Purification and Characterization of 2,&Diketo-~-gluconate Reductase from Corynebacterium Sp.*

2,5-Diketo-~-gluconate reductase, a novel enzyme that catalyzes the stereospecific NADPH-dependent reduction of 2,5-diketo-~-gluconate to 2-keto-L-gulon-ate, has been purified to homogeneity by sequential anion exchange, Cibacron blue F3GA affinity, and gel permeation chromatography from Corynebacterium sp. ATCC 3 1090. Molecular weight of the native form, determined by gel permeation chromatography, is 35,000 2 2,000. The subunit molecular weight, deter- mined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis is 34,000; thus, the enzyme is active as a monomer. A PI value of 4.4 is measured for the enzyme. Amino- and carboxyl-terminal sequences are consistent with that predicted by the DNA sequence of the reductase gene. At 25 OC, pH 6.4, the turnover number is 500 min", and the apparent K,,, values for 2,5-diketo-~-gluconate and NADPH are 26 mM and 10 PM, respectively. The enzyme is specific for NADPH, but the sugar binding site will also accept keto-^-fructose and dihydroxyacetone as substrates. The enzyme is active over a broad pH range (pH 5-8) for the reduction of 2,5-diketo-~-gluconate; a sharp optimum at pH 9.2 is observed for the oxidation of 2-keto-~-

2,5-Diketo-~-gluconate reductase, a novel enzyme that catalyzes the stereospecific NADPH-dependent reduction of 2,5-diketo-~-gluconate to 2-keto-L-gulonate, has been purified to homogeneity by sequential anion exchange, Cibacron blue F3GA affinity, and gel permeation chromatography from Corynebacterium sp. ATCC 3 1090. Molecular weight of the native form, determined by gel permeation chromatography, is 35,000 2 2,000. The subunit molecular weight, determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis is 34,000; thus, the enzyme is active as a monomer. A PI value of 4.4 is measured for the enzyme. Amino-and carboxyl-terminal sequences are consistent with that predicted by the DNA sequence of the reductase gene. At 25 OC, pH 6.4, the turnover number is 500 min", and the apparent K,,, values for 2,5-diketo-~-gluconate and NADPH are 26 mM and 10 PM, respectively. The enzyme is specific for NADPH, but the sugar binding site will also accept keto-^fructose and dihydroxyacetone as substrates. The enzyme is active over a broad pH range (pH 5-8) for the reduction of 2,5-diketo-~-gluconate; a sharp optimum at pH 9.2 is observed for the oxidation of 2-keto-~gulonate. A K. , value of 5.6 X 10"' M indicates that reduction of substrate by NADPH is highly preferred. An activation energy of 12.3 kcal mol" is measured. Enzyme turnover is slow relative to dehydration of the gem-diol at C-5 of the substrate.
2-Keto-~-gulonic acid (2-KLG)' is a key intermediate in the Reichstein-Griissner multistep synthesis of L-ascorbic acid, a commercial process which utilizes D-glucose as the initial starting material (1). Since 2-KLG can be readily converted into L-ascorbate, a great deal of research has been directed toward the improved synthesis of 2-KLG by both chemical and microbial methods using a wide variety of starting materials (2,3). Recently, Sonoyama et al. (4) reported on a tandem fermentation in which first a mutant strain of Erwinia herbicolu oxidatively converted D-glucose (G in Scheme I) to 2,5-diketo-~-gluconate (2,5-DKG) via D-gluco-* 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.
$ Present address: Research Dept., Genencor, Inc., 180 Kimball Way, South San Francisco, CA 94080. §To whom correspondence and reprint requests should be addressed.
In this paper, we report on the identification, purification, and characterization of Corynebacterium sp. (ATCC 31090) 2,5-DKG reductase, a cytosolic enzyme that catalyzes the NADPH-dependent reduction of 2,5-DKG to 2-KLG.' To our knowledge, this is the first example of an enzyme that catalyzes this reaction. The properties of this enzyme have allowed us to develop a one-step microbial bioconversion of D-glucose to 2-KLG (5). This was accomplished by cloning and expressing the 2,5-DKG reductase gene into E. herbicola, an efficient 2,5-DKG producer, thereby creating a new metabolic pathway (Scheme I).

2.5-DKG 2-KLG
in a final volume of 1 ml. The reaction was monitored for the initial linear decrease in absorbance at 340 nm ( e = 6.22 m"' cm"). Activity was proportional to the amount of 2,5-DKG reductase added. One unit of activity corresponds to the production of 1 pmol of NADP+ per min under conditions stated above. Assays in the oxidative direction were carried out in 0.2 M glycine/NaOH buffer (pH 9.2).

