyde for 10 min and subsequently blocked with 2% fetal calf serum in PBS (Gibco–Life Technologies) overnight. Fixed cells were incubated with mouse anti-CEACAM5 (1:300, clone 26/3/13, Aldevron) and rabbit anti-Hp (1:400, AK175) for 1 h at room temperature. After washing with DPBS, cells were incubated with secondary antibodies (1:1,000, goat anti-rabbit Alexa Fluor 488, goat anti-mouse Alexa Fluor 555; Invitrogen - ThermoFischer Scientific) and Alexa Fluor 647 phalloidin (300 U) (1:1,000; Invitrogen–ThermoFischer Scientific) for 1 h at room temperature in the dark. Afterwards, cell nuclei were stained with DAPI (5 µg ml–1; Sigma) for 5 min. Samples were mounted with Mowiol (Sigma). Pictures were taken with a CLSM (Leica SP5) with a ×63 oil immersion objective (HCX PL APO lamda blue ×63, 1.4 oil). Image processing was performed with LAS AF software (Leica).
Mouse and Mongolian gerbil infections
Specific-pathogen-free, transgenic C57BL/6_CEACAM5± mice (Mus musculus) were bred, infected and killed at the Max von Pettenkofer-Institut (Munich, Germany). The animals (males and females) were infected orogastrically three times on three subsequent days with a dose of 109 bacteria. Animals were infected with Hp strain PMSS1 at between 8 and 12 weeks of age. Littermates not expressing the CEA5 receptor (C57BL/6_CEACAM5−/−, here designated WT) were inoculated with identical volumes and preparation of bacteria. All experiments and procedures were conducted in accordance with the guidelines for the care and use of laboratory animals and approved by the Regierung von Oberbayern (AZ 55.2-1-54-2532-34-14). Groups of animals (n = 10) were killed after 6 weeks, 3 or 12 months.
Specific-pathogen-free Mongolian gerbils (Meriones unguiculatus, outbred) were bred, infected and killed at the Max von Pettenkofer-Institut. The animals (males and females) were infected orogastrically three times on three subsequent days with a dose of 109 bacteria. Animals were infected with the Hp strain P322wt (n = 10) and P322wtΔhopQ (n = 11) at between 8 and 14 weeks of age. All experiments and procedures were conducted in accordance with the guidelines for the care and use of laboratory animals and approved by the Regierung von Oberbayern (AZ-55.2-1-54-2532-1-11). Animals were killed after 8 weeks.
Sample-taking from animals
Animals were killed with CO2. Blood was taken via cardiocentesis and serum was obtained by centrifugation. Stomachs were opened along the greater curvature and washed with PBS. The whole stomach was divided into two equal halves. One half was homogenized (glass homogenizer, Ochs) in brucella broth (BB) medium for isolation of bacteria. For mice, the other half was fixed in Tissue-Tek O.C.T Compound (Sakura) and for Mongolian gerbils the other half was fixed in formalin for immunohistology. Appropriate dilutions of homogenates were spread on selective serum plates (GC agar Difco). The number of colony forming units (c.f.u.) was calculated per gram of gastric tissue.
Mouse stomachs were opened along the lesser curvature, flushed with ice-cold PBS and fixed in PBS-buffered 4% formaldehyde at room temperature (2 h) for paraffin-embedded tissues. Serial 3 µm paraffin sections were dewaxed and subjected to heat antigen retrieval at pH 9.0 (Target Retrieval Solution; Dako). After blockage of endogenous peroxidase, tissue sections were incubated with primary antibodies (α-CEACAM5, Dako, A011502, 1:350) for 1 h at room temperature or overnight at 4 °C, then with horse radish peroxidase-coupled secondary antibodies and 3-amino-9-ethylcarbazole (ImmPRESS Anti-Rabbit Ig Polymer Detection Kit; Vector Labs) and counterstained with haematoxylin.
Grading of inflammation
H&E-stained longitudinal paraffin sections of antrum and corpus were graded on the intensity of inflammation, metaplasia and ulcer development (scale 0–5) by a pathologist blinded to the treatment, according to Garhart and co-authors28.
Immunofluorescence of gastric biopsies
Paraffin sections of antrum and corpus were deparaffinized and treated with citrate buffer for antigen retrieval (microwave). Sections were blocked with 5% goat serum in PBS and incubated with mouse anti-CEACAM5 (Aldevron) and rabbit anti-Hp (AK175), or mouse anti-CagA (Antibodies Online). After washing, the secondary antibodies were applied (goat anti-rabbit Alexa488 and goat anti-mouse Alexa555; Invitrogen) and incubated. Cell nuclei were stained with DAPI. Micrographs were taken with an epifluorescence microscope (Olympus BX61) ×40 objective and a CLSM (Leica SP5) with a ×63 oil immersion objective.
Measurements were performed with Volocity (Perkin Elmer) using the images taken from the CLSM Leica SP5. Briefly, CEACAM5 expression was evaluated as follows. Gastric sections were screened for the highest expression of CEACAM5 and the area of the CEACAM5 signal was then normalized to the whole epithelial cell area (CEACAM5 per nm2) (Fig. 6a).
