TIFA Signaling in Gastric Epithelial Cells Initiates the cag Type 4 Secretion System-Dependent Innate Immune Response to Helicobacter pylori Infection

ABSTRACT Helicobacter pylori is a bacterial pathogen that colonizes the human stomach, causing inflammation which, in some cases, leads to gastric ulcers and cancer. The clinical outcome of infection depends on a complex interplay of bacterial, host genetic, and environmental factors. Although H. pylori is recognized by both the innate and adaptive immune systems, this rarely results in bacterial clearance. Gastric epithelial cells are the first line of defense against H. pylori and alert the immune system to bacterial presence. Cytosolic delivery of proinflammatory bacterial factors through the cag type 4 secretion system (cag-T4SS) has long been appreciated as the major mechanism by which gastric epithelial cells detect H. pylori. Classically attributed to the peptidoglycan sensor NOD1, recent work has highlighted the role of NOD1-independent pathways in detecting H. pylori; however, the bacterial and host factors involved have remained unknown. Here, we show that bacterially derived heptose-1,7-bisphosphate (HBP), a metabolic precursor in lipopolysaccharide (LPS) biosynthesis, is delivered to the host cytosol through the cag-T4SS, where it activates the host tumor necrosis factor receptor-associated factor (TRAF)-interacting protein with forkhead-associated domain (TIFA)-dependent cytosolic surveillance pathway. This response, which is independent of NOD1, drives robust NF-κB-dependent inflammation within hours of infection and precedes NOD1 activation. We also found that the CagA toxin contributes to the NF-κB-driven response subsequent to TIFA and NOD1 activation. Taken together, our results indicate that the sequential activation of TIFA, NOD1, and CagA delivery drives the initial inflammatory response in gastric epithelial cells, orchestrating the subsequent recruitment of immune cells and leading to chronic gastritis.

H elicobacter pylori is a Gram-negative bacterium that colonizes greater than 50% of the world's population. The only known niche of H. pylori is the human stomach, where the bacterium resides in intimate contact with gastric epithelial cells, causing inflammation that, in a subset of infected individuals, progresses to gastric and duodenal ulcers and cancer (1). It has been estimated that H. pylori has resided with its human host for at least 100,000 years, predating modern human migration out of Africa, making it the oldest known and one of the most successful human pathogens (2,3).
H. pylori is able to achieve chronic colonization due in part to its ability to evade host immune responses. A number of H. pylori pathogen-associated molecular patterns (PAMPs) avoid stimulation of their cognate pattern recognition receptors (PRRs). The lipopolysaccharide (LPS) of H. pylori is tetra-acetylated, making it a poor ligand for Toll-like receptor 4 (TLR4) (4), and divergent flagellin monomer sequences abrogate its interaction with TLR5 (5,6). The O-antigen of H. pylori LPS contains Lewis antigens which mimic host receptors and facilitate immune escape (7). Furthermore, the unique dephosphorylated lipid A structure in H. pylori confers resistance to host antimicrobial peptides, as well as to the antibiotic polymyxin B (8). Despite sophisticated evasion strategies, several virulence factors induce a robust immune response. H. pylori-mediated gastritis is initiated by proinflammatory cytokines released by gastric epithelial cells, as well as neutrophils and macrophages that are recruited to the site of infection (9). Cytotoxin-associated gene A (CagA) has been extensively characterized as a major contributor to epithelial cell transformation that can result in gastric cancer. Infection with cagA-positive (cagA ϩ ) strains is associated with increased inflammation and subsequent development of peptic ulcers and gastric adenocarcinoma (10,11). CagA is encoded within a region of the chromosome known as the cag pathogenicity island (cag-PAI). Along with cagA, cag-PAI contains genes required for assembly of a type IV secretion system (T4SS). CagA is injected directly into epithelial cells via the cag-T4SS, which acts as a molecular syringe. Once CagA is delivered into the host cytosol, it is phosphorylated by host kinases, leading to aberrant host cell signaling. After several months of experimental infection performed with a single H. pylori strain, mice and rhesus macaques accumulated bacterial strain variants that contained mutations in the cagY gene which rendered the cag-T4SS nonfunctional. Interestingly, strains with a functional cag-T4SS were also recovered from the same animals (12,13), suggesting that the ability to switch the cag-T4SS on or off is advantageous for the bacteria.
Along with CagA, the cag-T4SS delivers bacterial factors that are known to activate the global transcription factor NF-B, leading to the upregulation of proinflammatory genes and production of cytokines such as interleukin-8 (IL-8) (14). IL-8 is a major chemokine for neutrophil recruitment to the site of infection and has been extensively used as a physiologically relevant readout of the NF-B inflammatory pathway. Active H. pylori infection is characterized by extensive neutrophil infiltration of the gastric mucosa, where these cells not only recruit lymphocytes, which establish chronic inflammation, but also contribute to tissue damage through production of reactive oxygen species (15). Prior to this study, most of the cag-T4SS-dependent IL-8 response in gastric epithelial cells had been attributed to NOD1 activation on the basis of a foundational study demonstrating that NOD1 activation by H. pylori peptidoglycan (PG) leads to NF-B activation and IL-8 production and that Nod1-deficient mice are more susceptible to H. pylori infection (16). Although several groups have challenged the notion that NOD1-dependent responses predominate in H. pylori-infected gastric epithelial cells (17)(18)(19), comprehensive understanding of the relevant pathogen recognition pathways has remained elusive. Interestingly, recent work has suggested that additional bacterial products may be delivered by the cag-T4SS. TLR9-dependent inhibition of IL-8 induction in reporter cell lines and suppression of the H. pylori load in Tlr9 Ϫ/Ϫ mice suggest that bacterial DNA containing inhibitory CpG motifs can also transit the cag-T4SS, though in this case with the result of dampening NF-B signaling (20,21).
