Targeted Identification of Glycosylated Proteins in the Gastric Pathogen Helicobacter pylori (Hp)

Virulence of the gastric pathogen Helicobacter pylori (Hp) is directly linked to the pathogen's ability to glycosylate proteins; for example, Hp flagellin proteins are heavily glycosylated with the unusual nine-carbon sugar pseudaminic acid, and this modification is absolutely essential for Hp to synthesize functional flagella and colonize the host's stomach. Although Hp's glycans are linked to pathogenesis, Hp's glycome remains poorly understood; only the two flagellin glycoproteins have been firmly characterized in Hp. Evidence from our laboratory suggests that Hp synthesizes a large number of as-yet unidentified glycoproteins. Here we set out to discover Hp's glycoproteins by coupling glycan metabolic labeling with mass spectrometry analysis. An assessment of the subcellular distribution of azide-labeled proteins by Western blot analysis indicated that glycoproteins are present throughout Hp and may therefore serve diverse functions. To identify these species, Hp's azide-labeled glycoproteins were tagged via Staudinger ligation, enriched by tandem affinity chromatography, and analyzed by multidimensional protein identification technology. Direct comparison of enriched azide-labeled glycoproteins with a mock-enriched control by both SDS-PAGE and mass spectrometry-based analyses confirmed the selective enrichment of azide-labeled glycoproteins. We identified 125 candidate glycoproteins with diverse biological functions, including those linked with pathogenesis. Mass spectrometry analyses of enriched azide-labeled glycoproteins before and after cleavage of O-linked glycans revealed the presence of Staudinger ligation-glycan adducts in samples only after beta-elimination, confirming the synthesis of O-linked glycoproteins in Hp. Finally, the secreted colonization factors urease alpha and urease beta were biochemically validated as glycosylated proteins via Western blot analysis as well as by mass spectrometry analysis of cleaved glycan products. These data set the stage for the development of glycosylation-based therapeutic strategies, such as new vaccines based on natively glycosylated Hp proteins, to eradicate Hp infection. Broadly, this report validates metabolic labeling as an effective and efficient approach for the identification of bacterial glycoproteins.

Helicobacter pylori (Hp) 1 infection poses a significant health risk to humans worldwide. The Gram-negative, pathogenic bacterium Hp colonizes the gastric tract of more than 50% of humans (1). Approximately 15% of infected individuals develop duodenal ulcers and 1% of infected individuals develop gastric cancer (2). Current treatment to clear infection requires "triple therapy" (3), a combination of multiple antibiotics that is often associated with negative side effects (4). Because of poor patient compliance and the evolution of antibiotic resistance, existing antibiotics are no longer effective at eradicating Hp infection (4). New treatment methods are needed to eliminate Hp from the human gastric tract.
Recent work has focused on gaining insights into the pathogenesis of Hp to aid the development of new treatments. The most recent findings in this area have conclusively revealed that glycosylation of proteins in Hp is required for pathogenesis. Hp use complex flagella, comprised of flagellin proteins, to navigate the host's gastric mucosa (5,6). The flagellin proteins are heavily glycosylated with the unusual nine-carbon sugar pseudaminic acid, found exclusively in mucosal-associated pathogens (Hp (7), Campylobacter jejuni (8) and Pseudomonas aeruginosa (9)). This modification is absolutely essential for the formation of functional flagella on Hp (7,10). Deletion of any one of the enzymes in the pseudaminic acid biosynthetic pathway results in Hp that lack flagella, are nonmotile, and are unable to colonize the host's stomach (7). Although pseudaminic acid is critical for Hp virulence, it is absent from humans (11,12). Therefore, insights into Hp's pathogenesis have revealed that Hp's glycan pseudaminic acid is a bona fide target of therapeutic intervention. This is one of a number of examples linking protein glycosylation to virulence in medically significant bacterial pathogens (13,14).
Despite these findings, Hp's glycome remains poorly understood overall. Only the two flagellin glycoproteins have been firmly characterized in Hp (7) to date. Nine other candidate glycoproteins have been identified in Hp, but their glycosylation status has not been biochemically confirmed (15). The relative paucity of information regarding Hp's glycoproteins is due in part to the previously held belief that protein glycosylation could not occur in bacteria (13,16,17). However, even after Szymanski (18,19), Koomey (20), Guerry (21), Logan (7), Comstock and others (13,16,17) disproved this belief by firmly establishing the synthesis of glycoproteins in bacteria, the study of bacterial glycoproteins has presented unique challenges for analytical study (14,22). For example, the unusual structures of bacterial glycans, which often contain amino-and deoxy-carbohydrates exclusively found in bacteria (12,(23)(24)(25), hampers their identification using existing tools. Though methods such as the use of glycan-binding reagents (20,24,26,27) and periodic acid/hydrazide glycan labeling (15) have successfully detected glycoproteins in a range of bacteria, they present limitations. Glycan bindingbased methods are often limited because of the unavailability of lectins or antibodies with binding specificity for glycosylated proteins in the bacteria of interest (14,22). Periodic acid/hydrazide-based labeling is plagued by a lack of specificity for glycosylated proteins (15). Thus, an efficient and robust approach to discover Hp's glycoproteins is needed.
In previous work, we established that the chemical technique known as metabolic oligosaccharide engineering (MOE), which was developed by Bertozzi (28,29), Reutter (30), and others for the study of mammalian glycoproteins, is a powerful approach to label and detect Hp's glycoproteins (31). Briefly, Hp metabolically processes the unnatural, azidecontaining sugar peracetylated N-azidoacetylglucosamine (Ac 4 GlcNAz) (32), an analog of the common metabolic precursor N-acetylglucosamine (GlcNAc), into cellular glycoproteins ( Fig. 1). Elaboration of azide-labeled glycoproteins via Staudinger ligation (33) with a phosphine probe conjugated to a FLAG peptide (Phos-FLAG) (34) followed by visualization with an anti-FLAG antibody (Fig. 1) revealed a glycoprotein fingerprint containing a large number of as-yet unidentified Hp glycoproteins that merit further investigation (31).
Here we describe a glycoproteomic identification strategy for the selective detection, isolation, and discovery of Hp's glycoproteins. In particular, we demonstrate that glycan metabolic labeling coupled with mass spectrometry analysis is an efficient and robust chemical approach to identify novel glycoproteins in Hp. This work characterizes glycosylated virulence factors in Hp, thus opening the door to new vaccination and antibiotic therapies to eradicate Hp infection. Broadly, this work validates metabolic oligosaccharide engineering as a complementary method to discover bacterial glycoproteins.
Metabolic Labeling of Hp-Hp strain 26695 was grown on horse blood agar plates for 3-4 days in a microaerophilic environment (14% CO 2 ) at 37°C. The bacteria were then transferred to Brucella broth containing 10% FBS, 6 g/ml vancomycin, and 1 mM peracetylated N-acetylglucosamine (Ac 4 GlcNAc) or Ac 4 GlcNAz, as previously described (31). Alternatively, the bacteria were transferred to Brucella broth containing 10% FBS and 6 g/ml vancomycin supplemented with 5 mM sodium azidoacetate or 5 mM azidoacetyl ethyl ester. Liquid cultures were grown for 3-5 days in a microaerophilic environment with gentle rocking at 37°C, then harvested as described below.
