ADP heptose, a novel pathogen-associated molecular pattern associated with Helicobacter pylori type 4 secretion

The gastric pathogen Helicobacter pylori activates the NF-κB pathway in human epithelial cells via the α-kinase 1 (Alpk1)–TIFA axis. We and others have previously shown that heptose 1,7-bisphosphate (HBP) acts as a pathogen-associated molecular pattern (PAMP). HBP is an intermediate of lipopolysaccharide (LPS) synthesis in H. pylori and other gram-negative bacteria. Deletion of the hldE (rfaE) gene encoding the enzyme responsible for HBP synthesis, as well as deletion of further upstream genes, causes loss of NF-κB stimulation by H. pylori, while deletion of the downstream phosphatase encoding gene gmhB does not. This has led to the conclusion that HBP is the PAMP responsible for NF-κB induction. Here, our attempts to identify HBP in lysates of H. pylori revealed surprisingly low amounts that fail to explain NF-κB activation. Instead, we identified ADP heptose, a major downstream metabolite of HdlE, as the predominant PAMP in H. pylori lysates, exhibiting ∼100-fold stronger activity compared to HBP. It therefore appears that synthesis of ADP heptose from HBP in H. pylori occurs independently of GmhB. The data lead us to conclude that ADP heptose constitutes the key PAMP, secreted via the pathogen’s cagPAI encoded type 4 secretion (T4SS).


Introduction
The innate immune system provides mammalian cells with various means of self-defense against microbial infections. A key element in detecting possible threats is recognition of pathogen-associated molecular patterns (PAMPs) by host cells. These highly conserved structures are found in bacteria, viruses and fungi and can be detected via pattern recognition receptors (PRRs) (Medzhitov, 2007).
While a variety of PAMPs have been identified in the last decades, such as lipopolysaccharide (LPS), CpG nucleotides, or flagellin the picture is still incomplete (Baccala et al., 2009). Recognition of these structures alerts the host cell to the presence of potentially harmful microorganisms and triggers innate immune signaling pathways that induce production of inflammatory mediators and attract immune cells to fight an infection.
One of the main pathways activated in response to infection with pathogenic bacteria is nuclear factor κB (NF-κB) signaling, which has been suggested to play a role in inflammation and carcinogenesis (DiDonato et al., 2012). It comprises three pathways: the canonical, non-canonical and p105 pathway -all of which are important for the response to infections and the induction of inflammatory mediators but may also be relevant for the modulation of the inflammatory response (Banerjee et al., 2014;O'Reilly et al., 2018). Canonical NF-κB signaling can be activated by a variety of effectors, such as LPS or tumor necrosis factor α (TNF-α) (Newton and Dixit, 2012). Activation of the upstream receptors leads to phosphorylation of inhibitor of NF-κB kinase β (IKKβ) and subsequent phosphorylation of inhibitor of NF-κB α (IκBα). This induces polyubiquitination and proteasomal degradation of IκBα, which releases the p65/p50 heterodimer normally sequestered in the cytoplasm. Upon release, p65/p50 is translocated into the nucleus where it activates the expression of multiple genes involved in inflammation and proliferation (Jost and Ruland, 2007).
LPS is a well-studied PAMP found in most gram-negative bacteria. It has a tripartite structure consisting of lipid A, an oligosaccharide core comprising an inner and an outer core region, and the O antigen.
Lipid A is the hydrophobic component of LPS localized in the outer bacterial membrane while core polysaccharides and O antigen polysaccharide are presented on the bacterial surface (Wang and Quinn, 2010). The inner core typically consists of 3-deoxy-D-manno-octulosonic acid and L,D-heptose units (Kneidinger et al., 2002).
In 2015, the intermediate HBP was identified as a potent effector of NF-κB activation in epithelial cells upon infection with Neisseria spp. (Gaudet et al., 2015). More recently, we showed that the proteins α-kinase 1 (ALPK1) and TRAF-interacting protein with forkhead associated domain (TIFA) are key components in the response pathway to HBP and in activating canonical NF-κB signaling in gastric epithelial cells infected with H. pylori (Zimmermann et al., 2017). Moreover, activation of NF-κB was found to depend on cytosolic detection of HBP. In H. pylori infection, the type IV secretion system (T4SS) is needed for translocation of HBP into host cells (Zimmermann et al., 2017). In Neisseria infection, cytosolic delivery via phagocytosis followed by lysosomal degradation as well as endocytosis of HBP has been proposed (Gaudet et al., 2015).
