In vitro assembly of the undecaprenylpyrophosphate-linked heptasaccharide for prokaryotic N-linked glycosylation

Abstract

Campylobacter jejuni has a general N-linked glycosylation pathway (encoded by the pgl gene cluster), which culminates in the transfer of a heptasaccharide: GalNAc-α1,4-GalNAc-α1,4-(Glcβ1,3)-GalNAc-α1,4-GalNAc-α1,4-GalNAc-α1,3-Bac [where Bac is bacillosamine (2,4-diacetamido-2,4,6-trideoxyglucose)] from a membrane-anchored undecaprenylpyrophosphate (Und-PP)-linked donor to the asparagine side chain of proteins at the Asn-X-Ser/Thr motif. Herein we report, the cloning, overexpression, and purification of four of the glycosyltransferases (PglA, PglH, PglI, and PglJ) responsible for the biosynthesis of the Und-PP-linked heptasaccharide. Starting with chemically synthesized Und-PP-linked Bac and various combinations of enzymes, we have deduced the precise functions of these glycosyltransferases. PglA and PglJ add the first two GalNAc residues on to the isoprenoid-linked Bac carrier, respectively. Elongation of the trisaccharide with PglH results in a hexasaccharide revealing the polymerase activity of PglH. The final branching glucose is then added by PglI, which prefers native lipids for optimal activity. The sequential activities of the glycosyl transferases in the pathway can be reconstituted in vitro. This pathway represents an ideal venue for investigating the integrated functions of a series of enzymatic processes that occur at a membrane interface.

Although asparagine-linked protein glycosylation has long been considered to be a unique feature of the eukaryotic kingdom, recent studies have revealed the presence of N-linked glycoproteins in both bacterial and archaeal domains. In particular, the Gram-negative bacterium Campylobacter jejuni, which is acutely involved in human gastroenteric disorders, has been shown to carry out both N-linked and O-linked glycosylation (1, 2). Although the exact mechanisms of host pathogenicity are unknown, evidence suggests that the glycans play a major role in host adherence, invasion, and colonization (3). Similar to eukaryotes, C. jejuni attaches the glycan to the asparagine side chain at the Asn-X-Ser/Thr motif where X can be any amino acid except proline (4). In C. jejuni, the glycan transferred is the heptasaccharide GalNAc-α1,4-GalNAc-α1,4-(Glcβ1,3)-GalNAc-α1,4-GalNAc-α1,4-GalNAc-α1,3-Bac, where Bac is bacillosamine (2,4-diacetamido-2,4,6-trideoxyglucose) (Fig. 1) (5). Genetic studies have identified a locus (dubbed “pgl” for protein glycosylation) that is responsible for this process and showed that it can be functionally reconstituted in Escherichia coli, strongly suggesting that this locus contains all of the genes necessary for the N-linked glycosylation process (Fig. 2) (6, 7). The locus is proposed to include genes encoding three major classes of proteins: carbohydrate-modifying enzymes responsible for the synthesis of the unusual sugar Bac, glycosyltransferases, and most importantly the oligosaccharyl transferase (PglB), which is responsible for transferring the preassembled heptasaccharide to protein. PglB is homologous to the STT3 subunit of the yeast oligosaccharyl transferase complex.

Fig. 1.

Chemical structure of the Und-PP-linked heptasaccharide [GalNAc-α1,4-GalNAc-α1,4- (Glcβ1,3)-GalNAc-α1,4-GalNAc-α1,4-GalNAc-α1,3-Bac- α-1-PP-Und, where Bac is bacillosamine (2,4-diacetamido-2,4,6-trideoxyglucose)].

Fig. 2.

Schematic representation of the C. jejuni pgl N-glycosylation locus.

