A Novel Epimerase That Converts GlcNAc-P-P-undecaprenol to GalNAc-P-P-undecaprenol in Escherichia coli O157*

Escherichia coli strain O157 produces an O-antigen with the repeating tetrasaccharide unit α-d-PerNAc-α-l-Fuc-β-d-Glc-α-d-GalNAc, preassembled on undecaprenyl pyrophosphate (Und-P-P). These studies were conducted to determine whether the biosynthesis of the lipid-linked repeating tetrasaccharide was initiated by the formation of GalNAc-P-P-Und by WecA. When membrane fractions from E. coli strains K12, O157, and PR4019, a WecA-overexpressing strain, were incubated with UDP-[3H]GalNAc, neither the enzymatic synthesis of [3H]GlcNAc-P-P-Und nor [3H]GalNAc-P-P-Und was detected. However, when membrane fractions from strain O157 were incubated with UDP-[3H]GlcNAc, two enzymatically labeled products were observed with the chemical and chromatographic properties of [3H]GlcNAc-P-P-Und and [3H]GalNAc-P-P-Und, suggesting that strain O157 contained an epimerase capable of interconverting GlcNAc-P-P-Und and GalNAc-P-P-Und. The presence of a novel epimerase was demonstrated by showing that exogenous [3H]GlcNAc-P-P-Und was converted to [3H]GalNAc-P-P-Und when incubated with membranes from strain O157. When strain O157 was metabolically labeled with [3H]GlcNAc, both [3H]GlcNAc-P-P-Und and [3H]GalNAc-P-P-Und were detected. Transformation of E. coli strain 21546 with the Z3206 gene enabled these cells to synthesize GalNAc-P-P-Und in vivo and in vitro. The reversibility of the epimerase reaction was demonstrated by showing that [3H]GlcNAc-P-P-Und was reformed when membranes from strain O157 were incubated with exogenous [3H]GalNAc-P-P-Und. The inability of Z3206 to complement the loss of the gne gene in the expression of the Campylobacter jejuni N-glycosylation system in E. coli indicated that it does not function as a UDP-GlcNAc/UDP-GalNAc epimerase. Based on these results, GalNAc-P-P-Und is synthesized reversibly by a novel GlcNAc-P-P-Und epimerase after the formation of GlcNAc-P-P-Und by WecA in E. coli O157.

4-OH of GlcNAc-P-P-Und catalyzed by a novel epimerase encoded by the Z3206 gene in E. coli O157.
Thus, this paper describes a novel biosynthetic pathway for the assembly of an important bacterial cell surface component as well as a new biosynthetic route for the synthesis of GalNAc-P-P-Und. The potential of the bacterial epimerase as a new target for antimicrobial agents is discussed.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-E. coli strains PR4019 (13) and PR21546 (15) were generous gifts from Dr. Paul Rick, Bethesda, MD, and E. coli O157:H45 (16) was a gift from Dr. Claudio Zweifel, Veterinary Institute, University of Zurich. E. coli DH5␣ (Invitrogen) was used as the host for cloning experiments and for protein glycosylation analysis. Plasmids used are listed in Table 1.
Materials-[1,   (20 Ci/mmol) were obtained from American Radiolabeled Chemicals (St. Louis, MO). Quantum 1 silica gel G thin layer plates are a product of Quantum Industries (Fairfield, NJ), and Baker Si250 Silica Gel G plates are manufactured by Mallinckrodt Chemical Works. Yeast extract and Bacto-peptone were products of BD Biosciences. All other chemicals were obtained from standard commercial sources. Trimethoprim (50 g/ml), chloramphenicol (20 g/ml), ampicillin (100 g/ml), and kanamycin (50 g/ml) were added to the media as needed.
The gne gene was amplified from pACYCpgl (18), encoding Campylobacter jejuni pgl cluster, with oligonucleotides gne-Fw and gne-RV (AAACCATGGATGAAAATTCTTATTAGCGG and AAATCTAGATTAAGCGTAATCTGGAACATCGTA-TGGGTAGCACTGTTTTTCCCAATC; restriction sites are underlined). The PCR product was digested with NcoI and XbaI and ligated into the same sites of pMLBAD to generate plasmid pMLBAD:gne, which encodes Gne with a C-terminal hemagglutinin tag (Table 1).
