UDP-galactofuranose precursor required for formation of the lipopolysaccharide O antigen of Klebsiella pneumoniae serotype O1 is synthesized by the product of the rfbDKPO1 gene.

The O-side-chain polysaccharide in the lipopolysaccharide of Klebsiella pneumoniae O1 is based on a backbone structure of repeat units of [→3)-β-D-Galf-(1→3)-α-D-Galp-(1→]; this structure is termed D-galactan I. The rfb (O-antigen biosynthesis) gene cluster directs the synthesis of D-galactan I and consists of six genes termed rfbA-FKPO1. In this paper we show that rfbDKPO1 encodes a UDP-galactopyranose mutase (NAD(P)H-requiring) (EC 5.4.99.9), which forms uridine 5′-(trihydrogen diphosphate) P′-α-D-galactofuranosyl ester (UDP-Galf), the biosynthetic precursor of galactofuranosyl residues. The deduced amino acid sequence of rfbDKPO1 shows 85% and 37.5% identity to the rfbDKPO8 gene of K. pneumoniae serotype O8 and the glf gene of Escherichia coli, respectively. The molecular mass of the purified RfbDKPO1 enzyme is 45 kDa as determined by SDS-polyacrylamide gel electrophoresis, while gel filtration revealed a molecular mass of 92 kDa, suggesting a dimeric structure for the native protein. The rfbDKPO1 gene product interconverts uridine 5′-(trihydrogen diphosphate) P′-α-D-galactopyranosyl ester (UDP-Galp) and UDP-Galf Unlike Glf, RfbDKPO1 showed a requirement for NADH or NADPH, which could not be replaced by NAD or NADP. RfbDKPO1 was used to synthesize milligram quantities of UDP-Galf, allowing this compound to be purified and fully characterized in an intact form for the first time. The structure of UDP-Galf was proven by NMR spectroscopy.

Lipopolysaccharide (LPS) 1 is a major component of the outer membrane of Gram-negative bacteria. In enteric bacteria, the LPS molecule comprises a hydrophobic lipid A portion, which forms the outer leaflet of the outer membrane, a core oligosaccharide, and an O-side-chain polysaccharide. The O polysaccharide varies in structure from strain to strain, giving rise to unique antigenic epitopes (O antigens). In the genus Klebsiella, there exists a family of structurally related galactose-containing O polysaccharides. These are based on a backbone structure consisting of a disaccharide O-repeat unit [33)-␤-D-Galf-(133)-␣-D-Galp- (13] known as D-galactan I (1). Variations in O antigens arise from addition of side-chain ␣-D-Galp (2) and O-acetyl (2,3) residues, or by addition of domains of varying structure (1, 4 -6). Galactofuranosyl residues are present in a growing number of LPS O antigens e.g. in strains of Serratia marcescens (7,8), Shigella dysinteriae (9), Shigella boydii (10), Escherichia coli (11), Pasteurella hemolytica (12), Hemophilus pleuropneumoniae (13), and Actinobacillus pleuropneumoniae (14). The T1-antigen polysaccharide of Salmonella friedenau (15) and a variety of capsular or extracellular polysaccharides from both Gram-negative and Gram-positive bacteria (e.g. Refs. 16 -20) contain Galf. Galactofuranosyl residues are a central component in the mycolyl-arabinogalactan complex, which is characteristic of the cell walls of mycobacteria (21) and the related genera Nocardia and Rhodococcus (22), in the lipoglycan of Mycoplasma mycoides (23), in the paracrystalline S-layer glycoprotein of Clostridium thermohydrosulfuricum S102-70 (24), and in the cellulosome glycoproteins of Clostridium thermocellum (25) and Bacteroides cellulosolvens (26). In eukaryotes, galactofuranosyl residues are found in the lipophosphoglycan of Leishmania donovani (27), Leishmania major (28), and Leishmania mexicana (29), in the N-linked glycoproteins of Crithidia spp. (30), and in the lipopeptidophosphoglycan of Trypanosoma cruzi (31). A variety of fungal cell surface glycans, glycolipids and glycoproteins contain Galf residues in Penicillium spp. (32)(33)(34), Aspergillus spp. (35), Neurospora crassa (36), and Histoplasma capsulatum (37). Many of these microorganisms are important pathogens, and, for some, current therapies are limited. This observation, together with the absence of Galf residues in human glycoconjugates has fueled interest in the potential of generating novel therapeutic compounds directed against reactions involved in the formation of Galf precursors (38).
