Metabolic glycan labeling-based screen to identify bacterial glycosylation genes

Abstract

Bacterial cell surface glycans are quintessential drug targets due to their critical role in colonization of the host, pathogen survival, and immune evasion. The dense cell envelope glycocalyx contains distinctive monosaccharides that are stitched together into higherorder glycans to yield exclusively bacterial structures that are critical for strain fitness and pathogenesis. However, the systematic study and inhibition of bacterial glycosylation enzymes remains challenging. Bacteria produce glycans containing rare sugars refractory to traditional glycan analysis, complicating the study of bacterial glycans and the identification of their biosynthesis machinery. To ease the study of bacterial glycans in the absence of detailed structural information, we used metabolic glycan labeling to detect changes in glycan biosynthesis. Here we screened wild type versus mutant strains of the gastric pathogen Helicobacter pylori, ultimately permitting the identification of genes involved in glycoprotein and lipopolysaccharide biosynthesis. Our findings provide the first evidence that H. pylori protein glycosylation proceeds via a lipid carrier-mediated pathway that overlaps with lipopolysaccharide biosynthesis. Protein glycosylation mutants displayed fitness defects consistent with those induced by small molecule glycosylation inhibitors. Broadly, our results suggest a facile approach to screen for bacterial glycosylation genes and gain insight into their biosynthesis and functional importance, even in the absence of glycan structural information.

Graphical Abstract

Antibiotics that interfere with proper cell envelope construction¹ – including penicillin², vancomycin³, and bacitracin⁴ – have had remarkable success in the clinic. Their successes serve as a testament to the cell surface’s attractiveness as a drug target. Cell envelope glycans are quintessential drug targets due to their absence from human cells, their distinctive structures, and their role in bacterial fitness.^5–7 Despite the successes of blockbuster antibiotics that target the cell envelope, their efficacy has eroded due to the evolution and prevalence of antibiotic resistance mechanisms among bacterial pathogens.⁸ New cell envelope interference agents are urgently needed to disarm antibiotic resistant bacterial pathogens.

The bacterial cell envelope is covered with a dense array of glycan structures that are linked to fitness and pathogenesis. Gram-negative bacterial cell envelopes are comprised of peptidoglycan, lipopolysaccharide (LPS), and capsular polysaccharide (CPS) enmeshed on and between two membranes (Figure 1A).⁹ These complex structures are stitched together by glycosyltransferases adding one-sugar-at-a-time onto growing glycan structures, and in some instances, they are transferred en bloc onto protein or lipid substrates.¹⁰ These structures buttress bacterial cells and allow them to colonize the host, adhere to host cells, and evade the immune system. In addition to these canonical cell envelope glycans, bacterial glycoproteins have recently emerged as intriguing cell envelope denizens that are not broadly synthesized by all bacteria but instead are synthesized by select medically significant “priority pathogens,”⁸ such as Helicobacter pylori, Campylobacter jejuni, Pseudanonas aeruginosa, Haemophilus influenza, and Staphylococcus aureus^{5, 11}, as well as by select commensal¹² and environmental bacteria.¹³ The discovery of bacterial glycoproteins expands the list of potential targets of therapeutic intervention.

Figure 1. Bacteria synthesize distinctive cell envelope glycans that can be detected by metabolic oligosaccharide engineering (MOE); genetic disruption of glycoprotein biosynthesis yields a diminished MOE-based signal.

A) Gram-negative bacteria employ glycosyltransferases to construct cell envelope glycans, including peptidoglycan, lipopolysaccharide (LPS), and capsular polysaccharide (CPS), as well as glycoproteins in select bacterial strains. Each of these structures contains exclusively bacterial monosaccharides, shown as colored hexagons in this schematic, that are absent from human cells and complicate their study. B) Schematic overview of MOE, which facilitates detection and visualization of cellular glycan biosynthesis. Wildtype (WT) bacteria are first metabolically labeled with an unnatural, azide-containing sugar, then azides are covalently elaborated via Staudinger ligation with a phosphine-based probe to yield the full complement of detectable glycoproteins within cell lysates. C) MOE was used as a cell-based readout of interrupted glycoprotein biosynthesis in mutant versus wildtype bacteria. The effect of insertional inactivation of glycosyltransferase genes within mutant (ΔGT) strains was detected by diminished metabolic incorporation of unnatural, azide-containing sugars into cellular glycoproteins.

Bacterial glycoproteins have emerged as attractive targets because they are synthesized only by select bacteria, they contain distinctive structures that are markedly different from their eukaryotic counterparts, and they often play important roles in colonization and pathogenesis.^{5, 6} Indeed, an analysis of bacterial glycan structures reveals the presence of rare, exclusively bacterial amino deoxy monosaccharides that are absent from eukaryotes.¹⁴ Pseudaminic acid,¹⁵ bacillosamine,¹⁶ 2,4-diacetamido-2,4,6-trideoxygalactose,¹⁷ N-acetylfucosamine,¹⁸ and legionaminic acid¹⁵ are a sampling of the unusual building blocks used by medically significant priority pathogens. Pathogenic sampling of the unusual building blocks used by medically significant priority pathogens. Pathogenic bacteria that cannot synthesize glycoproteins have attenuated virulence. For example, P. aeruginosa and H. influenza glycosylation mutants have reduced adhesion to host cells,^{19, 20} S. aureus glycosylation mutants are defective at forming bio-films,²¹ and C. jejuni, H. pylori, and P. aeruginosa mutants with interruptions in flagellin glycosylation are immotile and cannot colonize the host.^{15, 22, 23} Taken together, these data indicate that the enzymes responsible for protein glycosylation in priority pathogens are potential therapeutic targets.

Despite their appeal as molecular targets, bacterial glycans and the machinery responsible for their construction are difficult to study. Glycans are challenging to study in any system due to their complex, branched, and often-heterogeneous structures. The study of bacterial glycoproteins is further compounded by the presence of rare exclusively bacterial monosaccharides,¹⁴ which are refractory to study using tools developed for the study of eukaryotic glycans.^{5, 24} For example, glycan-binding reagents such as lectins and antibodies often fail to bind to bacterial glycans, glycan-degrading enzymes typically do not recognize bacterial glycans as substrates, and glycan analysis techniques with automated analysis software, such as mass spectrometry, are programmed to detect eukaryotic rather than prokaryotic building blocks and structures.^25–27 Though rigorous biochemical, genetics, and analytical studies have yielded important insights into bacterial glycoproteins and how they are biosynthesized in C. jejuni^28–31 and Neisseria spp.,^32–34 for example, these approaches are time consuming and not readily generalizable. A lack of precise glycomics information has impeded pathway characterization of glycoprotein biosynthesis in some priority pathogens. As a result, identifying bacterial glycoprotein biosynthesis machinery and establishing how these components act together to yield higher order glycans has been challenging.

