Helicobacter pylori is a bacterial pathogen that colonizes the gastric niche of ∼ 50% of the human population worldwide and is known to cause peptic ulceration and gastric cancer. Pathology of infection strongly depends on a cag pathogenicity island (cagPAI)-encoded type IV secretion system (T4SS). Here, we aimed to identify as yet unknown bacterial factors involved in cagPAI effector function and performed a large-scale screen of an H. pylori transposon mutant library using activation of the pro-inflammatory transcription factor NF-κB in human gastric epithelial cells as a measure of T4SS function. Analysis of ∼ 3000 H. pylori mutants revealed three non-cagPAI genes that affected NF-κB nuclear translocation. Of these, the outer membrane protein HopQ from H. pylori strain P12 was essential for CagA translocation and for CagA-mediated host cellresponses such as formation of the hummingbird phenotype and cell scattering. Besides that, deletion of hopQ reduced T4SS-dependent activation of NF-κB, induction of MAPK signalling and secretion of interleukin 8 (IL-8) in the host cells, but did not affect motility or the quantity of bacteria attached to host cells. Hence, we identified HopQ as a non-cagPAI-encoded cofactor of T4SS function.
Helicobacter pylori is a microaerophilic, spiral-shaped Gram-negative bacterium that colonizes the gastrointestinal tract of ∼ 50% of the human population worldwide. Chronic infection always leads to gastritis and can also result in more serious gastric diseases, such as peptic ulceration, gastric cancer and mucosa-associated lymphoid tissue lymphoma (MALT) (Nomura et al., 1991; Cover and Blaser, 1992; Walker and Crabtree, 1998; Peek and Blaser, 2002; Farinha and Gascoyne, 2005; Peek and Crabtree, 2006).
H. pylori produces numerous virulence factors (Gressmann et al., 2005) that enable bacteria to adapt and multiply within the hostile environment of the human gastrointestinal tract (Blaser and Atherton, 2004; Bauer and Meyer, 2011). Among the known virulence determinants, the type IV secretion system (T4SS) stands out since its presence in the bacterial genome strongly correlates with serious H. pylori-induced pathologies (Ogura et al., 2000; Huang et al., 2003; Blaser and Atherton, 2004). The T4SS is encoded by a 40 kb pathogenicity island (cagPAI), which comprises 31 genes (Censini et al., 1996; Tomb et al., 1997) and was probably acquired by horizontal gene transfer. The T4SS serves to translocate the CagA effector protein from bacteria into gastric epithelial cells (Backert et al., 2000; Odenbreit et al., 2000; Stein et al., 2000), and provokes host cell responses, such as altered cell motility and signal transduction (Churin et al., 2001; 2003; Higashi et al., 2002; Mimuro et al., 2002; Tsutsumi et al., 2003; Suzuki et al., 2005). More recently, CagA has been implicated in the suppression of innate host cell responses (Bauer et al., 2012) and in the cause of an advanced invasive phenotype by disrupting cell-cell adhesion and perturbing epithelial cell differentiation (Bagnoli et al., 2005). Also the T4SS itself induces host cell responses in gastric epithelial cells such as the activation of the transcription factor NF-κB (Glocker et al., 1998). Together with the transcription factor AP-1 (activator protein 1), NF-κB regulates the expression of interleukin 8 (IL-8) (Aihara et al., 1997; Sharma et al., 1998) and is involved in cellular apoptotic, proliferative and immune responses (Hoffmann and Baltimore, 2006).
The H. pylori genome encodes numerous outer membrane protein (OMP) encoding genes. Most OMPs can be assigned to two protein families: Helicobacter OMPs (Hop) and Hop-related proteins (Hor) (Alm et al., 2000). Hops are reminiscent of autotransporter proteins (Jose et al., 1995), many of which are crucial adhesins and subject to phase variation and/or recombination (Alm and Trust, 1999; Odenbreit et al., 1999; Hauck and Meyer, 2003; Talarico et al., 2012). The Hop family adhesins include the H. pylori blood group antigen binding adhesin (BabA) (Ilver et al., 1998); the sialic acid binding adhesin SabA (Aspholm et al., 2006; Lu et al., 2007); the adherence-associated lipoproteins A and B (AlpA and AlpB) (Odenbreit et al., 1999); and the HopZ protein (Peck et al., 1999; Kennemann et al., 2011). Moreover, HopQ was shown to influence adherence of a subset of H. pylori strains to human epithelial cells (Loh et al., 2008). Another, related H. pylori outer membrane protein is reportedly involved in inducing IL-8 secretion in host cells (Yamaoka et al., 2000; 2002a) while studies by Odenbreit et al. (2002) failed to show correlations between CagA translocation, IL-8 induction and the expression of nine different OMPs. Another five members of the Hop family, named HopA through to HopE, have been speculated to function as porins (Doig et al., 1995; Exner et al., 1995). Therefore, the exact contribution of Hops to the modulation of host cell responses and pathogenesis remain fragmentary and contradictory in part.
To identify novel H. pylori factors that are required for cagPAI-mediated pathogenesis, we screened an H. pylori transposon mutant library (Salama et al., 2004) for mutants affecting a T4SS-dependent host epithelial cell response; namely, activation of the transcription factor NF-κB (Fischer et al., 2001; Maeda et al., 2001; Bauer et al., 2005; Bartfeld et al., 2010). From a total of 59 hits influencing T4SS-dependent NF-κB activation in H. pylori-infected cells, we identified HopQ as an important accessory factor. Our analysis also revealed a function of HopQ for other T4SS-dependent host cell responses in addition to NF-κB activation.
