Characterization of dTDP-4-dehydrorhamnose 3,5-Epimerase and dTDP-4-dehydrorhamnose Reductase, Required for dTDP-l-rhamnose Biosynthesis in Salmonella enterica Serovar Typhimurium LT2*

The thymidine diphosphate-l-rhamnose biosynthesis pathway is required for assembly of surface glycoconjugates in a growing list of bacterial pathogens, making this pathway a potential therapeutic target. However, the terminal reactions have not been characterized. To complete assignment of the reactions, the four enzymes (RmlABCD) that constitute the pathway in Salmonella enterica serovar Typhimurium LT2 were overexpressed. The purified RmlC and D enzymes together catalyze the terminal two steps involving NAD(P)H-dependent formation of dTDP-l-rhamnose from dTDP-6-deoxy-d-xylo-4-hexulose. RmlC was assigned as the thymidine diphosphate-4-dehydrorhamnose 3,5-epimerase by showing its activity to be NAD(P)H-independent. Spectrofluorometric and radiolabeling experiments were used to demonstrate the ability of RmlC to catalyze the formation of dTDP-6-deoxy-l-lyxo-4-hexulose from dTDP-6-deoxy-d-xylo-4-hexulose. Under reaction conditions, RmlC converted approximately 3% of its substrate to product. RmlD was unequivocally identified as the thymidine diphosphate-4-dehydrorhamnose reductase. The reductase property of RmlD was shown by equilibrium analysis and its ability to enable efficient biosynthesis of dTDP-l-rhamnose, even in the presence of low amounts of dTDP-6-deoxy-l-lyxo-4-hexulose. Comparison of 23 known and predicted RmlD sequences identified several conserved amino acid residues, especially the serine-tyrosine-lysine catalytic triad, characteristic for members of the reductase/epimerase/dehydrogenase protein superfamily. In conclusion, RmlD is a novel member of this protein superfamily.

Bacterial cell-surface glycoconjugates are essential for survival of pathogenic bacteria and interactions between bacteria and host. Consequently, there is reason to believe that inhibitors directed against target reactions in surface glycoconjugate assembly may provide viable alternate therapeutic approaches. However, bacterial cell surface glycoconjugates show remarkable structural diversity due to variations of the sugar components, linkages, and substitutions. A successful strategy requires identification of enzymes and pathways unique to bacteria, yet present within a wide spectrum of bacterial species. One such target is the synthesis of the activated form of L-rhamnose, dTDP 1 -L-rhamnose. L-Rhamnose is found in polysaccharides from strains of important human pathogens such as Salmonella, Shigella, Burkholderia, and streptococci, as well as in plant-associated bacteria including Xanthomonas and Rhizobium. The primary structures of many of these glycoconjugates have been reported (see the Complex Carbohydrate Structure Data base). L-Rhamnose is also found in many surface layer glycoproteins from Bacillaceae (1), in the linkage unit that joins the mycolylarabinogalactan complex to peptidoglycan in mycobacteria (2) and in some mycobacterial glycopeptidolipids (3). Examination of gene data bases also indicates the presence of the structural genes for enzymes involved in dTDP-L-rhamnose synthesis in strains where the rhamnosecontaining structure is not necessarily resolved, for example, in Enterococcus faecalis (4), Leptospira interrogans serovar Copenhageni (5), and some members of the archaea.
The pathway for the biosynthesis of dTDP-L-rhamnose from glucose 1-phosphate and thymidine triphosphate was proposed in the early 1960's by Glaser and Kornfeld (6,7) although the enzymes were not specifically identified. Genetic data indicates that the pathway requires four genes, rmlABCD with the prototypes being identified in the lipopolysaccharide O-antigen biosynthesis (rfb) gene cluster from Salmonella enterica serovar Typhimurium LT2 (8).
ʈ Supported by postdoctoral fellowships from the Medical Research Council of Canada and the Natural Sciences and Engineering Research Council.
Recently, the first two enzymes in the pathway, glucose-1phosphate thymidyltransferase (RmlA) (11) and dTDP-D-glucose 4,6-dehydratase (RmlB) (12) were analyzed in detail. Overexpressed gene products were used for enzymatic synthesis of dTDP-6-deoxy-D-xylo-4-hexulose and dTDP-L-rhamnose with varying yield (12,13). Although RmlC and D are required for the conversion of dTDP-6-deoxy-D-xylo-4-hexulose to dTDP-L-rhamnose, the definitive assignment of individual activities has not been done, and the mechanism has not been elucidated. This is a clear limitation for studies where inhibitor development is the ultimate goal.

