A Novel NDP-6-deoxyhexosyl-4-ulose Reductase in the Pathway for the Synthesis of Thymidine Diphosphate-d-fucose*

The serotype-specific polysaccharide antigen ofActinobacillus actinomycetemcomitans Y4 (serotype b) consists of d-fucose and l-rhamnose. Thymidine diphosphate (dTDP)-d-fucose is the activated nucleotide sugar form of d-fucose, which has been identified as a constituent of structural polysaccharides in only a few bacteria. In this paper, we show that three dTDP-d-fucose synthetic enzymes are encoded by genes in the gene cluster responsible for the synthesis of serotype b-specific polysaccharide in A. actinomycetemcomitans. The first and second steps of the dTDP-d-fucose synthetic pathway are catalyzed byd-glucose-1-phosphate thymidylyltransferase and dTDP-d-glucose 4,6-dehydratase, which are encoded byrmlA and rmlB in the gene cluster, respectively. These two reactions are common to the well studied dTDP-l-rhamnose synthetic pathway. However, the enzyme catalyzing the last step of the dTDP-d-fucose synthetic pathway has never been reported. We identified the fcd gene encoding a dTDP-4-keto-6-deoxy-d-glucose reductase. After purifying the three enzymes, their enzymatic activities were analyzed by reversed-phase high performance liquid chromatography. In addition, nuclear magnetic resonance analysis and gas-liquid chromatography analysis proved that the fcd gene product converts dTDP-4-keto-6-deoxy-d-glucose to dTDP-d-fucose. Moreover, kinetic analysis of the enzyme indicated that theK m values for dTDP-4-keto-6-deoxy-d-glucose and NADPH are 97.3 and 28.7 μm, respectively, and that the enzyme follows the sequential mechanism. This paper is the first report on the dTDP-d-fucose synthetic pathway and dTDP-4-keto-6-deoxy-d-glucose reductase.

Capsular polysaccharides are the outermost structures on a bacterial cell and play a critical role in the interactions of the cell with the environment (1). These molecules are prominent, structurally and serologically diverse antigens that are involved in pathogenic processes and in mediating resistance to important host defense mechanisms. In the case of human pathogens, a large number of different capsule serotypes have been identified (1). Previously, we reported that the serotype b-specific antigen of Actinobacillus actinomycetemcomitans Y4 is a capsular polysaccharide-like antigen consisting of two deoxyhexoses, L-rhamnose and D-fucose (2).
A. actinomycetemcomitans is a nonmotile, Gram-negative, capnophilic, fermentative coccobacillus that has been implicated in the etiology and pathogenesis of localized juvenile periodontitis (3)(4)(5), adult periodontitis (6), and severe nonoral human infections (7). A. actinomycetemcomitans strains isolated from the human oral cavity are divided into five serotypes, a, b, c, d, and e (8 -10). The serologic specificity is defined by the polysaccharide on the surface of this organism (11), and the serotype-specific antigen is one of the immunodominant antigens in the organism (12,13).
Recently, a large gene cluster associated with the biosynthesis of the serotype-specific polysaccharide antigen of serotype b A. actinomycetemcomitans was cloned and characterized by our group (22). It follows from the components of serotype b-specific antigen that some genes coding for enzymes responsible for the synthesis of dTDP-D-fucose may be located in the cloned gene cluster. This is the first report to identify genes coding for essential enzymes involved in the dTDP-D-fucose synthetic pathway and to characterize the purified enzymes.
DNA Manipulation-DNA isolation, restriction endonuclease digestion, ligation, and transformation were performed essentially as described by Sambrook et al. (23).
Plasmid Construction-Plasmid pARF400 was constructed from pARF211 containing the region responsible for the synthesis of serotype b-specific polysaccharide antigen (22). A 7.0-kilobase NheI-BamHI fragment of pARF211 was ligated to XbaI-BamHI-digested pHSG399 (25) to produce pARF400. Seven deletion derivatives of pARF400 were constructed by digestion of the plasmid with appropriate restriction enzymes ( Fig. 2A). These eight plasmids were used to detect the gene coding for a dTDP-4-keto-6-deoxy-D-glucose reductase.
