Functional and intricate interaction network connecting Helicobacter pylori Cag Type 4 Secretion System surface proteins with outer membrane proteins HopQ and HopZ

35 The Helicobacter pylori cag pathogenicity island (cagPAI) encodes a complex type IV secretion system 36 (CagT4SS) which is an important virulence factor of H. pylori. Recently, structural detail on the CagT4SS 37 has been substantially improved by cryo-EM. However, important structural and functional 38 information, in particular on protein interactions between T4SS surface proteins, and of T4SS surface 39 proteins with other proteins, is missing. In the present study, we followed the hypothesis that H. pylori 40 T4SS external proteins may form a surface protein assembly, together with other, non-CagT4SS 41 proteins, which may also be essential for T4SS function. Using interaction screens of H. pylori CagT4SS 42 surface proteins, followed by biochemical and functional characterization, we have enhanced the 43 knowledge on protein-protein interactions of CagT4SS extracellular proteins. This also includes newly 44 identified interactions of CagT4SS surface proteins, for instance the VirB2 homolog CagC, the VirB5 45 homolog CagL and the surface protein CagN of unknown function, with outer membrane proteins 46 HopQ and HopZ. We have further identified and quantitated direct interactions of T4SS surface 47 proteins with outer membrane proteins HopZ and HopQ, which play a role in T4SS functions, and of 48 both HopZ and HopQ with themselves and with host cell factors CEACAM and integrin. Furthermore, 49 we determined an influence of pH on interactions between HopQ/HopZ and CagT4SS components. 50 Utilizing protein tag insertions in H. pylori, we detected surface-exposed association of HopQ and HopZ 51 with T4SS components on bacteria without or with (for HopQ) human gastric epithelial cells. 52 Functionally antagonistic roles of HopQ and HopZ were uncovered in T4SS-dependent early pro-53 inflammatory human epithelial cell activation. In summary, we identified a network of interactions 54 between H. pylori outer membrane proteins and CagT4SS surface proteins and characterized them as 55 functionally important for transport processes. This will help to refine structural and functional details 56 regarding surface-exposed proteins of the CagT4SS. 57 58 59 60 61 62 63 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 3

64 Helicobacter pylori is a Gram-negative human stomach pathogen, which is strongly linked to the 65 causation of several gastric diseases in humans (1, 2). H. pylori infection is primarily acquired during 66 early childhood via fecal-oral or oral-oral transmission and affects approximately half of the world 67 population (3-6). The consequences of H. pylori infections depend on a number of environmental, 68 bacterial and host specific factors (3), and most infected persons never develop overt disease. 69 Nonetheless, about 15% of affected patients will develop serious long-term health consequences. Not 70 only do studies show a link to the development of gastric cancer (7, 8) but also the enhanced risk of 71 severe gastritis, peptic ulcer disease, gastric cancer, and MALT lymphomas (6). 72 H. pylori has co-evolved with its human host for at least 100,000 years, into different ethnic and further 73 admixed bacterial populations (9, 10). At some point in time, it has acquired, from an unknown source, 74 one of its major immune modulators, the Cag type IV secretion system (CagT4SS), a complex bacterial 75 membrane transport system (11-15). The individual virulence of H. pylori strains and severity of disease 76 caused are closely related to the presence of the CagT4SS (8, 10, 16). Together with its main effector 77 protein CagA, the T4SS is encoded on the cag pathogenicity island (cagPAI ((11, 17)), present in about 78 70% of strains and highly genetically variable between strains (10). The H. pylori cagPAI comprises 79 about 28 genes, roughly 13 of which are coding for homologs of the Agrobacterium tumefaciens Vir 80 T4SS model system (18). The remaining genes code for additional H. pylori-specific T4SS components 81 (16, 19-21) and also comprise genes of unknown function (22). 82 A high extent of genetic variability is found between different T4SS (reviewed in (18, 23, 24)), whose 83 paradigm is the VirB/D system of the plant pathogen A. tumefaciens (25). More complex Type 4 84 systems are, for instance, the Dot/Icm system (26) of Legionella pneumophila and the H. pylori CagT4SS 85 (27, 20, 28). The precise architecture of the various T4SS protein assemblies and machineries is not 86 fully understood yet, despite a growing number of excellent high-resolution Cryo-EM (18, 21, 28 - 30), 87 and protein-protein interaction studies (31, 32). As those findings show, the core structure of the H. 88 pylori CagT4SS associated with the bacterial outer membrane (outer membrane core complex, OMCC), 89 is composed of mainly five proteins (CagY, CagX, CagT, CagM and Cag3), all present in multiple copies 90 (30, 33). The CagT4SS OMCC (OMCC Cag)(18, 33) is localized between the bacterial inner and the outer 91 membranes, and CagY spans both the inner and outer membranes (33) and is therefore not restricted 92 to the OMCCCag. As far as it has already been visualized, the CagT4SS inner membrane complex (IMC) 93 consists of three ring structures around a central channel, which are, among other proteins, formed 94 by CagY, and contains the ATPases Cagβ, CagE, and Cagα at the inner membrane (20, 28). 95 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 4 In addition to the structurally important, membrane associated, IMC and OMC(C), subassemblies of 96 the H. pylori T4SS (30, 31, 34), some Cag proteins seem not to possess a structural role so far (22, 35-97 37). We and others previously characterized a CagT4SS protein, CagN, of unknown function (38), which 98 is partially surface-associated in the CagT4SS (21) and interacts with CagM, a component of the 99 OMCCCag (24, 33). The proteins CagL, CagC, CagH and CagI were reported before to be surface-100 associated in the bacteria (35, 36, 39). Of those, CagL, CagI and CagH are already known to interact 101 with each other (36, 40). Each of those proteins influence the expression as well as the stability of the 102 respective other proteins and are essential for the formation of a connecting structure that was 103 detected in scanning electron microscopy (EM) on the surface of H. pylori in contact with human gastric 104 epithelial cells (37, 40, 41). Although CagC alone appeared not to be strictly required for the formation 105 of such a structure (39), CagC was also located on the surface of H. pylori by EM and fluorescence 106 microscopy, and, similar to the other CagT4SS surface proteins, is essential for the translocation of 107 CagA and the induction of pro-inflammatory IL-8 secretion in human host cells (35-37, 42, 43). 108 The Cag proteins CagL and CagI, the oncogenic effector CagA, and the OMCC protein CagY interact 109 directly with host cell integrin receptors, for instance with α5β1 integrin (37, 44-46). While the integrin 110 interaction seems to be dispensable for the so far known T4SS functions, the interaction of the H. pylori 111 outer membrane autotransporter protein HopQ with human cell surface receptors carcinoembryonic 112 antigen-related cell adhesion molecules (CEACAMs) was found to be important for CagA transport into 113 human cells (15, 47). HopQ is one member of a large family (> 30 members) of outer membrane 114 proteins (OMPs) in H. pylori, making up roughly 4% of the H. pylori genome (48). The OMP family also 115 includes H. pylori’s main adhesins BabA and SabA (48-50). The HopQ proteins are subdivided into two 116 closely related but sequence-divergent allelic families, HopQI and HopQII (51). While the majority of 117 H. pylori strains carries only one HopQ allele, of which HopQI is more common, some strains carry both 118 HopQ alleles, HopQI and HopQII (52). The HopQI genotype is significantly linked to a cagPAI- and 119 therefore CagT4SS-positive status, as well as to the presence of the type s1-m1 VacA (which has higher 120 vacuolating activity and disease association than other VacA types (53, 54)) and is dominant in Asian 121 strains (47, 55, 56). The interaction of both HopQ alleles, HopQI and HopQII with human CEACAM was 122 studied previously in some detail (56, 57). This interaction, which can also dissolve CEACAM dimers on 123 the host cell side, has been shown to mediate both the induction of pro-inflammatory signaling and 124 translocation of CagA into the human host cell (47, 58, 59). 125 A highly sequence-related paralog of the same family of H. pylori OMPs is HopZ (60). Like HopQ, it 126 presents as two allelic genotypes, hopZI and hopZII, which are usually alternatively present in H. pylori. 127 The two hopZ alleles belong to the most strain-variable outer membrane proteins of H. pylori , 128 indicating that they are under strong diversifying selection (60, 61). While large parts of the N- and C-129 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 5 terminal regions of the HopZ amino acid sequence, which constitute the common beta-barrel structure 130 of the OMPs embedded in the outer membrane (48, 56), are highly conserved between different 131 strains and between different OMPs, especially the predicted surface-exposed segments in the 132 sequence of the mature outer membrane proteins are rather variable between strains, although they 133 belong recognizably to one of the two HopZ types (60). The most notable difference between the 134 protein sequences of the two HopZ alleles is a 20 amino acid stretch of HopZI (aa179 to aa198) that is 135 missing in HopZII (61). Other H. pylori OMPs are much less strain variable (56). 136 Potential host cell receptor(s) of HopZ are unknown, and conflicting results were reported regarding a 137 role of HopZ in human cell adhesion (60, 61). Interestingly, the hopZ coding region carries a phase-138 variable CT dinucleotide-repeat in its 5’ coding mRNA region that may function as an ON/OFF switch 139 for protein expression (60, 61). No clear correlation between hopZ status, cagPAI and clinical disease 140 was previously identified. While, in contrast to HopQ, no statistically significant correlation between 141 hopZ allele and cagPAI-positivity was reported so far, however a trend for hopZI to genomically co-142 occur with cagPAI-positivity, and a significant correlation for hopZI with the more active vacAs1m1, 143 cagPAI associated, genotype of the VacA toxin exists (60). hopZ (both alleles) was found frequently in 144 CT repeat-OFF status during chronic infection, but underwent a CT-ON-switch in one vaccine study 145 upon acute infection of healthy volunteers (62) 146 We hypothesized from the available evidence that further H. pylori Cag surface proteins, that we also 147 term “Cag outer proteins”, can interact with each other in a more complex network. A possible protein 148 assembly at the outer surface of the CagT4SS machinery may have a contact and conduit function 149 towards human host cells. In addition, we assumed that a functional and a direct physical interaction 150 of HopQ, for which a relevance for CagT4SS functionalities was shown before, and HopZ may exist with 151 CagT4SS proteins. Therefore, the initial aim of this study was to determine and characterize so far 152 unknown/novel protein-protein interactions of the H. pylori CagT4SS surface proteins, with the goal to 153 better understand the potential interaction network, especially of a proposed Cag surface protein 154 assembly. Bacterial Two-Hybrid (BACTH) assays were used to perform an interactome screen for 155 potential novel interaction partners in- and outside of the T4SS, in particular with respect to Cag 156 surface proteins and the two H. pylori outer membrane proteins HopQ and HopZ. Thereby, we have 157 identified novel interactions at the proposed surface assembly of the CagT4SS, and we obtained further 158 evidence for direct physical binding of Cag proteins with the two outer membrane proteins HopQ and 159 HopZ. The interaction network that results of the present study, jointly with functional changes that 160 we obtained with bacterial mutants, suggests new characteristics of the CagT4SS and a novel functional 161 role of HopZ with respect to the CagT4SS. 162 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 6

163 164 Bacterial Two-Hybrid (BACTH) assay reveals novel interactions of outer CagT4SS proteins in a 165 partially unbiased screen 166 In order to screen, in a partially unbiased manner, regarding the selected combinations of protein 167 expression constructs in the test array, for protein-protein interactions between members of the outer 168 proteins of the H. pylori T4SS, bacterial two hybrid assays were performed to detect novel binary 169 interactions. For this purpose, we used expression constructs for proteins of the CagT4SS OMCC and 170 Cag surface proteins. The existing BACTH set-up, which we had designed before for CagM interaction 171 analysis (22) was expanded by 39 new plasmids (different insertions in the four available BACTH 172 vectors; see Methods) encoding the proteins CagH, CagI, CagL, CagC, CagA, HopQI (S1 Fig), HopQII (S1 173 Fig), HopZI (S1 Fig), HopZII (S1 Fig) and human CEACAM1. 174 After screening 327 new plasmid combinations, in comparison with common negative and positive 175 controls (21) (S2 Table, Fig 1A, and Methods), we identified 8 novel CagT4SS protein-protein 176 interactions not previously known. All obtained -galactosidase results higher than 1.5-fold of the 177 respective negative control values were scored as positive (Methods) and are included in the all-results 178 matrix (Fig 1A) as fold-changes over the negative control of the respective assay. 179 Within the CagT4SS protein interaction network itself, and with a focus on the Cag proteins that might 180 be located at the surface of the T4SS, the novel set of interactions revealed included, among others, 181 the self-interactions of CagC and CagA, as well as the heterologous interactions of CagC with CagA, 182 CagL and CagI with CagM, CagI with CagH, and CagN with both CagL and CagI, respectively (Fig 1A, 1B; 183 S2A, S2B, S2C, S2F Fig). Some known direct interactions of CagT4SS surface proteins, such as those 184 between CagL and CagI, CagA and CagL, CagA and CagI, determined previously by biophysical methods 185 (46), were confirmed. Significantly positive, partially novel, BACTH interactions of the translocated 186 effector CagA with CagC (VirB2 homolog) or the CagL (VirB5 homolog), as well as CagC-CagC and CagA-187 CagA self-interactions, are separately depicted as -galactosidase assays (Fig 1B). 188 189 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 7 190 Fig 1. Bacterial-Two-Hybrid-Assay uncovers novel interactions between proteins of the H. pylori CagT4SS 191 surface proteins. A) intensity matrix of unbiased BACTH results ( -galactosidase assays, determined as Miller 192 units) to detect and quantitate novel protein-protein interactions between the CagT4SS outer proteins. Red color 193 indicates no detectable interaction above the negative control (ratio of 1 to the negative control), green shades of 194 color indicate detectable interactions (significantly above background levels; detection limits were defined as below 195 1.5-fold the average of the negative control background over all experiments (Methods)); the values in the matrix 196 boxes correspond to the ratio between the Miller unit results of each combination measured over the negative 197 control (fold change); for all tested interactions with positive outcome, exclusively the BACTH plasmid combination 198 for each interaction pair with the highest quantitative outcome of interaction is included in the matrix. If all assayed 199 interactions were negative, one negative result is shown. Negative and positive controls were run alongside each 200 separate assay and are explained in the Methods. Previously reported interactions of CagT4SS surface proteins 201 (33, 36, 40, 67) are boxed. Tabular depiction of the precise plasmid combinations used in the matrix is shown to 202 the right. Additional assay results and Western Blots are shown in S2 Fig). B) BACTH results of -galactosidase 203 activity of selected new interactions between CagT4SS outer proteins, depicted as bar graphs in absolute values 204 (Miller units). All assays were performed and quantitated in three-fold replicates. The first listed interaction partner 205 on the x-axis in each combination was expressed as an N-terminal fusion in the pUT18 BACTH plasmid; the second 206 listed interaction partner in each pair was expressed in pKT25 or pKNT25 (see also legend table and Methods). 207 The legend table included in panel B contains the full information about the BACTH plasmid fusions and 208 combinations used in each assay. Statistical analysis for significant difference was performed for the comparison 209 between each sample versus the negative control (Student’s t-test, two-tailed unpaired), and significant p values 210 are shown above each sample. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. CagC was fused as a shorter variant 211 missing a putative leader peptide sequence (aa1-29); CagC* is a fusion of the same N-terminally truncated protein 212 product which is expressed as a C-terminal fusion in BACTH plasmid pKT25. CagC-N designates an N-terminally 213 truncated version of CagC (aa1-44) and was cloned as a C-terminal fusion in pKT25. CagA was expressed as an 214 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 8 N-terminal fusion of a C-terminally truncated variant, only comprising the N-terminal domain consisting of aa1-892 215 in either pUT18 or in pKT25. CagA full-length was also cloned and tested, however was poorly expressed and 216 always gave negative results in BACTH. All used expression constructs including CagL, CagI and CagH are 217 described in S2 Table. Selected Western blots (e.g. CagC*) of the BACTH samples, which demonstrate protein 218 expression of fusion proteins in E. coli, are shown in the supplementary figures (S2 Fig). Detailed BACTH analyses 219 of CagM interactions with Cag surface proteins are shown in the supplementary figures (S2 Fig). 220 221 BACTH screen reveals novel interactions between proposed CagT4SS surface proteins and H. pylori 222 outer membrane proteins HopQ and HopZ 223 Owing to the previous characterization of a direct functional interaction between HopQ and host 224 CEACAM proteins, strongly influencing CagT4SS functionality, including CagA transport and 225 inflammatory processes by various transported molecules in the target human cells (47, 56), we also 226 included both HopQI and HopQII (expressed without the transmembrane β-barrel) in our BACTH 227 screen. We also hypothesized that both HopZI and HopZII alleles, due to the fact that HopZ is a closely 228 related paralog with substantial amino acid sequence and structural homology to HopQ (S1 Fig), are 229 additional candidates for being potential direct interaction partners of components of the H. pylori 230 CagT4SS. Hence, HopZ, in its two allelic types, was also included in our present BACTH test arrays in E. 231 coli. Thereby, we found, in addition to the 8 novel Cag-Cag interactions (above), that both alleles of 232 HopQ as well as HopZ exhibited multiple additional interactions in this assay (Fig 2). HopQ and HopZ 233 showed homotypic and heterotypic self-interactions (except for HopZII with HopQI) (Fig 2A, 2B). In 234 addition, the introduction of HopQ and HopZ as potential new interaction partners of the T4SS proteins 235 revealed 10 novel Hop interactions with specific H. pylori CagT4SS proteins (Fig 2G; S2G Fig). While 236 both Hop proteins showed positive interactions with CagA (Fig 2A, C) and CagL (Fig 2A, D) in various 237 BACTH combinations, assays involving either of both HopZ alleles also revealed additional interactions 238 with CagI, CagC, and CagN (Fig 2A, 2E; S2D, S2E Fig). 