In this study, we identified tumor necrosis factor receptor-associated factor (TRAF)interacting protein with forkhead-associated domain (TIFA) as a critical innate signaling component downstream of cag-T4SS-induced inflammation. TIFA has been extensively studied in the context of its interaction with TRAF6 and ability to activate NF-B (22)(23)(24). Recent work has revealed that TIFA is specifically activated by the cytosolic presence of the bacterial metabolite heptose-1,7-bisphosphate (HBP), a metabolic intermediate in LPS biosynthesis (25,26). HBP is highly conserved among Gram-negative bacteria but is absent from eukaryotic cells. Thus, TIFA-dependent detection of HBP represents a novel innate immune-sensing pathway that is distinct from classical TLR or NOD family pattern recognition receptor-driven signaling cascades. HBP is detected in the host cytosol and has been proposed to gain access to this compartment through dynamin-dependent endocytosis of extracellularly released HBP (Neisseria, Salmonella enterica serovar Typhimurium) and phagolysosomal degradation of engulfed bacteria within macrophages (Escherichia coli) as well as through intracellular bacterial replication (Shigella flexneri) (25)(26)(27)(28). Here, we reveal an additional mechanism whereby HBP is presented to the TIFA signaling pathway: type IV secretion system translocation. In the context of H. pylori infection, HBP is delivered to the host cytoplasm through the cag-T4SS, where it activates an acute TIFA-mediated innate immune response in gastric epithelial cells. Furthermore, we show that host NOD1 and the bacterial CagA toxin are sequentially responsible for the H. pylori-induced inflammation that follows the early TIFA response.

RESULTS
NOD1 gene targeting does not eliminate the cag-T4SS-dependent IL-8 response in gastric epithelial cells. To determine whether NOD1 was solely responsible for the IL-8 response in gastric epithelial cells, we used clustered regularly interspaced short palindromic repeat (CRISPR)/Cas9 gene editing tools to target NOD1 in a human gastric epithelial cell line (AGS cells). We designed a guide RNA targeting the N-terminal caspase recruitment domain (CARD) of NOD1, which is required for interaction with its downstream signaling partner RIP2 (29). We clonally selected two NOD1 knockout (KO) cell lines and characterized the deletions on both alleles of NOD1. The NOD1 KO#1 cell line contained identical 2-bp deletions adjacent to the PAM motif, while the NOD1 KO#2 cell line contained a 112-bp deletion on the first allele and a 2-bp deletion on the second allele (Fig. 1A). To assess the functional loss of NOD1, we treated control or NOD1-targeted cells with a NOD1-specific ligand (C12-iE-DAP). We found that the response to NOD1-specific ligand was completely abrogated in NOD1 KO cells compared to control targeted cells, and yet all cell lines retained the IL-8 response to IL-1␤, a NOD1-independent stimulus (Fig. 1B). When we cocultured control-targeted AGS cells with G27, a wild-type clinical isolate of H. pylori, we found that the cells produced IL-8 in response to infection and that, consistent with published data, the IL-8 response was completely dependent on the presence of an intact cag-T4SS. Surprisingly, NOD1 KO attenuated but did not abolish IL-8 production in response to infection with wild-type H. pylori (Fig. 1C), suggesting that an additional host pathway(s) is activated by bacterial factors delivered through the cag-T4SS.
TIFA gene targeting in epithelial cells suppresses the IL-8 response to H. pylori infection. Given the recent discovery of TIFA as a critical innate immune signaling component in response to Gram-negative bacteria, we hypothesized that TIFA may be playing a role in H. pylori detection by gastric epithelial cells. Using CRISPR, we targeted TIFA in wild-type or NOD1 KO AGS cells and assessed IL-8 responses following exposure to strain G27, a clinical isolate of H. pylori. We determined that in response to H. pylori infection, IL-8 responses were significantly attenuated in TIFA KO cells, particularly at the early time point (6 h), compared to the control targeted cells ( Fig. 2A). To address the reproducibility of NOD1-and TIFA-dependent IL-8 signaling in AGS cells infected with H. pylori, we analyzed our data in aggregate across three independent experiments. We found that H. pylori induced NOD1-dependent IL-8 production was attenuated at 24 h. However, IL-8 levels at 6 and 12 h in NOD1 KO cells differed across experiments and, analyzed across biological replicates, were not statistically significantly different from the levels seen with control AGS cells (see Fig. S1A in the supplemental material). TIFA KO AGS cells infected with H. pylori consistently showed significantly reduced IL-8 levels at 6, 12, and 24 h, both in analysis of a representative experiment ( Fig. 2A) and across three independent experiments (Fig. S1A). Because we observed some differences in analyses of representative experiments versus combined data from three independent experiments, we include statistical analysis of biological replicates (Fig. S1). To assess whether TIFA-dependent IL-8 induction was unique to the G27 strain, we cocultured TIFA KO or TIFA NOD1 DKO AGS cells with different H. pylori strains (G27, PMSS1, J99, and 26695). We found that, although the overall IL-8 levels differed across strains, all of the strains tested induced TIFA-dependent IL-8 production in AGS cells (Fig. S2A). In addition, we found that the effects of targeting both NOD1 and TIFA were additive, since the NOD1 TIFA DKO cells showed a more dramatic decrease in IL-8 responses than was seen in targeting either NOD1 or TIFA alone ( Fig. 2A; Fig. S1A; Fig. S2A). This indicates that these two pathogen recognition pathways contribute to H. pylori detection and yet are functionally independent. Interestingly, we observed residual IL-8 signal at the later time points, suggesting that in addition to peptidoglycan and HBP, at least one other bacterially derived factor delivered through the cag-T4SS is driving an immune response in gastric epithelial cells.
