Helicobacter pylori Type IV Secretion Apparatus Exploits β1 Integrin in a Novel RGD-Independent Manner

Translocation of the Helicobacter pylori (Hp) cytotoxin-associated gene A (CagA) effector protein via the cag-Type IV Secretion System (T4SS) into host cells is a major risk factor for severe gastric diseases, including gastric cancer. However, the mechanism of translocation and the requirements from the host cell for that event are not well understood. The T4SS consists of inner- and outer membrane-spanning Cag protein complexes and a surface-located pilus. Previously an arginine-glycine-aspartate (RGD)-dependent typical integrin/ligand type interaction of CagL with α5β1 integrin was reported to be essential for CagA translocation. Here we report a specific binding of the T4SS-pilus-associated components CagY and the effector protein CagA to the host cell β1 Integrin receptor. Surface plasmon resonance measurements revealed that CagA binding to α5β1 integrin is rather strong (dissociation constant, KD of 0.15 nM), in comparison to the reported RGD-dependent integrin/fibronectin interaction (KD of 15 nM). For CagA translocation the extracellular part of the β1 integrin subunit is necessary, but not its cytoplasmic domain, nor downstream signalling via integrin-linked kinase. A set of β1 integrin-specific monoclonal antibodies directed against various defined β1 integrin epitopes, such as the PSI, the I-like, the EGF or the β-tail domain, were unable to interfere with CagA translocation. However, a specific antibody (9EG7), which stabilises the open active conformation of β1 integrin heterodimers, efficiently blocked CagA translocation. Our data support a novel model in which the cag-T4SS exploits the β1 integrin receptor by an RGD-independent interaction that involves a conformational switch from the open (extended) to the closed (bent) conformation, to initiate effector protein translocation.


Introduction
Infection with the gastric pathogen Helicobacter pylori (Hp) is associated with a spectrum of pathologies, ranging from mild gastritis to peptic ulcers and gastric cancer [1]. However, the molecular mechanisms underlying the development of Hp-associated gastroduodenal diseases are still poorly defined. Two major virulence factors of Hp that have been associated with disease induction are the vacuolating cytotoxin (VacA) and the cytotoxinassociated antigen A (CagA), both of which are delivered into eukaryotic target cells. VacA, a secreted multifunctional protein toxin, induces intracellular vacuoles in epithelial cells, inhibits T lymphocyte proliferation and modulates T cell function [reviewed in 2]. Using the b2 integrin subunit CD18 as a cellular receptor for uptake [3], VacA efficiently down-regulates transcription of several cytokines or chemokines in T cells [4]. CagA, an immunodominant protein of 120-170 kDa, is encoded on the cag pathogenicity island (cag-PAI). The cag-PAI comprises a total of 27 genes, encoding the cag-Type IV Secretion System (T4SS) in Hp [5].
Upon direct contact with gastric epithelial cells, CagA is translocated into host cells via the cag-T4SS [6] and immediately tyrosine-phosphorylated at a variable number of so-called EPIYA motifs by kinases of the Src and c-Abl family [7,8]. CagA interacts with a large set of host proteins in phosphorylation-dependent and -independent ways and is considered as a bacterial oncoprotein that exerts multiple effects on host signal transduction pathways, the cytoskeleton and cellular junctions [reviewed in 9]. A further hallmark of cag-PAI positive Hp strains is their ability to induce the secretion of chemokines upon contact with epithelial cells, such as interleukin-8 (IL-8) [10]. The function of each cag-PAI-encoded component for CagA delivery and IL-8 secretion has been studied by a systematic mutagenesis approach [11]. Translocationcompetent Hp strains harbour membrane protrusions consisting of a central filament, carrying on its surface the cag-PAI encoded proteins CagY (HP0527), a VirB10 homologous protein [12], CagX (HP0528, VirB9-homologue) and CagT (HP0532, VirB7homologue) [13]. Although the ultrastructure and the biochemical composition of these protrusions has not been clarified yet, we would refer to these structures as type IV secretion system pili. Furthermore, the a5b1 integrin heterodimer has recently been identified as a receptor for the Hp pilus-associated adhesin CagL [14].
Integrins represent a family of about 24 different ab heterodimeric receptors that mediate cell-cell, cell-extracellular matrix and cell-pathogen interactions and govern migration and anchorage of almost all kinds of cells. Each of the non-covalently associated subunits contains a large N-terminal extracellular domain, a transmembrane segment and a short C-terminal cytoplasmic tail. Affinity for biological ligands is regulated by inside-out and outside-in signalling. The bent conformation of the integrin heterodimer represents the physiological low-affinity state, whereas inside-out signalling and ligand binding induces a largescale conformational rearrangement, in which the integrin extends from the bent into an extended, open conformation [15]. Hp T4SS-pilus-associated CagL was suggested to bind via its arginineglycine-aspartate (RGD) motif to a5b1 integrin, a process described as essential for CagA translocation and activation of focal adhesion kinase (FAK) and Src kinase [14]. In the present study we show that further components of the cag-T4SS, such as CagY (HP0527) and the effector protein CagA interact with distinct extracellular domains of b1 integrin. These components are located along or at the tip of the T4SS-pilus. We propose a model that suggests conformational changes of the integrin heterodimer as a basis for CagA translocation.