In the prototype system in K. pneumoniae O1, D-galactan I biosynthesis is directed by enzymes encoded by six genes in the chromosomal rfb locus. The polymer is synthesized in the cytoplasm by RfbCDEF activities and is then transported across the plasma membrane by a process involving an ATP-binding cassette (ABC-2) transporter, where RfbA is the transmembrane component and RfbB contains the consensus ATP-binding motifs (39). The synthesis of D-galactan I is initiated on a "primer" comprising undecaprenyl pyrophosphoryl N-acetylglucosamine. The primer is formed by the activity of the Rfe * This work was supported by grants (to C. W.) from the Canadian Bacterial Diseases Network, one of the federal Networks of Centers of Excellence. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) L31762 (for rfbCDEF KPO1 ).
The precursor for Galp residues is UDP-Galp (42), formed by the action of UDP-galactose 4-epimerase, encoded by the galE gene in the galactose operon (43). Early studies on the biosynthesis of the T1 antigen in S. enterica serovar Typhimurium (44,45) suggested that Galf precursors were derived from Galp at the level of UDP-linked sugars. The Galf precursor for a galactofuranosyl-containing polysaccharide in Penicillium charlesii was also proposed to be a UDP-linked derivative (46). More recently, the product of orf6, an enzyme encoded by the cryptic rfb locus in E. coli K-12 strains (47), was shown to have UDP-galactopyranose mutase (EC 5.4.99.9) activity, capable of the reversible formation of UDP-Galf from UDP-Galp (48). The product UDP-Galf was predicted from degradation products and this compound has not been isolated and fully characterized in an intact state.
This paper reports the cloning, expression, and purification of RfbD KPO1 from K. pneumoniae O1. We show that the enzyme has UDP-galactopyranose mutase activity but differs in several important aspects from the homologue from E. coli K-12. In addition, we report the isolation and first complete chemical characterization of intact UDP-Galf.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-The strains and plasmids used in this study are given in Table I. Strains were grown in Luria-Bertani (LB) medium containing the appropriate antibiotics. Antibiotics were used at the following concentrations: ampicillin, 100 g/ml; kanamycin, 50 g/ml; chloramphenicol, 50 g/ml.
DNA Manipulation and Analysis-Restriction endonuclease digestion, ligation, and transformation were performed essentially as described by Sambrook et al. (49). Plasmid DNA preparations were made with QIAGEN Inc. columns according to the manufacturer's instructions. CLUSTAL W was used for the multiple alignment (50) using default parameters.
Plasmid Construction-pWQ71 was constructed by cloning a XbaI-HindIII fragment of pWQ5 containing the complete rfb KPO1 gene cluster into pBBR1MCS. In order to delete an internal region in rfbD KPO1 , plasmid pWQ5 was mutagenized using a Transformer site-directed mutagenesis kit (Clontech), which is based on the method of Deng and Nickoloff (51). The 5Ј-phosphorylated primers 5Ј-CAT GTT TAT GGA TCC CAT ATT TTC C-3Ј and 5Ј-GAT TAC CAG GGA TCC GCA GTG ATG-3Ј were used to introduce a BamHI site (underlined) in the 5Ј and 3Ј-regions of rfbD KPO1 , respectively. The 5Ј-phosphorylated selection primer 5Ј-GCC TTT TTA CTT AGT AAT CCT AAA G-3Ј alters a unique CelII site in rfbF KPO1 and was used as selection primer. Digestion of this construct with BamHI, followed by religation, resulted in an in-frame deletion of 630 base pairs. The rfb KPO1 gene cluster containing the rfbD KPO1 deletion was subsequently cloned into pBBR1MCS on a XbaI-HindIII fragment, resulting in pWQ70. To clone the rfbD KP01 open reading frame, plasmid pWQ5 was mutagenized using the Transformer site-directed mutagenesis kit (Clontech) in order to introduce NdeI and HindIII sites overlapping with the start and stop codons of rfbD KPO1 , respectively. The NdeI-HindIII fragment containing the entire rfbD-KPO1 open reading frame was cloned into pET30a(ϩ) (Novagen), resulting in pWQ66.