Chemical tools have the potential to accelerate the study of bacterial glycoprotein biosynthesis. Indeed, metabolic oligosaccharide engineering (MOE), which was pioneered by Bertozzi, Reutter, and colleagues for the study of mammalian glycans,^{35, 36} has proven to be an enabling technology in bacterial systems. Since the initial demonstration that bacterial cells supplemented with unnatural monosaccharides bearing azides take up those sugars, process them via endogenous carbohydrate biosynthetic pathways, and install them into cellular glycans in place of natural monosaccharides (Figure 1B),³⁷ MOE has led to the discovery,^37–39 tracking,⁴⁰ and inhibition⁴¹ of bacterial glycosylation systems. Azide-containing sugars have facilitated tagging of labeled glycans with bioorthogonal reaction partners (e.g. alkynes, phosphines) to enable detection, identification, and visualization of bacterial glycans in live cells and animal infection models.^{6, 40, 42}

Motivated by the observation that azide-containing sugars provide a readout of bacterial glycan biosynthesis, we reasoned that metabolic substrates could effectively report on glycan biosynthesis defects and guide the identification of genes required for glycan biosynthesis, even in the absence of detailed pathway or glycan information. Guided by previous work in which azide-containing sugars led to visualization of the full complement of glycoproteins synthesized by bacteria (Figure 1B),^{37, 38} we hypothesized that disruption of genes involved in glycoprotein biosynthesis would yield a defective glycoprotein biosynthesis “fingerprint” relative to wildtype bacteria (Figure 1C). Here we report that MOE-based analysis of glycoprotein “fingerprints” across glycosyltransferase mutant strains implicated 13 genes in a novel glycoprotein biosynthesis pathway in the gastric pathogen H. pylori. Access to these glycoprotein biosynthesis-defective strains provided biosynthetic insights and allowed the scoring of fitness features, even in the absence of detailed structural information. This work unveils potential targets that could be harnessed for the treatment of H. pylori infection. Broadly, this work validates metabolic oligosaccharide engineering as a tractable method to uncover genes involved in bacterial glycan biosynthesis.

RESULTS AND DISCUSSION

To test the hypothesis that MOE could serve as the basis for a screen to discover glycosylation genes, we turned our attention to a bacterial protein glycosylation system in which metabolic labeling has been well established yet many unknowns about the biosynthetic pathway remain. The gram-negative gastric pathogen H. pylori synthesizes an array of glycosylated proteins.^{15, 37, 38, 43} Two of these are flagellin glycoproteins that are synthesized by a well-characterized, dedicated flagellin glycosylation system.¹⁵ In addition, H. pylori synthesizes a suite of 125 probe-enriched non-flagellin glycoproteins, two of which were biochemically validated as glycosylated and the rest of which may be considered putative glycoproteins. These species were initially detected and later identified using an MOE-based approach.^{37, 38} Though the structures of the glycans are unknown, cleavage and mass spectrometry analyses of O-linked glycans from enriched, azide-labeled glycoproteins led to the detection of Staudinger ligation-glycan adducts in samples, which supports the synthesis of O-linked glycoproteins in H. pylori.³⁸ Disruption of H. pylori’s general protein glycosylation system using metabolic glycan inhibitors precipitated functional defects in growth, motility and biofilm formation,⁴¹ suggesting that H. pylori’s general glycosylation system is linked to its ability to initiate and maintain infection. Although H. pylori synthesizes a large number of glycoproteins, the biosynthetic pathway responsible for synthesizing non-flagellin glycoproteins is unknown.

Targeted inactivation of glycan biosynthesis genes

Here we set out to employ metabolic glycan labeling to facilitate identification of genes responsible for protein glycosylation in H. pylori. Characterizing which H. pylori genes are responsible for carbohydrate assembly is challenging because, in contrast to most bacteria, the genes for H. pylori carbohydrate biosynthesis pathways (e.g. O-antigen biosynthesis) are spread throughout the chromosome.⁴⁴ Therefore, genetic inactivation of glycosylation islands is not possible. Based on bioinformatics studies using the carbohydrate active enzymes (CAZY) classification system,^{45, 46} twenty-eight open reading frames (ORFs) have been identified as glycosyltransferase genes potentially involved in glycan biosynthesis within H. pylori strain G27.⁴⁷ Though many of these genes have been assigned putative roles based on sequence homology to glycosyltransferase genes with known roles in other bacterial strains, only eight H. pylori glycosyltransferases (seven involved in lipopolysaccharide biosynthesis^48–54 and one involved in flagellin glycosylation⁵⁵) have been biochemically characterized. Therefore, there are numerous glycosyltransferases within H. pylori that may be responsible for the synthesis of non-flagellin glycoproteins.

We relied on CAZY^23,24 bioinformatics predictions to select nineteen putative glycosyltransferase genes in H. pylori strain G27 that have homology to known glycosyltransferase genes yet have not been biochemically characterized (Table 1); we also included HpG27_645 in our studies because it has homology to WlaA, a gene found in glycosylation clusters in other bacteria.⁵⁶ In addition to focusing on the role of putative glycosyltransferase genes in our studies, we included one glycosyltransferase gene, HpG27_1518, which has a confirmed role in initiating lipopolysaccharide (LPS) biosynthesis in this strain.⁴⁴ We also included several glycosyltransferase genes that are involved in LPS biosynthesis,⁵⁷ as there is known pathway overlap in glycoprotein biosynthesis and LPS biosynthesis in select bacteria.^{58, 59} We included several glycosyltransferase genes with established roles in peptidoglycan biosynthesis⁶⁰ and cholesterol modification⁶¹ to explore additional possible roles. Each putative glycosylation open reading frame (ORF) was insertionally inactivated with a chloramphenicol acetyl transferase marker (CAM^R) to yield a series of H. pylori G27 mutant strains, which have been previously reported.^{62, 63} Insertional inactivation of each target gene was confirmed by PCR amplification of target genes from wildtype versus mutant strain genomic DNA. In each mutant strain, an increase in target gene size consistent with CAM^R insertion was observed, confirming successful disruption of each ORF (Figure S1).

Table 1.