To ensure that the effect on IL-8 production was TIFA-dependent and did not result from CRISPR off-target effects, we transduced CRISPR-targeted AGS cells with lentivirus containing the complete TIFA coding sequence and showed that the IL-8 produced in response to H. pylori infection in TIFA-complemented cells was restored. We found that  Neisseria lysates. Supernatants were collected after 6 h, and IL-8 levels were measured by ELISA. (E) Control, TIFA KO, C-terminally tagged wild-type TIFA-complemented, N-terminally tagged wild-type, or T9A TIFA-complemented AGS cells were cocultured with H. pylori, and IL-8 levels were measured in culture supernatants after 6 h of infection. For coculture experiments (A to C and E), H. pylori was added at MOI ϭ 10, supernatant was collected at the indicated time points, and each condition was tested in triplicate, with bars showing means and standard deviations. (F) NF-B-luciferase activity in wild-type or TIFA-targeted 293T cells treated with TNF-␣ (5 ng/ml), N. meningitidesderived HBP and HMP, or wild-type H. pylori lysate. Data were normalized to signal from cotransfected Renilla luciferase plasmid and are represented as normalized fold changes from mock-treated samples. Each condition was tested in triplicate, with bars showing means with standard deviations. Data shown in panels A to F are representative of results from at least two independent experiments. Statistical significance was determined using ANOVA with Bonferroni correction for multiple comparisons. ns, not significant (P Ͼ 0.05); **, P Ͻ 0.01; ***, P Ͻ 0.001. at the 6-h time point, the level of IL-8 induced in TIFA-complemented cells was statistically significantly higher than in control targeted cells (P ϭ 0.001), which may have been due to TIFA overexpression since TIFA expression in the complemented cells is driven by the lentiviral MND promoter (Fig. 2B, Fig. S1B). Furthermore, since TIFA signaling relies on oligomerization, which can be caused by protein overexpression, we confirmed that no IL-8 was produced in mock-treated TIFA-complemented cells (Fig. 2B).
Activation of the TIFA-mediated inflammatory response to H. pylori is cag-T4SS dependent. The IL-8 response to H. pylori in AGS cells is dependent on a functional cag-T4SS, since infection with a cag-T4SS mutant (strain ΔcagE) resulted in complete loss of IL-8 production (Fig. 1C). Similarly to wild-type AGS cells, TIFA KO AGS cells do not produce IL-8 in response to ΔcagE mutant H. pylori (data not shown). Importantly, our data indicate that, without a functional cag-T4SS, the TIFA-dependent immune response to H. pylori is completely abrogated. To confirm that H. pylori activates a TIFA-mediated inflammatory response and that it is dependent on the cag-T4SS, we cocultured H. pylori with HCT116 cells, a colorectal adenocarcinoma cell line that, like AGS cells, expresses the ␣5␤1 integrin required for cag-T4SS attachment and CagA translocation (30,31) and has a functional TIFA signaling pathway (28). When we infected control or TIFA-targeted HCT116 cells with H. pylori, we found that in the absence of TIFA, the IL-8 response was completely abrogated, while the IL-8 response to a TIFA-independent ligand, IL-1␤, was intact in both control and TIFA-targeted cells. Consistent with our data determined in AGS cells, we found that IL-8 produced by colon epithelial cells in response to H. pylori is completely cag-T4SS dependent, since no IL-8 was produced when we performed a coculture with a ΔcagE mutant ( Fig. 2C; Fig. S1C). However, in contrast to AGS cells, the IL-8 induced following H. pylori detection in the context of this colon epithelial cell line appears to have been completely TIFA dependent and did not rely on NOD1 activation or sensing of the NOD1-and TIFAindependent bacterially derived factor.
Gastric epithelial cells robustly respond to bacterially derived HBP. To assess whether AGS cells are able to produce IL-8 in response to HBP stimulation, we reversibly permeabilized control or TIFA KO AGS cells with digitonin and treated them with Neisseria lysates containing HBP (ΔgmhB) or heptose-7-monophosphate (HMP; ΔhldA) (26). Control AGS cells treated with HBP showed an accumulation of IL-8 over a 24-h time period, whereas HMP-treated cells showed substantially lower IL-8 responses.