Harvesting Hp to Probe for Total Cellular Azide-Labeled Glycoproteins-Hp grown in the presence of Ac 4 GlcNAc, Ac 4 GlcNAz, sodium azidoacetate, or azidoacetyl ethyl ester were centrifuged at 3500 rpm using a Sorvall Legend RT ϩ centrifuge (Thermo Scientific) and washed two times with PBS. The cells were resuspended in ice cold lysis buffer (20 mM Tris-HCl, pH 7.4; 1% Igepal; 150 mM NaCl; 1 mM EDTA) containing protease inhibitors (protease inhibitor mixture, Sigma) for 5 mins at room temperature. Protein concentration of the samples was measured using BioRad's D C protein concentration assay (BioRad) per manufacturer's instructions. All samples were standardized to 3.0 mg/ml. To probe for azide incorporation, the lysates were diluted 1:1 with 500 M Phos-FLAG (34), reacted overnight at room temperature, and then analyzed via Western blot with anti-FLAG antibody (Sigma) or via SDS-PAGE with zinc stain (BioRad).
Subcellular Fractionation of Glycoproteins-Hp grown in the presence of Ac 4 GlcNAc or Ac 4 GlcNAz were harvested from Brucella broth by centrifugation at 6000 ϫ g, and the conditioned medium was set aside to access secreted proteins. The harvested cells were fractionated as described by Hoshino et al. to obtain periplasmic, inner membrane-associated, and cytoplasmic protein fractions (35). Secreted proteins were isolated from the conditioned Brucella broth via TCA-precipitation as described by Bumann et al. (36). An acid-glycine extraction was performed as described previously to isolate Hp's surface-associated proteins (37). An N-lauroylsarcosine extraction was performed as reported by Hopf et al. to access outer-membrane Hp proteins (15,38). The protein concentration of subcellular fractions was standardized to 3.0 mg/ml using a BioRad D C Protein Assay (BioRad). Standardized samples were incubated 1:1 with 500 M Phos-FLAG (34) overnight at room temperature to label azide-containing glycoproteins, and then analyzed by Western blot analysis with anti-FLAG antibody (Sigma) or via SDS-PAGE with zinc stain (BioRad). In addition, standardized periplasmic and cytoplasmic samples were assayed for malate dehydrogenase (35) activity alongside a positive control (malate dehydrogenase, Sigma) according to the manufacturer's (Sigma's) instructions to confirm the efficacy of subcellular fractionation.
Preparation and Analysis of Enriched Azide-Labeled Glycoproteins-Hp metabolically labeled with 1 mM Ac 4 GlcNAc or Ac 4 GlcNAz were centrifuged at 3000 ϫ g and washed two times with phosphate buffered saline (PBS). The cells were lysed according to Koenigs et al. and centrifuged at 3000 ϫ g for 10 min to remove insoluble debris (31). Following the protocol by Laughlin et al., Phos-FLAG-His 6 was added as a solid directly to the lysate at a final concentration of 500 M, and Staudinger ligation was run at room temperature for 24 h under argon (29,39). The unreacted Phos-FLAG-His 6 was removed by Bio-Rad P-10 size-exclusion column according to the manufacturer's protocol. Flow-through fractions with A 280 Ͼ 0.050 were combined and concentrated using a centrifugal filter device with 30 kDa molecular weight cutoff (Millipore). FLAG-His 6 -tagged glycoproteins were enriched using 2 ml of ␣-FLAG agarose beads (Sigma) followed by 5 ml of nickel-nitrilotriacetic acid (Ni-NTA)-agarose resin (Bio-Rad), as previously described (29,39). After rinsing away nonbinders, the bound glycoproteins were eluted from the nickel column with one column volume of a solution of 8 M urea, 0.1 M Na 2 HPO 4 , 10 mM Tris (pH 8.0), and 250 mM imidazole. Samples containing 15 g of material from Ac 4 GlcNAc and Ac 4 GlcNAz-treated Hp both before purification (input) and after purification (eluent) were analyzed by SDS-PAGE to confirm successful enrichment. In addition, 150 g of Hp's enriched azide-labeled glycoproteins were suspended in 10 mM Tris (pH 8) and focused on 11 cm pH 3-10 nonlinear IPG strips (GE Healthcare) using GE Healthcare's Multiphor ™ II system. These strips were reduced, alkylated and loaded onto 8 -18% polyacrylamide gels (GE Healthcare), which were run at 10°C and stained with zinc stain (BioRad).
Protein Identification by MudPIT-An azide-labeled enriched glycoprotein sample containing 150 g of protein and a mock enriched control (which was labeled with Ac 4 GlcNAc, reacted with Phos-FLAG-His 6 , and enriched via anti-FLAG and Ni-NTA affinity chromatography, as described above) were concentrated using a Microcon centrifugal filter device (Millipore) to ϳ10 l, suspended in 8 M urea and 100 mM Tris-HCl (pH 8.5), reduced with 5 mM tris(2-carboxyethyl)phosphine (TCEP), and reacted with 10 mM iodoacetamide. The samples were diluted by a factor of four with 100 mM Tris-HCl (pH 8.5) and 1 mM CaCl 2 , and then digested with 0.25 g of trypsin (Promega) overnight at 37°C. The digests were mixed with formic acid to a final concentration of 5%, desalted using C 18 Spec tips (Varian), fully dried by speed vacuum at room temperature, and analyzed by multidimensional protein identification technology (mudPIT) (40). The Vincent J. Coates Proteomics/Mass Spectrometry Laboratory at University of California Berkeley performed mass spectrometry analyses of the enriched, azide-labeled sample and mock-enriched control sample. A nano LC column was packed in a 100 m inner diameter glass capillary with an emitter tip. The column consisted of 10 cm of Polaris C18 5 m packing material (Varian), followed by 4 cm of Partisphere 5 SCX (Whatman). The column was loaded by use of a pressure bomb and washed extensively with buffer A (see below). The column was then directly coupled to an electrospray ionization source mounted on a Thermo LTQ XL linear ion trap mass spectrometer. An Agilent 1200 HPLC equipped with a split line so as to deliver a flow rate of 30 nl/min was used for chromatography. Peptides were eluted using an 8-step MudPIT procedure (40). Buffer A was 5% acetonitrile/0.02% heptaflurobutyric acid (HBFA); buffer B was 80% acetonitrile/0.02% HBFA. Buffer C was 250 mM ammonium acetate/5% acetonitrile/0.02% HBFA; buffer D was same as buffer C, but with 500 mM ammonium acetate.