Inactivated glycan biosynthesis genes^{*^}

G27 ORF	26695 ORF	proposed/established function	glycoprotein	LPS
HpG27_94	HP0102	fucosyltransferase involved in G27 LPS biosynthesis57	✕	✕
HpG27_146	HP0159	α−1,6 glucosyltransferase; involved in G27 LPS biosynthesis57	✓	✕
HpG27_190	HP0208	putative glycosyltransferase; no role in G27 LPS biosynthesis	✕	✓
HpG27_579	HP0619, JHP0562	putative β−1,3-galactosyltransferase; no role in G27 LPS	✕	✓
HpG27_580	HP0619, JHP0563	putative β−1,3-galactosyltransferase; no role in G27 LPS	✕	✓
HpG27_607	HP0645	exolytic transglycosylase in peptidoglycan biosynthesis60	✓	✓
HpG27_613	HP0379	α−1,3 fucosyltransferase invovled in G27 LPS biosynthesis57	✕	✕
HpG27_645	n/a	putative DNA polymerase; homology to WlaA in C.jejuni pgl operon56	✕	✓
HpG27_761	HP0805	putative glycosyltransferase	✕	✓
HpG27_785	HP0826	β−1,4-galactosyltransferase in G27 LPS biosynthesis57	✕	✕
HpG27_952	HP0421	α-glucosyltransferase in cholesterol modification61	✕	✕
HpG27_1018	HP0651	α−1,3 fucosyltransferase involved in LPS biosynthesis57	✓	✕
HpG27_1036	HP0360	putative UDP-glucose-4-epimerase based on homology to gne	✓	?
HpG27_1046	HP1105	β−1,3-galactosyltransferase in G27 LPS biosynthesis57	✓	✕
HpG27_1230	n/a	unknown glycosyltransferase; involved in G27 LPS biosynthesis57	✕	✕
HpG27_1235	HP1283	putative heptosyltransferase; involved in G27 LPS biosynthesis57	✓	✕
HpG27_1236	HP1284	heptosyltransferase involved in G27 LPS biosynthess57	✕	✕
HpG27_1339	HP1416	α−1,2-glucosyltransferase involved in core LPS biosynthesis	✓	✓
HpG27_1509	HP1572	endolytic transglycosylase in peptidoglycan biosynthesis60	✕	✓
HpG27_1515	HP1578	putative GlcNAc transferase; involved in G27 LPS biosynthesis57	✓	✕
HpG27_1518	HP1581	undecaprenyl N-acetylglucosaminyl transferase (WecA) in G2744	✕	✕

^*Shaded entries indicate genes involved in G27 LPS biosynthesis.⁵⁷

^{^}Intact (✓) versus interrupted (✕) biosynthesis observed our studies.

MOE-based screen of glycosylation mutants

Next, we screened H. pylori G27 mutants containing insertions in putative glycosyltransferase-encoding genes for their ability to synthesize the full suite of H. pylori glycoproteins. Following our published methods, glycoprotein production was monitored via metabolic labeling experiments with the azide-containing sugar peracetylated N-azidoacetylglucosamine (GlcNAz).^{37, 38} Briefly, H. pylori parental or mutant strains were treated with 1 mM GlcNAz for four days. Following metabolic labeling the relative abundance of azide-labeled glycoproteins in cell lysates was assessed by Staudinger ligation with a phosphine probe comprising a FLAG peptide (Phos-FLAG) and subsequent Western blot analysis with anti-FLAG.⁶⁴ To benchmark the most and least signal, respectively, that we would expect to observe in any given sample, we performed metabolic labeling of wild type H. pylori with GlcNAz as a positive control and with the azide-free sugar peracetylated N-acetylglucosamine (Glc-NAc) as a negative control. Consistent with previous work,^{37, 38} metabolic labeling of wild type H. pylori with GlcNAz yielded a robust array of azide-labeled glycoproteins, whereas control samples from cells treated with the azide-free sugar GlcNAc yielded negligible azide-dependent signal (Figure 2A; Figure S1).

Figure 2. Metabolic labeling-based assay reveals H. pylori’s glycoprotein profile is altered upon genetic inactivation of select glycosylation genes and by chemical removal of O-linked glycans.

A) Wildtype G27 H. pylori were screened for their ability to biosynthesize the full profile of azide-labeled glycoproteins upon treatment with the azide-containing sugar Ac₄GlcNAz. H. pylori were grown for four days in media supplemented with 1 mM Ac₄GlcNAz (GlcNAz) or the azide-free control sugar Ac₄GlcNAc (GlcNAc), then harvested by lysis. The presence of azides in cellular glycoproteins was detected by reacting lysates with 250 μM Phos-FLAG for 12 h at room temperature and then analyzing samples via western blot with anti-FLAG antibody. B) Wild type (WT) and mutant G27 strains bearing insertionally inactivated glycosylation genes strains were screened for their ability to biosynthesize the full profile of azide-labeled glycoproteins upon treatment with the azide-containing sugar Ac₄GlcNAz for four days. Following metabolic labeling, samples were lysed, reacted with Phos-FLAG, and analyzed via western blot with anti-FLAG antibody as indicated in (A). C) Wildtype G27 H. pylori were metabolically labeled with Ac₄GlcNAz for four days. Following metabolic labeling, samples were lysed, reacted with Phos-FLAG, and left untreated (−) or subjected to O-linked glycan removal via beta-elimination (+) prior to analysis via western blot with anti-FLAG antibody. The data shown are representative of three biological replicates.

With the MOE-based assay validated, we turned to our experimental samples. Several of the mutant strains (e.g. Δ607, Δ1339) exhibited robust azide-labeled glycoprotein production that was comparable to the glycoprotein fingerprint produced by wild type cells (Figure 2B; Figure S2). These data indicate that the inactivated glycosyltransferase genes (e.g. HpG27_607, HpG27_1339) in those strains are not required for glycoprotein biosynthesis. By contrast, thirteen mutant strains (e.g.Δ579, Δ580) displayed substantially reduced azide-labeled glycoprotein production relative to wild type cells (Figure 2B; Figure S2). Zinc or Coomassie staining of electrophoresed samples confirmed that all samples contain equivalent protein levels, indicating that the observed decreases in azide incorporation profiles in select samples were not due to low protein levels (Figure S2). Thus, thirteen H. pylori genes – those inactivated within glycoprotein-deficient mutant strains – appear involved in glycoprotein biosynthesis.

