Helicobacter pylori binds human Annexins via Lipopolysaccharide to interfere with Toll-like Receptor 4 signaling

Helicobacter pylori colonizes half of the global population and causes gastritis, peptic ulcer disease or gastric cancer. In this study, we were interested in human annexin (ANX), which comprises a protein family with diverse and partly unknown physiological functions, but with a potential role in microbial infections and possible involvement in gastric cancer. We demonstrate here for the first time that H. pylori is able to specifically bind ANXs. Binding studies with purified H. pylori LPS and specific H. pylori LPS mutant strains indicated binding of ANXA5 to lipid A, which was dependent on the lipid A phosphorylation status. Remarkably, ANXA5 binding almost completely inhibited LPS-mediated Toll-like receptor 4- (TLR4) signaling in a TLR4-specific reporter cell line. Furthermore, the interaction is relevant for gastric colonization, as a mouse-adapted H. pylori increased its ANXA5 binding capacity after gastric passage and its ANXA5 incubation in vitro interfered with TLR4 signaling. Moreover, both ANXA2 and ANXA5 levels were upregulated in H. pylori-infected human gastric tissue, and H. pylori can be found in close association with ANXs in the human stomach. Furthermore, an inhibitory effect of ANXA5 binding for CagA translocation could be confirmed. Taken together, our results highlight an adaptive ability of H. pylori to interact with the host cell factor ANX potentially dampening innate immune recognition.


Introduction
Infection with Helicobacter pylori (H. pylori) is responsible for chronic gastritis, gastroduodenal ulcers, and is also a high risk factor for the development of mucosa-associated lymphoid tissue (MALT) lymphoma as well as gastric adenocarcinoma [1]. Two major virulence factors of H. pylori are the active vacuolating toxin VacA [2] and a 40kb pathogenicity island (cag-PAI) [3] encoding the cag-type IV secretion system (cag-T4SS). The cag-T4SS is used by H. pylori to inject the 120-145 kDa immunodominant protein CagA into the host cell [4]. The presence of a functional cag-PAI correlates with the development of severe gastric inflammation, gastric or duodenal ulcer and gastric cancer [5].
The membranes of the H. pylori cell envelope consist of simple lipids (such as free fatty acids, triglycerides and cholesterol) and phospholipids [6] and harbor a set of outer membrane proteins acting as porins or specific adhesins [7]. The fatty acid composition of H. pylori is uncommon and distinct from the closely related Campylobacter jejuni (C. jejuni) [8]. As H. pylori is not able to synthesize cholesterol itself the bacteria extract cholesterol from the host cell and integrate it into their membrane after it has been α-glycosylated by its cholesterol-αglucosyltransferase (Cgt) [9]. Moreover, the lipopolysaccharide (LPS) of H. pylori is exceptional in many ways. It is appropriately adapted to evade the host immune system and exhibits low endotoxicity compared to the LPS of other pathogens. Therefore, it is considered as an important virulence factor in the persistent colonization of the human stomach [10]. The Oantigen side chains of H. pylori LPS are modified in many ways to resemble Lewis blood group antigens and therefore serve as molecular mimicry [11]. Due to modifications in the phosphorylation and acylation pattern (underphosphorylation and-acylation) lipid A of H. pylori is less negatively charged, and therefore more resistant to cationic antimicrobial peptides (CAMPs) and less biologically active [10,12]. Lipid A can also be modified by a phosphoethanolamine transferase and consists of longer fatty acid chains compared to other bacteria [13,14].
Annexins (ANXs) comprise a multigene superfamily and are known for their ability to bind to membranes in a calcium-dependent way. They are generally found in animal and plant kingdoms, and to date twelve ANX subfamilies are known in mammals, namely ANXA1-ANXA11 and ANXA13 [15,16]. Extracellularly, ANXs interact with bacteria and viruses as well as with components of the extracellular matrix. They may play a regulatory role in inflammation, coagulation and fibrinolysis [17]. Some ANXs are able to interact with viruses and therefore may influence the infection process [17]. The interaction between ANXA5 and influenza virus appears to be essential for successful infection [18], similar to the interaction of ANXA2 with the human cytomegalovirus (CMV), human papilloma virus type 16 (HPV16) and enterovirus 71 (EV71) [19][20][21]. ANXA5 was also shown to bind some Gramnegative bacteria, such as Pseudomonas aeruginosa, Shewanella putrefaciens and Haemophilus influenzae, but not the previously tested Gram-positive bacteria, such as Enterococcus faecalis, Streptococcus pyogenes or Streptococcus agalactiae. Both in vitro and in vivo experiments displayed a possible role of ANXA5 in inhibiting the effect of LPS in endotoxin activities [22].
Although ANXs are present in the human stomach, little is known about their physiological function or their relevance for the H. pylori infection and pathogenicity. However, some associations between ANXs and gastric cancer have already been published [23][24][25][26]. Hence, this study was conducted to investigate an interaction of H. pylori and human ANXs, especially ANXA5. After developing a specific binding assay, we characterized the nature of binding and identified H. pylori lipid A as a bacterial binding target for ANXA5. Furthermore, we investigated whether the recognition by the host innate Toll-like receptor 4 (TLR4) is altered in bacteria covered by ANXA5. Possible effects of the ANX interaction regarding cag-T4SS function and translocation of CagA were also studied. Finally, human gastric tissue samples were analyzed to study the potential role of ANXs in complex infection situations and to gain a better understanding of the physiological significance of the binding.

H. pylori binds ANXA5 and other members of the ANX family
To study a possible interaction of H. pylori with host cell ANXs, especially ANXA5, we first established and standardized an in vitro ANX binding assay. Therefore, GFP (green fluorescent protein)-expressing H. pylori strain P12 (P12-GFP) was incubated with fluorophore-coupled (Alexa-647) ANXA5 to determine and quantify ANX binding by flow cytometry. A strong increase of fluorescence indicated that H. pylori binds ANXA5 ( Fig 1A) and quantification showed binding of~85% of H. pylori P12 wild type (wt) and P12-GFP bacteria to fluorescent ANXA5 (Fig 1B).
To confirm that the events measured in the flow cytometer represented the intact bacteria binding ANXA5, P12-GFP was used to detect GFP-expressing bacteria together with Alexa-647 labelled ANXA5 (Fig 1C). The vast majority of bacteria showed a high 488 signal (Q1 and Q2), which indicates that P12 was expressing GFP, as expected. With ANXA5 addition, most events, which showed a positive signal in the 647-channel, were also positive in the 488-channel (83.4%), suggesting that intact bacteria bind the 647-labelled ANXA5. To further verify and strengthen the H. pylori-ANXA5 interaction by a biochemical method, H. pylori and ANXA5-Alexa 488 were co-incubated and samples were run by polyacrylamide gel electrophoresis. ANXA5 interaction was visualized by excitation and detection of its coupled fluorophore using a Gel Doc documentation system (ChemiDoc, BioRad) ( Fig 1D). These data confirmed the flow cytometry data, supporting that H. pylori is binding ANXA5.
To gain information where on the bacterial surface ANXA5 localizes and how it is distributed, confocal laser scanning microscopy was performed. Since the GFP signal (green) was variable in intensity H. pylori P12-GFP was additionally stained with DAPI after fixation ( Fig  1E). A high number of bacteria bound ANXA5 (magenta), correlating with the flow cytometry data. Whereas some bacteria bound only a small amount of ANXA5 in one spot (Fig 1F), others appeared completely "coated" with ANXA5 on their surface ( Fig 1G).
Next, we tested whether H. pylori would bind other members of the ANX family as well. ANXA1 and ANXA2 were selected, since they are described to play a role in gastric cancer and ANXA2 was found to be up-regulated in H. pylori-associated gastric cancer [23]. Fluorescently labelled ANXA1 or ANXA2 probes were not commercially available, therefore samples were run on a polyacrylamide gel and bound ANXs were detected by immunoblotting using corresponding antibodies. All H. pylori strains tested bound ANXA1 (Fig 2A) and ANXA2 (Fig 2B). In conclusion, by using different methods, such as flow cytometry, gel electrophoresis or

