Synthesis and Application of Rare Deoxy Amino l-Sugar Analogs to Probe Glycans in Pathogenic Bacteria

Abstract

Bacterial cell envelope glycans are compelling antibiotic targets as they are critical for strain fitness and pathogenesis yet are virtually absent from human cells. However, systematic study and perturbation of bacterial glycans remains challenging due to their utilization of rare deoxy amino l-sugars, which impede traditional glycan analysis and are not readily available from natural sources. The development of chemical tools to study bacterial glycans is a crucial step toward understanding and altering these biomolecules. Here we report an expedient methodology to access azide-containing analogs of a variety of unusual deoxy amino l-sugars starting from readily available l-rhamnose and l-fucose. Azide-containing l-sugar analogs facilitated metabolic profiling of bacterial glycans in a range of Gram-negative bacteria and revealed differential utilization of l-sugars in symbiotic versus pathogenic bacteria. Further application of these probes will refine our knowledge of the glycan repertoire in diverse bacteria and aid in the design of novel antibiotics.

Graphical Abstract

Bacteria coat themselves in a dense array of carbohydrate structures that are essential for their fitness and survival.¹ For example, the peptidoglycan provides mechanical stability,² lipopolysaccharide (LPS) and capsular polysaccharide (CPS) provide structural integrity and protect bacteria from external insults,^3–5 and glycoproteins mediate binding to host cells⁶ and evade immune recognition⁷ (Fig. 1A). As a testament to their functional importance, bacterial glycans have served as targets of blockbuster antibiotics^8–11 and highly effective vaccines.^12–14 Thus, taking stock of which glycan structures are present on which bacterial cells reveals fundamental insight into the role of these structures, as well as aids in antibiotic and vaccine development.

Figure 1.

Sampling of exclusively bacterial (A) glycans, (B) monosaccharides, and (C) glycoconjugates synthesized by Gram-negative pathogens.

Comprehensively surveying bacterial glycans is challenging due to their structural complexity.¹⁵ Although there are conserved components in bacterial glycans, they contain variable regions that differ dramatically between and within bacterial species.^{16, 17} The differences in these glycan epitopes often track with pathogenesis. For example, variation in LPS and CPS triggers differential host immune responses, with particular serotypes correlated to strain virulence.^18–20 Structural diversity of bacterial glycans arises from differential utilization of over 700 monosaccharides in bacteria.^21–23 These structures include rare, deoxy amino sugars that are absolutely absent from human cells.²⁴ For example, di-N-acetyl d-bacillosamine (d-Bac),²⁵ d-2,4-diacetamido-2,4,6-trideoxy galactose (d-DATDG), N-acetyl l-fucosamine (l-FucNAc),^{26, 27} N-acetyl l-pneumosamine (l-PneNAc),²⁸ N-acetyl-l-quinovosamine (l-QuiNAc),²⁹ and N-acetyl l-rhamnosamine (l-RhaNAc)²⁹ are a small sampling of exclusively bacterial d- and l-sugars (Fig. 1B) incorporated into higher-order glycans (Fig 1C). The presence of these monosaccharides on bacterial cells is rare, highly variable, and strain and serotype specific. These sugars have been described in serotypes of the life-threatening and emerging pathogens including Staphylococcus aureus,³⁰ Pseudomonas aeruginosa,²⁷ Vibrio vulnificus,²⁹ and Plesiomonas shigelloides²⁸, for example (Fig. 1C). However, there are limited reports of the distribution of these sugars across bacteria, largely due to challenges associated with detecting, isolating, and studying them.³¹

Metabolic oligosaccharide engineering (MOE) has emerged as a powerful method for the efficient investigation of bacterial glycans.^{31, 32} Originally pioneered by Bertozzi,^{33, 34} Reutter and colleagues^{35, 36} for studying glycans in mammalian systems, MOE is a two-step chemical approach that has been adapted to probe the bacterial glycome (Fig. 2A).^{32, 37} Briefly, bacteria treated with azide-containing sugar analogs remodel their glycocalyx to yield metabolically labeled cellular glycans. Then, bioorthogonal chemistry^{38, 39} is used to produce detectable signals that aid the tracking, enrichment, and characterization of bacterial glycans. Early studies in bacteria with peracetylated azide-containing analogs of common monosaccharides, including N-azidoacetylglucosamine (GlcNAc),^{40, 41} N-acetylgalactosamine (GalNAc),⁴² and fucose⁴³, enabled the investigation of glycans in select bacteria. However, these analogs suffered from limited utility in probing rare bacterial glycan structures in a broader subset of bacteria.⁴⁴ Thus, we and others developed azide-bearing analogs of rare exclusively bacterial sugars, including d-Bac, pseudaminic acid, N-acetyl muramic acid, and trehalose.^44–48 Access to rare sugar analogs opened the door to studying glycans in a broader range of species,⁴⁸ probing glycans in complex microbial communities,⁴⁹ labeling glycans in a bacteria-selective manner,⁴⁴ providing biosynthetic insights,⁶ and inspiring the development of metabolic glycan inhibitors.⁵⁰ These studies have validated MOE, paired with intentional choice of unnatural sugars, as a powerful strategy to examine bacterial glycans even in the absence of full structural information. Thus, MOE complements traditional glycan analysis tools that are stymied by rare bacterial glycan building blocks.^{51, 52}

Figure 2.