The current model for the pgl pathway (Fig. 3) is based on bioinformatic and mutational analyses and shares similarities with the dolichol pathway in yeast and the O-antigen pathway in bacteria (2, 8–11). The pathway begins with a uridine diphosphate (UDP) GlcNAc or GalNAc, which is converted sequentially by a dehydratase (PglF), aminotransferase (PglE), and acetyltransferase (PglD) to afford UDP-Bac. UDP-Bac is then linked to undecaprenyl phosphate through the action of PglC to create the first membrane-associated intermediate undecaprenylpyrophosphate-linked Bac (Und-PP-Bac). This product is then elaborated by PglA, PglJ, and PglH to produce the hexasaccharide followed by PglI, which completes the glycan assembly by addition of a single-branching glucose. These four glycosyltransferases have very similar sizes and topologies; the enzymes have a M_r of ≈40 kDa and do not contain any predicted transmembrane domains (TMHMM, Expert Protein Analysis System). However, small amounts of detergent are necessary for their solubility, indicating that these enzymes may have a hydrophobic patch, which interacts with the bacterial membrane. The proteins bear little homology to the alg enzymes in the yeast dolichol pathway, which are all integral membrane proteins. However, they do share a significant amount of homology with the enzymes involved in the synthesis of the O-antigen sugar core (12) and were readily assigned as glycosyltransferases by bioinformatics analyses (6).

Fig. 3.

Current model of the C. jejuni pgl N-glycosylation pathway.

Although mutagenesis analysis suggests that PglA adds the first GalNAc whereas PglJ adds the second GalNAc, how the trisaccharide is elaborated to the hexasaccharide and the exact role of PglH remains in question (10). Three probable hypotheses have been presented: (i) PglH transfers a single GalNAc followed by two yet to be identified glycosyltranferases, (ii) PglH has polymerase activity and transfers the remaining three GalNAc residues, or (iii) PglH and PglJ act alternately to create the hexasaccharide. Ultimately, PglI adds the branching glucose and the flippase, WlaB, flips the completed lipid-linked heptasaccharide into the periplasm (6). Once in the periplasm, PglB transfers the completed saccharide affording a β-linked glycan to asparagine side chains in the target proteins.

Our interest lies in defining the precise roles of the pgl glycosyltransferases to develop a tractable system for investigating the sequential biosynthetic assembly process that occurs at the membrane interface. To achieve this goal, we have overexpressed four of the glycosyltransferases (PglA, PglH, PglI, and PglJ) in E. coli, purified them to homogeneity, and observed their action on chemically synthesized Und-PP-Bac (13). Using various combinations of enzymes, we demonstrated by mass spectral and HPLC analysis that PglA adds the first GalNAc to create the disaccharide; PglJ adds the second GalNAc to create the trisaccharde; PglH adds the remaining three GalNAc residues to create the hexasaccharide; and PglI adds the branching glucose. Furthermore we have shown that PglA is very specific for Und-PP-Bac, but can accept Und-PP-6-hydroxybacillosamine and Und-PP-GlcNAc to lesser extents. Strikingly, these four enzymes act extremely efficiently in concert to fully elaborate the monosaccharide precursor to the ultimate heptasaccharide in a reconstituted system.

With high quantities of active enzymes on hand, biophysical and structural analyses of these glycosyltransferases can now be targeted. Furthermore, the biochemical validation of these transferases helps to complete the picture of heptasaccharide assembly, sets the stage for using these glycosyltransferases to create native and unique substrates for the bacterial glycosylation machinery, and provides specific targets for the development of inhibitors that may reduce C. jejuni pathogenicity.

Materials and Methods

Cloning. PglA was amplified by PCR from C. jejuni genomic DNA (American Type Culture Collection 700819D, designation NCTC 11168) with the oligonucleotides PglABamHI (CGCGGATCCATGAGAATAGGATTTTTATCACATGCA) and PglAXhoI (TACATTCTTAATTACCCTATCATAAAG) by using Pfu Turbo (Stratagene) polymerase. The PCR product was digested with BamH I and XhoI and cloned into the same sites of vector pET-24a (Novagen). The final gene construct encoded a protein product with an N-terminal T7-Tag and a C-terminal His-Tag. PglH, PglI, and PglJ were cloned in an identical fashion by using the oligonucleotides PglHBamHI (CGCGGATCCATGATGAAAATAAGCTTTATTATCGCAAC), PglHXhoI (CCG-CTCGAGGGCATTTTTAACCCTCGGCTATAAGCTTA), PglIBamHI (CGCGGATCCATGCCTAAACTTTCTGTTATAGTACCAAC), PglIXhoI (CCGCTCGAGATTTTTGCATAAAG-CCACCCGAATTTTTG), PglJBamHI (CGCGGATCCATGCAAAAATTAGGCATTTTTATTTATTC), and PglJXhoI (CCGCTCGAGTCCTAATAAATATTTCAAAGCATCGCGT).

