Synthesis of the Heteropolysaccharide O Antigen of Escherichia coli O52 Requires an ABC Transporter: Structural and Genetic Evidence

Abstract

The structural and genetic organization of the Escherichia coli O52 O antigen was studied. As identified by sugar and methylation analysis and nuclear magnetic resonance spectroscopy, the O antigen of E. coli O52 has a partially O-acetylated disaccharide repeating unit (O unit) containing d-fucofuranose and 6-deoxy-d-manno-heptopyranose, as well as a minor 6-deoxy-3-O-methylhexose (most likely, 3-O-methylfucose). The O-antigen gene cluster of E. coli O52, which is located between the galF and gnd genes, was found to contain putative genes for the synthesis of the O-antigen constituents, sugar transferase genes, and ABC-2 transporter genes. Further analysis confirmed that O52 employs an ATP-binding cassette (ABC) transporter-dependent pathway for translocation and polymerization of the O unit. This is the first report of an ABC transporter being involved in translocation of a heteropolysaccharide O antigen in E. coli. Genes specific for E. coli O52 were also identified.

The O antigen (O-specific polysaccharide) is a key component of the lipopolysaccharide (LPS) in the outer membrane of many gram-negative bacteria. It consists of repeats of an oligosaccharide unit (O unit), which usually contains two to eight residues of a broad range of both common and rarely occurring sugars and their derivatives. The O antigen is one of the most variable cell constituents, and there are variations in the types of sugars present, their arrangement within the O unit, and the linkages between the O units. The O antigen contributes the major antigenic variability to the cell surface and is subject to intense selection by the host immune system and other environmental factors, such as bacteriophages, which may account for the maintenance of diverse O-antigen forms within species, such as Escherichia coli (42).

Three mechanisms for the biosynthesis of O antigen have been recognized, as reviewed by Samuel and Reeves (50). These mechanisms are a Wzy/Wzx-dependent pathway, an ATP-binding cassette (ABC) transporter-dependent pathway, and a synthase-dependent pathway. All three pathways have the same initiation reaction, in which a sugar phosphate is attached to the carrier lipid, undecaprenyl phosphate (Und-P), on the cytosolic face of the plasma membrane. In the Wzy/Wzx-dependent pathway, additional sugars are added sequentially to form an O unit on the cytosolic face of the plasma membrane. The Und-PP-linked nascent O unit is then transferred across the plasma membrane by the O-unit flippase Wzx and is polymerized on the periplasmic face of the plasma membrane by the O-antigen polymerase Wzy (13). The O antigen is then attached to the independently synthesized lipid A core oligosaccharide to form the LPS (44). This pathway is utilized for the synthesis of the majority of E. coli, Shigella, and Salmonella O antigens studied. In the ABC transporter-dependent pathway, which has been found for only a few E. coli O antigens, additional sugars are added sequentially at the nonreducing end of the growing polymer to form an O antigen on the cytosolic face of the plasma membrane (9). The nascent Und-PP-linked O antigen is transferred across the plasma membrane by a member of the ABC-2 subfamily of ABC transporters and subsequently is ligated to the lipid A core (37). The ABC-2 transporters consist of an integral membrane protein, Wzm, and a hydrophilic protein containing an ATP-binding motif, Wzt. Involvement of the transporter with the translocation of the polymer is an attractive hypothesis, but it has not been proven experimentally and details of the process are not clear at this stage (41). This pathway for O-antigen biosynthesis has been observed only in E. coli O8 and O9 and Klebsiella pneumoniae O1 and O12 (12, 17, 23, 32). The synthase-dependent pathway has been reported only for the synthesis of the plasmid-encoded O54 antigen (a homopolymer of ManNAc) in Salmonella enterica (28). This pathway works like the ABC transport-dependent pathway, except that a synthase, which is an integral membrane protein, appears to catalyze a vectorial polymerization reaction by a processive mechanism, resulting in extension of the polysaccharide chain with simultaneous extrusion of the nascent polymer across the plasma membrane (28).

In E. coli, genes for O-antigen synthesis are normally located in a gene cluster, which typically contains genes for the synthesis of nucleotide sugar precursors and genes encoding glycosyltransferases for sequential and specific addition of sugars to form the O unit. Other genes involved in assembly of the O antigen are also located in the gene cluster; these genes include wzy (encoding O-antigen polymerase) and wzx (encoding O-antigen flippase) in the Wzy/Wzx system and wzm (encoding a membrane component of the ABC transporter) and wzt (encoding an ATP-binding component of the ABC transporter) in the ABC transporter system (50).

Of 186 O-antigen forms recognized in E. coli (including Shigella forms), at least 90 have known chemical structures (24, 30). The O-antigen gene cluster has been characterized for 22 of the E. coli O serotypes with known O-antigen chemical structures, including O6, O7, O8, O9, O26, O55, O75, O91, O104, O111, O113, O121, O157, Flexneri 2a, Sonnei, Dysenteriae O1, O7, Boydii O4, O5, O6, O9, and O13 (http://www.microbio.usyd.edu.au/BPGD/default.htm). In most of the serotypes the Wzy/Wzx-dependent pathway is used for synthesis of the O antigens; the exceptions are O8, O9 and 9a, in which the ABC transporter-dependent pathway is used. E. coli heteropolysaccharide O-antigen gene clusters usually map between galF and gnd in the genome, whereas the gene clusters for O8, O9, and 9a, with homopolysaccharide O antigens, map between gnd and hisI. Synthesis of a heteropolysaccharide O antigen via the ABC transporter-dependent pathway has been reported only for K. pneumoniae O12 (23).