Notably, when we assessed TIFA KO AGS cells treated with HBP, we found the same reduction in IL-8 responses as in control targeted AGS cells treated with HMP ( Fig. 2D;  Fig. S1D). These data suggest that AGS cells recognize HBP and activate TIFA to drive the NF-B-dependent IL-8 response.
Published reports have demonstrated that HBP-inducible phosphorylation of TIFA at the T9 amino acid drives intermolecular oligomerization, which recruits and activates its downstream signaling partner TRAF6 (25,27,28). Therefore, we assessed whether the T9 residue is essential for TIFA activation in AGS cells during H. pylori infection. We stably complemented TIFA KO cells with either wild-type N-terminally tagged TIFA or a T9A substitution containing TIFA. We found that, similarly to TIFA KO AGS cells, IL-8 responses were reduced in the T9A complemented cells. H. pylori infection of both Nand C-terminally tagged wild-type complemented cells resulted in control levels of IL-8 ( Fig. 2E; Fig. S1E). These data are consistent with previous studies and suggest that in AGS cells, H. pylori-induced inflammation requires phosphorylation of TIFA at T9 in order to activate TRAF6 and propagate the NF-B-driven immune response.
To determine whether H. pylori cell lysates containing HBP could activate NF-B in a TIFA-dependent manner, we treated NF-B-luciferase reporter human embryonic kidney (293T) cells with various ligands and measured luciferase reporter activity. 293T cells have high transfection efficiency and lack endogenous TLR and some nucleic acid-sensing pathways (32,33), allowing dissection of the TIFA-dependent response with little contribution of other PAMP sensing pathways. Consistent with published data (25, 28), we found that wild-type reporter cells robustly responded to HBP, but not to the upstream monophosphorylated derivative HMP, and that the induction of NF-B-luciferase activity was largely dependent on TIFA. When we treated the reporter cells with G27 H. pylori cell lysates ( Fig. 2F; Fig. S1F) or lysates from different H. pylori strain backgrounds (Fig. S2B), only the wild-type 293T cells responded and not the TIFA KO cells, showing a greater than 100-fold induction of NF-B-luciferase. Both wild-type and TIFA KO cells responded to a TIFA-independent stimulus, tumor necrosis factor alpha (TNF-␣) ( Fig. 2F; Fig. S1F). Together, these data confirm that cag-T4SS-dependent delivery of H. pylori HBP into host epithelial cells triggers the activation of TIFA, leading to NF-B activation and subsequent production of IL-8.
H. pylori heptose biosynthesis mutants are filamentous and have delayed growth. In Gram-negative bacteria, the biosynthesis of D,D-heptose and L,D-heptose, which are incorporated into the core oligosaccharide of LPS, occurs in the bacterial cytosol by the activity of highly conserved enzymes. In E. coli, the reaction begins with the conversion of sedoheptulose-7-P to D,D-heptose-7-P (HMP) by isomerase GmhA, followed by the heptokinase activity of bifunctional HldE to generate D,D-heptose-1,7-PP (HBP, the molecule that activates TIFA). HBP is then converted to D,D-heptose-1-P by phosphatase GmhB, which is further modified by the ADP-transferase activity of HldE to form ADP-D,D-heptose. ADP-D,D-heptose is then converted to ADP-L,D-heptose by epimerase HldD. WaaC then transfers the first L,D-heptose to 3-deoxy-D-mannooctulosonic acid (Kdo) for ongoing lipopolysaccharide synthesis (34). Sequenced H. pylori strains contain conserved homologues of the four E. coli enzymes involved in ADP-heptose biosynthesis (Fig. 3A). Using a phosphate release assay, Yu et al. showed that when recombinant H. pylori enzymes GmhA, HldE, and GmhB were combined with a sedoheptulose-7-P substrate, significantly more phosphate was released than was seen with the enzymes alone or with an enzyme mixture that was missing HldE, supporting the hypothesis that these proteins function similarly to the ADP-heptose biosynthesis pathway described for other Gram-negative bacteria (35). Furthermore, the crystal structure of H. pylori HldD revealed that both the substrate binding site and the overall structure were highly similar to those of the HldD of E. coli, suggesting a conserved enzymatic function (36). It has also been shown that deletion of several enzymes involved in assembly of the highly conserved LPS core in H. pylori results in mutants that are unable to colonize mice, typifying the essential nature of most PAMPs (37,38). Furthermore, despite the fact that waaC was the first LPS biosynthesis gene identified in H. pylori, attempts to generate deletion mutants proved unsuccessful (39).