Schematic of metabolic glycan labelling and the precedented and novel azide-containing monosaccharide analogs used in this study.

Although bacteria utilize over 700 monosaccharides in glycan biosynthesis,^21–23 only a small percentage of these sugars have been chemically modified into metabolic probes.³¹ Thus, the knowledge learned from existing probes does not adequately capture the enormous amount of glycan epitope variability across the bacterial domain. Herein we report an expedient route to access azide-containing analogs of the rare deoxy amino bacterial sugars l-FucNAc, l-PneNAc, l-RhaNAc and l-QuiNAc (Fig. 2C; compounds 5–8) and their evaluation as metabolic labeling agents for detection of these glycan epitopes in diverse bacteria. As described below, l-sugar analogs were incorporated in select bacteria, including some already known to express these monosaccharides. Moreover, these analogs were not efficiently used as metabolic substrates in the symbiotic intestinal bacteria Bacteroides fragilis and a gastric adenocarcinoma human cell line, both of which lack these monosaccharide scaffolds. The narrow incorporation of these analogs is indicative of the narrow distribution of these sugars. Knowledge about species-selective incorporation of sugar analogs sets the stage for further probing and characterizing glycans in these species. This work opens the door to unravelling structure-function relationships of these unusual sugars, designing monosaccharide-based interference agents, and developing new glycosylation-based strategies to eradicate pathogenic infections.

RESULTS AND DISCUSSION

Design of metabolic probes to study bacterial l-sugars

Motivated by the successful deployment of metabolic probes to study bacterial glycans yet the challenges with comprehensively surveying these structures, we sought to develop novel tools to profile the large pool of unusual glycan structures that are not efficiently probed with existing tools. We focused on derivatives of the rare bacterial monosaccharides l-FucNAc, l-PneNAc, l-RhaNAc and l-QuiNAc due to expedient syntheses of these scaffolds and their known utilization by select bacterial pathogens.⁵³ As a first design element based on successful precedents using monosaccharide analogs with N-azidoacetyl substituents in place of the natural N-acetyl moieties to probe bacterial glycans (Fig. 2B, 2–4),^{32, 44} we reasoned that the requisite bacterial carbohydrate biosynthetic enzymes would likely process analogs of l-sugars bearing N-azidoacetyl groups. As a second design element, we chose to employ peracetylated analogs, as we and others have demonstrated transient masking of hydrophilic hydroxyl groups on monosaccharide analogs with hydrophobic acetyl groups facilitates uptake and metabolic labeling in some bacteria.^{32, 40, 42, 44} Thus, we designed a panel of azide-containing analogs of rare bacterial l-sugars (Fig. 2C, 5–8) that embody these two design elements. In addition to these novel compounds, we included precedented azide-containing analogs of the monosaccharide scaffolds GlcNAc,⁶ GalNAc,⁴² and d-Bac⁴⁴ (Fig. 2B, 2–4) in our studies due to reports of their robust incorporation into bacterial glycans. We reasoned that these metabolic probes had the potential to act as competent metabolic substrates and yield insight into bacterial glycans in strains that utilize l-sugars.

Synthesis of metabolic probes of bacterial l-sugars

We synthesized azide-containing analogs 5–8 by adaptation of our published protocols for synthesis of l-FucNAc, l-PneNAc, l-RhaNAc and l-QuiNAc via nucleophilic displacements of l-rhamnose or l-fucose derived triflates.⁵³ Mixtures of anomers are suitable for biological experiments and were therefore employed without further purification.

Synthesis of l-FucNAz 5 starting from l-rhamnosyl 2,4-diol 9 is shown in Scheme 1. Diol 9 was transformed into l-fucosamine derivative 10 by C2,C4 inversion by employing a slight modification of our reported⁵¹ one-pot procedure. It should be noted that using a cheaper sodium nitrite (NaNO₂) as a nucleophile in place of originally employed and expensive tetrabutyl ammonium nitrite (TBANO₂) resulted in much improved yield of 10. Accordingly, compound 9 was treated with triflic anhydride (Tf₂O) and pyridine in CH₂Cl₂ to obtain the corresponding 2,4-bis-triflate, which, upon further treatment with a stoichiometric amount of tetrabutyl ammonium azide (TBAN₃) in acetonitrile at 0 °C for 1 h, underwent a facile, regioselective displacement of the C2-OTf by azide nucleophile. Subsequent addition of 3.0 equiv of NaNO₂ in the same pot displaced the remaining C4-OTf to afford the l-fucosamine derivative 10 in 70% yield over 3 steps after a single column chromatographic purification. Next, azide group in 10 was reduced by zinc in acetic acid and subsequent EDC mediated azidoacetic acid coupling⁵⁴ with the so formed amine furnished the amide derivative 11 in 52% over two steps. Debenzoylation of compound 11 using Na-OMe and subsequent acetylation with acetic anhydride furnished compound 12 in 77% yield over two steps. Finally, hydrolysis of the thioglycoside by using NBS and water generated the corresponding anomeric hemiacetal, which upon subsequent treatment with acetic anhydride in the presence of DMAP in THF afforded triacetate l-FucNAz 5 (α/β = 10:1) in 89% yield over two steps.