The programs SEQUEST, as embodied in BIOWORKS BROWSER (version 3.3.1 SP1), (41) and DTASELECT (version 1.9) (42) were used to identify peptides and proteins from a database consisting of all proteins (1560 total) from Hp 26695 plus a collection of 160 common contaminants (1720 protein entries in the database actually searched). The Hp database was the newest available from NCBI RefSeq on October 7, 2010. Search parameters were altered to DTA generation 600.00 -4000.00 with an absolute threshold of 500. The search was restricted to tryptic peptides and allowed three missed cleavages. Carboxyamidomethylation of cysteines as a fixed modification and oxidation of methionines as a variable modification were specified. The mass tolerance for precursor ions was 1.4 and for fragment ions was 1.0. Individual peptides were accepted with Xcorr of 1.8 for charge state ϩ1, 2.2 for charge state ϩ2 and 3.5 for charge state ϩ3. These values have given Ͻ 1% false positives over many data sets. Proteins were further required to have at least two peptides meeting acceptance criteria. Data was considered high confidence if assigned spectra were present in the Ac 4 GlcNAz samples but absent from the mock-enriched Ac 4 GlcNAc control samples.
Prediction of Subcellular Localization-Initial predictions of the subcellular localization of Hp proteins identified by mudPIT were based on literature reports, where available. Because of the shortage of subcellular location information for hypothetical proteins and functionally uncharacterized proteins, each identified protein was also subjected to PSORTb v.3.0 to predict subcellular location (43).
Phos-FLAG Labeling to Produce "Tagged Lysates"-Hp metabolically labeled with 1 mM Ac 4 GlcNAc or Ac 4 GlcNAz were centrifuged at 3000 ϫ g and washed two times with phosphate buffered saline (PBS). The cells were lysed according to Koenigs et al. (31). Lysates were rinsed extensively with PBS using a Microcon centrifugal filter device (Millipore, 10000 MWCO) to remove residual Ac 4 GlcNAc or Ac 4 GlcNAz. Lysates were reacted with 250 M Phos-FLAG at room temperature overnight. The unreacted Phos-FLAG was removed by extensive rinsing with ddH 2 O and passage of reacted lysates through a Microcon centrifugal filter device (Millipore, 10000 MWCO) to yield samples of "tagged lysates." These lysates were then subjected to beta-elimination, as described below.
Biochemical Purification of Azide-Labeled Urease-Urease was purified from Hp strain 26695 using ion exchange chromatography (IEC) and size exclusion chromatography (SEC) based on previously described purification schemes by Hu and Mobley (44,45) and Rokita et al. (46). A 250 ml culture of Hp metabolically labeled with 1 mM Ac 4 GlcNAz was inoculated and grown as described above. Cells were harvested by centrifugation at 10,000 ϫ g and the resultant pellet was frozen at Ϫ20°C. The cells were re-suspended in ice-cold buffer (20 mM sodium phosphate, pH 6.9) containing protease inhibitors (protease inhibitor mixture, Sigma) and 10 g/ml DNase and then sonicated to achieve complete cell lysis. Concentrated stock solutions of EDTA and ␤-mercaptoethanol (␤ME) were added to the lysis solution to make it consistent with the initial IEC buffer (20 mM sodium phosphate, pH 6.9, 1 mM EDTA, 1 mM ␤ME). Cellular debris was removed from the sample through centrifugation at 14,000 ϫ g. The supernatant was loaded onto a Q-Sepharose Fast Flow column (25 ml, 3 cm diameter, GE Healthcare) that had been pre-equilibrated with the above IEC buffer. Urease was eluted from the column at ϳ150 mM NaCl using a linear gradient from 0 M to 0.5 M NaCl in IEC buffer. The presence of urease in the elution fractions was detected qualitatively using a phenol-red colorimetric assay (44). Fractions that exhibited urease activity were further analyzed by SDS-PAGE with zinc-stain (BioRad) and via Western with anti-ureA and anti-ureB antibodies (Santa Cruz Biotechnology) to verify the presence of ureA and ureB. Fractions with the most urease and fewest contaminants were pooled for subsequent purification by SEC. The pooled fractions were exchanged into the SEC buffer (20 mM sodium phosphate, pH 6.9, 150 mM NaCl, 1 mM EDTA, 1 mM ␤ME) using a 9000 Da MWCO filter (Pierce) and concentrated to a final volume of 7 ml. This concentrated sample was loaded onto a pre-equilibrated Sepharose CL-6B (GE Healthcare) SEC column (200 ml total volume, 3 cm diameter) and eluted at 1 ml/min. Fractions containing urease were identified using the phenol-red assay and subsequently analyzed by SDS-PAGE. The purest urease-containing fractions (see supplemental Fig. S7) were pooled, concentrated to 0.5 mg/ml via ultrafiltration (Millipore, 100,000 MWCO), and mixed 1:1 with 500 M Phos-FLAG at room temperature overnight. The unreacted Phos-FLAG was removed by extensive rinsing with 50% MeOH in ddH 2 O, 15% acetonitrile in ddH 2 O, and 100% ddH 2 O over a Microcon centrifugal filter device (Millipore, 10000 MWCO) to yield samples of "tagged urease." This sample was then subjected to beta-elimination, as described below.
Beta-Elimination of Hp's O-Linked Glycans-Hp's enriched azidelabeled glycoproteins, Phos-FLAG-reacted Ac and Az lysates ("tagged lysates"), and purified azide-labeled urease ("tagged urease") were subjected to beta-elimination to remove glycans overnight at 4°C using GlycoProfile Beta-elimination Kit (Sigma-Aldrich) according to manufacturer's instructions. Released glycans were then analyzed by mass spectrometry, as described below.

HPLC-Chip/Q-TOF Mass Spectrometric Analysis-Samples of
Hp's enriched azide-labeled glycoproteins, Phos-FLAG-reacted Ac and Az lysates ("tagged lysates"), and purified azide-labeled urease ("tagged urease") were analyzed before and after beta-elimination of O-linked glycans using a 6530 High Performance Liquid Chromatographic-Chip Quadrupole Time-of-Flight Mass Spectrometer (HPLC-Chip/Q-TOFMS; Agilent Technologies, Santa Clara, CA). Chromatographic separation and nano-electrospray ionization (nanoESI) was performed with a 1260 Chip Cube system (Agilent Technologies) using a ProtID-chip with a 40 nL enrichment column and a 150 mm x 75 m analytical column (Agilent Technologies). The enrichment and analytical columns were packed with 300 Å, 5 m particles with C18 stationary phase. The mobile phases were 0.1% formic acid/H 2 O (A) and 0.1% formic acid/acetonitrile (B). Samples (0.2-2 l) were loaded on the enrichment column using 98:2 (A:B) at 4 l/min and eluted through the analytical column using a gradient from 98:2 (A:B) to 80:20 (A:B) over 2 min to 2:98 (A:B) over a period of 8 min at 0.3 l/min. Mass spectra (MS and MS/MS) were collected in positive ion mode; the ionization voltage ranged from 1850 -1950 V and the ion source temperature was held at 350°C. The fragmentor and collision cell voltages, as well as other tuning parameters, were optimized to minimize metastable water losses associated with ion sampling and transmission to the TOF analyzer. This was of particular importance for the analysis of HexNAz-Phos-FLAG-His 6 (1), for which metastable water losses became increasingly prominent for higher charge state ions. A similar effect, impacting metastable losses of sulfate from sulfated oligosaccharides, has been reported by Zaia and co-workers (47). Spectra were internally calibrated using methyl stearate (C 17 H 35 CO 2 CH 3 ) or dibutyl phthalate (C 16 H 22 O 4 ) and hexakis(1H, 1H, 4H-hexafluorobutyloxy)phosphazine (HP-1221; C 24 H 18 O 6 N 3 P 3 F 36 ), continuously introduced and detected as [MϩH] ϩ . Collision-induced dissociation (CID)-MS/MS experiments were executed with precursor selection determined using a targeted or data-dependent approach. Precursor ions were subjected to CID using nitrogen as the target gas with collision energies ranging from 20 -35 V.