PLOS PATHOGENS
H. pylori annexin interaction H. pylori is binding ANXA5. A) Representative flow cytometry histogram of the 647-fluorescence of H. pylori without (grey) and with ANXA5-Alexa 647 (A647) addition (red). B) ANX binding assay was performed with P12 or P12 GFP and ANXA5-A647. Flow cytometry data was gated to the ANXA5-positive events; the mean percentage of ANXA5-positive bacteria is shown with standard error of mean (SEM). Unpaired t-test was performed; n = 4-5. C) Dot plot showing ANXA5-A647-vs. GFP-488-channel of GFP-expressing H. pylori without (grey) and with (red) ANXA5-A647 addition. Quadrants are shown for the population with ANXA5-A647 addition (red). Q1 represents GFP-positive bacteria, Q2 represents GFP-and ANXA5 positive bacteria and Q3 represents ANXA5-positive bacteria. D) H. pylori and ANXA5-Alexa

PLOS PATHOGENS
H. pylori annexin interaction confocal laser scanning microscopy, the data show that H. pylori is capable of binding human ANXs.

High variability in ANX binding within H. pylori strains and other bacterial species
To study the nature of the H. pylori-ANX interaction in more detail, we now concentrated on ANAX5. Six additional H. pylori strains, type I (cag-PAI-positive) and type II (cag-PAI-negative) were analyzed with the ANX binding assay (for representative histograms see S1 Fig). All strains bound ANXA5, although the percentage of ANXA5-binding bacteria of a given strain was variable. While nearly~80% of H. pylori P12, P217, P145, G27 and X47 bound ANXA5, only~50% of 26695 and Tx30a bacteria showed binding ( Fig 2C). We next determined the capacity of ANXA5 binding by the average bacterium using the median fluorescence intensity (FI) of the ANXA5-binding group of bacteria (see Experimental procedures for details). Surprisingly, this method of evaluation revealed that the P12 strain bound significantly higher amounts of ANXA5 than all other strains tested ( Fig 2D) and that type II strains (X47 and TX30a) had the lowest binding. All together these data show that the ANXA5 binding is not unique for the P12 strain, but prevalent in all H. pylori strains tested. The ANXA5-binding capacity of individual bacteria varies considerably between independent isolates.
To clarify whether ANX binding is more common in bacteria, we included a set of 16 different (Gram-negative, Gram-positive, aerobic, microaerophilic, human pathogenic and nonpathogenic) bacteria in the binding assay (S1 Table for bacterial strains used, S2 Fig for  histograms). Surprisingly, C. jejuni, a close relative to H. pylori, did not bind ANXA5 at all (Figs 2E and S2). Statistical analysis was performed comparing all bacteria with both, H. pylori P12 as positive and C. jejuni as negative control. Less than half of the tested bacterial species showed a significant difference in binding to C. jejuni, meaning that those bacteria did bind ANXA5 (Fig 2E and S2 Table). All strains except Neisseria gonorrhoeae N356 showed a significant difference in binding to H. pylori as the positive control, suggesting that they bound significantly less ANXA5 compared to H. pylori P12, or did not bind at all (Fig 2E and S2 Table).

Growth conditions, but not cholesterol or serum affects the H. pylori-ANX binding
Next, we evaluated the effect of different growth conditions, especially liquid culture versus agar plate growth, which might influence the outer membrane lipid distribution or the "fitness" of the bacteria and could change the ANXA5-binding. The corresponding ANX binding assays showed a significant reduction in the ANXA5 binding capacity from~50% of ANXA5-bound bacteria from plate-grown conditions to~7% of bound bacteria grown in liquid culture. To exclude any interfering effect from the liquid culture medium on the binding, liquid culture medium was added to plate grown bacteria, which did not change the binding (S3A, S3C and S3D Fig). Taken together these experiments show that H. pylori most likely changes its surface binding characteristics depending on the growth conditions. These environmental changes apparently influence the binding of ANXA5. 488 (A488) were co-incubated and samples were loaded onto a polyacrylamide gel and analyzed under denaturing and reducing conditions. The ANXA5 band with a molecular weight of 37 kDa was detected by the Alexa 488 signal using the ChemiDoc MP system (BioRad). ANXA5-A488 was used as positive control. The stain-free gel is shown as a control for the amount of protein loaded. n = 3; one representative gel is shown. E) Analysis with the confocal SP8 microscope and STED technology; GFPexpressing H. pylori P12 strain (green) and ANXA5-Alexa 594 (magenta) were used; the fixed samples were stained with DAPI (blue). Scale bar 2.4 μm. F, G) Analysis with the confocal SP5 microscope and Volocity software; GFP-expressing H. pylori P12 strain and ANXA5-A647 were used. Scale bar equals 1 μm (F) and 1.3 μm (G). https://doi.org/10.1371/journal.ppat.1010326.g001

PLOS PATHOGENS
H. pylori annexin interaction H. pylori is known to extract cholesterol from its growth medium and integrates the glycosylated cholesterol into its membrane using the enzyme Cgt [27]. Furthermore, cholesterol can augment the binding of different ANXs, including ANXA5, to phospholipids [28]. To test whether cholesterol affects the binding of ANXA5, the ANX binding assay was performed with bacteria that had been cultivated on either serum, blood or cholesterol agar plates. Bacteria growing on cholesterol agar plates bound the most ANXA5, however, the quantitative analysis of the data showed no significant difference between bacteria grown on different plates (S3B and S3E-S3G Fig).
To further evaluate the role of cholesterol for the ANXA5 binding of H. pylori, a Δcgt mutant was tested, which is not able to incorporate cholesterol into the bacterial membrane when grown on cholesterol-containing agar plates. We did not observe any difference in ANXA5 binding between the P12 wt and the mutant strain, neither in the amount of positive bacteria nor in the median FI. (S3H- S3K Fig). We therefore concluded that cholesterol incorporation into the bacterial membrane does not play an important role in the interaction of H. pylori and ANXA5.