Culture Methods
Corynebacterium sp. cells (ATCC 31090) were grown in a 10-liter fermentor from a 5% inoculum (grown in shake flasks containing Luria broth for 24 h at 28 "C and 150 rpm) at 28 "C and 1000 rpm in the basal medium of Sonoyama et al. (4), modified to contain 1% glucose and 0.5% yeast extract adjusted to pH 7.0 with NH,OH. Cells were harvested at -2oASw, as measured after dilution on a Bausch and Lomb Spectronic 20, and pelleted using a Sorvall RC5B refrigerated centrifuge at 6000 X g for 30 min. Cell paste was stored at -20 "C.

Purification of 2,5-DKG Reductase
Step I: Cell Lysis and Extraction-All steps were performed at 4 'C unless noted otherwise. Frozen cell paste (288 g wet weight), stored at -20 "C, was thawed and resuspended in 1 volume of 20 mM Tris-HCl (pH 8.0) containing 0.5 M NaCI. The washed cells were recovered by centrifugation at 6000 X g for 30 min. The pellet was resuspended in 1 liter of 20 mM Tris-HC1 (pH 8.0) containing 2 mg/ml lysozyme and 0.1% Tween 80 and allowed to stand for -1 h at 25 "C. DNase (-1 Fg/ml) was added to reduce viscosity, and lysis was allowed to continue for an additional 2 h. Cell debris was separated by centrifugation at 20,000 X g for 16 h at 4 "C. The pellet was discarded, and the supernatant was used as the starting material for enzyme purification.
Step 2: DEAE-cellulose Chromatography-The crude cell extract was dialyzed against 20 mM Tris-HC1 (ph 8.0) and adsorbed batchwise onto 590 ml of DEAE-cellulose (Whatman DE52) that had been equilibrated in the same buffer. After stirring for 30 min, the resin was filtered, washed with buffer, loaded into a 5-X 60-cm column, and washed with additional buffer until a stable base line (Azso) was established. The enzyme was eluted with a 2-liter linear gradient of 0.75 M NaCl in 20 mM Tris (pH 8.0) at a flow rate of 1.8 ml/min. Two distinct peaks of NADPH-dependent 2,5-DKG reductase activity were observed (Fig. 1). The second peak of activity, which elutes at 0.4 M NaC1, was shown to produce 2-KLG as the reduction product (see below). This peak was pooled and dialyzed overnight against 20 mM Tris-HC1 (pH 8.0).
Step 3: Affinity Chromatography-The enzyme solution from the DEAE-cellulose column was adjusted to pH 7.2 with HC1 and applied to a 2.5-X 2.0-cm column of Cibacron blue F3GA agarose equilibrated with 20 mM Tris-HC1 (pH 7.2). The column was washed with 200 ml of buffer, and 2,5-DKG reductase was eluted in the same buffer containing 5 mM NADP+. Fractions that contained 2,5-DKG reductase activity were pooled and concentrated 16-fold by ultrafiltration using an Amicon YM-5 membrane. This pool was dialyzed overnight against 20 mM Tris-HC1 (pH 8.0).
Step 4: High Performance Gel Filtration Chromatography-The dialyzed enzyme pool from the affinity column was applied in 250-500-~1 aliquots to an Altex TSK preparative gel filtration column (0.5 X 60 cm) equilibrated with 0.2 M ammonium bicarbonate (pH 7.5) at 0.5 ml/min. The column elution profile showed a minor high molecular weight peak followed by a major symmetrical peak corresponding to 35,000; 2,5-DKG reductase activity was observed in the second peak (see below).