Immunopurification of Hp's ureA and ureB-Lysate from metabolically labeled Hp was standardized to 4.4 mg/ml and incubated 1:1 with 500 M Phos-FLAG overnight at room temperature to label azide-containing glycoproteins. Insoluble membrane material was separated from soluble proteins by centrifugation at 12,000 ϫ g for 5 min. To the soluble protein samples, 1 l of ␣-Hp ureA or ureB (Santa Cruz Biotechnology) was added per 100 l of protein volume. The samples were incubated at 4°C overnight with gentle shaking. After completion of the overnight incubation, 100 l of a 50% Protein-G agarose suspension in HNTG washing buffer (20 mM HEPES buffer pH 7.5, containing 150 mM NaCl, 0.1% (w/v) Triton X-100, and 10% (w/v) glycerol) was added to Hp lysates and allowed to incubate for 90 min at 4°C with gentle shaking. The immunoprecipated complexes were collected via centrifugation at 3000 ϫ g for 5 min at 4°C and the pellet was washed three times with ice-cold HNTG washing buffer. The final pellet was re-suspended in 40 l of ice-cold HNTG washing buffer, diluted with an equal volume of 2 ϫ SDS-loading buffer, and heated at 95°C for 5 min. The boiled samples were centrifuged at 12,000 ϫ g for 30 s to separate the agarose beads from the immunopreciptated proteins. Immunoprecipitated proteins were then analyzed by Western blot with anti-FLAG (Sigma) or anti-urease antibody (Santa Cruz Biotechnology).

Metabolic Labeling of Hp's Proteins With Azides Does Not
Appear to be Catabolic-Here we employed metabolic glycan labeling to characterize Hp's glycoproteins and unveil molecules that have the potential to serve as the basis of novel anti-Hp treatments. In previous work, our laboratory demonstrated that supplementation of Hp's media with 1 mM Ac 4 GlcNAz for three to 5 days leads to incorporation of azidedependent signal into a large number of glycoproteins (31). The majority, though not all, of this azide-dependent signal can be removed by glycosidases that catalyze the cleavage of certain N-linked and O-linked glycans (31). Thus, we sought to assess whether the glycosidase-resistant azide signal is because of an alternative metabolic fate of Ac 4 GlcNAz in Hp; specifically, catabolism of the azidosugar to the azidoacetyl unit followed by subsequent activation and covalent addition to proteins. This type of azidosugar catabolism has been observed in some other organisms. If this process were to occur in Hp, then supplementation of Hp's media with the azidoacetyl moiety would likely lead to metabolic labeling of Hp's proteins with azides. To assess this possibility, Hp cells were grown in media supplemented with 5 mM sodium azidoacetate or with the more bioavailable azidoacetyl ethyl ester ( Fig. 2A); Hang and coworkers demonstrated that structurally analogous compounds are cell permeable (48,49). Ac 4 GlcNAz or the azide-free sugar peracetylated GlcNAc (Ac 4 GlcNAc) were employed as positive and negative controls of metabolic labeling, respectively. Ac 4 GlcNAz treatment resulted in robust azide-dependent signal in lysates, whereas no azide-labeled proteins were observed in lysates from cells supplemented with Ac 4 GlcNAc, with sodium azidoacetate (Fig. 2B), or with azidoacetyl ethyl ester (Fig. 2C). SDS-PAGE analysis confirmed the presence of equivalent protein levels in all lanes (supplemental Fig. S1). These data suggest that azide-dependent labeling of Hp's proteins is not because of catabolic incorporation of the azidoacetyl moiety into proteins. The remaining azide-dependent signal could be glycosidase resistant because of the presence of glycans that are not substrates of the enzymes, because similar glycosidaseresistance, and even resistance to chemical cleavage, has been observed for glycoproteins produced by C. jejuni (24). With the confidence that the vast majority of Hp's azidelabeled proteins are glycosylated, we undertook experiments to study and identify these labeled glycoproteins.
Assessment of the Subcellular Distribution of Hp's Azide-Labeled Glycoproteins Reveals Glycoproteins Throughout Hp, Including on the Cell Surface-To study Hp's glycoproteins, we began by assessing the subcellular distribution of these species. In particular, we were curious to see whether any of Hp's secreted and cell surface proteins are glycosylated, as these proteins are likely to directly modulate interactions with the host and could be novel therapeutic targets. Hp cells supplemented with the azidosugar Ac 4 GlcNAz or with the azide-free control Ac 4 GlcNAc (31) were fractionated to acquire secreted (36), periplasmic, inner membrane-associated, cytoplasmic (35), and surface associated proteins (37). Further, fractions containing outer membrane proteins were also obtained (15,38). Western blot visualization of azide-labeled glycoproteins (Fig. 3) and SDS-PAGE analysis of total protein (see supplemental Fig. S2A-C) demonstrated that protein composition varied between fractions, consistent with successful subcellular separation of proteins. Further, malate dehydrogenase assays (50) of fractions revealed activities con- sistent with successful separation of cytoplasmic and periplasmic proteins (35) (supplemental Fig. S2D). No azidelabeled proteins were observed in fractions from cells supplemented with Ac 4 GlcNAc, whereas Ac 4 GlcNAz treatment resulted in robust azide-dependent signal in all subcellular fractions (Fig. 3). Our data demonstrate the remarkable selectivity of the Staudinger ligation between phosphines and azides, as these abiotic functional groups are nonreactive with endogenous functional groups yet display exquisitely selective reactivity with each other to form a covalent adduct (51,52). Moreover, these data indicate that glycoproteins are present throughout Hp and may serve diverse functions. For example, modification of internal glycoproteins may be important for the physiology of Hp, whereas glycosylation of secreted and surface proteins may be crucial for Hp to survive and cause harm within the gastric mucosa.
The prominent display of glycans on a subset of Hp's cell surface proteins might be harnessed for novel anti-Hp therapeutic strategies. For example, these glycosylated proteins might provide the basis of a carbohydrate-based vaccine; analogous glycoconjugate vaccines have been remarkably successful in the clinic (14,53). Moreover, if glycan modifications are important for the interactions of Hp with its host, small molecule inhibitors could target glycan synthesis to inactivate Hp. Finally, Hp's glycans could be selectively labeled with azides and targeted with phosphines to disrupt their function (14,54,55). Realizing these strategies would first require identification of Hp's glycoproteins.