239 Due to earlier findings of HopQ interaction with human CEACAMs (56), we also decided to use human 240 CEACAM1 as a positive control for HopQI binding to verify the validity of this assay for detecting the 241 human-bacterial protein interactions. As expected for HopQ due to previous findings (15, 47, 56), 242 HopQ large binding domain (lbd), contained in the cloned construct (amino acids 22 to 468), of which 243 segments between aa100 to aa160 (including clasped loop CL1 and loop CL1-H4 (56) – the latter 244 included in Loop2 as designated in the present study; S1A Fig) have been found to be crucially involved 245 in binding hCEACAM in a co-crystallization analysis (56), interacted with human CEACAM1 in the BACTH 246 set-up (Fig 2A). We also observed the known CEACAM1-CEACAM1 homotypic interaction. By this assay, 247 also three novel hCEACAM1 interactions (with HopZI, CagC and CagL; Fig 2A; S2G Fig) were suggested. 248 249 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 9 250 Fig 2. BACTH and newly devised Bacterial-Three-Hybrid-Assay (BAC3H) characterize novel direct 251 interactions between HopQ and HopZ proteins, the H. pylori CagT4SS surface proteins, and between human 252 CEACAM1, CagT4SS proteins and HopQ/HopZ, predicting an entirely novel interaction network and 253 competitive interactions. A) intensity matrix of BACTH results (-galactosidase assays, quantitated in Miller units) 254 to quantitate selected protein-protein interactions between various proteins of the CagT4SS outer proteins with 255 outer membrane proteins HopQ and HopZ. Red color indicates no detectable interaction above the negative control 256 (ratio of 1 over the negative control), green shades of color indicate detectable interactions; the values in the matrix 257 boxes correspond to the ratios between the Miller unit results measured for each new interaction and that of the 258 negative control (fold change). Positive controls with leucine zipper fusions of both BACTH plasmids were always 259 run alongside. Negative controls, containing the two BACTH plasmid types together but without inserts, and 260 selected pairs of BACTH plasmids with inserts combined with a respective empty plasmid of the complementary 261 compatibility type were also tested alongside (not shown). For all tested interactions with positive outcome, 262 exclusively the BACTH plasmid combination for each interaction pair with the highest quantitative outcome of 263 interaction is included in the matrix. If all assayed interactions were negative, one negative result is shown. Human 264 CEACAM1 self-interaction (homodimer) was used as a control condition for the human CEACAM1 functionality in 265 BACTH. The matrix fields designated with crosses refer to Cag-Cag interactions shown in Fig 1, which are not 266 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 10 relevant for HopQ or HopZ interactions and therefore not included in this panel again. B) BACTH results of -267 galactosidase activity of interactions between HopQ (types I and II) and HopZ (types I and II) as homo- and 268 heterodimers, depicted as bar graphs (absolute values in Miller units). Detection limit of positive interaction was at 269 1.5-fold the negative control value as in Fig 1. C) BACTH results of relevant β-galactosidase activities of interactions 270 between HopQ (types I and II) and CagT4SS outer proteins, depicted as bar graphs (Miller units); D) BACTH results 271 of -galactosidase activity of interactions between HopZ (types I and II) and CagT4SS outer proteins, depicted as 272 bar graphs in absolute values (Miller units). E) Shows a complete graphical network model of novel interactions 273 between H. pylori CagT4SS outer proteins and outer membrane proteins HopQ and HopZ, summarizing all results 274 from Fig 1 and Fig 2; green lines are novel interactions found by BACTH, blue line indicates one single interaction 275 not found in BACTH but verified by plate-binding assay (see S5 Fig), which confirms the potential of the CagC 276 (VirB2) and CagL (VirB5) orthologs of the T4SS to bind to each other. F) Novel Bacterial Three-Hybrid Assay 277 (BAC3H) results of relevant combinations of proteins (results shown in Miller units) suggest competitive interactions 278 between HopQ (types I and II) and CagL as well as between CagM, CagL and CagI. The legend tables included in 279 each of the panels B, C, D, F contain the full information about the plasmid combination used in each assay. 280 Detection limits of the BAC3H were determined by negative controls containing all three plasmids as empty versions 281 which were included in all experiments, and were at or below 1.5-fold of the Miller units of the negative control. 282 Each result consists of three data points from three independent measurements. Statistical analysis for significant 283 difference in B, C, D, F was performed for the comparison between each sample versus the negative control 284 (Student’s t-test, two-tailed unpaired), and significant p values are shown above each sample. *p<0.05; **p<0.01; 285 ***p<0.001; ****p<0.0001. G) Immunoblot from E. coli cell lysates (supernatant (s) and pellet (p) fractions) detecting 286 expression of the proteins assayed in the BAC3H. Blot incubated for detection with α-T25 antiserum (rabbit, 287 1:20,000) αCagM antiserum (rabbit, 1:10,000) and αCagL antiserum (rabbit, 1:20,000). Soluble (s) as well as 288 insoluble (p) fractions of the same samples were tested for each sample to determine solubility of the expressed 289 proteins. Fixed amounts of 10 µg protein were loaded on each lane. Antisera against bacterial GAPDH was used 290 as a loading and fractionation control (as in S2 Fig) and confirmed even loading on the blots. 291 292 Novel Bacterial-Three-Hybrid assay was developed to refine the new model based on CagT4SS and 293 Cag-Hop interactions 294 Next, we established and performed selected bacterial three-hybrid (BAC3H) assays for novel CagT4SS 295 and HopQ interactions, with the aim of achieving a better understanding of a selected set of protein-296 protein combinations. By co-transforming a third expression plasmid without an interactive protein 297 fusion (pAB148 derivative (Methods)), we aimed to indicate a potential competitiveness or synergistic 298 interactions, through a decrease or increase in comparison to the ß-galactosidase activity measured in 299 parallel in the corresponding two-hybrid combinations. 300 As shown in Fig 2F, the introduction of CagL as a third interaction partner to the HopQI - HopQI or 301 HopQII – HopQII self-interaction led to a significant reduction for both interactions, as quantitated by 302 ß-galactosidase activity (Fig 2F). This suggests that CagL can weaken HopQI (or HopQII) self-interactions 303 (Fig 2F). For testing the second hypothesis that the interaction between the two surface proteins CagI 304 and CagL interferes with the CagI - CagM interaction, we combined the latter three proteins by adding 305 non-tagged CagL, which indeed resulted in a strong reduction of the CagI-CagM galactosidase 306 measurement (Fig 2F), while soluble CagI presence was increased in Western blot (Fig 2G). Although 307 this has to be interpreted with caution, the apparent competition may imply that an alteration of 308 interactions occurs for selected transported Cag surface proteins, including CagI, CagL, CagH (40): we 309 suggest they may undergo transient CagM interactions (22), which can accompany their transport 310 through the system, before they coalesce on the bacterial surface. 311 312 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 11 Direct physical interactions of CagL, CagN, CagA, HopQI, HopQII, HopZI, and HopZII were verified and 313 quantitated in pair-wise combinations using Biolayer Interferometry (BLI) and plate-binding assays 314 In order to better characterize and quantify the newly discovered protein-protein interactions, the 315 genes coding for proteins CagL, CagN, CagA, HopZI, HopQI and HopQII (the latter cloned in two 316 variants, from strains PNGhigh12A [ cagPAI-positive strain; designated as HopQII throughout the 317 manuscript] and K26A [hpAfrica2 strain from Khoisan population, cag-negative] 318 (https://enterobase.warwick.ac.uk/)) were cloned into expression vectors, expressed in E. coli Rosetta 319 pLysS, and purified by affinity chromatography (S3A Fig and Methods) (22). The phylogeographic 320 population hpAfrica2 is always devoid of a cagPAI and thought to predate the acquisition of the cagPAI 321 by H. pylori (10) and is therefore an interesting variant to test for interactions of HopQ(II) with Cag 322 proteins. The hopQII allele of an hpAfrica2 strain should therefore not be affected by prior selection 323 pressure exerted by possible interactions with CagT4SS components. CagC was cloned and expressed 324 as a GST fusion for interaction assays (Methods), to improve its solubility, and purified alongside non-325 fused GST (negative control). 326 The purified recombinant proteins (S3A Fig) were used for Biolayer Interferometry measurements 327 using amine-reactive 2 nd generation biosensors with covalent coupling of the ligand. BLI (Octet) 328 measurement of protein-loaded biosensors for association and dissociation of an analyte in solution 329 added in increasing concentrations also permitted us to calculate KD values for the tested interactions 330 (Table 1). We performed those for newly detected CagT4SS protein interactions and also included 331 human proteins CEACAM1 and integrin- 51 (commercial purified proteins) for selected interaction 332 partners (Fig 3A, 3B). We tested most interactions at two different pH values, pH 6 and pH 7, since the 333 natural habitat of H. pylori deep in the gastric mucus alternates naturally between neutral and (slightly) 334 acidic (63, 64). Table 1 shows the calculated KD-values for the 23 protein-protein interactions tested in 335 BLI at both pH 7.0 and pH 6.0, assuming a 1:1 interaction model. 336 337 Protein used for immobilization Analyte protein KD at pH 7.0 [nM] KD at pH 6.0 [nM] HopQ and HopZ interactions (BLI) HopQI HopQI 76.00 25.85 R2 = 0.96 R2 = 0.99 HopQII_PNGhigh12A HopQII_PNGhigh12A 5.49 6.82 R2 = 0.99 R2 = 0.99 HopQII_PNGhigh12A HopQI 105.20 5.89 R2 = 0.98 R2 = 0.98 HopQII_K26A HopQII_K26A 0.000016 6.90 R2 = 0.99 R2 = 0.99 HopZI HopZI 98.66 31.58 R2 = 0.97 R2 = 0.85 HopQ and HopZ interactions with hCEACAM1 and hα5β1-integrin preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 12 hCEACAM1 HopQI 20.12 20.23 R2 = 0.97 R2 = 0.93 hα5β1-integrin HopQI 40.45 36.79 R2 = 0.99 R2 = 0.99 hCEACAM1 HopQII_PNGhigh12A 12.90 5.26 R2 = 0.99 R2 = 0.99 hα5β1-integrin HopQII_PNGhigh12A 8.171 5.91 R2 = 0.96 R2 = 0.98 hCEACAM1 HopQII_K26A 0.55 0.10 R2 = 0.99 R2 = 0.98 hα5β1-integrin HopQII_K26A 0.40 0.075 R2 = 0.98 R2 = 0.99 hCEACAM1 HopZI 9.89 38.08 R2 = 0.82 R2 = 0.99 hα5β1-integrin HopZI 9.38 31.05 R2 = 0.99 R2 = 0.99 HopQ and HopZ interactions with T4SS proteins HopQI CagA 54.68 17.79 R2 = 0.96 R2 = 0.89 CagL HopQI 1.72 0.73 R2 = 0.98 R2 = 0.99 HopQI CagN 33.21 112.2 R2 = 0.97 R2 = 0.97 HopQII_PNGhigh12A CagA 18.2 104.1 R2 = 0.94 R2 = 0.91 CagL HopQII_PNGhigh12A 22.58 110.10 R2 = 0.99 R2 = 0.99 HopQII_PNGhigh12A CagN 131.9 96.71 R2 = 0.94 R2 = 0.85 CagL HopQII_K26A 0.1 0.66 R2 = 0.92 R2 = 0.99 HopQII_K26A CagN 111.8 70.7 R2 = 0.95 R2 = 0.96 HopZI CagA 56.97 22.51 R2 = 0.92 R2 = 0.85 HopZI CagN 110.80 40.64 R2 = 0.99 R2 = 0.97 338 Table 1. Kinetic parameters of HopQ and HopZ homo- and heterodimerization and their interactions with T4SS 339 proteins CagL, CagN, CagA as well as host cell factors hCEACAM1 and hα5β1-integrin measured by Biolayer 340 Interferometry (BLI) using the Octet system (Methods). A 1:1 model was used for the data fitting and K D 341 determination. Two different variants of HopQII were purified and used, one variant cloned from the cagPAI-positive 342 strain PNGhigh12A, from Papua New Guinea, and from strain K26A, a Khoisan hpAfrica2 isolate from an ancient 343 H. pylori population which does not harbor a cagPAI (10). Only interactions that had shown a positive outcome in 344 at least one plasmid combination of the BACTH assay and for one of the HopQ or HopZ allelic variants were tested 345 in Octet. In addition, hα5β1-integrin interactions of HopQ and HopZ that could not be tested using BACTH, were 346 investigated. Coefficient of Determination (R 2) values (most were higher than 0.9) were calculated as a statistical 347 measure to determine the data fit to the model. Due to remaining protein degradation products after purification 348 (S4A Fig), KD determination of some interactions may not be reliable. 349 350 The BLI measurements permitted a quantitative assessment of important novel interactions revealed 351 in the BACTH screen. They also clearly demonstrated that the tested pH value has a significant 352 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 13 influence on the binding affinity of many interactions including T4SS proteins and HopQI or HopZII. 353 Many tested interactions were stronger at pH 6.0 when compared to pH 7.0 (higher Kd at pH 7) (Table 354 1; Fig 3A, 3B). For HopQII interactions, this difference was less apparent. HopQI and HopQII self-355 interactions and hetero-oligomerization were quantitated, with high affinities, as well, but showed less 356 pH-dependence in BLI, in particular the HopQII homotypic interactions. (Table 1). Results of analytical 357 size-exclusion chromatography (SEC) performed for purified recombinant HopQI protein further 358 supported the formation of homotypic oligomers (possibly including tetramers) in solution (S3B Fig). 359 HopQI was confirmed in BLI to interact directly with purified CagN, CagL (medium affinities), and with 360 CagA at high affinities (Table 1; Fig 3A). HopQII self-interaction and HopQI-CagN interactions were pH-361 dependent and stronger at pH 7, while HopQI-CagA and -CagL interactions were determined to be of 362 higher affinity at pH 6 (Table 1). Moreover, HopZI self-interaction was quantitated as high-affinity. 363 HopZI had lower affinities to CagA and CagN than HopQI, which were stronger at pH 6 (Table 1 and Fig 364 3A). The affinities of interactions with CagL and CagA were significantly stronger for HopQII 365 (PNGhigh12A) as compared to HopQI at the same conditions (pH 7). Both the homo-dimerization assay 366 of HopQII (variant PNGhigh12A) as well as the hetero-dimerization of HopQII (PNGhigh12A) and HopQI 367 in BLI had higher calculated affinities than the HopQI homotypic dimerization (Table 1; S4A Fig). 368 As for the additionally BLI-tested interactions with human receptor proteins, HopQI had a lower affinity 369 for both CEACAM1 and integrin- 51 than HopQII and HopZI (Table 1). While HopQI and 370 HopQII_PNG12A showed no or little pH-dependency for integrin or CEACAM binding, CEACAM1 and 371 integrin binding for HopQII_K26A (from cagPAI-negative hpAfrica2 strain K26A; see also below) and 372 HopZI were pH-dependent and exhibited lower or higher affinities, respectively, at neutral pH 7 (Fig 373 3B; Table 1). Since the non-membrane portion of HopZII which was also cloned and expressed for 374 purification in E. coli (S1 Fig) was extremely insoluble and therefore difficult to purify in sufficient 375 amounts, we did not perform BLI with this protein. To complement this deficit, we developed a multi-376 well plate assay (Methods) to test binding with smaller amounts of protein, which also allowed for a 377 direct comparative testing. CagC interaction with CagL, for which we could also not perform BLI due to 378 low yields of purified CagC, was likewise confirmed using a plate assay (S4B Fig). 379 380 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 14 381 Figure 3: Biolayer Interferometry and selected plate-binding assays to detect concentration-dependent 382 binding of purified recombinant HopZ and HopQ (variants) to T4SS outer proteins, human gastric epithelial 383 cells and human proteins (CEACAM1, integrin-51). Assays detect differential binding of HopQ and HopZ and 384 their type variants, and pH-dependent substrate binding. A) and B) Overview bar graphs of biolayer interferometry 385 assays quantitate different affinities of HopQ or HopZ interactions with H. pylori CagT4SS proteins (A), as well as 386 HopQ or HopZ interactions with human CEACAM1 or integrin 51 (B), and demonstrate that some affinities differ 387 between pH 7 and pH 6. A) Different HopQ and HopZ types were purified and assayed in BLI against T4SS outer 388 proteins CagL, CagA, and CagN. B) HopQ or HopZ interactions with human CEACAM1 or integrin51 were tested 389 by BLI. The interactions between ancient HopQII (strain K26A) or HopZI and either CEACAM1 or integrin51 were 390 pH-dependent, while the interactions between modern HopQI(26695) or HopQII(PNGhigh12A) types and the 391 human receptors were rather not. Ancient type HopQII_K26A (cloned from Africa2 strain without a cagPAI) showed 392 high-affinity interactions with human CEACAM1 or integrin-51. The first-named proteins on the X axis were bound 393 to the sensors, and the second-listed were used as soluble analytes in BLI. See full results of BLI assays in Table 394 1. We did not assay CagC and HopZII in BLI, which were only purified in small quantities due to low solubility. 