Scheme 1.

Synthesis of l-FucNAz 5.

To synthesize l–pneumosamine derivative 6, we started from the known l-fucosyl 2,4-diol 13⁵¹ (Scheme 2). Regioselective triflation of the more reactive equatorial C2-OH group of 13 by employing controlled amount of triflic anhydride generated the corresponding C2-OTf, which upon subsequent refluxing with sodium azide at 110 °C in DMF followed by concomitant removal of benzoyl groups provided l-pneumosamine 3,4 diol derivative 14 in 71% yield over three steps. The azido group in compound 14 was transformed into corresponding amine under hydrogenation condition using Pd(OH)₂/C and further coupling with azidoacetic acid afforded amide compound 15 in 69% over two steps. Acetylation of the free hydroxyl groups in 15 gave diacetylated azido compound 16 in 97% yield. Finally, the anomeric OPMP group was hydrolyzed by using CAN/H₂O followed by acetylation to furnish our target molecule l-PneuNAz 6 in 87% yield (α/β = 10:1).

Scheme 2.

Synthesis of l-PneNAz 6.

Synthesis of l-rhamnosamine derivative 7 was started from the same l-fucosyl 2,4-diol 13 (Scheme 3). First, the C2-OH of diol 13 was regioselectively protected with TBSCl and imidazole to obtain C2 silylated compound 17. Remaining C4-OH was triflated and subsequently inverted with NaNO₂ to generate the corresponding l-quinovose derivative, which upon benzoylation and subsequent removal of TBS group furnished compound 18 in excellent overall yield. Triflation of C2-OH in 18 and refluxing in DMF at 110 °C with sodium azide furnished l-rhamnosamine derivative 19 in 70% yield over two steps. When compound 19 was subjected to azide reduction, under hydrogenolysis conditions or with Zn, AcOH conditions, the desired amine 20 could not be isolated. Due to the syn orientation of the formed amine and C3-OBz ester group in 20, concomitantly the C3-O-benzoate group migrated to the more reactive nitrogen center to form 21 under the prevailing reaction conditions. Alternatively, removal of benzoate groups followed by hydrogenation using H₂, Pd(OH)₂/C to form the amine, and its subsequent EDC mediated coupling with azidoacetic acid led to the desired azido 3,4-diol compound 22 in 71% yield over three steps. Free hydroxy groups were acetylated with acetic anhydride, DMAP in THF to obtain 23 in 88% yield. Anomeric p-methoxyphenyl group was oxidatively hydrolyzed with ceric ammonium nitrate (CAN) in CH₃CN/H₂O and the so formed hemiacetal was immediately capped with acetate to form triacetate compound l-RhaNAz 7 in 95% yield (α/β = 3:1).

Scheme 3.

Synthesis of l-RhaNAz 7.

To synthesize l-quinovosamine derivative 8, we started from the known 4-OBz 2,3-diol l-rhamnosyl-β-thioglycoside derivative 24⁵⁵ (Scheme 4). A highly regioselective O3 benzoylation of 24 using Me₂SnCl₂ and BzCl in 91% yield, followed by C-2 triflation and concomitant inversion with sodium azide afforded l-quinovosamine derivative 25 in 63% yield over two steps. Reduction of azide using zinc and acetic acid in THF to obtain corresponding amine in 80% yield followed by its concomitant EDC coupling with azidoacetic acid generated azido compound 26 in 72% yield. Debenzoylation and subsequent acetylation with acetic anhydride delivered diacetylated thioglycoside compound 27 in 88% yield over two steps. Thioglycoside was converted to corresponding hemiacetal by NBS/H₂O in 88% and was capped with acetate to obtain target molecule l-QuiNAz 8 in 81% yield.

Scheme 4.

Synthesis of l-QuiNAz 8.

A variety of reducing conditions were employed for reduction of azides in these syntheses. l-fucosamine (10) and l-quinovosamine (25) are the two comparatively less polar azides which under Zn/AcOH reduction condition gave the corresponding amines in satisfactory yields after easy column purification (using 45% and 40 % ethyl acetate: pet ether, respectively). Whereas, l-pneumosamine (14) and the corresponding l-rhamnosamine azide compounds being 3,4-diols are quite polar and their corresponding amines were difficult to purify on silica gel column. Therefore, in these cases hydrogenation was used for reduction of azides instead of Zn/AcOH to circumvent column purification and proceed directly to the next step (amide formation reaction) after filtration. Hydrogenation conditions could also be applied for the l-fucosamine (10) and l-quinovosamine (26), if desired.