Expression. Starting from a 5-ml overnight culture, E. coli strains expressing PglA, PglH, PglI, or PglJ were grown at 37°C in LB broth to an OD₆₀₀ of 0.6–0.8. At that point, the temperature was reduced to 16°C, and protein production was induced by the addition of isopropyl β-d-thiogalactoside (1 mM). After 24 h, the cells were harvested by centrifugation (5,000 × g) and frozen at -80°C until needed.

Purification. All steps were performed at 4°C. Cell pellets of E. coli strains expressing PglA, PglH, PglI, or PglJ were thawed and resuspended in 5% of the original culture volume in buffer L (50 mM Tris-acetate, pH 8/20 mM imidazole). The cells were then subjected to sonication, followed by the addition of Triton X-100 (1%) and shaking for 10 min. Cellular debris and membrane proteins were removed by centrifugation (142,414 × g) for 1 h, and the supernatant was loaded on to a column containing Ni-NTA agarose equilibrated with buffer L. After washing with 10 column volumes of buffer W (50 mM Tris-acetate, pH 8/45 mM imidazole), the purified protein was eluted with buffer E (50 mM Tris-acetate, pH 8/250 mM imidazole). Fractions containing at least 500 μg/ml protein were used for enzyme assays, and no further purification of the proteins was undertaken.

Preparation of PglI Membrane Fraction. All steps were performed at 4°C. The cell pellet of PglI was thawed and resuspended in 5% of the original culture volume in buffer M (50 mM Tris-acetate, pH 8/1 mM EDTA). The cells were then subjected to sonication; unbroken cells were removed by centrifugation at 5,697 × g for 15 min, and the membrane fraction was collected by centrifugation at 142,414 × g for 60 min. The pellet was washed once with buffer M, centrifuged again, and resuspended in 0.25% of the original culture volume in buffer M. The final suspension was aliquoted and stored at -80°C.

Enzyme Assay. Buffer A (80 μl) [50 mM Tris-acetate, pH 7/3 mM DTT/5 mM MgCl₂/1% (vol/vol) Nonidet P-40/250 mM sucrose] was added to a tube containing 0.06 mg of dried Und-PP-Bac, UDP-GalNAc (2 mM final concentration), and/or UDP-glucose (0.4 mM final concentration). The mixture was vortexed vigorously and sonicated (water bath). Next, 5 μl each of the desired enzymes were added, and if necessary, buffer E was added to bring the total volume to 100 μl. Reactions were run at room temperature for 120 min and quenched by addition to a tube containing 800 μl of 2:1 chloroform:methanol and 160 μl of pure solvent upper phase (15 ml of chloroform/240 ml of methanol/1.83 g of potassium chloride in 235 ml of water). After vortexing for 20 s, the tubes were centrifuged briefly, and the organic layer (bottom) containing product was removed and dried.

Radioactive Enzyme Assay. To a tube containing 0.06 mg of dried Und-PP-Bac, Und-PP-6-hydroxybacillosamine, or Und-PP-GlcNAc, 3 μl of DMSO and 7 μl of 14.3% (vol/vol) Triton X-100 were added. After vortexing and sonication (water bath), 77 μl of H₂O, 5 μl of 1 M Tris-acetate (pH 8), 1 μl of 1 M MgCl₂, and 1 μl of PglA (660 μg/ml) were added. The reaction was initiated by the addition of 6 μl of UDP-GalNAc (2.25 μCi/nmol). Aliquots (15 μl) were taken at 2.5, 5, 10, and 15 min. Und-PP-6-hydroxybacillosamine and Und-PP-GlcNAc were prepared by means of the coupling of Und-P with the corresponding sugar phosphate similar to the procedure used in the synthesis of Und-PP-Bac (13). The 6-hydroxy Bac is an intermediate in the synthetic route to Bac (E.W., unpublished results) and GlcNAc-phosphate was synthesized according to a reported procedure (14).

