INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Interest in 2,5-diketo-D-gluconic acid reductases (DKGRs) is related to a new biotechnological process for the production of vitamin C (ascorbic acid) from glucose. Conversion of glucose to ascorbic acid is a complicated process that involves selective epimerization, oxidation, and lactonization reactions. The natural biosynthetic pathways are long and incorporate many energy-consuming reactions (8, 21, 38). The present commercial process for ascorbic acid production (the Reichstein process) couples a single biological stepthe microbial oxidation of sorbitol to sorbosewith a subsequent, multistep, chemical conversion of blocked derivatives of sorbose to ascorbic acid (7, 26). An alternative commercial process that has been proposed (2, 10, 34) consists of biological conversion of glucose to 2-keto-L-gulonic acid, which is then lactonized chemically to ascorbic acid (Fig. 1). The biological metabolism involved is simpler than that of natural biosynthetic routes and requires less metabolic energy (less ATP and NADPH). In this process, glucose is first converted to 2,5-diketo-D-gluconic acid by endogenous oxidases of a suitable bacterial strain, with molecular oxygen as the ultimate electron acceptor. 2,5-Diketo-D-gluconic acid is then reduced enzymatically to 2-keto-L-gulonic acid by a heterologous DKGR expressed in the production strain. The NADPH required for the reaction is generated by the metabolism of the host strain. Finally, chemical lactonization of 2-keto-L-gulonic acid generates ascorbic acid.

View larger version (8K): [in this window] [in a new window]   FIG. 1.   Alternative process for conversion of glucose to ascorbic acid. An alternative process for conversion of glucose to ascorbic acid effected by whole cells, a recombinant enzyme, and a chemical catalyst. Membrane-bound oxidases convert glucose to 2,5-diketo-D-gluconic acid (2,5-DKG), with oxygen as the electron acceptor. A heterologous 2,5-diketo-D-gluconic acid reductase, in practice obtained from another organism, reduces 2,5-diketo-D-gluconic acid to 2-keto-L-gulonic acid (2-KLG). Known reductases are NADPH dependent. Finally, chemically catalyzed lactonization of 2-keto-L-gulonic acid generates ascorbic acid.

To date, only two DKGRs have been extensively characterized, both isolated from a species of Corynebacterium (20, 23, 35). These enzymes, designated DKGR A and B, reduce 2,5-diketo-D-gluconic acid rather inefficiently, exhibiting K_m values for substrate greater than 1 mM and catalytic efficiencies (k_cat/K_m) less than 20 mM¹ s¹. The enzymes are members of the aldo-keto reductase superfamily (14, 30). Like almost all other aldo-keto reductases, the known DKGRs are specific for NADPH (14, 30). Recently, additional aldo-keto reductases that can convert 2,5-diketo-D-gluconic acid to 2-keto-L-gulonic acid have been isolated from Escherichia coli through a search of the genome sequence for homologues (40, 41). However, these enzymes also catalyze the reaction relatively inefficiently. The Corynebacterium DKGRs also lack stability (23, 35). Consequently, DKGRs with superior enzymatic activity and enhanced stability were sought to facilitate the development of the alternative production processes for ascorbic acid.

Rather than search through culture collections for a known microbe with a superior DKGR, we explored the vast reservoir of uncultured microorganisms. Molecular analyses of microbial communities have revealed that less than 1% of the microbes present are known from culture collections. These undiscovered microbes certainly contain many homologues of known enzymes, some of which may have valuable properties. Present methods for accessing these enzymes rely on construction of libraries of DNA fragments of uncultured organisms and screening of clones for the desired enzymes (12, 33). Here we describe an alternative approach that relies solely on PCR to obtain full-length, functional genes from environmental DNA.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Extraction and purification of environmental DNA. Environmental samples were collected in the vicinity of Argonne National Laboratory, Argonne, Illinois. Pond water was collected in plastic carboys, and the suspended matter was concentrated either by flowthrough centrifugation (model A5-16; Sharples) or for small volumes by filtration through 0.22-µm-pore-size nitrocellulose filters. The DNA was extracted from the concentrates using the Puregene whole blood and tissue DNA extraction kit (Gentra Systems). Soil samples from a deciduous forest and a residential garden were taken from 3 to 6 cm below the surface and extracted essentially as described by Selenska and Klingmüller (31). Two grams (wet weight) of soil were suspended in 4 ml of extraction buffer (120 mM Na₂HPO₄, pH 8.0, containing 1% sodium dodecyl sulfate) in a 50-ml conical tube, shaken at 200 rpm for 1 h at 70°C, and then centrifuged at 3,000 × g for 5 min at room temperature. The DNA-containing supernatant was collected, and the soil pellet was extracted twice more with 2 ml of extraction buffer. The combined supernatants were centrifuged at 20,000 × g for 10 min at room temperature to remove residual particles.