CagA Translocation into Epithelial Cells is Dependent on b1 Integrin Heterodimers
Hp translocates its effector protein CagA via the cag-T4SS into a number of different cell types in vitro [16]. Recently it was shown that CagA translocation is dependent on the interaction of the Hp T4SS-pilus-associated protein CagL, binding in an RGD-dependent way to a5b1 integrin on the host cell [14]. However, nothing is known about the mechanism of CagA translocation and the involvement of other T4SS components in this process. Using a different approach, we also identified b1 integrin as a cellular receptor for the cag-T4SS. Three independent Hp strains (P12, P145, P217) were applied to study host cell requirements for CagA translocation using several human and animal cell lines (data not shown). Of special interest were human promyelocytic leukaemia (HL60) cells, which were fully competent for CagA translocation ( Figure 1A, lanes 1-3 and 7), whereas, HL60 cells differentiated to a granulocyte-like phenotype (dHL60 cells) revealed only a very weak CagA-P signal ( Figure 1A, lanes 4-6 and 8). Thus, the capacity of Hp to translocate CagA varies considerably, even for the same type of cell, dependent on its cellular differentiation stage. Flow cytometry revealed elevated b2, but significantly reduced b1 integrin levels on the surface of dHL60, as compared to HL60 cells ( Figure 1C). We therefore concentrated on b1 integrin as a potential receptor for the T4SS. CagA translocation was completely absent for epithelial (GE11) or fibroblast-like (GD25) b1 integrin knockout cells, but was functional in genetically complemented GE11b or GD25b cells ( Figure 1B) [17]. In agreement with Kwok et al. [14], these data independently confirmed the important finding that b1 integrin is essential for CagA translocation.

CagA, CagY and CagI are b1 Integrin Interaction Partners
CagL is the only protein encoded on the cag-PAI which carries an RGD motif and therefore might be recognized by the a5b1 integrin receptor in a typical integrin/ligand-like fashion. Whereas Kwok et al. [14] specifically concentrated on the CagL/a5b1 integrin interaction, we chose a systematic approach to identify possible T4SS-integrin interactions and applied a yeast two-hybrid (YTH) screen using the GAL4 Matchmaker system (Clontech) (Figure 2A). Since various cell lines expressing different a/b1 integrin combinations proved successful for CagA translocation (data not shown), the b1 subunit was considered as important and the extracellular portion of the human b1 integrin gene was used as bait. As prey for the YTH screen each of the 27 cag-PAI-encoded proteins were assayed [18]. Positive interactions were obtained for the extracellular part of b1 integrin with the Nterminal region of CagA (HP0547 a ), the C-terminal (VirB10homologous) portion of CagY (HP0527 c ) and with CagI (HP0540) (Figure 2A, Figure S1). Similar results were obtained when bait and prey were exchanged (data not shown).
To confirm the YTH data, pulldown experiments using Hp T4SS-associated proteins were performed (see Figure S2A for preparation of extracts) using functional a1b1 and a5b1 integrin heterodimers (Chemicon) coupled to magnetic beads. CagA was specifically pulled down from wt cell lysates by a1b1 or a5b1 integrin beads, but not by controls (Tris-blocked beads) ( Figure 2B). The ectopic expression of cagA from the shuttle plasmid pJP66 [19] in a P12DPAI strain demonstrated that CagA alone is able to interact with b1 integrin without any other component of the cag-PAI. Preferentially the upper band of CagY and only small amount of the lower band of CagY was pulled down by the same procedure ( Figure 2B, lower panel). Again, precipitation of CagY from a cagA-negative Hp background confirmed an interaction of CagY with integrin, independent of CagA.
The putative interaction of b1 integrin with CagI could not yet be verified by pulldown assays, due to the lack of a specific functional antibody against CagI. We therefore used as a further method a cell-based assay to determine binding of the corresponding GST-Cag fusion proteins to b1 integrin-proficient (GE11b), versus b1 integrin-deficient cells (GE11) by flow cytometry ( Figure 2C and Figure S2B,C). GST did not bind b1integrin-dependent, but purified GST-CagA, GST-CagI and GST-CagYc bound significantly more efficiently to GE11b as