SDS-PAGE of LPS-LPS samples were routinely prepared by the SDS-proteinase K lysate method (52). Samples were analyzed by SDS-PAGE according to the conditions described by Lesse et al. (53), and gels were silver-stained by the method of Tsai and Frasch (54).
Overexpression of RfbD KPO1 -An overnight culture of E. coli CWG288 (pWQ66) was diluted 1:200 into fresh medium (LB, 250 mM sorbitol, 50 g/ml kanamycin) and grown to an A 580 of 0.5. Galactose and isopropyl-1-thio-␤-D-galactopyranoside were added to a final concentration of 10 mM and 0.4 mM, respectively. The cells were then grown for an additional 3 h at 37°C, collected by centrifugation at 7,800 ϫ g, 4°C for 10 min, and stored at Ϫ70°C in aliquots of 1 g of wet cell paste.
Extraction of Nucleotide Sugars-For the extraction of sugar nucleotides, 1 g of wet cell paste was resuspended in 5 ml of 75% ethanol and incubated at 100°C for 10 min. Cell debris was removed by centrifugation (20,000 ϫ g, 10 min) and reextracted with another 5 ml of 75% ethanol. The samples were evaporated to dryness and redissolved in a small volume of water. Insoluble material was removed by centrifugation.
Purification of the RfbD KPO1 Protein of K. pneumoniae Serotype O1-Wet cell paste (1 g) was resuspended in 9 ml of 50 mM HEPES, pH 7.0, 1 mM phenylmethylsulfonyl fluoride, 0.1 mM dithiothreitol, and the suspension was disrupted by sonication. Cell debris was removed by centrifugation at 20,000 ϫ g for 30 min, followed by ultracentrifugation at 100,000 ϫ g for 1 h. The 100,000 ϫ g supernatant was further fractionated by ammonium sulfate precipitation at 4°C. The fraction between 55% and 75% saturation contained the RfbD KPO1 protein and was dialyzed against 50 mM HEPES pH 7.0 overnight, and then concentrated to ϳ 0.5 ml using a ultrafiltration cartridge (Centriplus 30, Amicon). RfbD was further purified by dye-ligand affinity chromatography on 5-ml columns of Reactive Green 5-agarose (Sigma) and Cibacron Blue 3GA-agarose (Sigma) at room temperature. The concentrated protein fraction was applied to the Reactive Green 5-column preequilibrated with 50 mM HEPES pH 7.0 and washed with 25 ml of 50 mM HEPES pH 7.0. This column bound the majority of applied protein, while most of the RfbD KPO1 protein was not retained under this conditions. The unbound fraction was subsequently applied to a 5-ml column of Cibacron Blue 3GA-agarose equilibrated with 50 mM HEPES pH 7.0, washed with 25 ml of 50 mM HEPES pH 7.0. The rfbD KP01 protein was eluted with 10 ml of 1 mM UDP-Galp in 50 mM HEPES pH 7.0. Residual contaminating proteins were removed by anion exchange chromatography on MonoQ at room temperature, using a linear sodium chloride gradient (0 -300 mM) in 50 mM HEPES pH 7.0.