Metabolic probes are incorporated into select l-sugar bacteria

To ascertain whether the newly developed azide-containing analogs of rare bacterial l-sugars (Fig. 2C, 5–8) are effective metabolic substrates, we evaluated their metabolic incorporation into glycans in bacteria reported to express l-monosaccharides. In particular, we began our studies in P. shigelloides, V. vulnificus, and S. aureus, since P. shigelloides²⁸ serotype O1 bears l-FucNAc in its LPS, Vibrio vulnificus²⁹ M06–24 and BO62316 CPS contains l-QuiNAc and l-RhaNAc, and S. aureus³⁰ type 5 utilizes l-FucNAc in its CPS (Fig. 1C). Thus, we predicted l-sugar analogs would be incorporated in these bacteria in a manner consistent with literature reports of their distribution. We included established d-azidosugars^{56, 57} (Fig. 2B, 2–4) in our studies due to their potential to be incorporated into glycans in these bacteria.

Metabolic labeling experiments were performed following our published methods, with bacteria incubated with the azide-free sugar peracetylated N-acetylglucosamine (Ac₄GlcNAc, 1, Fig. 2B)⁵⁷ as a negative control or with one of the azide-containing l-sugar analogs (l-FucNAz (5), l-PneNAz (6), l-RhaNAz (7), l-QuiNAz (8)) or established d-sugar analogs (N-azidoacetylglucosamine (Ac₄GlcNAz, 2),⁵⁷ N-azidoacetylgalactosamine (Ac₄GalNAz, 3),⁵⁸ di-N-azidoacetylbacillosamine (Ac₂Bac-diNAz, 4)⁴⁴ as experimental samples. After three days of metabolic labeling, lysates were harvested from treated cells, reacted via Staudinger ligation with a phosphine probe comprising a FLAG peptide (Phos-FLAG),⁵⁹ and analyzed via western blot with anti-FLAG antibody. Control samples from cells treated with the azide-free compound A₄GlcNAc (1) yielded signal in each bacterial species that benchmarks the least signal that we would expect to observe in any given sample from that bacterium (Fig. 3A). Bands present in the negative control sample (e.g., 43 kD in P. shigelloides and 80 kD in S. aureus) reflect proteins within the tested bacteria that the anti-FLAG antibody binds to in an azide-independent manner. Of the previously reported azidosugar analogs, we observed modest to robust incorporation of Ac₄GlcNAz (2), Ac₄GalNAz (3), and Ac₂Bac-diNAz (4) into P. shigelloides and V. vulnificus, with Ac₄GlcNAz yielding a pronounced ladder of labeled glycans in P. shigelloides and V. vulnificus (Fig. 3A). Metabolic incorporation of l-sugar analogs was subtle yet detectable relative to the negative control, with all l-azidosugar analogs leading to labeling of glycans in P. shigelloides and V. vulnificus (Fig. 3A). Of note, l-FucNAz (5), l-PneNAz (6), l-RhaNAz (7), and l-QuiNAz (8) led to subtle labeling of glycans between 50–130 kDa in P. shigelloides and V. vulnificus (Fig. 3A). In contrast, western blot analysis did not reveal detectable incorporation of any established analogs or l-sugar probes in S. aureus (Fig. 3A; Fig. S1). Coomassie staining of electrophoresed samples confirmed that all samples from each bacteria contain protein (Fig. S1), indicating that variation in intensity of azide-dependent signal was not due to large differences in protein loading. Taken together, these findings indicate that choice of probe appeared to influence the extent and pattern of glycan labeling in a bacteria-selective manner, with l-sugar probes gaining access to a subset of glycans.

Figure 3.

Metabolic probes are incorporated in a bacteria-selective manner in some bacteria that express l-sugars. Metabolic probes 1–8 were screened for their ability to be processed into glycans in P. shigelloides, V. vulnificus, and S. aureus. Bacteria were grown for three days in media supplemented with 1 mM of the azide-free control sugar Ac₄GlcNAc (1, Ac), or with 1 mM of Ac₄GlcNAz (2, Az), Ac₄GalNAz (3, Gal), Ac₂Bac-diNAz (4, Bac), l-FucNAz (5, Fuc), l-PneNAz (6, Pne), l-RhaNAz (7, Rha), or l-QuiNAz (8, Qui), then probed for the presence of azide-labeled glycans within lysates or on cells. A) Lysates from treated cells were reacted with 250 μM Phos-FLAG for 12 h at room temperature and subsequently detected with anti-FLAG antibody via western blot analysis. A replicate S. aureus western blot revealed a uniform background band at 80 kD in all l-sugar samples (Fig. S1). B) Treated cells were reacted with 20 μM AF488-DBCO for 1 h at 37 °C and fluorescence intensity of each cell population was measured by flow cytometry analysis. The data shown are representative of replicates (n > 2).