395 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 15 Panels C) and D) illustrate results of concentration-dependent binding in plate-assays for all four purified HopQ and 396 HopZ types to human CEACAM1 (CEACAM1) at pH values of 7 and 6. For all those conditions, HopZI bound best 397 to CEACAM1, while HopQI had the lowest binding characteristics in this assay system. pH 7 enhanced HopZ 398 binding to CEACAM1 over pH 6. Panels E) and F) show results of concentration-dependent binding in plate-assays 399 for all four HopQ and HopZ allelic variants to human integrin- 51 at pH 7 and pH 6, respectively. HopZI showed 400 the strongest binding in both conditions and increased binding at pH 6. HopZII also exhibited stronger binding at 401 pH 6 than at pH 7, while HopQI and HopQII binding to integrin was less influenced by pH variation. Panels G) and 402 H) illustrate concentration-dependent comparative quantitative plate-binding assays of three different HopQ variants 403 (HopQI, HopQII modern [PNGhigh12A] and HopQII ancient [K26A]) to human CEACAM or integrin- 51, 404 respectively. All plate-binding assays were at least performed twice independently in triplicates. I), J) depict results 405 of comparative multi-well plate-binding assay of all four HopQ and HopZ allelic variants to gastric epithelial cells 406 (AGS) at two different pH settings; HopQI and HopZI showed the strongest binding at pH 7; while HopQI binding 407 was stronger at pH 7 compared to pH 6, it bound better than HopZI at pH 6. HopQII showed the weakest cell binding 408 under all conditions. For all panels: recombinantly expressed HopQI was cloned from strain 26695, modern HopQII 409 from strain PNGhigh12A (here designated as: PNG12A), HopZI from strain Su2 and HopZII from strain 26695 410 (Methods); purification results and quantitation for all used proteins are shown in supplemental materials, S4 Fig; 411 recombinantly expressed CagN, CagA (aa1-892), and CagL were cloned from strain 26695. For the plate-binding 412 assays, we used bovine serum albumin (BSA) as a negative control analyte against all bound proteins, which did 413 not indicate binding (not shown). Statistical differences in C, D, E, F, G, H, I, J were calculated using One-way 414 ANOVA, followed by Tukey’s post-test, and are included in each panel. *p<0.05; **p<0.01; ***p<0.001; 415 ****p<0.0001; ns = non-significant. 416 417 Multi-well plate-binding assays also confirmed, concentration-dependently, the interactions of all four 418 HopQ and HopZ alleles with human CEACAM1 and integrin (Fig 3C, 3D, 3E, 3F), which were not testable 419 for the integrin heterodimers with Hop proteins in BACTH. Direct comparative assays in multi-well 420 plate format performed for the four HopZ and HopQ variants with human CEACAM1 showed the 421 strongest interaction of HopZI, followed by HopZII and HopQII (PNGhigh12A), and finally HopQI (Fig 422 3C, 3D). This was the case at both tested pH values, with slightly higher HopZ binding detected at pH 423 7. Similar results as for hCEACAM1 in the plate-binding assays were also obtained for integrin (α5β1) 424 binding to both HopZ and HopQ variants, with the highest binding determined for HopZI and HopZII, 425 compared to lower binding for both HopQ allelic types (Fig 3E, 3F). The highest binding for the HopZ 426 variants to integrins was preserved at pH 6, while all of the four Hop proteins showed increased binding 427 at pH 6 compared to pH 7. The latter results, in addition to the newly discovered Cag to Cag protein 428 and Cag protein to HopZ or Cag protein to HopQ interactions, also add a complex layer of multiple 429 integrin-CEACAM-Cag-Hop interactions (S2G Fig) to the proposed Cag surface protein assembly 430 constituents. 431 432 Archetypal HopQII from Africa2 strains reveal higher-affinity protein-protein interactions to human 433 receptors integrin and CEACAM in comparison to current, Cag-associated, HopQI allelic variants 434 Different allelic variants of HopQ from different H. pylori populations exist, which have evolved in the 435 presence or absence of a CagT4SS. In order to test for potential differences between them, we tested 436 both, a current (“modern”) hopQII allele, cloned from a CagT4SS-positive H. pylori strain 437 (PNGhigh_12A, isolated from Papua New Guinea), and the ancestral (“ancient”) sequence-diverse 438 hopQII, from the cag-negative hpAfrica2 population strain K26A, for interactions with Hop and Cag 439 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 16 proteins and human receptors using BLI and plate-binding assays. Using BLI, binding of ancient HopQII 440 K26A to human receptors integrin and CEACAM1 was found to have higher affinities than binding of 441 modern HopQII or HopQI to these receptors (Table 1), and more pH-dependent for human receptor 442 binding (Fig 3B) than modern HopQII (PNGhigh12A). In multi-well plate binding assays, performed 443 comparatively for the three purified HopQ variants at pH 7, ancient HopQII bound much better to 444 integrin (Fig 3G) and CEACAM (Fig 3H) than modern HopQII and HopQI, corroborating the BLI assays. 445 Hence, our results suggest that ancient HopQII has higher affinity to human cell surface receptor 446 proteins than modern, presumably CagT4SS-adapted, HopQII. Modern HopQII exhibited an 447 intermediate binding phenotype (Table 1; Fig 3F). We therefore suppose (see also discussion), below 448 supported by functional assays, that modern HopQII alleles, if expressed, have sufficiently adjusted to 449 the CagT4SS to functionally complement HopQI in this task, possibly by evolving towards weaker 450 binding to human cell receptors. CagN binding affinities to the different HopQ alleles were not very 451 different in BLI (high nanomolar range), however slightly pH dependent (Table 1), with the highest 452 binding for HopQI at neutral pH. CagL-HopQ binding was more divergent, with very high affinities to 453 ancient HopQII (from cagT4SS-negative strain) at neutral pH, followed by HopQI, and with the lowest 454 affinity towards modern HopQII type (Table 1; Fig 3A). HopQI had lower affinity to CagL at pH 6, while 455 modern HopQII showed similar affinities to CagL at both pH values. 456 457 Differences between HopQ and HopZ in interaction with human cells 458 Since our above-described results of novel single-protein interactions also extended to novel 459 interactions of both HopQ and HopZ towards human cell-surface exposed receptors, we also designed 460 test systems to detect and roughly quantitate binding differences of the different Hop alleles to human 461 cells. For this purpose, gastric epithelial cell lines AGS and NCI-N87 and HEK293-T cells were cultured 462 to confluent monolayers in 96-well plates and preserved by aldehyde fixing. Subsequently, 463 recombinantly expressed and purified HopQ and HopZ variants were co-incubated with the fixed cell 464 layers in different concentrations in order to detect binding. The binding assays were performed at 465 both pH values of 7 (Fig 3I) and 6 (Fig 3J). For AGS cells, HopQI and HopZI bound more avidly to the 466 AGS cell surface than HopZII and modern HopQII (Fig 3I, 3J). HopQI and HopZI bound about equally 467 well in this assay at pH 7 (Fig 3I). At pH 6 (Fig 3J), HopQI maintained highest binding to AGS cells, 468 followed by HopZII, while HopZI showed relatively decreased binding. Highest HopQI binding of all the 469 four tested variants was also confirmed for a second gastric epithelial cell line, NCI-N87 (S4C, S4D Fig), 470 however in this line, HopZII was the second-best binder at both pH 7 and pH 6 (S4 Fig), indicating a 471 host or cell type specificity. For NCI-N87 cells, the effect was different, since all HopQ/HopZ variants 472 seemed to bind more strongly at pH 6. The cell type-specific divergent result points strongly to a cell 473 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 17 type and/or individual donor specificity for HopQ and/or HopZ binding, also in a pH-dependent 474 manner. We predicted that those differences may potentially impact on functionalities with regard to 475 the CagT4SS, which we tested in the functional assays described below. For plate binding assays with 476 HEK cells, which do not have CEACAM1 (47), we found that HopZ (both variants) were binding best, 477 followed by HopQI (S4E Fig). 478 479 Strain- and cagPAI-dependent transcriptional activity and expression of hopQ/HopQ and hopZ/HopZ 480 The isotype-specific binding results (above) and initial surface localization assays of HopQ and HopZ 481 HiBiT-tagged bacteria (Methods) prompted us to ask the question whether strain-specific differences 482 in HopQ and HopZ expression, for both allelic types, exist, which could instruct our strain selection for 483 further assays. This information may seem to be tangential at first glance, but, in the very variable 484 species H. pylori, this information is extremely important to interpret possible reasons for divergent 485 phenotypes between strains, for example with respect to surface localization, human cell interactions, 486 T4SS functionalities in vivo, or microscopic detection and localization. Transcript analysis using qPCR 487 indeed revealed extremely divergent strain-specific transcript amounts of hopQ and hopZ alleles, 488 despite the fact that the strains were all grown to mid-exponential growth phase under identical 489 conditions (S5 Fig). Among the tested strains, Africa2 (cagPAI-negative), P12 and L7 showed the lowest 490 hopQ expression levels (all normalized to 16S rDNA of the respective strain) (S5A, S5B, S5C Fig). Strains 491 SU2, J99, B3 and D3a also had relatively low hopQ levels, in comparison to high expression in strains 492 26695 and N6 that we used for most of our present experiments. For hopZ, the transcript amount 493 variation between strains was similarly strong (S5D Fig). The amounts by qPCR and Western blot were 494 generally at much lower levels than for hopQ. Isogenic deletion mutants in the cagPAI ((13); three 495 strains tested) showed minor changes in hopQ and hopZ transcript amounts (S5B, S5E Fig) in 496 comparison to parental strains. As already known, strain 26695 was switched OFF for hopZ/HopZ by 497 CT repeat length variation (61), but still maintained hopZ transcript by qPCR (S5D, S5E Fig). 498 Western blots using specific custom-produced antisera against HopZ (both alleles) (60) and HopQ (this 499 work) confirmed the strongly variable strain-specific protein expression of HopQ and HopZ at the 500 protein level in various wild type isolates (S5F Fig). As expected, in particular hopZ alleles with CT-501 dinucleotide repeats in predicted OFF status (frame-shifted ORF with early stop codon) did not yield 502 detectable full-length HopZ protein in the Western blots (S5F Fig). Of the tested wild type strains, the 503 cagPAI-negative Africa2 isolate showed the lowest protein expression of HopQ (S5F Fig). 504 505 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 18 High-resolution fluorescence microscopy with dual tagging verifies partial co-localization of OMPs 506 HopQ and HopZ with surface-localized CagT4SS protein CagN 507 The surprising determination of novel interactions between the outer membrane proteins HopQ and 508 HopZ with CagT4SS proteins in vitro prompted us to use high-resolution fluorescence microscopy to 509 verify whether the same proteins may colocalize in bacteria in situ. For this purpose, we initially 510 established single HiBiT-tag insertions and their surface localization in HopQ and HopZ. We generated 511 amino acid alignments and AlphaFold 3 structural models (S1A, S1B, S1C Fig) for full-length HopQ (56) 512 and HopZ to reveal potential surface-associated loops. Three loops in HopQ and two in HopZ were 513 selected for the insertion of tags (S1 Fig) that served for verification and quantitative detection of their 514 surface localization (Methods, S6 Fig). Loop2-insertion in HopQ and LoopY-insertion in HopZ turned 515 out to be the best candidates for surface detection of the proteins (S6A – S6G Fig). Engineered cag-516 negative (deleted cagPAI) isogenic mutants of strain 26695, containing a HiBiT insertion in HopQI (loop 517 2), showed no differences in surface expression of HopQI (S6H Fig), suggesting that Hop surface 518 expression is not per se related to or influenced by cagPAI presence or function. 519 Subsequently, in order to develop strains suitable to identify localized protein-protein interactions in 520 situ, in addition to the single HopZ and HopQ loop-tag fusions (above and Methods), we established 521 internal HiBiT fusions of CagN and CagL, which both have been reported as surface-located “outer” 522 proteins of the CagT4SS (22, 36, 40, 65), in strain N6. CagN-HiBiT was generated as a fusion in an 523 internal loop that has been assumed due to preliminary predictions (AlphaFold) to be surface-located 524 in the protein (S6 Fig and Methods) and was introduced into H. pylori (strain N6) as a plasmid construct, 525 providing about 5-fold overexpression. CagL is also located at the bacterial surface, where it interacts 526 with human integrins (37, 65, 66). Therefore, we designed CagL-HiBiT as a second tagged construct, 527 with the HiBiT tag inserted at the N-terminus of CagL, immediately downstream of its proposed leader 528 peptide cleavage site. CagL-HiBiT was chromosomally integrated in strain N6 using allelic exchange 529 mutagenesis. Both HiBiT insertions were shown by Western blot to be expressed in the respective 530 strains (S6G Fig; not shown for CagL). Both insertions were also verified using the HiBiT surface 531 detection methodology, using reconstituted luciferase (Methods), to be surface-associated in intact 532 bacteria (S6F, S6I Fig). The CagN-HiBiT insertion construct was highly expressed and strongly surface-533 detectable, using HiBiT surface detection (S6F Fig), and also in fluorescence microscopy of non-534 permeabilized bacteria (Fig 4A, 4B), making this construct more amenable for double-tagged analysis. 535 The N-terminal CagL-HiBiT fusion construct yielded markedly lower surface fluorescence intensity in 536 HiBiT surface detection (S6I Fig). Both CagN-HiBiT and CagL-HiBiT strains, albeit at lower intensity for 537 CagL, presented with a similar dot pattern on the surface of the bacterial cell, with three to four dots 538 on average detectable per bacterium in fluorescence microscopy (S6J Fig; and Fig 4). For the interaction 539 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 19 partners HopQI and HopZII, we created V5 epitope tag insertion constructs in the same predicted 540 extracellular loops described above for the HiBiT insertions, Loop2 and LoopY (S1B, S1C Fig), which are 541 predicted not to be relevant for the overall structure of the proteins (S2 Table for mutagenesis 542 plasmids). Using the Hop-V5 insertions and CagN-HiBiT constructs, ultimately, H. pylori strains with 543 double tags were generated by allelic exchange mutagenesis in strain N6 that contain both, a V5 tag in 544 HopQI (Loop2, see alignment in S1A Fig) or HopZII (LoopY; protein alignment in S1A Fig) and the HiBiT 545 insertion in CagN (S6F, S6G Fig, S1 Table). Using those NQ1 and NZ10 double-tagged strains, we first 546 verified that they are expressed and surface–exposed (S6F, S6G Fig), and do not lose CagT4SS 547 functionality with respect to heptose-dependent pro-inflammatory signalling on human epithelial cells 548 (S6J Fig). Subsequently, we performed fluorescence microscopy on double-labelled non-permeabilized 549 bacteria (Fig 4), detecting both CagN and Hop proteins on the bacterial surface in situ. CagN localized 550 on the bacterial surface in dot-like patterns (two to eight per single bacterium). HopQ was clearly 551 membrane associated, as expected for an outer membrane protein. It localized around the entire 552 periphery of the single bacterial cells (Fig 4A) with some dot-like accumulations, which partially co-553 localized with the CagN label (Fig 4E). HopZ, in contrast, showed a rather dot-like distribution in the 554 periphery of the single bacterial cells, with smaller dots also partially co-localizing with tagged CagN 555 dots (Fig 4C, Fig 4E)). Quantitative assessment (paired counts) of HopQ and CagN, HopZ and CagN, 556 respectively, showed that more than two thirds of the dots appeared to co-localize in each 557 combination (Fig 4E). 558 The results showed a partial co-localization of HopQ or HopZ with CagN (T4SS) in the absence of host 559 cells in situ, although HopZ dots were smaller and much more difficult to spot visually. In order to 560 detect closely appositioned, possibly interacting, proteins, rather than mere visual (partial) co-561 localizations, and to verify the results using another method, we employed Proximity Ligation Assay 562 (PLA, see Methods) on the same double-tagged, non-permeabilized, bacterial strains (Fig 4D), followed 563 by fluorescence microscopy. Thereby, we detected HopQ and CagN, and HopZ and CagN, respectively, 564 in very close neighbourhood by the PLA label, supporting the notion of direct interactions between 565 HopQI-CagN and between HopZII-CagN at the bacterial surface (Fig 4D). When we double-labelled 566 HopQI together with CagN, without permeabilization, after bacteria had been co-incubated with 567 gastric epithelial cells for 30 min, we detected an even more significant co-localization of most CagN 568 dots with HopQI in microscopy (Fig 4B; Fig 4F). This was not the case for HopZ, which rather lost the 569 majority of its bacteria-located immunolabel early (<= 30 min) during the cell co-incubation, for 570 reasons yet unclear (not shown). Despite several attempts, we did not succeed in generating doubly-571 tagged bacteria for CagL-HiBiT together with HopQI-V5, which hindered the further evaluation. 572 573 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 20 574 Figure 4: High-resolution microscopy colocalization of CagN and HopQ/HopZ in bacteria without and with 575 coincubation with human gastric epithelial cells. H. pylori (strain N6) bacteria were doubly tagged with V5 576 (inserted in either HopQ or HopZ) and with HiBiT tag in the CagN protein, which is a surface-associated protein of 577 the CagT4SS. The doubly-tagged strains, fixed and non-permeabilized, were immuno-labelled at their surface with 578 both rabbit anti-V5 antibody (1:250) and mouse anti-HiBiT (1:100) antibodies (Methods) and fluorescently-labelled 579 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 21 secondary antibodies (green, anti-mouse Alexa 488-coupled and red, anti-rabbit Alexa564-coupled); high-resolution 580 fluorescence microscopy was performed on the fixed samples. The labels are explained in the figure panels. A) 581 doubly-labelled N6 HopQI-V5, CagN-HiBiT; bacteria only B) bacterial co-incubation of doubly-labelled N6 HopQI-582 V5, CagN-HiBiT with gastric epithelial AGS cells for 30 min, showing bacteria attached to the cell surface; white 583 hyphenated line indicates the AGS cell periphery; C) doubly-labelled N6 HopZII-V5, CagN-HiBiT; bacteria only. D) 584 H. pylori N6 bacteria in the absence of cells, expressing CagN-HiBiT combined with HopQI-V5 or HopZII-V5, 585 subjected to proximity ligation assay (PLA) and subsequent high-resolution microscopy; red color of the PLA label 586 indicates direct interactions or very close apposition of the two proteins. Negative controls of bacteria not expressing 587 HiBiT or V5 tags and incubated with the same antibodies were negative for any antibody staining for V5 or HiBiT 588 (not shown). Panel E) quantitative dot counts in microscopy for double labelled bacteria (strain N6) with HiBiT-589 tagged CagN and V5-tagged HopQ (clone NQ1) or V5-tagged HopZ (clone NZ10), respectively (corresponding to 590 panels A and B). Non-permeabilized fixed bacteria alone were surface-stained as above (A, B, C). HiBiT dots are 591 highlighted in green. HopQ or HopZ dots (red), respectively, were only counted if they were colocalized (paired) 592 with green HiBit dots in the same bacteria. F) quantitative dot counts in microscopy for CagN-HibiT, HopQ-V5 593 doubly-tagged bacteria co-incubated with AGS cells, stained as in B (strain N6, clone NQ1). HiBiT-dots were 594 counted for CagN-HiBiT and all clearly recognizable dots for HopQ (red) were separately counted (not paired) in 595 the same bacteria, where the majority of dots were co-localizing with green HiBiT dots. Statistics in E) and F) were 596 performed by Student‘s t-test. Significance of differences: ****p<0.0001, ns = not significant. 