Mechanistic Studies of the Biosynthesis of 3,6-Dideoxyhexoses in Yersinia pseudotuberculosis PURIFICATION AND STEREOCHEMICAL ANALYSIS OF CDP-D-GLUCOSE OXIDOREDUCTASE*

An NAD+-dependent CDP-D-glucose oxidoreductase which catalyzes the first step of the biosynthesis of CDP-ascarylose (CDP-3,6-dideoxy-~-arabino-hex-ose), converting CDP-D-glucose to CDP-4-keto-6- deoxy-D-glucose, was isolated from Yersinia pseudo-tubercu~osis. A protocol consisting of DEAE-cellulose, Matrex Blue-A, hydroxylapatite, DEAE-Sephadex, Sephadex G-100, and NAD+-agarose column chromatography was used to purify this enzyme 6000-fold to homogeneity. This enzyme consists of two identical subunits, each with a molecular weight of 42,500.

The 3,6-dideoxyhexoses are an important class of carbohydrates. They are found, with few exceptions (l), only in the lipopolysaccharide component of the cell wall of Gram-negative bacteria in which they constitute the nonreducing terminal groups of the 0-antigen repeating units (2)(3)(4). Since the immunological specificities of Gram-negative species are mainly determined by the nonreducing sugar entities of these 0-antigen repeating units, the 3,6-dideoxyhexoses have been shown to confer unique serological specificities in many immunological active lipopolysaccharides. Thus, they are commonly referred as immunodominant sugars and/or antigenic determinants in bacteria (5)(6)(7)(8)(9). Biosynthesis of these unusual sugars follows a complex pathway starting with an internal * This work was supported by National Institutes of Health Grant GM-35906. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. oxidation-reduction step mediated by an NAD+-dependent oxidoreductase. This enzyme catalyzes the transformation of a nucleotidyl diphosphohexose (1) to the corresponding 4keto-6-deoxyhexose derivative (2) (10)(11)(12) which, upon further catalysis by a dehydrase and a reductase, is converted to a 3,6-dideoxyhexose as the final product (13-15). It has been shown that this oxidoreductase catalyzed reaction is a key step common to the biosynthesis of many unusual carbohydrates (4).
A number of this type of oxidoreductase have been found in nature. These include the TDP-D-glucose oxidoreductase from Pseudomonas aeruginosa (16)) Escherichia coli (17)(18)(19), Streptomyces rimosus (20)) and Sacchuropolyspora erythraea This class of enzymes is characteristic for an active-site bound nicotinamide coenzyme. As depicted in Scheme I (NDP' represents a generic nucleoside diphosphate group), studies of the homogeneous TDP-D-glucose oxidoreductase from E. coli had shown that this enzymatic reaction proceeds with an oxidation at C-4, a dehydration between C-5 and C-6, and a reduction at C-6 (10-12). Such a transformation is accompanied by an intramolecular hydrogen transfer from C-4 of the substrate to C-6 of the resulting 4-keto-6-deoxyhexose product (17, 28) with the enzyme-bound NAD+ serving as a hydride carrier. An examination of the stereochemical course of the E. coli enzyme (29) found that the displacement of the hydroxyl group at C-6 by the hydrogen from C-4 occurs with inversion, the dehydration from C-5 and C-6 is a syn elimination, and the internal hydrogen transfer to NAD+ is "si face"-specific (30). The same mechanistic conclusions were also reached for the GDP-D-mannose oxidoreductase isolated from an unidentified soil bacterium (ATCC 19241) (31). Thus, enzymes of this type obtained from different sources may have distinct substrate specificity, but they all seem to share a common reaction pathway.