In this study, we examined the structure of the E. coli O52 O antigen and also characterized its O-antigen gene cluster. The O antigen of E. coli O52 is a heteropolymer containing two unusual sugars, d-fucofuranose (d-Fucf) and 6-deoxy-d-manno-heptopyranose (d-6dmanHepp). We found that the O52 O-antigen gene cluster is located between galF and gnd and contains the genes for synthesis of the two constituent sugars, genes encoding putative sugar transferases, and ABC-2 transporter genes. We also found that the ABC-2 transporter genes are involved in the biosynthesis of the E. coli O52 O antigen. Five genes specific to E. coli O52 were identified by screening strains representing all E. coli O serotypes.

MATERIALS AND METHODS

Bacterial strains and plasmids.

All plasmids used in this study were maintained in E. coli K-12 strain DH5α, which was purchased from Beijing Dingguo Biotechnology Development Center (Beijing, People's Republic of China). E. coli O52 type strain G1066 was supplied by the Institute of Medical and Veterinary Science, Adelaide, Australia. All other Shigella and E. coli type strains used have been described previously (16).

Construction of a random DNase I shotgun bank.

Chromosomal DNA was prepared as previously described (6). Primers 1523 (5′-ATTGTGGCTGCAGGGATCAAAGAAATC-3′) and 1524 (5′-TAGTCGCGCTGNGCCTGGATTAAGTTCGC-3′) based on the galF and gnd genes, respectively, were used to amplify the E. coli O52 O-antigen gene cluster with the Expand Long Template PCR system (Roche). Each PCR cycle consisted of denaturation at 94°C for 10 s, annealing at 60°C for 30 s, and extension at 68°C for 15 min. To limit any PCR errors, five individual PCR products were pooled. The PCR products were digested with DNase I, and the resulting DNA fragments were cloned into pGEM-T Easy to produce a bank by using the method described previously (61).

Sequencing and analysis.

The plasmid DNA template used for sequencing was prepared by the method of Sambrook et al. (49). Sequencing was carried out with an ABI 3773 automated DNA sequencer. The Staden package (54) and the program Artemis (47) were used for sequence assembly and gene annotation, respectively. The program BLOCKMAKER (21) was used to search conserved motifs. BLAST and PSI-BLAST (2) were used to search databases, including the GenBank and Pfam protein motif databases (7), for possible gene functions. The algorithm of Eisenberg et al. (15) was used to identify potential transmembrane segments. Sequences were aligned and compared by using the program ClustalW (57).

Deletion of the wzm and orf16 genes from the E. coli O52 strain.

The wzm and orf16 genes were replaced by a chloramphenicol acetyltransferase (CAT) gene by using the RED recombination system of phage lambda (14, 63). The CAT gene was PCR amplified from plasmid pKK232-8 (Pharmacia) by using primers wl-1066 (5′-TGACAGTCTTTGTTAGAGCTCTATTGAGAAAAAGGTAACATGGAGAAAAAAATCACTGG) and wl-1067 (5′-AATTCTAATCTCTTAAATGCATACAACGAAATAGAAATCAAAAAATTACGCCCCGC) bound to the 5′ and 3′ ends of the wzm gene, respectively, and primers wl-1070 (5′-CAGGATAATAGCTTAAGGAAGGAGATGTGAGTTTTAAATATGGAGAAAAAAATCACTGG) and wl-1071 (5′-TTAGAGCACTTCTTGCATTGTGAGTTTATTATAGATTTGAAAAAATTACGCCCCGC) bound to the 5′ and 3′ ends of the orf16 gene, respectively; each primer carried 39 bp based on the E. coli O52 DNA which flanked the corresponding gene. The PCR product was transformed into the E. coli O52 strain carrying pKD20, and chloramphenicol-resistant transformants were selected after induction of the RED genes by using the protocol described by Datsenko and Wanner (14). PCR performed with primers specific for the CAT gene and E. coli O52 DNA flanking the objective gene was carried out to confirm the replacement.

Preparation of LPS.

Bacteria were grown to the late log phase in 10 liters of Luria-Bertani medium by using a 16-liter fermentor (BIOSTAT C-10; B. Braun Biotech International, Melsungen, Germany) with constant aeration at 37°C and pH 7.0. Bacterial cells were washed and dried as described by Robbins and Uchida (46). The LPS (860 mg) was isolated from dried cells (15.8 g) by the phenol-water method (62) and was purified by precipitation of nucleic acids and proteins with CCl₃CO₂H as described previously (64).

Preparation and O deacetylation of the O polysaccharide.

The LPS (150 mg) was hydrolyzed with aqueous 2% acetic acid at 100°C for 30 min, and a lipid precipitate was removed by centrifugation at 13,000 × g for 20 min. The water-soluble carbohydrate portion was fractionated by gel permeation chromatography on a column (56 by 2.6 cm) of Sephadex G-50 (S) in 0.05 M pyridinium acetate buffer (pH 4.5) with monitoring with a Knauer differential refractometer to obtain an O polysaccharide (51 mg).

The O polysaccharide (50 mg) was heated with aqueous 12% ammonia (4 ml) at 37°C for 16 h, and after evaporation an O-deacetylated polysaccharide (33 mg) was isolated by gel permeation chromatography on a Sephadex G-50 (S) column as described above.

Sugar and methylation analyses.