597 598 Functional characterization of interactions reveals role of novel interactions and a dampening 599 function of HopZ in heptose-dependent proinflammatory signalling in human gastric epithelial cells 600 We ultimately assayed, whether the newly identified direct protein-protein interactions, in particular 601 those involving the H. pylori outer membrane proteins HopQ and HopZ, had specific functional 602 implications during interactions with human cells. We therefore generated isogenic insertion-603 inactivation mutants of strain N6 by allelic exchange mutagenesis targeting hopZ (corresponding to 604 previous strategy in (60); Methods and S1 Table), and hopQ (including a hopQI and hopQII double-605 negative mutant in strain N6, which harbours both hopQI and hopQII alleles), by inserting antibiotic 606 resistance cassettes into the respective chromosomal loci. We also constructed HopZ-ON constitutive 607 mutants (complementants) in two strains, 26695 and L7 (10), for both of which the wild type is in the 608 OFF status for hopZ due to the frame-shifted CT dinucleotide repeat region in the 5’-terminus of the 609 gene (60, 61) (S1 Fig; S7A Fig; Methods). Both engineered HopZ-ON strains expressed HopZ in 610 comparison to the original wild types, which did not show detectable HopZ protein (Western blots in 611 S7B, S7D Fig). 612 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 22 613 Fig 5. Characterization of H. pylori hopQ and hopZ mutants in cell activation assays in human epithelial 614 cells shows that CagT4SS-dependent proinflammatory activation is dependent on both HopQ and HopZ 615 and is inversely regulated by HopQ and HopZ. Cell assays for pro-inflammatory activation were performed with 616 various H. pylori strains and isogenic hopZ and hopQ mutants in two different gastric epithelial cell lines (AGS and 617 NCI-N87, activation determined quantitatively as IL-8 release by ELISA) and HEK_luc NF- B luciferase reporter 618 cells (activation quantitated as reporter luciferase activity). A) AGS and B) NCI-N87 gastric epithelial cell lines were 619 co-incubated with H. pylori strains N6 and 26695 and isogenic hopZ and hopQI/II mutants for 4 h; results of cell 620 activation were quantitated using IL-8 ELISA from cell supernatants (at least four experiments were performed in 621 triplicates); statistical analysis was performed for the comparison between wild type-co-incubated versus mutant-622 co-incubated cells for each mutant (One-way ANOVA with pairwise comparisons and Tukey’s post-test). C) HEK-623 luc reporter cells were transfected with human CEACAM1 expression plasmid or empty vector control; 24 h post 624 transfection, cells were co-incubated with H. pylori strains and isogenic mutants for 3 h before recording activation 625 (luminescence as relative luminescence units [RLU]); strain-specific aspect in CEACAM1-dependent T4SS function 626 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 23 is revealed. D) HEK-luc reporter cells were transfected with human integrin-α5β1 expression plasmids or empty 627 vector control; 24 h after transfection, cells were co-incubated with H. pylori strains and isogenic mutants for 3 h 628 before measuring activation (luminescence as relative luminescence units [RLU]); for statistical analyses of 629 differences between conditions in C) and D), again Two-way ANOVA with pairwise comparisons with Tukey’s test 630 was calculated. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; ns = non-significant. Color code of statistics: blue 631 indicates significant differences between empty vector-transfected and expression vector-transfected; Black 632 indicates differences between independent strains co-incubated with cells transfected with only empty vector, and 633 green indicates differences between independent strains co-incubated with cells transfected with plasmids 634 expressing either human CEACAM1 (C) or integrin-α5β1 (D). Co-incubation assays were performed in at least four 635 biological replicates; All experiments were repeated at least twice on different days with similar results. E) CagA 636 translocation assay (hummingbird phenotype, see Methods) for bacteria co-incubated with human gastric epithelial 637 cells (AGS), developed in the following conditions: mock co-incubated, co-incubated with 26695 wild type (wt), 638 26695 HopZ-ON-complementant, 26695 cagA mutant, or 26695 cagPAI mutant. Statistical evaluation by One-way 639 ANOVA (Kruskal-Wallis test) with pairwise comparisons; **p<0.01; ****p<0.0001. 640 641 For further functional evaluations, we then performed comparative cell infections (Fig 5, S7C, S7E Fig), 642 both in AGS and NCI-N87 human gastric epithelial cells, measuring IL-8 cytokine secretion as a read-643 out. We obtained phenotypes for heptose-transport-dependent IL-8 secretion (13, 14) after short-term 644 co-incubation (4 h) of the strains with the cell lines. hopQI-hopQII double mutants in strain N6 and 645 hopQI single mutants in strain 26695, as expected (15), exhibited a significantly reduced IL-8 release 646 with AGS cell co-incubation (Fig 5A), supporting the known important role of HopQ in the functionality 647 of the CagT4SS, reported in the context of human CEACAMs (15). IL-8 induction in AGS cells was not 648 affected by the constitutive ON-switch engineered in hopZ in strain 26695 (Fig 5A, 5B) nor by the hopZ 649 inactivation mutation in strain N6. Surprisingly, in the second gastric cell type, NCI-N87, in contrast, 650 the HopZ-deficient isogenic mutant of N6 showed a significantly divergent phenotype compared to the 651 wild type, inducing about 30% higher IL-8 secretion in the cells, in comparison to the bacterial wild 652 type (Fig 5B). Furthermore, the constitutive HopZ-ON mutant in strain 26695 induced a significantly 653 diminished IL-8 secretion (Fig 5B, S7C Fig). This was also the case for HopZ-ON in the second strain L7 654 (S7E Fig). The latter results emphasized a cell-type dependent, direct, involvement of HopZ in CagT4SS 655 functionality. As another cell- and strain-specific phenotype, the hopQI-hopQII double mutant of strain 656 N6 showed a significantly reduced IL-8 secretion phenotype in NCI cells (Fig 5B), while the 26695 hopQ 657 mutant did not show an altered phenotype. Together with the newly detected direct interactions of 658 HopZ with several CagT4SS proteins and HopQ, we suggest a direct cooperative function of the two 659 Hop proteins with the T4SS outer proteins which can inversely affect the functionalities for HopQ 660 (decrease) versus HopZ (increase). In each strain-cell line combination, the quantitative effects were 661 very stable, however, effects were clearly strain- and cell-type-dependent. 662 We also carried out a functional characterization of cell activation by hopQ and hopZ mutants in 663 HEK293 cells (HEK_luc, containing a luciferase-based NF- B reporter, Methods) in which cell type H. 664 pylori major adhesins SabA and BabA (50, 64) play a less important role, due to the lack of specific cell 665 surface receptors, which also includes a lack of CEACAM1. Only for strain 26695, this setting revealed 666 a significant contribution of HopZ, but not HopQ, in CagT4SS-mediated heptose transport-induced NF-667 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 24 B activation, quantitated as luciferase induction, under all conditions; HopZ-ON mutants showed 668 reduced reporter cell activation by more than 60% during the early co-incubation of 3 h (Fig 5C, 5D) or 669 reduced IL-8 production (data not shown), in comparison to the HopZ-deficient 26695 wild type strain. 670 N6 and its isogenic mutants had no strong phenotype and rather low activation in the HEK cell model 671 when the cells were co-incubated plain or empty vector-transfected (Fig 5C, 5D). As a second T4SS-672 dependent, transport-dependent phenotype, determined by CagA translocation efficiency, that may 673 also be influenced by HopZ, we analyzed cell elongation (termed hummingbird phenotype (13, 68, 69)) 674 in AGS gastric epithelial cells (strain 26695). Converging with the reducing activity of HopZ on pro-675 inflammatory cell activation, a significant reduction of the exclusively CagA-dependent phenotype, 676 revealing diminished CagA translocation, was quantitated for the constitutive HopZ-complementant in 677 comparison with the hopZ-negative 26695 parent (Fig. 5E). cagA-mutant and cagPAI-deletion mutant 678 were negative for the effect. Consistently, for the HopZ-deficient isogenic mutant in strain N6, an 679 enhanced hummingbird phenotype was shown, indicating increased CagA transport (own data not 680 shown). 681 In parallel, HEK-NF-B reporter cells transiently transfected with plasmids to express human CEACAM1 682 or α5β1 integrin were also exposed to the same H. pylori strains, in order to verify a functional 683 relevance of those two human receptors for HopZ function in the T4SS context. Interestingly, this 684 revealed an important strain-specific difference: strain N6 was almost completely dependent on 685 CEACAM1 expression for NF-B activation and showed the known, enhanced, HopQ-dependent T4SS 686 functionalities only when CEACAM1 was expressed (Fig 5C). The effect of hopZ in this strain was minor 687 in HEK cells under all test conditions. For strain 26695, in contrast, T4SS pro-inflammatory function was 688 not CEACAM1-dependent at all and showed similar phenotypes, with a significant deficiency in 689 activation of the 26695 HopZ-ON mutant, in the empty-vector conditions as well as with either 690 CEACAM1 or integrin α5β1 expressed (Fig 5C, 5D). The HopZ-dependent decrease in cell activation for 691 26695 was not significantly influenced by additional CEACAM1 or integrin expression (Fig 5C, 5D). 692 We also tested other possibly relevant human CEACAMs (3, 5, and 6), as reported in (47, 57, 59), in the 693 HEK reporter cells after transient transfection, of which hCEACAM3 promoted significantly elevated 694 NF-B activation with HopZ-ON mutant (S7F Fig), while hCEACAM5 and hCEACAM6 did not show 695 significant divergence compared to wild type, as with CEACAM1 transfection (S7G, S7H Fig). This set-696 up also confirmed that HopQ-mediated T4SS increase for strain N6 (but not 26695) was largely 697 CEACAM-dependent in HEK cells. Integrin expression in the HEK reporter cells generally dampened 698 CagT4SS-induced responses but did not alter the relative contributions of HopQ or HopZ to the strain-699 specific phenotypes (Fig 5D). 700 701 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 25

702 The cagPAI genetic island geared for host cell manipulation by assembling a functional T4SS was 703 presumably acquired by H. pylori in a single event during evolution more than 60,000 years ago, and 704 has since further evolved to be genetically highly diverse, with particularly the genes coding for 705 surface-exposed proteins of the presumed Cag outer protein complex underlying strong diversifying 706 selection (10). These surface proteins of the H. pylori CagT4SS, reported to form a pilus-like structure 707 under certain conditions of human host cell co-incubation (35), mainly CagL (VirB5 homolog), CagI, 708 CagH, possibly supported by CagC (VirB2 homolog), are known so far to be important for the function 709 of the export machinery and to directly or indirectly interact with host cell factors (35, 39, 44, 67). CagY 710 with CagL, as well as CagL with CagI, and CagI with CagH were previously reported to interact directly 711 (36, 40, 46). 712 Although in recent years, more extensive structural studies, utilizing the improving technology of high-713 resolution Cryo-EM or Cryo-ET, have contributed to a better understanding of the overall CagT4SS 714 structure. However, even if it has already been proposed that a CagT4SS surface assembly or structure 715 might exist (40, 67), in particular the composition and interaction of the surface-exposed Cag proteins, 716 for instance CagI, CagL, CagN, CagC, and CagA, the definition of a potential CagT4SS surface protein 717 complex or structure, and the interplay of Cag proteins with regard to host cell interaction and the 718 export functions of the CagT4SS are still poorly understood (21, 30, 33). 719 In this study, bacterial two- and three-hybrid assays were initially used to fill the gaps of so far 720 uncharted territory in CagT4SS protein-protein interactions, in particular of the CagT4SS surface 721 proteins. Thereby, we aimed to uncover potential unknown interactions and interaction partners, in 722 order to improve the understanding of the Cag apparatus and possibly also shed more light on the 723 entire CagT4SS transport process. Positive interactions identified in these assays were further 724 characterized by analyzing purified recombinant proteins for interactions by biolayer interferometry 725 and further methods. In addition to HopQ (two alleles), we also included the sequence-variable and 726 phase-variable closely related HopQ paralog HopZ (60, 61) (two alleles) as a possible functional 727 interaction partner of the CagT4SS. While the interaction of both HopQ alleles with human CEACAMs, 728 in particular CEACAM1, was studied before to some detail biochemically, structurally as well as 729 functionally (56, 57), a receptor for HopZ on the human side was not yet known (59, 61). We newly 730 identified here that HopZ interacts with itself, with HopQ, and with human CEACAM1, using BACTH, 731 multi-well plate-binding assays and direct BLI interaction assays using purified recombinant HopZ. 732 Surprisingly, we also identified and quantitated both HopQ and HopZ to interact with human integrin 733 (α5β1) in binding assays. Despite our positive binding results and the established evidence that 734 CagT4SS proteins such as CagL, CagI, CagY (44, 46, 66), and CagA (46, 70) bind to integrins, our 735 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 26 functional results so far corroborate previous findings that tested integrins are not involved in 736 important CagT4SS functionalities (15). 737 738 The Hop protein HopQ was revealed as new interaction partner of the CagT4SS surface proteins CagL, 739 CagN, CagI, CagC, CagA, possibly contributing to a proposed CagT4SS surface assembly, and of human 740 integrins (51). Since we only cloned and tested the larger portions of HopQ and HopZ (lbd and sbd 741 as defined in S1 Fig) that are not associated with the outer membrane-integrating beta-barrel 742 structures, we can only safely summarize that the detected novel interactions are supposedly located 743 predominantly in the extracellular (binding) domains of those Hops. HopQ (both alleles) was shown in 744 the present study to directly interact in vitro with CagA and CagL at nanomolar affinities, in a strongly 745 pH-dependent manner. HopQ also exhibited high-affinity interactions in vitro with CagN. While both 746 proteins, HopQ and CagL, can play a role in bacterial adhesion to host cells (59-61, 65, 66, 71), and the 747 interaction of HopQ with human CEACAMs plays a vital role in the translocation of CagA in some cell 748 models, and also for the induction of early pro-inflammatory signaling (47), mediated by heptose 749 metabolite transport (13), little has been known about the functional details of those processes. HopQ 750 close paralog HopZ interacted with CagA, CagL, CagN, and CagC, as well as with integrin and CEACAM1. 751 Interestingly, ancient HopQII, in contrast to modern, Cag-adapted HopQs, interacted more strongly 752 with human receptors and less strongly with Cag proteins. Considering that frequent recombination 753 between H. pylori strains occurs, which supports fast adaptation, this likely has led to the situation that 754 most modern HopQI or HopQII protein variants have frequently interacted with Cag proteins in their 755 strain genetic environment. This presumably created opportunities to adapt specific Hops over time 756 to interaction with CagT4SS proteins in the highly recombinogenic species. HopQI bound best to gastric 757 epithelial cells at all conditions in contrast to the other HopQ and HopZ variants, although this was not 758 true for isolated human receptor CEACAM1- and integrin-binding. 759 The results of this study also suggest an interaction and partial co-localization of both the Hop proteins 760 HopQ and HopZ in situ with proteins of the CagT4SS, and specifically Cag proteins such as CagN at the 761 bacterial surface. CagN is linked to the other Cag surface proteins by the newly identified CagN 762 interactions with both CagL and CagI, and by prior positive interaction studies of CagL with CagI, CagI 763 with CagH, and CagL with CagY (36, 40, 46), the latter being an integral protein of the OMCC (33). 764 Double immunofluorescence labelling with dual-tagged HopQ, or HopZ, and CagN proteins in intact 765 non-permeabilized bacteria demonstrated partial co-localization of HopQI and, to a lesser extent, 766 HopZII, with CagN (T4SS) in situ on the bacterial surface. HopQI in particular, which partially associated 767 with CagN on the bacterial surface in the absence of host cells, co-localized more strongly during co-768 incubation conditions with human gastric epithelial cells. Close apposition of HopQI (and HopZII) with 769 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 27 CagN by PLA assay, likewise suggesting interaction, was detectable. This corroborates and expands 770 previous fluorescence microscopy showing focal dot-like localization of CagY and CagT (VirB7) on the 771 bacterial surface of H. pylori co-incubated with AGS cells (72). The propensity of the HopQ (and HopZ) 772 proteins to form at least dimers, if not oligomers, also indicates a protein assembly, which may be 773 associated with some Cag surface proteins. Taking all the findings into account, it seems well 774 conceivable that Hop proteins may be clustering with CagT4SS surface proteins on the bacterial 775 membrane, possibly even triggered by host cell interaction. The latter hypothesis needs to be explored 776 in more detail in a future study. The finding also implies an important role of HopQ and HopZ for the 777 function of the secretion system, not only through adhesions to host cell receptors, but also via direct 778 interaction with the surface-exposed proteins of the proposed CagT4SS surface protein complex. It still 779 remains to be proven that the interacting proteins actually form a structure or complex at the bacterial 780 surface and close to the membrane-spanning secretion system under some relevant conditions. Future 781 efforts at isolating a surface protein assembly, for instance by crosslinking and/or pull-down 782 approaches in H. pylori , should help to clarify such questions, although manipulation may subvert 783 assembly and it is challenging to distinguish between direct and indirect interaction in the complex 784 setting. Very recently, a study reported a successful immune-purification experiment using the CagA 785 chaperone CagF as bait which enriched CagA, low amounts of CagN, CagL, CagI, CagH, and high 786 amounts of OMCC components from H. pylori solubilized fractions containing the OMCC Cag structure, 787 but this was so far not linked to the bacterial surface (73). This may be further evidence for the 788 assembly of a larger complex, but its existence and localizations need to be verified. 789 The strong, pH-dependent, interactions of HopQ and HopZ with themselves as well as with CagA and 790 CagL might indicate a structural change of the proteins in response to environmental conditions in the 791 stomach niche, which might affect Hop, Cag and host cell proteins in a functionally convergent manner. 792 A similar pH-dependent binding characteristic, mediated by a structural change, was shown before for 793 BabA-receptor interaction (64) and for CagL tertiary structure and its adherence to human integrins 794 (65). We also detected a pH-dependence of the binding of the four HopQ and HopZ variants to cells. 795 In addition, cell-specific phenotypes of so far unknown causality were observed, as all four protein 796 variants bound better to NCI-N87 cells at pH 6, while, to AGS cells, they bound better at pH 7. 797 Interestingly, lower pH, which is supposed to elongate the structure of CagL (65), led to a reduced 798 CagL-integrin binding, while, derived from our new results, the interaction of CagL with HopQ seems 799 to be stronger at pH 6.0 than at pH 7.0 in solution. This may suggest a competitive action between host 800 cell integrins versus CagL towards HopQ at different pH values, possibly pointing to some dynamics of 801 the secretion apparatus components at the bacterial surface. 802 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 28 The initial screening results also revealed a previously unknown interaction of CagM with CagH, and 803 CagM with CagI. CagL appears to compete with the CagM-CagI interaction. Together with our own 804