In our study, using the H. pylori G27 strain, we attempted to generate deletion mutants for all four predicted heptose biosynthesis genes but were unable to delete gmhA or hldE. HPG27_814 is annotated as a hypothetical protein in the fully sequenced G27 genome (40). Using the Web-based Phyre2 protein structure prediction tool (41), we found that HPG27_814 showed 35% sequence identity and mapped with 100% confidence as a homologue of E. coli GmhB (data not shown). Deletion of hldD and the putative gmhB gene in G27 resulted in highly filamentous mutants that had a pronounced growth defect. Complementation of hldD and gmhB at the neutral rdxA locus restored the wild-type shape and growth rates ( Fig. 3B and C). Our observations are consistent with a previous report showing that deletion of hldD in strain 26695 resulted in bacteria with truncated LPS (deep-rough phenotype), decreased growth rates, increased susceptibility to novobiocin, and decreased adherence to AGS cells (42). To our knowledge, analysis of gmhB mutants had not been undertaken in H. pylori. However, in our study, deletion of gmhB in strain G27 resulted in an even more pronounced growth defect than deletion of hldD. This suggests that, although the heptose biosynthesis enzymes are part of the same pathway, deletions of individual enzymes have differing effects on H. pylori. Furthermore, another recent study characterized a gmhA deletion mutant in strain 26695 (35). The results showed that, similarly to the hldD mutant, ΔgmhA bacteria displayed truncated LPS, decreased growth rates, and increased susceptibility to novobiocin and detergents and were less able to adhere to AGS cells. Our inability to generate a ΔgmhA mutant may reflect strain-specific differences between G27 and 26695. It is also possible that the ΔgmhA 26695 mutant generated by Yu et al. contains a suppressor mutation that partially alleviates the lethality of deleting gmhA and thus results in viable, albeit sickly, bacteria.
H. pylori HldE drives HBP synthesis and TIFA-dependent NF-B activation. In order to determine whether H. pylori HBP is driving the TIFA-dependent NF-B activation, we used an E. coli overexpression system where we placed H. pylori hldE, which encodes the enzyme that produces HBP, under the control of an IPTG (isopropyl-␤-Dthiogalactopyranoside)-inducible promoter. We collected uninduced and IPTG-induced cell lysates and used them to treat NF-B-luciferase reporter cells. We found that only the E. coli cells that expressed H. pylori HldE potently activated NF-B, a response that was further enhanced by IPTG treatment. In contrast, E. coli expressing H. pylori GmhB, which converts HBP to HMP, showed background levels of NF-B activation that were equivalent to those seen with bacteria expressing vector only in a representative experiment (Fig. 3D) and across three independent experiments (Fig. S3A). To confirm that the enhanced NF-B activation by lysates from H. pylori HldE-expressing E. coli was TIFA dependent, we treated wild-type or TIFA KO reporter cells with a dilution series of lysates from IPTG-induced GmhB-or HldE-expressing bacteria. We standardized our input based on total protein levels measured by a bicinchoninic acid (BCA) protein , or H. pylori cell lysates from strains of the indicated genotype. Data in panels D and E were normalized to signal from cotransfected Renilla luciferase plasmid and are represented as normalized fold changes from mock-treated samples. Each condition was tested in triplicate, with bars showing means with standard deviations, and data are representative of results from at least two independent experiments. Statistical significance was determined using ANOVA with Bonferroni correction for multiple comparisons. ns, not significant (P Ͼ 0.05); *, P Ͻ 0.05; ***, P Ͻ 0.001. assay and found that across the dilution series, lysates from HldE-expressing E. coli consistently showed increased NF-B activation relative to GmhB-expressing bacteria and that the NF-B signal was completely TIFA dependent (Fig. S3B). Since the ADP-heptose biosynthesis enzymes appear to function in H. pylori similarly to the manner in which they function in other Gram-negative bacteria and since HldE, the enzyme that synthesizes HBP, appears essential, we focused our analysis on enzymes downstream of HBP. We hypothesized that targeting gmhB might result in HBP accumulation within the bacteria and, thereby, in increased TIFA-dependent NF-B activation. Using the NF-B-luciferase reporter cells, we treated wild-type or TIFA KO cells with lysates generated from wild-type, ΔgmhB, or ΔhldD (encoding an enzyme downstream of GmhB) bacteria. We normalized our input based on the total protein estimation using a BCA protein assay, since optical density (OD) is not an accurate reflection of total cell number in working with filamentous mutants. We found that lysates from ΔhldD mutants induced statistically significantly more NF-B (P ϭ 0.05) than those from wild-type H. pylori (Fig. 3E). However, assessing data across three independently conducted experiments, we found no significant differences among the levels of NF-B activation induced by the wildtype, ΔgmhB, and ΔhldD strain lysates and the levels induced by their respective complemented strain lysates (Fig. S3C). It is possible that H. pylori controls metabolic pathways to prevent excess cytosolic HBP accumulation in the absence of GmhB. Nevertheless, our overexpression data determined with E. coli suggest that H. pylori HldE generates HBP, which signals through TIFA, leading to NF-B activation.