Growth Conditions, Protein Expression, and Immunodetection-E. coli strains were cultured in Luria-Bertani medium (1% yeast extract, 2% Bacto-peptone, 0.6% NaCl) at 37°C with vigorous shaking. Arabinose-inducible expression was achieved by adding arabinose at a final concentration of 0.02-0.2% (w/v) to E. coli cells grown up to an A 600 of 0.05-0.4. The same amount of arabinose was added again 5 h post-induction, and incubation continued for 4 -15 h.
Total E. coli cell extracts were prepared for immunodetection analysis using cells at a concentration equivalent to 1 A 600 unit that were resuspended in 100 l of SDS loading buffer (19). Aliquots of 10 l were loaded on 10% SDS-PAGE. Periplasmic extracts of E. coli cells were prepared by lysozyme treatment (20), and 10 l of the final sample (corresponding to 0.2 A 600 units of cells) was analyzed by SDS-PAGE. After being blotted on nitrocellulose membrane, sample was immunostained with the specific antiserum (21). Anti-AcrA (18) antibodies were used. Anti-rabbit IgG-HRP (Bio-Rad) was used as secondary antibody. Detection was carried out with ECL TM Western blotting detection reagents (Amersham Biosciences).
For the preparation of membrane fractions, bacterial cells were collected by centrifugation at 1000 ϫ g for 10 min, washed once in ice-cold phosphate-buffered saline, once with cold water, and once with 10 mM Tris-HCl, pH 7.4, 0.25 M sucrose. The cells were resuspended to a density of ϳ200 A 600 units/ml in 10 mM Tris-HCl, pH 7.4, 0.25 M sucrose, 10 mM EDTA containing 0.2 mg/ml lysozyme, and incubated at 30°C for 30 min. Bacterial cells were recovered by centrifugation at 1000 ϫ g for 10 min, quickly resuspended in 40 volumes of ice-cold 10 mM Tris-HCl, pH 7.4, and placed on ice. After 10 min the cells were homogenized with 15 strokes with a tight-fitting Dounce homogenizer and supplemented with 0.1 mM phenylmethylsulfonyl fluoride and sucrose to a final concentration of 0.25 M. Unbroken cells were removed by centrifugation at 1000 ϫ g for 10 min, and cell envelopes were recovered by centrifugation at 40,000 ϫ g for 20 min. The membrane fraction was resuspended in 10 mM Tris-HCl, pH 7.4, 0.25 M sucrose, 1 mM EDTA and again sedimented at 40,000 ϫ g and resuspended in the same buffer to a protein concentration of ϳ20 mg/ml. Membrane fractions were stored at Ϫ20°C until needed.
Assay  1% Triton X-100, final concentration 0.1%) in a total volume of 0.05 ml. After incubation at 37°C, reactions were terminated by the addition of 40 volumes of CHCl 3 /CH 3 OH (2:1), and the total lipid extract containing [ 3 H]HexNAc-P-P-undecaprenols was prepared as described previously (22). After partitioning, the organic phase was dried under a stream of nitrogen and redissolved in 1 ml CHCl 3 /CH 3 OH (2:1), and an aliquot (0.2 ml) was removed, dried in a scintillation vial, and analyzed for radioactivity by liquid scintillation spectrometry in a Packard to an A 600 of 0.5-1. [ 3 H]GlcNAc was added to a final concentration of 1 Ci/ml, and the incubation was continued for 5 min at 37°C. The incorporation of radiolabel into glycolipids was terminated by the addition of 0.5 gm/ml crushed ice, and the cultures were thoroughly mixed. The bacterial cells were recovered by centrifugation at 4000 ϫ g for 10 min, and the supernatant was discarded. The cells were washed with ice-cold phosphate-buffered saline two times, resuspended by vigorous vortex mixing in 10 volumes (cell pellet) of methanol, and sonicated briefly with a probe sonicator at 40% full power. After sonication, 20 volumes of chloroform were added, and the extracts were mixed vigorously and allowed to stand at room temperature for 15 min. The insoluble material was sedimented by centrifugation, and the pellet was re-extracted with a small volume of CHCl 3 /CH 3 OH (2:1) twice. The combined organic extracts were then processed as described below.