The H. pylori-ANXA5 interaction depends on calcium ions
Generally, ANXs can interact with eukaryotic membranes in at least three different ways: i) by calcium-dependent binding of their C-terminal domain to lipids, ii) by calcium-independent binding mediated by lipids and iii) directly via the N-terminal domain to soluble or membrane-associated proteins [15]. To determine the nature of the ANXA5 interaction with H. pylori these options were evaluated.
First, the role of calcium ions was tested, since the commonly known way of interaction between ANXs and membrane phospholipids is calcium-dependent [29]. Hence, the ANX binding assay was performed with the chelator EGTA, which binds divalent ions and shows a high affinity for calcium ions. When EGTA was added, the median FI decreased drastically, indicating that H. pylori bound hardly any ANXA5 ( Fig 3A).
Additionally, we tested whether an increased calcium concentration would have any effect on the ANXA5 binding. Therefore, CaCl 2 was added to RPMI, so that RPMI had the same calcium concentration as the standard ANX binding buffer (1.8 mM) specifically designed to investigate ANX-interactions [30]. Under these conditions higher amounts of ANXA5 were bound on average, as indicated by the increase of the median FI ( Fig 3B). In conclusion, when more calcium is present, bacteria bind higher amounts of ANXA5. These findings strongly H. pylori binds ANXA1, ANXA2, and ANXA5, with ANXA5 binding being highly variable among different bacteria. A) H. pylori wt strains P12, 26695 and Tx30a were incubated with ANXA1, the samples were prepared, and gel electrophoresis under denaturing and reducing conditions and Western blots were performed using an anti-ANXA1 antibody (S3 Table) to detect the protein at 37 kDa. ANXA1 only was used as a positive control. The stain-free gel is shown as a control for the amount of protein loaded. n = 3; one representative gel is shown here. B) H. pylori P12 wt strain was incubated with ANXA2, the samples were prepared, and gel electrophoresis under denaturing and reducing conditions and Western blot were performed using an anti-ANXA2 antibody (S3 Table). ANXA2 only (37 kDa) was used as a positive control. The stain-free is shown as a control for the amount of protein loaded. One representative example is shown here. C-E) The ANX binding assay was performed using different H. pylori wt strains and ANXA5-A647 and analyzed by flow cytometry. C) Flow cytometry data was gated to the ANXA5-positive events; the mean percentage of ANXA5-positive bacteria is shown with standard deviation (SD). One-way ANOVA and Dunnett's multiple comparison test against P12 as control were performed. n = 5 (n = 4 for 26695). � p<0.05; �� p<0.01; ��� p<0.001. Representative histograms of the 647-fluorescence of the different bacteria without (grey) and with (red) ANXA5-A647 addition are shown in S1D Fig) The graphs show the mean value with SD of the median 647-fluorescence intensity of the gated "ANXA5-positive group". One Way ANOVA, Post-hoc-test: Bonferroni's Multiple Comparison Test, n = 5 (n = 4 for 26695). � p<0.05; �� p<0.01; ��� p<0.001. E) ANX binding assay using different bacterial species and ANXA5-A647. The flow cytometry data were gated to the ANXA5-positive events; the mean percentage of bacteria in this group is shown here with SEM. C. jejuni is used as negative control for ANXA5 binding. All bacteria showing a significant ( � ) or highly significant ( ��� ) difference to the negative control are considered to bind ANXA5. A complete statistical analysis of ANXA5 binding of the different bacterial species shown shown in S2

PLOS PATHOGENS
H. pylori annexin interaction

Further binding studies suggest ANXA5 binding to lipopolysaccharide of H. pylori
To discriminate between a protein and a lipid as binding target of ANXA5 on the surface of H. pylori, the bacteria were exposed to protein denaturation by heating to different temperatures. Heated bacteria displayed a slightly higher ANXA5 binding ( Fig 3C) and a significantly higher number of ANX5-positive bacteria ( Fig 3D). The successful heat denaturation of proteins was controlled by using the P12-GFP strain showing a positive GFP signal, whereas the temperaturetreated bacteria did not show any GFP signal (see representative histogram, Fig 3E). Taken together these data suggest that the ANXA5 binding of H. pylori is not dependent on the integrity of bacterial proteins and is therefore most likely lipid mediated. The increased binding of ANXA5 to protein-denatured bacteria might be explained by its better access to membrane lipids.
Binding of ANXA5 to LPS of a few Gram-negative bacteria was described previously [22]. H. pylori as a Gram-negative bacterium contains LPS on its outer membrane, although its structure is unique [31]. We used highly purified LPS from H. pylori strain G27, the structure of which has been determined recently [32] (Fig 4A). In a dot blot assay, the binding of ANXA5 was compared to LPS of E. coli K-235 as a negative control ( Fig 4B). An anti-Lipid A antibody verified the binding of LPS on the membrane (0.5 μg, 5 μg or 10 μg), whereas the anti-Lewis Y antibody specifically recognized the corresponding epitope present on H. pylori G27, but not on E. coli K-235 LPS (Fig 4B), which is devoid of this epitope. Incubation with ANXA5 and detection with an anti-ANXA5 antibody revealed binding to H. pylori, but not E. coli LPS, demonstrating a specific, concentration-dependent interaction of ANXA5 with H. pylori LPS ( Fig 4B).

The H. pylori lipid A/core region is a ligand for ANX binding
To further localize and narrow down the ANXA5 binding target on H. pylori LPS, we applied a set of H. pylori strains with inactivation of defined genes encoding key proteins in the LPS biosynthesis pathway (see Fig 4A for a schematic overview of mutant LPS structures). First, the LPS mutant strains G27ΔwaaL and G27ΔrfaE were used, which either lack the complete LPS O-antigen (ΔwaaL) or are deficient in the heptose sugars (ΔrfaE), connecting the O-antigen and core structure to the lipid A membrane anchor, respectively (Figs 4A and S4A).
The G27ΔwaaL and rfaE mutant strains bound only slightly more ANAX5 (97.1% and 99.8%) as compared to the G27 wt strain (94.9%) (S5A Fig). However, the median FI of the events from the ANXA5-positive group was significantly higher in the G27ΔrfaE versus the G27 wt or G27ΔwaaL mutant strain (Fig 4C), indicating that ANXA5 can bind H. pylori G27 significantly better when the core structure and the complete O-antigen of LPS are absent. To verify these results with other H. pylori strains, we used H. pylori P12, 26695 and the the left graph ± SEM. Unpaired t-test was performed. n = 4. � p<0.05; �� p<0.01; ��� p<0.001. One representative histogram of the 647-fluorescence of the bacteria with (blue) and without (red) EGTA is shown on the right. Unstained bacteria were used as a negative control (grey). B) ANX binding assay was performed with P12 in RPMI without (red) and with (dark grey) CaCl 2 (final Ca 2+ -concentration: 0.42 and 1.8 mM, respectively) with ANXA5-A647. The flow cytometry data was gated to the ANXA5-positive events; the median FI of bacteria in this group is shown in the left graph ± SEM. Unpaired t-test was performed. n = 4. � p<0.05; �� p<0.01; ��� p<0.001. A representative histogram of the 647-fluorescence of the bacteria with (dark grey) and without (red) CaCl 2 is shown on the right. Unstained bacteria were used as a negative control (light grey). C-E) ANX binding assay was performed with untreated bacteria (grey) or with bacteria heated to 56˚C (red), 65˚C (blue) or 80˚C (green) with ANXA5-A647. C) A representative histogram of the 647-fluorescence of the heated bacteria. D) Flow cytometry data were gated to the ANXA5-positive events; the mean percentage of bacteria in this group are displayed here ± SEM. Non-heated bacteria (grey) were compared to bacteria with protein denaturation. Statistical analysis with one-way ANOVA and Dunnett's multiple comparison post-test against the non-treated control; n = 3-4. � p<0.05; �� p<0.01; ��� p<0.001. E) Representative histogram of the 488-fluorescence of the bacteria. This was used as a control for protein denaturation, as the GFP protein was destroyed by heating. Same color code used as in (C/D). The median FI of the events from the ANXA5-positive group was slightly higher in the P12ΔrfaE, but not significantly different to the P12 wt strain ( Fig 4D). A striking difference in ANXA5 binding was seen between the 26695 wt (22.0%) and the 26695ΔrfaE mutant strain (85.9%) (S5C Fig), which also was supported by a significant difference between the median FI of the ANXA5-positive group of both strains ( Fig 4E).
These results could be corroborated further by using H. pylori strains harboring a deletion in the heptose III transferase gene (ΔHP1284) in strains G27 and 26695. These mutations result in an incomplete core region, like the rfaE mutants (Figs 4A and S4A and S4C). ANXA5 binding revealed a significantly higher median FI from the ANXA5-positive group in the ΔHP1284 versus the corresponding wt strains (Figs 4F-4G and S5D-S5E), indicating that ANXA5 can bind H. pylori G27 and 26695 significantly better when part of the core structure, especially the third heptose and the adjoining Glc-Gal residues are missing, although the Oantigen of LPS is still present (Fig 4A) [31]. In conclusion, the corresponding ΔHP1284 mutants behaved like the rfaE mutants concerning ANAX5 binding, whereas the waaL mutant, which harbors a complete core structure, but lacks the complete O-antigen, behaves like the wt strain concerning ANXA5 binding. These data suggest that the O-antigen of LPS does not significantly influence ANXA5 binding, but changes in the core region (ΔHP1284 and rfaE deletions) seem to expose the binding target of ANXA5 on LPS.