CagA contributes to the late NF-B-dependent IL-8 response in gastric epithelial cells. Assessing NOD1 TIFA DKO AGS cells for IL-8 production in response to H. pylori infection, we found that IL-8 levels were substantially reduced at the earlier time points. However, in the absence of both NOD1 and TIFA signalling, we found that residual IL-8 accumulated over time, being especially evident at the 24-h time point (Fig. 2A). We hypothesized that CagA may have been driving this residual IL-8 signal since the N-terminal domain (amino acids 24 to 221) of CagA has been proposed to activate NF-B (17,43,44). Indeed, when we cocultured wild-type or TIFA KO AGS cells with the H. pylori ΔcagA mutant, we found a significant decrease in the IL-8 response at the 24-h time point (Fig. 4A; Fig. S4), which was restored to wild-type levels when cells were cocultured with ΔcagA complemented H. pylori (Fig. S4). The reduced IL-8 levels were particularly evident in TIFA KO AGS cells, confirming that both TIFA activation and CagA detection contribute to the IL-8 response in AGS cells. We also showed that CagA is present and becomes phosphorylated within AGS cells cocultured with wild-type or ΔcagA complemented H. pylori (Fig. 4B). Examining HCT116 colon epithelial cells, where the IL-8 response to H. pylori appears to be almost entirely TIFA dependent, we did not observe a significant difference in IL-8 levels in cells treated with wild-type or isogenic ⌬cagA mutant strains (Fig. 4C). We also observed that, while in AGS cells wild-type H. pylori delivered CagA into the host cell, where it was phosphorylated by host kinases, in HCT116 cells CagA was not phosphorylated, suggesting either that CagA is not delivered into these cells or that HCT116 cells are unable to phosphorylate translocated CagA (Fig. 4D).

DISCUSSION
Using CRISPR/Cas9 gene targeting of specific pathogen recognition pathways, we revealed that TIFA is involved in the human cellular response to H. pylori, and we were able to dissect the relative contributions of the H. pylori-derived factors to the NF-Bmediated immune response in gastric epithelial cells. While it was previously accepted that NOD1 activation following H. pylori cell wall fragment (PG) detection is the main proinflammatory signal that leads to IL-8 production, here we show that H. pylori HBP-dependent TIFA activation also plays a critical role in the early immune response to H. pylori (Fig. 5). Furthermore, we reveal a previously unrecognized mechanism by which HBP gains access to the cytosolic TIFA pathway, since H. pylori HBP is delivered directly to the host cell cytosol through the cag-T4SS. It remains an open question whether HBP delivery to the host is an inadvertent consequence of forming the   factors that are delivered through the cag-T4SS into gastric epithelial cells to initiate the NF-B-mediated inflammatory response is shown. The response is initiated by cytosolic delivery of HBP that activates TIFA, which may be phosphorylated at the Thr9 residue by kinase ALPK1 (27). Phosphorylated TIFA forms oligomers that act as a scaffold to recruit and activate TRAF6, which in turn activates transcription factor NF-B, which translocates into the nucleus and upregulates a suite of proinflammatory genes, including the IL-8 gene. Following TIFA activation, peptidoglycan (PG) delivered through the cag-T4SS activates NOD1, which ultimately amplifies NF-B activation. Finally, CagA associates with host TAK1 and enhances its polyubiquitination, which is mediated by the E3 ligase activity of TRAF6. Polyubiquitinated TAK1 then activates NF-B, further amplifying the proinflammatory cascade (43,49). The HBP that is delivered through the cag-T4SS to activate TIFA is generated through the ADP-heptose biosynthesis pathway in the bacterial cytosol. Four H. pylori enzymes are involved in converting sedoheptulose-7-P into ADP-L,D-heptose. Two of these enzymes (marked with an asterisk), GmhA and HldE, appear to be essential in the G27 H. pylori strain. ADP-L,D-heptose and ADP-D,D-heptose are incorporated into the highly conserved core structure of H. pylori LPS, which consists of a hexa-saccharide cag-T4SS that connects the bacterial and host cytosol or whether H. pylori HBP delivery to the host instigates an inflammatory response that benefits the bacteria, such as by the recruitment of regulatory T cells and activation of other tolerogenic mechanisms that prevent bacterial clearance and support chronic colonization (14,45).
It was recently demonstrated on the basis of the rates of cytosolic replication by invasive bacteria that TIFA activation fine-tunes the downstream inflammatory response (28). That study suggested that TIFA signaling is one mechanism by which intestinal epithelial cells can potentially distinguish commensal bacteria from pathogenic invaders. Interestingly, the temporal pattern of NOD1 and TIFA activation in the context of invasive Shigella infection is the inverse of what we observed during H. pylori infection: the NOD1 response constitutes the initial response to Shigella, and TIFA is activated later (28), while the HBP response occurs earlier with H. pylori. This difference may be attributable to the very different lifestyles of these bacteria. While HBP slowly accumulates within the mammalian cell as Shigella grows, a bolus of the cytosolic HBP present within H. pylori can presumably be rapidly delivered as the cag-T4SS penetrates the host cell, rapidly activating the TIFA pathway.
The impact of the CagA toxin on H. pylori-induced inflammation is somewhat controversial. Because ΔcagA H. pylori strains induce a robust inflammatory response, a number of studies concluded that CagA is dispensable for the cag-T4SS-dependent NF-B-driven immune response (46)(47)(48). However, a consensus has recently emerged supporting the idea that CagA itself can activate the NF-B inflammatory pathway since transfection with CagA expression constructs leads to NF-B activation (17) and that CagA physically binds the immune signaling component TAK1 to promote NF-B activation (43,49). Our data support the latter studies, showing that CagA can drive an NF-B-mediated immune response. However, we found that CagA likely activates this pathway later than TIFA and NOD1. Because the TIFA and NOD1 pathways converge on TRAF6 activation, it is possible that initial activation of these pathways potentiates CagA interaction with TAK1 (Fig. 5) and that CagA then further amplifies and/or sustains NF-B activation. Further work is needed to fully elucidate the mechanism by which CagA activates NF-B.