Purification of GlcNAc-P-P-Und and GalNAc-P-P-Und-GlcNAc/GalNAc-P-P-Und was extracted with CHCl 3 /CH 3 OH (2:1) and freed of water-soluble material by partitioning as described elsewhere (22). The organic extract was then dried under a stream of nitrogen, and the bulk glycerophospholipids were destroyed by deacylation in toluene/methanol (1:3) containing 0.1 N KOH at 0°C for 60 min. The deacylation reaction was neutralized with acetic acid, diluted with 4 volumes of CHCl 3 /CH 3 OH (2:1), and washed with 1/5 volume of 0.9% NaCl. The organic (lower) phase was washed with 1 ⁄ 3 volume of CHCl 3 , CH 3 OH, 0.9% NaCl (3:48:47), and the aqueous phase was discarded. The organic phase was diluted with sufficient methanol to accommodate the residual aqueous phase in the organic phase and applied to a DEAE-cellulose column (5 ml) equilibrated with CHCl 3 /CH 3 OH (2:1). The column was washed with 20 column volumes of CHCl 3 /CH 3 OH/H 2 O (10: 10:3) and then eluted with CHCl 3 /CH 3 OH/H 2 O (10:10:3) containing 20 mM ammonium acetate. Fractions (2 ml) were collected and monitored for either radioactivity or GlcNAc/ GalNAc-P-P-Und using an anisaldehyde spray reagent (23) after resolution by thin layer chromatography on borate-impregnated silica plates (as described earlier).
GlcNAc/GalNAc-P-P-Und elute in fractions 3-10. These fractions were combined, supplemented with sufficient CHCl 3 and H 2 O to give a final composition of CHCl 3 /CH 3 OH/H 2 O (3:2:1), and partitioned. The aqueous layer was aspirated, and the organic layer was dried under a stream of nitrogen, spotted on a Quantum 1 thin layer plate (silica gel G, not borate-impregnated), and developed with CHCl 3 /CH 3 OH/H 2 O (65: 25: 4). GlcNAc/GalNAc-lipid zones were located by staining with iodine vapors and by comparison with radioactive marker lipids (if nonradioactive lipids were being purified) or detected by Bioscanning if radiolabeled GlcNAc/GalNAc-lipids were present. The desired zones were scraped from the plate, loaded into a Pasteur pipette, equipped with a glass wool plug, and washed sequentially with 2 column volumes of CH 3 OH and then with CHCl 3 /CH 3 OH/H 2 O (65:35:6) (ϳ10 column volumes). The eluate was adjusted to CHCl 3 /CH 3 OH/H 2 O (3:2:1) by the addition of CHCl 3 and H 2 O, and the phases were separated by a brief centrifugation. The aqueous layer was discarded, and the lower phase was dried under a stream of nitrogen, spotted on a borate-impregnated Quantum 1 thin layer plate, and developed with CHCl 3 , CH 3 OH, H 2 O, 0.2 M sodium borate (65:25:2:2). GlcNAc/GalNAc-P-P-Und were located by exposure of the TLC plate to iodine vapors and by comparison to radioactive marker lipids and recovered from the silica gel as described above, dried under a stream of nitrogen, and stored at Ϫ20°C until use.
Preparation of Borate-impregnated Thin Layer Plates and Whatman No. 1 Paper-Silica gel thin layer plates were impregnated with sodium borate by briefly immersing the plates in 2.5% Na 2 B 4 O 7 ⅐10 H 2 O in 95% methanol as described by Kean (24). The borate-impregnated TLC plates were dried overnight at room temperature and stored in a vacuum desiccator over Drierite until use. Immediately before chromatography, the plates were activated by heating briefly (ϳ10 -15 min) to 100°C. Whatman No. 1 paper was impregnated with sodium borate by dipping 20 ϫ 30-cm sheets of Whatman 1 paper in 0.2 M Na 2 B 4 O 7 ⅐10 H 2 O. The Whatman No. 1 paper sheets were pressed firmly between two sheets of Whatman No. 3MM paper and allowed to dry at room temperature for several days, as described by Cardini and Leloir (25).