Despite the fact that the mechanistic similarity of enzymes within this class has been well established, studies of the oxidoreductases isolated from S. erythruea (21) and P. pseudotuberculosis (22, 32) showed an absolute requirement of NAD+ for activity that is fundamentally distinct from most of their counterparts. Since the enzyme-bound nicotinamide cofactor in this type of enzyme is known to play a pivotal role in subunit association and its redox state is crucial in regu- The abbreviations used are: NDP, nucleoside diphosphate; El, CDP-4-keto-6-deoxy-~-glucose-3-dehydrase; E3, CDP-6-deo~y-A~~~glucoseen reductase; PMSF, phenylmethanesulfonyl fluoride; ee, enantiomeric excess; THF, tetrahydrofuran; NMO, N-methylmorpholine N-oxide; TEAB, tetraethylammonium bromide. lating substrate binding and product releasing (33, 34), the observation of weak NAD+ binding in the Pasturella and Saccharopolyspora enzymes throws into doubt whether their catalyses still follow the established mechanism. As most of the early mechanistic studies of the Pasturella enzyme were performed on partially purified proteins, and the catalytic properties of the Saccharopolyspora enzyme had not been characterized, a re-examination of these enzymes at the homogeneous stage appeared to be in order. In an effort to study the mechanism of 3,6-dideoxyhexose formation in Yersinia pseudotuberculosis (35-37),' we have purified a CDP-D-glucose oxidoreductase from this bacterial strain whose lipopolysaccharide structure is known to contain ascarylose (3,6dideoxy-L-arabino-hexose) as the end group of its terminal repeating unit (38). Aimed at elucidating its reaction course in more detail, we describe, in this paper, the purification of Y. pseudotuberculosis enzyme 6000-fold to homogeneity and report the stereochemical analysis of this enzymatic process.

Enzyme Purification
Our initial attempts to purify CDP-D-glucose oxidoreductase from Y. pseudotuberculosis followed the procedures reported in the purification of analogous enzymes from other microbial origins, especially the P. pseudotuberculosis enzyme (22). However, examination of SDS-polyacrylamide gels of the proteins isolated by these well documented protocols clearly showed that our initial efforts were futile. This prompted us to develop a new separation procedure initiated by a DEAE-cellulose chromatography. Although most of the key enzymes involved in the biosynthesis of ascarylose were collected during the gradient elution (36, 39), the desired oxidoreductase was eluted only after washing isocratically with 200 mM of the potassium phosphate buffer. As summarized in Table I, further separation by Matrex Blue-A, hydroxylapatite, DEAE-Sephadex, Sephadex G-100, and NAD+-agarose chromatography permitted a nearly 6000-fold overall purification of CDP-D-glucose oxidoreductase to homogeneity. This highly purified enzyme exhibited a single band upon SDS-polyacrylamide gel electrophoresis (Fig. 1).
Portions of this paper (including "Materials and Methods,"  ND, not determined.
e Crude extracts were obtained from 95 g of wet cells. Properties of CDP-D-glucose Oxidoreductase Molecular Weight-The native molecular weight of the highly purified CDP-D-glucose oxidoreductase was estimated by gel filtration to be 86,000, while SDS-polyacrylamide gel electrophoresis of this pure protein showed a single band with a molecular weight of 42,500. It is apparent that this oxidoreductase is a dimeric protein consisting of two identical subunits.
UV-uisible Spectrum-The electronic spectrum of CDP-Dglucose oxidoreductase is a spectrum of a single polypeptide which shows no absorption at wavelength above 300 nm.
Amino Acid Composition and Amino-terminal Sequence-The amino acid composition and the amino-terminal sequence are summarized in Table 11. This sequence has been used in designing an oligonucleotide primer as a probe for cloning this enzyme's gene.
Substrate Specificity-The specificity of CDP-D-glucose oxidoreductase for CDP-D-glucose was examined by substituting alternate substrates for CDP-D-glucose in the normal assay. In this study, ADP-D-glucose, GDP-D-glucose, TDP-D-glu-from Y. pseudotuberculosis

Residue
Residues  cose, UDP-D-glucose, and the P-epimer of CDP-D-glucose showed no measureable activity, suggesting that CDP-D-glucose oxidoreductase is completely specific for CDP-D-glucose.
Cofactor Requirement-The purified enzyme alone was unable to catalyze the conversion of CDP-D-glucose to CDP-4keto-6-deoxy-~-glucose. The enzyme became competent only after the addition of NAD+. NADP+ could also stimulate activity, albeit at 20% of the rate obtained with the NAD+ under identical conditions. Incubation of the purified protein with NAD+ followed by gel filtration to remove the excess NAD' failed to reconstitute the enzyme activity in full. Thus, the purified oxidoreductase showed an absolute requirement for NAD+ (Fig. 2).