The O polysaccharide was hydrolyzed with 2 M CF₃CO₂H at 120°C for 2 h, and the monosaccharides were identified as the alditol acetates (51) on an Ultra 2 capillary column by using a Hewlett-Packard 5880 chromatograph, a temperature gradient consisting of 180°C for 1 min followed by an increase to 290°C at a rate of 10°C min⁻¹, and a Finnigan MAT ITD-700 gas-liquid chromatograph (GLC)-mass spectrometer (MS) with a temperature gradient consisting of 150°C for 1 min followed by an increase to 280°C at a rate of 5°C min⁻¹. The absolute configurations of the monosaccharides were determined by GLC of the acetylated (S)-2-octyl glycosides as described previously (35). Methylation was performed with CD₃I by the method of Hakomori (20), the methylated polysaccharide was recovered by extraction with ethyl acetate after dilution of the reaction mixture with water, and the partially methylated aldito acetates were derived and analyzed by GLC-MS as described above for the sugar analysis.

NMR spectroscopy.

Samples were deuterium exchanged by freeze-drying twice from D₂O and then examined as solutions in 99.96% D₂O at 50°C. Nuclear magnetic resonance (NMR) spectra were recorded with Bruker DRX-500 spectrometers by using internal acetone (δ_H 2.225, δ_C 31.45) as a reference. Two-dimensional NMR spectra were obtained by using standard Bruker software, and the Bruker XWINNMR 2.6 program was used to acquire and process the NMR data. Mixing times of 200 and 100 ms were used in TOCSY and ROESY experiments, respectively.

Other methods.

Membrane preparation, sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and silver staining for visualizing the LPS were carried out as described by Wang and Reeves (60).

Specificity assay with PCR.

Chromosomal DNA was prepared from 186 E. coli strains, including Shigella strains having different O-antigen serotypes. The quality of DNA was examined by PCR amplification of the mdh gene (coding for malate dehydrogenase and present as a housekeeping gene in E. coli) by using primers as described previously (40). A total of 26 DNA pools of E. coli and Shigella O serotypes were prepared, and each pool contained between 6 and 10 strains (16). The pools were screened by PCR by using primers based on specific genes of E. coli O52. Each PCR was carried out by using a total volume of 25 μl, 15 μl of which was loaded on an agarose gel to check for the presence of amplified DNA.

Nucleotide sequence accession number.

The DNA sequence of the E. coli O52 O antigen gene cluster has been deposited in the GenBank database under accession number AY528413.

RESULTS AND DISCUSSION

Elucidation of the structure of the E. coli O52 O polysaccharide.

The LPS was isolated from dried bacterial cells of E. coli O52 by the phenol-water procedure (62). The O polysaccharide was obtained by mild acid degradation of the LPS, followed by gel permeation chromatography on Sephadex G-50. Sugar analysis of the polysaccharide by GLC-MS of the acetylated alditols revealed a 6-deoxy-3-O-methylhexose, fucose, glucose, galactose, a deoxyheptose, and 2-amino-2-deoxyglucose (GlcN) at a ratio of 0.03:1:0.14:0.15:0.99:0.08. The deoxyheptose showed a characteristic fragmentation pattern in the electron impact mass spectrum, including C-1—C-2, C-1—C-3, C-1—C-4, C-5—C-7, and C-4—C-7 fragment ions at m/z 145, 217, 289, 159 and 231, respectively, which indicated that the molecule was either a 2-deoxyheptose or a 6-deoxyheptose. Glucose, galactose, and GlcN, which are present in small amounts, are common monosaccharides in enterobacterial LPS and could be attributed to the core region (22), whereas fucose and deoxyheptose seem to be components of the O unit. GLC analysis of the acetylated (+)-2-octyl glycosides indicated that fucose has the d configuration. Further studies demonstrated that the other sugar is 6-deoxy-d-manno-heptose (d-6dmanHep). Methylation analysis with CD₃I enabled identification of major 3-substituted Fucf and 3-substituted 6-deoxyheptopyranose, as well as minor amounts of terminal Fucf, terminal 6-deoxy-3-O-methylhexofuranose, and terminal 6-deoxyheptopyranose.

The ¹H-NMR spectrum of the polysaccharide (Table 1) showed signals for one CH₃-C group (H-6 of Fuc) at δ 1.27, one C-CH₂-C group (H-6 of 6dmanHep) at δ 1.75 and 2.13, one O-acetyl group (CH₃CO) at δ 2.14, and other protons at δ 3.39 to 5.34, including five signals in a lower field than δ 4.79 (i.e., in a resonance region for anomeric and other deshielded protons [e.g., protons at acetoxylated carbons]). The ¹³C-NMR spectrum of the polysaccharide (Fig. 1B and Table 2) showed four signals for anomeric carbons at δ 100.3 to 105.9, as well as signals for one CH₃-C group (C-6 of Fuc) at δ 19.9, one C-CH₂-C group and one HOCH₂-C group (C-6 and C-7 of 6dmanHep) at δ 35.1 and 59.5, respectively, other sugar carbons at δ 68.0 to 87.3, and one O-acetyl group (CH₃ at δ 21.7, CO at δ 174.4).

TABLE 1.

¹H-NMR data^a

Sugar residue	δ (J)b
	H-1	H-2	H-3	H-4	H-5	H-6 (6a)	H-6b	H-7a,7b
O-Deacetylated polysaccharides
→3)-β-d-6dmanHepp-(1→	4.80 (<2)	4.23 (∼2)	3.75 (9.5)	3.52 (9.5)	3.43	2.12	1.74	3.76
→3)-β-d-Fucf-(1→	5.18 (<2)	4.29 (<2)	4.17 (5.2)	4.06 (5.2)	3.97 (6.4)	1.26
O polysaccharide non-O-acetylated units
→3)-β-d-6dmanHepp-(1→	4.79	4.23	3.75	3.52	3.39	2.13	1.75	3.76
→3)-β-d-Fucf-(1→	5.18	4.29	4.17	4.06	3.98	1.27
O polysaccharides, O-acetylated units
→3)-β-d-6dmanHepp-(1→	4.86	4.24	3.75	3.52	3.43	2.13	1.75	3.76
→3)-β-d-Fucf2Ac-(1→	5.34	5.11	4.28	4.13	4.0	1.27

^aA signal for the O-acetyl group is at δ 2.14.