As a complementary means to detect metabolic incorporation of analogs into glycans, we probed cell surface azide expression using a flow-cytometry based assay.⁴² Following metabolic labeling with analogs as described above, surface accessible azides were probed on live cells using strain-promoted azide-alkyne cycloaddition with fluorescent Alexa Fluor 488-dibenzocyclooctyne (AF488-DBCO).^{42, 50} Treatment with Ac₄GlcNAc (1) followed by AF488-DBCO served as a negative control that captured background due to non-specific association of the DBCO reagent. Some of the azidosugar treatments led to cells that exhibited a fluorescence intensity comparable to the negative control, corresponding to minimal incorporation of that azidosugar into surface glycans. Indeed, all azidosugar treatments in P. shigelloides and S. aureus were comparable to the Ac₄GlcNAc-treated negative control, indicating no detectable surface-accessible azides on these cells (Fig. 3B; Fig. S1). By contrast, treatment with azide-bearing analogs in V. vulnificus led to a subtle yet significant enhancement in the fluorescence intensity relative to azide-free controls, suggesting the presence of azide-labeled glycans on these cell surfaces (Fig. 3B; Fig. S1).

Taken together, the results of these studies indicate the ability of select bacteria to metabolically process l-sugar probes. Integration of western blot and flow cytometry results has the potential to yield insight into the accessibility of glycan epitopes on the cell envelope. Metabolic incorporation detected by both western blot and flow cytometry, as in V. vulnificus, likely corresponds to azides present in peripheral surface epitopes. In contrast, azides detected by western blot but not flow cytometry, as observed in P. shigelloides, likely correspond to probe incorporation into intracellular or otherwise surface inaccessible epitopes. Finally, lack of detectable azides by western blot and flow cytometry experiments, as observed in S. aureus, suggests no appreciable metabolic incorporation into glycans. These two analyses, in tandem, may prove particularly fruitful for probing the presence and accessibility of rare l-sugar epitopes on bacteria.

l-sugar probes are utilized in additional pathogens, but not by symbiotic Bacteroides fragilis or human cells

Once we established that l-sugar analogs are incorporated into glycans on select bacteria that express these epitopes, we next explored whether l-sugar probes are incorporated into bacteria for which there are no reports of l-FucNAc, l-PneNAc, l-RhaNAc and l-QuiNAc. Such information is critical to survey monosaccharide utilization and has the potential to reveal a more comprehensive picture of glycan epitope expression. For these studies, we turned to the pathogens Helicobacter pylori and Campylobacter jejuni, as well as the commensal Bacteroides fragilis, as these bacteria are amenable to metabolic glycan labeling.⁴⁴ Consistent with literature reports,⁴⁴ surface glycans were robustly azide-labeled upon supplementation of H. pylori with Ac₄GlcNAz (2) and Ac₂Bac-diNAz (4), C. jejuni with Ac₂Bac-diNAz (4), and B. fragilis with Ac₄GalNAz (3). H. pylori used l-sugar analogs 5–8 as metabolic substrates and incorporated these analogs into discrete species (Fig. 4A) that correspond to surface-accessible glycans (Fig. 4B). Supplementation of C. jejuni with the panel of l-sugar analogs 5–8 led to no apparent detection of azide-labeled glycans by western blot relative to azide-free controls (Fig. 4; Fig. S2), yet low but significant incorporation of l-RhaNAz (7) and l-QuiNAz (8) observed by flow cytometry (Fig. 4; Fig. S2). By contrast, supplementation of B. fragilis with the panel of l-sugar analogs 5–8 led to no detection of azide-labeled glycans by western blot relative to azide-free controls (Fig. 4; Fig. S2), despite high protein content in these samples (Fig. S2). These data indicate that although it is possible to incorporate azides into glycans within B. fragilis, incorporation is negligible with the l-sugar analogs reported here. Moreover, l-sugar probes led to some unexpected utilization in H. pylori and C. jejuni, suggesting our current understanding of glycan epitope expression is incomplete. These data further support the relatively rare distribution of these monosaccharides and suggest the potential for glycan-based interference of l-sugar-expressing bacteria in a highly selective manner.

Figure 4.

l-sugar analogs are not efficiently used as metabolic substrates in some bacterial and human cells that lack these monosaccharide scaffolds. Metabolic probes 1–8 were screened for their ability to be processed into glycans in H. pylori, C. jejuni, B. fragilis, and human gastric adenocarcinoma (AGS) cells. Bacteria were grown for three days in media supplemented with 1 mM of the azide-free control sugar Ac₄GlcNAc (1, Ac), or with 1 mM of Ac₄GlcNAz (2, Az), Ac₄GalNAz (3, Gal), Ac₂Bac-diNAz (4, Bac), l-FucNAz (5, Fuc), l-PneNAz (6, Pne), l-RhaNAz (7, Rha), or l-QuiNAz (8, Qui), then probed for the presence of azide-labeled glycans within lysates or on cells. AGS cells were incubated with 10 μM analogs 1–8 for three days. A) Lysates from treated cells were reacted with 250 μM Phos-FLAG for 12 h at room temperature and subsequently detected with anti-FLAG antibody via western blot analysis. B) Treated cells were reacted with 20 μM AF488-DBCO and fluorescence intensity of each cell population was measured by flow cytometry analysis. The data shown are representative of replicates (n > 2).