High throughput quantitative assay for H. pylori mutant analysis
To identify novel bacterial factors involved in the induction of T4SS-mediated host cell responses, we screened an H. pylori transposon-based mutant library for mutants defective in activating the eukaryotic transcription factor NF-κB. The H. pylori type I strain G27 (cagPAI-positive), the strain from which the mutant library was derived (Salama et al., 2004), stimulated NF-κB nuclear translocation in human AGS gastric epithelial cells upon infection, as determined by immunofluorescent staining of the NF-κB subunit p65 (Fig. 1A). Similarly, G27 infection stimulated NF-κB activation in recombinant AGS-SIB02 cells – a cell line in which p65-GFP nuclear translocation can be analysed as a high-throughput-compatible quantitative read-out (Bartfeld et al., 2009). NF-κB activation by G27 peaked at 45 min post infection (p.i.), whereas a T4SS-defective congenic G27ΔcagE strain failed to activate NF-κB (Fig. 1B). Moreover, infection of AGS-SIB02 cells with the H. pylori strain P12 (cagPAI-positive) led to the induction of an oscillating activation phenotype exhibiting two peaks (45 and 150 min p.i.), very similar to the pattern induced by infection with G27 (Fig. 1C). These studies confirmed the T4SS-dependent activation of NF-κB by H. pylori G27 and validated our read-out system for high-throughput investigations. The assay was adapted to a semi-automated screening protocol, to facilitate the analysis of ∼ 3000 Tn7 transposon mutants using wild type G27 and the library-derived mutant G27ΔvacA as NF-κB-activating controls and the G27ΔcagE mutant as a negative control (Fig. S1A and B).
H. pylori mutations affecting NF-κB activation map both within and distal to the cagPAI
A total of 579 candidate mutants passed initial statistical analyses, determined using a modified R-script (see Experimental procedures; Figs S2 and S3), and these mutants were validated in two further screening rounds (Fig. S1A), to yield 59 high-confidence hits in genes that exhibited decreased T4SS-dependent p65 translocation upon transposon mutagenesis. All 59 transposon mutants were sequenced using transposon-specific primers with the aim of identifying the loci of transposon insertions within the H. pylori G27 genome. The vast majority (56) of hits carried the insertion in 16 different cagPAI genes, underscoring the strong T4SS-dependency of NF-κB activation in AGS epithelial cells (Table 1). Infection with all cagPAI mutants led to reduced nuclear translocation of p65 at all time points investigated (Fig. 2A).
The remaining three transposon mutants contained insertions in non-cagPAI genes (Fig. 2B and Table 2). One insertion (tn-57) mapped to a region overlapping two coding sequences for two hypothetical proteins, the second (tn-8) mapped to a lipopolysaccharide (LPS) 1,2-glycosyltransferase and the third (tn-35) mapped to the OMP HopQ. As expected, tn-8 mutant bacteria showed substantially altered LPS (Fig. S4A). Although all three transposon mutants exhibited reduced NF-κB activation (45 min p.i.) and IL-8 secretion (6 h p.i.) (Fig. 2C and D), host cell adherence (Fig. 2E) for all mutants was similar to wild type. Moreover, all mutants retained normal morphology, as shown by electron microscopy, and functioning flagella, as observed from soft agar assays (showing that the ability to form halos was unaffected in mutants) (Fig. S4B and C). These observations suggested that mutations at non-cagPAI H. pylori genomic loci directly affected a T4SS-dependent host cell response.
Table 2. Identified non-cagPAI mutants
Orthologue (Hp 26695)
Outer membrane protein
Flagellar associated protein/Toxin
Flagellar associated protein/Toxin
Association of HopQ polymorphism with cagPAI status
Previous work revealed a degree of ‘co-inheritance’ between hopQ and cagPAI (Salama et al., 2000), and the hopQ gene exists in two allelic forms: alleles I and II. Allele I is generally present in H. pylori strains possessing the cagPAI. However, some strains harbour both alleles (Cao and Cover, 2002). On the basis of this interesting background information we decided to study a possible contribution of HopQ to T4SS-mediated pathogenesis in greater detail. A blast search revealed single hopQ genes in the published genome sequences of strains P12 (Fischer et al., 2010) and G27 (Baltrus et al., 2009). However, the allelic form of hopQ in these two strains had, until now, not been classified. Here, we repeated the sequence analysis of P12 hopQ and identified three nucleotide deviations from the published genome sequence (Fischer et al., 2010); namely, the absence of an adenine (position 1188) and two guanine residues (positions 1141 and 1209). This led to an amendment of the internal reading frame (between amino acid residues 380 and 401; Table S1) with a net loss of one amino acid from the P12 HopQ protein sequence (JF303025). An obvious discordance in a short (21 amino acid) fragment of the published sequence (ACJ08294) with aligned sequences of known HopQ proteins of allele I and II strains (27 and 8 respectively) was thus excluded in our revised P12 HopQ sequence (Table S2). Phylogenetic analysis with MEGA 4.0 software (Tamura et al., 2007) subsequently revealed that the P12 and G27 hopQ genes, as well as their gene products, group with the type I allele (Fig. S5). The P12 and G27 hopQ genes thus constitute genuine type I alleles.