^bData for δ are expressed in parts per million, and data for J are expressed in hertz.

FIG. 1.

125-MHz ¹³C-NMR spectra of the O-deacetylated (A) and initial (B) O polysaccharide of E. coli O52.

TABLE 2.

¹³C-NMR data^a

Sugar residue	δ (ppm)
	C-1	C-2	C-3	C-4	C-5	C-6	C-7
O-Deacetylated polysaccharides
→3)-β-d-6dmanHepp-(1→	100.5	68.7	78.5	70.1	74.2	35.1	59.5
→3)-β-d-Fucf-(1→	105.9	81.2	85.8	87.3	68.4	19.9
O polysaccharide non-O-acetylated unit
→3)-β-d-6dmanHepp-(1→	100.5	68.6	78.5	70.1	74.3	35.1	59.5
→3)-β-d-Fucf-(1→	105.9	81.2	85.8	87.3	68.4	19.9
O polysaccharides, O-acetylated units
→3)-β-d-6dmanHepp-(1→	100.3	68.5	78.5	70.0	74.2	35.1	59.5
→3)-β-d-Fucf2Ac-(1→	103.7	83.3	83.7	87.7	68.0	19.9

^aSignals for the O-acetyl group are at δ 21.7 (CH₃) and 174.4 (CO).

Some ¹H- and ¹³C-NMR sugar signals were approximately twice as intense as other signals, including the signals for the O-acetyl group, and this suggested that O acetylation of the polysaccharide is nonstoichiometric. This was confirmed by mild basic O deacetylation of the polysaccharide, which significantly simplified the ¹H- and ¹³C-NMR spectra (Fig. 1A). In particular, the spectra of the O-deacetylated polysaccharide contained signals for only two anomeric atoms (δ_H 4.80 and 5.18; δ_C 100.5 and 105.9) and, hence, indicated that there was a disaccharide repeating unit.

The ¹H-NMR spectrum of the O-deacetylated polysaccharide was assigned by performing two-dimensional ¹H,¹H COSY and TOCSY experiments (Table 1), and spin systems for Fuc and 6dmanHep were identified. The TOCSY spectrum showed the following correlations: H-1/H-2-H-6 and H-6/H-5-H-2 for Fuc; and H-1/H-2-H-3, H-6/H-5-H-2, and H-6/H-7 for 6dmanHep. Signals for each proton within the spin systems were assigned by using the COSY spectrum. J_2,3, J_3,4, and J_4,5 coupling constant values of ∼2, 9.5, and 9.5 Hz demonstrated the manno configuration and the pyranoid form of 6dmanHep. A set of J_H,H coupling constants for fucose (Table 2) suggested that this sugar occurs in the furanoid form.

The ROESY spectrum of the O-deacetylated polysaccharide exhibited intense intraresidue H-1/H-3 and H-1/H-5 cross-peaks for 6dmanHep, which are typical of β-linked pyranosides. The spectrum also showed intense interresidue 6dmanHep H-1/Fuc H-3 and Fuc H-1/6dmanHep H-2 and H-3 cross-peaks; therefore, the polysaccharide is linear, and both constituent monosaccharides of the O unit are 3-substituted (strong nuclear Overhauser effects on both H-2 and H-3 are characteristic of 3-substituted, but not 2-substituted, manno-pyranosides [38]).

With the ¹H-NMR spectrum assigned, the ¹³C-NMR spectrum of the O-deacetylated polysaccharide (Fig. 1A) was assigned by performing an H-detected ¹H,¹³C HMQC experiment (Table 2). A relatively low-field position of the signal for C-1 of Fuc at δ 105.9 indicated the β-configuration of the fucofuranosidic linkage (compare previously published data for the α- and β-furanosides [8]). Down-field displacements of the signals for C-3 of Fuc and C-3 of 6dmanHep to δ 85.8 and 78.5, respectively, compared with their positions in the spectra of the nonsubstituted monosaccharides (8, 31), confirmed the glycosylation pattern. The positions of the signals for C-1 of Fucf and C-2 of 6dmanHepp at δ 105.9 and 68.7, respectively, indicated the same absolute configuration of these β-(1→3)-linked sugar residues (53) (i.e., the d configuration of 6dmanHepp). The ¹H,¹³C HMQC spectrum also had a minor cross-peak at δ_H/δ_C 3.46/59.1 for the O-methyl group of the 6-deoxy-3-O-methylhexose.

These data showed that the O-deacetylated E. coli O52 polysaccharide has the following structure: →3)-β-d-Fucf-(1→3)-β-d-6dmanHepp-(1→. This structure was finally confirmed by a ¹H,¹³C HMBC experiment, which demonstrated that there were correlations between the anomeric protons and the linkage carbons (Fuc H-1 and 6dmanHep C-3; 6dmanHep H-1 and Fuc C-3) and between the anomeric carbons and the protons at the linkage carbons (Fuc C-1 and 6dmanHep H-3; 6dmanHep C-1 and Fuc H-3).