generated prior to this study for CagN and CagL (22), CagM is thereby shown to interact, we 805 assume at least transiently, with the majority of CagT4SS proteins associated with the bacterial surface. 806 Other novel interactions identified here were CagC-CagC (also reported in a recent preprint (34)), CagA 807 with CagC, and CagC with CagI, with CagL, and with HopZ. These results may support a previously 808 proposed function of CagM in guiding and transporting Cag proteins to the bacterial surface and 809 suggest that affinity-driven competition could provide export-order prioritization (22). Despite 810 extensive screening, no interaction of CagM with CagA was found here, in agreement with previous 811 studies (31-33), which supports the hypothesis that CagA export may be CagM independent. 812 Quite surprisingly, cell- and strain-specific functionalities of the CagT4SS in combination with HopZ 813 revealed that HopZ has the potential to act antagonistically to HopQ with respect to the pro-814 inflammatory function of the CagT4SS (in NCI-N87 and HEK cells), which depends on bacterial 815 metabolite transport. This was also the case for CagA-dependent cell shape changes (hummingbird 816 phenotype (68, 69)) which was decreased in the HopZ constitutive-ON complementant, while HopQ 817 enhances CagA translocation (47, 59). This functional divergence between HopQ and HopZ may also 818 be linked to our unexpected finding of strong expression differences of both HopQ and, in particular, 819 HopZ between different strains. HopZ expression can be enforced by the previously identified CT-820 dinucleotide repeat-dependent switch mechanism (61). We speculate that HopZ might be selected for 821 an ON switch when a changed condition such as pH or transmission to a new host requires its 822 dampening function, or to OFF, when reduction of T4SS function is not needed in a stomach niche or 823 altered environment of the host, e.g. during chronic infection. To expand our previous report that in a 824 smaller strain collection, hopZ allelic type did not significantly correlate with the cagPAI (60), we found 825 when analyzing a larger strain collection of the H. pylori Genome Project (74) that in particular the 826 hopZI type shows a significant correlation with cagPAI presence (own data not shown). The dampening 827 function of HopZ was not dependent on human CEACAM1 or integrin- 51 receptors in our assays. 828 Human CEACAM3, but not CEACAM1, CEACAM5, CEACAM6, or human integrin- 51, seemed to 829 influence HopZ-dependent cell activation. Strain differences comprised a strong dependence of 830 CagT4SS functionality on CEACAM1 in specific strains (strain N6, similar to previously characterized 831 strain P12 (15, 47)), versus a CEACAM-independent T4SS activity in other strains, such as 26695. In 832 contrast to previous reports (15, 47, 58), HopQ-dependent, T4SS-mediated cellular activation was only 833 promoted by CEACAMs in one (N6) of two tested strains. We can only speculate that HopQ, which 834 shows sequence variation between strains, is more versatile than previously reported and can also use 835 other cellular receptors to promote CagT4SS functionality. 836 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 29 Taken together, we have identified novel interactions between Cag outer proteins, including CagC, 837 CagA, CagL, CagN and CagI. Furthermore, novel direct interactions between H. pylori outer membrane 838 proteins HopQ and HopZ and multiple Cag proteins point to a direct functional involvement of those 839 interactions in contact with and during transport processes towards human cells. This may render the 840 possibility of a larger-order complex of proteins on the bacterial surface, mediating those functions, 841 more plausible to further look into. Intriguingly, CEACAM1, as previously reported for human integrin 842 (44, 46, 65, 67) not only bound HopQ and HopZ, but also interacted with some CagT4SS surface 843 proteins (e.g. CagC, CagL), that we propose may form a CagT4SS surface protein assembly, possibly 844 together with outer membrane proteins. 845

of the present study that should be covered in future investigations include clarification of 846 a proposed assembly or structure at the surface of the CagT4SS and its possible components. A 847 mechanism for the HopZ-mediated dampening functions on the T4SS-mediated activites is still lacking. 848 There is also a persistent lack of information on other potential binding factors and receptors on the 849 host cell side, since not all effects and cell-directed phenotypes that we detected here could be 850 explained by CEACAM and/or integrin binding, in particular for HopZ, but also for HopQ. We also do 851 not know yet whether other H. pylori outer membrane proteins can be involved. We assume this might 852 be the case since multiple paralogs with structural homology to HopZ or HopQ exist in the H. pylori 853 Hop family (48, 75), which indicates a certain functional redundancy. This may also help explain strain- 854 and cell type-specific effects. Using other H. pylori mutants defective in known cell adhesins, specific 855 Hop/OMP or CagT4SS proteins, or site-directed mutants of those proteins deficient in novel 856 interactions, the knowledge about the interplay of interaction partners and their role in the transport 857 of effectors through the T4SS should be intensified. Strain differences in the versatility of CagT4SS 858 functions and the underlying causes should be better characterized, taking H. pylori strains of diverse 859 origins into account. Furthermore, the role of pH and glycan modifications of cell surface proteins and 860 potential receptors, which also might influence cell line-, cell type-, and individual binding or functional 861 characteristics, still have to be investigated in larger molecular detail. Uncharted territory also still has 862 to be filled for detailed molecular interaction analyses (including site-and domain-specific interactions 863 between proteins), and stoichiometric considerations. With this newly generated knowledge of 864 functional protein interactions and complex composition of CagT4SS outer protein assemblies, further 865 functional research should be implemented to characterize and visualize structural detail and export 866 processes of the T4SS in contact with human epithelial cells. 867 868 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 30

869 870 Bacterial culture conditions (H. pylori) 871 All described H. pylori wild type strains and mutants (Table S1) were grown on blood agar plates (Oxoid 872 Blood Agar Base No. 2) under microaerobic conditions created by humidified Anaerocult C bags 873 (Merck) in gas tight jars. Agar plates contained 10 % horse blood, amphotericin (4 mg/L), polymyxin B 874 (2,500 U/L), vancomycin (10 mg/L), trimethoprim (5 mg/L) and, depending on the resistance markers 875 in the mutant strains, also chloramphenicol (5 mg/L) and/or kanamycin (10 mg/L). H. pylori bacteria, 876 which are rather slow growing (minimal duplication time ca. 4 h) were cultured for 20 to 24 h at 37°C 877 on plates (reaching mid-log phase) before being used or passed to fresh plates for continuous 878 cultivation. Bacteria were discarded or thawed freshly latest after passage number 10. E. coli strains 879 (S1 Table) were grown on LB plates supplemented with relevant selective antibiotics, or in LB or TB 880 broth for propagation of plasmids, BACTH assays (76), or protein expression. For comparative protein 881 and RNA preparations from H. pylori wild type strains, some of which grow very poorly in liquid culture, 882 we used 20 h plate-grown bacteria, as the bacteria auto-synchronize under those conditions, are highly 883 proliferative and motile (checked by motility assay and microscopy). 884 885 Protein methods and Western blot 886 Samples from H. pylori were prepared by harvesting the bacteria directly from 20 h grown blood agar 887 plates into PBS with sterile cotton swabs. Cell disruption was performed by ultrasonication for 2 cycles 888 of 45 s at 4°C (Branson sonifier, Power setting = 5). Soluble and insoluble fractions were separated by 889 centrifugation (12.000 x g, 10 min, 4°C) and the protein concentration per fraction was determined by 890 BSA assays. 10 µg of total protein per sample was loaded on 11.8% or 14% SDS-PAGE Gels run in 891 Laemmli buffer (25 mA per gel) before tank blotting onto BA85 nitrocellulose membranes (Schleicher 892 & Schuell) in Towbin buffer (300 mA, 2 h). Western blot membranes were blocked in 5 % skim milk in 893 TBS-T (0.1% Tween) for 1 h at room temperature. Specific primary antibodies (see figure legends) 894 diluted in 5 % skim milk in TBS-T were incubated at 4°C overnight. Secondary POX-conjugated goat-895 anti rabbit or POX goat-anti mouse antibodies were incubated for 1 h at room temperature in a 896 1:10.000 dilution in 5% skim milk in TBS-T. Immobilon HRP chemiluminescence substrate (Merck 897 Millipore) was used for signal detection, and blots were imaged using a chemiluminescence imager 898 (BioRad). 899 900 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 31 Generation of hopQI and hopQII single and double insertion-inactivation mutants 901 The isogenic hopQI mutants (N6 and 26695) were generated by initially constructing a plasmid (Table 902 S2) which contains an insertion of a kanamycin resistance cassette (AphA3’-III) in pUC19 between 903 about 400 bp flanking arms of hopQI in a single overlap PCR cloning step (primers in Table S3). 904 Subsequently, a PCR product was derived from the resulting plasmid, pSUS3322, using the flanking 905 primers for hopQI. This PCR product was transformed into H. pylori wild type strain (N6). Clones were 906 selected on kanamycin, expanded and characterized by PCR with relevant primer combinations and by 907 Western blot using HopQ-specific antiserum for detecting lack of HopQI expression (not shown). 908 hopQII isogenic inactivation mutants were generated by cloning of hopQII[aa22-455] from H. pylori 909 Khoisan26A into the pUT18c-vector backbone by enzymatic digestion with BamHI and KpnI. By reverse 910 amplification of the construct via PCR (Roche, Expand High fidelity PCR system) a 299-bp deletion in 911 the mid-segment of the gene was achieved as well as the introduction of a SpeI digestion site. Utilizing 912 this SpeI site, a chloramphenicol resistance cassette was inserted for religation of the final construct 913 (S1 and S2 Tables). H. pylori hopQ single- and double-insertion mutant strains were then generated by 914 allelic exchange mutagenesis after natural transformation of H. pylori N6 wild type. Clones were 915 selected on blood agar plates with chloramphenicol, kanamycin, or chloramphenicol and kanamycin 916 for correct selection. The correct insertion-inactivation of the mutated hopQII gene or double insertion 917 mutation was again checked via PCR and Western blotting. The existing hopZ mutant in strain N6 was 918 used (60) 919 920 Isogenic HopZ-constitutive-ON mutants for complementing HopZ loss of function by natural OFF 921 switch 922 We also constructed constitutive HopZ-ON mutants (complementants) in two H. pylori strains, 26695 923 and L7 (10), for both of which the wild type is in the OFF status for hopZ due to the frame-shifted CT 924 dinucleotide repeat region in the 5’-terminus of the gene (S1 Fig; S7A Fig). Briefly, wild type hopZ 925 sequences together from each target strain with flanking nucleotide sequences up- and downstream 926 of the CT-repeats were cloned into pUT18c vector using primers with suitable restriction enzymes (S3 927 Table). After confirmation of the correct plasmid insertions, each vector was amplified using specific 928 primer pairs with the shortened CT-repeats (here: 7 CTs) and religated afterwards using T4 ligase. E. 929 coli was transformed with the religated plasmids, and clones were screened via direct Sanger 930 sequencing. Correct plasmids with shortened 7CT repeats (Table S2) were again PCR-amplified using 931 the initial primer pairs. A PCR product with the correct CT-repeat length and flanking sequences was 932 then used for transformation in H. pylori together with the rdxA-cm selection marker, according to the 933 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 32 MUGent strategy (77). Correct identity of the engineered strains, including expression of HopZ, was 934 verified by PCR, Sanger-sequencing the altered nucleotide segment, and by Western blot (S7 Fig). 935 936 Construction of H. pylori HopQ- and HopZ-tag insertion mutants 937 Amino acid alignments and AlphaFold 3 structural models (S1A, S1B, S1C Fig) were generated for full-938 length HopQ (56) and HopZ to reveal potential loops and surface-associated regions. Three loops in 939 HopQ and two in HopZ were selected for the insertion of tags (S1 Fig) that could serve for verification 940 of surface localization and for surface quantification of the proteins. We chose two short epitope tags 941 (HiBiT and V5) to insert into HopQ and HopZ at the respective loop locations. For the generation of H. 942 pylori strains with V5- and HiBiT-tagged HopQ and HopZ proteins (S1 Table), wild type sequences of 943 hopQI from H. pylori 26695, hopQII from PNGhigh12A and hopZII from H. pylori 26695 cloned into 944 pUT18c Vector (as used in BACTH assays) were utilized. By reverse amplification of the constructs via 945 PCR (Roche, Expand High fidelity PCR system), a SpeI digestion site along with the 42 bp long V5-tag or 946 33 bp long HiBiT-tag sequence was introduced into pre-defined surface-exposed loop regions of the 947 proteins (S1 Fig). By enzymatic digestion and subsequent religation of the PCR product, the tag 948 insertion in the genes was achieved in the final plasmids (S2 Table). H. pylori mutant strains were then 949 generated by natural transformation of N6 or 26695 wild type with a PCR product of the tag-inserted 950 hop gene along with an PCR product containing a chloramphenicol or kanamycin resistance cassette 951 with arms of homology of the H. pylori rdxA gene for homologous recombination into the chromosome 952 and resistance selection of clones. The correct introduction of the tags was checked via PCR, using tag-953 specific primers (S3 Table), and Western blotting. 954 955 Detection and validation of H. pylori HopQ and HopZ tag insertion mutants 956 The HopZ and HopQ loop HiBiT-tag insertions (highlighted in S1B, S1C Fig), were then tested for surface 957 HiBiT-tag localization in intact bacteria using an extracellular detection system with luciferase 958 reconstitution. All those loops showed strong surface exposure with HiBiT extracellular detection in 959 intact live bacteria (several different clones tested of each insertion; S6 Fig). HopQ-Loop1 insertions 960 had weaker reconstituted surface luciferase signals than HopQ-Loop2 and HopQ-Loop3 insertions (S6A 961 Fig). HopQII-Loop2 insertions gave weaker signals than HopQI-Loop2 insertions, indicating lower 962 expression of HopQII (double-HopQ-expressing strain N6). Western blot confirmed expression of the 963 HiBiT-fused constructs (S6B Fig). Loop1 insertion reduced the overall expression of HopQI (S6B Fig). 964 For HopZII, LoopX insertions gave relatively stronger surface luciferase signals at the surface than the 965 LoopY tag variants (S6C Fig), while they showed slightly lower expression in Western blot (S6D Fig), 966 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 33 indicating that LoopY insertions may be impaired in membrane insertion. Relevant fusions of HopQ 967 and HopZ were also similarly constructed as single V5 fusions and tested (see e.g. S6E Fig). The positive 968 tests for expression and surface localization of single and double-tagged loop insertions for HopQ, 969 HopZ and CagN and confirmed surface localization (S6 Fig) and offered the possibility to use such loop 970 insertions for microscopy applications at the bacterial surface. 971 972 Generating and validating double-tagged mutants CagN-HiBiT with HopQ-V5 and HopZ-V5 (NQ & NZ) 973 For double tagging of both CagN and Hop proteins in H. pylori N6, first an expression plasmid containing 974 cagN (strain 26695) with an insertion of a HiBiT tag between amino acids 214 and 215 (both duplicated) 975 was generated (S1 Table, S2 Table) and transformed into N6 wild type strain. The transformants were 976 checked by resistance, PCR with relevant primers targeting the plasmid, plasmid re-isolation, and by 977 CagN-HiBiT detection in Western blot and using Nano-Glo HiBiT Extracellular Detection System 978 (Promega #N2420) for bacterial surface detection of HiBiT-tagged proteins (S6 Fig). Subsequently, 979 plasmids, containing V5 tags to express in exposed loops of HopQ (Loop2) and HopZ (LoopY) (S1 Fig; 980 S2 Table) were created. The V5-inserted hopQ and hopZ genes were amplified by PCR. The mutants 981 were shuttled into the H. pylori N6 chromosome of the CagN-HiBiT expression strain by natural 982 transformation and recombination, using the MuGENT principle (77), co-transforming with a PCR of an 983 AphA3’-III kanamycin resistance cassette insertion in the rdxA gene of H. pylori. The double-labelled 984 clones were selected on kanamycin and chloramphenicol, verified by PCR sequencing, and 985 characterized for conserved HiBiT surface detection. They were also characterized for correct HiBiT 986 and V5 expression by Western blot (S6 Fig). 