The lipid A phosphorylation status determines ANXA5 binding of H. pylori
Since the ANXA5 interaction with H. pylori lipid A is strongly calcium-dependent, we speculated that negatively charged structures on lipid A might be crucial for ANXA5 binding. We therefore concentrated on the phosphate groups of lipid A. In the process from lipid A synthesis to outer membrane insertion an originally bisphosphorylated form of lipid A is modified to result in a mature major lipid A species lacking both, the 4´-phosphate (removed by LpxF) and the 1-phosphate group (cleaved by LpxE), the latter being subsequently substituted with a phosphoethanolamine residue catalyzed by EptA [10]. We generated lpxE and lpxF mutants as well as lpxE/F double mutants in H. pylori G27 (see Fig 4A for a structural overview on lpx mutant strains), since this strain is best characterized concerning its LPS structure [33]. In addition to lpxE and lpxF, we also inactivated the eptA gene encoding the enzyme which transfers phosphoethanolamine to the H. pylori lipid A. The intact LPS structure of these mutants was verified by SDS PAGE and silver staining (S4B Fig). ANXA5 binding studies revealed a significant increase in ANXA5 binding for both, the G27ΔlpxF mutant and a G27ΔlpxE/F double mutant as compared to the G27 wt strain ( Fig   Fig 4. Characterization of lipid A as binding target for ANXA5. A) Overview of LPS structures of G27 wt, G27ΔHP1284, G27ΔwaaL and G27ΔrfaE (left, adapted from [31]) and lipid A phosphorylation status of H. pylori lpx mutant strains (right, adapted from [10]). GlcNglucosamine; Gal-galactose; KDO-3-deoxy-D-manno-oct-2-ulosonic acid; Glc-glucose; Hep-heptose; GlcNAc-N-acetylglucosamine. B) The binding of ANXA5 to LPS from E. coli and H. pylori was analyzed by dot blot. Different amounts of LPS (10, 5, and 0.5 μg) on a PVDF membrane were incubated with anti-lipid A antibody, anti-Lewis Y, or ANXA5 (2 μg/ml, A9460, Sigma) followed by anti-ANXA5 antibody (for antibodies used see S3 Table). Lipid A and Lewis Y were used as loading controls. Shown is a representative image from at least 3 independent experiments. C-G) Different H. pylori wt and LPS mutant strains were analyzed for ANXA5-Alexa647 binding to G27 and respective mutants (C, F), P12 and respective mutants (D), and 26695 and respective mutants (E, G). Results are depicted as mean ± SD. Statistics were calculated using one-way ANOVA with Bonferroni's multiple comparison test (C) or Student's unpaired t-test (D-G). n = 8 (C), n = 5 (D), n = 7± (E), n = 6 (F), n = 5 (G). � p < 0.05, �� p < 0.01, ��� p < 0.001. Representative histograms are shown in Fig S5. H) Median FI of ANXA5-positive group of G27 wt, ΔlpxE, ΔlpxF, ΔlpxE/F and ΔeptA strains was analyzed as described before. Results are depicted as mean ± SD. Statistics were calculated using one-way ANOVA with Bonferroni's multiple comparison test. n = 6 (n = 7 for ΔlpxE; n = 5 for ΔlpxF). � p < 0.05, �� p < 0.01, ��� p < 0.001. https://doi.org/10.1371/journal.ppat.1010326.g004

PLOS PATHOGENS
H. pylori annexin interaction 4H). A G27 lpxE deletion strain showed a stronger tendency for ANXA5 binding without reaching statistical significance. The absence of the eptA gene in G27 resulted in an ANXA5 binding pattern similar to the lpxE mutant.
In conclusion, the binding assays suggest that ANXA5 binding of strain G27 is significantly enhanced when its lipid A is phosphorylated at the 4´-sugar position. However, the 1-phosphate, as well as the lack of a phosphoethanolamine at this position, both seem to slightly support ANXA5 binding.

ANXA5 binding inhibits LPS-mediated TLR4 signaling
In response to bacterial infection, the infected host produces antimicrobial proteins, such as lactoferrin, lipocalin or calprotectin (CP), which sequester soluble metal ions [34,35]. Especially CP can sequester the transition metals iron, manganese and zinc via chelation. To promote resistance to CP H. pylori alters its lipid A molecules and changes its growth behavior, which also results in increased biofilm formation [36]. Interestingly, H. pylori evades recognition of the host's immune system by modifying lipid A, especially by removing the phosphate groups, as described above [10]. Structural and experimental infection studies indicate that especially the 4´-phosphate group is important for TLR4-MD2 recognition of lipid A [10, 37]. We therefore hypothesized that H. pylori might use ANXs binding to shield its phosphorylated lipid A molecules against the recognition by the TLR4-MD2 complex from cells of the innate immune system or epithelial cells in the stomach mucosa.
To study the role of ANX binding to H. pylori lipid A for its recognition by the TLR4-MD2 complex we monitored TLR4 activation using HEK293 cells stably transfected with the TLR4 machinery (HEK-Blue-hTLR4) and a secreted alkaline phosphatase reporter gene controlled by an NF-kB inducible promoter (Invivogen). This colorimetric assay allows to quantify TLR4-MD2 complex activation. To eliminate the contribution of the cag-PAI on activation of the NF-kB promoter, we generated a ΔlpxE/F double mutant with the subsequent deletion of the entire cag-PAI (G27ΔlpxE/FΔcag-PAI). Co-incubation of H. pylori G27ΔlpxE/FΔcag-PAI with the reporter cell line clearly showed a significantly higher activation of the NF-kB promoter as compared to the G27Δcag-PAI strain. Remarkably, a significant, concentrationdependent reduction in activation of hTLR4-MD2 in G27ΔlpxE/FΔcag-PAI was observed upon addition of ANXA5 from 1 to 5 μg/ml (Fig 5A).
To determine the specificity of the hTLR4-MD2 activation by H. pylori we incubated the reporter cell line with purified E. coli LPS (1 μg/ml), which clearly stimulated the reporter cell activity. In contrast, addition of ANXA5 (1 or 5 μg/ml), which does not bind E. coli LPS ( Fig  4B), did not have any effect on the NF-kB promoter activity, ruling out any direct inhibitory activity of ANXA5 on the hTLR4-MD2 complex (Fig 5B).

Gastric mouse colonization significantly enhances the H. pylori ANXA5 binding capacity to interfere with TLR4 signaling
To test whether H. pylori grown under in vitro conditions on agar plates would change its surface in favor of ANXs binding during a stomach passage we infected mice with the mouse adapted H. pylori strain PMSS1 for three weeks and tested the ANXA5 binding before infection and immediately after re-isolation. Notably, the ANXA5 binding capacity was twice as high in the re-isolates compared to the pre-infection strain. This was the case for both the median FI of the ANXA5-positive group (Fig 5C) and the total percentage of ANXA5-binding bacteria (Fig 5D). The ability of PMSS1 re-isolates to stimulate TLR4 was increased compared to the pre-infection strain (Fig 5E), but the difference was not significant, presumably because the cag-T4SS of this wt strain also strongly contributes to the stimulation of NF-kB.

PLOS PATHOGENS
H. pylori annexin interaction Nevertheless, TLR4-MD2 specific stimulation in the reporter cell line could be significantly inhibited with ANXA5 in the re-isolates, but not in the pre-infection strain (Fig 5E). Taken together these data suggest that the ability of H. pylori to bind ANXs can be upregulated in the mouse model in vivo, presumably by changes in the lipid A phosphorylation status.