Purification of Enzymes-The rmlA, rmlB, and fcd products were purified using the expression vector, pTYB2 (New England Biolabs). To purify the rmlA, rmlB, and fcd products, E. coli ER2566 was transformed with pARF514, pARF509, and pARF517, respectively. Optimally, cells were grown in 500 ml of 2ϫ TY broth with ampicillin at 30°C to an optical density of 0.7 at 600 nm. Cultures were induced with 1 mM isopropyl-␤-thiogalactopyranoside. The cells were harvested 4 h after induction and lysed by ultrasonication. The cell extracts were obtained by centrifugation at 12,000 ϫ g for 30 min at 4°C. Binding to chitin beads (New England Biolabs), cleavage of the fusion protein (in 20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.1 mM EDTA, and 30 mM dithiothreitol at 4°C in 20 h), and elution of each product was done according to the manufacturer's instructions. The protein fraction was collected, and glycerol was added to a final concentration of 50%. The purified proteins were stored at Ϫ20°C until used. The purity of the proteins was checked by SDS-PAGE.
Enzyme Assay-The overall function of the rmlA, rmlB, rmlC, and rmlD products was ascertained by the detection of rhamnose. Moreover, the gene coding for a dTDP-4-keto-6-deoxy-D-glucose reductase was identified by the production of fucose. The reaction mixture (20 l), containing 50 mM Tris-HCl (pH 7.6), 12 mM MgCl 2 , 4 mM ␣-D-glucose-1-phosphate, 4 mM dTTP, 0.45 unit of inorganic pyrophosphatase (Roche Molecular Biochemicals), 6 mM NADPH, and appropriate amounts of the crude enzyme extracts, was incubated at 37°C for 3 h. The solution was mixed with 180 l of 0.5 M HCl. After the tube was sealed under vacuum, the mixed solution was heated at 80°C for 1 h to hydrolyze the reaction products.
Identification of Products-Sugar components in the hydrolyzed solution were coupled with 2-aminopyridine, and the pyridylamino sugars were analyzed by HPLC with an anion-exchange column (PALPAK type A column; Takara, Kyoto, Japan) as described previously (27,28). The identification of dTTP, dTDP-D-glucose, dTDP-4-keto-6-deoxy-D-glucose, or dTDP-D-fucose was confirmed using reversed-phase HPLC as described by Tonetti et al. (29). Samples (10 l) diluted 10-fold with distilled water were injected onto a TSKgel ODS-80Ts column (0.46 by 15 cm; Tosoh, Tokyo, Japan) with 0.5 M KH 2 PO 4 as the mobile phase at a flow rate of 1.0 ml/min at 40°C. The eluate was monitored with a UV detector at 260 nm.
NMR Analysis-The fractions containing dTDP-D-fucose were collected from the reaction mixture using the HPLC method described above. The reaction mixture (4 ml) contained 50 mM Tris-HCl (pH 7.6), 12 mM MgCl 2 , 4 mM ␣-D-glucose-1-phosphate, 4 mM dTTP, 90 units of inorganic pyrophosphatase, 8 mM NADPH, and 2 g of the purified rmlA, rmlB, and fcd products/ml. To remove the KH 2 PO 4 , four volumes of ethanol were added to the solution collected. The mixture was then centrifuged at 12,000 ϫ g for 10 min. After ethanol precipitation, the supernatant (60 ml) was concentrated to approximately 1 ml by evaporation and desalted by gel filtration on a Sephadex G-10 column (1.5 ϫ 69 cm; flow rate 30 ml/h; Amersham Pharmacia Biotech Inc.) with distilled water as the eluent. The sample was lyophilized and redissolved in 0.5 ml of D 2 O and transferred to a 5-mm NMR tube. 1 H NMR analysis was performed on a Bruker AM400 spectrometer. The measurement was made at 298 K. The chemical shifts were referenced to 3-trimethylsilylproionate-D 4 at 0.0 ppm. The 1 H spectrum of 128 scans was recorded with presaturation of HOD resonance at 4.72 ppm.