Kinetic Parameters-As detailed under "Materials and Methods," the enzyme was assayed with 1.15 pg of enzyme, 5 PM NAD', and an appropriate amount of CDP-D-glucose in 100 mM Tris-HC1 buffer (pH 7.5) at 37 "C. The dependence of the reaction rate on substrate concentration was found to follow the classic Michaelis-Menton saturation kinetics with  a K,,, of 222 pM, V,, The deuterium at C-4 was introduced by repetitive incubation of 6 in a pyridine/[2Hz]H20 (5:l) solution followed by evaporation (40). Although similar deuterium exchange proceeded rapidly for D-psicose (41), it occurred much more sluggishly for 6, probably due to the formation of the hydrate form (42). The deuterium content estimated by NMR by reference to the unlabeled sample of the reduced product 7 (43) was greater than 95%. Inversion at C-3 of 7 to afford the D-gluco configuration was effected by displacing the C-3 tosyl group with sodium benzoate in hot N,N-dimethylformamide (44). As the benzoate substituent of 8 was found to be labile in the subsequent reactions, it was replaced with the more stable benzyl protecting group. Stereospecific tritium incorporation at C-6 was the most challenging step since the C-5 hydroxymethyl group is not rigidly held within the sugar's ring structure. The method eventually chosen was the one established by Kakinuma (45,46), which is in many ways analogous to the classical synthesis of chiral acetates developed by Cornforth et al. (47). The common precursor, a 5,6-yne derivative 13, was obtained from the dibromo olefin 12 upon treatment with n-butyllithium in THF at -78 "C followed by quenching with [3H]Hz0 (30 mCi, 0.3 ml) (48). Transformation of 13 with a specific radioactivity of 0.46 Ci/mol to the (E)-olefin 14 was achieved by reduction using chromous sulfate in aqueous N,N-dimethylformamide (49). 'H NMR analysis of a control using the 6-'H1-labeled 13 as the reactant revealed that this reaction proceeded with 91% stereospecificity. Dihydroxylation with a catalytic amount of OsOl in the presence of N-methylmorpholine N-oxide (NMO) afforded the desired cis diol 15 in 72% yield (50). A small albeit noticeable amount of the (5S,GR)-isomer was also formed, but was readily removed by flash chromatography. The protected glucose was diluted 5-fold with 9 and was converted to free (6S)-~-[4-2H,6-3H]glucose (3a) by hydrogenolysis (10% Pd/C) and subsequent acid hydrolysis. The 6R-labeled glucose was prepared analogously from 13 via the (2)-olefin 16 which was produced by hydrogenation over Lindlar catalyst in the presence from Y. pseudotuberculosis

Chirality analysis of the methyl group of the product isolated from the reaction of CDP-D-glucose oxidoreductase with (65')-and (6R)-CDP-~-C4-~H,6-~H/glucose
Compound analyzed   (1:9), and then incubated with homogeneous CDP-D-glucose oxidoreductase in the presence of NAD' . The CDP-4-keto-6-deoxy-D-glucose products isolated by paper chromatography (EtOH/ BuOH/H20 = 5:5:2) were subjected to Kuhn-Roth oxidation (51,52). The nascent acetic acid samples were formed in radiochemical yields of 52-54%, and their chiralities were determined by the method of chiral methyl analysis (47,(53)(54)(55). An F value (56) of 71 corresponding to a 72% ee R configuration and an F value of 30 corresponding to a 69% ee S configuration were obtained for the two acetates derived from the 6s-and 6R-labeled glucose, respectively. The same analysis was also preformed on chiral acetate standards which were prepared from tritium labeled glycine. The numbers obtained for all acetates analyzed are summarized in Table  111. DISCUSSION The conversion of nucleotidyl diphosphohexose to its 4keto-6-deoxyhexose derivative has been shown to be the first step unique to the biosynthesis of several naturally occurring deoxy sugars. In an attempt to investigate the formation of ascarylose, a bacterial antigentic determinant, we have purified a CDP-D-glucose oxidoreductase catalyzing the first biosynthetic step of this 3,6-dideoxyhexose from Y. pseudotuberculosis. Although evidence for the occurrence of isozymes had been reported for TDP-D-glucose oxidoreductase found in Salmonella typhimurium, gel electrophoresis of our enzyme revealed only a single enzymatically active band with identical mobility at all stages of purification. This purified enzyme alone was unable to catalyze the conversion of CDP-D-glucose to CDP-4-keto-6-deoxy-~-glucose, and incubation of the protein with NAD' followed by gel filtration failed to reconstitute the enzyme activity in full. Such an absolute requirement of NAD' for its activity, as illustrated in Fig. 2, makes this enzyme a rare example among its class of oxidoreductases (12).