Assignment in the same way of the ¹H- and ¹³C-NMR (Fig. 1B) spectra of the initial polysaccharide (Tables 1 and 2) revealed two series of signals in each spectrum. The signals of one series coincided with the signals of the repeating unit of the O-deacetylated polysaccharide, whereas the other series appeared to belong to an O-acetylated O unit. The acetylation at O-2 of Fuc was inferred from the significantly lower field position of the signal for H-2 of Fuc in the O-acetylated unit compared with the position in the non-O-acetylated unit (δ 5.11 versus δ 4.29), which was due to a strong deshielding effect of the O-acetyl group. The position of the O-acetyl group was confirmed by characteristic displacements of the signals for C-1—C-3 of Fuc caused by 2-O-acetylation (δ 105.9, 81.2, and 85.8 in the ¹³C-NMR spectrum of the O-deacetylated polysaccharide versus δ 103.7, 83.3 and 83.7 in the spectrum of the initial polysaccharide) (compare previously published data [25]).

On the basis of the data obtained, we concluded that the O polysaccharide of E. coli O52 has the following structure:

The O polysaccharide contains d-6dmanHep, a rarely occurring monosaccharide that has previously been found only in a few bacterial polysaccharides, including O polysaccharides of Yersinia pseudotuberculosis (10) and Burkholderia (Pseudomonas) pseudomallei (30). The polysaccharide includes a terminal 6-deoxy-3-O-methylhexose, which, most likely, is 3-O-methylfucose. The terminal modification with a mono-O-methylated monosaccharide has been previously reported for various homopolymer O antigens, including d-mannans of E. coli O8 and K. pneumoniae O3 and O5 with 3-O-methylmannose (26, 36, 59) and a d-rhamnan of Campylobacter fetus serotype B with 3-O-methyl-d-rhamnose (52).

Sequencing of the region between galF and gnd from E. coli O52.

A 18,900-base sequence including galF and gnd was obtained. Sixteen open reading frames other than galF and gnd were identified, and all were transcribed from galF to gnd (Fig. 2). The open reading frames were assigned functions based on their similarities to genes in databases (Table 3).

FIG. 2.

O-antigen gene cluster of E. coli O52.

TABLE 3.

Characterization of E. coli O52 O-antigen genes

Gene	Location in sequence	G + C content (%)	Conserved domain(s)	Similar protein (accession no./ no. of amino acids)	% identical/% positive (amino acid overlap)	Putative function
rmlB	1186-2274	41.5	NAD-dependent epimerase/dehydratase family (PF01370; E = 3.9 × e−203)	dTDP-glucose 4,6-dehydratase of Escherichia coli O91 (AAK60448/358)	75/86 (284)	dTDP-glucose 4,6-dehydratase
rmlA	2287-3156	39.5	Nucleotidyl transferase (PF00483; E = 3.1 × e−117)	Glucose-1-phosphate thymidylytransferase of Actinobacillus actinomycetemcomitans strain NCTC9710 (BAA28132/290)	80/89 (283)	Glucose-1-phosphate thymidylyltransferase
fcf1	3156-4106	34.2	NAD-dependent epimerase/dehydratase family (PF01370; E = 2.7 × e−15)	UDP-N-acetylglucosamine-4-epimerase of Yersinia enterocolitica O3 (CAA87706/336)	28/47(274)	Reductase
fcf2	4109-5242	36.6	UDP-galactopyranose mutase (PF03275; E = 7.1 × e−98)	UDP-galactopyranose mutase of Klebsiella pneumoniae O1:K20 (AAC98417/384)	60/76 (375)	Mutase
wbrV	5239-6093	33.1	Glycosyl transferase (PF00535; E = 3.9 × e−12)	Putative glycosyltransferase of Salmonella enterica serovar Typhimurium strain LT2 (AAL19664/298)	28/49 (243)	Glycosyltransferase
wzm	6095-6868	35.1	ABC-2-type transporter (PF01061; E = 3.2 × e−40)	Wzm of Klebsiella pneumoniae O8 (AAC98405/259)	26/49 (240)	Permease
wzt	6865-7839	35.5	ABC transporter (PF00005; E = 1.3 × e−33)	Wzt of Klebsiella pneumoniae O8 (AAC98406/246)	42/62 (219)	ATP-binding protein
dmhB	7913-8746	34.3	NAD-dependent epimerase/dehydratase family (PF01370; E = 3.1 × e−4)	DmhB of Yersinia pseudotuberculosis IA (AAN23043/277)	63/79 (276)	6-Deoxy-d-mannoheptose pathway protein
dmhA	8753-9781	39.7	NAD-dependent epimerase/dehydratase family (PF01370; E = 2.2 × e−14)	DmhA of Yersinia pseudotuberculosis IA (AAN23044/342)	91/96 (342)	6-Deoxy-d-mannoheptose pathway protein
hddA	9798-10826	40.4	GHMP kinases (PF00288; E = 4.2 × e−34)	d-Heptose 7-phosphate kinase of Yersinia pseudotuberculosis IA (AAN23045/342)	85/93 (342)	d-Heptose 7-phosphate kinase
gmhA	10814-11404	39.4	SIS domain (PF01380; E = 1.9 × e−16)	Glyceromannoheptose-7-P isomerase of Yersinia pseudotuberculosis IA (AAN23046/195)	85/93 (196)	Glyceromannoheptose-7-P isomerase
hddC	11407-12084	36.7	Nucleotidyl transferase	d-Heptose 1-phosphate guanosyltransferase of	71/83 (225)	d-Heptose 1-phosphate
			(PF00483; E = 5.6 × e−10)	Yersinia pseudotuberculosis IA (AAN23047/225)		guanosyltransferase
gmhB	12081-12611	36.5		d-Heptose 1,7-biphosphate phosphatase of Yersinia pseudotuberculosis IA (AAN23048/176)	61/76 (176)	d-Heptose 1,7-biphosphate phosphatase
wbrW	12608-13936	36.1	Glycosyl transferase group 1 (PF00534; E = 2.5 × e−4)	Glycosyltransferase of Agrobacterium tumefaciens C58 (AAL44360/364)	29/52 (236)	Glycosyltransferase
wbrX	13929-16088	35.9	Glycosyl transferase (PF00535; E = 2.1 × e−33)	Putative sugar transferase of Campylobacter jejuni subsp. jejuni NCTC 11168 (CAB73858/445); SAM-dependent methyltransferase of Methanopyrus Kandleri AV19 (AAM01852/225)	27/46(360)	Bifunctional enzyme (glycosyl transferase and methyltransferase)
					34/51 (164)
orf16	16244-17389	35.1	Glycosyl transferase group 1 (PF00534; E = 2.4 × e−39)	WbbO galactosyl transferase of Klebsiella pneumoniae O1:K20 (AAC98419/377)	44/67 (374)	Galactosyl transferase