HPLC and MALDI-MS Analysis of Saccharides. The saccharide sample was hydrolyzed by dissolving the dried sample in n-propanol:2M trifluoroacetic acid (1:1), heating to 50°C for 15 min, and then evaporating to dryness. The hydrolyzed sugars were then labeled with 2-aminobenzamide (2AB). To prepare the labeling reagent, a solution of 2AB (5 mg) in 100 μl of acetic acid:DMSO (1:2.3) was prepared. This dye solution was added to 6 mg of sodium cyanoborohydride, and aliquots of 5 μl of this reagent were added to dried samples of hydrolyzed, desalted glycans and heated to 60°C for 2–4 h. Postlabeling clean-up was accomplished by using GlykoClean S cartridges (ProZyme, San Leandro, CA) according to the manufacturer's instructions. The labeled glycans were separated on a normal phase analytical HPLC column (GlykoSepN, ProZyme) by using 50 mM ammonium formate (pH 4.4) (solvent A) and acetonitrile (solvent B) as eluents. A gradient of 20–52% solvent A over 80 min was used at a flow rate of 0.4 ml/min. The peaks were detected by using a fluorescence detector (λ_ex = 330 nm and λ_em = 420 nm), collected, and characterized with MALDI-MS by using a matrix composed of 2,5-dihydrobenzoic acid in acetonitrile/water with Nafion perfluorinated resin (Aldrich) and trifluoroacetic acid as additives.

Results

PglA, PglH, PglI, and PglJ were overexpressed, purified by using Ni²⁺ affinity chromatography, and confirmed by SDS/PAGE (Coomassie-stained) and Western blot analysis (Fig. 4). Various combinations of the purified enzymes were incubated in the presence of Und-PP-Bac, UDP-glucose, and UDP-GalNAc. Next, the Und-PP-linked saccharide products were isolated from the reaction mixture by extraction into organic solvent and hydrolyzed to remove the Und-PP moiety. The free saccharide was then fluorescently labeled by reductive amination (2AB/sodium cyanoborohydride) for HPLC analysis.

Fig. 4.

Ni-NTA purified glycosyl transferases. (A) Coomassie-stained polyacrylamide gel. (B) Anti-T7-Tag Western blot analysis. M_r markers (lane 1), PglA 43 kDa (lane 2), PglH 41 kDa (lane 3), PglI 36 kDa (lane 4), and PglJ 41 kDa (lane 5). The higher M_r band observed in lane 4 is due to protein oligomerization. The higher M_r band in lane 3 does not react with the anti-T7 Ab and represents an impurity brought through the purification.

PglA. When PglA alone is reacted, a peak with a retention time of 22 min is observed (Fig. 5A). When this peak is collected and subjected to MS analysis, a [M + Na]⁺ m/z of 592.27⁺ is observed, which corresponds to the sodium ion adduct of the disaccharide 2AB-Bac-GalNAc (Fig. 6A). PglH, PglI, and PglJ all failed to show activity on the lipid-linked monosaccharide (data not shown). To examine the prenylpyrophosphate-linked substrate specificity, two nonnative glycosyl acceptors were prepared through chemical synthesis: Und-PP-6-hydroxybacillosamine and Und-PP-GlcNAc. Und-PP-GlcNAc shows product formation that is several fold less efficient than the Und-PP-Bac whereas Und-PP-6-hydroxybacillosamine, which contains an extra hydroxyl group, is accepted significantly more efficiently than Und-PP-GlcNAc (Fig. 7).

Fig. 5.

HPLC traces of the 2AB-labeled saccharide products. (A) PglA. (B) PglA + PglJ. (C) PglA + PglJ +PglH. (D) PglA + PglJ + PglH first then extracted followed by addition of PglI (bacterial membrane fraction). Asterisks denote major saccharide peaks. The two peaks at 25 and 28 min are present in each trace and represent unknown byproducts of the 2AB-labeling process.

Fig. 6.

MALDI-MS of 2AB-labeled saccharide products. (A) PglA. (B) PglA + PglJ. (C) PglA + PglJ +PglH. (D) PglA + PglJ + PglH first then extracted followed by addition of PglI (bacterial membrane fraction). Major peaks represent sodium ion adducts. Additional peaks seen in B and C are due to potassium ion adducts.

Fig. 7.

Plot of Und-PP-disaccharide product formation vs. time by using PglA, UDP-GalNAc, and the three Und-PP-monosaccharides shown; dotted line, Und-PP-GlcNAc; dashed line, Und-PP-6-hydroxybacillosamine; solid line, Und-PP-Bac. Results represent a typical experiment.