Gas-Liquid Chromatography Analysis-dTDP-D-Fucose as a sample was obtained by the same method as described under "NMR Analysis." About 2 mg of dTDP-D-fucose was dissolved in 300 l of 0.1 M HCl. An ampoule containing the solution was sealed under vacuum and heated at 80°C for 1 h to hydrolyze dTDP-D-fucose. The solution was then evaporated to remove water and HCl. The pellet, D-fucose, or L-fucose was converted into the corresponding D-(ϩ)-2-octylglycoside acetate by the method of Leontein et al. (30). Each product was characterized by gas-liquid chromatography (model GC-14B; Shimadzu Works, Tokyo, Japan) with a fused silica capillary column (CP Sil-88, 0.25 mm by 50 m; Chrompack Inc., Bridgewater, NJ) at 200°C. Approximately 1 l of the sample was injected, and the split ratio was 1:20. Helium was used as a carrier gas at a flow rate of 0.9 ml/min.

FIG. 2. Restriction map of the gene cluster involved in the synthesis of A. actinomycetemcomitans serotype b-specific antigen and deletion analysis of pARF400 (A) and HPLC profiles of monosaccharides obtained by acid hydrolysis of the reaction mixtures (B).
A, a linearized restriction map of the 25-kilobase chromosomal XbaI-SalI fragment containing the gene cluster associated with the synthesis of A. actinomycetemcomitans serotype b-specific antigen is shown. The open arrows indicate the positions of ORFs. The horizontal lines below the map show the DNA inserts carried by the recombinant plasmids. The production of fucose is shown to the right of each fragment by a plus sign, and a minus sign indicates undetectable fucose production. Ba, BamHI; Bg, BglII; Bs, BspHI; Bss, BssHII; Ec, Eco47III; Ev, EcoRV; M, MunI; N, NheI; Nd, NdeI; P, PstI; X, XbaI; kb, kilobases. B, the reaction mixture contained the crude enzyme extract from E. coli harboring pARF425 (top) or pARF400 (bottom). The pyridylamino sugars were analyzed by HPLC using a PALPAK type A column. Arrowheads labeled Xyl, Glu, and Fuc indicate the elution times of pyridylaminated xylose, pyridylaminated glucose, and pyridylaminated fucose, respectively. mM Tris-HCl (pH 7.6), 12 mM MgCl 2 , 4 mM ␣-D-glucose-1-phosphate, 4 mM dTTP, 450 units of inorganic pyrophosphatase, and 2 g of the purified rmlA and rmlB products/ml. The dTDP-4-keto-6-deoxy-D-glucose in the reaction mixture was purified by HPLC, ethanol precipitation, and gel filtration as described above.
The general approach for determining the kinetic properties of dTDP-4-keto-6-deoxy-D-glucose reductase is as described by Cleland (31) and Segel (32). The experiments were designed to provide data by varying the concentration of dTDP-4-keto-6-deoxy-D-glucose at several fixed concentrations of NADPH. The enzyme assay was carried out in a microcuvette containing 0.5 ml of a reaction mixture composed of 50 mM Tris-HCl (pH 7.6), 12 mM MgCl 2 , 80 ng of the purified protein/ml, 18.8 -300 M of dTDP-4-keto-6-deoxy-D-glucose, and 25-150 M of NADPH. The kinetic constants were determined by the estimating the amount of NADPH oxidation during the conversion of dTDP-4-keto-6deoxy-D-glucose to dTDP-D-fucose. The rate of the decrease in optical density at 340 nm was determined spectrophotometrically.