HPLC Analysis-Sugar nucleotides were analyzed on a CarboPac PA1 column (4 ϫ 250 mm) according to the method of Palmieri et al. (55) with minor modifications. Sugar nucleotides were separated using a linear ammonium acetate (pH 7.0) gradient from 200 to 500 mM in 50 min at a flow rate of 0.6 ml/min. UDP-sugars were detected at 262 nm. For the purification of larger quantities of UDP-Galf, a CarboPac PA1 column (9 ϫ 250 mm) operated at a flow rate of 3 ml/min was used. Separated compounds were recovered by lyophilization. NMR Analysis-The 1 H, 13 C, and 31 P spectra were recorded on a Bruker AMX-600 spectrometer, at 285 K using a 5-mm broadband tunable probe. The sample was prepared from 0.5 mg of UDP-Galf in ammonium acetate buffer, lyophilized, and redissolved in 0.6 ml of 10 mM potassium phosphate buffer (pH 7.2), containing 0.5 mM EDTA. The pH of the sample dropped below 6; therefore, additional K 2 HPO 4 was immediately added to adjust the pH to 6.8. The sample was then lyophilized and redissolved in 0.6 ml of D 2 O and transferred to a 5-mm NMR tube. Acetone was added to the sample to provide the internal proton chemical shift reference at 2.225 ppm. The methyl resonance of an external acetone in D 2 O, set at 31.07 ppm, was used for the 13 C chemical shift reference. The chemical shift reference for 31 P was that of external phosphoric acid (25%) in D 2 O set at 0.0 ppm. All of the experiments were carried out without sample spinning and with the standard software and pulse programs provided by Bruker. The J (H,H) and J (P,H) coupling constants were measured directly from the onedimensional 1 H or 31 P spectra processed with a digital resolution of 0.2 Hz/point. The 1 H spectrum of 256 scans was recorded with presaturation of the HOD resonance at 4.945 ppm. The 13 C spectrum with proton decoupling was recorded overnight with 50,000 scans. The 31 P spectrum of 4000 scans was recorded without proton decoupling. The two-dimensional homonuclear magnitude COSY, phase-sensitive TOCSY, 1 H-31 P HMQC, 1 H-13 C HMQC experiments were recorded and processed as described previously (56). The one-dimensional TOCSY z-filtered spectra of 2000 scans was performed as described previously (57).

RESULTS
The K. pneumoniae Serotype O1 rfbD KPO1  The predicted translational product of rfbD KPO1 is a 384-amino acid protein with a molecular weight of 44,454 and a theoretical pI of 6.06. The N-terminal region between Lys-5 and Asp-33 contains a signature for an ADP-binding ␤␣␤-fold involved in FAD or NAD binding (58). The sequence deviates only at Gly-19 from the fingerprint. Consequently, a BLAST search (59) revealed striking similarities of the N terminus of rfbD-KPO1 to flavin-containing oxidases and dehydrogenases. Across the entire predicted polypeptide, 85% and 37.5% identity were found to the rfbD KPO8 gene of K. pneumoniae serotype O8 (also comprising D-galactan I (3, 60) and the glf (orf6) gene of E. coli K-12, respectively (Fig. 1).
Direct evidence that rfbD KPO1 encodes a Glf homologue was obtained from complementation experiments. The rfb KPO1 gene cluster was cloned into pBBR1MCS resulting in pWQ71. Sitedirected mutagenesis was used to generate plasmid pWQ70, which contains an in-frame deletion of 630 base pairs in the rfbD KPO1 gene of the D-galactan I gene cluster. A physical map of the DNA fragments cloned into pWQ70 or pWQ71 is given in Fig. 2. The LPS phenotypes conferred by these plasmids were analyzed by SDS-PAGE in DH5␣ and a derivative of strain SØ874 (CWG287). The glf gene, located in the cryptic E. coli K-12 rfb region (47,61), is deleted in E. coli SØ874. The LPS of DH5␣ or CWG287 both containing the complete cluster (pWQ71) showed a ladder of smooth LPS (S-LPS) (Fig. 3). Deletion of the internal region of rfbD KPO1 in pWQ70 resulted