Targeted Isolation of Hp's Azide-Labeled Glycoproteins was Facilitated by the Staudinger Ligation-To identify Hp's glycoproteins, we began by taking advantage of the metaboli-cally incorporated azide, in conjunction with the Staudinger ligation, to enrich these species. Lysates from Hp supplemented with Ac 4 GlcNAz or Ac 4 GlcNAc were subjected to Staudinger ligation with a phosphine comprising a tandem FLAG-hexahistidine peptide (Phos-FLAG-His 6 , Fig. 4A) (29,56). Tagged glycoproteins containing both the FLAG and His 6 epitopes were enriched by purification via anti-FLAG chromatography followed by Ni 2ϩ affinity chromatography (29, 56) (Fig. 4A). SDS-PAGE analysis enabled comparison of protein profiles from samples before purification (input) versus after enrichment of azide-labeled species (eluent). A large number of proteins were present in both the azide-labeled and acetyllabeled samples before purification (Fig. 4B). In contrast, after purification, proteins were present in substantially higher abundance in the enriched, azide-labeled sample relative to the mock-enriched control (Fig. 4B). These data highlight the selectivity of the Staudinger ligation between phosphines and azides. The ladder of proteins observed in the enriched, azide-labeled fraction is consistent with the large number of azide-labeled proteins observed by Western blot analysis of lysate from Ac 4 GlcNAz-treated cells (Figs. 2B, 2C). These data indicate a dramatic enrichment of Hp's azide-labeled glycoproteins using Phos-FLAG-His 6 . Analysis of the enriched, azide-labeled glycoprotein sample by two-dimensional gel electrophoresis yielded greater than 50 detectable species (Fig. 4C), indicating that Hp synthesizes myriad glycoproteins. The large number of Hp glycoproteins detected is similar in magnitude to the number of glycoproteins identified in large-scale studies of the related species C. jejuni (57,58) and in more distantly related bacteria, including Mycobacterium tuberculosis (26), Neisseria sp. (20)., and Bacteroides sp.  (27). Hp appears to dedicate considerable metabolic energy to glycoprotein synthesis, suggesting glycoproteins play an important role in Hp's physiology.

Identification of Hp's Azide-Labeled Glycoproteins by Mass Spectrometry Indicates Glycosylation of a Number of Viru-
lence Factors-Motivated by these results, we next sought to identify Hp's glycoproteins by mass spectrometry analysis. Hp's enriched, azide-labeled glycoproteins and a mock-enriched control (which was labeled with Ac 4 GlcNAc, reacted with Phos-FLAG-His 6 , and enriched via anti-FLAG and Ni-NTA affinity chromatography, as described above) were digested with trypsin, then analyzed by multi-dimensional protein identification technology (mudPIT) (Fig. 4A) (40). Experimental detection of unique peptides enabled the unambiguous identification of 127 proteins in the azide-labeled sample. In contrast, the mock-enriched control that lacked the azide contained only two Hp proteins (HP0721, HP0570), both of which were also present in the azide-labeled sample. These proteins were subtracted from the list of putative glycoproteins identified in the azide-labeled sample, narrowing the pool of candidate glycoproteins to 125 (Table I and Table  II). All proteins were identified based on two or more peptides meeting acceptance criteria. Detailed information on all of the proteins identified in the azide-labeled sample and the mockenriched control, including detected peptides that made assignments possible, can be found in supplemental Table S1. These results demonstrate efficient enrichment and identification of azide-labeled proteins, suggesting this metabolic labeling strategy is a robust method to profile bacterial glycoproteins.
Of the putative glycoproteins identified, 70% have housekeeping functions, 21% are linked to Hp's pathogenesis, and 9% have unknown function. From a basic biology perspective, the identification of a number of putative glycoproteins that have housekeeping functions suggests that glycosylation could play a role in proper folding and function (59), or perhaps that glycosylation provides a selective advantage for survival in the harsh conditions of the gastric tract. From a therapeutic perspective, it is intriguing that 26 of Hp's glycoproteins have known roles in pathogenesis ( Table I) that include enabling survival in the host stomach via iron acquisition (pfr) (60), pH resistance (ureA, ureB, tufB) (61), oxidative protection (tsaA, sodB, katA, tagD, msrA) (62, 63), motility (flgH) (64), adhesion (babC, babA, tlpA, dnaK, htrA) (65), eliciting an immune response (napA, msrA, icd, frdA) (66,67), and causing harm to the host (cag7, groES, groEL, hopQ, fusA) (68,69). The large percentage of virulence factors in Hp's glycoproteome, relative to the low percentage of virulence factors in Hp's proteome (70,71), implicates Hp's glycans in pathogenesis. This observation suggests that the known link between protein glycosylation and virulence in Hp, which Logan firmly established with her studies of Hp's flagellin glycosylation (7), may be a widespread phenomenon within this bacterium. We next considered the subcellular distribution of Hp's putative glycoproteins. Based on literature reports and in silico predictions, candidate glycoproteins appear to be expressed throughout Hp (Fig. 5), including intracellular locations such as the cytoplasm, inner membrane, and periplasm, and extracellular locations such as the outer membrane, cell surface (surface-associated), and the secreted milieu. Moreover, the predicted subcellular distribution of Hp's putative glycoproteins (Fig. 5) appears to reflect the subcellular distribution that we detected experimentally (Fig. 3). Follow-up experiments need to be conducted to establish whether these localization predictions are correct and to confirm that Hp synthesizes glycoproteins throughout the cell, as our data suggest.
We next compared the results of our analyses of Hp's glycoproteins to a recent study by Creuzenet and coworkers. In their study, they employed a periodic acid/hydrazide-based screen to identify nine candidate Hp glycoproteins (15). Five of the putative glycoprotein hits identified here (katA, gltA, tsf, atpA, atpD) were also identified as candidate glycoproteins in that screen, further validating our hit list. Moreover, their analyses indicated the presence of glycoproteins in soluble and membrane fractions of Hp (15), consistent with the subcellular distribution observed here. These findings underscore the value of our targeted strategy to detect and discover bacterial glycoproteins in a manner that is complementary to alternative techniques.
We note the absence of the flagellin proteins FlaA and FlaB, glycoproteins confirmed to be modified with pseudaminic acid, from our list of identified azide-labeled glycoproteins. This result is consistent with our previous studies, which demonstrated that Ac 4 GlcNAz treatment does not lead to detectable incorporation of azide into Hp's flagellin proteins (31). Our previously reported data indicate that Ac 4 GlcNAz is not converted to the complex sugar pseudaminic acid at detectable levels. Thus, though the studies described here identity a large number of putative Hp glycoproteins, they do not exhaustively identify all Hp glycoproteins.