H. pylori-ANXA5 interaction reduces CagA translocation into AGS cells but does not affect cell binding or IL-8 induction
Earlier experiments reported that the addition of ANXA5 to an infection of H. pylori with the gastric epithelial cell line AGS resulted in a~20% decrease of CagA translocation [38]. When H. pylori was pre-incubated with ANXA5 and was then used to infect AGS cells, CagA translocation was significantly reduced (~33% on average) (Fig 6A and 6B). To verify these results, we applied phosphorylation-independent, quantitative CagA translocation assays. The TEM1-beta-lactamase reporter assay [39] was not applicable, due to the necessity of high calcium concentrations needed for ANXA5 binding, which interfered with the reporter assay. Therefore, we next applied the HiBiT-CagA translocation reporter assay developed and adapted for H. pylori recently [40]. This luciferase-based assay, making use of an N-terminal fusion of CagA with an 11-residue split-luciferase (HiBiT) tag, demonstrated a significant reduction of CagA translocation into AGS cells (~20%) in an ANXA5 concentration-dependent manner (Fig 6C). We next used a P12[HiBiT-CagA]ΔrfaE mutant strain for determination of CagA translocation into AGS cells. Notably, this mutation, which results in a major disturbance of the H. pylori cell envelope, apparently also affected the integrity of the cag-T4SS, resulting in a lower CagA translocation efficiency compared to the wt strain. However, an even stronger effect of roughly 50% reduction in CagA translocation was seen in comparison of the P12 wt versus the P12[HiBiT-CagA]ΔrfaE mutant loaded with 10 μg/ml ANXA5 (Fig 6D), suggesting that the unmasking of lipid A and the higher amount of ANXs binding directly affects CagA translocation efficiency of H. pylori.
We next studied the possible effects of H. pylori ANXA5 binding to the bacteria-host cell interaction. Notably, the binding of the bacteria to AGS cells was not at all affected when ANXA5 was added simultaneously or by pre-incubation of the bacteria (S6A and S6B Fig). Moreover, ANXA5 addition had no effect on IL-8 production of AGS cells by the cag-T4SS (S6C and S6D Fig). In summary, our data indicate that ANXA5 binding to H. pylori LPS can reduce the CagA translocation efficiency without affecting the bacterial binding capacity to host cells.

H. pylori infected gastric tissue reveals elevated levels of ANXA2/5 and its binding/recruitment to the bacterial surface
To investigate whether the H. pylori infection in humans might affect production and release of ANXs in the gastric mucosa, gastric sections were stained with primary antibodies against ANXs A1, A2 and A5 and the amount of ANXs present was determined by the fluorescence intensity of the anti-ANXA 1, 2 and 5 antibodies calculated per cell (see Experimental procedures for details). Analysis of gastric antrum and corpus tissue sections from two non-infected and three infected individuals showed a slightly enhanced amount of ANXA1 in H. pylori-positive versus H. pylori-negative tissue, which did not reach statistical significance (Fig 7A). For signaling was determined using QuantiBlue (Invivogen) and OD was measured at 665 nm. Results are depicted as mean ± SD. C-D) ANXA5-binding was analyzed for H. pylori PMSS1 or a mouse-isolated PMSS1 strain by flow cytometry. Data were evaluated for median FI (C) and percentage of ANXA5-positive bacteria (D). Results are depicted as mean ± SD (A). Statistics (A-B) were calculated using one-way ANOVA with Bonferroni's multiple comparison test or two-way ANOVA (E) with Bonferroni's post hoc test. Statistics (C-D) were analyzed using Student's t-test. n = 5 (A, B-D), n = 4 (E). � p < 0.05, �� p < 0.01, ��� p < 0.001. https://doi.org/10.1371/journal.ppat.1010326.g005

PLOS PATHOGENS
H. pylori annexin interaction ANXA2 and ANXA5 a significantly higher quantity of green fluorescence per cell nucleus was detected, indicating a higher amount of ANXA2 and ANXA5 in gastric corpus, but not gastric antrum tissue of infected versus non-infected individuals (Fig 7B and 7C).
To test whether H. pylori would also bind ANXs under in vivo conditions in the gastric mucosa, human H. pylori infected gastric corpus sections were stained with anti-H. pylori antiserum (magenta), anti-ANXA5 antibody (green), and DAPI to visualize host cell tissue. As controls, sections were stained for H. pylori and the secondary antibodies of the ANXA5 staining (goat anti-mouse IgG-Alexa488) or with both secondary antibodies only (S7A and S7B  Fig). Indeed, co-localization of H. pylori and ANXA5 was detected in H. pylori-positive patient sections using confocal laser scanning microscopy (Fig 7D-7F). The green signal (ANXA5) overlaps with the magenta signal (H. pylori), demonstrating an interaction of the bacteria with ANXA5 in vivo. High resolution confocal images of two individual bacteria were taken with z-stacks and 3D projections rendered images. Displayed images with different rotated angles (y-axis) show cell volume with close binding of ANXA5 mostly covering the bacterial pole ( Fig 7G and S1 Movie).
In conclusion, these data suggest that the H. pylori infection results in a higher abundance of ANXs in the gastric tissue and the colonizing H. pylori are able to recruit these ANXs to their surface.

Discussion
ANXs are soluble proteins and are mainly found in the cytosol of eukaryotic cells. However, some ANXs such as A1, A2, A4 and A5 can also be detected extracellularly. They are primarily known to interact with eukaryotic membranes and modulate membrane trafficking and organization, including exo-, endo-, and phagocytic processes [15]. Recombinant ANXA5 is commonly used to detect the presentation of phosphatidylserine (PS) on the outer leaflet of the cell membrane, since externalized PS serves as a recognition and clearance signal for phagocytes to identify apoptotic cells [41].
We describe here for the first time the intense interaction between ANXs and H. pylori. After having initially established the binding of H. pylori to ANXA5, we standardized the assay and investigated the nature of this interaction using different methods. In addition to ANXA5, ANXA1 and ANXA2 were also shown to be bound by H. pylori. To demonstrate a more general relevance of our results, we extended the binding studies to a number of other H. pylori strains as well as some other Gram-negative and Gram-positive bacterial species. Interestingly, some species showed significant binding, while others showed no binding at all (Fig 2E and S2  Table). We then identified the lipid A portion of H. pylori LPS as a binding target for human ANXA5. Earlier work already showed that LPS from cag-PAI positive type I H. pylori strains induced a stronger TLR4 signaling in cognate target cells as compared to LPS of type II strains [42]. Intrinsic differences in the structure or bioactivity of LPS dependent on the presence of the cag-PAI might explain the generally lower ANXA5 binding of type II strains in our assays.

Fig 7. Elevated levels of ANXs in H. pylori infected versus non-infected human stomach biopsy sections and co-localization in vivo. A-C) ANXA1 (A), ANXA2 (B)
and ANXA5 (C) levels were quantified in H. pylori positive (red) and negative (black) gastric biopsy samples. Gastric sections from antrum (N1-107-A1) or from corpus (N1-14-C, N1-17-C, N1-4-C2, N1-24-C2) were stained with antibodies as described in Material and Methods (Immunohistochemistry of tissue slides. . .). Sections were imaged by the Leica TCS SP5 confocal microscope and quantification was performed using Fiji software. Data analysis was as described in Materials and Methods. Statistical analysis was performed using one-way ANOVA with a Bonferroni's Multiple Comparison test. � p < 0.05, �� p < 0.01, ��� p < 0.001. D-G) Images were acquired by super resolution confocal microscope LSM 880 (Zeiss) with Airyscan module.