Several groups have recently provided insight into the LPS structure of various commonly studied H. pylori strains (37,(50)(51)(52)(53). Li et al. proposed a redefinition of the H. pylori LPS structure to contain lipid A, a short, highly conserved core hexa-saccharide domain composed of glucose-galactose-D,D-heptose-L,D-heptose-L,D-heptose-2-KDO (keto-3-deoxyoctulosonic acid) and the direct attachment of the variable O-antigen to the D,D-heptose molecule (37). This is in contrast to the LPS structures of other Gramnegative bacteria, which contain an inner core resembling the hexa-saccharide domain of H. pylori and an outer core to which the O-antigen is attached. Studies have shown that deletion of enzymes involved in the assembly of the highly conserved LPS core in H. pylori results in bacteria that are unable to colonize mice (14,37,38,45). Our own data suggest that hldE and gmhA, at least in the G27 strain background, appear to be essential in H. pylori and that deletion of hldD and gmhB results in highly filamentous and slow-growing bacteria. The enzymes involved in the biosynthesis of ADP-heptose and assembly of the core LPS oligosaccharide present bacterial pathways that can potentially be therapeutically targeted.
Our data suggest that TIFA, NOD1, and CagA all contribute to the NF-B-driven inflammatory response in gastric epithelial cells; however, how much each of these pathways contributes to natural infection remains an open question. Although AGS cells represent a useful and tractable model system to study H. pylori interaction with gastric epithelial cells, they do not fully recapitulate the complexity of primary gastric epithelial cells. Interestingly, gene expression analysis of primary human stomach tissue validates the idea that TIFA and NOD1 are both expressed, supporting their physiological relevance to H. pylori infection observed in our study. Furthermore, primary stomach tissue has much higher expression levels of TIFA than of NOD1, which may reflect the relative contributions of the two pathways to H. pylori detection (28). Future studies will focus on assessing NOD1 and TIFA innate immune signaling pathways in primary gastric organoids and animal models of infection to further elucidate their respective contributions to H. pylori detection.
How H. pylori is able to maintain a chronic infection despite a robust host immune response has remained a mystery. To address that question, we must first understand the underlying host inflammatory pathways that H. pylori activates and subverts. Our study revealed a previously unappreciated role of TIFA in the innate immune response initiated in H. pylori-infected gastric epithelial cells. Furthermore, we demonstrate, for the first time, that HBP can be directly delivered to host cells through a bacterial secretion system where it potently activates TIFA and drives the downstream NF-Bmediated inflammatory response. Gaining a comprehensive view of the early signaling events that take place at the interface of H. pylori and gastric epithelial cells is ultimately required to determine how this host-pathogen interaction can be manipulated in the host's favor. Our report adds a critical piece to our understanding of the host immune response to an important human pathogen.

MATERIALS AND METHODS
CRISPR/Cas9 gene targeting and complementation. CRISPR/Cas9 gene targeting of AGS cells was performed as described by Gray et al. (54). Briefly, we used a single lentiviral plasmid that expresses the gene targeting guide RNA under the control of a U6 promoter, as well as an MND promoter-driven Cas9-T2A-puromycin or -blasticidin resistance cassette (pRRL lentiviral vector). 293T cells were transfected with 1.5 g pVSV-G, 3 g psPAX-2, and 6 g pRRL lentiviral vector. Lentiviral supernatants were collected 48 h later and used to transduce AGS cells. AGS cells were then selected with puromycin (0.5 g/ml) or blasticidin (12.5 g/ml) for 7 to 10 days. As a control, cells were transduced with pRRL expressing Cas9 only or a human nontargeting control guide RNA, as described by Sanjana et al. (55). NOD1-targeted cells were clonally selected, and the gene targeting was verified by TOPO cloning and Sanger sequencing. TIFA targeting was verified using an in vitro RNA-guided engineered nucleaserestriction fragment length polymorphism (RGEN-RFLP) assay, as previously described (56). The following guide RNA sequences were used: for NOD1, 5=-GTTTTCAGTAATTGAATGTGG-3=; for TIFA, 5=-CAGATGAC GGTTTACCATCC-3=. To restore TIFA in targeted cells, the TIFA coding DNA sequence (CDS) was cloned into the pRRL lentiviral backbone to generate an MND promoter-driven TIFA-T2A-zeocin plasmid. TIFA-targeted cells were subsequently transduced with pRRL-MND-TIFA-T2A-zeocin plasmid, as described above, and selected with 250 g/ml zeocin for 10 to 14 days. To ensure that the complementation construct was CRISPR resistant, we introduced single nucleotide polymorphisms (SNPs) into the guide RNA-targeted genomic sequence that resulted in synonymous mutations, which maintain the TIFA amino acid sequence but prevent guide RNA binding and Cas9 cutting. The following single oligonucleotide was used in a site-directed mutagenesis reaction to alter the TIFA CDS: 5=-GAGACAGTAACTTGTCTCCAa ATGACaGTcTAtCAcCCaGGCCAGTTGCAGTGTGGAATA-3= (the guide RNA sequence is underlined, and the introduced SNPs are in lowercase). In addition, C-terminal 6ϫHis and V5 tags and N-terminal FLAG tag were added, as indicated, to the TIFA CDS to allow downstream protein expression analysis.