PglJ. When PglA and PglJ are combined and the products analyzed by HPLC, a peak at 31 min is observed that has a [M + Na]⁺ m/z of 795.25 (Figs. 5B and 6B). This corresponds to the sodium ion adduct of the trisaccharide 2AB-Bac-GalNAc-GalNAc and is consistent with PglJ adding a single GalNAc residue. PglJ is very specific for this step as the combination of PglA and PglH failed to yield anything larger than the disaccharide (data not shown). These results are complementary with the in vivo studies, which demonstrated build up of a disaccharide intermediate when PglJ was genetically knocked out (10).

PglH. When PglA, PglH, and PglJ are combined, HPLC analysis reveals a peak with a dramatically increased retention time of 51 min (Fig. 5C). By MS analysis, this peak has a [M + Na]⁺ m/z of 1403.26, which corresponds to the hexasaccharide 2AB-Bac-GalNAc-GalNAc-GalNAc-GalNAc-GalNac (Fig. 6C). In a parallel experiment, trisaccharide is prepared by using PglA and PglJ and then reacted singly with PglH, a hexasaccharide is also observed (data not shown). Together, these two studies reveal that PglH is adding the three remaining GalNAc residues.

PglI. PglI is proposed to add the final branching glucose to complete the heptasaccharide (Fig. 1). When PglI is purified by Ni-NTA chromatography in the presence of Triton X-100, and subjected to SDS/PAGE analysis, two major bands are observed by Coomassie staining and Western blot analysis (Fig. 4). The lower M_r band corresponds well to the predicted M_r of PglI, whereas the higher band indicates the presence of protein oligomerization. When added in combination with the other transferases, PglI purified in this manner failed to show any glucosyl transferase activity (data not shown). However, when a bacterial membrane fraction prepared from cells that are overexpressing PglI is combined with PglA, PglJ, and PglH, a new peak at 56 min is observed (Fig. 5D). This peak has a [M + Na]⁺ m/z of 1,567.08 that corresponds to the full heptasaccharide (Figs. 2 and 6D). Control studies with bacterial membranes prepared by using the same overexpression conditions but by using a plasmid, which did not have the PglI gene insert, did not show transferase activity confirming that PglI is responsible for the Glc transfer (data not shown).

Discussion

The processes by which nature assembles complex saccharides is of fundamental biological importance. In the C. jejuni system, a heptasaccharide is assembled in a stepwise fashion on an Und-PP carrier, by a process that has many parallels with the dolichol pathway in yeast (15). PglA acts first on the Und-PP-Bac substrate to yield the disaccharide. Although in vivo reconstitution studies in E. coli suggested that PglA had a relaxed substrate specificity (7), our studies revealed that PglA has strong preference for Bac over GlcNAc in vitro consistent with the observation that only Bac is observed in C. jejuni in vivo (5). Of interest, the intermediate 6-hydroxybacillosamine is accepted almost as well as Bac, indicating that the additional N-acetyl group at C-4 is the most important determinant in the specificity of PglA. PglJ transfers a single sugar whereas PglH has a polymerase activity adding the three remaining GalNAc residues. Defining the roles of PglJ and PglH was crucial for a complete understanding of the glycan assembly process. In vivo studies were not able to account for the complete assembly of hexasaccharide and concluded with three possible scenarios (10). The studies presented herein clearly indicate that PglH precisely transfers the three terminal GalNAc residues independent of the PglI glucosyl transferase activity. Finally, PglI is considerably more hydrophobic than the other transferases. In particular, the protein has a modest probability of having a C-terminal transmembrane domain (TMHMM, Expert Protein Analysis System), and when purified in the presence of detergents it shows oligomeric behavior by SDS/PAGE, which is indicative of a poorly stabilized protein. Consequently, when this preparation was used for transferase studies no activity was observed. However, when a bacterial membrane preparation with PglI was generated it revealed very good activity, highlighting the important role played by native lipids in the stability and activity.

Remarkably, when the various transferases are combined, the intermediate saccharides are not observed; the transfer is complete. These data indicate that the sequential activities of the pathway have been reconstituted and suggest that the enzymes closely interact with each other. Having the exact roles of PglA, PglJ, PglH, and PglI clearly defined helps to complete the picture of glycan assembly in C. jejuni and sets the stage for using this pathway to probe the more fundamental questions surrounding multienzyme processes as well as engineering the enzymes to create unnatural saccharides.