HopQ deficient mutants do not exhibit altered bacterial morphology, cell adherence or motility
Several Hop family members in H. pylori have been implicated in both adherence and phase variable expression (Alm and Trust, 1999; Talarico et al., 2012). Although previous work on four different H. pylori strains (26695, J99, J178 and 87-29) suggested that HopQ can contribute to bacterial adhesion, this activity appears to be strain-specific, since two strains showed no changes in adherence (Loh et al., 2008). In order to determine whether our results might be caused by decreased bacterial adhesion, we analysed the G27-derived HopQ transposon mutant, but this did not reveal any detectable alteration in adherence as compared with wild type bacteria (Fig. 2E). Similarly, H. pylori P12ΔhopQ did not show detectable differences in cell adherence. To further validate this observation, we monitored the binding behaviour of P12ΔhopQ and isogenic wild type to AGS cells at different time points of infection (30 and 90 min p.i., MOI 50) using immunofluorescence microscopy (Figs 3A, B and S6), and we determined the colony-forming units (cfu) of adherent bacteria (Fig. 3C). This analysis clearly demonstrated that the HopQ defect did not alter the number of adherent P12 or G27 bacteria on the human epithelial cells. Nonetheless, this result does not exclude a likely function of HopQ as an adhesin, since adhesin phenotypes could be masked in strains expressing several adhesins at a time (Robertson and Meyer, 1992; Yamaoka et al., 2002b).
To further exclude that altered NF-κB activation was due to impaired bacterial motility or differences in bacterial morphology, we performed soft agar motility assays and transmission electron microscopy using the P12ΔhopQ strain. The results of these analyses confirmed the transposon mutant observations that HopQ deficient bacteria exhibit similar motility and morphology (bacterial cell shape and surface structure) to wild type strains (Figs S7 and S8). Thus, our data indicate that mechanisms other than decreased bacterial attachment and/or altered motility/morphology are responsible for the reduced capacity of hopQ-mutants to activate NF-κB.
The hopQ defect broadly impairs activation of cag-T4SS-dependent host cell signalling routes
A hallmark of H. pylori infections is the induction of a strong pro-inflammatory host cell response via the activation of two major parallel signalling routes: the NF-κB and MAP kinase pathways (reviewed in Backert and Naumann, 2010). To substantiate the observed effect of HopQ on T4SS-dependent NF-κB activation, we extended the time-course experiments of AGSSIB02 cell infection using the hopQ deletion mutant (P12ΔhopQ), and the wild type P12 and P12ΔcagPAI congenic strains as positive and negative controls respectively. Cells were infected at a multiplicity of infection (MOI) of 100 for up to 225 min. Wild-type P12 infection induced p65-GFP nuclear translocation from 30 min p.i., reaching a peak at 45 min p.i. while no p65-GFP nuclear translocation was observed upon infection with the mutant P12ΔcagPAI strain (Fig. 4A), which was consistent with previously described T4SS-dependency of NF-κB activation (Schweitzer et al., 2010). Interestingly, infection with the P12ΔhopQ mutant caused a ∼ 2.5-fold decrease in p65-GFP nuclear translocation compared with wild type. Additionally, the peak of activation was delayed, shifting to 75 min p.i. (Fig. 4A). In order to confirm the HopQ dependency of T4SS-mediated NF-κB activation, we complemented the P12ΔhopQ knock-out mutant with an intact hopQ gene. This restored the p65 translocation phenotype in infected cells (Fig. S9) corroborating the notion that HopQ is functionally linked with the T4SS.
We then investigated the activation of the extracellular signal-regulated kinases 1 and 2 (ERK1/2), p38, and c-Jun N-terminal kinase 1/2 (JNK1/2) (Keates et al., 1999) by determining the phosphorylation of respective pathway components. Infection with the wild type strain resulted in phosphorylation of ERK1/2, p38 and JNK1/2, which reached its maximum peak between 30 and 60 min p.i. (Fig. 4B). By contrast, infection with the P12ΔcagPAI strain resulted in marked attenuation of JNK1/2 and p38 activation and decreased as well as delayed ERK1/2 activation (Fig. 4B). These findings demonstrate that HopQ also facilitates T4SS-dependent MAPK pathway activation.
Both MAPK and NF-κB pathways culminate in the activation of several pro-inflammatory target genes, including the IL-8 gene (Aihara et al., 1997). Hence, we determined the potential of P12ΔhopQ to stimulate IL-8 secretion in AGS cells compared with P12 wild type and P12ΔcagPAI. IL-8 levels were monitored in cell culture supernatants 6 h p.i. by ELISA. Consistent with the results obtained with transposon mutant-35 (Fig. 2D), IL-8 secretion in AGS cells was ∼ 2-fold lower following infection with the mutant P12ΔhopQ compared with P12 wild type (Fig. 4C). This observation correlated with reduced levels of NF-κB activation upon infection with the P12ΔhopQ mutant strain. Taken together, the isogenic deletion mutant P12ΔhopQ stimulated a reduced T4SS-dependent inflammatory host cell response, which ranged between the levels of the wild type strain P12, and the congenic P12ΔcagPAI T4SS mutant.