H. pylori annexin interaction
At this point, it is unclear why E. coli and C. jejuni did not bind ANXA5, even though they carry a bis-phosphorylated lipid A disaccharide, which in C. jejuni is masked by phosphoethanolamine that might interfere with ANXs binding. The fact that Gram-positive S. aureus and S. pneumoniae clearly showed binding of ANXA5 argue against a general or exclusive binding of ANXs to lipid A. Possible other targets might be lipoteichoic acid in case of Gram-positive bacteria, or membrane components, such as PS or other membrane lipids. Thus, also for H. pylori, lipid A might not be the only possible target for ANX interaction, since P12 did not significantly change its binding pattern upon deletion of rfaE (Fig 4D), suggesting that ANXA5 might have an additional binding target in this strain besides lipid A.
To evade the host, innate immune system some pathogenic bacteria, including H. pylori, apparently change their surface characteristics and therefore employ different strategies. During LPS biogenesis in H. pylori and some other bacteria, the negatively charged phosphate groups on the lipid A disaccharide backbone (1-phosphate) are subsequently removed and replaced by the more neutral substituent phosphoethanolamine [10, 43]. As a result, negative surface charges are lost, leading to a bacterial membrane that is more resistant to cationic antimicrobial peptides and proteins [44].
We show that the conserved LPS core region, but not the O-antigen chain, interferes with ANXA5 to access its lipid A binding target by using a series of defined H. pylori mutant strains in key genes of the LPS biosynthetic pathway. Due to the strict calcium dependence of ANXA5 binding to H. pylori, a negative-charged structure, such as the phosphorylated lipid A sugars, were considered as possible target, and therefore lpxE/F mutants were generated and tested. As expected, the 4´-phosphate, but also the 1-phosphate group of lipid A play an important role for ANX binding. The gene deletion-mediated prevention of H. pylori lipid A dephosphorylation by phosphatases LpxE, LpxF or both together, which results in mono-or bis-phosphorylated lipid A, respectively [10], significantly increased the binding of ANXA5 in strain G27 (Fig 4H). Furthermore, the presence of phosphorylated lipid A clearly enhanced the recognition of H. pylori LPS by the innate immune receptor TLR4, as already shown by Cullen et al. [10]. In addition to lpxE and lpxF, we also inactivated the eptA gene, which encodes the enzyme that transfers a phosphoethanolamine residue to the 1-phosphate after its dephosphorylation by LpxE. This resulted in a similar slightly enhanced ANXA5 binding comparable to the G27ΔlpxE mutant (Fig 4H), suggesting that the negatively charged phosphate groups, rather than the neutral phosphoethanolamine residue, plays an important role in strain G27 ANX binding.
This strategy of H. pylori to evade innate immunity by reducing the surface charge of lipid A seems to be generally straight forward. However, in the human stomach H. pylori comes in contact with activated neutrophils, which release large amounts of antimicrobial proteins, such as CP [36], comprising as much as 60% of the soluble protein content of the cytosol of a neutrophil [45]. CP sequesters soluble metal ions at the pathogen-host interface, designated as nutritional immunity [35]. This host response causes H. pylori to alter its lipid A molecules, apparently by a posttranscriptional mechanism, including the inactivation of LpxF, which results in an increased phosphate decoration of the outer membrane. This programmed surface change of H. pylori in the presence of CP was reported to induce a better bacterial resistance and fitness against the antimicrobial protein and induces a strong biofilm formation [36].
Thus, under infection conditions and in contact with CP released by activated neutrophils, the bacteria might have a high risk to be strongly recognized and eventually eliminated by the innate immune system. Structural and experimental studies showed that especially the 4 -phosphate group of lipid A is important for TLR4-MD2 recognition [10,37]. We therefore hypothesized that under these conditions H. pylori might use ANX binding in the host to shield its phosphorylated lipid A molecules against the recognition by the innate TLR4 receptor expressed on neutrophils, macrophages or dendritic cells resident in the stomach mucosa. Indeed, we could show that H. pylori mutant strains producing phosphorylated lipid A (lpxE/F mutants) are significantly protected from TLR4 recognition when they were incubated with ANXA5, suggesting a role of ANX binding for an immune escape mechanism (Fig 5A). These results were corroborated by a strongly induced ANXA5 binding capacity and a significant reduction of TLR4-mediated NF-kB activation of the ANXA5-treated mouse colonizing H. pylori strain PMSS1 after mouse infection, as compared to the same strain grown for several passages in vitro (Fig 5C-5E).
Recently, the interaction of H. pylori with human conventional dendritic cells (cDCs) such as CD1c + cDCs (cDC2s), which are among the first immune cells to encounter H. pylori in the gastric lining, has been studied [46]. These innate immune cells express the TLR4 receptor, which upon interaction with H. pylori drives the secretion of inflammatory Th1/Th17 cytokines (IL-1β, IL-12, IL-18 and IFN gamma) to shape a chemokine milieu for an effective immune response for bacterial clearance. Antibody-mediated blocking of TLR4 signaling in a mouse model resulted in down regulation of MyD88 expression, NF-kB activation and an increase in the numbers of CD4+CD25+Foxp3+Tregs, which increased the H. pylori colonization density [47]. Thus, our finding that the H. pylori-ANX interaction efficiently blocks TLR4 signaling might contribute to the bacterial strategy to efficiently subvert host cell clearance mechanisms.
CagA, the major pathogenicity factor of H. pylori, is injected into the host cell via the cag-T4SS, where it causes various effects [4]. We show here that binding of ANXA5 to H. pylori reduces the CagA translocation by up to~33%, which is consistent with the data from Murata-Kamiya et al., who previously described that ANXA5 can reduce the CagA translocation by approximately 20% [38] (Fig 6A and 6B). They hypothesized that the reduced translocation is due to the blocking of phosphatidylserine (PS), which is involved in binding of CagA before translocation, since an anti-PS antibody showed the same effect. This does not exclude a possible direct interaction between ANXA5 and H. pylori, as described here. Thus, not only the host cell, but also the bacteria is a target for ANXA5 binding to modulate the efficiency of the cag-T4SS and translocation of CagA. A reduction of the CagA translocation as a consequence of a reduced binding of bacteria to the host cells was ruled out (S6A and S6B Fig).
In conclusion, our data clearly expand the functions of ANXs. In addition to their role as modulators of eukaryotic membrane trafficking and phagocytosis, novel functions for ANXs emerge as tools exploited by bacterial pathogens such as H. pylori to actively modulate its surface and interfere with recognition by the host immune system.

Ethics statement
The study was approved by the Ethics Committee of the Nigerian Institute of Medical Research (IORG0002656) and of the Ludwig-Maximilians-Universität München (Registration number 335-08). All participants provided informed written consent.
Bacterial strains and cell line maintenance. H. pylori strains were grown on GC agar plates (Oxoid) supplemented with vitamin mix (1%) and horse serum (8%) (serum plates) or cholesterol (cholesterol plates) and cultured for 16 to 48 h in a microaerobic atmosphere (85% N 2 , 10% CO 2 , 5%O 2 ) at 37˚C. E. coli strains were grown on Luria-Bertani (LB) agar plates supplemented with antibiotics, as appropriate. Plasmids were introduced into H. pylori by transformation as described previously [48], and transformants were selected on serum plates containing 6 mg/l chloramphenicol, 8 mg/l kanamycin, 10 mg/l erythromycin, or 250 mg/l