Characterization of Glycan Products Formed in in Vitro
Reactions-The glycans of the individual glycolipids were characterized by descending paper chromatography after release by mild acid hydrolysis. The GlcNAc/GalNAc lipids were dried under a stream of nitrogen in a conical screw-cap tube and heated to 100°C, 15 min in 0.2 ml of 0.01 M HCl. After hydrolysis the samples were applied to a 0.8-ml mixed-bed ion-exchange column containing 0.4 ml of AG50WX8 (H ϩ ) and 0.4 ml AG1X8 (acetate form) and eluted with 1.5 ml water. The eluate was dried under a stream of nitrogen, redissolved in a small volume of H 2 O (0.02 ml), spotted on a 30-cm strip of borate-impregnated Whatman No. 1 paper, and developed in descending mode with butanol/pyridine/water (6:4:3) for 40 -50 h. After drying, the paper strips were cut into 1-cm zones and analyzed for radioactivity by scintillation spectrometry. GlcNAc and GalNAc standards were detected using an anilinediphenylamine dip reagent (26).
Glycan products were converted to their corresponding alditols by reduction with 0.1 M NaBH 4 in 0.1 M NaOH (final volume 0.1 ml) after mild acid hydrolysis as described above. After incubation at room temperature overnight, the reactions were quenched with several drops of glacial acetic acid and dried under a stream of nitrogen out of methanol containing 1 drop of acetic acid, several times. The alditols were dissolved in water, desalted by passage over 0.5-ml columns of AG50WX8 (H ϩ ) and AG1X8 (acetate), dried under nitrogen, and spotted on 30-cm strips of Whatman No. 3MM paper. The Whatman No. 3MM strips were developed overnight in descending mode with ethyl acetate, pyridine, 0.1 M boric acid (65: 25:20), dried, cut into 1-cm zones, and analyzed for radioactivity by scintillation spectrometry. GlcNAcitol and GalNAcitol standards were visualized using a modification of the periodate-benzidine dip procedure (27). The paper strips were dipped in acetone, 0.1 M NaIO 4 (95:5), allowed to air dry for 3 min, and then dipped in acetone/acetic acid/H 2 O/o-tolidine (96:0.6:4.4:0.2 gm). Alditols containing cis-diols stained as yellow spots on a blue background.
Mass Spectrometry of Glycolipids-Purified glycolipids were analyzed using an ABI/MDS Sciex 4000 Q-Trap hybrid triple quadrupole linear ion trap mass spectrometer with an ABI Turbo V electrospray ion source (ABI/MDS-Sciex, Toronto, Canada). In brief, samples were infused at 10 l/min with ion source settings determined empirically, and MS/MS information was obtained by fragmentation of the molecular ion in linear ion trap mode.
Analytical Procedures-Protein concentrations were determined using the BCA protein assay (Pierce) after precipitation of membrane proteins with deoxycholate and trichloroacetic acid according to the Pierce Biotechnology bulletin "Eliminate Interfering Substances from Samples for BCA Protein Assay." Samples were analyzed for radioactivity by scintillation spectrometry in a Packard Tri-Carb 2100TR liquid scintillation spectrometer after the addition of 0.5 ml of 1% SDS and 4 ml of Econosafe Economical Biodegradable Counting Mixture (Research Products International, Corp., Mount Prospect, IL).

UDP-GalNAc Is Not a Substrate for E. coli WecA (GlcNAcphosphotransferase)-
To determine whether E. coli WecA will utilize UDP-GalNAc as a GalNAc-P donor to form GalNAc-P-P-Und, membrane fractions from E. coli strains K12, PR4019, a WecA-overexpressing strain, and O157, which synthesize a tetrasaccharide O-antigen repeat unit with GalNAc at the reducing terminus presumably initiated by the synthesis of GalNAc-P-P-Und, were incubated with UDP-[ 3 H]GalNAc. As seen in Table 2, no labeled glycolipids were detected after the incubation with UDP-[ 3 H]GalNAc.
Moreover, neither the addition of exogenous Und-P to incubations with membranes from PR4019, the WecA-overexpressing strain, or the addition of cytosolic fractions from O157 cells resulted in the formation of GalNAc-P-P-Und from UDP-GalNAc (data not shown). These results demonstrated that UDP-GalNAc is not a substrate for WecA and suggested that GalNAc-P-P-Und is formed by an alternative mechanism.