Analytical Techniques
Nucleotide-activated sugars were analyzed on a CarboPac PA-1 column by the method of Köplin et al. (14). Monosaccharides were analyzed on a CarboPac PA-1 column as described previously (15). Determination of the molecular weight of native proteins was done on a UltroPac TSK G3000SW column (LKB, 7.5 ϫ 300 mm) in 0.2 M ammonium acetate, pH 7.0. Bovine serum albumin (67,000), ovalbumin (43,000), chymotrypsinogen A (25,000), and RNase A (13,700) were used as reference proteins. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed using slight modifications (16) of the original method of Laemmli (17). Nondenaturing PAGE was performed in 50 mM Tris/citrate buffer at pH 7.0. The resolving gels contained 12.5%, and the stacking gels 5% polyacrylamide (3% crosslinking). Ammonium persulfate and N,N,NЈ,NЈ-tetramethylethylenediamine were removed by prerunning gels at 200 V for 1 h with two changes of Tris/citrate buffer. Electrophoresis was performed using a constant voltage (200 V) at 10°C for 4 h. Gels were silver-stained using a method described elsewhere (18). Protein concentration was determined by the method of Bradford (19). Fluorescence spectroscopy was performed at 25°C in a Hitachi F-2000 spectrofluorometer and at excitation and emission wavelengths as indicated. Radioactivity counting was done in Ultima Gold liquid scintillation mixture (Packard Instrument Co., Meriden, CT) on a Packard Tri-Carb 1900 CA liquid scintillation counter.
Polymerase chain reaction amplifications were performed using a GeneAmp polymerase chain reaction system 2400 (Perkin-Elmer, Norwalk, CT). Chromosomal DNA from S. enterica LT2 (1 g/l) was obtained from Dr. Wendy J. Keenleyside. For the amplification reactions, 2.5 units of PwoI DNA polymerase (Roche Molecular Biochemicals), 0.5 g of template DNA, 200 mol of each deoxynucleotide triphosphate, and 25 mol of the corresponding synthetic nucleotide primers were used. Twenty cycles were used. Optimal MgSO 4 concentration and reaction temperatures and times were individually determined for each polymerase chain reaction product. The product sizes for rmlA (943 base pairs), rmlB (1, 222 base pairs), rmlC (627 base pairs), and rmlD (966 base pairs) were those predicted from their respective nucleotide sequences.
The amplified fragments were digested with the appropriate restriction endonucleases, according to the manufacturer's recommendations and cloned in plasmids pET-28a(ϩ) (rmlB) or pET-30a(ϩ) (rmlA,C,D) (Novagen, Madison, WI). The procedures for manipulation of DNA and for ligation were those described by Sambrook et al. (20). Electrotransformations of E. coli DH5␣ and BL21 (DE3) were performed using published protocols (21) with a Gene Pulser apparatus (Bio-Rad). Transformants were selected on LB/K m plates. Plasmids were columnpurified using QIAGEN spin columns (QIAGEN, Chatsworth, CA) according to the manufacturers instructions. The sequence of each cloned rml gene was determined and confirmed to be identical to the authentic chromosomal copy. Plasmid DNA sequencing was performed by automated sequencing at the Guelph Molecular Supercentre.

Sequence Analyses
Nucleotide and protein sequences were analyzed using online analysis tools, including BLAST (Basic local alignment search tool; Ref. 22) and Clustal W 1.6 (Refs. 23 and 24) using default parameter settings. For protein family analysis and motif analysis, the Prosite data base (25) was used.

Expression of Genes in the rml Operon
The pET vector-based constructs place the cloned rml genes under the control of a T7-promoter. For overexpression of the rml gene products, an isopropyl-1-thio-␤-D-galactopyranoside-inducible T7 RNA polymerase is supplied by the host, E. coli BL21 (DE3). Isopropyl-1-thio-␤-D-galactopyranoside induction of E. coli BL21 (DE3) transformed with pET constructs carrying each rml gene, led to expression of RmlA, RmlB, RmlC, and RmlD as the predominant proteins in the induced cultures. For large-scale enzyme preparations, up to 500 ml of LB/K m medium was inoculated with 1% of an overnight culture and cultivated at 37°C on a rotary shaker. After 3 h, isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 1 mM and incubation was continued for 3 h. The bacterial cells were then collected by centrifugation at 7,000 rpm for 6 min at 4°C. The cell pellets were washed twice with cold 10 mM Tris-HCl buffer, pH 8.0. Before sonication, the buffer was adjusted by addition of dithiothreitol and MgCl 2 to final concentrations of 1 and 10 mM, respectively. Sonication on ice was carried out 6 times for 10 s with intervals of 20 s. Large debris were removed by centrifugation at 5,000 rpm for 10 min at 4°C. The cell-free supernatants were carefully removed and centrifuged at 60,000 rpm for 40 min at 4°C in a Ti 70.1 rotor in a Beckman LE-80 ultracentrifuge. Supernatants were stored at Ϫ20°C following addition of glycerol to a final concentration of 44%.