We were struck by the similarity of the glycoprotein fingerprints across mutant strains displaying glycosylation defects; this shared defective glycoprotein fingerprint may indicate a common biosynthetic pathway, along which interruptions at multiple points have the same net effect on observed glycoprotein biosynthesis. Chemical removal of O-linked glycans from azide-labeled wild type H. pylori samples yielded a similar glycoprotein fingerprint (Figure 2C). Metabolically labeled wild type H. pylori samples exhibited a robust azide-dependent glycoprotein profile prior to beta-elimination. By contrast, the majority of azide-dependent signal was lost following removal of O-linked glycans via beta-elimination (Figure 2C). Coomassie staining of electrophoresed samples confirmed equivalent protein levels, indicating that the observed decrease in azide-labeled glycoproteins in the beta-eliminated sample was not due to low protein levels (Figure S2). The similarity in glycoprotein profiles observed in samples stripped of O-linked glycans and in glycosylation mutants implicate the thirteen genes within these mutant strains in a general O-linked protein glycosylation system.

Glycoprotein biosynthesis likely occurs by a lipid-carrier pathway

Protein glycosylation in bacteria proceeds by one of two general types of mechanisms: a lipid carrier-mediated assembly or a stepwise protein glycosylation pathway.¹⁰ In the lipid carrier-mediated assembly pathway, glycosyltransferases sequentially transfer nucleotide-activated monosaccharides one-sugar-at-a-time onto a lipid carrier to construct lipid-linked oligosaccharides in the cytoplasm. The lipid-linked oligosaccharides are then flipped across the inner membrane into the periplasm, where an oligosaccharyltransferase (OST) transfers the glycans onto protein substrates. The alternative pathway involves glycosyltransferases (GTs) sequentially adding nucleotide-activated monosaccharides one-sugar-at-a-time directly onto protein substrates to yield glycans of increasing complexity. Lipid carrier-mediated assembly occurs in general glycosylation systems in C. jejuni,²⁹ Neisseria spp.,⁶⁵ and A. baumanii⁶⁶ and in a dedicated pilin glycosylation system in P. aeruginosa,⁶⁷ whereas direct protein glycosylation pathways are employed for dedicated flagellin and adhesin glycosylation in H. pylori,⁶⁸ C. jejuni,²² and H. influenzae.⁶⁹ Whether one of these models or a blend of the two (initial glycan construction on a lipid carrier followed by transfer and further tailoring on the protein) is an apt descriptor of how general protein glycosylation occurs in H. pylori is unknown.

One of the mutant strains that displayed glycoprotein biosynthesis defects (Δ1518, Figure 2A) is insertionally inactivated in HpG27_1518 (wecA), a gene that initiates O antigen synthesis⁴⁴ for production of fully elaborated lipopolysaccharide (LPS, Figure 3B). Given the known role of this gene in addition of GlcNAc-1-P onto an undecaprenyl phosphate lipid carrier, the protein glycosylation defect observed in Δ1518 provides support for the lipid carrier-mediated assembly pathway. Moreover, this dual role for HpG27_1518 in LPS and glycoprotein biosynthesis suggests that these pathways share common genes in H. pylori.

Figure 3. Glycoprotein and LPS biosynthesis appear to share select biosynthetic genes and the Lewis Y glycan epitope, yet not azide-labeling, indicating an overlapping lipid-carrier mediated assembly pathway that bifurcates to yield glycoproteins and LPS.

A) Analysis of LPS isolated from wild type and mutant strains of H. pylori. LPS was biochemically purified from the indicated strains via a hot aqueous phenol extraction, and isolated LPS was electrophoresed and stained with Pro-Q Emerald LPS stain. The data shown are representative of triplicate experiments. B) Working model of general O-linked protein glycosylation pathway in H. pylori, in which glycosyltransferases (GTs) sequentially transfer nucleotide-activated monosaccharides (shown as hexagons in this schematic) one-sugar-at-a-time onto a lipid carrier (shown as a phosphorylated squiggle) to construct lipid-linked oligosaccharides in the cytoplasm. The lipid-linked oligosaccharides are then flipped across the inner membrane by a flippase into the periplasm, where a pathway bifurcation occurs: either O-antigen ligase transfers the glycan en bloc onto the LPS core to produce fully-elaborated LPS, or an oligosaccharyltransferase (OST) transfers the glycan onto protein substrates to produce glycoproteins. Modifying glycosyltransferases may tailor glycoproteins further. This working model is built on the established role of the gene product of HpG27_1518 (WecA) in O-antigen biosynthesis, coupled to the effect of mutating known LPS biosynthesis genes on glycoprotein biosynthesis. The Lewis Y (Le^y) tetrasaccharide is at the terminus of LPS. C, D) LPS-depleted and LPS-enriched samples were probed for the presence of the Lewis Y antigen (C) or for the presence of detectable azides (D). Wild type (WT) G27 was metabolically labeled with the azide-containing sugar Ac₄GlcNAz for four days. Following metabolic labeling, samples were either lysed directly to yield a whole-cell lysate (control), LPS was enriched (+LPS), or LPS was removed from lysate using the LPS-toxin eraser (−LPS). C) Samples were electrophoresed and probed for total protein by Coomassie blue stain or for the Lewis Y epitope via western blot with anti-Lewis Y antibody. D) Samples were probed via Staudinger ligation with Phos-FLAG, then electrophoresed and probed for total protein with Coomassie blue stain or for the FLAG epitope via western blot with anti-FLAG antibody.

To evaluate the extent of glycoprotein and LPS pathway overlap in H. pylori, we analyzed LPS samples that were isolated from wild type and mutant strains of H. pylori. Briefly, LPS was biochemically purified via a hot aqueous phenol extraction from wild type and mutant strains, treated with proteinase K to remove contaminating proteins, and isolated samples were electrophoresed and visualized using ProQ-Emerald LPS stain. Wild type cells produced a characteristic LPS profile that included a ladder of high molecular weight species >25 kDa, consistent with previous reports (Figure 3A). Eight of the mutant strains (e.g. Δ190, Δ579, Δ580) produced an LPS profile that essentially mirrored the profile observed with wild type LPS (Figure 3A), suggesting that the mutated genes within each of these strains are not involved in LPS biosynthesis. By contrast, seven of the mutant strains, including Δ1518, produced a defective LPS profile with a paucity of high MW species and an over-abundance of low MW species relative to wild type cells. The observed LPS profiles for Δ94, Δ613, Δ785, Δ1230, and Δ1236 are consistent with a recent report by Benghezal and coworkers, which established a role for HpG27_94, HpG27_613, HpG27_785, HpG27_1230, and HpG27_1236 in LPS biosynthesis.⁵⁷ In addition, our data implicate the involvement of HpG27_952 in G27 LPS biosynthesis.