987 988 Surface labelling of HiBiT-tagged proteins in situ (H. pylori) 989 To verify the membrane localization of the HiBiT-tagged HopQ, HopZ, CagN and CagL proteins and the 990 surface exposure of the insertion loops in intact bacteria, the Nano-Glo® HiBiT Extracellular Detection 991 assay (Promega #N2420) was performed with all generated HiBiT strains. This is based on the 992 reconstitution of the small HiBiT luciferase segment by the larger LgBit segment which is added in 993 solution. H. pylori strains were grown on blood agar plates for 24 h and subsequently harvested in PBS 994 with sterile cotton swabs. In white 96-well F-bottom plates (Thermo Scientific, #236105), 50 µl of this 995 bacterial suspension diluted to OD 600 = 0.1 were mixed with equal amounts of the Nano-Glo® HiBiT 996 Extracellular Buffer containing LgBiT Protein (1:1,000) and Nano Glo® HiBiT Extracellular Substrate 997 (1:50). Plates were incubated for 10 min at room temperature on an orbital shaker at 750 rpm. 998 Luminescence was measured in a Victor Nivo Multimode microplate reader (PerkinElmer). All samples 999 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 34 were measured in triplicates. Bacteria without the tags were used as negative controls for background 1000 luminescence. Bacteria were always used after one-day growth and freshly harvested, to avoid lysis or 1001 non-specific exposure of HiBiT label on the bacterial surface. Washing intact bacteria with PBS, which 1002 was performed as an additional non-specific control to remove any potential non-specific HiBiT label 1003 with possible surface adherence from the bacterial surface, yielded the same values as without 1004 washing. To avoid misinterpretation due to potential lysis of the bacteria, we also performed specific 1005 controls, using a strain which expresses CagN-HiBiT in combination with a knock/out exchange 1006 mutation in the CagT4SS ATPase Cag (28), using an aphA-III kanamycin resistance cassette insertion, 1007 or a HiBiT fusion of the intracellular RecA protein. Those strains were directly compared with the CagN-1008 HiBiT strain in wild type background, testing for HiBiT surface localization in intact bacteria and at the 1009 same time, for total HiBiT yield in bacterial lysates, generated by ultrasonication. The ratio between 1010 total HiBiT content (measured upon ultrasonication) and surface label was calculated for the test 1011 conditions, and yielded an up to 20-fold higher ratio of total versus surface label for the control HiBiT 1012 fusion strains, verifying surface detection of CagN in wild type background. 1013 1014 Expression and purification of proteins in E. coli 1015 For recombinant protein expression and subsequent purification, the CagL gene was amplified from H. 1016 pylori 26695 without its predicted N-terminal signal sequence (amino acids 1-20). CagA was amplified 1017 from H. pylori 26695 as a truncated protein (CagA[aa1-892]), with omission of the 294 C-terminal 1018 amino acids. hopQI (sequence from H. pylori 26695), and hopQII (H. pylori K26A) were amplified as a 1019 truncated gene to code for a partial protein (HopQI[aa128-424], HopQII_K26A[aa132-412]), lacking the 1020 transmembrane bound β-barrel structure as well as hydrophobic C- and N-terminal regions in order to 1021 prevent excessive insolubility in E. coli . The cagL, cagA, and hopQI genes were cloned into the 1022 expression vector pET28a(+) (EMD Biosciences, Novagen) in frame with a C-terminal 6xHis-tag, using 1023 the enzymes NcoI and XhoI, while the hopQII gene from strain K26A was cloned with two additional N- 1024 and C-terminal 6xHis-tags and N-terminal thrombin cleavage site using the enzymes BamHI and XhoI. 1025 hopQII from H. pylori strain PNGhigh12A was designed and amplified as the same truncated binding 1026 domain as described for HopQI, but cloned into pET28a(+) in frame with a N-terminal 6xHis-tag and a 1027 thrombin site using enzymes BamHI and XhoI. Similarly, hopZI (sequence from H. pylori SU2) and hopZII 1028 (sequence from H. pylori 26695) were amplified as gene segments to express partial proteins 1029 (HopZI[aa37-531], HopZII[aa37-501], hopZ-ON sequence as reference; S1 Fig) and cloned into 1030 pET28a(+) in frame with a C-terminal 6xHis-tag using digestion enzymes SacI and XhoI. CagC was cloned 1031 as an N-terminal GST fusion protein, omitting the first 24 amino acids (forming a predicted leader 1032 peptide) initially in pGEX-4T2 (Addgene/GE Healthcare), and then recloned in between SacI and XhoI 1033 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 35 sites of pET28a(+) (all constructs in S2 Table; cloning primers in S3 Table). The cloning success of all 1034 expression constructs was checked via restriction enzyme digestions as well as sequencing. CagN was 1035 expressed and purified as described in (22). 1036 1037 Expression of HopQI[aa128-424] 1038 HopQI (sequence from H. pylori 26695) was expressed in E. coli Rosetta™(DE3) pLysS (Novagen, Merck, 1039 Germany). Expression cultures were inoculated with a starting OD 600 of 0.1 from an overnight culture 1040 in LB medium, into TB medium, and cultured further over night at 30°C, following induction with 0.1 1041 mM IPTG at a culture OD 600 of 0.6 to 1.0, before culture pellets were harvested by centrifugation at 1042 9.000xg at 4°C. 1043 Expression of HopQII[aa133-418] from PNGhigh12A and HopQII[aa132-412] from K26A 1044 HopQII (sequence from H. pylori PNGhigh12A) was expressed in E. coli Rosetta™(DE3) pLysS (Novagen, 1045 Merck, Germany) from expression cultures, inoculated with a starting OD 600 of 0.1 (from LB medium 1046 overnight culture) in TB medium and cultured further at 30°C for 4.5 h, following induction with 0.1 1047 mM IPTG at an OD600 between 0.6 and 1.0. 1048 Expression of CagL 1049 CagL expression cultures in E. coli Rosetta™(DE3) pLysS were inoculated from an overnight preculture 1050 in LB-Medium (50 μg/ml Kanamycin (Sigma)) to a starting OD 600 of 0.1 in fresh LB medium (50 μg/ml 1051 Kanamycin (Sigma)). Cultures grown at 37°C were induced with 0.5 mM IPTG (Sigma-Aldrich) at an 1052 OD600 between 0.6 and 1. Expression was performed further at 16°C in LB medium, with 180 rpm 1053 shaking overnight. 1054 Expression of CagA, GST-CagC, and GST 1055 Expression cultures were inoculated with a start OD 600 of 0.1 and induced with 0.1 mM IPTG between 1056 OD600 0.6 – 1.0 after initial growth at 37°C, with 180 rpm shaking. Expression of CagA[aa1-892], GST-1057 CagC, and GST (from empty pGEX-4T2) was further carried out for 4.5 h at 30°C and 180 rpm shaking 1058 in E. coli Rosetta™(DE3) pLysS. 1059 Expression of HopZI and HopZII 1060 HopZI and HopZII were expressed in E. coli Rosetta™(DE3) pLysS (Novagen, Merck, Germany) from 1061 expression cultures, inoculated with a starting OD 600 of 0.1 and grown at 37 °C in LB medium or TB 1062 medium respectively. Cultures were induced at an OD 600 between 0.6 and 0.8 with 0.1 mM IPTG and 1063 then grown further at 30 °C for 4.5 h. 1064 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 36 1065 Protein purification from the soluble fraction (CagL, CagA, HopQI, HopQII_K26A, and GST-CagC) 1066 Expression culture pellets were resuspended in Lysis buffer (50 mM Tris/HCl pH 8.0, 300 mM NaCl, 2% 1067 Triton-X-100, 0.1 mM DTT) and cells were mechanically disrupted with a French press at 1.9 kBar 1068 pressure for 1 cycle (One Shot cell disruptor, Constant Systems LTD). Following subsequent 1069 centrifugation at 12.000 x g for 20 min at 4°C, the proteins were purified from the soluble fraction via 1070 a Protino Ni 2+-NTA (Ni-NTA) affinity column (Macherey&Nagel) using an Äkta Prime Plus system (GE 1071 Healthcare). The system and column were first equilibrated in purification buffer (50 mM Tris/HCl pH 1072 8.0, 300 mM NaCl, 0.1 mM DTT) and loaded with protein. Following an extensive wash step to remove 1073 non-specifically bound protein, an additional high-salt wash step with loading buffer including 1 M 1074 NaCl was performed (to remove any remaining nucleotide contamination), followed by two mild 1075 imidazole wash steps, by step-wise increasing the Imidazole concentration in the running buffer to 25 1076 mM and then 50 mM. The target protein was eluted by increasing the imidazole concentration in a 1077 linear gradient from 50 mM to 500 mM in the running buffer. 1 ml elution fractions were collected and 1078 analyzed on SDS gels. High protein-containing fractions were pooled, dialyzed against purification 1079 buffer and concentrated to an appropriate protein concentration using centrifugal filter devices 1080 (Merck Millipore). Protein purity as well as concentration were assessed via SDS-PAGE (S4 Fig), loading 1081 the final protein pools in various amounts along a BSA standard with defined protein concentrations. 1082 Purification of GST proteins (GST-CagC, GST) from the soluble fractions of expression culture pellets 1083 was performed similarly to the Ni-NTA purifications. Pellets were resuspended in GST-Lysis buffer (50 1084 mM Tris/HCl pH 8.0, 300 mM NaCl, 2% Triton-X-100, 5 mM DTT) for mechanical cell disruption as 1085 described. Using the Äkta Prime plus system (GE Healthcare) a Protino GST/4B 1 ml FPLC column 1086 (Macherey&Nagel) was equilibrated in GST-purification buffer (50 mM Tris/HCl pH 7.5, 150 mM NaCl, 1087 5 mM DTT) before loading of the protein. Following an extensive wash step with equilibration buffer, 1088 to remove non-specifically bound protein, the target protein was eluted using a linear gradient from 1089 0% to 100% GST elution buffer (50 mM Tris/HCl pH 8.0, 10 mM reduced glutathione, 5 mM DTT). 1 ml 1090 elution fractions were collected and analyzed as described before for Ni-NTA purified protein. An SDS 1091 gel of CagC-GST after purification is included in the supplement, S4 Fig. 1092 1093 Protein purification from the insoluble fraction (HopZI, HopZII, HopQII_PNGhigh12A) 1094 Expression culture pellets were resuspended in denaturing lysis buffer (50 mM Tris/HCl pH 8.0, 300 1095 mM NaCl, 6 M Urea, 0.1 mM DTT) and cells mechanically disrupted with a French press at 1.9 kBar 1096 pressure for 1 cycle (One Shot cell disruptor, Constant Systems LTD). The crude cell extract was 1097 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 37 incubated at 4°C for 1 h under constant shaking on a spinning wheel prior to separation of soluble and 1098 insoluble fraction by centrifugation at 12.000 x g for 20 min at 4°C. 1099 HopZ proteins and HopQII from PNGhigh12A were purified from the insoluble fraction via Protino Ni-1100 NTA affinity columns (Macherey&Nagel) using an Äkta Prime plus system (GE Healthcare) equilibrated 1101 in purification buffer (50 mM Tris/HCl pH 8.0, 300 mM NaCl, 6 M Urea, 0.1 mM DTT). Following an 1102 extensive wash step with running buffer and two mild imidazole wash steps, by stepwise increasing 1103 the imidazole concentration in the running buffer to 25 mM and then 50 mM, the target protein was 1104 eluted by increasing the imidazole concentration in a linear gradient from 50 mM to 500 mM in the 1105 running buffer. 1 ml elution fractions were collected and analyzed via SDS-Page. High protein-1106 containing fractions were pooled and step-wise dialyzed against purification buffer with decreasing 1107 urea concentrations using centrifugal filter devices (Merck Millipore). Purified HopZI and HopZII 1108 proteins were dialyzed to final urea concentrations of 0.5 M and 2 M, respectively, which was tested 1109 before in small aliquots to still support solubilization, to avoid proteins to precipitate. For subsequent 1110 assays, proteins were diluted further (at least 1:10), to appropriate concentrations, in the respective 1111 assay buffers, immediately before start of the assay. 1112 For HopQII_PNGhigh12A specifically, high protein containing fractions after purification were pooled 1113 and concentrated to a protein concentration of 10 mg/ml using centrifugal filter devices (Merck 1114 Millipore). The concentrated protein pool was then dropwise diluted and subsequently incubated in 1115 ice-cold refolding buffer (50 mM Tris/HCl pH 8.0, 300 mM NaCl, 210 mM urea, 0.1 mM DTT) under 1116 constant stirring for 45 min. The resulting fraction was again loaded onto a Protino Ni-NTA affinity 1117 column (Macherey&Nagel) using an Äkta Prime plus system (GE Healthcare) equilibrated in native 1118 purification buffer (50 mM Tris/HCl pH = 8.0, 300 mM NaCl, 0.1 mM DTT). Following the protein 1119 loading, a wash step with a flow rate of 0.1 ml/min was employed for 45 min for a final on-column 1120 refolding of the HopQII protein. Elution was subsequently performed by a single step increase of the 1121 imidazole concentration to 500 mM in the running buffer. High protein containing fractions were again 1122 pooled, dialyzed against native purification buffer to reduce the urea content as far as possible, and 1123 concentrated to an appropriate protein concentration using centrifugal filter devices (Merck 1124 Millipore). As a final step for all proteins, protein purity as well as concentration were determined on 1125 SDS gels (S4 Fig), loading the final protein pools in various amounts along a BSA standard with defined 1126 protein concentrations. 1127 1128 Protein analysis using analytical size exclusion chromatography 1129 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 38 Quantitative analytical size-exclusion chromatography (SEC) was performed to determine the native 1130 state of highly purified HopQI protein. SEC data were collected on a Cytiva ÄKTAmicro instrument 1131 equipped with a Superdex 200 Increase (10/300) column (Cytiva). Data were evaluated with the 1132 OmniSEC software package supplied with the instrument. A commercial molecular mass standard 1133 (BioRad Gel Filtration Standard) was run before HopQ to calibrate the SEC for mass determination with 1134 the same column and buffer. Using the peak positions of the standard, we generated a calibration 1135 curve that was then used to calculate the molecular masses of the peaks in the HopQ elution profile. 1136 For analysis of protein eluted in each SEC peak of a preparative SEC run in parallel, 2 ml fractions were 1137 collected and analyzed on SDS-PAGE (S4 Fig). 1138 1139 Bacterial two-hybrid (BACTH) assay 1140 Protein-Protein interactions were studied by generation of C- or N-terminal fusions of all proteins of 1141 interest with either one of the two (T18 or T25) domains of the Bordetella pertussis adenylate cyclase 1142 (Cya) enzyme (76). By co-transformation of two plasmids each carrying a protein of interest fused to 1143 one of two domains of the Bordetella pertussis adenylate cyclase enzyme, a protein-proteins 1144 interaction can be made visible and quantifiable. Upon close interaction of the studied proteins the 1145 Cya-Enzyme activity is restored leading to cAMP synthesis and ultimately to the transcription of the β-1146 galactosidase enzyme in the E. coli BTH101 host. β-galactosidase activity as a readout for the studied 1147 protein-protein interactions was visualized by cleavage of the artificial substrate X-Gal (5-Bromo-4-1148 chloro-3-indolyl-β-D-galactopyranoside, Sigma-Aldrich) in LB-Agar (blue coloring after cleavage) as well 1149 as quantitated by an enzyme activity assay in solution measuring cleavage of ONPG (yellow coloring, 1150 Miller units). 1151 All studied proteins of interest were cloned in frame (leaving out canonical membrane domains and 1152 predicted leader peptides) with either one of the two domains of the Cya gene in the Vectors pUT18c 1153 or pKT25 (for C-terminal fusions) and pUT18 or pKNT25 (for N-terminal fusions). Expression constructs 1154 and primers used for cloning are listed in Tables S2 and S3, respectively. A number of BACTH expression 1155 constructs, designated in Table S2 as (LT) were generously donated by Laurent Terradot, UMR 5086, 1156 Lyon, France. Inserts in all generated constructs were verified via control digestion, selected PCR, and 1157 partial sequencing. For each BACTH assay, 10 ng plasmid DNA encoding a protein of interest with a T18 1158 domain fusion was co-transformed into E. coli BTH101 with 10 ng plasmid DNA of a T25 domain-fused 1159 protein. Transformed bacteria were plated on Luria Bertani (LB) plates with 200 µg/ml ampicillin, 50 1160 µg/ml kanamycin, 0.5 mM IPTG (Sigma-Aldrich) and 40 µg/ml X-Gal and incubated at 30°C for 48 h. 1161 Three single clones per transformation were jointly inoculated in 600 µl LB medium with 200 µg/ml 1162 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 39 ampicillin and 50 µg/ml kanamycin in 2 ml Eppendorf tubes. Cultures were grown at 30°C for 24 h at 1163 180 rpm shaking. 50 µl per preculture were then used to inoculate the expression overlay-culture on 1164 top of 800 µl solid LB agar in individual wells of a 24-well cell culture plate (Greiner) in duplicates. 1165 Expression cultures were incubated at 30°C for 24 h and subsequently harvested by washing the 1166 bacteria from the agar with 500 µl 0.9% NaCl into Eppendorf tubes. Bacteria were pelleted by 1167 centrifugation at 5.000 x g at 4°C for 10 minutes and pellets were stored at -20°C until further use. In 1168 total, we tested 327 new plasmid combinations (negative and positive controls not counted). In the 1169 bar graph figures, for any specific positive result, only the combination with the highest outcome is 1170 shown. The full matrix of plasmid combinations is available upon request. 1171 To analyse the expression of each fusion protein, one duplicate pellet per sample was resuspended in 1172 100 µl 0.9% NaCl before cell lysis by sonication (4°C, 2x 45 sec at 100% Output on Level 5, Branson 1173 sonifier, Emerson, Danbury, USA). Soluble and insoluble fractions were separated by centrifugation at 1174 4°C for 20 min at 9.000 x g. Total protein concentration was determined by bichinchoninic acid assays 1175 (BCA) before 10 µg total protein were loaded per lane on a 11.8% or 14% SDS-PAGE Gel run in Laemmli 1176 buffer (25 mA) for subsequent tank blotting on nitrocellulose membranes (BA85, Schleicher & Schuell) 1177 at constant 300 mA for 2 h in Towbin buffer. Antibodies specific for either the expressed proteins or 1178 the Cya enzyme domains T18 or T25 used in the fusion constructs were used to verify expression of 1179 both interaction partners. Expression testing was done for all combinations. 