The specific assay for determination of RmlC typically contained 45 mM potassium phosphate buffer, pH 7.0, 9 mM MgCl 2 , 0.18 mM dTDP-6-deoxy-D-xylo-4-hexulose, 0.072 mM NAD(P)H, a 20-fold molar excess of RmlD for determination of RmlC, and an appropriate volume of RmlC in a total volume of 550 l. Assays for RmlD were done using a similar reaction mixture but incorporating a 100-fold molar excess of RmlC. Time absorption plots were recorded at 25°C in a Beckman DU65 spectrophotometer or a Hitachi U-2010 spectrophotometer. Enzyme activities were calculated from the linear decrease of the absorption at 340 nm. Assays with low amounts of NAD(P)H were analyzed in a Hitachi F-2000 spectrofluorometer with an excitation at 340 nm and recording emission at 460 nm. One unit of RmlC or RmlD activity refers to 1 mol of NAD(P)H consumption per minute using these assay conditions. Controls without addition of dTDP-6-deoxy-D-xylo-4-hexulose, or enzyme, were analyzed by the same procedure.

Equilibrium Analysis
For determination of the thermodynamic equilibrium constants of the reaction described in reaction steps 3 and 4, the reverse reaction starting from dTDP-L-rhamnose was used. Assays contained 0.05 to 0.18 mM dTDP-L-rhamnose, 0.03 to 0.4 mM NAD ϩ , 0.16 M RmlD in 50 mM potassium phosphate buffer, pH 7.0, or 50 mM ethanolamine-HCl buffer, pH 9.0. The assays were performed with, and without, addition of RmlC (0.6 M). Equilibrium constants were calculated as:

pH Optima
For determination of the pH dependence of enzyme activity, RmlC and RmlD were assayed in different buffers ranging from pH 5.5 to 11.0 (50 mM potassium phosphate buffer, pH 5.5-8.5, and 50 mM ethanolamine-HCl buffer, pH 8.5-11.0) without addition of MgCl 2 . To determine the profile of enzyme stability versus pH, enzymes were diluted in various buffers (potassium phosphate buffer, pH 5.5-8.0) and incubated at 37°C for 20 min. Residual activities were determined at pH 7.0 after exchange of the buffer.

Kinetic Analyses
Measurements of kinetic data were performed using the spectrophotometric and spectrofluorometric assays described above. The kinetic constants, K m and k cat , where K m is the apparent Michaelis constant and k cat is the turnover number, were obtained by fitting the experimental data to the equation (27), using the program Sigma Plot (Jandel, Erkrath, Germany). k cat (1/s) was calculated from the maximum initial velocity v max (mol/(liter⅐s)) as, where E is the total enzyme concentration (mol/liter). A molecular mass of 20.6 and 32.6 kDa was used for RmlC and RmlD, respectively. For RmlC, the initial velocities were recorded for the concentration range 0.036 to 0.36 mM dTDP-6-deoxy-D-xylo-4-hexulose. The RmlC concentration was 16.6 nM and RmlD was used in 20-fold molar excess. For RmlD, NADH and NADPH concentrations were 0.0015 to 0.109 mM and dTDP-6-deoxy-D-xylo-4-hexulose concentration was 0.18 mM. The RmlD concentration was 7.5 nM and RmlC was used in 100-fold molar excess. Activities at NAD(P)H concentrations below 0.01 mM were measured spectrofluorometrically. For the analysis of the reaction mechanism of RmlD the concentrations of NADPH and dTDP-6-deoxy-D-xylo-4-hexulose were varied between 0.007 and 0.11 mM, and 0.091 and 0.36 mM, respectively. The experimental data were fit to, for a sequential reaction mechanism (27,28). For an ordered mechanism, k I represents the dissociation constant of the first substrate. The correlation coefficients of nonlinear regression were usually 0.98 or better. Variance analysis and other statistics provided by the program showed that Equation 3 adequately fits the data.