Author Summary
Integrins are single transmembrane proteins present on almost all types of cells. They are composed of an a and a b subunit, which together form the ligand binding pocket, able to interact with extracellular matrix proteins. The best known binding domain on integrin ligands is the RGD domain. Many bacterial, but also viral pathogens exploit this ligand-binding domain to interact with integrins on the host cell. Helicobacter pylori, a common bacterial pathogen associated with gastric diseases, was recently added to this list. One of H. pylori's most important factors associated with gastric pathologies is the CagA protein. This protein is directly injected into host cells through the Cag Type IV Secretion System (cag-T4SS). Previous studies demonstrated that the cag-T4SS requires integrins for the injection (translocation) of CagA into cells. We provide evidence that three proteins, CagA, CagI and CagY, interact with integrins in an RGD-independent way. Additionally, our data point out that the Cag apparatus needs the physical capacity of a b1 integrin heterodimer to change from an active/extended conformation to a closed/ bent conformation. This novel kind of integrin interaction opens a new way in which pathogens can use receptors on cells.   Figure S1). (B) For pulldown assays, magnetic beads (SiMAG) were loaded with purified functional a1b1 or a5b1 integrin or Tris buffer (control) and incubated with processed Hp P12 wt or defined mutant strains, as indicated (fraction Soluble II, Figure S2A). Beads were recovered by magnetic forces, washed, boiled and run for SDS-PAGE. Immunoblotting with a-CagA or a-CagY antibody detects precipitated proteins (arrowheads). (C) Quantification of binding of purified GST-CagA, GST-CagYc, GST-CagI or GST-Inv397 (30mg/1610 6 cells) to integrin-deficient GE11 versus integrin-proficient GE11b cells by flow cytometry. Binding is analysed by a-GST antibody (fold increase in binding versus GST, the values for binding of GST alone has been set to 1). RPI, repeat region I, RPII, repeat region II, (*, p,0.05; **, p,0.01; ***, p,0.001; students T-test). MF, mean fluorescence, (n = 20). Oligonucleotides for construction of GST fusion proteins see Table S2 Figure 2C). Interestingly, GST-CagY, but not the other GST fusion proteins, bound more efficiently when the integrins were activated by Mn 2+ . The Yersinia invasin (GST-Inv397) [20], known to specifically bind b1 integrin, showed a similar behaviour in these binding assays as the GST-Cag proteins ( Figure 2C). To exclude that these GST fusion proteins would bind unspecifically to the GE11b cells, we also generated unrelated cag-PAI GST fusion proteins, such as GST-Cagb (HP0524), GST-CagG (HP0542) and GST-CagZ (HP0526), which did not show b1 integrin-specific binding to the cells ( Figure S2B). These data confirmed the b1 integrin-specific binding of CagI from the YTH assay and verified CagI as an additional cag-T4SS component interacting with b1 integrin. Unexpectedly, neither a1b1, nor a5b1 integrin beads specifically pulled down CagL from membrane or soluble fractions of Hp wt cells using our precipitation conditions (data not shown). We therefore also generated GST-CagL and GST-CagL-RAD mutant protein. Although both purified proteins revealed a rather weak interaction to b1 integrin, the binding was completely independent from the RGD motif of CagL ( Figure S2C).
Taken together, these data verified CagA (the translocated effector protein), CagY and CagI as direct interaction partners of different b1 integrin heterodimers. The fact that b1 integrin in combination with different integrin a chains (a1, a5) precipitated the Cag proteins confirmed that these proteins bind to the b1 subunit, rather than the a subunit of the heterodimer and strongly support the YTH results.

Localization of CagY and CagA Along the T4SS Pilus or Tip
To allow binding of CagA, CagY and CagI to the integrin receptor, these T4SS components should be accessible at the surface of the T4SS pilus. CagY is known as an essential component of the membrane-spanning T4SS complex, but in addition its surface-or T4SS pilus-association has also been demonstrated [12,13]. In addition to CagY, we also verified CagA on the pilus by field emission scanning electron microscopy (FESEM) (Protocol S1) ( Figure S3) [12]. Anti-CagA-coupled gold particles preferentially labelled the pilus tip, with one or rarely two gold grains only, but no staining of the pilus base and only rare background staining of the bacterial or eukaryotic cell surfaces was visible ( Figure S3A-I). To investigate a binding of the cag-T4SS pilus to b1 Integrin during the infection process, confocal laser scanning microscopy (CLSM) and life cell imaging were applied. The gfp-expressing Hp P12 wild type (wt) strain, but not the equally well binding P12DPAI mutant strain, showed a rapid colocalization with b1 Integrin upon infection of AGS cells ( Figure 1E). Quantification data revealed that roughly 10% of Hp P12 wt or P12DcagA, but only 2.5% of P12DPAI bacteria colocalized with b1 Integrin ( Figure 1D). A P12DcagA deletion mutant still showed co-localization to b1 integrin, which can be explained by its binding via CagY or CagI.

The RGD Motif in CagL is not Essential for T4SS Function
We wondered why CagL was detected neither in our YTH screen, nor the pulldown assays. To reassess the described RGDdependent interaction of CagL and a5b1 integrin [14], we first generated a defined cagL deletion mutant in Hp P12. The strain was genetically complemented in the Hp recA locus by mutated cagL genes encoding RAD, RGA or DRGD versions of CagL ( Figure 3A). AGS cells infected with the P12DcagL strain failed to translocate CagA and to induce IL-8, as described earlier [11,14], but surprisingly all three distinct Hp cagL mutant strains (cagL-RAD, cagL-RGA and cagLDRGD) behaved identical to the P12 wt strain concerning CagA translocation, as well as IL-8 induction ( Figure 3B). This clearly indicated that under infection conditions an RGD-mediated interaction of CagL with a5b1 integrin either is not existent, or not necessary for CagA translocation, or for IL-8 induction.