HopQ is essential for CagA translocation
A prominent function of the cagPAI-encoded T4SS is the translocation of the effector CagA into gastric epithelial cells, resulting in the activation of several cellular responses (Censini et al., 1996; Tomb et al., 1997; Keates et al., 1999). One striking phenotype of CagA is its effect on cell morphology of AGS gastric epithelial cells (hummingbird phenotype) and cell motility (Segal et al., 1999; Stein et al., 2002; Moese et al., 2004). Immunofluorescence microscopy revealed hummingbird formation in AGS cells only when infected with the CagA-expressing P12 wild type strain, but no notable changes of cell morphology occurred during infection with either P12ΔhopQ or P12ΔcagPAI (MOI 100; 6 h p.i.; Fig. 5A). Consistently, phase contrast microscopy revealed increased cell scattering, indicative of motility, after infection with wild type P12, whereas minimal scattering occurred after infection with either hopQ or cagPAI mutant strains (Fig. 5B). To assess the ability of the P12ΔhopQ strain to translocate CagA into infected host cells, we determined CagA phosphorylation as a marker for its translocation (Odenbreit et al., 2000; Stein et al., 2000). In accordance with our phenotypic analyses, no, or very minor levels of, CagA phosphorylation was observed in P12ΔhopQ infected AGS cells, although the hopQ mutant expressed similar levels of CagA protein as wild type bacteria (Fig. 5C). As depicted in Fig. 5C, genetic complementation of the mutant hopQ restored the ability of H. pylori to translocate CagA. Furthermore, in an equivalent fashion, we also complemented the P12ΔhopQ mutant with an intact hopQ gene from strain Tx30a that contains hopQ allele II (Fig. 5D). The result suggests that hopQ allele II, too, can functionally restore P12ΔhopQ.
To address the question whether hopQ deletion can affect T4SS assembly we performed immunostaining experiments of CagY, one of the major proteins of the T4SS which is proposed to bind the T4SS-specific host cell receptor β1α5 integrin and therefore represents an important bacterial surface-located factor for T4SS function (Tanaka et al., 2003; Jiménez-Soto et al., 2009). As expected, CagY expression was only detectable in bacteria carrying the intact cagPAI. Both the wild type H. pylori P12 and P12ΔhopQ strains CagY revealed a similar patchy pattern of staining, reminiscent of local accumulation of CagY on the bacterial cell surfaces (Fig. 5E). This suggests T4SS formation was not fully distorted in hopQ-deleted bacteria and supports the notion of an accessory, rather than a central structural function of HopQ in the activation of the H. pylori T4SS (Fig. 5D).
The cagPAI-encoded T4SS is a major H. pylori virulence determinant (Tomb et al., 1997; Odenbreit et al., 1999; Alm et al., 2000). Its function has been implicated in severity of disease and increased risk of gastric cancer (Hatakeyama and Brzozowski, 2006). One key effector role of the T4SS is translocation of the CagA protein, which is phosphorylated upon arrival in the host cells and triggers the activation of host cell signalling pathways. Major outcomes include the suppression of innate defence mechanisms (Bauer et al., 2012) and putatively oncogenic events involving cytoskeletal rearrangements (Segal et al., 1999; Odenbreit et al., 2000). However, even in the absence of CagA, the T4SS can activate pro-inflammatory signalling events in infected cells (reviewed in Backert and Naumann, 2010). T4SS-mediated delivery of bacterial peptidoglycan into host cells could be a mechanism of T4SS-dependent, but CagA-independent, activation of the pro-inflammatory response (Viala et al., 2004). Recent studies have also suggested a role for β1 integrins in docking of the T4SS to host cells (Kwok et al., 2007; Jiménez-Soto et al., 2009), an interaction that could mediate downstream perturbation of host cell signalling, as well as integrin linked kinase (ILK)/integrin β5 signalling complex, which could play a role in the activation of pro-inflammatory cascades in a T4SS-dependent, CagA-independent manner (Wiedemann et al., 2012). The proposed involvement of peptidoglycan and prominent host cell receptors did not rule out the existence of other potential, as yet unknown bacterial effectors or T4SS-associated determinants, which could function in pathogen-induced host cell responses. This possibility is supported by studies showing that pathways other than NOD-1 are involved in H. pylori-induced activation of, e.g. NF-κB and miR-155 (Maeda et al., 2000; Neumann et al., 2006; Backert and Naumann, 2010; Koch et al., 2012).
Here, we screened an H. pylori transposon mutant library for putative factors that modulate the pathogen-induced pro-inflammatory response in host cells using nuclear translocation of GFP-coupled p65 in human AGS epithelial cells as a measure of NF-κB activation. We identified 19 genes whose products influenced T4SS-dependent NF-κB activation and IL-8 secretion. The majority of transposon insertions that substantially reduced NF-κB activation occurred in 16 distinct cagPAI genes, thus confirming the central importance of the cagPAI for NF-κB activation. Interestingly, five cagPAI genes of H. pylori G27 (HPG27_479, HPG27_480, HPG27_485, HPG27_502, HPG27_504), previously described as having modest effects on H. pylori-induced IL-8 secretion (Fischer et al., 2001), were identified here as strong contributors to T4SS-dependent NF-κB activation. This discrepancy could be due to strain-specific differences (such as H. pylori 26695 vs G27) or transposon-based polar effects on neighbouring genes. Of the three mutants carrying defects in non-cagPAI genomic regions in the H. pylori G27 genome, we could not clearly assign functions for two of them, being affected in hypothetical and LPS genes respectively. Instead, we focused our analysis on the most intriguing hit from our screen, HopQ, for which a functional association with T4SS has already been reported (Cao and Cover, 2002). Our studies thus extend current knowledge on the existence of T4SS accessory factors that appear to be intricately associated with H. pylori pathogenicity, including flagellar motility (Asakura et al., 2010), bacterial attachment via OipA (Dossumbekova et al., 2006), and the regulation of cagPAI gene expression (Hp1451, Hare et al., 2007).