Purification of dTDP-4-dehydrorhamnose 3,5-Epimerase
To prevent oxidative damage of proteins during the purification process, all buffers contained 0.5 mM dithiothreitol. Fractions were collected on ice.
Step 1: Hydroxyapatite Chromagraphy-Cell-free lysates were applied to a Bio-Gel HT column (1.6 ϫ 10 cm) at a flow rate of 1 ml/min. Proteins were eluted using the following gradient: 0 -20 ml, 10 mM potassium phosphate buffer, pH 6.8; 20 -100 ml, 0 -100% 200 mM potassium phosphate buffer, pH 6.8. Two-ml fractions were collected. Fractions showing enzyme activity eluted at approximately 0.06 M potassium phosphate and these were combined and adjusted to 1 M ammonium sulfate.
Step 2: Hydrophobic Interaction Chromatography-Pooled enzyme fractions from step 1 were applied to phenyl-Superose HR 5/5 column equilibrated with 50 mM potassium phosphate buffer, pH 7.0, containing 1 M ammonium sulfate. Proteins were eluted at a flow rate of 0.5 ml/min using the following gradient: 0 -10 ml, 1 M ammonium sulfate; 10 -30 ml: 1-0 M ammonium sulfate. One-ml fractions were analyzed for enzyme activity. Enzymatically active fractions eluted at 0.2 M ammonium sulfate and these were pooled and dialyzed overnight at 4°C against several exchanges of 20 mM Tris-HCl buffer, pH 7.7.
Step 3: Ion Exchange Chromatography-The enzyme preparation from step 2 was applied to a Mono Q HR 5/5 column equilibrated with 20 mM Tris-HCl buffer, pH 7.7. Proteins were eluted at a flow rate of 1 ml/min by the following gradient: 0 -10 ml, 20 mM Tris-HCl buffer, pH 7.7; 10 -30 ml, 0 to 0.5 M KCl. Fractions showing enzyme activity eluted at 0.28 M KCl. These were combined, adjusted to 44% glycerol, and stored at Ϫ20°C until use. The purified RmlC enzyme preparation (86% yield; approximately 12.5 mg/liter culture) showed a specific activity of approximately 21 units/mg.

Purification of dTDP-4-dehydrorhamnose Reductase
Step 1: Cibacron Blue-Sepharose Chromatography-Cell-free Lysates-Cell-free lysates (stored without addition of glycerol) were diluted 1:1 in 50 mM Tris-HCl buffer, pH 7.7, and up to 10-ml aliquots were applied to a Cibacron blue-Sepharose CL-4B column (1.5 ϫ 5 cm) at a flow rate of 2.5 ml/min. The column was washed with 50% ethylene glycol in 50 mM Tris-HCl buffer, pH 7.7, and 0.3 M KCl in the same buffer. Most of the proteins in the lysate did not bind to the matrix, while RmlD was absorbed and subsequently eluted in 1.5 mM KCl. The active fractions were dialyzed overnight as described.

Synthesis of Nucleotide Activated Monosaccharides
Reaction mixtures for synthesis of dTDP-6-deoxy-D-xylo-4-hexulose (12,13) contained approximately 20 mol of dTDP-D-glucose and 2 units of RmlB in 1 ml of 20 mM Tris-HCl buffer, pH 7.7. Incubation was performed at 25°C for 1 h. The reaction was stopped by addition of 1 ml of absolute ethanol and precipitated proteins were removed by centrifugation. The product was desalted on a Sephadex G-10 column (1.5 ϫ 120 cm), lyophilized, and stored at Ϫ20°C until use. UV spectroscopy at 320 nm performed in 0.1 M NaOH was used to determine concentration of dTDP-6-deoxy-D-xylo-4-hexulose. Quantitation is based on the characteristic absorption of the 4-keto group. Following reduction with NaBH 4 and hydrolysis, the monosaccharides obtained from dTDP-6deoxy-D-xylo-4-hexulose were analyzed by high performance anion exchange chromatography with pulsed electrochemical detection (HPAEC/PED) (15).