PLOS PATHOGENS
H. pylori annexin interaction streptomycin, as appropriate. For liquid culture plate-grown H. pylori were resuspended in PBS and added in an OD of 0.075 into 6-well plates prepared with 2 ml Brucella Broth and 8 μl cholesterol (Gibco) as previously published [49]. After 24h at 37˚C in a 10% CO 2 -atmosphere vitality was checked under the microscope before use. All cell lines were maintained at 37˚C and 5% CO 2 with the appropriate media. Generally, cells were grown in 75 cm 2 tissue culture flasks (BD Falcon) and in 6-, 12-and 24-well plates (tissue culture treated cell culture clusters, Costar, Corning Inc.).
H. pylori mutant strain construction. H. pylori LPS mutants G27ΔHP1284, G27ΔwaaL, 26695ΔHP1284 were generated as described earlier [31]. P12ΔrfaE, G27ΔrfaE and 26695ΔrfaE were constructed by transformation of the wt strains with plasmid pCL1. Therefore, a pSMART-hcKan vector harboring bp 908247-910341 of the P12 chromosome (= HP2kb04_M12) was amplified by inverse PCR using primers CL1 (ACG TGT CGA CTA GGA TTT TTT TCA TGC AC, SalI) and CL2 (CAG GAT CCA AAA TTA AAA GGA CAC ATA ATG ATT GAT G, BamHI) to generate the flanking regions for homologous recombination and ligated with a chloramphenicol resistance cassette (SalI/BamHI). Successful deletion mutants were confirmed by PCR using primers CL23 (ACA ATC AAA GCC ATA TCG CT) and CL24 (CCG CTG TTT CTG ATA CGA CCA).
Infection with H. pylori. Before infection, AGS cells were synchronized by serum deprivation overnight. 1 h prior to infection, the medium was changed back to complete medium (CM). Cells enter the G 0 quiescence when they grow in medium without serum. When the medium is again complemented with serum the cells ideally all enter the early G 1 phase simultaneously [50]. Cells were infected with an MOI of 60 and incubated at 37˚C, 5% CO 2 for 1-3 h. 1 h prior to infection, ANXA5 was added to the H. pylori pre-culture. Alternatively, ANXA5 was added simultaneously at infection. To harvest the cells, they were scraped off in PBS supplemented with protease inhibitors (1 mM PMSF, 1 mM sodium vanadate, 1 μM leupeptin, 1 μM pepstatin). To prepare the samples for Western blotting, the pellet was resuspended in 20 μl PBS/protease inhibitors and 30 μl of 2X SDS loading buffer and boiled for 10 min.
H. pylori gastric colonization and re-isolation. 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 (ROB-55.2-2532.Vet_02-18-189). The PMSS1 strain was cultured on GC agar plates from frozen stocks (with 1% vitamin mix and 8% horse serum) for three days under microaerophilic condition (85% N 2 , 10% CO 2 , 5% O 2 ). Mice were infected using an oral gavage needle for three consecutive days using approximately 6 x 10 7 CFUs in Brucella Broth. 3 weeks post infection, H. pylori was re-isolated from the antrum and corpus following tissue homogenization using Wheaton homogenizations and selected on DENT plates (Oxoid H. pylori selective supplement-SR0147E containing Vancomycin, Cefsulodin Trimethoprim, Amphotericin B with Polymyxin B (Simga: P9602), Nalidixin acid (Roche), Bacitracin and 1% vitamin mix and 8% horse serum).
Protein separation by polyacrylamide gel and blotting. An adapted 6% single gel system based on Ahn et al [51] containing 0.01% trichloroethanol was used to separate the proteins based on their molecular weight, as previously published [52]. For a detailed protocol see https://dx.doi.org/10.17504/protocols.io.gipbudn). As a molecular weight marker, the Precision Plus Protein Allblue standards (BioRad) was used. The total number of proteins was detected in the ChemiDoc MP System (BioRad) using the stain-free technology [53]. In a semi-dry blotting chamber, the proteins were transferred to a PVDF membrane at a current of 1 mA/cm 2 for 90 min. After blotting, the membrane was dried for a minimum of 1 h at 37˚C or overnight at RT. The dry membrane with the proteins was re-activated by methanol. Blocking of unspecific binding sites was performed with TBS/0.075% Tween-20/3% bovine serum albumin (BSA) for 1 h at RT. The primary antibody was added and incubated for 1 h at room temperature (RT). After washing four times in TBS/0.075% Tween-20, the secondary antibodies were incubated for 45 min. at RT. For detection of horse radish peroxidase (HRP)-coupled secondary antibodies, the Millipore Immobilon Western solution was used. The signal was detected in the ChemiDoc MP System. For detection of phosphorylated CagA, the α-phosphotyrosine antibody (4G10) was added (1:10,000). Quantification of protein bands for CagA was performed using ImageLab software (Version 6.0.1, BioRad) and normalized to the stain-free gel. Alternatively, detection of fluorophore-coupled proteins was performed by exciting and detecting the fluorophore in the Gel Doc documentation system. ANX binding assay. Bacteria were resuspended in PBS and incubated at an OD 550 of 0.1 in an Eppendorf tube with 1 ml RPMI 1640 (Life Technologies) and 2.5 μl of ANXA1, ANXA2 or ANXA5. ANXA5 coupled with Alexa Fluor 647, 594 or 488 (Thermo Fisher). After 1 h incubation at 37˚C in 10% CO 2 , the samples were prepared for further analysis by flow cytometry, Western blot or confocal microscopy. For flow cytometry analysis the samples were washed three times in ANXA5 binding buffer modified after Kenis et al., (10 mM HEPES, 150 mM NaCl, 5 mM KCl, 5 mM MgCl 2 , 1.8 mM CaCl 2 , pH 7.4) [30] with centrifugation at 4,000 rpm at 4˚C for 5-10 min. The bacteria were resuspended in ANXA5 binding buffer and ANXA5 binding was measured as the increase of fluorescence in the 647-channel using the BD FACS Canto II. Flow cytometry data was analyzed using FlowJo software (version 7.6.1 or 10). Gates were set using the negative control (~2.5% of events in the positive gate). For statistical evaluation, the values of the negative controls (percentage and median FI) were subtracted from the obtained values of the other samples. For analysis with Western blot the samples were washed twice in ANXA5 binding buffer with centrifugation at 4,000 rpm at 4˚C for 5-10 min and were then resuspended in 5 μl of PBS/Protease inhibitors and 7 μl of 2x SDS loading buffer.
Dot blot assay. LPS samples were diluted in PBS and applied to a PVDF membrane. Unspecific binding sites were blocked with TBS/0.05% Tween-20/5 mM CaCl 2 /3% BSA for 1 h at RT. All washing and incubation steps were performed with TBS/0.05% Tween-20/5 mM CaCl 2 . To detect ANXA5 binding, membranes were incubated with 2 μg/ml ANXA5 (Sigma, A9460) at 4˚C overnight. After washing three times, the primary antibody (S3 Table) was incubated for 4 h at RT to detect ANXA5 binding. For detection of Lewis Y or lipid A (S3 Table), membranes were incubated with the respective primary antibodies at 4˚C overnight. Blots were incubated with HRP-coupled secondary antibodies for 1 h at RT and detection was performed using Immobilon Chemiluminescent HRP Substrate (Millipore).
H. pylori binding assay. AGS cells were infected with GFP-expressing H. pylori at an MOI of 60. Bacteria were either pre-incubated with ANXA5 or ANXA5 was added simultaneously with the bacteria to the cells. After 1 h of incubation, the cells were put on ice, washed twice with ANXA5 binding buffer and harvested with a cell scraper directly before measurement in the flow cytometer (BD FACS Canto II). The GFP-signal was measured and used to determine the amount of bacteria that had bound to the cells.