Here we biochemically analyzed the amounts of HBP in lysates of H. pylori. Surprisingly, we observed only minor amounts of HBP, which hardly explained the strong inflammatory response of host epithelial cells. We thus hypothesized that yet another compound involved in LPS synthesis might account for the innate immune response observed in H. pylori infection. Further analysis revealed abundant ADP-heptose in the lysate. Here, we demonstrate that ADP-heptose is a novel, potent NF-κB-activating PAMP that is recognized by the ALPK1-TIFA signaling axis. Since ADP-heptose is present in all gram-negative bacteria, this molecule might be of major relevance in other gram-negative infections.

Identification of ADP heptose as novel NF-κB activator
In order to confirm the presence of HBP in H. pylori lysates, we tried to separate the heptose by solid phase extraction (SPE) methods based on the protocol described by Gaudet et al. (2015). Detection of HBP was performed by ion-pairing reversed phase chromatography via electrospray ionization mass spectrometry (ESI-MS). However, in SPE-extracts derived from at least 1 ml of H. pylori lysate we could not detect HBP as judged by the appearance of a mass with m/z 369 Da [M-H] -(data not shown). Since the detection limit of HBP was relatively high (≥100 pM), we aimed to increase the sensitivity by derivatizing the heptose by reductive amination using 3-amino-9-ethylcarbazole (AEC) after treatment with either alkaline phosphatase (CIP), which cleaves both the 7'-O-phosphate and the glycosidic phosphate ester, or mild acid, which hydrolyzes only the glycosidic one (Han et al., 2013). This creates an aldehyde group enabling Schiff's base generation between carbohydrates and primary amines, a reaction which does not occur if the sugar is esterified to phosphate by an O-glycosidic bond. We then determined the sensitivity of our assay using chemically synthesized β-HBP as a reference.
Derivatization of β-HBP allowed us to detect the AEC-sugar derivates by LC-MSMS (MALDI) 1A) but it also revealed that HBP was present in quantities too low to account for the stimulation of NF-κB. In our experiments with synthetic HBP, a concentration of at least 10 µM was required (Zimmermann et al., 2017). By contrast, in extracts HBP was only detected at concentrations in the fM range. To our surprise, in the lysates we detected a mass of m/z 405 Da (fragment ions of m/z 210, 223, 315) after AEC-derivatization, which appeared after acid treatment (Fig. 1B), indicative of a heptose with all phosphate residues being acid labile.
Improved chromatographic resolution of the H. pylori carbohydrate compounds from the SPE-extract was achieved by ion-pairing reversed phase-UPLC. We tested the fractions for their ability to stimulate NF-κB in luciferase reporter cells and for the appearance of AEC-derivatized heptoses by MALDI mass spectrometry. By transfecting the fractions into reporter cells we found that only a single faction in any given run had a strong capacity for inducing NF-κB ( Fig. 2 A). For these NF-κB stimulating fractions, the most prominent MSMS-peak intensities (m/z 223 Da) of the AEC-precursor masses was quantified

De novo synthesis of ADP heptose
To firmly establish the identity of the 618-peak as the NF-κB-stimulating factor, we decided to synthesize β-ADP heptose both chemically and enzymatically. For chemical synthesis, we applied two strategies. We coupled β-H1P either directly to electrophilic AMP morpholidate (Wagner et al., 2009) or indirectly to activated AMP by in situ generation of 2-imidazolyl-1,3-dimethylimidazolinium chloride, which converts the 5'phosphate ester of AMP to the reactive phosphorimidazolide (Tanaka et al., 2012). Both methods worked well, yielding moderate amounts of the expected β-ADP-heptoses (L-and D-Isomers), which eluted at the same time from reversed-phase UPLC as compared to the active compound from H. pylori extracts.