For synthesis of dTDP-6-deoxy-L-lyxo-4-hexulose, 1 mol dTDP-Dglucose was incubated at 25°C for 30 min with 1 unit each of RmlB and RmlC in 500 l of 30 mM Tris-HCl buffer, pH 7.0, at 25°C for 30 min. The enzymes were removed by ultrafiltration and the reaction mixture containing both dTDP-6-deoxy-D-xylo-4-hexulose and dTDP-6-deoxy-Llyxo-4-hexulose was desalted as described above. The amount of dTDP-6-deoxy-L-lyxo-4-hexulose in the resulting preparation was determined by conversion to dTDP-L-rhamnose in the presence of RmlD and spectrofluorometric measurement of the decrease of NADH. Direct proof for the presence of free dTDP-6-deoxy-L-lyxo-4-hexulose was obtained by reducing the 4-keto group with NaB[ 3 H] 4 (100 mCi/mmol, American Radiolabeled Chemicals, St. Louis, MO). Following hydrolysis and HPAEC/PED of the resulting monosaccharides, fractions (0.25 ml) were analyzed for radioactivity. dTDP-L-rhamnose was synthesized from dTDP-D-glucose using RmlB, RmlC, and RmlD in the presence of low amounts of NAD ϩ . NADH was regenerated using NAD ϩ -dependent formate dehydrogenase from Candida boidinii (ASA Spezialenzyme GmbH, Braunschweig, Germany). A typical reaction mixture contained 20 mol of dTDP-Dglucose, 1 mol of NAD ϩ , 100 mol of ammonium formate, 1 unit of each RmlB, RmlC, and RmlD, and 3.5 units of formate dehydrogenase in a total volume of 5 ml of 0.1 M Tris-HCl buffer, pH 7.0. After reaction at 25°C for 2 h, proteins were removed by ultrafiltration in an Amicon model 8050 cell using a Millipore PLGC 10-kDa ultrafiltration membrane. dTDP-L-rhamnose was purified by anion exchange chromatography on a DEAE-Sephacel column as described previously (15) and desalted as described above.

Expression of the Cloned Protein Products in E. coli-Isopro-
pyl-␤-D-thio-galactopyranoside-induction of E. coli BL21 (DE3) cells, transformed with pET-derivatives carrying each of the rmlABCD genes, led to expression of RmlA, RmlB, RmlC, and RmlD as the predominant proteins in the induced cultures (Fig. 1). Limited amounts of these proteins were also present in the noninduced cultures (not shown). The majority of each Rml protein was found in the cell-free supernatant fraction. This, together with their activities (see below), indicates that these proteins are located in the cytoplasm. The residual amounts of RmlC and RmlD were sedimented during removal of cellular debris from the ultrasonicated cell lysate. Electron microscopy of ultrathin-sectioned intact cells confirmed formation of inclusion bodies (not shown), presumably containing the enzyme confined to the particulate fraction. Since sufficient amounts of soluble protein was obtained, inclusion bodies were not pursued further for protein purification.
The expressed Rml enzymes showed in SDS-PAGE apparent molecular masses of 31.8, 42.9, 23.4, and 34.8 kDa, respectively. These values correspond closely to the predicted translation products based upon available nucleotide sequence data (32.5, 40.7, 20.6, and 32.6 kDa, respectively) (8). By gel permeation chromatography on an UltroPac TSK G3000SW column, the native molecular mass of RmlC was determined to be 40.6 kDa. This is consistent with RmlC forming a dimer (calculated molecular mass ϭ 41.2 kDa). The molecular mass of RmlD, determined on a silica column was 41.5 kDa, a value inconsistent with the calculated molecular mass for either the monomer (32.6 kDa), or the dimer (65.2 kDa). The aberrant molecular mass estimated for RmlD may result from interactions between RmlD and the chromatographic resin. For example, the hydrophobic protein chymotrypsinogen A has been shown to bind tightly to silica matrices resulting in retarded elution and accordingly a lower apparent molecular mass (29). Comparison of chymotrypsinogen A and RmlD by analytical hydrophobic interaction chromatography on a phenyl-Superose HR 5/5 column revealed that RmlD elutes at even lower (NH 4 ) 2 SO 4 concentrations than chymotrypsinogen A (0.5 and 0.1 M, respectively), indicating higher hydrophobicity of RmlD. In conclusion, the apparent molecular mass of RmlD, as determined by size exclusion chromatography to be 41.5 kDa, may correspond to at least a dimer.
Activity of the Overexpressed Rml Enzymes-The correct folding and functionality of the overexpressed dTDP-L-rhamnose biosynthesis enzymes was tested using individual spectrophotometric assays for each enzyme. All four enzymes are active, and the activities in supernatants of the expression assays ranged from 2 to 10 units/mg of protein.