Binding of CagA to b1 Integrin with High Affinity
To judge the specificity and the strength of CagA binding to b1 integrin, surface plasmon resonance measurements were performed. The expression of recombinant CagA is problematic because the protein is rapidly degraded [21,22]. We considered this property as we purified a stable N-terminal fragment (100 kDa) of CagA, lacking the C-terminal 33kDa domain [22]. The protein was used for binding studies with purified a5b1 integrin and aVb3 integrin (Clontech) as a negative control. CagA binds with high affinity to a5b1 integrin (K D = 0.153+/ 20.096 nM) ( Figure 3C) and with a 2-log higher K D value to aVb3 integrin (K D = 33.4+/218 nM) ( Figure 3D), demonstrating the avidity and the specificity of CagA for b1 integrin binding. Thus, the K D value of a5b1 and CagA is approximately hundredfold lower as compared to the same integrin with its cognate ligand, fibronectin, which is dependent on an RGD motif (K D = 15 nM) [23]. This strong and specific binding suggests an important function for this interaction.
Integrin Clustering is Essential for CagA Translocation, but not Signalling via the b1 Integrin Cytoplasmic Tail Specific binding of CagA or CagY to the b1 integrin subunit of the heterodimer on the host cell surface might stimulate integrin clustering and internalization. Cholesterol depletion of AGS cells by methyl-b-cyclodextrin strongly reduced CagA translocation in a dose and time-dependent manner ( Table 1). Calpeptin inhibits the Ca 2+ -dependent protease calpain, which is required for the release of integrins from the cytoskeleton and for clustering in lipid rafts. Calpeptin treatment completely abrogated CagA translocation ( Figure S4, Table 1) and together with the methyl-bcyclodextrin data strongly suggested that the organization of b1 integrins into lipid rafts and integrin clustering is essential for CagA translocation, as recently confirmed [24].
To clarify whether b1 integrin-mediated signalling might be necessary for CagA translocation, we used CHO cells stably transfected with either a full-length human b1A gene, a deletion comprising the complete cytoplasmic tail (b1TR) or constructs containing only the transmembrane and the common region of the b1A tail (b1COM) [25]. Surface expression of b1A integrin was verified by flow cytometry (data not shown). With exception of the CHO vector control, all cell lines were competent for CagA translocation by Hp P217 and with lower efficiency in the P12 strain ( Figure 4A). The generally low efficiency of CagA translocation into b1 reconstituted CHO cells might be due to a combination of human b1 integrin with the endogenous hamster a integrin chains. Nevertheless, these data suggest that the cytoplasmic tail of b1 integrin and therefore outside-in signalling via the integrin b1 chain is not essential for CagA translocation, although these data do not exclude that such a signalling occurs under in vivo conditions.
To further substantiate these findings and to prove whether CagA translocation via the T4SS may occur independently of b1 integrin signalling, we applied an integrin linked kinase (ILK) gene knockdown. ILK binds to the b1 integrin cytoplasmic domain, thereby directly coupling outside-in integrin signalling to a variety of downstream signal transduction pathways [26]. Production of the ILK protein was reduced by ,80% at 60h after transfection of the ILK siRNA ( Figure 4B). Interestingly, knockdown of ILK had no effect on the capacity of CagA translocation by any of the three Hp strains used ( Figure 4B), providing strong evidence that CagA translocation solely depends on the extracellular part of b1 integrin and its clustering in lipid rafts, but does not necessarily need the b1 integrin/ILK downstream signalling pathway.
Interference with CagA Translocation b1 integrin heterodimers binding to ligands, such as fibronectin, collagen or laminin, involves the a chain and the b chain [27]. Integrin head domains are able to adopt two alternative conformations, termed open (high affinity) and closed (low affinity), which are modulated via binding of metal ions, such as Ca 2+ (stabilising closed conformation, deactivating), or Mg 2+ or Mn 2+ (stabilising open conformation, activating). [27]. Deactivation of integrins by treatment with Ca 2+ or EDTA did not interfere, but the intracellular Ca 2+ -chelator BAPTA significantly reduced either translocation or tyrosine-phosphorylation of CagA in AGS cells. In contrast, extracellular activation of integrins (Mn 2+ ) significantly enhanced CagA translocation and its tyrosinephosphorylation ( Table 1). Treatment of AGS cells with proteases, such as trypsin or thrombin, which leads to detachment of the cells, but not cleavage of integrin heterodimers, resulted in slightly enhanced, rather than abrogated CagA translocation efficiency (Table 1). This observation might possibly be explained by an indirect activation of b1 integrin through Proteinase-Activated Receptor-2 (PAR-2) [28]. Natural ligands of a5b1 integrin, such as fibronectin, Yersinia enterocolitica invasin [29] or RGD peptides, even in high quantities, did not alter CagA translocation efficiency (Table 1).
We next used a set of specific anti-b1 monoclonal antibodies (mAbs), including stimulatory (N29, 8E3, 12G10, 9EG7, B3B11), inhibitory (JB1A, AIIB2), and neutral ones (K20, LM534), to check for interference with the binding of Cag proteins to b1 integrin and thus eventually block CagA translocation (Table 1). Antibodies targeting different domains of b1 integrin were shown to bind to AGS cells ( Figure 5A,D). These well-defined mAbs cover essentially all domains and conformations of the b1 integrin chain ( Figure 5A), but with the exception of 9EG7, none of them was able to interfere with CagA translocation ( Figure 5B,C and Table 1). According to its function, mAb AIIB2 detached AGS cells from the tissue culture plate. This is due to its interaction with the ligand binding domain and its b1 integrin deactivation, but this binding did not block CagA translocation ( Figure 5B). Thus, our data show a novel type of interaction of the cag-T4SS with b1 integrins, which is independent of the integrin a chain and a typical integrin/ligand interactions, as well as RGD-motifs in any of the Cag proteins.