Bacterial strains. G27, a clinical isolate of H. pylori that contains the cag pathogenicity island (57), or isogenic mutants were grown at 37°C in a trigas incubator equilibrated with 10% oxygen, 10% carbon dioxide, and 80% nitrogen on solid media containing horse blood agar or in liquid culture containing brucella broth (BD Biosciences) with 10% heat-inactivated fetal bovine serum (FBS) (Gemini-Benchmark), as previously described (58). Isogenic knockout mutants were constructed using a vector-free allelic replacement strategy where Ͼ60% of the CDS was replaced by a chloramphenicol resistance cassette and selected with 15 g/ml chloramphenicol, as previously described (59). To generate complementation mutants, we integrated a wild-type copy of the deleted gene at a neutral locus: either rdxA with metronidazole selection (60) or an intragenic region between HPG27_186 and HPG27_187 (McGee locus) with kanamycin selection (61). The primers used for generating H. pylori mutants are listed in Table S1 in the supplemental material. E. coli DH5␣, HST08, or BL21 cells were grown and transformed according to standard methods (62). For induced protein overexpression experiments, H. pylori gmhB and hldE genes were amplified and cloned into pET15b vector (Novagen). E. coli BL21 cells were transformed, selected with 100 g/ml ampicillin, and grown to an optical density at 600 (OD 600 ) of 0.6. Cells were then induced with 1 mM IPTG for 4 h and harvested by centrifugation. Cell pellets were then processed to obtain cell lysates as described above for NF-B-luciferase assays or were resuspended in Laemmli buffer with ␤-mercaptoethanol for protein expression analysis by Western blotting.
Cell treatments and analysis. AGS cells, from a human gastric adenocarcinoma cell line (ATCC CRL-1739), were grown in Dulbecco's modified Eagle's medium (DMEM) (Gibco) supplemented with 10% heat-inactivated FBS (Gemini-Benchmark). HCT116 cells, from a human colon adenocarcinoma cell line (ATCC CCL-247), were grown in McCoy's 5A media (ATCC) supplemented with 10% FBS. TIFA KO HCT116 cells were generated using CRISPR/Cas9 gene targeting, as previously described (28). For coculture with H. pylori, AGS or HCT116 cells were seeded at 1 ϫ 10 5 cells/well in 24-well plates 16 h prior to infection. The day of infection, medium was removed from human cells and mid-log-phase H. pylori was added at multiplicity of infection of 10:1 resuspended in either DMEM-10% FBS-20% brucella broth for the AGS cells or McCoy's 5A-10% FBS-20% brucella broth for the HCT116 cells. Supernatants from three individual wells per experimental condition were collected at the indicated time points and assayed for the IL-8 concentration using a human IL-8 enzyme-linked immunosorbent assay (ELISA) kit according to the instructions of the manufacturer (BioLegend). As indicated, cells were also treated with C12-iE-DAP (InvivoGen), IL-1␤ (BioLegend), or TNF-␣ (PeproTech). NF-B-luciferase assays were performed as previously described (25) using a Dual-Glo luciferase assay system (Promega). Wild-type or TIFA KO human embryonic kidney (HEK) 293T cells were plated in triplicate and cotransfected with 90 ng NF-Bluciferase and 10 ng Renilla luciferase plasmids and then treated the following day with Neisseria lysates containing HBP, HMP, or cell lysates from H. pylori or isogenic mutants in 5 g/ml digitonin-containing permeabilization buffer (63). To obtain cell lysates, a volume of mid-log-phase H. pylori culture equivalent to an OD 600 of 1 was pelleted by centrifugation, washed once with phosphate-buffered saline (PBS), resuspended in 100 l of water, boiled for 10 min, and centrifuged to remove cell debris and the supernatant was passed through a 0.22-m-pore-size syringe filter. Input for luciferase assays was normalized using a BCA protein assay kit (Pierce), and 1 l of normalized lysate was used to stimulate 293T reporter cells.
Immunoblotting. After 24 h of coculture with H. pylori, AGS cells were washed with ice-cold PBS and lysed in Laemmli buffer with ␤-mercaptoethanol and boiled for 5 min. For CagA translocation and phosphorylation analysis, cell lysates were separated on Mini-Protean TGX polyacrylamide gels (Bio-Rad) and blotted for anti-phosphotyrosine (4G10; EMD Millipore) and beta-actin (Cell Signaling Technology, Inc.) overnight at 4°C. Membranes were then blocked for several hours in 5% nonfat milk-Tris-buffered saline with Tween 20 (TBST) and then incubated with anti-CagA antibody (Santa Cruz; sc-28368) overnight at 4°C. Statistical analysis. The statistical significance of differences between groups was assessed using one-way analysis of variance (ANOVA) with Bonferroni correction for multiple comparisons. Statistical differences were assessed within individual experiments based on mean values from 3 independent samples per experimental condition, as well as across at least two independently performed experiments. P values greater than or equal to 0.05 were considered statistically significant. All analyses were performed using Prism v7.0 software (GraphPad).