When membranes from strain K12 were incubated with UDP-[ 3 H]GlcNAc, [ 3 H]GlcNAc-P-P-Und was synthesized as expected (13). However, when membranes from strain O157 were incubated with UDP-[ 3 H]GlcNAc, in addition to [ 3 H]GlcNAc-P-P-Und, a second labeled lipid shown to be [ 3 H]GalNAc-P-P-Und (see below) was observed. When the time course for the formation of the two glycolipids was examined, the incorporation of radioactivity into [ 3 H]GlcNAc-P-P-Und (Fig. 1, O) occurred more quickly and to a higher extent than into [ 3 H]GalNAc-P-P-Und (Fig. 1, F), compatible with a precursor-product relationship (Fig. 2).
The observation that E. coli O157 membranes do not utilize UDP-GalNAc as a GalNAc-P donor for the synthesis of GalNAc-P-P-Und prompted us to propose an alternative biosynthetic pathway for the formation of GalNAc-P-P-Und illustrated in Fig. 2. In this scheme GlcNAc-P-P-Und is formed by the transfer of GlcNAc-P from UDP-GlcNAc, catalyzed by WecA, and then GlcNAc-P-P-Und is epimerized by the action of a previously undescribed 4-epimerase to produce GalNAc-P-P-Und. The research described in this report was designed to test this hypothesis and to identify the novel GlcNAc 4-epimerase (E. coli O157 Z3206) catalyzing the formation of GalNAc-P-P-Und.

Characterization of [ 3 H]GalNAc-P-P-Und Formed in Vitro with Membrane Fractions from E. coli Strain O157-Consistent
with the additional O157-specific glycolipid product detected in Fig. 1, as GalNAc-P-P-Und, it was stable to mild alkaline methanolysis (toluene/methanol 1:3, containing 0.1 N KOH, 0°C, 60 min), retained by DEAE-cellulose equilibrated in CHCl 3 /CH 3 OH/H 2 O (10:10:3), and eluted with CHCl 3 / CH 3 OH/H 2 O (10:10:3) containing 20 mM ammonium acetate as reported previously for [ 3 H]GlcNAc 1-2 -P-P-Dol (28). The putative [ 3 H]GalNAc-P-P-Und was clearly resolved from [ 3 H]GlcNAc-P-P-Und by thin layer chromatography on borate-impregnated silica gel G (24) and purified by preparative TLC as shown in Fig. 3, panel A and B. When the glycolipid was treated with mild acid (0.01 N HCl, 100°C, 15 min), the watersoluble product co-chromatographed with [ 3 H]GalNAc on descending paper chromatography with borate-impregnated Whatman No. 1 paper (Fig. 3, panel C). In addition, when the labeled sugar was reduced, it was converted to [ 3 H]alditol, GalNAc-OH (Fig. 3, panel D). Moreover, negative-ion MS analysis yielded the [M-H] Ϫ ion of m/z ϭ 1128, expected for GalNAc-P-P-Und, and the MS/MS daughter ion spectrum showed a prominent ion at m/z ϭ 907, expected for a glycolipid containing P-P-Und (data not included) (29). The identification of the glycolipid product formed by strain O157 as GalNAc-P-P-Und is also supported by its formation from exogenous GlcNAc-P-P-Und (see below).