Locking Integrin in its High Affinity Conformation Blocks CagA Translocation
Several studies have indicated that a close apposition of the a and b subunits in the membrane-proximal region and the so-called bent structure of the heterodimer are characteristic for the low affinity state of integrins [30,31]. In contrast, the extended conformation, characterized by separated legs (comprising the b1 I-EGF1-4/b-tail and the a chain Calf-1/Calf-2 domains), represents the high affinity state [15]. Certain allosteric b1 integrin antibodies are able to modulate integrin activity rather specifically by stabilizing a distinct affinity state of the integrin [32]. 9EG7 is a b1-specific mAb with a binding epitope in the I-EGF2-4 region, which is strongly exposed upon manganese treatment or ligand binding ( Figure 5E) [33]. The mAb 9EG7 stabilizes and probably fixes conformational changes in the integrin heterodimer, which means that 9EG7 binding of activated integrin will no longer allow its inactivation/bending. Interestingly, mAb 9EG7 completely blocked CagA translocation in AGS cells, whereas Mn 2+ alone enhanced, rather than reduced CagA translocation ( Figure 5B). A papain-generated Fab fragment of mAb 9EG7 binds in a Mn 2+ -dependent way to the integrin receptor ( Figure 5F), but is unable to block CagA translocation ( Figure 5C). A direct competition for binding of 9EG7 and GST-Cag proteins to the same integrin epitope was excluded, as measured by FACS analysis ( Figure 5G).
The important function of the integrin activation state was supported by the human cervix cell line HeLa. This cell line produced normal levels of b1 integrin on the cell surface, as determined by flow cytometry (data not shown), but was only very inefficiently, or not at all able to act as host cell for CagA translocation by certain Hp strains ( Figure 6A). An in vitro phosphorylation assay ruled out a possible defect in CagA tyrosine phosphorylation (data not shown). The binding capacity of mAb 9EG7 revealed a 20% difference between non-activated and activated (Mn 2+ ) state of the cells, whereas for AGS cells the difference was approx. 65%. These data suggest that, due to an unknown mechanism, HeLa cells apparently produce constantly activated b1 integrin, which might be locked in the active state unable to switch back to the closed, inactive conformation, similar to the situation obtained by 9EG7 binding. This could explain the limited ability of Hp to translocate CagA.