Identification of the Gene Encoding dTDP-4-Keto-6-deoxy-Dglucose Reductase-To identify the gene encoding dTDP-4-keto-6-deoxy-D-glucose reductase, eight deletion derivatives of pARF211 (pARF400, pARF420, pARF421, pARF422, pARF423, pARF424, pARF425, and pARF426) ( Fig. 2A) were constructed, and a crude enzyme extract from E. coli SØ874 harboring each plasmid was obtained. Each crude enzyme extract was added to a reaction mixture containing D-glucose-1-phosphate and dTTP as substrates and crude enzyme extracts from E. coli SØ874 expressing rmlA and rmlB as enzymes. Fucose was not detected in the hydrolysate of the reaction mixture containing the crude enzyme extract from E. coli transformant harboring pARF425 by HPLC after being coupled with 2-aminopyridine (Fig. 2B, top). The HPLC analysis using the crude enzyme extract from E. coli transformant harboring pARF421, pARF422, or pARF424 instead of pARF425 exhibited a similar pattern (data not shown). On the other hand, pyridylaminated fucose was detected in the hydrolysate of the reaction mixture containing the crude enzyme extract from E. coli transformant harboring pARF400 by HPLC (Fig. 2B, bottom). A similar HPLC profile was exhibited when using the crude enzyme extract from E. coli transformant harboring pARF420, pARF423, or pARF426 instead of pARF400 (data not shown).
To determine relative retention times, pyridylaminated xylose was added to each sample as an internal standard. These results suggested that dTDP-4-keto-6-deoxy-D-glucose reductase is encoded by ORF14. Fucose was not detected in the reaction mixtures lacking one of the crude enzyme extracts from E. coli SØ874 expressing rmlA, rmlB, or ORF14. We therefore renamed ORF14 fcd according to the new gene nomenclature of Reeves et al. (33). These results were summarized in Fig. 2A.
Purification and Characterization of Enzymes Involved in the Synthesis of dTDP-D-Fucose-To unequivocally determine the function of the rmlA, rmlB, and fcd products, these proteins were purified by affinity chromatography as described in detail under "Experimental Procedures." The proteins were homogeneous as judged by SDS-PAGE (Fig. 3). The molecular masses of the denatured polypeptides determined by SDS-PAGE of 32, 39, and 27 kDa agree well with the predicted molecular masses of the rmlA, rmlB, and fcd gene products, respectively.
An in vitro assay was carried out with these purified enzymes. The conversion of D-glucose-1-phosphate and dTTP to dTDP-D-glucose was observed in the reaction catalyzed by the rmlA product (Fig. 4B). dTDP-4-Keto-6-deoxy-D-glucose was detected as a broad peak in the reaction mixture containing the purified products of both rmlA and rmlB as enzymes (Fig. 4C). The reason why the peak of this intermediate is apparently broad and poor is unknown. dTDP-D-Fucose was detected in the reaction mixture containing all three purified proteins, the rmlA, rmlB, and fcd products (Fig. 4D). The retention time of the dTDP-D-glucose formed was compared with that obtained for a commercially available standard. However, a standard for dTDP-4-keto-6-deoxy-D-glucose could not be purchased from any company. Therefore, the standard for dTDP-4-keto-6-deoxy-D-glucose was synthesized using the purified rmlA and rmlB products of E. coli K-12 (34). 1 H NMR analysis and gas-liquid chromatography analysis were performed to confirm the structure and configuration of this final product. NADPH was not detected under the HPLC condition used.
Determination of the Absolute Configuration of the Fucosyl Residue in the Final Product-On the basis of the structure and configuration of dTDP-4-keto-6-deoxy-D-glucose, we can predict that the absolute configuration of the fucosyl residue in dTDPfucose is D. However, the prediction was not supported by direct evidence. Gas-liquid chromatography analysis was performed to prove the hypothesis. After dTDP-fucose was hydrolyzed, fucosyl residue in dTDP-fucose was detected as a D-(ϩ)-2-octylglycoside acetate. For this fucosyl residue, four peaks were obtained, the two pyranosides and the two furanosides. Relative retention times of four peaks obtained from fucosyl residue in dTDP-fucose agreed with those obtained not from L-fucose but from D-fucose (Fig. 6). Thus the D absolute configuration of the fucose in dTDP-fucose was determined.