in no phenotypic alterations in DH5␣, while in CWG287 the synthesis of S-LPS was completely abolished (Fig. 3), suggesting that a gene of the cryptic rfb cluster of E. coli K-12, most likely glf, is functionally equivalent to rfbD KPO1 . An E. coli K-12 ⌬rfb Strain Overexpressing RfbD KPO1 Accumulates UDP-Galp and a Novel Galactose-containing UDP Derivative-The function of RfbD KPO1 was further addressed by analyzing the UDP-sugars synthesized in vivo. In order to achieve high level expression of RfbD KPO1 , a NdeI and a Hin-dIII site overlapping with the start and stop codon of rfbD KPO1 , respectively, were introduced by site-directed mutagenesis. The NdeI-HindIII fragment containing only the rfbD KPO1 open reading frame was cloned into the T7 expression vector pET30a(ϩ), resulting in pWQ66. The functionality of the rfb-D KPO1 gene was proven by complementation of CWG287 (pWQ70). When pWQ66 was transformed into CWG287 (pWQ70), the synthesis of S-LPS was restored, even without isopropyl-1-thio-␤-D-galactopyranoside induction (Fig. 3). To avoid further metabolism of in vivo synthesized UDP-Galp or UDP-Galf, CWG288 (E. coli K-12 ⌬rfb DE3 galE) was constructed as a galE derivative of CWG287. Additionally, sorbi-tol, which has been reported to suppress elevated levels of UDP-sugar hydrolase activity as a result of UDP-galactose accumulation (62), was added to the medium to avoid enzymatic degradation. Nucleotide sugars were extracted from CWG288 (pWQ66) and from the control strain CWG288 (pET30a(ϩ)), and extracts were analyzed by HPAEC. From  Fig. 4 it can be seen that both strains accumulated UDP-Galp (retention time 34.5 min), while an additional peak (p37) with a retention time of 37.5 min was unique to extracts of CWG288 (pWQ66). In experiments using [1-14 C]galactose, p37 was found to contain about 5% of the radioactivity, with the remaining 95% confined to the UDP-Galp peak (data not shown), suggesting that p37 is a novel galactose-containing sugar nucleotide formed by the activity of RfbD KPO1 .
Purification and Characterization of the RfbD KPO1 Protein-In order to unequivocally determine the function of Rfb-D KPO1 and the nature of p37, the RfbD KPO1 protein was purified to homogeneity from CWG288 (pWQ66) by ammonium sulfate precipitation, dye-ligand affinity chromatography, and anion exchange chromatography as described in detail under "Experimental Procedures." The protein was homogeneous as judged by SDS-PAGE (Fig. 5). The molecular mass of 45,000 Da of the denatured polypeptide determined by SDS-PAGE is in good agreement with the predicted molecular weight of 44,454. However, gel filtration on Superose 12 revealed a molecular weight of 92,000, suggesting a dimeric structure for the native protein (data not shown).
An in vitro assay was used to gain more information on the RfbD KPO1 -mediated reaction. The enzyme reaction was carried out in a total volume of 100 l, and the reaction products obtained after incubation at 37°C for 30 min were analyzed by HPAEC on a CarboPac PA1. The enzyme showed an absolute requirement for NADH or NADPH (data not shown). Under these reaction conditions, a peak well separated from UDP-Galp or UDP-Glc with the same retention time as p37 was obtained (Fig. 6B). In all experiments about 5% of UDP-Galp was converted to p37. No reaction could be seen in assays lacking the cofactor, or in assays containing NAD or NADP. When pooled p37 was used as substrate in the enzyme assay, ϳ95% was converted back to UDP-Galp (Fig. 6C), while no reaction was observed in assays lacking a cofactor or in assays with NAD or NADP (data not shown). Other than the cofactor requirements, these data are in agreement with the data de- scribed for Glf of E. coli (48) and further support the conclusion that p37 indeed is UDP-Galf.