(6S)-CDP-~-[4-*H,6-~H]glucose (6R)-CDP-~-[4-~H,6-~H
Since studies of the catalysis of two other sugar oxidored- uctases, TDP-D-glucose oxidoreductase (29) and GDP-Dmannose oxidoreductase (31), have shown a remarkable stereochemical convergency in which the displacement of C-6 hydroxyl group by C-4 hydrogen proceeds intramolecularly with inversion of configuration, an analogous analysis directed at elucidating the stereochemical course of the reaction mediated by CDP-D-glucose oxidoreductase may provide unique mechanistic insights that are not available from other experimental approaches. In order to carry out the proposed stereochemical study, a set of stereospecifically labeled sugar 'nucleotide substrates and acetate standards were synthesized, mainly by chemical reactions. Although these compounds had been prepared heretofore by methods based on complex enzymatic manipulations (29), the large quantities of material and the well defined chiral purity of the sample provided by a chemical synthesis made this new approach uniquely appealing. As shown in Scheme 11, the precursors of the substrate, 3a and 3b, were prepared so that every tritiated molecule also carried a deuterium at C-4. If the displacement of the C-6 hydroxyl group by the migrating C-4 deuterium is indeed stereospecific, the resulting methyl group at C-6 is expected to contain hydrogen, deuterium, and tritium in a chiral arrangement of either S or R configuration. Kuhn-Roth oxidation of the incubation product 2 will give an acetic acid derived from the Cs and C5 of the sugar moiety, and the chirality of the methyl group can then be determined by the method of Cornforth et al. (53) and Luthy et al. (54). The key step of this well established analysis is the malate synthasemediated coupling of the acetate in its coenzyme A form with glyoxylate to form malate. Due to an isotope effect of 3.7-3.8 exhibited by this condensation, the malate derived from R and S labeled acetate will have a level of tritium labeling at C-3 dependent on the chirality of the original acetate. The unsymmetrical tritium distribution between the two diastereotopic hydrogens at C-3 of malate can be determined by incubating with fumarase, which catalyzes the stereospecific exchange of the pro-3R hydrogen with solvent protons. Using this method, chirally pure acetates of R and S configuration are expected to give F values (percentage of tritium retention in the fumarase reaction) of 79 and 21, respectively (56). As shown in Table 111, the malate derived from 6R-labeled glucose (3b) retained only 30% of its tritium after equilibration with fumarase, while the malate derived from 6s-glucose (3a) retained 71% of its tritium. These F values clearly indicate that incubation of the 6 s isomer (3a) with the purified oxidoreductase gave a product containing a stereospecifically labeled methyl group of R configuration, whereas the 6R isomer (3b) yielded material bearing a chiral methyl group of S configuration. In light of the large dilution of the labeled precusor by the unlabeled species prior to the incubation with oxidoreductase, the aforementioned results confirm that this enzymatic conversion involves an intramolecular hydrogen migration from C-4 to C-6, since the methyl from Y. pseudotuberculosis SCHEME 111.
group of product 2 is chiral only if the migration of the deuterium from C-4 to C-6 is strictly intramolecular and stereospecific. The data shown in Table I11 also establish that the displacement of the C-6 hydroxyl group by the hydrogen from C-4 occurs with inversion, which is identical to that found earlier for both TDP-D-glucose oxidoreductase and GDP-D-mannose oxidoreductase.