Depletion of HopQ substantially reduced both the induction of cagPAI-associated pro-inflammatory signalling and the translocation of the pathogen's main effector protein CagA, even though this protein and CagY, a structural component of T4SS, appeared to be properly produced and positioned in the mutant. While the pro-inflammatory host cell response to H. pylori infection culminates in the production of cytokines such as IL-8, the precise role of CagA in IL-8 induction remains controversial. Whereas CagA has been shown to induce IL-8 expression in a strain-dependent fashion (Brandt et al., 2005; Argent et al., 2008) which stimulates only minor increases in IL-8 expression (Ritter et al., 2010), NF-κB activation has been demonstrated to be independent of CagA (Schweitzer et al., 2010). Here, we found that the depletion of HopQ reduced both pro-inflammatory signalling and CagA translocation, thus indicating that HopQ broadly influenced T4SS activation. Moreover, in agreement with defective CagA translocation in the HopQ mutant, there was no development of the characteristic host cell phenotype resulting from CagA translocation: host cell elongation, ‘hummingbird phenotype’, and increased host cell motility (Segal et al., 1999; Higashi et al., 2002; Tsutsumi et al., 2003; Suzuki et al., 2005). At the same time, weak and delayed activation of MAPK p38 and JNK1/2, NF-κB and IL-8 secretion observed after infection with the P12ΔhopQ mutant could be facilitated by bacterial factors in a T4SS-dependent, but CagA-independent manner, as mentioned above. The mild host cell responses induced by the hopQ-deletion mutant could not be explained by any decrease in the number of adherent bacteria or motility because these respective measurements were comparable to those of the parental strain. Finally, genetic complementation of the mutant P12ΔhopQ restored the wild type infection phenotype indicating that deletion of the hopQ gene was responsible for the observed phenotype.
Both H. pylori strains (P12 and G27) investigated carried the hopQ allele I, previously found to be ‘co-inherited’ with cagPAI, which is contained in highly pathogenic H. pylori strains (Salama et al., 2000; Cao and Cover, 2002). Our current observations support the notion that a functional link exists between HopQ and the T4SS and nurture speculations that allelic divergence of HopQ could have an impact on T4SS function. However, our preliminary studies on the phenotypic rescue of the P12ΔhopQ mutant with hopQ allele II from the H. pylori Tx30 did not support this hypothesis in the infection model that we have used.
Several previous observations are in line with the idea of HopQ being an adhesive factor. The hopQ gene exhibits pronounced homology to several well characterized adhesion genes in the H. pylori genome (Alm et al., 2000). The sabA and sabB genes, known to encode prominent sialyl-Lewis specific adhesins (Mahdavi et al., 2002), and hopQ further are part of a combinatorial family that undergoes gene conversion in the genome (Talarico et al., 2012). Both these facts strongly suggest a functional relationship between the three factors. Moreover, the structure of H. pylori Hops is reminiscent of the large family of autotransporter proteins of Gram-negative bacteria, the majority of which represent adhesins (Henderson and Nataro, 2001). Intriguingly, a subclass of autotransporters occurs as trimers (Cotter et al., 2005), which could, in principle, allow the formation of hetero-complexes. Whether SabA, SabB and HopQ are involved in such complex formation is currently unknown, but it could provide a rationale even for increased adherence upon HopQ depletion from hetero-complexes (Loh et al., 2008). Therefore, our findings are consistent with the notion that HopQ exhibits adhesive properties, but has no net effect on the number of bacteria attaching to host cells. However, a site-specific adherence property of HopQ could be instrumental in conferring contacts of H. pylori's T4SS to the host cell surface, providing a tantalizing explanation for the observed role of HopQ in T4SS function.
Bacterial strains, cell culture and cultivation
Helicobacter pylori wild type strains G27 and P12 (MPIIB strain collection numbers P409 and HP8880), and the mutants P12ΔcagPAI (Backert et al., 2001; Kwok et al., 2002), G27ΔflaA and P12ΔhopQ were grown on horse serum agar plates containing vancomycin (10 μg ml−1), trimethoprim (5 μg ml−1), and nystatin (1 μg ml−1) under microaerobic conditions (85% N2, 10% CO2, 5% O2) at 37°C. The mutant P12ΔcagPAI was grown on agar plates supplemented with kanamycin (8 μg ml−1); for cultivation of P12ΔhopQ, G27ΔflaA and G27-derived transposon mutants including G27ΔvacA, agar medium was supplemented with chloramphenicol (4 μg ml−1). Escherichia coli strain DH5α (Invitrogen) was grown on Luria–Bertani (Sambrook and Russel, 2001) agar plates or in liquid medium supplemented with ampicillin (100 μg ml−1) or chloramphenicol (40 μg ml−1). Conjugation experiments were performed using E. coli strain β2155 (Dehio and Meyer, 1997), grown on the same media supplemented with diaminopimelic acid (0.2 mM). AGS cells (human gastric adenocarcinoma cell line, ATCC CRL-1739) and AGSSIB02 cells were maintained in RPMI-1640 medium (Gibco) supplemented with 10% fetal calf serum (FCS) in a humidified 5% CO2 atmosphere at 37°C.