If these epitopes are to serve as antibiotic or vaccine targets, their absence from human cells is critical. Thus, we set out to evaluate the incorporation of these analogs in the well-studied human gastric adenocarcinoma (AGS) cell line, a cell line that is an established model for epithelial cells that line the gastrointestinal tract.⁶⁰ Previous reports indicate that 10 μM monosaccharide analog is standard for metabolic labeling in mammalian cells,^{33, 34} presumably due to efficient uptake across the cell membrane, so we used this concentration in AGS cell experiments. Treatment of AGS cells with the positive control sugar Ac₄GalNAz (3) led to an array of azide-labeled glycans in lysates and on cells (Fig. 4), consistent with incorporation of this substrate into cell surface mucin-type O-linked glycoproteins.⁵⁸ In contrast, none of the novel l-sugar probes 5–8 resulted in appreciable display of azide-labeled glycans in AGS lysates or on AGS cells (Fig. 4). Instead, AGS cells treated with 5–8 exhibited cellular fluorescence on par with cells treated with the azide-free negative control sugar Ac₄GlcNAc (1). These results are consistent with the negligible incorporation of bacterial sugars by these human cells and bode well for the potential of these sugars to form the basis of selective antibiotic therapies or vaccines.

Discussion

Traditionally, the study of bacterial glycans has been extremely challenging due to their complex, branched, and often-heterogeneous structures that contain rare monosaccharides refractory to glycomics analyses.^{14, 15, 61} Given their important roles in mediating host-pathogen interactions and their potential to form the basis of selective perturbation strategies, expanding the repertoire of tools to probe their structures is of critical importance.^{31, 62} New probes will aid in unravelling bacterial glycan structure-function relationships and have the potential to guide the development of urgently needed antibiotics and vaccines. Complementary approaches, including the use of carbohydrate-binding proteins^63–65 and biosynthetically installed probes,^{31, 66, 67} have greatly eased the study of glycans on live bacterial cells. However, given the profound structural complexity and diversity of bacterial glycans, there is a paucity of probes well suited to their study. Inspired by successful precedents from Grimes,⁴⁵ Swarts,⁴⁷ Vauzeilles,⁶⁸ Logan and Tanner,⁶⁹ and others,^{46, 48} along with our own successes,⁴⁴ of using rare bacterial monosaccharide analogs to metabolically label and study bacterial glycans, we produced a series of azide-containing analogs of rare bacterial l-monosaccharides to build upon the existing toolkit and gain insight into these seemingly rare epitopes.

Here we relied upon metabolic oligosaccharide engineering as an approach to report on the distribution of l-FucNAc, l-PneNAc, l-RhaNAc and l-QuiNAc in glycans across a suite of bacterial species. Our data indicate that l-sugars are utilized in a bacteria-selective fashion in bacterial strains reported to express l-sugars as well as in H. pylori and C. jejuni, yet they are minimally processed by symbiotic B. fragilis and human cells that lack these monosaccharides. Moreover, choice of probe appeared to influence the extent and pattern of glycan labeling in a bacteria-selective manner. Treatment with select l-sugars led to modest labeling of discrete glycans, in contrast to ladders of labeled glycans observed with some previously reported analogs. For instance, treatment of P. shigelloides with l-PneNAz (6) led to a robustly labeled species at ~50 kDa, whereas treatment with Ac₄GlcNAz (2) led to an array of labeled species at a variety of molecular weights (Fig. 3A). These observations may reflect distinct utilization of monosaccharide probes for the construction of different glycan classes and epitopes. In other instances, the fingerprint of labeled molecules looked very similar for the l-sugar probes; for example, the profiles of P. shigelloides, V. vulnificus, and H. pylori appear similar for many of the l-sugar treatments and, to a lesser extent, share patterns with the Ac₂Bac-diNAz treatments. This observation may reflect the presence of complex glycans bearing a combination of d- and l-sugars (Fig. 1C).