NMR Analysis of UDP-Galf-In order to unequivocally prove that p37 is UDP-Galf, about 1.5 mg of p37 was pooled from several HPLC runs on a semipreparative CarboPac PA1 (9 ϫ 250 mm). After lyophilization, a small amount was redissolved in 100 mM ammonium acetate, pH 7.0, and analyzed by HPAEC. The material was found to be Ͼ95% pure (Fig. 6D) as judged by peak integration. During preparation of the sample for NMR analysis, the pH of the sample dropped below pH 6.0, which resulted in partial degradation of the nucleotide sugar. However, a complete assignment by two-dimensional methods of all proton, carbon, and phosphorous NMR resonances permitted the identification of intact p37 and of all the major components of the sample. By integration of the 1 H spectrum, the sample was found to be a mixture of the following major components: 26% UDP-Galf, 53% 5Ј-UMP, and 21% Galf-1,2-P. The 1 H spectrum is shown in Fig. 7, together with subspectra for the Galf moiety of UDP-Galf and Galf-1,2-P. The 1 H-13 C HMQC is shown in Fig. 8A along with the 1 H and 13 C spectra. The 1 H-31 P HMQC spectrum is shown in Fig. 8B, along with the 1 H and 31 P spectra. Proton, 13 C and 31 P chemical shifts are given in Tables II-IV. Coupling constants for 31 P are in Table  V and J (H,H) coupling constants in Table VI. Due to limited quantities of material, the low signal to noise in the 13 C spectrum did not permit the measurement of J (C,P) couplings.
UDP-Galf-The 31 P spectrum showed the presence of various phosphorylated compounds, one of which contained a pyrophosphate group that has characteristic shifts at Ϫ11.2 ppm and Ϫ12.4 ppm and a J (P@,P?) coupling constant of 21 Hz (63). A 1 H-31 P HMQC spectrum indicated the proton resonances, which were coupled to the pyrophosphate group. The proton resonances at 5.629 ppm and 4.152 ppm, which were coupled to the 31 P resonance at Ϫ12.4 ppm had J (P,H) coupling constants of 5.5 Hz and 2.3 Hz, respectively. The complete coupled spin system for these proton resonances was identified by COSY, TOCSY, and one-dimensional TOCSY experiments. The proton-proton coupling constants (J (H,H) ) could be obtained from the one-dimensional TOCSY spectrum. A 1 H-13 C HMQC permitted the assignment of all the carbon atoms directly bonded to these protons (C-H). The proton chemical shifts, J (H,H) coupling constants, 13 C chemical shifts (64), and 31 P chemical shifts were all characteristic of a terminal ␣-Galf residue bonded to a pyrophosphate group.
From a similar analysis, the resonances at 4.24 and 4.21 ppm, which bonded to the pyrophosphate resonance at Ϫ11.2 ppm, were found to belong the H5Ј and H5Љ resonances of the ribose moiety of UDP-Galf. The proton chemical shifts, J (H,H) coupling constants, and J (P,H) coupling constants of the UDP moiety were found to be characteristic for those found for UDP sugars (65).
5Ј-UMP-In the 1 H-31 P HMQC spectrum, the 31 P resonances at 3.5 ppm bonded to the proton resonances at 4.032 ppm and 3.970 ppm were found to belong to 5Ј-UMP by a complete assignment of the 1 H and 13 C spectra. The proton chemical shifts, J (H,H) coupling constants, and J (P,H) coupling constants of the 5Ј-UMP were found to be similar to those previously reported for this compound (66).