Tetramethyl-rhodamine (TAMRA) labelling of Hp and ex vivo gastric tissue infections
Labelling and infection was performed as described earlier40. In brief, a suspension of Hp in PBS (500 µl, OD550 = 1.0) was mixed with 2 µl TAMRA (5-(and 6)-carboxytetramethylrhodamine succinimidyl ester, mixed isomers, Molecular Probes; 10 mg ml–1 in DMSO) and incubated in the dark for 30 min. Bacteria were then washed three times with PBS/0.05% Tween20/1% bovine serum albumin (BSA) and resuspended in 500 µl PBS for direct use. Biopsy sections (cryo-sections) were cut and fixed on microscope slides with ice-cold acetone, rehydrated in PBS (10 min, 4 °C) and blocked with BSA (1%, 1 h, 37 °C). TAMRA-labelled Hp was added and incubated for 1 h at 37 °C and 5% CO2. To remove unbound bacteria, the glass slides were rinsed six times with PBS in excess and sealed in Entellan (Merck) with a cover slide for fluorescence microscopy.
Quantification of Hp ex vivo binding on mouse and Mongolian gerbil gastric sections
To quantify the ex vivo binding of Hp to gastric tissue, the area of red fluorescence was measured and normalized to the whole tissue size. In brief, images were taken on an Olympus microscope with ×10 magnification and saved as multipage-TIFF files (DAPI and Cy3 channel). For analysis of Hp ex vivo binding, images were blinded and evaluated with ImageJ (version 1.50e). All images were calibrated (width, 1,376 pixels; height, 1,032 pixels; 1.55 pixels per µm) and analysed individually. The size of the tissue was measured by applying the free-hand selection tool on the DAPI channel and subsequently measuring the selected area (in µm2). The selection was applied to the Cy3 channel and the picture was cropped accordingly. A binary image was generated using the ImageJ Threshold tool. The threshold was set individually for each picture to compensate for individual background fluorescence. The area within the binary image corresponds to the red fluorescent Hp bound to the tissue. The size of this area was measured in µm2 and the ratio between the area of red fluorescence (Cy3 channel) and the whole tissue size was calculated and considered as the degree of Hp ex vivo binding to tissue samples.
Human biopsy samples
Patients who underwent upper gastrointestinal endoscopy from August 2010 to December 2013 were enrolled in the study in Nigeria, which was supported by the Africa Infectiology Initiative of the Deutsche Forschungsgemeinschaft (DFG). Upper gastrointestinal endoscopy was performed in Lagos State University Teaching Hospital, Ikeja, University College Hospital, Ibadan, Obafemi Awolowo University Teaching Hospitals Complex, Ile-Ife, Jos University Teaching Hospital, Jos, and Lagos State University Teaching Hospital, Campus Idi-Araba, Lagos. Endoscopies were performed for medically indicated reasons, not for research only. Written informed consent was obtained from each patient before enrolment, in agreement with the Helsinki Declaration. The study was reviewed and approved by an Ethics Committee of the Nigerian Institute of Medical Research (IORG0002656) and of the Ludwig-Maximilians-Universität München (registration no. 335-08). During endoscopic examination of the patients, two biopsy specimens were obtained from the gastric antrum and two from the gastric corpus. One of the samples was used for isolation of bacteria, and the other specimen was fixed in 4% formalin and embedded in paraffin for histopathological examination. Hp status was assessed by C14-urease test, bacterial culture and rapid urease test (CLO test) (Tri-Med Distributors). Histology with H&E staining was performed to determine pathological changes such as inflammation, metaplasia and dysplasia. Patients were considered positive for Hp infection when at least one of these examinations yielded positive results. Patients were defined as Hp-negative if all test results were negative.
The results of image analyses were transferred to GraphPad Prism5 (GraphPad Software) to determine the statistical significance with appropriate tests. In in vitro experiments, to determine if the samples were normally distributed, the Shapiro–Wilk test was performed as a pre-test (≥3 independent, biological replicates) and statistical significance was assessed by the one- or two-sided Student's t-test. For more than two groups a one-way analysis of variance (ANOVA) was performed. In in vivo experiments, data that were not normally distributed were analysed by the Mann–Whitney U-test for unpaired groups. For the contingency tables the chi-square test was used. The size of the animal group (n) in all experiments was ≥9. The calculation of the size was based on a standard deviation (δ/σ) of 1.25, power of 80% and one-tailed Student's t-test. (δ = biological relative difference, σ = estimate of variation, s.e.m. = s.d./√n). Animal group sizes were as low as possible and empirically chosen. The criteria for the exclusion of animals were pre-established. No animals were excluded from any analysis. Randomization was done by a non-involved person (animal care keeper) and blinding was performed during the experiment and when assessing the outcome. Quantitative data are depicted as mean ± s.e.m. Results were considered statistically significant if the P value was below 0.05.
The data that support the findings of this study are available from the corresponding author upon reasonable request; see author contributions for specific data sets.
The authors acknowledge support from collaborating gastroenterologists who collected gastric biopsies in the framework of the Africa Infectiology Study in Nigeria, especially at Lagos University Teaching Hospital, the University College Hospital (UCH) Ibadan and the University Teaching Hospital Complex Ile-Ife. The authors thank V. Naegele for CEACAM1 and CEACAM5 transfected HEK293 cells, E. Vetter for staining of CEA immunohistology sections, M. Schiemann and L. Henkel for FACS sorting of transfected cells and E. Weiss for technical support. This work was supported by grants from DFG (HA2856/6-2 to C.R.H. and HA2697/16-1, 17-1 and 18-1 and SFB914 Project B05 to R.H.) and in part by an Alexander von Humboldt Foundation Experienced Research Fellowship (to E.J.S.).
Supplementary Figures 1-10, Supplementary Table 1-4, Original gel images