Using cell-free lysates from the strain overexpressing RmlA, dTDP-D-glucose was synthesized from glucose-1-P and dTTP. dTDP-6-deoxy-D-xylo-4-hexulose was synthesized from dTDP-D-glucose using RmlB, as expected from work reported by others (12,13). The yield of dTDP-6-deoxy-D-xylo-4-hexulose was 100% after reaction. The product was purified by anion exchange high performance liquid chromatography on a Car-boPac PA-1 column and the final yield after purification was approximately 95%. Following reduction and subsequent hydrolysis, 6-deoxyglucose and 6-deoxygalactose (fucose) were formed from the reaction product. Each component was identified by comparison with internal and external standards. The products are consistent with reduction of the 4-keto group of To prove the ability of RmlC to produce dTDP-6-deoxy-L-lyxo-4-hexulose, dTDP-D-glucose was reacted with RmlB and RmlC. The purified reaction mixture contained both dTDP-6-deoxy-Dxylo-4-hexulose and dTDP-6-deoxy-L-lyxo-4-hexulose in a ratio of approximately 97:3, as shown by a spectrofluorometric assay. Consistent with this result, radiolabeling of the 4-keto group with NaB[ 3 H] 4 , followed by hydrolysis of the labeled nucleotide-bound monosaccharides, resulted in two major and one minor radioactive peaks that could be separated by HPAEC/PED (Fig. 2). Using external and internal standards, the major peaks were identified as fucose and 6-deoxyglucose (products of dTDP-6-deoxy-D-xylo-4-hexulose), and the minor one as rhamnose (product of dTDP-6-deoxy-L-lyxo-4-hexulose). However, 6-deoxytalose, the second product of dTDP-6-deoxy-L-lyxo-4-hexulose, was not detected; this may coelute with one of the other monosaccharides.
For synthesis of dTDP-L-rhamnose, dTDP-D-glucose was reacted with RmlB, C, and D. For in situ regeneration of NADH, a formate dehydrogenase catalyzing the NAD ϩ -dependent oxidation of formate was used (Fig. 3, track A). The yield of dTDP-L-rhamnose in this reaction was 100%. After removal of the proteins and salts the reaction mixture contained approximately 5% of NADH (Fig. 3, track B), and this was removed from the reaction mixture using anion exchange chromatography. Purified dTDP-L-rhamnose (yield 80%) contained no other nucleotide as evidenced by the single peak in HPAEC analysis on a CarboPac PA-1 column (Fig. 3, track C). Identity of dTDP-L-rhamnose was demonstrated by monosaccharide analysis of the hydrolyzed product.
To prove the substrate specificity of the three enzymes converting dTDP-D-glucose to dTDP-L-rhamnose (RmlB, RmlC, and RmlD), UDP-D-glucose, CDP-D-glucose, ADP-D-glucose, and GDP-D-glucose were used as substrates instead of dTDP-D-glucose. The activities with all these substrates were lower than 1% of the activity with dTDP-D-glucose.
Distinction between dTDP-4-dehydrorhamnose 3,5-Epimerase and dTDP-4-dehydrorhamnose Reductase-The existing assignments of RmlC and RmlD to specific enzymatic activities are based on sequence data and there is some ambiguity in the literature. Direct assays for the separate measurement of RmlC activity or RmlD activity are not available. However, by using a coupled assay in which RmlC and RmlD are utilized together, and by varying the conditions in the assay, the activities of RmlC and RmlD were clearly distinguished. Quantitation in the assays is based on the consumption of NADPH by the reductase, which is conveniently monitored by a time-dependent decrease in the absorbance at 340 nm. When RmlD was used in 20-fold molar excess over RmlC, the initial reaction velocity of RmlC, determined with 0.18 mM dTDP-6-deoxy-Dxylo-4-hexulose, is not dependent on the NADPH concentration in a range of 0.007-0.109 mM. This result identified RmlC as the epimerase and RmlD as the reductase. Accordingly, with RmlC being used in 100-fold molar excess, the activity of RmlD can be specifically measured.
Conditions for Enzyme Assays and Synthesis of dTDP-activated Monosaccharides-Under the assay conditions employed, RmlC shows maximum activity at pH 7.5 and is stable over a wide pH range. Its pH activity profile is similar to that of RmlD, indicating that their inclusion in a coupled RmlCD assay is not compromised by pH considerations. The pH optimum of RmlD is 6.5, but stability is highly dependent on the pH conditions. Even at pH 7.0 approximately 75% of RmlD activity is lost during incubation at 37°C for 20 min. However, RmlD is more stable at temperatures below 30°C (data not

FIG. 3. High performance liquid chromatography analysis of enzymatic synthesis of dTDP-L-rhamnose.