Discussion
The cag-Type IV Secretion System (cag-T4SS) of Hp constitutes one of the most important virulence factors of this gastric bacterial pathogen. The mechanism by which Hp translocates CagA into host epithelial cells is still not well understood. An important finding was that the cag-T4SS apparently does not inject its effector protein CagA randomly into target cells, but uses the a5b1 integrin as a cellular receptor for the pilus-associated adhesin CagL [14]. CagL is the only cag-PAI encoded protein carrying an RGD sequence, which is present in certain extracellular matrix proteins and known as a typical integrin/ligand interaction motif [34].
In the present study, we describe the mammalian b1 integrin in different combinations with integrin a chains as a receptor for the Hp T4SS. Convincing data for a functional role of b1 integrin were obtained by the promyelocytic HL60 cell line. Nondifferentiated promyelocytic HL60 cells, producing high levels of b1 integrin on their cell surface, but not differentiated dHL60 cells, with low levels of surface-associated b1, translocated CagA very efficiently ( Figure 1A,C). These data were substantiated by using integrin-deficient murine fibroblast (GD25) or epithelial (GE11) cells, which were completely resistant to CagA transloca-  PLoS Pathogens | www.plospathogens.org tion by Hp, but could be functionally restored upon re-expression of the b1 integrin (GD25b, GE11b) ( Figure 1B).
Here, we performed a systematic YTH screen to identify all proteins of the cag-PAI interacting with the b1 integrin receptor. We identified the translocated effector protein CagA, the Cterminal domain of the secretion apparatus component CagY and CagI as binding partners of the integrin receptor. Biochemical evidence for a receptor function of a1b1 (a collagen and laminin receptor) or a5b1 integrin (a fibronectin receptor) was obtained by (i) pulldown experiments using Hp lysates ( Figure 2B), (ii) direct binding of the corresponding GST-Cag fusion proteins to b1 integrin as determined by FACS analysis (Figure 2C), or (iii) by surface plasmon resonance measurements ( Figure 3C,D). CagA binds b1 integrin with a significantly lower K D value as a5b1 integrin binds in a RGD dependent way to fibronectin, its natural ligand. CagA affinity for aVb3 is significantly lower (approx. 100fold) as compared to a5b1, demonstrating the high specificity of CagA for the b1 heterodimer. CagA also binds with significantly higher affinity as postulated for CagL to a5b1 integrin. CagA carries a C-terminal translocation signal, but also the N-terminus is essential [19]. So far, the role of the N-terminal portion of CagA had remained elusive, but this specific binding to b1 could explain its important role in the translocation process. The exceptionally high affinity of CagA for the integrin receptor might compensate for the relative low abundance of CagA at the tip of the cag-T4SS pilus and suggests an important function for the surface-associated CagA. The integrin binding might have a structural role in triggering integrin rearrangements, whereas only the cytoplasmic (non-pilus-associated) CagA might act as the translocated effector molecule when a translocation-competent configuration has been established (see model Figure 7).
Whereas our data support the essential role of b1 integrin for the process of CagA translocation, the RGD-dependent binding of CagL to the integrin receptor could not be verified. Using Hp extracts, we were able to show here a direct interaction of native (non-recombinant) CagA or CagY with integrin heterodimers, however we could not confirm an interaction of native CagL with the a5b1 integrin. It is possible that in the native pilus-associated CagL the RGD motif is buried within the protein and not accessible to interaction. In recombinant overexpression systems, this motif could be exposed, due to partially incorrect folding. Our genetic complementation data support this theory. We are able to successfully rescue CagA translocation with the complementation of CagL mutants, independently of the RGD status of the protein.
Possible failures in complementation are known for Hp due to frequent secondary mutations, often in the cag-PAI [35]. In support of these genetic data we also showed that binding of GST-CagL protein to b1 integrin is very low, as compared to CagA, CagI or CagYc. More important, the binding of purified GST-CagL to b1 integrin was completely independent from its RGD motif. Thus, we show on the functional as well as on the binding level that the RGD motif of CagL is not essential for the protein function.
To further study the type of interaction between the b1 integrin heterodimer and the Cag proteins, we used typical b1 integrin ligands, such as RGD peptide, fibronectin or Yersinia invasin protein, to possibly interfere with CagA translocation ( Table 1). None of these known ligands was able to block CagA translocation, indicating that the Cag proteins use different sites on the integrin for interaction. Kwok et al showed that Escherichia coli strain HB101 that expresses Yersinia invasin inhibited CagA phosphorylation in AGS cells. It might be possible that E. coli expressing invasin on the surface could sterically inhibit H. pylori to bind and translocate CagA, just by blocking the access to the integrins.
Interaction of integrins with its ligands results in integrin clustering. Inhibition of integrin clustering into lipid rafts using methyl-b-cyclodextrin or calpeptin strongly reduced CagA translocation, indicating that Hp-mediated clustering of b1 integrin heterodimers on the cell surface might be essential for this process. To determine whether integrin signalling might play a role for CagA translocation, signalling-deficient, truncated versions of b1 integrin receptor were used. Unexpectedly, neither the b1 integrin cytoplasmic tail, nor signalling via the integrin linked kinase was necessary for CagA translocation, indicating that only the extracellular domains of the b1 integrin is important. CHO cells derived from hamster generally showed a lower CagA translocation efficiency as compared to human AGS cells ( Figure 4A). We assume that human/hamster integrin heterodimers, generated upon transfection are the reason, an effect also seen for murine GE11 cells (human/mouse integrin heterodimers) ( Figure 1B). H. pylori P217 shows a very strong CagA phosphorylation in AGS cells due to its high number of EPIYA motifs (8 motifs as To obtain insight which domains of the extracellular part of the integrin receptor are important, a set of defined monoclonal antibodies against various b1 integrin domains were applied ( Figure 5A, Table 1). None of these antibodies, even those blocking the integrin ligand interaction (AIIB2, 12G10), were able to block CagA translocation, except mAb 9EG7. This is in contrast to many viruses, which use integrins as receptors or co-receptors for entry into different host cells [36][37][38]. Most viruses known to use integrins as entry receptors have been shown to do so by extracellular matrix (ECM) protein mimicry, which means that viral proteins contain an RGD or any other conserved integrin recognition motif. Thus, specific antibodies, which block integrin ligand interaction, usually abrogate virus infection in vitro [36][37][38]. Taken together our data suggest that the cag-T4SS uses the extracellular portion of integrin to mediate entry of the effector protein into the cell by a different mechanism, probably independent from ECM protein mimicry and the usual integrin ligand interaction.
What is the difference in the effect of mAb 9EG7 in comparison to all the other b1 specific mAbs used in this study? 9EG7 binds an epitope in the b1 integrin which is close to the fulcrum at the genu and is buried in the inactive (bent) state of the integrin receptor. Mn 2+ treatment or ligand binding opens the integrin into the extended conformation and the epitope is free for antibody binding. When 9EG7 is bound, the integrin cannot move back into the bent conformation, probably due to sterical problems with the bulky Fc part of the antibody or its ability to crosslink integrin chains. In addition, we cannot exclude the possibility that 9EG7 might prevent the interaction of the b1 integrin subunit with a co-receptor necessary for the translocation process, although there is no evidence for a coreceptor being involved. The 9EG7 Fab fragment still needs activation of the integrin for binding ( Figure 5E,F). This indicates that its binding is unchanged, but the lack of the Fc chain will not cause the effects presumed for the complete mAb, due to the smaller size of the Fab fragment and its inability to crosslink. Our data lead us to propose a model whereby the rearrangement of the integrin from the extended, open conformation, binding the T4SS components, to a bent conformation (bent closed) is an essential step in the process of CagA translocation (see Figure 7 for a model). We propose that this mechanics of the integrin, which is associated with a closer approximation of the integrin head to the cellular membrane [27], brings the pilus closer to the cellular membrane. When this conformational change is inhibited, CagA translocation cannot occur.