Double reciprocal plots of the initial velocity as a function either of dTDP-4-keto-6-deoxy-D-glucose concentration at different NADPH concentrations or of NADPH concentration at different dTDP-4-keto-6-deoxy-D-glucose concentrations yielded apparently nonparallel lines (Fig. 7A). This pattern indicates a sequential mechanism. From the secondary plots (Fig. 7, B and C), the K m values for dTDP-4-keto-6-deoxy-D-glucose and NADPH and the V max value (turnover number) were estimated to be 97.3 M, 28.7 M, and 7.74 M⅐min Ϫ1 , respectively. Because the concentration of the enzyme added to the reaction mixture was 2.97 nM, the K cat value was estimated to be 2.61 ϫ 10 3 min Ϫ1 . DISCUSSION Previously, we identified a gene cluster associated with the biosynthesis of the serotype-specific polysaccharide antigen consisting of L-rhamnose and D-fucose from A. actinomycetemcomitans Y4 (serotype b) (22). D-Fucose is a component of some bacterial polysaccharides (19,20). This sugar, however, is not widespread in nature. The biosynthetic precursor of D-fucose is believed to be dTDP-D-fucose, but little attempt has been made to isolate the enzymes involved in the conversion of D-glucose-1-phosphate and dTTP to dTDP-D-fucose or the genes coding for the enzymes. On the other hand, the synthesis of dTDP-Lrhamnose has been investigated extensively (15)(16)(17)(18). The dTDP-L-rhamnose synthetic pathway consists of four enzymes catalyzing the reaction sequence from D-glucose-1-phosphate and dTTP to dTDP-L-rhamnose through three intermediates, and the rmlA, rmlB, rmlC, and rmlD genes coding for the four enzymes have been identified (24, 34, 36 -38) (Fig. 1). In this paper, we identified the genes required for the synthesis of dTDP-D-fucose and dTDP-L-rhamnose and showed that dTDP-D-fucose was formed by the reduction of dTDP-4-keto-6-deoxy-D-glucose, an intermediate in the dTDP-L-rhamnose biosynthetic pathway, at C-4 ( Fig. 1).
We initially identified the four genes required for the synthesis of dTDP-L-rhamnose in A. actinomycetemcomitans. As expected from the amino acid sequences of their products, ORF7, ORF6, ORF9, and ORF8 were suggested to encode D-glucose-1-phosphate thymidylyltransferase, dTDP-D-glucose 4,6dehydratase, dTDP-4-keto-6-deoxy-D-glucose 3,5-epimerase, and dTDP-4-keto-L-rhamnose reductase, respectively. On the other hand, the identity of dTDP-4-keto-6-deoxy-D-glucose reductase, which forms dTDP-D-fucose, was not predicted from amino acid sequence homology. Instead, the dTDP-D-fucose synthesis gene was located by detecting fucose coupled with 2-aminopyridine in the reaction mixture containing crude extracts from E. coli strains harboring rmlA, rmlB, and one of the eight deletion plasmids (Fig. 2). This experiment led us to conclude that the ORF14 gene encodes a dTDP-4-keto-6-deoxy-D-glucose reductase. We renamed ORF14 the fcd gene. The amino acid sequence deduced from the fcd gene showed an identity of Ͻ10% with previously reported proteins in a homology search using the program FASTA. However, a BLAST search showed that the amino acid sequence of the fcd gene product shared limited homology with that of CDP-abequose synthase of Salmonella enterica LT2 (39). CDP-Abequose synthase catalyzes a similar reaction, in which the keto group on carbon 4 of a hexosyl group is attacked. The difference between the two reductases lies in their substrates. dTDP-4-Keto-6deoxy-D-glucose is converted to the L-manno configuration by dTDP-4-keto-6-deoxy-D-glucose 3,5-epimerase and is then reduced by dTDP-4-keto-L-rhamnose reductase, the rmlD product, to form dTDP-L-rhamnose (Fig. 1). The fcd product did not share homology either with the amino acid sequence of the rmlD product or with those of any previously reported NDP-6deoxyhexosyl-4-ulose reductases except CDP-abequose synthase of S. enterica. An NADPH binding motif GXGXXA (40) was found in the N-terminal region in the fcd product but not in the corresponding region of CDP-abequose synthase.