Identification of an E. coli O157 Gene Encoding GlcNAc-P-P-Und 4-Epimerase-Genomic sequences of different bacteria encoding O antigen repeating units having a GalNAc at the reducing terminus were screened. One group with a repeating unit containing a GalNAc at the reducing terminus and a second group lacking a terminal GalNAc in the repeating unit were compared to identify potential epimerases. Using these criteria Z3206 was identified as a candidate GlcNAc-P-P-Und 4-epimerase (Table 3). The gene encoding a candidate for the GlcNAc-P-P-Und 4-epimerase was identified by a combination of genetic and bioinformatic approaches. The genomic location of the Z3206 gene is consistent with a role in this pathway, as it resides between galF of the O-antigen cluster and wcaM, which belongs to the colanic acid cluster. The GlcNAc 4-epimerase genes present in E. coli strains with O-antigen repeat units containing GalNAc can be separated into two homology groups as shown in Table 3. One homology group (containing gne1) clearly is correlated with the presence of GalNAc as the initiating sugar on the O-antigen repeat unit. The second group (containing gne2) exhibits a high degree of similarity to the UDP-Glc epimerase, GalE, and is found in E. coli strains that do not initiate O-antigen repeat unit synthesis with GalNAc. Z3206 in E. coli O157, a gene with a high degree of homology to gne1, was selected for further study as a candidate GlcNAc-P-P-Und 4-epimerase.   H]GlcNAc/GalNAc-P-P-Und formation. E. coli strain 21546 was selected as the host for the Z3206 expression studies because a mutation in UDP-ManNAcA synthesis results in a block in the utilization of GlcNAc-P-P-Und for the synthesis of the enterobacterial common antigen. Because E. coli 21546 is derived from E. coli K12, it does not synthesize an O-antigen repeat as well (30), and thus, larger amounts of GlcNAc-P-P-Und accumulate for the conversion to GalNAc-P-P-Und. When strain 21546 and the transformant expressing the Z3206 gene were labeled with [ 3 H]-GlcNAc and the radiolabeled lipids were analyzed by thin layer chromatography on borate-impregnated silica gel plates, the parental strain (Fig. 4, panel A) synthesized only one labeled lipid, GlcNAc-P-P-Und. However, 21546 cells expressing the Z3206 gene (Fig. 4, panel B) also synthesized an additional labeled lipid shown to be GalNAc-P-P-Und (see above).
Membrane Fractions from E. coli Cells Expressing the Z3206 Gene Synthesize GalNAc-P-P-Und in Vitro-To corroborate that the protein encoded by the E. coli O157 Z3206 gene catalyzed the synthesis of GalNAc-P-P-Und, membrane fractions from E. coli cells expressing the Z3206 gene were incubated with [ 3 H]UDP-GlcNAc, and the [ 3 H]glycolipid products were analyzed by thin layer chromatography on borate-impregnated silica gel plates as shown in Fig. 5. When membrane fractions from E. coli K12 or the host strain E. coli 21546 cells were incubated with UDP-[ 3 H]GlcNAc, only [ 3 H]GlcNAc-P-P-Und was observed (Fig. 5, panels A and C). However, membrane fractions from E. coli O157 and E. coli 21546 expressing Z3206 formed GalNAc-P-P-Und as well (Fig. 5, panels B and D).
Formation of GlcNAc-P-P-Und, but Not GalNAc-P-P-Und, Is Reversed in the Presence of UMP-To provide additional evidence that GalNAc-P-P-Und is synthesized from GlcNAc-P-P-Und and not by the action of WecA using UDP-GalNAc as a glycosyl donor, the effect of discharging endogenous, pre-labeled [ 3 H]GlcNAc-P-P-Und and [ 3 H]GalNAc-P-P-Und with UMP was examined. The GlcNAc-phosphotransferase reaction catalyzed by WecA is freely reversible by the addition of excess UMP re-synthesizing UDP-GlcNAc and releasing Und-P.
In this experiment membrane fractions from the E. coli strain 21546 expressing Z3206 were pre-labeled for 10 min with UDP-[ 3 H]GlcNAc followed by the addition of 1 mM UMP, and the amount of each labeled glycolipid remaining was determined. The results illustrated in Fig. 6, panel A, show the relative amounts of [ 3 H]GlcNAc-P-P-Und and [ 3 H]GalNAc-P-P-Und at the end of the 10-min labeling period. After incubation with 1 mM UMP for 1 min, it can be seen that there is a substantial loss of [ 3 H]GlcNAc-P-P-Und, whereas the [ 3 H]GalNAc-P-P-Und peak is relatively unchanged (Fig. 6, panel B). This obser-  sylated protein, which migrates slower than the unglycosylated form, was formed only when cells expressing pgl locus ⌬gne were complemented by Gne (lane 2). Z3206 was unable to restore glycosylation of the reporter glycoprotein (Fig. 8, lane  1). Expression of Gne and membrane-associated Z3206 were confirmed by immunodetection (data not shown).