Mass Spectrometry Analyses Reveal Staudinger Ligation-Glycan Adducts-We sought to characterize the nature of Hp's azide-modified glycans. A sample of Hp's enriched, azide-labeled glycoproteins was analyzed by HPLC-Chip/Q-TOFMS before and after being subjected to beta-elimination to cleave O-linked glycans. For these preliminary experiments, we focused on identifying the simplest Staudinger ligation-glycan adduct that would be formed as a result of metabolic labeling with Ac 4 GlcNAz -addition of a single GlcNAz residue or, if the stereochemistry is changed, an N-azidoacetylhexosamine (HexNAz) residue onto a glycosylated protein. Thus, we directed our analysis at the identification of the Staudinger ligation-glycan adduct that would be formed by protein glycosylation with a single HexNAz residue followed by release via beta-elimination -HexNAz-Phos-FLAG-His 6 (1) (Fig. 6). Because this adduct has not been     Table S3). We also detect triply charged versions of product ions at m/z 741.63 and 722.61 (ions Ib 3ϩ and IIb 3ϩ , respectively) (Figs. 6 and 7A), which may be formed by loss of the positively charged glycan via pathway (b) (Fig. 6). This pathway may also be responsible for formation of the glycan-derived product ions that appear in the spectrum at m/z 162.08, m/z 144.07, m/z 126.06, and m/z 96.04 (see proposed structures in Figs. 6 and 7B). We also detected FLAG-His 6 -derived product ions. These include abundant immonium ions at m/z 110.07 (His) and m/z 129.10 (Lys), as well as y-type (72) fragments containing the FLAG-His 6 C terminus (see Fig. 7E) and b-type (72) fragments; the latter are denoted by b n * and contain the glycan-eliminated N terminus (see Fig.  7E) that is attributed to subsequent fragmentations from ion IIa 4ϩ . To determine if the Staudinger ligation-glycan adduct HexNAz-Phos-FLAG-His 6 (1) is released from Hp's enriched, azide-labeled glycoproteins on beta-elimination, a sample was analyzed by HPLC-Chip/Q-TOFMS before and after being subjected to beta-elimination to cleave O-linked glycans.
Although analysis of the pre-beta-eliminated sample showed no evidence for product 1, LC-nanoESI analysis of the betaeliminated sample showed a chromatographic peak appearing at the retention time characteristic of synthetic 1, and producing a mass spectrum with m/z values consistent with the mass of product 1 (M ϭ 2,300.90 Da, mass measurement error Ͻ 5 ppm) (supplemental Fig. S4 and supplemental Table  S2). Furthermore, when the [Mϩ4H] 4ϩ ion at m/z 601.24 from putative product 1 was subjected to MS/MS analysis, the measured mass spectrum showed excellent agreement with the synthetic standard (Figs. 7C, 7D; supplemental Table S4). These results suggest that one metabolic fate of Ac 4 GlcNAz in Hp is conversion to HexNAz (presumably GlcNAz) and incorporation into Hp's O-linked glycoproteins.
Although the above data indicate the presence of a Staudinger ligation-glycan adduct in an azide-labeled glycoprotein sample, we sought to confirm that these adducts are absent from acetyl-labeled controls. Because of the metastable water losses observed with the Phos-FLAG-His 6 tag and the multiplicity of charge states, we chose to conduct fol- low-up experiments with the less complex Phos-FLAG tag. We first characterized synthetic GlcNAz-Phos-FLAG (2) (M ϭ 1578.54 Da) by LC-nanoESI and CID. With the basic His 6 chain eliminated, 2 yielded a nanoESI mass spectrum dominated by the lower charge state [Mϩ2H] 2ϩ ion at m/z 790.28 (supplemental Fig. S5; supplemental Table S5). As was observed for 1, the CID-MS/MS spectrum for GlcNAz-Phos-FLAG (2) (Fig. 8A and 8B) was dominated by peaks resulting from neutral glycan loss via pathway (a) to form product ions Ia 2ϩ and IIa 2ϩ (Figs. 6 and 8A); however, glycan-derived fragments (m/z ϭ 96.04, 126.05, 144.07 and 162.08) were weaker, whereas y-and b-type sequence ions (Figs. 8A and 8E) were more abundant in the spectrum of GlcNAz-Phos-FLAG (2), a change that may result from the lower charge state of this precursor.
With characterization of the targeted analyte in hand, lysates from Hp supplemented with Ac 4 GlcNAz or Ac 4 GlcNAc were rinsed extensively to remove residual free sugar, and then subjected to Staudinger ligation with Phos-FLAG. After extensive washing to remove unreacted Phos-FLAG from labeled lysates via ultrafiltration, samples were analyzed by HPLC-Chip/Q-TOFMS before and after being subjected to beta-eliminations to cleave O-linked glycans. LC-nanoESI analysis of the beta-eliminated glycans from the azide-labeled  Fig. 6) resulting from glycan loss, and ions IIa 4ϩ and IIb 3ϩ following net loss of CO, NH ϭ CH2 (Fig. 6). As summarized in (E), subsequent fragmentation of ion IIa 4ϩ is proposed to yield y-type fragment ions that include the C terminus of the FLAG-His 6 peptide and b-type ions, labeled with an asterisk, that include the Phos-modified N terminus. B, Expansion (magnified 5x along the y axis) of the low-mass region from (A) displays a variety of peaks that include glycan-derived fragments with proposed structures represented. Immonium ions (H and K) are also detected in this mass range. C, CID-MS/MS spectra of beta-eliminated glycans from enriched azide-labeled glycoprotein sample showing mass spectral features that agree with those of the synthetic standard (A). D, Expansion (magnified 5x along the y axis) of the low-mass region from (C), showing the production of glycan-derived peaks from the beta-eliminated putative HexNAz-Phos-FLAG-His 6 (1). E, Schematic representation of ions (y-and b-type) derived from further IIa 4ϩ fragmentation. The asterisk associated with the b-type ions denotes a Phos-modified N terminus. Charge states are labeled when greater than ϩ1. All CID-MS/MS spectra were measured using nitrogen as the collision gas, a 4 Da window for ion isolation, and collision energy of 20 V. sample revealed a doubly-charged peak at m/z 790.28 that corresponds to the mass of HexNAz-Phos-FLAG (2) (M ϭ 1578.54 Da, mass measurement error Ͻ 5 ppm; supplemental Fig. 6E and supplemental Table S2). The CID-MS/MS spectrum of the peak at m/z 790.28 ( Fig. 8C and 8D; supplemental Table S6) is consistent with the Staudinger ligation product between GlcNAz and Phos-FLAG -HexNAz-Phos-FLAG (2). This HexNAz-Phos-FLAG adduct is not detectable in the beta-eliminated sample from the corresponding acetyl-labeled control (supplemental Fig. S6C), nor is it present at detectable levels in tagged azide-or acetyl-labeled lysates before beta-elimination (supplemental Fig. S6A and S6B). The exclusive detection of this adduct in beta-eliminated azide-labeled samples indicates that it is derived from glycosylated Hp proteins and that beta-elimination conditions released this glycan adduct. The absence of this adduct in samples before beta-elimination further supports the conclusion that this adduct is derived from Hp glycoprotein conjugates and is released by the beta-elimination reaction. These data provide strong evidence that some Hp glycoproteins are covalently modified with O-linked HexNAz.