The m/z 618 [M-H]precursor ion was further analyzed by LC-MSMS (MALDI-TOF) using chemically synthesized α-ADP-heptose as reference (Zamyatina et al., 2003). The same fragment ions could be MALDI-MSMS of the UPLC fractionated SPE-extract from H. pylori led to weak intensities of product ions (Fig 3A, 4). Alkaline phosphatase treatment of these ADP-heptoses, which were generated from the unprotected sugar phosphate, led to small amounts (1-5%) of a product with a mass of m/z 538 resulting from the loss of the heptose-phosphate group (data not shown). We presume that coupling of activated AMP to the unprotected sugar led to small amounts of an adenylylated heptose 1phosphate product, which became esterified at the 5'-phosphate of AMP at one of the hydroxyl groups of the heptose.

Pro-inflammatory activity of ADP heptose
We hypothesized that β-ADP-heptose might share the same signaling pathway as HBP, activating NF-κB via the ALPK1-TIFA axis. We therefore transfected AGS wild type as well as AGS TIFA -/and ALPK1 -/knockout cells with β-ADP-heptose and measured the induction of IL-8. We observed that β-ADPheptose-triggered NF-κB activation was completely abrogated in ALPK1 and TIFA knockout cells ( Figure   4A). To further prove that β-ADP-heptose activates the ALPK1-TIFA axis, we incubated a stably tdTomato-TIFA overexpressing cell line with β-ADP-heptose. Induction of TIFAsomes could be observed in H. pylori wt infected cells (data not shown), cells transfected with β-ADP-heptose, and cells transfected with HBP ( Figure 4B, left panel). We therefore concluded that β-ADP-heptose does indeed activate the ALPK1-triggered TIFA phosphorylation that leads to NF-κB activation.
For HBP it has been shown that delivery to the host cell cytoplasm is necessary for recognition by the host cell and subsequent NF-κB activation (Gall et al., 2017;Gaudet et al., 2015;Zimmermann et al., 2017). To our surprise, when we incubated an AGS reporter cell line with ADP-heptose in the absence of transfection reagent we observed a robust NF-κB response, whereas HBP only stimulated the cells when delivered into the cytoplasm by a transfection reagent ( Figure 4C and D). This observation was confirmed in TIFA reporter cells incubated without transfection ( Figure 4B right panel). Additionally, we observed that β-ADP-heptose activates NF-κB at concentrations as low as 1-10 pmol -around 100fold lower than the concentration required for HBP ( Fig. 4C and D). We conclude that β-ADP-heptose rather than HBP is the main pro-inflammatory PAMP in H. pylori.

Discussion
Identifying immune stimulatory pathways is crucial for understanding how cells can sense infection and initiate an immune response. Recently, we and other groups described the ALPK1-TIFA axis stimulated by recognition of HBP as a novel pathway for the induction of an innate immune response in epithelial cells (Gall et al., 2017;Stein et al., 2017;Zimmermann et al., 2017). Yet, when analyzing H. pylori extracts we only detected minor amounts of HBP that appear to be insufficient for eliciting the observed NF-κB activation. Thus, we searched for additional bacterial triggers of NF-κB induction. By ion-pairing reversed phase-UPLC analysis of H. pylori extracts, we identified a fraction that is highly efficient in inducing an NF-κB response. Further analysis of this fraction by LC-MSMS (MALDI) revealed a prominent compound with a mass of m/z 618 [M-H]that was identified as β-ADP-heptose. The capacity of this compound to stimulate NF-κB was confirmed with chemically and enzymatically synthesized β-ADP-heptoses (D-and L-glycero isomers), both of which led to a strong NF-κB induction.
Interestingly, this activation occurred even without permeabilization of the cell membranes. Activation was dependent on the presence of ALPK1 and TIFA. Hence, we show that β-ADP-heptose is present in H. pylori and acts as a novel and potent stimulator of ALPK1-TIFA controlled innate immune responses in epithelial cells.