Identification of Reaction Products
Purified 2,5-DKG reductase (0.14 mg) was incubated in 100 mM bis-Tris-HC1 (pH 6.4) buffer containing 0.84 mM 2,5-DKG and 1.4 mM NADPH. After a 12-h incubation, 0.79 mM NADPH had been oxidized as measured by the decrease in absorbance in AW (z = 1.25 mM" cm"). The sample was analyzed by HPLC on a Bio-Rad HPX-87H organic acids column (7.8 X 300 mm) using 0.01 N HzSO, at 0.5 ml/min at 25 "C (10). Peaks were monitored by a Waters Model 681 variable wavelength detector set at 210 nm, and the area was calculated by a Hewlett-Packard HP3390 integrator. Standard curves of 2-KLG concentration were found to be linear over the range investigated. Product identification of the reaction mix was confirmed by HPLC analysis on a 4.6-X 250-mm column of Aminex A27 (Bio-Rad) using 0.2 M ammonium formate (pH 4.0) as the eluant at 0.8 ml/min at 45 "C (10). Detection of peaks was performed by a Knauer Model 98 refractive index detector. The HPLC analyses were verified by analysis of the pertrimethylsilylated derivatives using gas-liquid chromatography and mass spectroscopy (10).
The lyophilized reaction mixture was incubated with trimethylsilylimidazole/pyridine (50:50) for 30 min at 60 "C. Analysis of the derivatized mixture was performed on a Perkin-Elmer Sigma Model 115 using a 25 meter 5% cross-linked phenylmethylsilicone-fused silicon-bonded capillary column (DB-5, J and W) and a flame ionization detector. A Hewlett-Packard Model 5985 gas chromatographymass spectrometer clearly identified the correct peaks of the pertrimethylsilyl derivatives of the known standards and reaction products. The gas chromatography analysis was capable of detecting 2-KDG. trimethylsilyl, at less than 1% of the amount of 2-KLG. trimethylsilyl,. The reduction product of 5-keto-~-fructose was monitored by ion exclusion chromatography (HPX-87H), since L-sorbose (RT = 13.5 min) and D-frUCtOSe (RT = 14.0 min) were clearly separated using 0.01 H,SOI as the eluant at 0.4 ml/min.

Determination of Native Molecular Weight
The purified enzyme (20 Irg) was applied to an LKB 2135 Ultrapak TSK G 3000 SW gel filtration column (7.5 X 600 mm) fitted with an LKB 2135 Ultrapak TSK-GWSP guard column (7.5 X 75 mm) equilibrated in 200 mM potassium phosphate (pH 6.5). The column was eluted at 0.5 ml/min and calibrated for native molecular weight

Purification and Characterization of 2,5-DKG Reductase
Protein Determination Protein concentrations were determined using the Bradford dye binding method using bovine serum albumin as a standard (11).

Amino Acid Sequence Analyses
Amino-terminal sequence was determined by the sequential Edman degradation method using a modified Beckman Model Sequencer 890 B as previously described (12).

Electrophoresis
Purity of the enzyme, as well as molecular weight, was determined by polyacrylamide gel electrophoresis according to the method of Laemmli (14) in 10% vertical slab gels (1.0 mm thick) containing sodium dodecyl sulfate using a Hoefer Model PS 500 power supply.
Proteins were detected either by staining with Coomassie Brilliant Blue R-250 or by silver staining as described by Oakley et al. (15).
Isoelectric focusing gel electrophoresis was carried out on an LKB Model 2117 Multiphor bed using an LKB 2197 power supply. The gel (LKB 1804-102) was run from pH 4 to pH 6.5 using 0.1 M glutamate in 0.5 M H,PO, for the anode solution and 0.1 M B-alanine for the cathode solution. Approximately 15 pg of 2,5-DKG reductase was applied using a paper filter wick and run at 10 "C for 2.5 h at 83 watts, 666 V, and 83 mA. The gel was stained with Coomassie R-250. Standards corresponding to pH 4.1,4.9, and 6.4 were run.

DKG][NADPH][H+]/[2-KLG][NADP] at 25 "C. Reactions were in-
The equilibrium constant (K,) is defined as being equal to [2,5itiated by adding known amounts of 2-KLG (20-200 mM) and NADP (0.2-0.8 mM) to 0.1 mg of purified 2,5-DKG reductase in either 0.1 M glycine/NaOH or 40 mM sodium carbonate from pH 9.2-9.6. The increase in Aa0, which corresponds to [NADPH] or [2,, was measured at equilibrium. The hydrogen ion concentration was calculated from the measured pH. The K, that was subsequently calculated is the average of six separate determinations.