LPS purification. LPS from H. pylori G27 was purified using the EDTA-promoted ultracentrifugation method as previously described [31]. Briefly, the bacterial cells harvested from 25 serum plates (24 h growth) were suspended into 15 ml of 10 mM Tris-HCl buffer (pH 8.0) containing 2 mM MgCl 2 . The bacterial cells in the suspension were disrupted by sonication and treated with DNase I and RNase A for 4 h in a 37˚C water bath. Subsequently, 5 ml 0.5 M EDTA, 2.5 ml 20% SDS, and 2.5 ml 10 mM Tris-HCl buffer (pH 8.0) were added, and the pH was adjusted to 9.5. After centrifugation at 50,000 g for 30 min, the supernatant was transferred into a new tube and treated with proteinase K overnight in a 55˚C water bath. After the addition of 2 volumes of 0.375 M MgCl 2 in 95% ethanol, the mixed solution was centrifuged at 12,000 g for 15 min at 4˚C, and the obtained LPS pellet was resuspended in 10 ml of 10 mM Tris-HCl (pH 8.0) containing 25 mM MgCl 2 . The solution was then ultracentrifuged at 100,000 g for 12 h at 15˚C to obtain the purified LPS pellet, which was resuspended in H 2 O and lyophilized.
Confocal microscopy to study H. pylori-ANX interaction. After incubation of the bacteria with ANXA5 as described for the ANX binding assay, samples were fixed with 2.5% PFA for 15 min at RT and stained with DAPI (5 μg/ml) for 10 min at RT. Samples were settled in a 12 mm rounded cover slips laid in 24 wells plates by centrifugation at 3.000 rpm (Heraeus Megafuge 3.0) for 20 min at 4˚C. The samples were mounted with mounting media on an object slide. Microscopy samples were examined with a Leica TCS SP5 confocal microscope, or Leica TCS SP8 confocal microscope and STED technology. Routine image processing was done with ImageJ 1.45h (National Institutes of Health) while Volocity 6.0.1 (Perkin Elmer) was used for co-localization analysis.
Immunohistochemistry of tissue slides and analysis by high resolution confocal microscopy. Paraffin sections were de-paraffinized and treated with sodium citrate buffer (10 mM sodium citrate, 0.05% Tween-20, pH 6.0) for antigen retrieval (microwave). Blocking was performed with 1% BSA, 2.5% goat serum and 0.5% saponin in PBS overnight at 4˚C. Primary antibodies were added 1:50 (ANX) or 1:1000 (H. pylori) (S3 Table) in PBS/1% BSA/2.5% goat serum/0.5% saponin and samples were incubated overnight at 4˚C. After washing, secondary antibodies (1:1000 in PBS/0.05% Tween-20/2.5% goat serum) were applied for 1 h at RT in the dark. Sections were stained with DAPI (5 μg/ml) for 10 min at RT in the dark and after washing were embedded in Fluorescence Mounting Medium (Dako). Sections were visualized using the Leica TCS SP5 confocal microscope for quantification of ANX signal or the confocal laser scanning microscope LSM880 (Zeiss) with Airyscan Module for co-localization studies. Imaris Bitmap 8.7 was used for co-localization analysis and Imaris software was used for the 3D reconstruction and rendering images.
Quantitative analysis of immunohistochemistry data. To analyze the stained sections, FIJI software (https://pubmed.ncbi.nlm.nih.gov/22743772) was used. For this, two macros were written, one counting the nuclei on each image and one measuring the fluorescence intensity of the stained protein of interest. The threshold for measuring the fluorescence intensity was chosen, so that no signal was detected in the isotype control samples. Fluorescence intensity was normalized to the number of nuclei on the respective image.
IL-8 secretion. The production of IL-8 by AGS cells after infection with H. pylori strains for 4 h was determined from cell supernatants by a sandwich enzyme linked immunosorbent assay (ELISA) (R&D Systems, USA), according to the manufacturer's instructions, as previously described [56].
TLR4 assay. HEK-Blue hTLR-4 cells were obtained from Invivogen. Cells were cultured in DMEM/10% FCS/1% glutamine (Life Technologies) and HEK-Blue selection antibiotics (Invivogen) and sub-cultured every two to three days. Cells were seeded in a clear flat bottom 96-well plate (BD Falcon) one to two days prior to infection without selection antibiotics. Bacterial strains were incubated at an OD 550 of 0.075 in 500 μl of F12 (Sigma)/10% FCS/5 mM CaCl 2 for 2 h at 37˚C and 10% CO 2 while shaking. 1 h prior to infection, 1, 2, or 5 μg/ml ANXA5 (origene) were added to the bacteria. Cells were infected with an MOI of 100 for 20-24 h at 37˚C and 5% CO 2 . QuantiBlue (Invivogen) was used for quantification according to the manufacturer's instructions. After 30-120 min, the OD was measured at 655 nm using the Clariostar plate reader (BMG lab tech). As a positive control, cells were also incubated with 1 μg/ml of E. coli LPS (K-235, Sigma). For analysis, the value of mock infected cells was subtracted from all other values. Samples were measured in duplicates.
HiBiT assay. AGS [LgBiT] cells were seeded in a black, clear bottom 96-well plate (4titude) one to two days prior to infection without selection antibiotic. Bacteria, expressing HiBiT-CagA, were pre-incubated at an OD 550 of 0.019 in 500 μl of F12 (Sigma)/5% FCS/5 mM CaCl 2 for 2 h at 37˚C, 10% CO 2 and 100 rpm. 1 h prior to infection, 2, 5 or 10 μg/ml ANXA5 (origene) were added to the respective cultures. The cells were infected with an MOI of 25 and incubated for 2.5 h at 37˚C and 5% CO 2 . After discarding the supernatant, 50 μl of luciferase substrate (Promega; 0.5 μl substrate, 9.5 μl assay buffer, 40 μl medium) were added and the luminescence signal was measured immediately using the Clariostar plate reader (BMG lab tech) as described previously.
Human biopsy samples. Gastric biopsy material was obtained from a gastrointestinal endoscopy study in Nigeria, which was supported by the Africa Infectiology Initiative of the Deutsche Forschungsgemeinschaft (DFG). Written informed consent of the patient was obtained prior to enrolment, in agreement with the Helsinki Declaration [57].
Statistical analysis. All statistical analyses were performed using the Graph Pad Prism 5 software. When two samples/conditions were compared, an unpaired Student´s t-test was used. When more than two samples/conditions were tested, a one-way ANOVA statistical analysis was performed. Either Bonferroni's multiple comparison test of all columns or Dunnett's multiple comparison test against a control column was used as post hoc test. Statistical analysis of two or more groups was performed using two-way ANOVA with Bonferroni's post hoc test. P-values < 0.05 were considered as significant.
Supporting information S1 Fig. ANXA5 binding characteristics of different H. pylori wt strains. To assess ANXA5 binding of different H. pylori strains (type I and type II), bacteria were incubated with ANXA5-A647 (red line) or left untreated (grey) and subsequently analyzed by flow cytometry. A representative histogram is shown for each strain. Median FI of the ANXA5-positive group was determined as described. pylori infected patient (N2-97-A1) was stained for H. pylori (magenta) and DAPI (blue). For ANXA5, only the secondary antibody (green) was applied to control for unspecific binding. Confocal image was taken as a z-stack and the three-dimensional projection was reconstructed. B) Confocal image of gastric tissue of an H. pylori infected patient (N2-97-A1). Section was stained with secondary antibodies only (goat anti-mouse IgG-Alexa-488 and goat anti-rabbit IgG-Alexa-555) and DAPI to control for unspecific binding. (TIF) S1 Table. Bacterial strains used in this study. (DOCX) Fig 2E. Oneway ANOVA (p< 0.0001) was performed. Subsequently, Dunnett's multiple comparison posttest was performed against the control columns H. pylori P12wt (as positive control for ANXA5 binding) and C. jejuni (as negative control for ANXA5 binding). All bacteria except N. gonorrhoeae N356 showed a highly significant reduction of ANXA5 binding when compared to P12wt. Only N. gonorrhoeae (N356, N302 and N309), S. aureus, S. pneumoniae and M. catarrhalis (25238) showed a significant or highly significant difference to the negative control, C. jejuni, and therefore are considered to bind ANXA5. (DOCX) S3 Table. Antibodies used in this study. (DOCX) S1 Movie. H. pylori in human gastric tissue imaged in diverse rotated angles, for the distribution of annexin on the membrane surface of H. pylori, as shown in Fig 7G. (AVI)