DISCUSSION
E. coli O157 synthesizes an O-antigen with the repeating tetrasaccharide structure (4-N-acetyl perosamine 3 fucose 3 glucose 3 GalNAc). This study was conducted to determine whether the biosynthesis of the lipid-linked tetrasaccharide intermediate was initiated by the enzymatic transfer of GalNAc-P from UDP-GalNAc to Und-P catalyzed by WecA, as indicated by earlier genetic studies (12). The results described here obtained from genetic, enzymology, and metabolic labeling experiments indicate that WecA does not utilize UDP-GalNAc as a substrate but that WecA is required to synthesize GlcNAc-P-P-Und, which is then reversibly converted to GalNAc-P-P-Und by a novel epimerase encoded by the Z3206 gene in strain O157.
The Z3206 gene was selected as a candidate to encode the epimerase because it belongs to a family of genes present in several strains that produce surface O-antigen repeat units containing GalNAc residues at their reducing termini (Table 3). Previous reports identified two genes from E. coli O55 (33) and E. coli O86 (34), gne and gne1, respectively, that are 100% identical to Z3206 (Table 3). We conclude that these genes also encode an epimerase capable of converting GlcNAc-P-P-Und to GalNAc-P-P-Und in strains O55 and O86, which also produce O-antigen repeat units with GalNAc at the reducing termini ( Table 3).
The gne and gne1 genes were previously proposed to encode a UDP-GlcNAc 4-epimerase (33,34). The gne and gne1 proteins from strains O55 and O86 were not examined in this study. However, two experimental approaches in this study indicate that the Z3206 protein does not catalyze the epimerization of UDP-GlcNAc to UDP-GalNAc in strain O157. First, when membranes from strain O157 were incubated with [ 3 H]UDP-GalNAc, neither [ 3 H]GlcNAc-P-P-Und nor [ 3 H]-GalNAc-P-P-Und was detected ( H]GlcNAc-P-P-Und should be observed. Second, we have shown that hemagglutinin-tagged Z3206 was incapable of complementing the UDP-GalNAc-dependent C. jejuni N-glycosylation reporter system (Fig. 8).
Although it cannot be excluded that accessory proteins in strains O55 and O86, not expressed in strain O157, might alter the specificity of the Gne and Gne1 proteins, it is also possible that the assays used to detect the epimerization of UDP-GlcNAc/GalNAc in the earlier studies (33,34) produced misleading results. For example, Gne1 was assayed by incubating crude cell extracts from strain O86 (34), which could have contained membranes capable of forming GlcNAc/GalNAc-P-P-Und, with UDP-GlcNAc and following a decrease in the reaction with p-dimethylaminobenzaldehyde. Because the extracts were treated with 0.1 N HCl, the loss of reactivity with the reagent could plausibly be due to the conversion of GlcNAc-P-P-Und to GalNAc-P-P-Und and the subsequent release of GalNAc from the lipid intermediate during acid hydrolysis.
Gne from strain O55 (33) was also assayed for epimerase activity by incubating crude extracts with UDP-GalNAc and indirectly assaying the conversion to UDP-GlcNAc by measuring an increase in reactivity with p-dimethylaminobenzaldehyde after acid hydrolysis. In both studies the formation of the putative product was based on changes in reactivity with p-dimethylaminobenzaldehyde and not a definitive characterization of the sugar nucleotide end product. A 90% pure polyhistidine-tagged Gne1 was also shown to have a low level of UDP-glucose epimerase activity relative to Gne2 in a coupled assay. Z3206 from strain O157 was not assayed for UDP-glucose epimerase in the study described here.
It is significant that E. coli O86, which synthesizes an O-antigen containing two GalNAc residues, which would presumably require UDP-GalNAc as the glycosyl donor for the additional, non-reducing terminal GalNAc, also possesses an additional putative GlcNAc 4-epimerase gene, termed gne2, within the O-antigen gene cluster (34). This additional epimerase gene has high homology with the galE gene of the colanic acid gene cluster and appears to be a UDP-GlcNAc 4-epimerase capable of synthesizing UDP-GalNAc.