Amplification of complete DKGR genes. Homologues of known DKGR genes were amplified from environmental DNA samples in three steps. In the first step, internal fragments were generated using degenerate primers based on known DGKR gene sequences (Table 1). Following sequencing of the internal fragments, their flanking regions were amplified in a series of nested PCRs in which one primer was specific for the internal fragment and the other was a random primer that targeted the unknown flanking DNA. For subsequent amplifications, the product of the previous PCR was used as the template and a different, internally nested primer was paired with the same random primer. Once distinct PCR products were generated in good yield (three or four rounds of PCR were required), the products were isolated and sequenced, allowing prediction of the sequence of the full-length gene (Fig. 2). The entire gene was then amplified in a single PCR with primers specific for the flanking regions of the assembled, putative gene using the original environmental DNA as the template.

View larger version (69K): [in this window] [in a new window]   FIG. 2.   Assembled putative DKGR genes and flanking regions.The putative DKGR genes, I-14 (A) and I-28 (B), and their flanking sequences were assembled from the three fragments generated by PCR of environmental DNA. The coding sequence is shown in uppercase letters, and below it the predicted amino acid sequence of the DKGR homologue is shown. Flanking sequence is shown in lowercase letters, and flanking ORFs are italicized. The arrows indicate primers used in the PCR amplifications (Table 1). In each case, the regions of overlap span from the degenerate primers DU1 and DL1 to the nearest opposing nested primer.

Cloning of internal gene fragments and DKGR genes. The internal gene fragments generated in the initial amplifications were purified by electrophoresis in 1.0% agarose gels in Tris-borate-EDTA buffer (28). The bands of interest were excised, purified with a QiaQuick gel purification kit (Qiagen), and ligated into pBluescript SK+ (Stratagene) that had been treated with EcoRV and Taq polymerase plus dTTP to introduce a single unpaired T residue at its 3' ends (3). PCR products were ligated into the vector with T4 DNA ligase (Promega) and were transformed into E. coli DH5 (MaxEfficiency; GIBCO/BRL). White transformants that arose on Luria-Bertani agar plates containing ampicillin, isopropyl--D-thiogalactopyranoside (IPTG) (U.S. Biochemicals), and 5-bromo-4-chloro-3-indolyl--D-galactopyranoside (X-Gal) (28) were analyzed for the presence of inserts of the expected size by PCR with primers specific for the T3 and T7 promoter regions of the vector.

Analysis of PCR products and genes. The cloned internal fragments, flanking-region PCR products, full-length PCR products, and cloned DKGR genes were sequenced using the ABI Prism Dye Terminator Cycle Sequencing Ready Reaction kit (Perkin-Elmer Applied Biosystems) in a Perkin-Elmer GeneAmp PCR system 9600 thermal cycler. All component concentrations and the incubation and cycling conditions followed the manufacturer's instructions. Samples were separated on a 6% acrylamide gel containing 8 M urea in an Applied Biosystems 373A DNA sequencer (Perkin-Elmer Applied Biosystems). Sequences were analyzed with the Seqman program (DNAStar). The flanking regions amplified from environmental DNA were not cloned. Rather, their PCR products were sequenced directly using the clone-specific primers.

Expression and purification of DKGRs. Cultures of E. coli DH5 containing pI-14jf and of E. coli JM109 containing pI-28jf were grown at 37 and 30°C, respectively, in 500 ml of Luria broth (28) in 2-liter, notched Erlenmeyer flasks shaken at 250 rpm. IPTG was added to 1 mM at an optical density at 600 nm of 0.5, and cells were harvested after 4-h (37°C experiment) or overnight growth (30°C) and were then washed once with Tris-EDTA buffer. Cells were resuspended in approximately 2 volumes of 10 mM Tris-HCl (pH 7.5) containing 1 mM EDTA, 0.5 mM dithiothreitol, and 0.001% phenylmethylsulfonyl fluoride and were then lysed by passing the suspension twice through a French press. Cell debris and membranes were removed by centrifugation at 950 × g, followed by ultracentrifugation at 435,000 × g in a Beckman TL-100 ultracentrifuge. Both reductases were purified by affinity chromatography (Matrix Red A; Amicon) followed by ion exchange on a MonoQ column (Pharmacia). All buffers contained 0.5 mM dithiothreitol.