HPLC analysis of dTDP-D-fucose formed by the purified enzymes showed that D-glucose-1-phosphate was converted to dTDP-D-glucose by the rmlA product, and this was then converted into dTDP-4-keto-6-deoxy-D-glucose by the rmlB product. These reactions are common to the first two steps of the dTDP-L-rhamnose biosynthetic pathway (Fig. 1). The retention times of the sugar nucleotide formed by the rmlA product and that of its derivative converted by the rmlB product agreed with those of dTDP-D-glucose formed by purified glucose-1phosphate thymidylyltransferase from E. coli DH5␣ and dTDP-4-keto-6-deoxy-D-glucose formed by E. coli dTDP-D-glucose 4,6- dehydratase, respectively (data not shown). In addition, dTDP-4-keto-6-deoxy-D-glucose formed by E. coli glucose-1-phosphate thymidylyltransferase and dTDP-D-glucose 4,6-dehydratase was converted to a sugar nucleotide that was eluted at the same retention time as that of dTDP-D-fucose produced by the three A. actinomycetemcomitans enzymes. These results also support the prediction that the rmlA and rmlB products of A. actinomycetemcomitans Y4 are glucose-1-phosphate thymidylyltransferase and dTDP-D-glucose 4,6-dehydratase, respectively, and that the fcd product reduces dTDP-4-keto-6-deoxy-D-glucose to form dTDP-D-fucose.
dTDP-4-Keto-6-deoxy-D-glucose reductase catalyzes a bimolecular group reduction reaction. The mechanisms of such reactions are divided into two major categories, namely, the sequential and ping-pong mechanisms, based on the order of interaction between the enzyme and its substrates (31,32). In contrast to the ping-pong mechanism, in the sequential mechanism all the substrates must combine to form a ternary complex before the product is formed. Our kinetic studies of NADPH and dTDP-4-keto-6-deoxy-D-glucose are consistent with its action by a sequential mechanism (Fig. 7A). The K m and V max values cannot be compared with those of other similar enzymes because no other NDP-6-deoxyhexosyl-4-ulose reductases have been characterized.
It is very interesting to note that the primary structure of dTDP-4-keto-6-deoxy-D-glucose reductase from A. actinomycetemcomitans Y4 exhibits little homology with other NDP-6deoxyhexosyl-4-ulose reductases when considering its origin and evolution. There are only a few reports on D-fucose as a component of bacterial polysaccharide. Furthermore, of the five serotype antigens (a-e) of A. actinomycetemcomitans strains, only the serotype b antigen contains D-fucose. Moreover, Southern blot analysis indicated that the fcd gene is specific to serotype b (data not shown). It is, however, unlikely that this gene evolved in A. actinomycetemcomitans Y4. The GϩC content (37.7%) of the operon essential for producing serotypespecific antigen is lower than the average GϩC content (45.6%) of A. actinomycetemcomitans Y4 chromosomal DNA (22). This low GϩC content suggests that the gene cluster was transferred from another species. A lower GϩC content has been found in many gene clusters involved in the synthesis of various bacterial polysaccharides (34,37,41,42). In particular, the region containing ORF14-ORF19 has the lowest GϩC content (27.0%), and the discrepancy of the GϩC content within the gene cluster suggests that the fcd and dTDP-L-rhamnose synthetic genes have different origins. In addition, the amino acid sequences of the proteins encoded by genes around the fcd gene are not homologous with any reported proteins. These findings suggest that the fcd gene evolved in an unknown gene cluster involved in polysaccharide synthesis in some other bacteria and was transferred to A. actinomycetemcomitans.
6-Deoxy-D-talose, a component of serotype a-specific antigen, and 6-deoxy-L-talose, a component of serotype c-specific antigen, are also rare 6-deoxyhexoses (43). The physiological significance of the rarity of 6-deoxyhexoses in the serotype-specific antigens of A. actinomycetemcomitans remains to be elucidated.