In view of these results, CDP-D-glucose oxidoreductase exhibits a mechanism analogous to that proposed for other enzymes of this class, as depicted in Scheme 111. The three chemically discrete steps include the oxidation of the substrate to a 4-ketosugar intermediate, dehydration to give a 4-ket~-A~~~-glucoseen species, and reduction to yield the 4-keto-6-deoxyglucose product. Since the nicotinamide cofactor serves as a hydride carrier in this catalysis, the intramolecular hydrogen transfer from C-4 to C-6 is likely to be suprafacial. On the basis of the suprafacial hydrogen migration and inversion of configuration at C-6, the removal of water from C-5 and C-6 can be perceived to proceed through a syn elimination. Since the mechanism of an elimination reaction is determined mainly by the lifetime of the initially formed carbanion (57), and in many enzymatic reactions where the elimination is syn the initially formed anion is commonly a to a keto or a thioester group that is capable of stabilizing the adjacent carbanion (58), the stereochemical mode of the dehydation deduced from our results suggests that the elimination catalyzed by CDP-D-glucose oxidoreductase may proceed through a stepwise mechanism. In order to completely define the stereochemical course of this enzyme-catalyzed reaction, incubation of CDP-[4-3H]6-deoxy-~-glucose with the purified enzyme was also attempted, following a procedure developed by Wang and Gabriel (30). Unfortunately, the 6-deoxyglucose derivative was found to be a competitive inhibitor5 and no [4-3H]NADH could be detected.
Although this reaction appears to follow the same pattern seen in other enzymes of this class, the observed F values, which may be regarded as indices of the stereospecificity of the catalysis, are lower than those expected for a fully stereospecific transformation. More precisely, the F value of 71.4 for the acetate from 3a corresponded to 72% enantiomeric excess (ee) of R configuration for the methyl group of this sample; while the F value 30.2 translated to a 69% ee of S configuration of the acetate derived from 3b. The enantiomeric purity of these materials was clearly lower than those generated by TDP-glucose oxidoreductase, which exhibited an 82% ee for the S and 88% ee for the R configurations of the derived acetates, respectively. This discrepancy may be ascribed to the different affinity of the two enzymes for the H. Liu and L. Liu, unpublished results. nicotinamide cofactor during turnover. Namely, the lack of rigid fixation may allow the cofactor to assume different orientations in the active site and thus lead to hydride transfer from either sides of the coenzyme. However, the fact that incubation with [3H]NAD+ resulted in no tritium incorporation in the product stands against this possibility. The relatively low ee found in this study may simply reflect a relaxed binding of the glycosyl portion of the 4-ket0-A~*~-glucoseen intermediate in the enzyme's active site, a situation akin to that observed in UDP-galactose-4-epimerase (59) and other related enzymes (12). It is also possible that the dehydration from C-5 and C-6 is not fully stereospecific. Obviously, more experiments are needed to gain further insights into this mechanistic ambiguity. It is worth mentioning that the chiral methyl group at C-6 generated by GDP-D-mannose oxidoreductase showed an even lower degree of chiral purity (31). Thus, this enzyme may have an even weaker binding for its coenzyme. However, since the later analysis was performed with a cell-free extract, an alternate explanation is available. Namely, occasional cofactor dissociation may have led to lower chiral purity of the product due to the replacement of deuterated cofactor midway through the catalytic cycle with unlabeled NADH from the cellular pool.
In conclusion, an NAD+-linked oxidoreductase catalyzing the conversion of CDP-D-glucose to the corresponding 4-keto-6-deoxy-~-glucose derivative was isolated from Y. pseudotubercubsis. It belongs to a small group of enzymes that are NAD+-dependent even though the overall catalysis is redox in nature (12, 60). Unlike most other enzymes in this class that have a tightly bound NAD+ in the active site, the purified enzyme showed an absolute requirement for exogenous NAD+ following purification. Despite its low cofactor affinity, the stereochemical analysis presented herein using chemically synthesized labeled substrate of high enantiomeric purity clearly showed that this purified enzyme undergoes a mechanism consistent with that followed by other members of its class. Aside from the moderate ee found for the acetate samples, the mechanistic and stereochemical convergency exhibited by CDP-D-glucose oxidoreductase with all of the other sugar oxidoreductases characterized so far suggested that this class of enzymes, regardless of their source, may evolve from a common progenitor whose catalytic course has persevered throughout the enzyme's subsequent diversification.