Enzymatic analyses. Standard assays were performed in the direction of reduction of 2,5-diketo-D-gluconic acid at 30°C in 1.0 ml of 100 mM Tris-HCl buffer, pH 7.2, containing 0.1 mM NADPH and 1 mM 2,5-diketo-D-gluconic acid (provided by Genencor International). The decrease in absorbance due to the oxidation of NADPH was measured with a Shimadzu UV 160U spectrophotometer. One unit of enzyme is defined as the amount of enzyme that catalyzes the oxidation of 1 µmol of NADPH per min. Alternative substrates, obtained from Sigma, were evaluated at 1 mM. For determination of the pH optima, solutions of 100 mM bis-Tris and bis-Tris propane were prepared in increments of 0.5 pH unit from pH 5.5 to 9.0. The enzymes were assayed at each pH level to determine optimal activity.

Characterization of reaction product. To a 1-ml solution that contained 10 µmol of highly pure 2,5-diketo-D-gluconic acid in 65 mM bis-Tris buffer at pH 7.0, 5 µmol of NADPH and approximately 40 U of purified reductase were added. The progress of the reaction was monitored by determination of the concentration of NADPH that remained. When the reaction mixture reached an optical density at 340 nm of less than 2.0, an additional 5 µmol of NADPH and 40 U of purified reductase were added and the incubation continued. The product was then evaporated to dryness under reduced pressure and derivatized as follows. Approximately 0.3 mg of dried material was mixed with 0.40 ml of pyridine plus 0.1 ml of solution containing 60 mg of hydroxylamine hydrochloride in 3 ml of pyridine. The sample was heated for 30 min at 75°C in a sealed vial, and then 1 ml of N,O-bis-(trimethylsilyl)trifluoromethylacetamide (BSTFA; Pierce) containing 1% trichloromethylsilane was added slowly while the vial was agitated by hand. The resulting solution was heated for 30 min at 75°C to induce silylation. After cooling, 1-µl samples were injected directly into a Finnigan TSQ-700 gas chromatograph-mass spectrometer (GC/MS) equipped with an electron impact detector and a 0.25-µm-thickness film DB-5 capillary column (0.23-mm inside diameter, 30-m length). The column temperature was ramped from 50 to 290°C at 15°C/min and was then held at 290°C for 14 min. Other parameters were as follows: injector, 280°C; transfer line, 280°C; splitless injection; valve A open after 0.5 min; multiplier, 1,200 V; and filament, 300 mA. The mass/charge ratio (m/z) range of 35 to 650 was scanned in 1-s intervals. The trimethylsilyl oxime derivatives of 2-keto-L-gulonic acid (syn and anti) were eluted as a doublet at approximately 13 min.

Thermal inactivation of DKGRs. The thermal stability of each reductase was evaluated at low protein concentration (0.085 mg/ml) in 100 mM bis-Tris buffer, pH 7.0. The half-life at 45°C was determined by incubating 30-µl aliquots of purified enzyme in thin-walled PCR tubes at 4°C. The temperature was shifted rapidly to 45°C by means of a Robocycler Gradient 96 thermal cycler (Stratagene) and was held at 45°C for 0.5, 5, 10, 20, 30, or 60 min before the sample was returned to 4°C. Each tube was assayed later by the standard procedure. The rate constant for loss of activity was determined by fit to the equation for exponential decay using Prism (GraphPad). The midpoint temperature of inactivation of DKGR D was determined by incubating the enzyme for 10 min over a range of temperatures defined by the Robocycler. The Robocycler was programmed to move samples from 4°C to a gradient of defined temperatures ranging from 30 to 52°C in 2°C increments. After 10 min the samples were returned to 4°C. All samples were assayed in duplicate.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Amplification of internal gene fragments. In a control PCR using a plasmid bearing the Corynebacterium spp. DKGR A gene as the template, the degenerate primers DU1 and DL1 (Table 1) generated a well-defined band of the expected 380 bp. When various environmental DNA extracts were used as the template, PCR gave broad bands between 350 and 400 bp in size. These bands were excised from the gel and cloned, resulting in a total of six independently derived, putative gene fragment clones that were studied further (Table 2). Sequencing revealed that all six clones were different from one another. A basic local sequence alignment tool, BLASTX (1), search of the GenBank database indicated that all six possessed significant homology with known members of the aldo-keto reductase superfamily. Each was most highly homologous to a gene of unknown function: fragments I-14, I-28, and II-4 to the yvsB gene of Bacillus subtilis; fragment III-6 to the E. coli yafB gene; fragment III-19 to ytbE of B. subtilis; and fragment III-24 to protein CAA22355 of Streptomyces coelicolor. Two of the fragments, I-14 and I-28, were very similar to each other (76% nucleotide sequence identity, excluding the primer sequences). Of the six, these two were the most homologous to the Corynebacterium DKGR A gene (GenBank accession no. M12799 [2]), possessing 45 and 44% nucleotide sequence identity, respectively.