β-ADP-heptose is a metabolite of gram-negative and also some gram-positive bacteria in the synthesis of the inner core-region of LPS (Kneidinger et al., 2002). In H. pylori, the rfaE gene encodes a bifunctional enzyme (HldE) that harbors the kinase and the ADP transferase activity responsible for generation of D-glycero-β-D-manno-heptose 1-P and ADP-D-glycero-β-D-manno-heptose (Stein et al., 2017). Previous studies have shown that deletion of gmhA and rfaE abolishes the induction of NF-κB (Gall et al., 2017;Stein et al., 2017;Zimmermann et al., 2017). Yet, a mutant in GmhB, the phosphatase converting HBP into heptose 1-phosphate, a precursor for β-ADP-heptose synthesis, still showed NF-κB activation (Gall et al., 2017;Stein et al., 2017). It is possible that this mutant is still capable of ADPheptose synthesis via an alternative pathway, explaining the preserved NF-κB activation. Such a compensatory mechanism is known from E. coli, where the gmhB gene function can at least partially be compensated by other phosphatases (Kneidinger et al., 2002).
Former studies, including ours, have focused on HBP as the activating molecule of the ALPK1-TIFA axis, showing that permeabilization of the membrane either by lipofection or by electroporation is necessary for a robust HBP-triggered NF-κB response (Gall et al., 2017;Gaudet et al., 2015;Stein et al., 2017;Zimmermann et al., 2017). By contrast, β-ADP-heptose also activates NF-κB in epithelial cells when added extracellularly. In infection experiments, however, either a functioning type IV secretion system (Stein et al., 2017;Zimmermann et al., 2017) or cellular uptake of bacteria (Milivojevic et al., 2017) is needed to induce an immune response. This suggests that β-ADP-heptose is not normally released by bacteria but needs to be actively transported into the cell. The exact mechanism of uptake or translocation of ADP-heptose to host cells remains speculative.
Compared to HBP, β-ADP-heptose showed NF-κB activation at 100-fold lower concentrations, making this the more likely direct simulator of the ALPK1-TIFA axis. Whether HBP stimulates the same receptor only with lower affinity, or if it is converted before receptor binding, needs to be determined. Very recently, ALPK1 was identified as the direct receptor for β-ADP-heptose and shown to directly phosphorylate TIFA upon binding (Zhou et al., 2018). The same study also suggested an intracellular conversion of HBP to heptose 7-phosphate, which is also able to activate ALPK1 (Zhou et al., 2018). In support of this theory, we also observed activation in cells treated with ADP-heptose 7-phosphate (data not shown).
Here we demonstrate the presence of β-ADP-heptose in lysates of H. pylori and show that this compound has a substantially higher potency than HBP for inducing an NF-κB response via the ALPK1-TIFA axis. The relevance of this pathway in vivo, the potential usefulness of β-ADP-heptose as an adjuvant or its importance in other clinical settings are questions awaiting urgent further analysis.

Bacterial Culture and Lysis Preparation
H. pylori WT strain P12 was grown on horse serum agar plates supplemented with vancomycin (10 µg/ml) and cultivated for two passages at 37 °C and 5% CO2 (Backert et al., 2000). For lysate preparation, bacteria were harvested by resuspension in water and diluted to OD550 1. Cells were lysed by heating to 95 °C for 15 min. Lysates were centrifuged at 4000 x g for 3 min and supernatants filtered through a 0.22 µm syringe filter.

Carbohydrate Delivery and Luciferase Assay
Stably tdTomato-TIFA overexpressing cells were starved in serum-free medium for 2 h at 37 °C and 5% CO2 prior to sugar delivery. Delivery of sugar into the cytoplasm was achieved by treating cells at 70% confluence with a preparation of 1 µl sugar solution mixed with 7.5 µl OptiMEM (GIBCO), 6.25 µl ATP (20 mM; Roche) and 1.25 µl lipofectamine 2000 (Thermo Fisher Scientific) after 20 min pre-incubation at room temperature (RT). Alternatively, sugars were added to the starvation medium without transfection reagent. Cells were incubated for 3 h at 37 °C and 5% CO2 and the subsequent luciferase assay performed according to the manufacturer's instructions (Promega).