Narrow incorporation of l-sugar analogs into discrete glycans within select bacteria is consistent with the apparent rare distribution of these epitopes. We note that while we explored metabolic incorporation of l-sugar probes into bacterial species reported to express these epitopes, the specific strains and serotypes used in these studies were not matched to reported strains due to lack of access. Thus, it is difficult to precisely match patterns of incorporation in the strains screened here with literature reports. In some instances, metabolic incorporation of l-sugars was in line with previous reports. For example, western blot data hinted at subtle processing of l-FucNAz (5) and l-PneNAz (6) into glycans in P. shigelloides, consistent with the presence of these epitopes in P. shigelloides serotype O1 (Fig. 1C).²⁸ Moreover, subtle yet detectable incorporation of l-FucNAz (5), l-RhaNAz (7) and l-QuiNAz (8) into V. vulnificus surface glycans matched reports of these monosaccharides in V. vulnificus M06–24 and BO62316.^{29, 70} Further, the reported d-Ac₂Bac-diNAz (4) probe was robustly incorporated into P. shigelloides, in line with reports of bacillosamine in this bacteria.²⁸ We noted some surprises, as well. V. vulnificus utilized d-Ac₂Bac-diNAz as a metabolic substrate, though we found no literature reports of bacillosamine in this bacterium. Likewise, H. pylori and C. jejuni processed l-sugar analogs into surface epitopes, yet these monosaccharide scaffolds have not been observed in these bacteria before. Taken together, our results suggest that there is more to learn about the distribution of rare bacterial sugars. Metabolic glycan labeling with this expanded set of probes offers inroads to studying distinctive glycans in these bacteria.

In all bacteria screened except for S. aureus, azidosugar substrates were appreciably incorporated into cellular glycans. These findings set the stage to use the azide as a chemical handle to enrich labeled species, ultimately enabling glycomics analyses that yield structural information. Indeed, Woo and coworkers have reported isotopic recoding approaches to tag, enrich, and characterize azide-labeled glycans from a variety of cell types.^71–73 Further, these findings open the door to identifying glycosylation genes and assessing functional roles, as recently described by Moulton et al. in an MOE-based approach.⁶ Moreover, access to these probes provides a means to monitor bacterial glycan dynamics and take stock of rare glycan epitope expression in more complex microbial communities.⁷⁴ A metabolic labeling-based approach to pan glycan structures in gut microbiota reported by Chen and colleagues⁴⁹ could offer a means to study l-sugar distribution more broadly. Expanding the scope of metabolic glycan labeling probes is an important step toward capturing the enormous amount of glycan epitope variability across the bacterial domain.

Given the absence of l-sugars from human cells and their variability across bacteria, these structures have the potential to form the basis of selective antibiotic therapies or vaccines. Differential incorporation of metabolic substrates across bacterial strains opens the door to targeting azide-covered bacteria with damage-inducing agents.⁸ For example, azide-coated bacteria could be reacted via strain-promoted azide-alkyne cycloaddition with cyclooctyne-based photosensitizers to induce photothermal lysis⁷⁵ or via Staudinger ligation with phosphine-based immune stimulants to trigger selective cell death.⁷⁶ As demonstrated via Kasper and coworkers, imaging agents delivered via click-chemistry could be employed to track bacterial glycans in an animal infection model.^{42, 74} Finally, substrate-based monosaccharide analog inhibitors⁵⁰ could be developed based on l-monosaccharide scaffolds to interfere with proper construction of the glycocalyx and disarm bacteria by weakening their cell envelope. An impediment to covalent targeting of azide-labeled rare bacterial monosaccharide epitopes is the high concentration of monosaccharide analogs required for even modest levels of labeling. Thus, tailoring of analogs to enhance the efficiency of uptake and utilization by bacteria is an area of need.

The probes designed in this study employed acetyl groups to transiently mask hydrophilic hydroxyls and ease analog uptake by bacterial cells. However, esterase activity to remove acetyl protecting groups from monosaccharides is not uniformly present across bacterial cells.^77–79 Inefficient deacetylation of peracetylated analogs to yield the corresponding free sugar analog within cells could account for the lack of incorporation in some bacteria. Indeed, our studies with peracetylated GlcNAz (2) revealed no detectable metabolic labeling in S. aureus (Fig. 4), yet a report by Memmel et al.⁸⁰ demonstrated that S. aureus is efficiently labeled with the free sugar GlcNAz. Thus, low or no deacetylation of peracetylated analogs could account for the lack of incorporation observed in S. aureus. The presence of esterase activity in the bacteria of interest is critical for the utility of the probes developed in this study.

A limitation of any metabolic labeling approach is the potential for off-target effects. Recent work by Chen and coworkers in mammalian cells demonstrated the potential for peracetylated monosaccharides, such as Ac₄GlcNAz (2), to undergo a nonenzymatically mediated process when used for long periods at high concentrations that leads to S-glycosylation of cysteine residues.^{81, 82} Thus, the possibility exists that some off-target labeling of intracellular proteins is occurring within the bacteria studied here, which may be overcome by use of free sugar analogs.⁷⁶ Despite the possibility of off-target effects, Pratt, Woo and coworkers concluded these probes are generally reliable and yield valuable insights into physiologically-relevant glycosylation.⁸³ As with any metabolic labeling strategy, validation of hits will be a necessary step to confirm probe incorporation into the desired glycan targets.⁸³ Additional experiments, including glycan degradation and mass spectrometry analysis, should be undertaken to confirm that l-sugar probes undergo metabolism-dependent incorporation into glycans.