Screening of H. pylori G27 Tn7-mutant library and p65 translocation assay of individual H. pylori P12 strains
Preparation of bacterial inocula
An H. pylori G27 mini-Tn-7 transposon (Tn7) mutant library comprising c. 10 000 clones (Salama et al., 2004) was used for the screening assay. H. pylori G27 wild type and the transposon mutant G27ΔvacA2 were used as positive controls, whereas the isogenic deletion mutant G27ΔcagE was used as negative control. The Tn7 library and control strains were grown on GC agar under microaerophilic conditions at 37°C. GC agar plates used for library incubation were supplemented with chroramphenicol (4 μg ml−1). After 2 days, 96 single clones per plate (containing 3 different controls) were picked and transferred to new plates containing no chloramphenicol (Fig. S1). Colonies were incubated again for 2 days under microaerophilic conditions at 37°C. Single clones were resuspended in 100 μl BHI broth containing 10% FCS in a 96-well tissue culture plate (TPP), followed by incubation for 1 h at 37°C under microaerophilic conditions.
Automated infection and p65 translocation assay
The monoclonal p65-GFP expressing cell line (AGSSIB02) (Bartfeld et al., 2009) was used for automated infection. Cells were seeded in 100 μl RPMI containing 10% FCS into 96-well plates 1 day prior to activation to obtain a confluency of approximately 70%. Bacteria (25 μl) were added at each time point by a Biomek FP Workstation (Fig. S1). Infection was stopped 45 min, 120 min and 180 min p.i and cells were fixed with 4% paraformaldehyde, stained with Hoechst 33342 (2 μg ml−1) and stored in PBS until analysis. In each well, four pictures were taken by the automated microscopy system Scan∧R (Olympus) using autofocus on nuclei. Automated image analysis to quantify translocation of p65-GFP was performed as described previously (Bartfeld et al., 2010). Briefly, activated cells were assigned by quantification of cells showing nuclear p65-GFP using Scan∧R image analysis software (Olympus).
Graphs were compiled using Microsoft Excel and Graph-Pad Prism. R-Script based analyses were performed to calculate activated cells (percent), NF-κB activation intensity per well, Z-score and Z′factor (Fig. S2). The Z-score expresses the divergence of the experimental result x from the most probable result μ as a number of standard deviations δ.
The Z′factor is a measure of statistical effect size, and is used in high-throughput screening to judge whether the response measured warrants further analysis (Zhang et al., 1999). Z-scores were calculated for controls and mutants (Fig. S2). Hits were then classified as activating or inhibitory (−0.5 and −6 respectively) (Fig. S3) and further validated by two additional screening rounds.
p65 translocation assay after infection with individual H. pylori P12 strains
Generally, this was performed as above using parental strain P12, mutants P12ΔhopQ and P12ΔcagPAI at MOI 100.
Identification of transposon mutation loci
Genomic bacterial DNA was isolated using the DNeasy tissue kit (Qiagen, Germany) following the manufacturer's instructions. Transposon loci were identified by DNA sequencing using the Primer S2 (5′-CAGTTCCCAACTATTTTGTCC-3′; Salama et al., 2004). Sequences were analysed by BLASTN search against the NCBI non-redundant (nr) database (http://blast.ncbi.nlm.nih.gov/Blast.cgi) and exact transposon loci were determined by in silico genome analysis of identified gene (Fig. S2B).
Deletion and complementation of hopQ gene
The P12ΔhopQ mutant was constructed by replacing the hopQ gene with a chloramphenicol resistance gene cassette (CmR, 1 kb), which was isolated from the plasmid pDH80 using endonuclease BamH I. Briefly, DNA fragments of c. 0.5 kb in length were amplified from chromosome of the parental strain P12 using following primer set (synthesized by TIB Molbiol, Germany) carrying BamH I restriction sites (underlined): (i) hp1141f (5′-CTCGGTGGCAATGTTAGTGTTGTGTTCTTT-3′) and hp1142r (5′-GACTGGATCCGGCTTTATAGCGTGTATCTC-3′) for upstream, and (ii) hp1142iGf (5′-GACTGGATCCAGATGTTCCTTAAAGTAATG-3′) and hp1143r (5′-CTTATGCTCAGTCTCAGATCACTTAATCAC-3′) for downstream DNA fragments flanking hopQ gene respectively. The amplicons were subsequently cloned into a pGEM-T Easy vector (Promega, Germany) and separated by the CmR gene at the BamH I restriction site. Genetic complementation of isogenic H. pylori mutant P12ΔhopQ was performed using a derivative of the E. coli-H. pylori shuttle plasmid pHel3, designated pIB6 (R. Haas et al., in preparation). For that, the hopQ gene was amplified from chromosome of the H. pylori strain P12 using the forward primer CE68 5′-GATCGTCGACATGAAAAAAACGAAAAAAAC-3′ containing a Sal I restriction site (underlined) and the reverse primer CE69 5′-GATCAGATCTTTTAATACGCGAACACATAA-3′ containing a Bgl II restriction site (underlined). The purified PCR product was directly digested with Sal I and Bgl II and ligated into the vector pIB6 at the corresponding restriction sites that resulted in plasmid pCE4. Conjugation of these shuttle vectors, carrying the hopQ gene, from E. coli β2155 to H. pylori strain P76 was performed under previously described mating conditions (Heuermann and Haas, 1998). Since growth of strain β2155 is dependent on exogenously supplied diaminopimelic acid, inoculation of serum plates without diaminopimelic acid is a strong counter selection against the donor strain. Shuttle vector was re-isolated from H. pylori strain P76 using Wizard Plus SV Minipreps Kit (Promega, USA). Purified shuttle vector and plasmid carrying the construct for hopQ deletion were introduced to H. pylori P12 strain by natural transformation (Haas et al., 1993). HopQ gene knockout was confirmed by PCR amplification using primers s1141f (5′-TGGTGATAAAGGTCGTTAAACCCGC-3′) and s1143r (5′-GCCTTAGCGTTTAGTTCCATCGCCG-3′) spanning a DNA fragment of c. 3 kb in length up- and downstream of sites targeted by primers hp1141f and hp1143r. PCR products were purified with the QIAquick Gel Extraction Kit (Qiagen) and were subjected to restriction analysis with BamH I endonuclease; additionally, amplicons were sequenced using the s1141f and s1143r primers.