Integrating the glycoprotein profiles and LPS profiles observed in these experiments (Figure 2B, Figure 3), there appears to be some, but not perfect, overlap between defects in glycoprotein biosynthesis and LPS biosynthesis in H. pylori. Seven strains indicated by an asterisk (*) had defects in both glycoprotein biosynthesis and LPS production (Figure 3), suggesting a role for the mutated genes in both pathways. By contrast, four strains – Δ190, Δ579, Δ580 and Δ761 – appeared to have defects in glycoprotein biosynthesis but not in LPS biosynthesis. Indeed, Δ190, Δ579, and Δ580 appear to produce LPS of higher molecular weights relative to wild type cells, indicating a possible enhancement of chain elongation of the O-antigen upon interruption of these genes. These results indicate that HpG27_190, HpG27_579, HpG27_580, and HpG27_761 may be directly involved only in glycoprotein biosynthesis. Taken together, these data provide evidence that, like LPS biosynthesis, glycoprotein biosynthesis appears to involve assembly of a glycan on a lipid carrier. Moreover, there is likely pathway overlap between glycoprotein and LPS biosynthesis in H. pylori, with a bifurcation event at some point and further tailoring of glycan structures after this bifurcation event (see Figure 3B for a proposed model). Insight into the biosynthetic intermediates produced along the pathway coupled to the exact roles of these genes in glycan biosynthesis will be necessary to fully define the glycoprotein biosynthesis pathway.

To query the plausibility of this model, we turned to the established presence of the Lewis Y (Le^y) tetrasaccharide at the terminus of O-antigen in H. pylori.⁷⁰ We wondered if the Le^y glycan epitope is present in both LPS and in glycoproteins or if it is exclusively found in LPS. To investigate these possibilities, we prepared LPS samples from wildtype H. pylori cells, as well as lysate samples from which LPS was removed, and probed the presence of the Le^y tetrasaccharide via Western blot with an anti-Le^y antibody. Western blot analysis revealed robust signal in a whole cell lysate control as well as in the LPS sample (Figure 3C). These findings are consistent with the biosynthesis of Le^y and its presence in LPS. The sample lacking LPS, but containing proteins, also had Le^y signal corresponding to discrete molecular weights (Figure 3C, Figure S3). These data indicate that the Le^y epitope is detectable in both LPS and glycoprotein samples. These results suggest that H. pylori glycoproteins may bear the Le^y tetrasaccharride and provide support for the pathway overlap model for LPS and glycoprotein biosynthesis (Figure 3B).

With evidence that there are structural similarities between glycan epitopes resident within LPS and glycoproteins, we wondered whether the azidosugar GlcNAz is metabolically incorporated into both of these classes of biomolecules. MOE-based detection of azide-labeled glycans revealed robust azide-dependent signal in metabolically labeled whole cell lysates and in samples from which LPS had been removed, but no detectable azide-dependent signal in LPS purified from metabolically labeled cells (Figure 3D). These data suggest that the azidosugar is incorporated into glycoproteins but not into LPS. These findings are consistent with the observation that proteinase K treatment of metabolically labeled whole cell lysates ablated azide-dependent but not Le^y signal (Figure S4). Together, these results support a pathway bifurcation event that occurs after Le^y synthesis and before azidosugar installation. One possibility is that the azidosugar is processed by a modifying glycosyltransferase that tailors glycans on glycoproteins after a pathway bifurcation event (Figure 3B). An alternative possibility is that the incorporation of the azidosugar is preventing labelled subunits from being incorporated into LPS. Structural glycomics will be required for a more incisive model.

H. pylori’s glycosylation mutants display fitness defects

Regardless of how precisely these genes work together to construct H. pylori’s glycoproteins, access to strains with defects in glycoprotein biosynthesis provided an opportunity to probe the effects of altered glycoprotein biosynthesis on strain fitness. To assess functional significance, we turned to a series of phenotypic assays, including growth, adhesion to host cells,⁷¹ motility,¹⁵ and biofilm formation.⁷² These critical functions are absolutely required for H. pylori to colonize the host and sustain an infection.^{73, 74}

We focused our attention on two mutant strains (Δ579 and Δ580) that exhibit defects only in glycoprotein biosynthesis, not in LPS biosynthesis, to mitigate confounding factors. Toward this end, we scored wild type and mutant strains for growth by monitoring OD₆₀₀, adhesion to host cells by scoring percent adhered cells relative to wild type,⁷¹ motility by quantifying halo formation on soft agar,¹⁵ and biofilm formation by measuring crystal violet staining.⁷² Wild type H. pylori grew to stationary phase (Figure 4A), were viable (Figure S5), adhered robustly to gastric adenocarcinoma (AGS) cells (Figure 4B), were motile (Figure 4C), and formed a dense biofilm that stained with crystal violet (Figure 4D). Glycoprotein biosynthesis mutants Δ579 and Δ580 displayed growth and viability comparable to wild type cells (Figure 4A, Figure S5). By contrast, H. pylori mutants Δ579 and Δ580 displayed diminished adhesion to AGS cells, decreased motility, and impaired biofilm formation relative to wild type H. pylori (Figure 4). Genetic disruption of glycoprotein biosynthesis induced functional defects in motility and biofilm formation that recapitulate those caused by chemical inhibition of H. pylori’s general protein glycosylation system.⁴¹ Taken together, these data support the functional significance of H. pylori’s general protein glycosylation system and its potential suitability as a drug target. In the absence of complementation studies, however, we cannot rule out the possibility that the changes in glycosylation and corresponding phenotypic changes observed in these studies could be due to a secondary effect such as phase variation.

Figure 4. Protein glycosylation mutants exhibit robust growth, yet defects in adhesion to host cells, motility, and biofilm formation.

Wild type (WT) and glycoprotein biosynthesis mutants (Δ579, Δ580) were scored for (A) growth by monitoring optical density at 600 nm (OD₆₀₀) over time, (B) percent of cells adhered to host gastric adenocarcinoma cells (AGS) relative to wild type G27 H. pylori in a cell adhesion assay (C) motility on soft agar (left: motility plate on day 3; right: motility quantified over time), and (D) biofilm formation using a crystal violet assay (left: side-view of stained biofilm in triplicate wells of a 96-well plate on day 4; right: quantification of biofilm formation on day 4).