A, plot showing substrates dTDP-D-glucose and NAD ϩ before addition of enzyme. B, reaction products obtained after addition of RmlB, RmlC, and RmlD to substrates and using formate dehydrogenase for regeneration of NADH. C, purified reaction product, dTDP-L-rhamnose.
shown). Treatment with chelators such as EDTA (up to 0.3 mM) causes a reversible loss of up to 70% of enzyme activity for both RmlC and RmlD. Upon addition of MgCl 2 in concentrations exceeding that of EDTA full activity is restored.
Kinetic Constants for RmlC and Its Potential Interaction with RmlD-For determination of RmlC activity, the assay was coupled with RmlD (20-fold molar excess) producing NADP ϩ from NADPH. The K m value of RmlC for dTDP-6-deoxy-D-xylo-4-hexulose was 0.71 Ϯ 0.17 mM and k cat was 39 Ϯ 6.5 s Ϫ1 (Table  I). There is clear evidence for product formation from dTDP-6deoxy-D-xylo-4-hexulose by RmlC acting in the absence of RmlD. However, dTDP-6-deoxy-L-lyxo-4-hexulose was extremely unstable, decomposing too rapidly to allow its isolation on a preparative scale. Therefore, no kinetic data were obtained for the reverse reaction of RmlC.
One explanation for the very low amount of dTDP-6-deoxy-L-lyxo-4-hexulose, if RmlC is examined in isolation, is that RmlC and RmlD form a complex, converting dTDP-6-deoxy-Dxylo-4-hexulose to dTDP-L-rhamnose. This was proposed in the early work of Melo and Glaser (30), in a scenario where only dTDP-6-deoxy-L-lyxo-4-hexulose bound to the enzyme complex can be reduced by the reductase. dTDP-6-deoxy-L-lyxo-4-hexulose would therefore not exist in free form. To address this hypothesis, mixtures of RmlC and RmlD, with and without substrate, were examined in nondenaturing anionic PAGE. No evidence was found for tight complexes of both enzymes, even in overloaded silver-stained gels (Fig. 4). Additionally, enzyme assays were performed in the presence of Ficoll. Ficoll can act as a macromolecular crowding agent to inhibit the diffusion of proteins (31,32) and should strengthen the interaction of any macromolecular assemblies involving a dTDP-6-deoxy-L-lyxo-4-hexulose⅐RmlC complex. Enzyme activity was not influenced by addition of Ficoll at concentrations of up to 30%, making a complex of RmlC and RmlD, formed during catalysis, very unlikely.
Kinetic Measurements for RmlD-The pH optimum for the reduction of dTDP-6-deoxy-L-lyxo-4-hexulose by RmlD is 6.5. At pH values above 9.0 NAD(P) ϩ -dependent oxidation of dTDP-Lrhamnose was detectable. The activity for the reverse reaction was less than 1% of the activity of the reduction reaction, with a maximum at pH 10.0. In the presence and absence of RmlC, similar initial reaction velocities were detected, although the equilibrium levels of NADH were different. The equilibrium concentration of substrates and products formed in the absence of RmlC allowed calculation of the K eq,RmlD to be 3.6 ϫ 10 13 (Ϯ 1.5 ϫ 10 13 , n ϭ 6) in reaction 4. When RmlC was added, the calculated K eq,RmlC ϫ K eq,RmlD value was 4.5 ϫ 10 11 (Ϯ 1.1 ϫ 10 11 , n ϭ 6). From these results K eq, RmlC was estimated to be 0.013. This value is in agreement with the detection of low amounts of free dTDP-6-deoxy-L-lyxo-4-hexulose in equilibrium with dTDP-6-deoxy-D-xylo-4-hexulose.