RESULTS
Purification of 2,5-DKG Reductase-A 2,5-DKG reductase from Corynebacterium sp. ATCC 31090 was purified to apparent homogeneity from cytosolic extracts of lysed cells by consecutive chromatography on DEAE-cellulose, Cibacron blue F3GA agarose, and HPLC TSK gel filtration columns. A typical purification scheme is summarized in Table I, and technical details of the purification are outlined under "Methods." Initial screening for the 2,5-DKG reductase activity indicated that the enzyme was NADPH-linked since 2-KLG producing 2,5-DKG-dependent NADH oxidation was not observed in crude lysates. The elution profile of the DEAE-~ ~~ a As described under "Methods." ' The total number of 2,5-DKG reductase (2-KLG-producing) units in the crude lysate was calculated by multiplication of the total number of units times the fraction of activity due to 2,5-DKG reductase (0.4) as determined by integration of the peak areas in Fig. 1. e The specific activity of pure protein is lower than that in Table   I1 since assays were carried out at subsaturating levels of 2,5-DKG. cellulose column (Fig. 1) indicated two distinct peaks of 2,5-DKG-dependent reductase activity. The second peak, which eluted at 0.4 M NaC1, was found to catalyze the stereospecific reduction of 2,5-DKG to 2-KLG. This was further purified using Matrix Gel Blue A dye ligand chromatography with elution by NADP+, followed by an HPLC TSK gel filtration step. The first peak of activity on the DEAE-cellulose column, which eluted at 0.25 M NaC1, was found to reduce 2,5-DKG at the 2-position to give 5-KDG as the reaction product (see below). Beyond tentative identification of a band at M, = 49,000 by SDS-polyacrylamide gel electrophoresis after purification steps similar to those outlined above, this protein was not further characterized. Purity, Moleculur Weight, Subunit Structure-The purified 2,5-DKG reductase eluted as a single, symmetrical peak of active protein during gel filtration on an HPLC TSK column.
The native molecular weight was calculated to be 35,000 * 2,000 when compared to standards (Fig. 2). SDS-polyacrylamide gel electrophoresis of the purified protein revealed a single band corresponding to a molecular weight of 34,000 (Fig. 3). This is in good agreement with the size (Mr = 29,992) calculated from the DNA sequence of the gene (5). These results suggest that the enzyme is active as a monomer. The UV absorption spectrum of pure 2,5-DKG reductase displayed a maximum at 280 nm with a molar absorption coefficient of NH2-terminaE Analysis and COOH-terminal Analysis-Amino-terminal sequence analysis (40 residues) of 2,5-DKG reductase was derived by Edman degradation. An identical sequence is predicted from the DNA sequence of the cloned 2,5-DKG reductase gene ( 5 ) . Carboxyl-terminal analysis revealed an aspartic acid which is also consistent with the amino acid sequence derived from the DNA sequence ( 5 ) .
Substrate Specificity, Kinetic Constants, and pH-rate Profik-2,B-DKG reductase is relatively specific in its cofactor requirements. The rate of reduction of 2,5-DKG (8.5 mM) is favored by a factor of 170 for saturating levels of NADPH over NADH as a cofactor. An apparent K , of Table 11. From this table, a turnover number of -500 min" is calculated for 2,5-DKG. In addition to those compounds listed in Table 11, the following compounds showed no reduction activity with NADPH and the enzyme: D-fructose, L-sorbose, 5-keto-~-gluconate, 2-keto-L-gulonate, 2keto-D-gluconate, pyruvate, or hydroxypyruvate.
The effect of 2,5-DKG reductase concentration on the rate of reduction of 2,5-DKG by saturating NADPH revealed the following: 1) initial burst of product formation is observed and 2) both the initial zero order ([2,5-DKG] > K,) and subsequent first order ([2,5-DKG] < K,,,) decrease in absorbance are directly proportional to enzyme concentration. These results imply that dehydration of the gem-diol a t C-5, which is the major tautomeric structure in aqueous solution (7,16), is rapid with respect to enzyme turnover.
The enzyme is active over a broad pH range with maximal activity at pH 6.4 in the reduction of 2,5-DKG (Fig. 4). The oxidation of 2-KLG (K,,, = 204 mM; V,,, = 0.4 pmol/min/ mg) by NADP ( K , = 125 p~) has a pH optimum at pH 9.2 and narrow range of activity between pH values of 8.0-10.3. The kinetic parameters for glycerol oxidation at pH 9.2 at 0.5 mM NADP were determined to yield an apparent K , = 2.2 M and V,,, = 1.4 pmol/min/mg.
Equilibrium and Temperature Dependence-The equilibrium constant (K,) of the reduction of 2,5-DKG by NADPH was measured spectrophotometrically by measuring the increase in A340 of incubations of 2-KLG and NADP a t several different pH values. A K, value of 5.6 X M indicates that reduction of 2,5-DKG to 2-KLG is thermodynamically highly favored. The measured K, defines the standard reduction potential (Eo') for 2,5-DKG/Z-KLG to be -0.165 V. An Arrhenius plot of enzyme activity versus temperature was linear between 20 "C and 36 "C, from which an activation energy of 12.3 kcal mol" was calculated.