Genetic complementation of isogenic H. pylori mutant P12ΔhopQ with hopQ allele II was performed using hopQ amplified from the strain Tx30a with primers CE68 (as above) CE62 (ACTTTTTACAACCAGCCAG), CE67 (GATCGTCGACATGTACCCATACGATGTTCCAGATTACGCTATGAAAAAAACGAAAAAAAC) containing a Sal I restriction site (underlined) and an HA tag (bold), and the reverse primer CE70 (GATCAGATCTTTTAATAGGCAAACACATAA) containing a Bgl II restriction site (underlined). Further procedures were similar to those described above.
For infection experiments, AGS cells were grown in 12-well tissue culture plates for 1–2 days to reach 70–80% confluence. Further, AGS cells were incubated in serum-free RPMI medium for 3 h before infection. Bacteria harvested from agar plates were washed, resuspended in RPMI medium and added to AGS cells at an assay-specific multiplicity of infection (MOI). Infection was carried out at 37°C in a humidified atmosphere supplemented with 5% CO2. Samples were taken 30, 60, 90, 180 min and 6 h p.i.
Adherence assay, cfu quantification, immunofluorescence and phase-contrast microscopy
The adherence of the H. pylori parental strain P12 and the P12ΔhopQ mutant was analysed by counting cfu on agar plates and by direct enumeration of bacterial cells using immunofluorescence microscopy. Infections with P12 and P12ΔhopQ were performed at an MOI of 50 for 30 and 90 min. Adherent H. pylori strains were recovered from AGS cells by dispersing in RPMI medium supplemented with 0.5% saponin (Serva) as described previously (Wunder et al., 2006). Serial dilutions of recovered bacteria were plated onto horse serum agar plates to determine cfu. For immunofluorescence microscopy, AGS cells were grown on coverslips under conditions described above. After infection, AGS cells with adherent H. pylori were fixed in 4% (w/v) paraformaldehyde (PFA) solution in phosphate-buffered saline (PBS) (137 mM NaCl; 2.7 mM KCl; 10 mM Na2HPO4; 2 mM KH2PO4) for 20 min at room temperature (RT) followed by three washes with PBS.
Samples were incubated in blocking buffer without detergent containing 3% cold water fish gelatine, 4% donkey serum, 1% (w/v) bovine serum albumin (BSA) and 0,2% BSAc for 30 min at 37°C to block non-specific binding. For CagY visualization, bacteria were incubated with CagY-specific antibody, followed by washing three times with PBS and incubation with goat anti-rabbit Cy3-linked (Dianova) secondary antibody. Cells were also stained with Phalloidin 488 and DRAQ5 (Enzo Life Sciences), washed three times with PBS and mounted on slides with mowiol. H. pylori cells were detected, by incubating in 1% (w/v) BSA in PBS for 1 h at RT to block non-specific binding followed by incubation with primary rabbit anti-H. pylori antibody (Biomol, Germany), followed by three washes in PBS and incubation with secondary goat anti-rabbit Cy2-linked antibody (Jackson ImmunoResearch, UK), for 1 h at RT. Cells were also stained with DRAQ5 (Enzo Life Sciences, Germany), washed three times with PBS and mounted on slides with mowiol. A monoclonal mouse anti-p65 antibody (Santa Cruz, USA) was used to monitor nuclear p65 translocation of AGS cells infected with G27.
For the calculation of adherence properties in total, images of at least 30 random fields of view from three independent experiments were acquired using a Leica TCS SP laser confocal microscope (Leica Microsystems, Germany). AGS cells were counted manually from obtained images. Adherent H. pylori cells were quantified using ImageJ 1.37 software (http://rsbweb.nih.gov/ij/) with thresholds manually set for individual images and circularity 0.05–1.00; each count was verified manually. Parameters for bacterial counts were established in preliminary experiments using a pure culture of H. pylori (data not shown). Statistical analysis of bacterial counts was then performed by t-test. Significance was set at P < 0.05.
Hummingbird phenotype and cell scattering
Development of the hummingbird phenotype in AGS cells was assessed upon infection with P12, P12ΔhopQ and P12ΔcagPAI at MOI of 100. Samples were analysed after 0, 2, 4 and 6 h of infection. AGS cells were fixed as described above, treated with 0.2% Triton X-100 in PBS for 20 min at RT and incubated with Alexa Fluor® 546 (A546) phalloidin (Invitrogen, Germany) and DRAQ5 to stain actin and DNA respectively. Cells were visualized using a Leica TCS SP laser confocal microscope. For scattering assay, AGS cells were infected as above described and their motility was assessed by phase-contrast microscopy using an inverted microscope (model IX50-S8F; Olympus) at the same time points.