(i) Sugar biosynthetic pathway genes.

As described above, the O52 O unit has two constituent sugar residues, d-Fucf and d-6dmanHepp. orf8 to orf13 showed 63, 91, 85, 85, 71, and 61% identity to the putative GDP-d-6dmanHep pathway genes dmhB, dmhA, hddA, gmhA, hddC, and gmhB of Yersinia pseudotuberculosis IA, IIA, and IVB, respectively (39). gmhA, hddA, gmhB, and hddC convert d-sedoheptulose 7-phosphate to GDP-d-glycero-α-d-manno-heptose (58), while it has been proposed that dmhA and dmhB convert the latter compound to GDP-d-6dmanHep. Therefore, orf8 to orf13 can be confidently identified as genes responsible for the synthesis of GDP-d-6dmanHep in E. coli O52 and are designated dmhB, dmhA, hddA, gmhA, hddC, and gmhB, respectively.

orf1 and orf2 showed 75 and 80% identity to rmlB and rmlA, respectively, of the E. coli K-12 (O16) O-antigen gene cluster. They also showed high identity to the rmlB and rmlA genes of a number of other bacterial strains (data not shown). Both rmlA and rmlB have been well characterized in many gram-negative bacteria, including E. coli. While RmlA converts d-glucose 1-phosphate to dTDP-d-glucose, RmlB converts the latter compound to dTDP-6-deoxy-d-xylo-hexos-4-ulose, which is a common intermediate of many different sugars (19). orf1 and orf2 can be confidently identified as rmlA and rmlB, respectively. In Actinobacillus actinomycetemcomitans serotype f, RmlA and RmlB catalyze the formation of dTDP-6-deoxy-d-xylo-hexos-4-ulose, which is converted to dTDP-d-fucopyranose (dTDP-d-Fucp) by dTDP-4-keto-6-deoxy-d-glucose C-4 reductase (Fcd) (27). Synthesis of galactofuranose (Galf) is known to involve mutase action on UDP-galactopyranose (UDP-Galp) (33), and the pathway genes for paratofuranose (Parf) include the genes for a full pathway for CDP-paratopyranose (CDP-Parp) and a mutase gene (43). One might have expected d-Fucf to be synthesized by mutase action on dTDP-d-Fucp. However, the O52 gene cluster includes no close homolog of fcd. Orf3 showed 23% identity to the GalE (UDP-glucose 4-epimerase) protein of E. coli K-12 (34). Although it showed only 16% identity to Fcd, Orf3 and other known C-4 reductases, including Fcd, are all GalE homologs, and it seems that Orf3 is the required C-4 reductase, which converts dTDP-6-deoxy-d-xylo-hexos-4-ulose to dTDP-d-Fuc. Orf4 belongs to the mutase family (COG0562; E = 2 × e⁻¹³⁰). It also exhibited 60% identity or 76% similarity to UDP-galactopyranose mutase encoded by glf of the K. pneumoniae O1 O-antigen gene cluster. Orf4 was identified as the required mutase based on its high level of amino acid identity to Glf. The enzymes of the dTDP-d-Fucf biosynthetic pathway have not been characterized either genetically or biochemically. We propose that Orf3 is a C-4 reductase and Orf4 is a mutase and that they are responsible for the conversion of dTDP-6-deoxy-d-xylo-hexose-4-uloseto d-Fucf. It is too early to propose a specific pathway, as the functions of Orf3 and Orf4 still have to be studied biochemically. We propose that orf1, orf2, orf3, and orf4 encode enzymes of the dTDP-d-Fucf biosynthetic pathway. As there are not enough data to confirm the functions of orf3 and orf4, these genes are designated fcf1 and fcf2, respectively, to indicate that they are in the d-Fucf pathway, but they should be renamed fcfA and fcfB, respectively, when their functions are known. orf1 and orf2 are designated rmlB and rmlA, respectively.

(ii) ABC transporter genes.

The protein encoded by orf6 belongs to the ABC-2 transporter family (PF01061; E = 3.2 × e⁻⁴⁰), which comprises permease components of ABC-type polysaccharide-polyol phosphate export systems. orf6 was most similar (26% identity) to wzm, which encodes an integral ABC-2 transport system protein in the K. pneumoniae O1 O-antigen gene cluster. Hydrophobicity analysis indicated that Orf6 is an integral membrane protein with six transmembrane segments, the average number for Wzm proteins. Orf7 belongs to an ABC transporter family (PF00005; E = 1.3 × e⁻³³), which comprises ATPase components of the ABC-type polysaccharide-polyol phosphate export system. orf7 also showed 42% identity to wzt, which encodes an ATP-binding protein of the K. pneumoniae O1 O-antigen gene cluster. The presence of the sequence GRNGAGKS, an ATP-binding protein motif (Walker box), as well as the ABC transporter family signature YSSGMLARLGFSIA, further indicates that orf7 encodes an ATP-binding protein. Therefore, orf6 and orf7 were identified as wzm and wzt, respectively. Evidence concerning the role of these genes is discussed below.