Bacterial strains, cell lines and culture conditions
Hp Strains. Hp strains P12, P145 and P217 and the isogenic knockout mutants P12DcagA, P12DcagE, P12DcagY as well as P12DPAI and P217DPAI, lacking the entire cag pathogenicity island, have been constructed as previously described for the corresponding mutant strains in Hp 26695 [11]. Hp strains were grown on GC agar plates (Difco) as described previously [11].
Eukaryotic cell lines. Eleven different eukaryotic cell lines were analyzed for their capacity to tyrosine-phosphorylate Hp translocated CagA and IL-8 induction (Table S1). Cells were grown in media as indicated and subcultured every 2 to 3 days.
Yeast two hybrid assay The Invitrogen system consisting of the entry vector pDONR207 and the destination vectors pDEST-GADT7 (prey vector) and pDEST-GBKT7 (bait vector) were used. Yeast twohybrid bait and prey libraries were generated comprising the external b1 integrin gene sequence and the cag-PAI genes. For the cag-PAI genes, 22 full-length open reading frames (excluding Nterminal signal sequences), and 10 partial open reading frames were amplified from chromosomal DNA of strain 26695 by nested PCR, and cloned in the bait and prey vectors exactly as described [18]. Bait and prey plasmids were transformed into the haploid Saccharomyces cerevisiae strains Y187 and AH109. Diploid yeast cells were selected after mating and selection on SD medium lacking tryptophan (Trp 2 ) and leucine (Leu 2 ), thus generating all possible combinations of bait and prey plasmids. After growth on SD-Trp 2 Leu 2 medium, yeast colonies were transferred to SD-Trp 2 Leu 2 His 2 medium in order to select for interactions. Growth after 3 to 6 days indicated bait-prey interactions. Additionally, the stringency of this screen was enhanced by selection on SD-Trp 2 Leu 2 His 2 medium containing the competitive inhibitor 3-aminotriazole (5 mM).

Phosphorylation assays
Cells were infected with Hp at 70-90% confluency with a multiplicity of infection (MOI) of 60. For synchronization experiments, cells were detached with PBS/2 mM EDTA, seeded and after 24 hours synchronized overnight in serum free media. Before infection, RPMI (GIBCO) complete media (CM) containing 10% Fetal Calf Serum (GIBCO) was added to cells, counting this point as time 0. To test different inhibitors, 1, 2 and 4h infections were performed after 60 min from addition of CM. After infection, supernatants from 2 & 4h experiments were collected, cells were harvested in PBS with protease inhibitors (pepstatin 1mM, leupeptin 1mM, PMSF 1mM) and phosphatase inhibitor sodium vanadate (1mM). Harvested cells were centrifuged at 5006g for 10 min at 4uC and pellets lysed in RIPA buffer with protease and phosphatase inhibitors and DNase I for later SDS-PAGE and immunoblot analysis under non-reducing conditions.