Confirmation of altered cell envelope glycan architecture

Finally, we sought to confirm via an orthogonal method that H. pylori’s glycosylation is disrupted upon insertional inactivation of protein glycosylation genes. Toward this end, we assessed the binding of the carbohydrate-binding lectins concanavalin A (ConA) and Ulex europaeus agglutinin (UEA) to H. pylori. In particular, we assessed lectin binding to wild type cells relative to two mutant strains, Δ579 and Δ580, that exhibited defective glycoprotein biosynthesis but intact LPS biosynthesis. Flow cytometry analysis of intact cells revealed that ConA and UEA bound to wildtype H. pylori (Figure 5). Preincubation of ConA and UEA with their known ligands, mannose and fucose, respectively, mitigated binding to wild type H. pylori, consistent with the carbohydrate dependent nature of these binding interactions (Figure 5). By contrast, Δ579 and Δ580 exhibited enhanced binding to UEA relative to wild type cells (Figure 5). In addition, Δ580 exhibited enhanced binding to ConA relative to wild type cells (Figure 5). Enhanced lectin binding is consistent with altered cell envelope glycan architecture, perhaps by unmasking underlying glycan epitopes or dysregulation of glycosyltransferases in the mutant strains giving rise to hyper/hypo glycosylation. The combination of western blot and flow cytometry data suggest that interruption of HPG27_579 and HpG27_580 precipitate alterations in surface glycan architecture by interfering with glycoprotein biosynthesis. Collectively, these data yield insight into genes involved in proper construction of H. pylori’s cell envelope glycan architecture. Complementation studies will be necessary to confirm these effects are causative, not correlative.

Figure 5. Flow cytometry analyses indicate protein glycosylation mutants display altered cell surface glycans relative to wild type H. pylori.

Wild type (WT) and glycoprotein biosynthesis mutants (Δ579, Δ580) were grown for four days in liquid culture, then probed with ConA or UEA conjugated to Alexa Fluor 488 or FITC, respectively, and analyzed by flow cytometry. Alternatively, ConA and UEA were pre-treated with 400 mM mannose or fucose, respectively (carbo block) prior to probing WT H. pylori. A) Flow cytometry overlays of raw data. B) Histograms of mean fluorescence intensity (MFI) of samples measured by flow cytometry in three independent experiments. Error bars represent the standard deviation of replicate samples.

Discussion

Bacterial glycans are intriguing molecular targets as they contain rare, exclusively bacterial monosaccharides that are absent from human cells and are frequently linked to pathogenesis.^{5, 6, 14} Thus, identifying the genes required for their biosynthesis is a high priority. Such efforts will aid in defining the functions of these molecules and have the potential to guide the development of urgently needed antibiotics. Although systematic efforts to define bacterial glycan biosynthetic pathways have been undertaken, there is a relative paucity of information about recently discovered bacterial protein glycosylation pathways. The progress that has been made in this area has typically required in-depth glycan structure characterization.^{17, 29, 66, 75} However, precise glycomics information is not always available and gathering this information is a non-trivial undertaking. Metabolic reporters have the potential to accelerate the study of bacterial glycan biosynthetic pathways since they report on the output of these pathways without mechanistic information about the underlying pathway as a prerequisite.

Here we explored metabolic glycan labeling as an approach to guide the targeted discovery of genes required for glycan biosynthesis. Western blot analyses indicate that MOE effectively reported on glycoprotein biosynthesis defects in the priority pathogen H. pylori and helped guide the identification of genes required for glycoprotein biosynthesis (Figure 2). Additional LPS analyses revealed that some genes are required for both glycoprotein and O-antigen biosynthesis (Figure 3). Taken together, these data led us to develop a working model for glycoprotein biosynthesis that involves a shared lipid carrier-mediated assembly pathway for both classes of glycans that ultimately bifurcates to produce LPS and glycoproteins with tailored glycans. Complementary evidence for this model arose from the apparent presence of the Le^y tetrasaccharide in both LPS and glycoproteins, yet the presence of azide-labeled glycans only in glycoproteins. Glycoprotein and LPS pathways overlap in P. aeruginosa, with shared glycan epitopes across both classes of biomolecules.^{58, 75} Structural analyses will be required to yield molecular-level information about H. pylori’s glycoprotein biosynthesis pathway. Without glycomics-level information, however, metabolic glycan labeling-based unveiled glycosylation machinery and provided insights about the underlying biosynthetic pathway.

Genetically-induced functional defects in adhesion to host cells, motility, and biofilm formation – three attributes which are absolutely critical for establishment and maintenance of infection within a host – suggest the potential functional importance of H. pylori’s general protein glycosylation machinery and ensuing suitability as a drug target. It is well established that flagellin glycosylation is a virulence factor in H. pylori, as targeted genetic disruption of the dedicated flagellin protein glycosylation system impedes flagella formation, motility, and colonization of the host.¹⁵ In contrast, the link between H. pylori’s general protein glycosylation system and virulence has been elusive, as the general protein glycosylation system is not well characterized. Genetically-induced functional defects provide support for the hypothesis that H. pylori’s general protein glycosylation system may be a virulence factor. These data are consistent with a recent report that small molecule inhibitors of H. pylori’s glycosylation system precipitate defects in motility and biofilm formation.⁴¹ Thus, this approach has yielded insights about the suitability of glycan biosynthesis as a virulence factor and potential drug target.

Our studies probed how interfering with protein glycosylation impacted lectin binding. Genetic inhibition of HpG27_579 and HpG27_580 enhanced binding to UEA, whereas insertional inactivation of HpG27_580 impacted binding to ConA. These enhancements in lectin binding may correspond to the unmasking of accessible glycan epitopes, presumably by altering cell envelope composition. HpG27_579 and HpG27_580 are homologous to H. pylori J99 genes JHP0562 and JHP0563, respectively. JHP0562 is a β−1,3-galactosyltransferase involved in production of type 1 Lewis antigens (Le^a and Le^b),^{76, 77} and JHP0563 is a galactosyltransferase required for the synthesis of both type 1 and type 2 (Le^y) Lewis antigens.^76–78 If HpG27_579 and HpG27_580 were involved in Lewis antigen production, a diminishment in fucose epitopes and a decrease in UEA binding would be expected upon inactivation of these genes. However, here we observed enhanced UEA binding to Δ579 and Δ580 relative to wildtype cells, as well as intact LPS biosynthesis and Lewis Y production in these cells. Therefore, HpG27_579 and HpG27_580 do not appear to be involved in Lewis antigen production. The precise function of these genes is an open question. Enhanced ConA binding to Δ580 but not Δ579 indicates differential roles of the two inactivated genes. How these observations relate to the function of these genes in glycan remodeling events is unknown. These observations reveal unknowns worthy of further exploration, in particular the precise details of glycan composition and assembly within H. pylori.