PCR amplification of upstream and downstream regions. The 5' and 3' flanking sequences for clones pI-14 and pI-28 were obtained by RSPCR (29). Nested, clone-specific primers (Table 1) were paired with several different RSPCR primers. The initial amplification, using environmental DNA as the template, generated a diffuse smear of products with a few faintly discernible bands. Subsequent rounds of PCR used the product of the previous reaction as the template, the same RSPCR primer, and a nested primer. With each round, increasingly discrete products were generated. After three or four rounds, discrete products were formed in good yield. For both the pI-14 and pI-28 clones, an approximately 500-bp fragment of 5' flanking sequence was generated using the BglI RSPCR primer and nested primers 14nL4 and 28nL4, respectively, and an approximately 800-bp fragment of 3' flanking sequence was generated using the XhoI RSPCR primer and nested primers 14nU3 and 28nU3, respectively.

Direct amplification and expression of complete genes. To establish that the assembled I-14 and I-28 genes are truly present in the environment and are not chimeras of multiple homologous genes, the full-length genes were amplified in a single PCR using the original environmental DNA as the template and primers specific for the flanking regions of the assembled fragments (Table 1; Fig. 2). In each case, a single band of the expected size, 1.34 and 0.98 kb, respectively, was obtained in good yield. Sequencing of these PCR products confirmed their identity as the assembled I-14 and I-28 genes. To allow expression of the genes, restriction sites were introduced adjacent to the start and termination codons by PCR. The primers (Table 1) also changed the GTG start codon of I-14 to ATG and added a second termination codon to each gene. These PCR products were cloned into pJF118EH (9) to give the expression vectors pI-14jf and pI-28jf, respectively.

View larger version (22K): [in this window] [in a new window]   FIG. 3.   Amino acid sequences of expressed DKGRs. The complete amino acid sequences of the final, expressed genes, obtained from direct amplification of environmental DNA, are given here. Only the residues of DKGR D (pI-28jf) that differ from those of DKGRC (pI-14jf) are shown. The numbering of DKGR D matches that of DKGR A, whose structure has been determined.

Purification of reductases. The two DKGRs were purified by affinity chromatography with Matrix Red agarose followed by ion-exchange chromatography on a MonoQ column (Table 3). DKGR C, which was more highly overexpressed, was purified to homogeneity in two steps. This reductase bound tightly to the Matrix Red agarose and was eluted at relatively high purity with 2.5 M NaCl, dialyzed to remove salt, and then purified to homogeneity on the MonoQ column. The enzyme was eluted as a sharp, symmetrical, well-isolated peak at approximately 0.4 M NaCl.

Determination of product of DKGRs. Reduction of 2,5-diketo-D-gluconic acid can generate any of four products depending on the carbonyl reduced and the stereochemistry of the resulting alcohol. We established the product of both DKGRs to be 2-keto-L-gulonic acid by high-pressure liquid chromatography and GC/MS. A 10 mM solution of 2,5-diketo-D-gluconic acid was prepared and converted to product by each purified reductase as described in Materials and Methods. High-pressure liquid chromatography analysis revealed that all of the 2,5-diketo-D-gluconic acid had been converted to a compound that coeluted with authentic 2-keto-L-gulonic acid. The reaction mixture was subsequently derivatized to give the trimethylsilyl oximes of the product and was analyzed by GC/MS. The product of both reactions was eluted as a doublet (syn and anti oximes) at 13.1 min, as was the derivative of authentic 2-keto-L-gulonic acid (data not shown). The mass spectra of the three derivatives were identical (Fig. 4). All other components present in the chromatogram were identified as derivatives of buffer components or components arising from the derivatization reagents (data not shown). No other product attributable to reduction of 2,5-diketo-D-gluconic acid was observed.