Pull-down assays and immunoblotting
SiMAG magnetic beads (Chemicell) were coated with 50 mg a5b1 integrin/10 mg beads following the manufacturer's instructions. Beads were saturated with 1 M Tris-HCl (pH 7.5). 1 ml Hp (OD 550 of 2) in PBS with protease inhibitors was treated with lysozyme (10 mg/ml, 4 mM EDTA) for 30 min at RT, DNase I was added (1 mg/ml) and bacteria were lysed by ultrasonication on ice. Soluble proteins (Soluble I) and membranes were separated by ultracentrifugation. Membranes were resuspended in 1 ml HSL (High Salt Lysis, 25mM Tris-HCl, pH 7.4, 0.05% Triton-X100, 4 mM MgCl 2 , 3 mM MnCl 2 , 150 mM NaCl) buffer with protease inhibitors, sonicated and centrifuged at 4uC, 13.000 rpm for 1min to collect the soluble fraction (Soluble II). Soluble II was used for protein pulldown. After pulldown, 3ml beads were incubated at 4uC for 1 h, washed 3 times with HSL 400 (HSL with 400 mM NaCl) buffer, boiled and used for SDS-PAGE (non-reducing conditions). Proteins were transferred to PVDF membranes, and blotted with the antisera indicated. Blots were routinely stripped and reprobed with the indicated antisera (a-actin or a-b1 integrin antibodies) as loading controls. Blots shown are representative of three independent experiments.

Live cell imaging
For co-localization experiments the integrin b1-specific monoclonal antibody 4B7 was labelled with AlexaFluor 568 according to the manufacturer's instructions (10 mol Alexa Fluor568/mol antibody). AGS cells were grown in 35 mm glass bottom dishes (MatTek) to 60-70% confluency. Cells were washed once with PBS (without Ca 2+ and Mg 2+ ) and infected with GFP-expressing P12 wt or P12DPAI grown on serum-free media at an MOI of 60. 2 mg/ml Alexa Fluor 568-labelled antibody against b1 integrin was added. Infection was performed for 7 min. at 37uC and PBS was exchanged before microscopy studies. For quantification of colocalization, assays were recorded over a time range of 50s and picture sequences were analyzed for co-localization events of single bacteria and integrin b1 clusters. Percent co-localization was calculated from ratio of bacteria co-localizing with integrin to total adherent bacteria.
Imaging was done using an UltraView LCI spinning disc confocal system (PerkinElmer) fitted on a Nikon Eclipse TE300 microscope equipped with a temperature-and CO 2 -controllable environment chamber. Images were taken with the black/white ORCA ER Camera (Hamamatsu). Pictures were taken and edited using LCI UltraView software. For immunofluorescence assays, an Olympus BX 64 microscope and Cell ' P software were used.

Purification of CagA 100 kDa fragment
CagA gene was cloned into a vector expressing an TEV cleavable His-tag fusion CagA (pHAR3011-CagA) as described in [22]. Briefly, the protein was expressed in BL21 cells induced o/n at 20uC with 1mM isopropyl-b-D-galactopyranoside (IPTG). Harvested cells were lysed by sonication in buffer A (10 mM Na Phosphate pH7.5, 5mM immidazole, 500 mM NaCl, 10% glycerol) containing DNaseI, lysozyme, one mini complete EDTA-free protease inhibitor cocktail tablet (Roche). After centrifugation, the supernatant was loaded onto a His-trap column (GE Healthcare) and eluted with a linear gradient of immidazole (5 to 500 mM) in buffer A. Two major N-terminal fragments (29 kDa and 100kDa, assessed by the His-tag presence) were eluted together with minor degradation products. Fractions were concentrated and the buffer was exchanged to 10mM Tris pH 7.5, 150mM NaCl. The two fragments were separated by gel filtration using Superdex S75 10/300 GL column (GE Healthcare). Fractions containing ,95% pure 100 kDa fragment (residue 1 to approximately 885) were pooled, concentrated and reloaded on the same column to ascertain stability. The protein was finally concentrated to 1.8 mg/ml.

Surface plasmon resonance
Using a BiacoreX unit, the purified 100 kDa N-terminal fragment of CagA was attached to a CM5 chip (BiaCore) using the standard amine coupling procedure. The flow-cell 1 was treated similarly without coupling of a protein and was used as a reference. Integrin binding to the reference was negligible. For the evaluation of the interaction of proteins, a Tris buffer was used containing 24 mM Tris-HCl pH 7.5, 137 mM NaCl and 2.4 mM KCl. Injection of the integrin proteins was for 1 min using a flow of 60 ml/min. Subsequently, dissociation was evaluated for 200 sec. Regeneration of the chip took place between each measurement using a solution of Tris 20 mM pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 and 0.0125% Triton, resulting in a stable baseline and retaining activity. BiaEvaluation software (version 4.1) was used for the evaluation of the dissociation constant using a 1:1 langmuir model of binding.
Generation of 9EG7 Fab fragments 1 mg of 9EG7 mAb (rat) (BD Biosciences) was digested using beads coupled to Papain (Pierce) following the manufacturers' instructions. Fab fragments were detected by western blotting using a rabbit anti rat Fab (Rockland Immunochemicals). Binding capacity of the Fab fragments to AGS cells was evaluated by flow cytometry. Secondary antibodies anti-rat Alexa 488 and anti-rabbit Alexa 488 were from Molecular Probes.

Statistical analysis
Data are presented as mean+/2SEM. Differences between groups were assessed by the paired, two-tailed Student's t-test, or by the Mann-Whitney U test for unpaired groups depending on the data set of concern (see figure legends).