The Z3206 gene appears to be highly conserved in E. coli O-serotypes initiated with GalNAc. In a recent study, 62 E. coli strains with established O-antigen repeat unit structures were screened for expression of Z3206 by a polymerase chain reaction-based method using nucleotide primers designed to specifically detect the E. coli O157 Z3206 gene (33). In this study Z3206 was detected in 16 of the 22 E. coli strains that were known to contain GalNAc and in only 4 of the 40 strains lacking GalNAc. Moreover, a similar screen of the 22 GalNAc-containing strains with primers designed to detect an alternative epimerase with UDP-GlcNAc 4-epimerase activity (the GalE gene of E. coli O113) detected no strains carrying this gene, indicating that Z3206 is the GlcNAc 4-epimerase gene most commonly associated with the presence of a reducing-terminal GalNAc in O-antigen repeat units of E. coli.
Analysis of the Z3206 protein sequence by a variety of webbased topological prediction algorithms indicates that the Z3206 protein is not highly hydrophobic. The majority of the topological prediction algorithms indicate that Z3206 is a soluble 37-kDa protein, although TMPred (35) predicted a single weak N-terminal transmembrane helix. However, Western blotting after SDS-PAGE of cellular fractions from E. coli cells expressing hemagglutinin-tagged Z3206 clearly shows that the tagged protein is associated with the particulate fraction after hypotonic lysis of the cells (data not shown). Preliminary experiments show that the protein remains associated with the particulate fraction after incubation of the membrane fraction with 1 M KCl but is solubilized in an active form by incubation with 0.1% Triton X-100 (data not shown). Further studies will be required to determine whether the Z3206 protein associates with the membrane fraction via a transmembrane helix or by some other means, perhaps through association with a membrane-bound binding partner.
E. coli O157 Z3206 has significant sequence homology with the short-chain dehydrogenase/reductase family of oxidoreductases including the GXXGXXG motif (Rossman fold), consistent with the NAD(P) binding pocket (36) and the conserved S X 24 YX 3 K sequence, involved in proton abstraction and donation (37). Molecular modeling based on crystal structures of UDP-Glc 4-epimerase, another member of the short-chain dehydrogenase/reductase family, suggests that, after hydride abstraction, the 4-keto intermediate rotates around the ␤ phosphate of UDP to present the opposite face of the keto intermediate and allow re-insertion of hydride from the opposite side, thus inverting the configuration of the hydroxyl at carbon 4. The presence of these conserved sequences suggests that Z3206 probably functions via a similar mechanism, but more experimentation will be required to verify this. Although the equilibrium distribution of the epimerase products, seen in Fig. 7, seems to favor the formation of GlcNAc-P-P-Und, the utilization of GalNAc-P-P-Und for O-antigen repeat unit assembly would drive the epimerization reaction in the direction of GalNAc-P-P-Und by mass action.
Epimerization of the glycosyl moieties of polyisoprenoid lipid intermediates has not been widely reported in nature. In one previous study the 2-epimerization of ribosyl-P-decaprenol to form arabinosyl-P-decaprenol, an arabinosyl donor in arabinogalactan biosynthesis in mycobacteria, was reported (38). Arabinosyl-P-decaprenol is formed via a twostep oxidation/reduction reaction requiring two mycobacterial proteins, Rv3790 and Rv3791. Although epimerization was modestly stimulated by the addition of NAD and NADP, neither Rv3790 nor Rv3791 contain either the Rossman fold or the SX 24 YXXXK motif, characteristic of the short-chain dehydrogenase/reductase family (36,37).
In summary, a novel biosynthetic pathway for the formation of GalNAc-P-P-Und by the epimerization of GlcNAc-P-P-Und is described. Several antibiotics have been shown to inhibit the synthesis of GlcNAc-P-P-Und but are limited in their utility because they also block the synthesis of GlcNAc-P-P-dolichol, the initiating dolichol-linked intermediate of the protein N-glycosylation pathway. Although GlcNAc-P-P-dolichol is a structurally related mammalian counterpart of the bacterial glycolipid intermediate, GlcNAc-P-P-Und, there is no evidence for a similar epimerization reaction converting GlcNAc-P-P-dolichol to GalNAc-P-P-dolichol. Thus, this raises the possibility that in strains where the surface O-antigen containing GalNAc at the reducing termini are involved in a pathological process, O-antigen synthesis could potentially be blocked by inhibiting the bacterial epimerases.