H. pylori exists in two morphologies, an infective spiral form and a nonculturable, viable metabolizing coccoid form. Khin et al. reported considerable enhancements in ConA binding profiles for H. pylori’s coccoid versus spiral forms, whereas the fucose-binding lectin Lotus-A bound equally to both morphologies.⁷⁹ Their data indicate that the carbohydrate coat is selectively remodeled upon morphological transformation. Our studies did not probe morphology. However, the glycosylation mutants produced in our studies, coupled to glycan analysis methods and morphological studies, could be leveraged to unveil how glycan structure corresponds to morphology, as well as the functional consequences of masking and unmasking particular glycan epitopes.

Finally, a limitation of any metabolic labeling approach is the potential for off-target effects. Recent work by Chen and colleagues demonstrated that peracetylated monosaccharides, such as Ac₄GlcNAz, undergo a nonenzymatically mediated process when used for prolonged periods at high concentrations that leads to the formation of S-glycosylation at cysteine residues.^{80, 81} Thus, the possibility exists that some off-target labeling is present within H. pylori. The substantial diminishment of metabolic labeling upon inactivation of particular glycosytransferase genes, however, indicates that many of these labeled species are glycosylated proteins. Thus, off-target labeling does not appear to dominate in this system.

CONCLUSION

Bacterial glycans and their associated biosynthetic machinery are blockbuster antibiotic targets that are still rife with untapped potential. This work represents the first demonstration, to our knowledge, of using MOE with mutant strains to unveil bacterial glycosylation genes. While this work is focused on one bacterial pathogen, the general approach presented here has the potential to pave the way to yield insights into glycan biosynthesis in diverse species. Broadly, this work introduces a new approach to study glycan biosynthesis in priority pathogens and unveil potential targets that could form the basis of urgently needed antibiotics.

METHODS

Materials.

All reagents and solvents were used as received from Sigma-Aldrich, Fisher Scientific, Invitrogen, New England Biolabs, or Qiagen. Ac₄GlcNAz, Ac₄GlcNAc, and Phos-FLAG (Phos-DYKDDDDK) were prepared as previously described.⁸² DNA manipulation (restriction digests, PCR, and agarose gel electrophoresis) were performed according to standard procedures.

H. pylori growth media and growth conditions.

H. pylori strain G27⁸³ was used for all experiments and was received from Manuel Amieva (Stanford University). Horse blood agar (HBA) plates were prepared with Colombia agar base and 5% (w/v) horse blood, 1% (w/v) vancomycin, 0.5% (w/v) cefulosidan, 0.033% (w/v) polymyxin, 0.5% (w/v) trimethoprim, and 0.8% (w/v) amphotericin B. H. pylori was stored in freezer media (10% fetal bovine serum, 20% glycerol, 70% Brain Heart Infusion broth) at −80°C. H. pylori aliquots were spread onto warmed HBA plates via sterile Q-tip and allowed to grow for three days under incubation of 14% CO₂ at 37°C. Liquid H. pylori cultures were prepared by suspending H. pylori grown on HBA plates in a liquid growth media composed of Brucella broth, 10% fetal bovine serum and 6 μg/mL vancomycin.

Construction of H. pylori mutant strains.

Using a combination of transposon mutagenesis⁶² and targeted insertional inactivation,⁸⁴ an ordered collection of H. pylori G27 mutants containing the Campylobacter coli chloramphenicol acetyl transferase (cat) resistance cassette (CAM^R) were produced, reported,^{62, 63} and shared by Nina Salama (Fred Hutchinson Cancer Research Center). Polymerase Chain Reaction (PCR) amplification of target genes from genomic DNA was used to confirm the presence and size of each insertionally inactivated target gene within the H. pylori mutant strains.

Metabolic labeling.

H. pylori G27 wild type and the constructed mutant strains were grown on horse blood agar plates for 3–4 days in a microaerophilic environment (14% CO₂) at 37 °C. The bacteria were then transferred to Brucella broth containing 10% FBS, 6 μg/mL vancomycin, and 1 mM Ac₄GlcNAz or 1 mM of the azide-free control sugar Ac₄GlcNAc, as previously described.⁶⁴ Liquid cultures were grown for 4 days in a microaerophilic environment with gentle rocking at 37 °C, and then harvested, washed with phosphate buffered saline (PBS), and lysed for subsequent western blot analysis, as described below.

Western blot of azide-labeled glycoproteins.

Lysates from metabolically labeled cells were standardized to a protein concentration of 2.5 mg/mL and reacted 1:1 with 500 μM Phos-FLAG⁸⁵ (for a final concentration of 250 μM) overnight at room temperature for detection of azide-labeled glycans. Reacted lysates were loaded onto a 12% Tris-HCl SDS-PAGE gel, separated by electrophoresis, and transferred to nitrocellulose paper. Anti-FLAG-HRP was employed to visualize FLAG-tagged proteins via chemiluminescence.

Beta-elimination of O-linked glycans.

Lysates from metabolically labeled cells were standardized to a protein concentration of 2.5 mg/mL and reacted 1:1 with 500 μM Phos-FLAG⁸⁵ (for a final concentration of 250 μM) overnight at room temperature for detection of azide-labeled glycans. Reacted lysates were left untreated or subjected to beta-elimination to remove glycans overnight at 4 °C using GlycoProfile Beta-elimination Kit (Sigma-Aldrich) according to manufacturer’s instructions. Treated samples were then loaded onto a 12% Tris-HCl SDS-PAGE gel, separated by electrophoresis, transferred to nitrocellulose paper, and probed with anti-FLAG-HRP.

LPS purification and analysis.

Wild type and mutant strains were grown on large scale (500 mL liquid media) for LPS enrichment experiments. Following growth to confluence in liquid cultures, cells were pelleted by centrifugation, washed in dH₂O, pelleted by centrifugation, and resuspended in 10 mL dH₂O. LPS was then extracted from resuspended pellets using the hot-phenol water extraction method.⁸⁶ Briefly, cell suspensions were heated to 68 °C, mixed with the same volume of hot-phenolic water (90% w/v) at 68 °C, and the mixture was stirred for 30 min at 68 °C prior to centrifugation at 3500 rpm at 10 °C for 45 min. The aqueous phases were collected and this process was repeated. The aqueous phases were pooled and dialyzed against dH2O for 4 days to remove phenol. The dialyzed samples were lyophilized, resuspended in 5 mL dH₂O, and soluble nucleic acids and proteins were removed using ultracentrifugation at 105,000 × g for 4 h at room temperature. The LPS pellets were resuspended in 5 mL dH₂O and lyophilized. LPS samples were treated with Proteinase K (NEB) and heated at 60 °C for 2 h in order to remove any contaminating proteins from the sample. Purified LPS samples were then loaded onto a 15% Tris-HCl SDS-PAGE gel, separated by electrophoresis, and visualized with ProQ-Emerald LPS stain.