Since isolation of dTDP-6-deoxy-L-lyxo-4-hexulose was not feasible due to instability of the product, this intermediate was generated in situ for kinetic analysis of RmlD using a 100-fold molar excess of RmlC. Apparent K m and k cat values for NADH and NADPH were determined with a constant concentration of dTDP-6-deoxy-D-xylo-4-hexulose (Table I). RmlD shows dual coenzyme specificity for NADH and NADPH with a slight preference for NADH. The initial velocities were determined with several fixed concentrations of dTDP-6-deoxy-D-xylo-4-hexulose and varying concentrations of NADPH. Double-reciprocal plots obtained from these data showed an intersecting pattern (not shown), indicating a sequential, ternary complex mechanism of RmlD. A fit of the experimental data to Equation 3 by nonlinear regression is shown in Fig. 5. The constants derived from this analysis were: K m,NADH ϭ 0.21 Ϯ 0.004 mM, K m,hexulose ϭ 0.106 Ϯ 0.018 mM, k cat ϭ 53 Ϯ 4 s Ϫ1 , and k I ϭ 0.012 Ϯ 0.005 mM. At pH 7.0, no reverse reaction was detectable. Therefore, the dissociation constants for NAD ϩ , NADP ϩ , and dTDP-Lrhamnose were determined by fluorescence titration ( excitation at 280 nm and emission at 350 nm). The addition of each component to a solution of RmlD (7.5 nM in 50 mM potassium phosphate buffer, pH 7.0) caused a significant quenching of the intrinsic tryptophan fluorescence of the enzyme. Scatchard analysis of the data allowed calculation of dissociation constants and determination of the numbers of binding sites for NAD ϩ , NADP ϩ , and dTDP-L-rhamnose. K d values were 1.0, 2.3, and 0.37 mM, respectively, and one binding site was predicted for both NAD(P) ϩ and dTDP-L-rhamnose. When both  NAD(P) ϩ and dTDP-L-rhamnose were used sequentially for fluorescence titration they showed additive quenching effects. RmlD Resembles the Reductase/Epimerase/Dehydrogenase (RED) Protein Superfamily-The search for conserved motifs rather than overall sequence identity may reveal family relationships, even when overall primary sequence identity/similarity is not higher than 15-20%. To find conserved regions in RmlD sequences, a multiple sequence alignment was performed using sequences from 23 different bacteria and archaea (Fig. 6). The sequences used for Shigella flexneri 2a (33), Xanthomonas campestris (34), and Saccharopolyspora erythrea (35) have been described as dTDP-4-dehydrorhamnose 3,5-epimerases but are, according to sequence similarities, the corresponding reductases. Some of the sequences have not yet been definitively designated as RmlD or dTDP-4-dehydrorhamnose reductase, but have distinct homologies to the S. enterica LT2 RmlD sequence. All of the RmlD sequences contain a strictly conserved (Y-X 3 -K) motif that is characteristic for the single domain reductase/epimerase/dehydrogenase protein superfamily (RED family, see Refs. 36 and 37). The RED family includes the short-chain dehydrogenases/reductases (38). The (Y-X 3 -K) motif is located within a larger conserved domain, identified in Fig. 6 as "motif 3." Furthermore, the typical coenzyme-binding Rossman fold, containing a modified Wierenga motif (G-X 2 -G-X 2 -G), was identified previously in the amino-terminal part of the RmlD sequence (39). The same motif is present in all other members of this protein family. The crystal structures of two functionally related members of the RED family, UDP-galactose epimerase (40,41) and GDP-4-keto-6-deoxy-D-mannose epimerase/reductase (42,43), have been elucidated recently. The active site of both enzymes was shown to consist of a catalytic triad of serine, tyrosine, and lysine (40,43). The catalytic serine residue is located upstream of the Y-X 3 -K loop. A conserved serine residue has also been found in the RmlD sequences, located in a highly conserved region labeled "motif 2" in Fig. 6. The Swiss Protein data base was searched for the sequence STDYVF, but no additional protein containing this motif was identified. Alignments of RmlD sequences with UDPgalactose epimerase and GDP-4-keto-6-deoxy-D-mannose epimerase/reductase sequences revealed distant relationships (and limited similarity) among these proteins, but the catalytic amino acids are conserved in all of these proteins (not shown). As expected from differences in substrate and coenzyme specificitiy, residues known to be important for substrate or coenzyme-binding in UDP-galactose epimerase and GDP-4-keto-6deoxy-D-mannose epimerase/reductase were not conserved in RmlD. However, presence of the Rossman-fold and the catalytic center residues identify RmlD as a novel member of the RED protein superfamily. DISCUSSION L-Rhamnose is a common constituent of different bacterial glycoconjugates. The biosynthetic precursor, dTDP-L-rhamnose, is formed by the sequential action of the RmlABCD enzymes. Detailed mechanisms have been proposed for synthesis of dTDP-D-glucose by RmlA (11) and its conversion to dTDP-6deoxy-D-xylo-4-hexulose by RmlB (12). The final two steps are catalyzed by RmlC and RmlD and convert dTDP-6-deoxy-Dxylo-4-hexulose to dTDP-L-rhamnose. These activities are now assigned by the studies reported here.