H. pylori motility assay
The motility of H. pylori parental strains G27 and P12, mutants P12ΔhopQ, G27ΔflaA and four G27 transposon mutants was assessed using soft agar plates, as described previously (Worku et al., 2004), with minor modifications. Briefly, approximately 109H. pylori cells in a final volume of 3 μl were spotted onto horse serum soft agar plates containing agar (0.5%) and FCS (10%). Plates were incubated for three days under microaerobic conditions. The diameters of the halos were measured by using a ruler.
Cells were lysed in 1× Laemmli buffer (60 mM Tris (pH 6.8), 2% SDS, 10% glycerol, 0.1% bromophenol blue). Proteins (25 μg) were separated by SDS-polyacrylamide gel (10%) electrophoresis and then transferred onto PVDF membranes (Perkin-Elmer) in buffer (25 mM Tris, 192 mM glycine, 20% methanol). Non-specific binding sites were blocked by incubation in blocking buffer [0.05 M Tris (pH 7.4), 0.2 M NaCl, 3% Tween, 3% BSA] for 1 h at RT before probing with primary antibodies against phospho-JNK1/2, phospho-p38 (Cell Signaling Technology, Germany), phospho-ERK1/2, β-actin, (Sigma, Germany), phospho-Tyr (PY99 and CagA, Santa Cruz Biotechnology, Germany; Odenbreit et al., 2000), all at a final dilution of 1:1000 in blocking buffer supplemented with 3% BSA. Secondary HRP-labelled goat anti-rabbit or goat anti-mouse antibodies (GE Healthcare, Germany) were used at a dilution of 1:3000 in the same buffer. The blots were detected on the X-ray film X-OMAT (Kodak, USA) after treatment with enhanced chemiluminescence reagent (ECL) (Pierce, USA). Band intensities were quantified from processed images using Photoshop CS3 (Adobe) and normalized to respective loading controls.
Interleukin 8 production by AGS cells was assessed following infection with H. pylori strains (MOI 100, 6 h and 8 h respectively). IL-8 was measured in the supernatants by ELISA using a sandwich ELISA kit (R&D Systems, USA), according to the manufacturer's instructions. Cytokine secretion was measured in at least three independent experiments carried out in triplicate.
Scanning electron microscopy
Fresh cultures of H. pylori wild type strain P12 and the isogenic mutant P12ΔhopQ, were harvested from agar plates, washed twice in PBS and then fixed in 2.5% glutaraldehyde /PBS solution for 20 min at RT. H. pylori wild type strain G27 and the transposon mutants (Tn-8, Tn-35, Tn-57) were inoculated in BHI broth (OD550: 0.2) and incubated for 12 h in a shaking incubator at 37°C (110 r.p.m.). Samples were washed once in PBS and fixed in 2.5% glutaraldehyde/PBS by RT for 30 min. After washing again in PBS and water, samples were postfixed in 0.5% Osmiumtetroxide, contrasted with 1% tannic acid, and incubated in 0.5% OsO4, with intermediate washing. Samples were then dehydrated in a graded ethanol series and subjected to critical point drying followed by coating with a layer of 10 nm carbon. The samples were analysed with Leo 1550 scanning and transmission electron microscope.
In silico analyses of hopQ gene and HopQ protein sequences
The hopQ gene and protein sequences were downloaded from the NCBI database. The hopQ gene of H. pylori strain P12 was fully sequenced using the primers 1142-F (5′-AAAAGCTCATCTTGTAAGGTAATTTT-3′) and 1142-R (5′-TTTTTTGCAATAAAACATTACTTT-3′), and with OP5136c (5′-CAACGATAATGGCACAAACT) and OP4829d (5′-GTCGTATCAATAACAGAAGTTG) (Cao and Cover, 2002). In silico translation of the sequenced hopQ and multiple alignments of nucleotide and protein sequences were carried out with BioEdit software (Hall, 1999). Phylogenetic analysis was performed with the MEGA 4.0 software (Tamura et al., 2007) using the neighbour-joining method and the Kimura 2-parameter model for gene sequences and Poisson correction for protein sequences. The sequence of P12 hopQ obtained from this work was deposited in GenBank/EMBL/DDBJ under accession numbers JF303025.
Quantification of Western blots
Western blot quantification of band intensities was performed directly from digital pictures which were processed using the computer software Photoshop CS3 (Adobe). Pictures were inverted and region of interest was defined. Histograms of selected regions displayed median of intensity and region size. All bands were normalized to background regions and respective band intensities of loading control.
The authors gratefully thank Drs Lesley Ogilvie, Kate Holden-Dye and Rike Zietlow for excellent editorial assistance on this manuscript. We also thank David Manntz, for the exceptional technical assistance at the Biomek FP Workstation and Dr André Maeurer for his obliging technical support at the automated microscopy system Scan∧R and Meike Soerensen, Dagmar Frahm, Joerg Angermann and Britta Laube for general technical support. This work was supported by research grants of the DFG (HA2697/10-1) to R. H. and funding under the Sixth Research Framework Programme of the European Union, Project INCA (LSHC-CT-2005–018704) and BMBF (ERA-NET PathoGenoMics – FunGen GA no. 0315435A) to T. F. M. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.