LPS enrichment from metabolically labeled H. pylori.

Wild type G27 H. pylori were metabolically labeled on a large scale (500 mL Brucella broth containing 1 mM Ac₄GlcNAz) to perform an LPS enrichment experiment. Following metabolic labeling, cells were harvested and LPS was extracted using the hot-phenol water extraction method described above.⁸⁶ . LPS content in samples was quantified using the ToxinSensor Chromogenic LAL Endotoxin Assay Kit (GenScript) (Figure S3). Enriched LPS was treated with both 2X SDS loading buffer and Proteinase K (NEB) and heated at 60°C for 1 hr in order to remove any contaminating proteins from the sample

LPS removal from metabolically labeled H. pylori.

Wild type G27 H. pylori were metabolically labeled on a large scale (500 mL Brucella broth containing 1 mM Ac₄GlcNAz) and lysed in lysis buffer prior to performing an LPS removal experiment. LPS removal was conducted using the ToxinEraserTM Endotoxin Removal Kit (GenScript) according to manufacturer’s instructions. Briefly, azide labeled HPG27 cell lysate was run through the LPS-binding affinity column, and fractions containing HPG27 LPS-depleted sample were pooled and concentrated using Amicon-Ultra-0.5 centrifuged at 14,000 × g. The pooled and concentrated fractions were assessed for protein concentration using the DC Protein Assay (Biorad). Protein concentrations were standardized to 3 mg/mL. LPS content in samples was quantified using the ToxinSensor Chromogenic LAL Endotoxin Assay Kit (GenScript) (Figure S3).

Detection of Lewis Y antigen and azides in LPS-depleted and LPS-enriched samples.

Following quantification of LPS in enriched and depleted samples (see Figure S3), all samples were reacted 1:1 with 500 μM Phos-FLAG overnight at room temperature for detection of azide-labeled glycans. Reacted lysates were loaded onto a 12% Tris-HCl SDS-PAGE gel, separated by electrophoresis, and transferred to nitrocellulose paper. An anti-Lewis Y antibody (GeneTex) and secondary anti-mouse IgM antibody (HRP) (GeneTex) were used to detect the Lewis Y antigen, whereas an anti-FLAG-HRP was employed to visualize FLAG-tagged proteins via chemiluminescence. Total protein loading was assessed by staining electrophoresed gels with Coomassie Brilliant Blue stain.

Growth curves.

Growth was measured over the course of 4 days. Bacteria were inoculated at a starting OD₆₀₀ of 0.1 into culture tubes containing 3 mL of Brucella broth with appropriate selection antibiotics. Cultures were kept at 37 °C and 14% CO₂ with gentle shaking. The OD₆₀₀ of each culture was measured using spectrophotometry at the indicated timepoints.

Soft agar motility assay.

The ability of H. pylori to swarm was measured over the course of 5 days, though plates were monitored for up to 12 days to ensure no substantial changes. H. pylori strains were standardized to an OD₆₀₀ of 0.4, grown overnight in Brucella Broth, then concentrated and inoculated into freshly prepared Brucella Broth plates containing 0.4% agar and 10% fetal bovine serum. Plates were incubated at 37 C and 14% CO2, and colony diameter was measured daily.

Biofilm formation assay.

Biofilm formation was measured after 4 days of growth in liquid media following a literature protocol.⁷² Briefly, H. pylori strains were standardized to an OD₆₀₀ of 0.4 in Brucella broth, plated in triplicate in the side wells of a 96 well plate, and incubated at 37 °C and 14% CO₂. Following 4 days of growth, media was removed from the wells and the remaining biofilm was stained with 0.15% crystal violet. Sideview images of the triplicate wells were taken after staining. Following photographic documentation, crystal violet-stained biofilms were dissolved in 30% acetic acid and quantified via absorbance at 562 nm via spectrophotometry.

Adhesion assay.

Adhesion assays were performed according to a protocol adapted from Loh et al.⁸⁷ and Kwok et al.⁸⁸ Human gastric adenocarcinoma (AGS) cells were plated into a 24-well tissue culture plate (3.2 × 10⁵ AGS cells/well), adhered overnight, then inoculated with standardized H. pylori samples (OD₆₀₀ ~1.0, 20 μL added per well). The samples were co-cultured at 37 °C and 14% CO₂ for 5 hours, then non-adherent bacteria were rinsed away with PBS rinses and AGS cell membranes were disrupted with 0.1% saponin (in PBS) for 20 minutes at 37 °C. Serial dilutions of resultant suspensions were plated on horse blood agar plates containing appropriate selection antibiotics for 5 days under microaerophilic conditions, then colony forming units (CFUs) were enumerated. Brain heart infusion medium alone was used as a negative growth control in the adhesion assay. Serial dilutions of the starting inoculum of H. pylori strains were plated on horse blood agar plates, with appropriate selection antibiotics, to ensure that no significant differences in CFUs were observed across strains.

Lectin binding.

H. pylori wild type and mutant strains were probed with Alexa Fluor 488-conjugated Concanavalin A (ConA; Life Technologies) or FITC-conjugated Ulex europaeus agglutinin (UEA; Vector Laboratories), then analyzed by flow cytometry on a BD Accuri C6 Plus (BD Biosciences) instrument. As a negative control, ConA and UEA were pre-incubated (carbo block) with 400 mM mannose and fucose, respectively, prior to binding to wild type H. pylori.

ABBREVIATIONS

LPS

lipopolysaccharide

CPS

capsular polysaccharide

WT

wild type

GT

glycosyltransferase

MOE

metabolic oligosaccharide engineering

CAZY

carbohydrate active enzymes

ORF

open reading frame

CAM^R

chloramphenicol acetyl transferase marker

PCR

polymerase chain reaction

Ac₄GlcNAz

peracetylated N-azidoacetylglucosamine

Ac₄GlcNAc

peracetylated N-acetylglucosamine

OST

oligosaccharyltransferase

Le^y

Lewis Y

AGS

gastric adenocarcinoma

ConA

concanavalin A

UEA

Ulex europaeus agglutinin

HBA

horse blood agar

PBS

phosphate buffered saline

OD

optical density

CFU

colony forming unit