View larger version (23K): [in this window] [in a new window]   FIG. 4.   Mass spectra of product of DKGRs and of authentic 2-KLG. Mass spectra of the trimethylsilyl oxime derivative of the product of the reduction of 2,5-diketo-D-gluconic acid by DKGR C (A) and DKGR D (B) compared to a library spectrum of the derivative of authentic 2-keto-L-gulonic acid (C).

Kinetic analyses of environmental reductases. The purified reductases were evaluated for their ability to reduce keto-sugars other than 2,5-diketo-D-gluconic acid. No reduction of 2-keto-D-gluconic acid, 5-keto-D-gluconic acid, 2-keto-L-gulonic acid, xylulose, xylulose-5-phosphate, or D-fructose was observed. Kinetic parameters of the enzymes with 2,5-diketo-D-gluconic acid as the substrate were determined at 30°C (Table 4). The K_m values determined for 2,5-diketo-D-gluconic acid were 57 and 67 µM for DKGR C and D, respectively. These values are much lower than the values of 2 mM and 13 mM reported previously for the Corynebacterium reductases (35). The observed k_cat values for both environmental reductases were close to that of the more active Corynebacterium enzyme, DKGR B (Table 4). As a result, the k_cat/K_m values were much higher for the new enzymes. The new DKGRs had catalytic efficiencies more than 1,000 times higher than those of the Corynebacterium DKGR A enzyme and 20 times higher than those of DKGR B. 

NADH-dependent activity. With rare exceptions, aldo-keto reductases are specific for NADPH as a cosubstrate, including the Corynebacterium DKGRs (24, 37). Purified DKGR C and D both catalyzed the NADH-dependent reduction of 2,5-diketo-D-gluconic acid (Table 5). The reduction was much less efficient with NADH as the cosubstrate, however. The K_m for NADH was nearly 3 orders of magnitude higher than for NADPH, and the apparent k_cat was much lower with NADH as the cosubstrate. These combined effects result in catalytic efficiencies (k_cat/K_m) 600- and 130-fold lower than those measured with NADPH for DKGR C and DKGR D, respectively. Replacement of NADPH by NADH also affected the apparent K_m for 2,5-diketo-D-gluconic acid, increasing it 40- and 17-fold, respectively (Tables 4 and 5). The NADH-dependent activity was enhanced by inclusion of inorganic phosphate in the reaction buffer (Table 5). The stimulation was saturable, indicating that the phenomenon was due to binding of inorganic phosphate to the enzyme.

Thermal stability of DKGRs. The Corynebacterium reductases are thermally labile and lose activity rapidly at temperatures above 40°C (23, 35). To determine the thermal stability of the environmental reductases, we incubated each at 44°C for various periods of time using a robotic PCR thermal cycler (Robocycler Gradient 96; Stratagene) to establish temperatures rapidly and precisely (see Materials and Methods). Under these conditions, DKGR C was quite labile, losing over half its activity at the earliest time point, 0.5 min. In contrast, DKGR D was relatively stable under these conditions, with a half-life of 53.4 min. The thermal inactivation temperature of DKGR D was determined by incubating the enzyme for 10 min over a temperature gradient established by the Robocycler (Fig. 5). The enzyme retained nearly complete activity up to 45°C, after which the activity declined rapidly. The temperature under which half the activity was lost under these conditions was estimated to be 47°C.

View larger version (15K): [in this window] [in a new window]   FIG. 5.   Thermal inactivation of DKGR D. Residual activity of DKGR D was determined after 10 min of incubation at various temperatures established in a Robocycler (Stratagene). Samples were chilled to 4°C and were then assayed under standard conditions (30°C). The protein concentration was 0.08 µg/ml.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Because of their high specificity and selectivity, enzymes are potentially ideal catalysts for industrial processes. However, under the conditions required by many processes, known enzymes are often deficient in some way. The discovery of enzymes with better reaction rates, pH optima, thermal stability, solvent tolerance, substrate specificity, or other properties may be critical to the successful commercialization of new industrial processes. A proposed source of better enzymes is the vast reservoir of microbes that have not as yet been cultured successfully. Useful enzymes have been recovered from uncultured microbes by screening environmental DNA expression libraries for the desired enzymatic activities (12, 33). As an alternative approach, internal fragments of a variety of genes have been amplified directly from environmental DNA by PCR using degenerate primers to conserved sequences (6, 11, 13, 18, 19, 22, 27, 31, 32).