N-acetylhexosamine (HexNAc) residues are found in a number of characterized bacterial glycoproteins, including those synthesized by Campylobacter jejuni (73) and Helicobacter pullorum (74). Indeed, H. pullorum has an N-linked protein glycosylation system that makes use of a pentasaccharide  Fig. 6) resulting from glycan loss, and ions IIa 2ϩ following net loss of CO, NH ϭ CH2 (Fig. 6). As summarized in (E), subsequent fragmentation of ion IIa 2ϩ is proposed to yield y-type fragment ions that include the C terminus of the FLAG peptide and b-type ions, labeled with an asterisk, that include the Phos-modified N terminus. B, Expansion of the low-mass region from (A) displays a variety of peaks that include glycan-derived fragments with proposed structures represented. Immonium ions (K) are also detected in this mass range. C, CID-MS/MS spectra of beta-eliminated glycans from an azide-labeled, Phos-FLAG-ligated, Hp cell lysate showing mass spectral features that agree with those of the synthetic standard (A). D, Expansion of the low-mass region from (C), showing the production of glycan-derived peaks from the beta-eliminated putative HexNAz-Phos-FLAG (2). E, Schematic representation of ions (y-and b-type) derived from further IIa 4ϩ fragmentation. The asterisk associated with the b-type ions denotes a Phos-modified N terminus. Charge states are labeled when greater than ϩ1. All CID-MS/MS spectra were measured using nitrogen as the collision gas, a 4 Da window for ion isolation, and collision energy of 35 V. with a core HexNAz. Structural studies are underway in our laboratory to assess whether Hp's O-linked HexNAz is elaborated on to form a higher order glycan.
In our experiments, detection of HexNAz-Phos-FLAG-His 6 and HexNAz-Phos-FLAG adducts by HPLC-Chip/Q-TOFMS analysis was greatly aided by the presence of the peptide moiety, which has favorable chromatographic and ionization characteristics. Thus, MOE facilitated detection of Hp's labeled glycans in unanticipated ways. If chromatographic resolution and signal enhancement are generalizable, metabolic glycan labeling coupled to reaction with a readily ionizable tag has the potential to amplify glycan detection in myriad bacteria and overcome challenges associated with the poor resolution and ionizablility of glycans.
ureA and ureB are Biochemically Validated Glycoproteins-Of the putative glycoproteins identified, two of themurease alpha (ureA) and urease beta (ureB) -are surfaceassociated colonization and persistance factors (75). We sought to evaluate their glycosylation status, because this has important implications for host-pathogen interactions and the development of novel anti-Hp therapeutics. To assess the glycosylation status of ureA and ureB, these proteins were immunopurified from Hp treated with 1 mM Ac 4 GlcNAz or Ac 4 GlcNAc, reacted with Phos-FLAG, and then analyzed by Western blot. Probing immunopurified ureA with anti-ureA revealed the presence of a band at ϳ27 kDa (Fig. 9A), indicating successful enrichment of ureA. Further, probing with anti-FLAG antibody revealed the presence of azide-dependent signal in ureA (Fig. 9A). Similar analyses of immunopurified ureB indicated successful enrichment of ureB (Fig. 9B) and the presence of azide-dependent signal in ureB (Fig. 9B). These results suggest that ureA and ureB are glycosylated with azide-modified glycans.
Based on our glycan mass spectrometry results, we were curious to see whether urease (composed of alpha and beta subunits) is among the glycoproteins modified with O-linked HexNAz in Hp. To assess this possibility, urease was biochemically purified from Ac 4 GlcNAz-labeled Hp. Purified ure-ase (supplemental Fig. S7) was reacted with Phos-FLAG and rinsed extensively to remove residual Phos-FLAG. Hp's labeled urease was analyzed by HPLC-Chip/Q-TOFMS before and after being subjected to beta-elimination to cleave Olinked glycans. Although analysis of the pre-beta-eliminated sample showed no evidence for product 2 (supplemental Fig.  S8A), LC-nanoESI analysis of the beta-eliminated sample showed a chromatographic peak appearing at the retention time characteristic of synthetic HexNAz-Phos-FLAG (2) (supplemental Fig. S8B) and producing a mass spectrum (supplemental Fig. S8C) with m/z values consistent with the mass of product 2 (M ϭ 1578.54 Da, mass measurement error Ͻ 5 ppm; supplemental Table S2). Furthermore, when the [Mϩ2H] 2ϩ ion at m/z 790.28 from putative product 2 was subjected to MS/MS analysis, the measured mass spectrum showed excellent agreement with the synthetic standard (supplemental Figs. S8D and S8E; supplemental Table S7). These results suggest that urease is glycosylated with Olinked HexNAz and have implications for the development of urease-based therapeutic strategies to prevent and treat Hp infection. Taken together, these data validate urease as a glycosylated protein and indicate that MOE is a robust approach to profiling bacterial glycoproteins.

CONCLUDING REMARKS
Glycosylation of Hp's flagellin proteins has been directly linked to Hp's pathogenesis, suggesting that Hp's glycoproteins are potential targets of therapeutic intervention. Despite their potential therapeutic importance, only the two flagellin glycoproteins have been firmly characterized in Hp (7). Mounting evidence from our lab (76) and others (15) suggests protein glycosylation is common in Hp. Employing experimental approaches to selectively study and identify Hp's glycoproteins should facilitate the discovery of new therapeutic targets and shed light on the role of protein glycosylation in Hp's physiology. In this article, we investigated the production of glycosylated proteins in Hp using a combination of metabolic glycan labeling and mass spectrometry analysis. Metabolic labeling of glycans with unnatural sugars and subsequent enrichment has been applied to identify glycoproteins in eukaryotic systems (29,56,77). Though metabolic glycan labeling has been used to study bacterial glycolipids (78), to our knowledge this study is the first to demonstrate the applicability of this chemical approach to bacterial glycoproteomics (79).
Metabolic incorporation of azide-labeled sugars, followed by reaction with phosphine probes via Staudinger ligation, facilitates detection, visualization, and enrichment of cellular glycoproteins (29). We began our studies by assessing the subcellular distribution of Hp's azide-labeled glycoproteins to gain insight into their potential functions. Western blot data indicated that glycoproteins are present throughout Hp, including on the cell surface, embedded in both the inner and outer membranes, and within the periplasm and cytoplasm. FIG. 9. Secondary screens to assess the glycosylation status of urease alpha (ureA) and urease beta (ureB) subunits. A, Western blot analysis of immunopurified ureA from Hp treated with Ac 4 GlcNAz (Az) or Ac 4 GlcNAc (Ac). These samples were reacted with Phos-FLAG, electrophoresed, and then probed with anti-ureA or anti-FLAG antibody. B, Western blot analysis of immunopurified ureB from Hp treated with Ac 4 GlcNAz (Az) or Ac 4 GlcNAc (Ac). These samples were reacted with Phos-FLAG, electrophoresed, and then probed with anti-ureB or anti-FLAG antibody.