Real Time-qPCR RT-qPCR was performed using the Power SYBR Green RNA-to-CT 1-Step Kit according to the manufacturer's instructions (Applied Biosystems). As described previously (Zimmermann et al., 2017), 1 ng/µl of the isolated RNA was added per reaction, the final concentration of the primers was 166 nM. RT-qPCR was performed using a StepOnePlus Real-Time PCR system (Applied Biosystems) with a one-step protocol for both RT reaction and PCR reaction. The following conditions were used during the protocol: an initial cycle for the generation of cDNA at 48 °C for 30 min; second cycle for the activation of the hot-start Taq polymerase at 95 °C for 10 min; and 40 consecutive denaturation cycles at 95 °C for 15 s, each followed by primer annealing at 60 °C for 1 min. Baseline and Cq values were determined automatically by the StepOne Software v2.3 (Applied Biosystems). Fold-changes were calculated using the 2 −ΔΔCT method (Livak and Schmittgen, 2001). The following primers were used: IL-8 5´-ACACTGCGCCAACACAGAAAT-3´ and 5´-ATTGCATCTGGCAACCCTACA-3´, GAPDH 5´-GGTATCGTGGAAGGACTCATGAC-3´ and 5´-ATGCCAGTGAGCTTCCCGTTCAG-3´.
Enzymatic Synthesis of β-ADP-heptose from β-H1P or H7P: Up to 10 nmol D-β-H1P (triethyl ammonium salt) were incubated at 35 °C with 1 µg recombinant HldE from H.pylori in 20 µl HldE-Buffer (50 mM Tris HCl, pH 8.0, 150 mM NaCl, 5 mM MgCl2, 5 mM ATP) for 1 h. The reaction was stopped by addition of 50 µl chloroform/methanol (2:1). The aqueous phase was dried in a SpeedVac evaporator and dissolved in 50 µl water. The product was analyzed by LC-MS as described below. When H7P was used instead of H1P, the above reaction was performed in the presence of 1 µg of recombinant GmhB from H. pylori. Da. ADP-heptose measurements were acquired in negative mode and the MS mass range was set to 250-1000 Da. All data were evaluated manually.
The QDa was operated in an electrospray negative ion mode by applying a voltage of 0.8 kV. The cone voltage was set at 15 V. The probe temperature was set at 600 °C. A full mass spectrum between m/z 100 and 1200 was acquired at a sampling rate of 8.0 points/sec. Isolation of ADP-heptose from H. pylori H. pylori strain P12 was grown on horse serum agar plates supplemented with vancomycin (10 µg/ml) and cultivated for two passages at 37 °C and 5% CO2 (Backert et al., 2000). For lysate preparation, bacteria were harvested by resuspension in water and diluted to OD550 1. Cells were lysed by heating to 95 °C for 15 min. Lysates were centrifuged at 4000 x g for 3 min and supernatants filtered through a 0.22 µm syringe filter.
One ml of the cleared lysate was extracted by the addition of 2 ml chloroform/methanol (2:1). After vortexing and centrifugation (5 min, 10,000 x g), the aqueous phase was passed over a HyperSep SPE-Aminopropyl cartridge (200 mg/3 ml) which was equilibrated with 50 mM acetic acid in 50% methanol.
Bound compounds were eluted with 500 mM TEAB, pH 8.5 in 50% methanol. Eluates were dried in a SpeedVac evaporator, solubilized in 1 ml 10 mM ammonium bicarbonate, pH 8.0 and passed over a graphite carbon column (Supelclean ENVI-Carb, 1 ml). Sugars were eluted with 30% acetonitrile. The eluate was dried in a SpeedVac evaporator, dissolved in 10 µl 10 mM TEAB-buffer and loaded onto a HSS-T3 reversed phase UPLC column (3x150 mm, 1.8 µm) equipped with a HSS T3 VanGuard Precolumn. Compounds were separated at a flow rate of 0.4 ml/min at 45 °C with a linear gradient from 2% acetonitrile to 80% acetonitrile containing 10 mM TEAB as ion-pairing reagent. Fractions were analyzed by UV (PDA-Detector).