(iii) Sugar transferase genes.

Glycosyltransferases are specific for different sugars and sugar linkages. In E. coli, synthesis of the O unit is generally initiated by linkage of GlcNAc-P or GalNAc-P to Und-P. This reaction is catalyzed by UDP-GlcNAc:Und-P GlcNAc-1-phosphate transferase (WecA), whose gene is located outside the O-antigen gene cluster (4). In E. coli O8 and O9, while neither GlcNAc nor GalNAc is present as the first O-unit sugar, a single GlcNAc is present as the O unit is built on Und-PP-GlcNAc and wecA is required. In O8 and O9 there is also an adaptor sugar residue between the Und-PP-GlcNAc and the O-unit domain (29, 41, 45). E. coli O52 has a →3)-β-d-Fucf-(1→3)-β-d-6dmanHepp-(1→disaccharide repeating unit and, presumably, also requires an adaptor for the synthesis of O52 O antigen. Therefore, three sugar transferases are expected in the O-antigen gene cluster.

The proteins deduced from orf5 and orf14 showed 49 and 52% similarity to putative glycosyltransferases of Salmonella enterica serovar Typhimurium LT2 and Agrobacterium tumefaciens strain C58, respectively. They also showed similarity to many other putative glycosyltransferases (data not shown). Both Orf5 and Orf14 belong to glycosyltransferase family 2 (PF00535; E = 3.9 × e⁻¹² and E = 2.5 × e⁻⁴, respectively), which includes many glycosyltransferases involved in the biosynthesis of polysaccharide antigens. orf5 and orf14 were identified as glycosyltransferase genes and designated wbrV and wbrW, respectively.

The protein deduced from orf15 has 720 amino acids and thus is bigger than most monofunctional glycosyltransferases involved in the biosynthesis of bacterial polysaccharide antigens. It has two distinct domains; the C-teminal half of the protein shows homology to methyltransferases, and the N-terminal half shows homology to putative glycosyltransferases and also has a glycosyltransferase motif (PF00535; E = 2.1 × e⁻³³). This finding suggests that Orf15 is bifunctional. Other larger glycosyltransferases involved in O-antigen synthesis, including Orf8 of K. pneumoniae O12, WbbM of K. pneumoniae O1, WbbB of K. pneumoniae O5, WbbA of Serratia marcescens O4, Orf704 of E. coli O8, and WbdA of E. coli O9a (3, 5, 9, 11, 23, 48, 55, 56), also have two distinct domains and have been proposed to be bifunctional. Moreover, all these organisms also employ an ABC-2 transporter system for export of their O antigens. In the Wzy/Wzx-dependent pathway, Wzz acts as the chain length terminator, which controls the strain-specific distribution of the O-antigen chain lengths (18). In the ABC transporter-dependent pathway, the O-antigen chain growth appears to be stopped by a terminal modification (37). In E. coli O8 and K. pneumoniae O3 and O5, the O polysaccharides terminate with a 3-O-methylmannose residue (26, 59). It has been suggested that Orf704 of E. coli, the O8 O-antigen gene cluster, and WbbB of K. pneumoniae O5 are involved in the terminal modification, and both Orf704 and WbbB are bifunctional proteins and have a methyltransferase domain in the C-terminal half (3). Detection of a terminal 3-O-methylfucose residue as a minor component of the E. coli O52 antigen suggests that a methyl group is involved in terminal modification of the O polysaccharide in this strain. Therefore, it has been suggested that Orf15 encoded by the E. coli O52 O-antigen gene cluster is bifunctional and is involved in both O-unit sugar transfer and termination of the O polysaccharide. orf15 was designated wbrX.

(iv) A nonfunctional open reading frame.

Orf16 belongs to a glycosyltransferase family (COG0438; E = 8 × e⁻²⁷), and it exhibited 44% identity and 67% similarity to WbbO, a galactosyltransferase encoded in the O-antigen gene cluster of K. pneumoniae O1. In K. pneumoniae O1, WbbO is involved in transfer of a d-galactose residue (56). The level of identity between orf16 and wbbO is relatively high compared to the levels of identity for other glycosyltransferases, which usually exhibit low levels of homology. A deletion test showed that loss of orf16 from E. coli O52 did not interfere with the normal production of the LPS (Fig. 3A). These data and the absence of galactose from the O antigen of E. coli O52 indicate that orf16 is nonfunctional.

FIG. 3.

Deletion analysis of E. coli O52 orf16 (A) and wzm (B). Membrane extracts were electrophoresed on sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels and stained by silver staining. (A) Lane 1, G1066 (E. coli O52 type strain); lane 2, H1044 (G1066 missing the orf16 gene); lane 3, G1066 (E. coli O52 type strain). (B) Lane 1, H1042 (G1066 missing the wzm gene); lane 2, H1043 (H1042 with plasmid pLW1001 carrying the E. coli O52 wzm gene); lane 3, G1066 (E. coli O52 type strain).