Product Characterization, Stereospecificity, and Stoichiometry-The product of the enzymatic reduction of 2,5-DKG by NADPH was found to be 2-KLG based on the fact that retention times observed using HPLC ion exclusion (HPX-87H), HPLC ion exchange (Aminex A27), or gas chromatography-mass spectrometry of the petrimethylsilyl derivative are identical with 2-KLG. A stoichiometric conversion to 2-KLG and NADP+ was observed based on quantitative analysis of both of the products. Reduction of 2,5-DKG occurs only at the 5-position and is stereospecific for 2-KLG since no 2-KDG, the other potential product from reduction at the 5position, could be detected by GC (10). In addition, 2-KDG does not serve as a substrate in the oxidative reaction.
The product of the NADPH-linked enzymatic reduction of 5-keto-D-fructose, which differs from 2,5-DKG only in that carbon 1 is an alcohol instead of an acid, is specifically Lsorbose; no D-fructose could be observed. Therefore, the reduction proceeds with the same stereochemistry as that observed for 2,5-DKG reduction to 2-KLG.
The reduction product of 2,5-DKG by NADPH catalyzed by the first peak of activity of the DEAE-cellulose column, which elutes at 0.25 M NaC1, was determined to be 5-KDG based on chromatographic analysis of the product and comparison with authentic 5-KDG. Reduction therefore occurs a t the 2-position with this enzyme.
Enzyme Inhibition and Stability-Product inhibition by NADP+ was found to be competitive, with an apparent KI =  no inhibition by 2-KLG was observed ( 4 0 0 mM). The enzyme was strongly inhibited by 0.5 mM solutions of Zn2+, Fe3+, Cu2+, or NiZ+. There was no observed effect on activity by 0.5 mM solutions of M P , Mn2+, Ca2+, or Co2+, 14 mM pmercaptoethanol, 1 mM-dithiothreitol, or 1 mM EDTA. The enzyme (2 mg/ml) was stable for over 6 months in 20 mM Tris-HC1 (pH 7.5) a t -70 "C. Enzyme solutions were stable at 4 "C from pH 6.5-7.5 for at least 2 months. DISCUSSION In this report, we present data on the purification and characterization of 2,5-DKG reductase, a novel enzyme from Corynebacterium sp. To our knowledge, it is the first enzyme that has been purified to apparent homogeneity that catalyzes the stereospecific reduction of 2,5-DKG to 2-KLG. The purified enzyme has an apparent native molecular weight of 35,000 based on gel permeation chromatography and runs with a molecular mass of 34,000 on an SDS-polyacrylamide gel, suggesting that the monomeric form of the enzyme is active. A strict requirement for NADPH as the cofactor is observed; however, both 5-keto-~-fructose and dihydroxyacetone can serve as substrates. The true physiological substrate for 2,5-DKG reductase remains unknown since Corynebacterium sp. is not known to produce 2,5-DKG as a metabolite (17).
The major form of 2,5-DKG in aqueous solution as determined by 13C NMR is the gem-diol hydrate at C-5 of the pyranose tautomer; little if any keto component at C-5 is present (7,16). Likewise, 5-keto-~-fructose has been shown to exist predominantly (>95%) as a gem-diol hydrate (18). Since reduction by NADPH occurs at the C-5 position, the hydration state and rate of dehydration to the keto form may influence the observed kinetic parameters (Scheme 11). In addition, the chemical nature of the substrate form, i.e. the keto form of the pyranose, furanose, or open-chain tautomer is unknown. Therefore, the apparent K, obtained from initial rate kinetics may grossly overestimate the true K,.
The tautomeric equilibrium of 5-keto-D-gluconate (5-KDG) has been studied by I3C NMR and shows the presence of two furanose forms (89%) with the remaining 11% existing as the open-chain keto form (16). Using the value of the apparent K,,, of 0.9 mM for 5-KDG obtained with 5-keto-~-ghconate reductase from Gluconobacter suboxydans (19), we calculate a corrected K,,, of 0.1 mM for the keto form of 5-KDG. If a similar binding constant is assumed for the keto form of 2,5-DKG with 2,5-DKG reductase, then the keto tautomer would represent less than 1% of the total species present in solution, consistent with the observed 13C NMR results (16).
A strict dependence of rate of product formation versus 2,5-

2.5-DKG 2-KLG
DKG reductase concentration is observed both where 2,5-DKG is above or below its apparent K,. In addition, no initial burst in the decrease of the A340 is found. These results imply either that dehydration of 2,5-DKG a t C-5 is fast relative to enzyme turnover, i.e. a pre-equilibrium of [keto] = [gem-diol] exists or that 2,5-DKG reductase is able to catalyze the dehydration of 2,5-DKG. This case is in contrast with the reduction of aldehydes by alcohol dehydrogenase, where hydration to form the gem-diol results in an initial enzymedependent burst followed by a slow enzyme-independent first order dehydration rate (20).