1180 For quantification of the protein-protein interactions, ß-galactosidase activity measurements were 1181 performed from each sample as described before (76), and analyzed in triplicates (calculated in Miller 1182 units). Positive and negative controls were incorporated in each individual assay to control for 1183 potential biological or technical variation. A leucine zipper from yeast protein GCN4 with a high affinity 1184 to itself, fused to both T25 and T18 in the vectors pKT25 and pUT18 (76), respectively, was used as a 1185 positive control. The empty vector constructs pKT25 and pUT18, co-transformed, served as the 1186 negative control. Additional negative controls tested incorporated one plasmid with a gene insertion 1187 together with one empty plasmid. The detection limit for positive interaction was determined at above 1188 1.5-fold the negative control over at least 10 experiments. Western Blots for all experiments were 1189 performed, using specific custom-generated antibodies against H. pylori CagN, CagM, CagI, CagH, HopZ 1190 and HopQ, or alternatively against the T25 ( T25-antiserum generously donated by Daniel Ladant at 1191 Pasteur Institute, Paris) and T18 ( T18-antibody obtained from Santa Cruz Biotechnology, mouse 1192 monoclonal antibody 3D1 against Bordetella pertussis Cya, #sc-13582) fusion fragments of B. pertussis 1193 adenylate cyclase Cya. The latter two antisera also detect the T25 and T18 fragments alone, produced 1194 by empty control plasmids. Results of BACTH experiments are provided as -galactosidase 1195 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 40 quantification (bar graphs) and as matrix heat-maps, providing fold-change values over negative 1196 control mean (explained also in the figure legends). 1197 1198 Bacterial three hybrid (BAC3H) assay 1199 The bacterial three hybrid assay is based on the principle of the BACTH (described above) with the 1200 addition of a third protein of interest via the arabinose-inducible compatible vector pAB184a (78). 1201 By addition of the third protein, expressed from a compatible plasmid backbone without a Cya-enzyme 1202 domain fusion, the outcome of the initial two-hybrid interactions can be altered. A reduction of ß-1203 galactosidase signal can either be the result of a higher affinity of one domain fused protein to the 1204 third interaction partner or the formation of a three-protein complex that sterically hinders the 1205 restoration of Cya-enzyme activity. A weak two-hybrid interaction may also be stabilized by the 1206 formation of a three-protein complex without steric hindrance, thereby increasing the measured ß-1207 galactosidase activity. For the BAC3H, similar as for the BACTH, 10 ng plasmid DNA of each plasmid 1208 were co-transformed into E. coli BTH101. In addition to the supplements used in the BACTH procedure, 1209 50 µg/ml chloramphenicol was used in LB-plates for selecting clones after transformation as well as 1210 for preculture media and expression agar. A final concentration of 0.1% arabinose was used in the 1211 transformation plates as well as in the expression agar for the pAB184a vector expression induction. 1212 Protein expression control of all three potential interaction partners as well as the ß-galactosidase 1213 measurements were performed as described for the BACTH assays. Positive and negative controls 1214 were the same as described for the BACTH above, likewise with additional Western blots performed 1215 for each experiment. Additional negative controls consisted in the third plasmid pAB184a introduced 1216 as empty vector alongside the two relevant pUT18 and pKT25 clones with inserts. Results of BAC3H 1217 are provided as -galactosidase values. 1218 1219 Biolayer interferometry measurements (BLI) for quantitation and affinity of direct protein-protein 1220 interactions 1221 For the determination of kinetic and affinity parameters of selected protein-protein interactions 1222 biolayer interferometry (BLI) measurements with the Octet RED96 System (Sartorius) were performed. 1223 Highly pure proteins (our own purified proteins for the Cag and Hop proteins, see above) and 1224 commercial preparations of hCEACAM1 (Abnova, #P6737) and α5β1-Integrin (R&D Systems, #3230-1225 A5)) were immobilized on Amine Reactive 2nd-Generation (ARG2G) biosensors activated with 20 mM 1226 EDC, 10 mM NHS in H 2O. Protein coupling was achieved by EDC-catalyzed amine-bond formation at a 1227 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 41 concentration of 20 mg/ml for 600 sec in 10 mM sodium acetate buffer at pH = 5.0. Sensors were 1228 subsequently inactivated in 1 M ethanolamine at pH 8.5. Analyte interactions were measured at 1229 different concentrations, in most experiments for at seven concentrations between 0.125 µM to 2 µM 1230 in 1x kinetic buffer (Sartorius), separately at both pH 7.0 or pH 6.0. Association and dissociation steps 1231 were carried out for 300 sec each. An activated reference sensor with immobilized protein but 1232 measuring a buffer blank during association and dissociation steps was used and subtracted for all 1233 kinetic measurements. All experiments were performed with 200 µl filling volume per well of a black 1234 96-well plate (Greiner #655209) at a constant temperature of 30°C and constant shaking at 1000 rpm. 1235 All parameters were analyzed and fitted (1:1 model) and Coefficient of Determination R 2 calculated 1236 using the ForteBio Octet red analysis software. 1237 1238 Protein-protein interaction plate assays 1239 20 ng protein were coated per well of 96-well plates (Greiner #650001) in 100 µl 1xPBS overnight at 1240 4°C. Following a blocking step in 200 µl PBS + 10% FCS for 1 h at room temperature, the analyte protein 1241 was incubated for 2 h. For pH-dependent binding, analytes were incubated in the protein-coated 1242 blocked plates with 50 mM Tris buffer adjusted to either pH 6 or 7, with testing after the incubation 1243 period that the pH remained constant. Subsequently, primary antibodies (in most tests except for CagL, 1244 we used HIS.H8 α-His (commercial mouse monoclonal, Invitrogen)) or α-CagL antisera (rabbit 1245 polyclonal sera) were diluted 1:4,000 or 1:5,000 (α-CagL) in PBS with 2% or 5% (α-CagL) FCS and 1246 incubated for 1 h at room temperature. Secondary antibodies (peroxidase-coupled goat-α-rabbit or 1247 goat-α-mouse) were diluted 1:10,000 in PBS + 1% BSA and incubated for 1 h at room temperature. 1248 Signal was developed by incubation in 100 µl tetramethylbenzidine (TMB) substrate (Sigma Aldrich) 1249 for 30 min at room temperature in the dark. The reaction was stopped by the addition of 50 µl 1250 phosphoric acid and the signal was detected by measuring the absorbance at 450 nm in a Victor Nivo 1251 Multimode microplate reader (PerkinElmer). Three subsequent wash steps with wash buffer (1xPBS + 1252 0.05% Tween20) were performed between all described incubation steps. All incubations were 1253 performed at room temperature on an orbital shaker operated at 180 rpm, except for the TMB 1254 substrate incubation which was kept without shaking in the dark. For all experiments, wells coated 1255 with ligand protein and antibodies but not incubated with analyte served as reference blank values 1256 (background), which were subtracted from all other values. Additional controls were performed with 1257 commercial preparation of pure BSA as unrelated analyte in solution, which showed no binding and no 1258 concentration dependence towards the coupled ligands. Furthermore, plate-bound human integrin 1259 was incubated with custom-purified His-tagged H. pylori CagM (22) as a Cag control, followed by the 1260 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 42 same antibody combination which was not showing binding for CagM. All samples were measured at 1261 least in triplicate experiments. 1262 1263 Binding assay of purified proteins to human cells 1264 The binding of purified proteins to human cells was performed similarly as described under protein-1265 protein binding above, in 96 well plates. As preparation, human cells were grown in cell culture 96-1266 well plates to confluency and fixed twice for 1 h in 2% paraformaldehyde in sodium phosphate buffer, 1267 pH = 7.0. Subsequently, the fixing agent was washed out, cells treated twice with quenching buffer 1268 (0.1% glycine in PBS) and then blocked in PBS + 1% BSA or 50 mM Tris-HCl + 1% BSA with different pH 1269 values, pH 6.0 or pH 7.0. After blocking, protein co-incubation was performed in the same blocking 1270 buffer. Detection with antibodies was conducted as described above under protein-protein binding 1271 plate assays. All binding conditions were performed at least in triplicates. 1272 1273 Immunofluorescent labelling, Proximity Ligation Assay (PLA) and fluorescence microscopy 1274 The bacterial specimens were fixed two times 1 h using 2% PFA in potassium phosphate buffer, 100 1275 mM, pH = 7, on gelatin-coated coverslips. After a subsequent quenching step, the specimens were 1276 washed twice in PBS, blocked in blocking buffer (non-permeabilizing; PBS, 1% BSA, 2% goat serum), 1277 and then exposed to primary antibodies in blocking buffer (α-HiBiT, mouse monoclonal 30E5, Promega 1278 #N7200, 1:100 diluted; or α-V5, rabbit, Invitrogen # MA5-15253, 1:500 diluted). After an overnight 1279 staining step, the samples were washed three times and then incubated for 2 h in secondary 1280 antibodies, tagged with fluorescent dyes, diluted in blocking buffer (Molecular Probes Thermo 1281 Scientific, anti-mouse, Alexa488, 1:2.500-diluted, anti-rabbit, Alexa564, 1:2.500-diluted). Finally, the 1282 samples were washed again carefully three times, counterstained with DAPI (staining bacterial nucleic 1283 acids) in PBS for 10 min, and finally embedded in Mowiol on glass slides. Negative control bacteria not 1284 bearing any tag which we stained in parallel did not show any fluorescent label with either the anti-V5 1285 or anti HiBiT-tag antibodies. Bacteria only bearing one tag did not show any non-specific background 1286 staining when incubated with the respective other antibody (not shown). 1287 Proximity Ligation Assay (PLA) on similar specimens with the same dual-tag insertions prepared in 1288 parallel was performed according to the manufacturer’s instructions by DuoLink (Sigma Aldrich, 1289 DuoLink in Situ PLA Kit Red #DUO92101-1KT), using two different secondary antibodies (anti-mouse-1290 PLUS and anti-rabbit-MINUS) in combination with the -HiBiT and -V5 antibodies, and the PLA 1291 labelling Red chemistry. Briefly, the specimens were incubated, after standard blocking as for 1292 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 43 immunofluorescence, with primary antibodies, washed, and then incubating with the secondary 1293 antibodies in PLA blocking buffer. The ligation between the PLUS and MINUS probes, followed by wash 1294 steps and amplification reaction with DuoLink PLA Red detection reagent, all in a humid chamber at 1295 37°C, were performed subsequently. After additional, final wash steps, PLA specimens were also 1296 counterstained with DAPI for visualization of bacterial DNA. Negative control bacteria without one or 1297 both of the tags did not show any PLA label, confirming its specificity. Routine immunofluorescence 1298 microscopy on all specimens was performed in an Olympus IX-40 microscope at 100-fold 1299 magnification. For high-resolution imaging, samples were exposed to line scanning in a high-resolution 1300 Zeiss Thunder Imager microscope with scanning functions (Mi8 microscope, Quantum Stage, highly 1301 sensitive K8 camera, and multi-line, high-intensity fluorescence LED light source), equipped with a 63-1302 fold magnification lens. Tif-type images were generated in the different single color channels as black 1303 and white images, alternatively overlayed from the different channels in the Thunder software, and 1304 exported as single-color or multi-color tif images. 1305 1306 RNA isolation of bacterial samples, cDNA synthesis and qRT-PCR 1307 For RNA isolation of bacterial samples, H. pylori strains were grown on blood agar for approximately 1308 20 h (mid-log phase) and biomass was collected with a cotton swab and shock-frozen in liquid nitrogen 1309 and stored at -80°C until further use. For collecting bacterial RNA, bacterial pellets were disrupted 1310 quickly by homogenizing with lysing matrix B (MP Biomedicals #116540425) in an MP bead beater. 1311 Total RNA of the samples was isolated using the RNeasy Mini Kit (QIAGEN, Germany) following the 1312 manufacturer’s instructions. cDNA synthesis and qRT-PCR were performed as described before in (79), 1313 applying the respective established MIQE quality standards. All sample results were normalized to 1314 bacterial 16S rDNA standards run for each sample in parallel. 1315 1316 Cultivation of human cells 1317 The human cell lines AGS (ATCC CRL-1739, human gastric adenocarcinoma cell line) and NCI N87 (ATCC 1318 CRL-5822, human gastric carcinoma cell line, kindly provided by Michael Naumann, University of 1319 Magdeburg, Germany) were cultured in RPMI 1640 medium (20 mM Hepes and GlutaMAX stable 1320 glutamine (Gibco, Thermo Fisher Scientific) and 10 % FCS (PromoCell, Germany.) and split routinely 1321 every third day. The cell line HEK-NF-B_luc (BPSBioscience #60650, USA, luciferase reporter cell line) 1322 was cultured in DMEM supplemented with GlutaMAX (Gibco, Thermo Fisher Scientific, USA), 50 µg/µl 1323 hygromycin B (Invivogen, USA), and 10% FCS. Hygromycin was omitted from the co-incubations with 1324 bacteria. All cell cultures were grown at 37°C in a 5% CO 2 incubator and passaged using 0.05% buffered 1325 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 44 Trypsin-EDTA (Gibco, Thermo Fisher Scientific, USA). For NCI N87 cells to reach polarization, they were 1326 cultured for at least 7 days in the cell culture plates. 1327 1328 Transient transfection of human cells 1329 For transient transfection of human HEK-NF-B_luc cells, Lipofectamine 2000 transfection (Invitrogen 1330 Thermo Scientific) was applied according to the manufacturer’s instructions, in 96-well plates, using 1331 50 ng per plasmid per well. Before transfection, cell wells were supplied with 50 µl fresh Optimem 1332 medium (Invitrogen-Gibco) supplemented with 5% FCS. Plasmids transfected were: human CEACAM1 1333 (origene); human CEACAM3 (origene); human CEACAM5 and CEACAM6 expression plasmids, kindly 1334 provided by Wolfgang Zimmermann (see ref. 47 for details). Integrin- 5 (Addgene #54970), integrin-1335 1 (Addgene #54129) expression plasmids were a kind gift from Michael Davidson’s lab. All plasmids 1336 were used as endotoxin–free preparations. 20 h post-transfection, used transfection medium was 1337 removed from the wells, and cells were again supplied with 50 µl of fresh growth medium (DMEM with 1338 10% FCS). Cell co-incubations with bacteria for luciferase reporter activation assays were initiated at 1339 24 h post-transfection and carried out for 3 h, before starting the luciferase assay. 1340 1341 Co-culture of cells with live bacteria and read-out of cell activation 1342 For co-cultures of human cells with live bacteria, human cells were seeded in 6-well, 24-well or 96-well 1343 plates at a confluency of 60-80% one day prior to the co-cultivation in antibiotic –free cell culture 1344 medium. The medium was changed 1 h prior to the infection to fresh RPMI or DMEM + 10% FCS. H. 1345 pylori was harvested after ca. 20 h fresh growth from blood agar plates into the respective cell culture 1346 medium, the OD 600 of the bacterial suspension was measured, and the multiplicities of infections (MOI) 1347 were calculated. MOI = 25 was used for routine co-cultivation experiments of AGS and NCI N87 cells in 1348 24-well plates, and MOI = 50 was used for co-cultivation experiments of HEK-NF- B_luc reporter cells 1349 in 96-well plates. Bacterial suspensions were added to the cells and plates were centrifuged to start 1350 and synchronize the co-incubation (300 x g, 5 min at room temperature), and then further incubated 1351 at 37°C and 5% CO2 for different times (indicated in figures and legends). Samples were harvested after 1352 taking off the supernatant (kept for ELISA, measuring IL-8 secretion as previously described (13) using 1353 human IL-8 ELISA Set, R&D Systems), either by scraping the cells from the bottom of the plate (for RNA 1354 isolation), or by adding luminescence substrate to the cells and medium (for NF-KB-activated luciferase 1355 quantification). Luciferase production by the reporter cells was measured according to the 1356 manufacturer’s protocol using SteadyGlo Luciferase Assay (Promega). Routinely, 50 µl of cell culture 1357 medium in in each well of the 96-well plates were provided together with the adherent cells after 1358 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 45 bacterial co-incubation (for co-incubation times, see figure legends), and mixed in-plate with an equal 1359 volume of SteadyGlo luciferase cell lysis and detection reagent. After 10 min of lysis with mixing, all 1360 wells were measured in replicates in a victor Nivo multi-well plate reader in luminescence mode. 1361 1362 Hummingbird phenotype (CagA translocation by H. pylori CagT4SS) 1363 The quantitation of CagA-dependent phenotype (read-out for CagA translocation) was performed by 1364 H. pylori co-incubation of AGS cells, followed by induction and detection of cytoskeletal elongation of 1365 cells (termed the hummingbird phenotype), similarly as previously described (13). We applied the 1366 assays for two strains, 26695 (Fig 5E) and N6 (further controls, not shown). Briefly, AGS cells were 1367 seeded in wells of a 24 well plates to medium confluency (1 x 10 5 cells per well) in standard media. 1368 Cells were let attach and grow overnight. The next day, medium was changed and bacteria were co-1369 incubated with the attached cells at an MOI of 50, centrifuged to synchronize the infection, and let the 1370 incubation continue for 9.5 h, while development of hummingbird phenotype was monitored visually 1371 over time. Subsequently, cells were washed once to remove non-adherent bacteria, and cells with 1372 attached bacteria were fixed in 2% PFA in 100 mM potassium phosphate buffer, pH=7, overnight. PFA 1373 was changed once and let incubate for another hour. Cells were shifted to 1xPBS and images were 1374 taken in an Olympus IX40 microscope at 20x magnification, with a size marker. Tif images were 1375 exported and imported into ImageJ software, where they were further processed. At least 100 cells 1376 per condition, counting all cells randomly in each image, were quantitated for cell length in an 1377 evaluator-blinded fashion. Cell lengths with mean were calculated for each condition and dot graphs 1378 generated from the results in GraphPad Prism. Statistics were performed by One-way ANOVA with 1379 pairwise comparisons and Kruskal-Wallis test in GraphPad. 1380 1381 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 46 Acknowledgments 1382 Parts of the work were financially supported by a CRC 900 consortium grant awarded by the German 1383 Research Foundation (DFG) to CJ (project no. 158989968/B6) and German Center for Infection 1384 Research (DZIF) projects no. 06.809 and no. 06.820 to CJ. Bettina Sedlmaier-Erlenfeld is gratefully 1385 acknowledged for expert technical help and support throughout this study. We thank Michael 1386 Naumann for generously providing human gastric epithelial cell line NCI-N87. We are also very grateful 1387 for additional technical support by Monia Camboni. We are grateful to the intramural graduate 1388 program “Infection Research on Human Pathogens@MvPI” at the Max von Pettenkofer Institute, LMU 1389 (MMRS), and to the iLIFE RTG program initiative at LMU for supporting FM and JB. We thank all 1390 colleagues from the Josenhans laboratory for discussions and helpful suggestions. 1391 1392 Author contributions 1393 FM: performed, analysed and designed experiments, contributing to draft writing, figure design, final 1394 reading and editing; JB: performed, analyzed and designed experiments, contributing to writing 1395 (methods), figure design, final reading and editing; AL: performed and analyzed experiments, final 1396 reading and editing; GW: designed and analyzed experiments, final reading and editing; KPH: funding, 1397 experimental analysis, final reading and editing; RTAM: analyzed experiments, funding, final reading 1398 and editing; SB: performed, analyzed and designed experiments, final reading and editing; LT: designed 1399 experiments, supervision, provided materials, final reading and editing; KT: provided and analyzed 1400 large-scale data, final reading and editing. WF: designed experiments and provided materials, final 1401 reading and editing; SS: designed experiments, performed formal analysis, supervision, funding, final 1402 reading and editing; CJ: concept of study, performed, analyzed and designed experiments, supervision, 1403 funding, figure design, drafting and writing the paper, final reading and editing. 1404 1405 Declaration of interests 1406 The authors declare no competing interests. 1407 1408 Supplemental information titles 1409 1410 Supplemental information 1411 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 47 Supplementary Information S1: Supplementary Figures S1 to S7. Supplementary Information S2: S1, 1412 S2, S3, Supplementary Tables and References. 1413 S1 Table.: H. pylori and E. coli strains used and generated in this study 1414 S2 Table.: plasmids (BACTH, protein mutagenesis, expression) 1415 S3 Table.: Primers used for cloning, PCR and sequencing 1416 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 48