Galf-1,2-P-The 31 P resonance at 17.1 ppm is characteristic of a five-membered cyclic ester (63). In the 1 H-31 P HMQC spectrum, this 31 P resonance was coupled to the proton resonances at 6.000 ppm and 4.813 ppm, with J (P,H) coupling con- stants of 14 Hz and 6.8 Hz, respectively. From the assignment of the proton spin system for this residue obtained from the COSY and one-dimensional TOCSY experiment, the H1 and H2 resonances were found to be coupled to the cyclic phosphate. DISCUSSION Galactofuranosyl residues are found in a growing number of surface glycoconjugates from both Gram-negative and Grampositive bacteria, as well as protozoa and fungi. These include well documented pathogens of humans and livestock. In bacteria (this work, and Refs. 44, 45, and 48) and fungi (46), the biosynthetic precursor for Galf residues is believed to be UDP-Galf. Galactofuranosyl-containing glycoconjugates have not been reported in humans so far, which makes UDP-Galf formation an interesting target for novel therapeutic compounds (38). However, development of strategies for inhibitors has been limited by the lack of fundamental information regarding the biosynthesis of galactofuranosyl residues. Galactocarolose, a ␤-D-(135)-linked polygalactofuranosid produced by P. charlesii, was the first polysaccharide found to contain galactofuranosyl residues (34,67). Trejo et al. (46) isolated a nucleotide sugar formed in cell-free enzyme preparations of P. charlesii, which was capable of acting as a donor of galactofuranosyl residues in the biosynthesis of a galactofuranosylcontaining polymer. Based on chemical analysis of the reaction products obtained by acid and alkaline hydrolosis as well as periodate oxidation, Trejo et al. (46) predicted that the sugar nucleotide is UDP-Galf. More recently, Nassau et al. (48) cloned the glf gene of E. coli and showed that it encodes UDP-galactopyranose mutase (EC 5.4.99.9). Nassau et al. (48) concluded that the reaction product of the UDP-galactopyranose mutase reaction is UDP-Galf based on HPLC analysis of the sugar 1-phosphate obtained after phosphodiesterase treatment of the reaction product. The modification of an HPAEC procedure, to give a method that separates UDP-Galp from UDP-Galf, allowed us for the first time to purify milligram quantities of UDP-Galf and to unequivocally prove by NMR the structure to be uridine 5Ј-(trihydrogen diphosphate) PЈ-␣-D-galactofuranosyl ester. This will provide an essential reagent for biochemical analyses of synthetic systems for galactofuranosyl-containing glycoconjugates.
The data presented here establish that RfbD KPO1 catalyzes the interconversion of UDP-Galp and UDP-Galf. The LPS profiles of E. coli K-12 strains containing pWQ70 and pWQ71 clearly demonstrated that Glf (48) of E. coli can complement the rfbD KPO1 deletion in pWQ70. Glf and RfbD KPO1 are therefore functionally equivalent. Although the K. pneumoniae Rfb-D KPO1 protein shows 37.5% identity to Glf of E. coli, the two proteins do exhibit remarkable differences. Glf has been reported to be relatively unstable, resulting in complete loss of activity after storage of the purified protein at 4°C for Ͼ24 h (48). In contrast, RfbD KPO1 can be stored at 4°C for periods in excess of 1-2 weeks without significant loss of activity (data not shown). Moreover, no cofactor requirement other than FAD has been reported for Glf, while RfbD KPO1 has an absolute requirement for NADH or NADPH. These differences are difficult to explain. One possibility is that the binding constants for NADH (NADPH) are different in both proteins. Alternatively, the differences may be a result of the purification procedure used. Cibacron Blue 3GA used in the purification procedure reported here is thought to bind to NAD binding sites. Consequently, binding of the enzyme to this resin may displace the essential cofactor. Analysis of the predicted sequence of both proteins revealed a ADP-binding ␤␣␤-fold at its N terminus, which could be involved in binding FAD or NAD. The motif in K. pneumoniae differs at Gly-19 from the fingerprint. However, the known structure of p-hydroxybenzoate hydroxylase, which has a glycine at the same position (position 15 of the finger- Ϫ7.9  print), demonstrates that peptides with a glycine at this position can form a functional ␤␣␤ unit (68). The presence of FAD in the purified Glf protein has been reported (48). Although we have not analyzed this cofactor in RfbD KPO1 , the yellow color of the pure protein is consistent with its containing FAD. Gel filtration of the RfbD KPO1 protein on Superose 12 revealed a molecular weight of 92,000 for the native protein. The possible dimerization of Glf has not been investigated.
Little is known about the mechanism of the UDP-galactopyranose mutase (NAD(P)H-requiring) reaction. Stevenson et al. (47) suggested that the reaction may proceed via a 2-keto intermediate. Since in our hands the UDP-galactopyranose mutase reaction shows an absolute requirement for NADH or NADPH and was found to be inactive with NAD or NADP, an alternative is that the first step of the reaction may be a reduction step, although there is no net oxidation/reduction in the interconversion of UDP-Galp and UDP-Galf. The stability of the RfbD KPO1 protein and the possibility to synthesize and purify larger quantities of UDP-Galf will facilitate a detailed analysis of the reaction mechanism and determination of the protein structure.