In the present examples we extended the latter approach by applying PCR-based gene walking techniques to internal fragments. By a series of PCRs using nested primers specific for the internal fragment paired with a random primer, the less conserved, flanking regions of the genes were obtained. Putative full-length genes were predicted from the amplified fragments, and complete, functional genes encoding the targeted enzymatic activity were obtained from environmental DNA in single PCRs. The approach was successful using a convenient method of DNA extraction that has been shown recently to yield less DNA per gram of soil than alternative methods (36). As such, the environmental DNA that we used most likely represents a subsample of the total DNA in the habitats sampled, but our success shows that it still contained sufficient diversity to yield valuable genes by this approach.

The targeted enzyme, DKGR, catalyzes a critical step in a proposed biotechnological process for the production of ascorbic acid, and the goal of our research was to discover new variants with better properties. Because DKGR is a member of the aldo-keto reductase superfamily of enzymes, it is possible that genes isolated based on homology to known examples could be other aldo-keto reductases, not true DKGRs. We found that both purified enzymes catalyzed the reduction of 2,5-diketo-D-gluconic acid to 2-keto-L-gulonic acid. The reaction was highly specific; several other structurally related keto-sugars were not reduced. The enzymes also possessed three properties of potential value in vitamin C productionhigher catalytic efficiency, use of NADH as the cosubstrate, and in the case of DKGR D, higher thermostability.

Both DKGR C and DKGR D reduced 2,5-diketo-D-gluconic acid more efficiently than the Corynebacterium enzymes (35), exhibiting >20-fold-higher k_cat/K_m values (Table 4). The increased efficiency was primarily due to much higher affinity for 2,5-diketo-D-gluconic acid. The environmental reductases possessed K_m values of 57 and 67 µM, whereas those for the Corynebacterium enzymes were 2 and 13 mM. Two recently described aldo-keto reductases encoded by the E. coli genes yafB and yqhE also reduce 2,5-diketo-D-gluconic acid to 2-keto-L-gulonic acid (41), but these enzymes also are inefficient catalysts with millimolar K_m values (L. Stols, unpublished observations).

The environmental DKGRs also accepted NADH as the cosubstrate, a rare characteristic among aldo-keto reductases (14, 17, 24). The Corynebacterium reductases are specific for NADPH (35). For a commercial process, such as the production of ascorbic acid, use of NADH in place of NADPH could result in improved performance. NADH is more abundant physiologically than NADPH and is the reduced cofactor most commonly generated from metabolism of glucose. For possible in vitro processes, NADH is more stable and much less expensive. However, the NADH-dependent activity of the new DKGRs is much less than the NADPH-dependent activity (Tables 4 and 5), and improved variants of these DKGRs would be required for efficient use of NADH. The crystal structures of aldo-keto reductases (14) and of the Corynebacterium DKGR A (15) indicate that two highly conserved, positively charged residues interact with the 2'-phosphate of NADPH. Based on sequence alignments, those residues are present in the environmental reductases as well. Apparently other differences allow these enzymes to accept NADH.

A common limitation of enzymes is their sensitivity to high temperatures. The previously known DKGRs are sensitive to thermal denaturation, exhibiting midpoint denaturation temperatures of less than 40°C (23). The two new environmental reductases differ considerably in their thermal stability in spite of being 82.5% percent identical in amino acid sequence. DKGR C is more labile than the Corynebacterium enzymes, but DKGR D is the most stable DKGR characterized to date. However, with an inactivation temperature of only 47°C (Fig. 5), the enzyme is still relatively labile. Modeling and crystallographic studies of these enzymes should provide insights into the determinants of their thermal stability, cofactor specificity, and high affinity for substrate, providing a basis for construction of mutants or chimeras optimized for commercial use.

The new DKGRs provide potential catalysts for industry but also offer a glimpse into the possible metabolic diversity of uncultured organisms. Considering the much higher efficiencies of the new enzymes, they may represent the first true examples of enzymes whose physiological function is reduction of 2,5-diketo-D-gluconic acid to 2-keto-L-gulonic acid. This reaction, however, is not known to occur in any established metabolic pathway. ORFs flanking the DKGR genes suggest that they are part of an operon, possibly encoding sequential metabolic reactions. BLAST searches found these ORFs and the DKGRs to be most homologous to hypothetical proteins of S. coelicolor or E. coli, but inspection of those genomes showed that none of these homologues was located in a similar operon. The true physiological role of DKGRs remains to be determined. Regardless, the results support the hypothesis that enzymes from uncultured organisms can provide new, alternative catalysts with desirable properties. The success of this PCR-based approach provides a valuable complement to the established, library-based methods for extracting those catalysts. Future studies targeting different enzymes and using diverse methods of DNA preparation will define the true scope and utility of the method.