The observed subcellular distribution suggests that protein glycosylation is abundant, widespread, and critical for Hp's physiology. Broadly, glycans modulate protein function in one of two ways: by providing epitopes for binding and recognition by other proteins (80,81), and by stabilizing proteins (e.g. folding, solubility, protease susceptibility) (82)(83)(84). Hp's intracellular glycans are likely involved in stabilizing proteins, whereas those exposed to the extracellular milieu (surface and secreted glycoproteins) may stabilize proteins as well as modulate interactions with the host.
Based on our results, the subcellular distribution of glycoproteins in Hp appears to be broader than that observed in other bacteria to date. For example, in C. jejuni (85,86), Neisseria sp. (20,87), and Bacteroides sp. (27)., general protein glycosylation systems are present within the periplasm and modify proteins that traffic through this compartment (e.g. periplasmic, membrane-bound, and secreted proteins). In Actinobacillus, in contrast, there is a general protein glycosylation system within the cytosol that modifies cytosolic proteins (88). Finally, in Haemophilus influenzae, monosaccharides are added to proteins in the cytoplasm and then transported to the cell envelope (89). Future research will have to be conducted to confirm the localization of our candidate glycoproteins. If the observed, broad distribution is confirmed by follow-up experiments, this result suggests that Hp may have two general protein glycosylation systems (one within the cytosol and one within the periplasm) or an exclusively cytoplasmic glycosylation pathway that modifies proteins before sorting. Based on predictions by the carbohydrate-active enzymes (CAZy) database (90), there are 22 glycosyltransferases within Hp strain 26695, and half of these have unknown functions. Therefore, there are numerous glycosyltransferases within Hp that may be responsible for the synthesis of the observed glycoproteins. Characterizing which of these genes are responsible for carbohydrate assembly pathways is challenging because, in contrast to most bacteria, the genes for Hp carbohydrate biosynthesis pathways (e.g. O-antigen biosynthesis) are spread throughout the chromosome (91). Efforts are underway in our laboratory to identify the glycosyltransferases responsible for protein glycosylation in Hp.
We were excited to observe that a subset of Hp's azidemodified glycoproteins are present on the cell surface, as these surface-exposed glycoproteins and the glycans that modify them are potential therapeutic targets. For example, these glycosylated proteins could provide the basis for a carbohydrate-based vaccine (14). Unusual bacterial monosaccharides, such as di-N-acetylbacillosamine produced by Neisseria sp. (92), can themselves be immunogenic. These glycans, in the context of a glycosylated protein, may serve as particularly effective conjugate vaccines. Alternatively, glycosylation pathways could be targeted with small molecule inhibitors to interrupt interactions between Hp and its host. Finally, our demonstration that azide-labeled glycoproteins are present on the cell surface suggests that azide-covered Hp could be targeted with therapeutic phosphines. Covalent modification with properly designed phosphines could disrupt glycan function or render Hp innocuous within the host (93). Indeed, Kaewsapsak et al. designed phosphine therapeutics conjugated to the immune stimulant 2,4-dinitrophenyl to selectively kill Hp covered with azide-labeled glycans (55).
Staudinger ligation with Phos-FLAG-His 6 allowed for selective and efficient isolation of Hp's azide-labeled glycoproteins. MudPIT analyses of enriched azide-labeled proteins and a mock-enriched control enabled the unambiguous identification of 125 putative glycoproteins. We identified a large number of glycoprotein hits in this study relative to the small number (nine) identified in Creuzenet and coworkers' recent periodic acid/hydrazide-based screen of Hp's glycoproteins (15). We attribute much of this difference to the high incidence of background signal associated with periodic acid/hydrazide chemistry (15), rendering the identification of glycoprotein signal relative to nonglycoprotein noise difficult. In contrast, metabolic glycan labeling has a very high signal-to-noise ratio, enabling the facile detection and selective enrichment of glycoproteins. Thus, metabolic glycan labeling offers a complementary and efficient alternative to existing approaches for bacterial glycoprotein detection and identification.
In our studies, we detected Staudinger ligation-HexNAz adducts in beta-eliminated glycan samples from enriched azide-labeled glycoproteins, azide-labeled Hp, and purified azide-labeled urease. The exclusive presence of these adducts in samples from azide-labeled Hp provides evidence that the Staudinger ligation is exquisitely selective for azidebearing molecules. Moreover, these results indicate that one metabolic fate of Ac 4 GlcNAz in Hp is conversion to HexNAz. The addition of the Phos-FLAG moiety onto the HexNAz epitope greatly enhanced the chromatographic resolution, ionization, and detection of this glycan species. Optimization of the phosphine probe to contain isotopic tags or characteristic fragment ions will further facilitate glycoprotein identification and glycan characterization. Thus, MOE is a robust approach for glycoprotein discovery.
A large number of proteins with known links to colonization, persistence and virulence were identified as glycoprotein hits (Table I), and two of these colonization factors, ureA and ureB, were biochemically validated as glycoproteins in our studies. This finding is consistent both with previous studies indicating that flagellin glycosylation is important to Hp's ability to survive within the host (94), and more broadly, with the known links between protein glycosylation and virulence in a number of medically significant bacterial pathogens (13,14). Further, this observed link underscores the importance of Hp's glycoproteins as potential therapeutic targets.
The discovery that the urease subunits are glycosylated has potential implications for the development of urease-based therapeutic strategies to prevent and treat Hp infection. Vaccines based on recombinant urease have had some success in preclinical (95,96) and clinical studies (97,98). However, these vaccines are based on recombinant proteins produced in Escherichia coli or Salmonella in their nonglycosylated state. New vaccines based on Hp's natively glycosylated urease have the potential to stimulate a robust immune response and thus protect patients more effectively than those based on recombinant urease.
In summary, we were able to use metabolic glycan labeling to enrich and identify 125 putative Hp glycoproteins. Further, we validated the glycosylation status of two of these hits. Finally, we characterized the resulting glycan adducts. These results reveal that glycosylated proteins are abundant in Hp, indicate that Hp's glycoproteins have a wide range of functions, and reveal new links between Hp's pathogenesis and glycosylation. Armed with this information, the stage is set for the development of glycosylation-based therapeutic strategies to eradicate Hp infection. In particular, this work opens the door to phosphine-based therapeutics that target Hp's azide-labeled glycans.
Broadly, this work validates MOE as a viable approach to discover and characterize bacterial glycoproteins. In addition to enabling the discovery of Hp's glycoproteins, MOE has the potential to facilitate the study of glycoproteins in other bacteria. Indeed, reports of metabolic labeling of glycoproteins with azide-modified pseudaminic acid in C. jejuni (54) and with alkyne-bearing fucosylated sugars in Bacteroidales sp. (99) suggest that this glycoproteomic approach should transfer readily to these organisms. The ability to use a general metabolic precursor rather than a specific glycan-binding reagent will facilitate the detection and discovery of glycoproteins in a broad range of bacteria. Thus, MOE will propel bacterial glycoproteomics forward and, in the process, unveil novel therapeutic targets.