1417 1418 1. Marshall B, Warren JR. UNIDENTIFIED CURVED BACILLI IN THE STOMACH OF 1419 PATIENTS WITH GASTRITIS AND PEPTIC ULCERATION. The Lancet. 1420 1984;323(8390):1311-5. 1421 2. Robin Warren J, Marshall B. UNIDENTIFIED CURVED BACILLI ON GASTRIC 1422 EPITHELIUM IN ACTIVE CHRONIC GASTRITIS. The Lancet. 1983;321(8336):1273-5. 1423 3. Go MF. Review article: natural history and epidemiology of Helicobacter pylori infection. 1424 Aliment Pharmacol Ther. 2002;16 Suppl 1:3-15. 1425 4. Suerbaum S, Michetti P. Helicobacter pylori Infection. New England Journal of 1426 Medicine. 2002;347(15):1175-86. 1427 5. Hooi JKY, Lai WY, Ng WK, Suen MMY, Underwood FE, Tanyingoh D, et al. Global 1428 Prevalence of Helicobacter pylori Infection: Systematic Review and Meta-Analysis. 1429 Gastroenterology. 2017;153(2):420-9. 1430 6. Malfertheiner P, Camargo MC, El-Omar E, Liou J-M, Peek R, Schulz C, et al. 1431 Helicobacter pylori infection. Nature Reviews Disease Primers. 2023;9(1):19. 1432 7. Nomura A, Stemmermann GN, Chyou PH, Kato I, Perez-Perez GI, Blaser MJ. 1433 Helicobacter pylori infection and gastric carcinoma among Japanese Americans in Hawaii. N 1434 Engl J Med. 1991;325(16):1132-6. 1435 8. Parsonnet J, Friedman GD, Vandersteen DP, Chang Y, Vogelman JH, Orentreich N, et 1436 al. Helicobacter pylori infection and the risk of gastric carcinoma. N Engl J Med. 1437 1991;325(16):1127-31. 1438 9. Linz B, Balloux F, Moodley Y, Manica A, Liu H, Roumagnac P, et al. An African origin 1439 for the intimate association between humans and Helicobacter pylori . Nature. 1440 2007;445(7130):915-8. 1441 10. Olbermann P, Josenhans C, Moodley Y, Uhr M, Stamer C, Vauterin M, et al. A global 1442 overview of the genetic and functional diversity in the Helicobacter pylori cag pathogenicity 1443 island. PLoS Genet. 2010;6(8):e1001069. 1444 11. Censini S, Lange C, Xiang Z, Crabtree JE, Ghiara P, Borodovsky M, et al. cag, a 1445 pathogenicity island of Helicobacter pylori , encodes type I-specific and disease-associated 1446 virulence factors. Proceedings of the National Academy of Sciences of the United States of 1447 America. 1996;93(25):14648-53. 1448 12. Covacci A, Rappuoli R. Helicobacter pylori: molecular evolution of a bacterial quasi-1449 species. Curr Opin Microbiol. 1998;1(1):96-102. 1450 13. Stein SC, Faber E, Bats SH, Murillo T, Speidel Y, Coombs N, et al. Helicobacter pylori 1451 modulates host cell responses by CagT4SS-dependent translocation of an intermediate 1452 metabolite of LPS inner core heptose biosynthesis. PLoS Pathog. 2017;13(7):e1006514. 1453 14. Zimmermann S, Pfannkuch L, Al-Zeer MA, Bartfeld S, Koch M, Liu J, et al. ALPK1- and 1454 TIFA-Dependent Innate Immune Response Triggered by the Helicobacter pylori Type IV 1455 Secretion System. Cell Rep. 2017;20(10):2384-95. 1456 15. Zhao Q, Busch B, Jiménez-Soto LF, Ishikawa-Ankerhold H, Massberg S, Terradot L, 1457 et al. Integrin but not CEACAM receptors are dispensable for Helicobacter pylori CagA 1458 translocation. PLoS Pathog. 2018;14(10):e1007359. 1459 16. Tran SC, Bryant KN, Cover TL. The Helicobacter pylori cag pathogenicity island as a 1460 determinant of gastric cancer risk. Gut Microbes. 2024;16(1):2314201. 1461 17. Denic M, Touati E, De Reuse H. Review: Pathogenesis of Helicobacter pylori infection. 1462 Helicobacter. 2020;25 Suppl 1:e12736. 1463 18. Costa TRD, Patkowski JB, Macé K, Christie PJ, Waksman G. Structural and functional 1464 diversity of type IV secretion systems. Nat Rev Microbiol. 2024;22(3):170-85. 1465 19. Fischer W. Assembly and molecular mode of action of the Helicobacter pylori Cag type 1466 IV secretion apparatus. Febs j. 2011;278(8):1203-12. 1467 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 49 20. Hu B, Khara P, Song L, Lin AS, Frick-Cheng AE, Harvey ML, et al. In Situ Molecular 1468 Architecture of the Helicobacter pylori Cag Type IV Secretion System. mBio. 1469 2019;10(3):10.1128/mbio.00849-19. 1470 21. Sheedlo MJ, Ohi MD, Lacy DB, Cover TL. Molecular architecture of bacterial type IV 1471 secretion systems. PLOS Pathogens. 2022;18(8):e1010720. 1472 22. Bats SH, Bergé C, Coombs N, Terradot L, Josenhans C. Biochemical characterization 1473 of the Helicobacter pylori Cag Type 4 Secretion System protein CagN and its interaction 1474 partner CagM. Int J Med Microbiol. 2018;308(4):425-37. 1475 23. Costa TRD, Felisberto-Rodrigues C, Meir A, Prevost MS, Redzej A, Trokter M, et al. 1476 Secretion systems in Gram-negative bacteria: structural and mechanistic insights. Nature 1477 Reviews Microbiology. 2015;13(6):343-59. 1478 24. Green ER, Mecsas J. Bacterial Secretion Systems: An Overview. Microbiol Spectr. 1479 2016;4(1). 1480 25. Fronzes R, Schäfer E, Wang L, Saibil HR, Orlova EV, Waksman G. Structure of a type 1481 IV secretion system core complex. Science. 2009;323(5911):266-8. 1482 26. Böck D, Hüsler D, Steiner B, Medeiros JM, Welin A, Radomska KA, et al. The Polar 1483 Legionella I cm/Dot T4SS Establishes Distinct Contact Sites with the Pathogen Vacuole 1484 Membrane. mBio. 2021;12(5):e0218021. 1485 27. Li YG, Hu B, Christie PJ. Biological and Structural Diversity of Type IV Secretion 1486 Systems. Microbiol Spectr. 2019;7(2). 1487 28. Lin AS, Dooyema SDR, Frick-Cheng AE, Harvey ML, Suarez G, Loh, JR, et al. Bacterial 1488 energetic requirements for Helicobacter pylori Cag Type IV Secretion System-dependent 1489 alterations in gastric epithelial cells. Infect Immun. 2020;88(2):e00790-19. 1490 29. Chang YW, Shaffer CL, Rettberg LA, Ghosal D, Jensen GJ. In Vivo Structures of the 1491 Helicobacter pylori cag Type IV Secretion System. Cell Rep. 2018;23(3):673-81. 1492 30. Chung JM, Sheedlo MJ, Campbell AM, Sawhney N, Frick-Cheng AE, Lacy DB, et al. 1493 Structure of the Helicobacter pylori Cag type IV secretion system. eLife. 2019;8:e47644. 1494 31. Kutter S, Buhrdorf R, Haas J, Schneider-Brachert W, Haas R, Fischer W. Protein 1495 subassemblies of the Helicobacter pylori Cag type IV secretion system revealed by localization 1496 and interaction studies. J Bacteriol. 2008;190(6):2161-71. 1497 32. Busler VJ, Torres VJ, McClain MS, Tirado O, Friedman DB, Cover TL. Protein-protein 1498 interactions among Helicobacter pylori cag proteins. J Bacteriol. 2006;188(13):4787-800. 1499 33. Sheedlo MJ, Chung JM, Sawhney N, Durie CL, Cover TL, Ohi MD, et al. Cryo-EM 1500 reveals species-specific components within the Helicobacter pylori Cag type IV secretion 1501 system core complex. eLife. 2020;9:e59495. 1502 34. Ryan ME, Damke PP, Bryant C, Sheedlo MJ, Shaffer CL. Architectural asymmetry 1503 enables DNA transport through the Helicobacter pylori Cag type IV secretion system. bioRxiv. 1504 2023. 1505 35. Andrzejewska J, Lee SK, Olbermann P, Lotzing N, Katzowitsch E, Linz B, et al. 1506 Characterization of the pilin ortholog of the Helicobacter pylori type IV cag pathogenicity 1507 apparatus, a surface-associated protein expressed during infection. J Bacteriol. 1508 2006;188(16):5865-77. 1509 36. Pham KT, Weiss E, Jiménez Soto LF, Breithaupt U, Haas R, Fischer W. CagI is an 1510 essential component of the Helicobacter pylori Cag type IV secretion system and forms a 1511 complex with CagL. PLoS One. 2012;7(4):e35341. 1512 37. Bönig T, Olbermann P, Bats SH, Fischer W, Josenhans C. Systematic site-directed 1513 mutagenesis of the Helicobacter pylori CagL protein of the Cag type IV secretion system 1514 identifies novel functional domains. Scientific Reports. 2016;6(1):38101. 1515 38. Bourzac KM, Satkamp LA, Guillemin K. The Helicobacter pylori cag pathogenicity 1516 island protein CagN is a bacterial membrane-associated protein that is processed at its C 1517 terminus. Infect Immun. 2006;74(5):2537-43. 1518 39. Johnson EM, Gaddy JA, Voss BJ, Hennig EE, Cover TL. Genes required for assembly 1519 of pili associated with the Helicobacter pylori cag type IV secretion system. Infect Immun. 1520 2014;82(8):3457-70. 1521 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 50 40. Shaffer CL, Gaddy JA, Loh JT, Johnson EM, Hill S, Hennig EE, et al. Helicobacter pylori 1522 exploits a unique repertoire of type IV secretion system components for pilus assembly at the 1523 bacteria-host cell interface. PLoS Pathog. 2011;7(9):e1002237. 1524 41. Merino E, Flores-Encarnación M, Aguilar-Gutiérrez GR. Functional interaction and 1525 structural characteristics of unique components of Helicobacter pylori T4SS. Febs j. 1526 2017;284(21):3540-9. 1527 42. Fischer W, Püls J, Buhrdorf R, Gebert B, Odenbreit S, Haas R. Systematic mutagenesis 1528 of the Helicobacter pylori cag pathogenicity island: essential genes for CagA translocation in 1529 host cells and induction of interleukin-8. Molecular Microbiology. 2001;42(5):1337-48. 1530 43. Faass L, Hauke M, Stein SC, Josenhans C. Innate activation of human neutrophils and 1531 neutrophil-like cells by the pro-inflammatory bacterial metabolite ADP-heptose and 1532 Helicobacter pylori. Int J Med Microbiol. 2023;313(4):151585. 1533 44. Jiménez-Soto LF, Kutter S, Sewald X, Ertl C, Weiss E, Kapp U, et al. Helicobacter pylori 1534 type IV secretion apparatus exploits beta1 integrin in a novel RGD-independent manner. PLoS 1535 Pathog. 2009;5(12):e1000684. 1536 45. Backert S, Tegtmeyer N. Type IV Secretion and Signal Transduction of Helicobacter 1537 pylori CagA through Interactions with Host Cell Receptors. Toxins. 2017;9(4):115. 1538 46. Koelblen T, Bergé C, Cherrier MV, Brillet K, Jimenez-Soto L, Ballut L, et al. Molecular 1539 dissection of protein-protein interactions between integrin α5β1 and the Helicobacter pylori 1540 Cag type IV secretion system. Febs j. 2017;284(23):4143-57. 1541 47. Königer V, Holsten L, Harrison U, Busch B, Loell E, Zhao Q, et al. Helicobacter pylori 1542 exploits human CEACAMs via HopQ for adherence and translocation of CagA. Nature 1543 Microbiology. 2016;2(1):16188. 1544 48. Alm RA, Bina J, Andrews BM, Doig P, Hancock RE, Trust TJ. Comparative genomics 1545 of Helicobacter pylori : analysis of the outer membrane protein families. Infect Immun. 1546 2000;68(7):4155-68. 1547 49. Ilver D, Arnqvist A, Ogren J, Frick IM, Kersulyte D, Incecik ET, et al. Helicobacter pylori 1548 adhesin binding fucosylated histo-blood group antigens revealed by retagging. Science. 1549 1998;279(5349):373-7. 1550 50. Mahdavi J, Sondén B, Hurtig M, Olfat FO, Forsberg L, Roche N, et al. Helicobacter 1551 pylori SabA adhesin in persistent infection and chronic inflammation. Science. 1552 2002;297(5581):573-8. 1553 51. Cao P, Cover TL. Two Different Families of hopQ Alleles in Helicobacter pylori. Journal 1554 of Clinical Microbiology. 2002;40(12):4504-11. 1555 52. Yakoob J, Abbas Z, Jafri W, Khan R, Salim SA, Awan S, et al. Helicobacter pylori outer 1556 membrane protein and virulence marker differences in expatriate patients. Epidemiol Infect. 1557 2016;144(10):2200-8. 1558 53. Atherton JC, Cao P, Peek RM, Jr., Tummuru MK, Blaser MJ, Cover TL. Mosaicism in 1559 vacuolating cytotoxin alleles of Helicobacter pylori . Association of specific vacA types with 1560 cytotoxin production and peptic ulceration. J Biol Chem. 1995;270(30):17771-7. 1561 54. Rudi J, Kuck D, Rudy A, Sieg A, Maiwald M, Stremmel W. Helicobacter pylori vacA 1562 genotypes and cagA gene in a series of 383 H. pylori -positive patients. Zeitschrift fur 1563 Gastroenterologie. 2000;38(7):559-64. 1564 55. Ohno T, Sugimoto M, Nagashima A, Ogiwara H, Vilaichone RK, Mahachai V, et al. 1565 Relationship between Helicobacter pylori hopQ genotype and clinical outcome in Asian and 1566 Western populations. J Gastroenterol Hepatol. 2009;24(3):462-8. 1567 56. Moonens K, Hamway Y, Neddermann M, Reschke M, Tegtmeyer N, Kruse T, et al. 1568 Helicobacter pylori adhesin HopQ disrupts trans dimerization in human CEACAMs. The EMBO 1569 Journal. 2018;37(13):e98665. 1570 57. Bonsor DA, Zhao Q, Schmidinger B, Weiss E, Wang J, Deredge D, et al. The 1571 Helicobacter pylori adhesin protein HopQ exploits the dimer interface of human CEACAMs to 1572 facilitate translocation of the oncoprotein CagA. EMBO J. 2018;37(13). 1573 58. Belogolova E, Bauer B, Pompaiah M, Asakura H, Brinkman V, Ertl C, et al. Helicobacter 1574 pylori outer membrane protein HopQ identified as a novel T4SS-associated virulence factor. 1575 Cell Microbiol. 2013;15(11):1896-912. 1576 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 51 59. Javaheri A, Kruse T, Moonens K, Mejías-Luque R, Debraekeleer A, Asche CI, et al. 1577 Helicobacter pylori adhesin HopQ engages in a virulence-enhancing interaction with human 1578 CEACAMs. Nature Microbiology. 2016;2(1):16189. 1579 60. Kennemann L, Brenneke B, Andres S, Engstrand L, Meyer TF, Aebischer T, et al. In 1580 vivo sequence variation in HopZ, a phase-variable outer membrane protein of Helicobacter 1581 pylori. Infect Immun. 2012;80(12):4364-73. 1582 61. Peck B, Ortkamp M, Diehl KD, Hundt E, Knapp B. Conservation, localization and 1583 expression of HopZ, a protein involved in adhesion of Helicobacter pylori. Nucleic Acids Res. 1584 1999;27(16):3325-33. 1585 62. Nell S, Estibariz I, Krebes J, Bunk B, Graham DY, Overmann J, et al. Genome and 1586 Methylome Variation in Helicobacter pylori With a cag Pathogenicity Island During Early 1587 Stages of Human Infection. Gastroenterology. 2018;154(3):612-23.e7. 1588 63. Schreiber S, Konradt M, Groll C, Scheid P, Hanauer G, Werling HO, et al. The spatial 1589 orientation of Helicobacter pylori in the gastric mucus. Proceedings of the National Academy 1590 of Sciences of the United States of America. 2004;101(14):5024-9. 1591 64. Bugaytsova JA, Björnham O, Chernov YA, Gideonsson P, Henriksson S, Mendez M, 1592 et al. Helicobacter pylori Adapts to Chronic Infection and Gastric Disease via pH-Responsive 1593 BabA-Mediated Adherence. Cell Host Microbe. 2017;21(3):376-89. 1594 65. Bonsor DA, Pham KT, Beadenkopf R, Diederichs K, Haas R, Beckett D, et al. Integrin 1595 engagement by the helical RGD motif of the Helicobacter pylori CagL protein is regulated by 1596 pH-induced displacement of a neighboring helix. J Biol Chem. 2015;290(20):12929-40. 1597 66. Kwok T, Zabler D, Urman S, Rohde M, Hartig R, Wessler S, et al. Helicobacter exploits 1598 integrin for type IV secretion and kinase activation. Nature. 2007;449(7164):862-6. 1599 67. Blanc M, Lettl C, Guérin J, Vieille A, Furler S, Briand-Schumacher S, et al. Designed 1600 Ankyrin Repeat Proteins provide insights into the structure and function of CagI and are potent 1601 inhibitors of CagA translocation by the Helicobacter pylori type IV secretion system. PLoS 1602 Pathog. 2023;19(5):e1011368. 1603 68. Segal ED, Cha J, Lo J, Falkow S, Tompkins LS. Altered states: involvement of 1604 phosphorylated CagA in the induction of host cellular growth changes by Helicobacter pylori. 1605 Proc Natl Acad Sci U S A. 1999;96(25):14559-64. 1606 69. Backert S, Moese S, Selbach M, Brinkmann V, Meyer TF. Phosphorylation of tyrosine 1607 972 of the Helicobacter pylori CagA protein is essential for induction of a scattering phenotype 1608 in gastric epithelial cells. Mol Microbiol. 2001;42(3):631-44. 1609 70. Kaplan-Türköz B, Jiménez-Soto LF, Dian C, Ertl C, Remaut H, Louche A, et al. 1610 Structural insights into Helicobacter pylori oncoprotein CagA interaction with β1 integrin. 1611 Proceedings of the National Academy of Sciences of the United States of America. 1612 2012;109(36):14640-5. 1613 71. Cao P, Cover TL. Two different families of hopQ alleles in Helicobacter pylori. J Clin 1614 Microbiol. 2002;40(12):4504-11. 1615 72. Rohde M, Püls J, Buhrdorf R, Fischer W, Haas R. A novel sheathed surface organelle 1616 of the Helicobacter pylori cag type IV secretion system. Mol Microbiol. 2003;49(1):219-34. doi: 1617 10.1046/j.1365-2958.2003.03549.x. 1618 73. Bryant KN, Frick-Cheng AE, Solecki LE, Kroh HK, McDonald WH, Lacy DB, McClain 1619 MS, Ohi MD, Cover TL. Species-specific components of the Helicobacter pylori Cag type IV 1620 secretion system. Infect Immun. 2025;93(5):e0049324. doi: 10.1128/iai.00493-24. 1621 74. Thorell K, Muñoz-Ramírez ZY, Wang D, Sandoval-Motta S, Boscolo Agostini R, 1622 Ghirotto S, et al. The Helicobacter pylori Genome Project: insights into H. pylori population 1623 structure from analysis of a worldwide collection of complete genomes. Nat Commun. 1624 2023;14(1):8184. 1625 75. You S, Li J, Chua EG, Tay A, Benghezal M, Marshall B, Tang H, Li H. Helicobacter 1626 pylori outer membrane protein families and related pathogenesis. Microb Pathog. 1627 2025;206:107740. 1628 76. Karimova G, Pidoux J, Ullmann A, Ladant D. A bacterial two-hybrid system based on a 1629 reconstituted signal transduction pathway. Proceedings of the National Academy of Sciences. 1630 1998;95(10):5752-6. 1631 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint 52 77. Bubendorfer S, Krebes J, Yang I, Hage E, Schulz TF, Bahlawane C, et al. Genome-1632 wide analysis of chromosomal import patterns after natural transformation of Helicobacter 1633 pylori. Nature Communications. 2016;7(1):11995. 1634 78. Kolber NS, Fattal R, Bratulic S, Carver GD, Badran AH. Orthogonal translation enables 1635 heterologous ribosome engineering in E. coli. Nat Commun. 2021;12(1):599. 1636 79. Estibariz I, Overmann A, Ailloud F, Krebes J, Josenhans C, Suerbaum S. The core 1637 genome m5C methyltransferase JHP1050 (M.Hpy99III) plays an important role in orchestrating 1638 gene expression in Helicobacter pylori. Nucleic Acids Res. 2019;47(5):2336-48. 1639 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 1, 2025. ; https://doi.org/10.1101/2025.09.01.673027doi: bioRxiv preprint