CONCLUSION

Bacterial glycans are blockbuster antibiotic targets and vaccine candidates that have enormous untapped potential. This work describes novel metabolic probes to detect bacterial glycans bearing rare deoxy amino l-sugars and validates their utility in a range of bacterial pathogens. The narrow incorporation of l-sugar analogs is consistent with the apparent rare distribution of these epitopes and sets the stage to further probe and perturb these structures. Broadly, this work expands the toolkit to survey bacterial glycans and gain critical insight into glycan epitope variability across the bacterial domain.

METHODS

Materials and Chemical Synthesis.

Organic chemicals and anti-FLAG antibodies were purchased from Millipore Sigma. H. pylori strain G27⁸⁴ was a gift of Manuel Amieva (Stanford University). Bacterial cells (P. shigelloides ATCC 51903; V. vulnificus ATCC 43382; S. aureus ATCC BAA-1708; C. jejuni ATCC 33560; B. fragilis ATCC 23745) and AGS cells (ATCC CRL-1739) were purchased from ATCC and grown according to the supplier’s instructions. Ac₄GlcNAc (1), Ac₄GlcNAz (2), Ac₄GalNAz (3), Ac₂Bac-diNAz (4), and Phos-FLAG were synthesized as previously described.^{44, 85, 86} l-FucNAz (5), l-PneNAz (6), l-RhaNAz (7), and l-QuiNAz (8) were synthesized using standard organic chemistry procedures and characterized by standard techniques including ¹H and ¹³C NMR spectroscopy and mass spectrometry. Analogs 5–8 were purified using flash silica gel chromatography.

Metabolic Labeling.

Bacterial cells were grown in rich liquid media supplemented with 1 mM⁸⁷ Ac₄GlcNAz (2), Ac₄GalNAz (3), Ac₂Bac-diNAz (4), l-FucNAz (5), l-PneNAz (6), l-RhaNAz (7), l-QuiNAz (8), or the azide-free control Ac₄GlcNAc (1) for 3 days. P. shigelloides, V. vulnificus, and S. aureus were grown in 1 mM sugar-containing media for 3 days under aerobic conditions at 37 °C, H. pylori and C. jejuni, were metabolically labeled for 3 days under microaerophilic conditions (14% CO₂, 37 °C), and B. fragilis were metabolically labeled for 3 days under anaerobic conditions (created by an Oxoid AnaeroGen Sachet in an airtight container; 37 °C), as appropriate for each bacterial strain.⁴⁴ AGS cells were grown in Ham’s F12 Glutamax supplemented with 10 μM of Ac₄GlcNAc (1), Ac₄GalNAz (3), Ac₂Bac-diNAz (4), l-FucNAz (5), l-PneNAz (6), l-RhaNAz (7), or l-QuiNAz (8) for 3 days in 5% CO₂ at 37 °C. Cells were then harvested, rinsed with phosphate buffered saline (PBS), and prepared for western blot or flow cytometry analyses as described below.

Western Blot.

Following metabolic labeling, cells were lysed and resultant protein lysates were standardized (BioRad’s DC protein concentration assay) to a protein concentration of ~3 mg mL⁻¹ prior to reaction with 250 μM Phos-FLAG for 12 h at room temperature. 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.

Flow cytometry.

Metabolically labeled intact cells were reacted with Click-IT Alexa Fluor 488 DBCO Alkyne (20 μM) for 1 hour in the dark for strain-promoted azide-alkyne cycloaddition detection of azides (Click Chemistry Tools; ex: 488/em: 519). Cells were analyzed by flow cytometry on a BD Accuri C6⁺ (BD Biosciences) instrument, with 10,000 live cells gated for each replicate experiment. Data were analyzed using FlowJo software.

ABBREVIATIONS

LPS

lipopolysaccharide

CPS

capsular polysaccharide

MOE

metabolic oligosaccharide engineering

Ac₄GlcNAz

peracetylated N-azidoacetylglucosamine

Ac₄GalNAz

peracetylated N-azidoacetylgalactosamine

Ac₄GlcNAc

peracetylated N-acetylglucosamine

d-DATDG

di-N-acetyl d-bacillosamine, d-Bac, d-2,4-diacetamido-2,4,6-trideoxy galactose

l-FucNAc

N-acetyl l-fucosamine

l-PneNAc

N-acetyl l-pneumosamine

l-QuiNAc

N-acetyl-l-quinovosamine

l-RhaNAc

N-acetyl l-rhamnosamine

NaNO₂

sodium nitrite

TBANO₂

tetrabutyl ammonium nitrite

Tf₂O

triflic anhydride

TBAN₃

tetrabutyl ammonium azide

CAN

ceric ammonium nitrate

AF488-DBCO

Alexa Fluor 488-dibenzocyclooctyne

AGS

gastric adenocarcinoma

ATCC

American Type Culture Collection

PBS

phosphate buffered saline