In the E. coli O52 O antigen, d-Fucf is partially O acetylated. However, no putative gene responsible for the acetylation was found in the E. coli O52 O-antigen gene cluster. It is known that genes for acetylation may or may not be located in the O-antigen gene cluster (1). In case of E. coli O52, the gene(s) for the O antigen acetylation appears to be located outside the O-antigen gene cluster.

ABC transporters involved in biosynthesis of E. coli O52 O antigen.

Wzm and Wzt are required for the synthesis of E. coli O8 and O9 and K. pneumoniae O1 and O12 O antigens in an ABC-2 transporter-dependent pathway (17, 29, 56). The presence of the wzm and wzt genes in E. coli O52 and the absence of the wzx and wzy genes suggested that the E. coli O52 O antigen also is synthesized by an ABC-2-dependent pathway. We therefore performed a deletion experiment to confirm this. We replaced wzm with a CAT gene and observed that the mutant strain without wzm produced no O antigen in its LPS, while the wild-type strain produced normal LPS (Fig. 3B). The mutation was complemented by plasmid pLW1001 containing the E. coli O52 wzm gene. This result confirmed the involvement of wzm in biosynthesis of the E. coli O52 O antigen.

The involvement of an ABC transporter-dependent pathway is supported (i) by the lack of either GlcNAc or GalNAc in the O52 O unit, whereas one of these sugars is usually present as the first O-unit sugar in E. coli O antigens synthesized by the Wzy/Wzx-dependent pathway, and (ii) by the presence of a putative bifunctional glycosyltransferase/methyltransferase gene typically involved in terminal modification of the O-polysaccharide chains of this type.

O-antigen gene cluster of E. coli O52 is located between galF and gnd.

The O-antigen gene cluster in most E. coli strains studied to date is located between galF and gnd; the only exceptions are the O8, O9, and O9a gene clusters, which are located between gnd and hisI. The type 1 capsular antigen gene clusters are located between galF and gnd (http://www.microbio.usyd.edu.au/BPGD/default.htm). Since the O antigens of E. coli O8 and O9 are identical to those of K. pneumoniae O5 and O3, respectively, it has been suggested that they may have arisen by lateral transfer of the corresponding K. pneumoniae O-antigen clusters (26, 55). In contrast to other E. coli strains, which utilize the ABC transporter system, E. coli O52 has a heteropolymer O antigen, and its O-antigen gene cluster is located between galF and gnd, which is the normal location for E. coli strains. This finding suggests that a more complicated mechanism was involved in the origin of the ABC transporter system in E. coli.

Identification of the E. coli O52-specific genes.

In general, sugar transferase genes of the O-antigen gene clusters are strain specific. While wzt homologues are more conserved, especially in the region encoding the nucleotide-binding domain, wzm homologues for O-polysaccharide biosynthesis display little primary sequence identity (37). Primer pairs were designed based on wzm and the putative glycosyltransferase genes wbrV, wbrW, wbrX, and orf16 (Table 4), and PCR was performed to screen DNA pools from representatives of the 186 known O-antigen forms of E. coli and Shigella strains. Except for the pool containing E. coli O52 DNA (pool 28), which gave predicted PCR products for each primer pair, none of the other pools produced the expected PCR products. Therefore, all five genes are specific for E. coli O52 strains and, hence, potentially useful in PCR-based methods for detection of E. coli O52 strains.

TABLE 4.

PCR specificity test with E. coli O52 genes

Gene	Base positions	Forward primer (base positions)	Reverse primer (base positions)	No. of pools giving band(s) of the correct size	Annealing temp of PCR (°C)
wbrV	5239-6093	329 (5266-5283)	330 (5970-5988)	0a	58
		331 (5266-5283)	332 (5651-5668)	0	58
		333 (5362-5379)	334 (5783-5800)	0	53
wzm	6095-6868	311 (6314-6331)	312 (6840-6857)	0b	58
		313 (6180-6197)	314 (6689-6705)	0	55
		315 (6422-6439)	316 (6711-6728)	0	58
wbrW	12608-13936	305 (12940-12958)	306 (13869-13886)	0	53
		307 (13003-13020)	308 (13743-13762)	0	56
		309 (13853-13872)	310 (13345-13363)	0a	50
wbrX	13929-16088	323 (14317-14335)	324 (15064-15080)	0	55
		325 (14518-14535)	326 (15248-15267)	0	50
		327 (14421-14438)	328 (14845-14864)	0	55
orf16	16244-17389	317 (16286-16304)	318 (16935-16952)	0	58
		319 (16305-16321)	320 (16813-16831)	0	55
		321 (16588-16605)	322 (17257-17273)	0	56

^aOne pool gave two bands of the wrong size.

^bOne pool gave a band of the wrong size.

Conclusion.

The structure of the E. coli O52 O antigen and the sequence of its gene cluster were determined. The O antigen is composed of two rarely occurring sugars, d-Fucf and d-6dmanHepp. A minor component of the O52 O antigen, a 6-deoxy-3-O-methylhexose (probably 3-O-methylfucose), appears to be the terminal modification of the O52 O polysaccharide, which can be involved in chain length determination. The O52 O-antigen gene cluster consists of 16 open reading frames, and 15 of them were putatively identified on the basis of homology and account for all functions expected for the synthesis and processing of the O antigen. The last gene appeared to be not related to O-antigen synthesis. We obtained evidence that the ABC transporter system participates in the synthesis of the E. coli O52 O antigen. This is the first report showing that the ABC transporter system is involved in the synthesis of a heteropolymer O antigen in E. coli. It is also the first time that an O-antigen gene cluster containing ABC transporter genes was found to be located between galF and gnd, the normal location of E. coli O-antigen gene clusters for the Wzy/Wzx-dependent pathway.