Modification of the Nucleotide Cofactor-binding Site of Cytochrome P-450 Reductase to Enhance Turnover with NADH in Vivo *

NADPH-cytochrome P-450 reductase is the electron transfer partner for the cytochromes P-450, heme oxygenase, and squalene monooxygenase and is a component of the nitric-oxide synthases and methionine-synthase reductase. P-450 reductase shows very high selectivity for NADPH and uses NADH only poorly. Substitution of tryptophan 677 with alanine has been shown to yield a 3-fold increase in turnover with NADH, but profound inhibition by NADP+ makes the enzyme unsuitable for in vivo applications. In the present study site-directed mutagenesis of amino acids in the 2′-phosphate-binding site of the NADPH domain, coupled with the W677A substitution, was used to generate a reductase that was able to use NADH efficiently without inhibition by NADP+. Of 11 single, double, and triple mutant proteins, two (R597M/W677A and R597M/K602W/W677A) showed up to a 500-fold increase in catalytic efficiency (k cat/K m ) with NADH. Inhibition by NADP+ was reduced by up to 4 orders of magnitude relative to the W677A protein and was equal to or less than that of the wild-type reductase. Both proteins were 2–3-fold more active than wild-type reductase with NADH in reconstitution assays with cytochrome P-450 1A2 and with squalene monooxygenase. In a recombinant cytochrome P-450 2E1 Ames bacterial mutagenicity assay, the R597M/W677A protein increased the sensitivity to dimethylnitrosamine by ∼2-fold, suggesting that the ability to use NADH afforded a significant advantage in this in vivo assay.

ferredoxin-NADP ϩ reductase and the FMN domain homologous with flavodoxin, suggesting that P-450 reductase was formed from an ancestral gene fusion event (2,3). A fourth domain acts as a hinge to orient the two flavin-containing domains for electron transfer, and an NH 2 -terminal membrane-binding domain anchors the protein to the cytosolic side of the endoplasmic reticulum. The NADP(H) domain adopts the typical dinucleotide fold structure consisting of alternating ␣-helices and ␤-strands, with the nicotinamide cofactor binding in a cleft between the FAD and NADP(H) domains (4).
Although expression of mammalian P-450/reductase systems in bacteria has been shown to be a valuable research tool and has promise as a tool for biodegradation (5), enzyme activity in recombinant bacterial systems appears to be limited by glucose availability (6). One explanation for this might be that the generation of NADPH is largely dependent on the glucose shunt pathway, which is operative primarily in high glucose environments. Supporting this hypothesis are data from Bochner and Ames (7) that indicate that bacterial NAD(H) levels exceed NADP(H) levels by ϳ4:1. To address the hypothesis that heterologously expressed P-450 systems in bacterial cells are limited by the availability of NADPH, we have constructed a series of reductase proteins designed to use NADH more efficiently. Specific mutations in the nicotinamide cofactor 2Јphosphate-binding site were generated based on the crystal structure of the reductase from rat (4) and from sequence alignments of rat cytochrome P-450 reductase (8) with other NADPH-and NADH-specific reductases including spinach ferredoxin-NADP ϩ reductase (9), nitrate reductases from Neurospora crassa (NADPH) (10) and corn (NADH) (11), and NADH-cytochrome b 5 reductase (12).
Tryptophan 677, the penultimate amino acid in P-450 reductase, blocks access of the nicotinamide portion of the NADPH cofactor to the flavin, and it has been proposed that this amino acid plays a role in removal of NADP ϩ after hydride transfer (13,14). This amino acid is suggested to contribute to cofactor specificity by blocking the nucleotide binding site from both NADPH and NADH; NADPH, with higher binding affinity caused by a 2Ј-phosphate-binding pocket, is evidently able to displace the tryptophan and form a productive electron transfer complex, whereas NADH cannot bind tightly enough to displace the tryptophan. Consistent with this hypothesis, removal or substitution of Trp 677 greatly increases turnover with NADH but leads to inhibition by NADP ϩ at a level that would restrict in vivo applications of this enzyme. We hypothesized that mutations in the 2Ј-phosphate-binding site might reduce the affinity of the enzyme for NADP(H) and reduce or prevent the inhibition by NADP ϩ . Several highly conserved residues make up the 2Ј-phosphate-binding motif, including serine 596, arginine 597, and lysine 602; the positions of these amino acids in the NADP(H)-binding domain are shown in Fig. 1 binding site mutations with the W677A substitution yields a cytochrome P-450 reductase that uses NADH more effectively than the native enzyme without inhibition by NADP ϩ .

EXPERIMENTAL PROCEDURES
Generation of P-450 Reductase Mutants-Site-directed mutagenesis was carried out in either pIN3OR6ϫHis, a P-450 reductase bacterial expression plasmid in which a His 6 tag was placed at the N terminus of the reductase, or in the pIN3ER plasmid coexpression vectors containing the cDNAs for rat P-450 reductase and human P-450 2E1 (15). Single amino acid mutations were introduced using the QuikChange TM site-directed mutagenesis kit (Stratagene) by whole plasmid amplification in Escherichia coli XL1-Blue cells. An eight-amino acid substitution was introduced by overlap extension polymerase chain amplification as described (16) to generate a reductase with the following substitutions: S596D, R597R, E598A, Q599P, A600E, H601A, K602W, and V603D. All of the mutations were confirmed by sequencing.
Purification and Quantitation of Proteins-With all of the constructs, a fresh colony was selected for protein expression and grown overnight in LB medium at 37°C. Overnight culture was used to inoculate up to 6 liters of Terrific Broth at 37°C followed by the addition of 1 mM isopropyl ␤-D-1-thiogalactopyranoside when the A 600 reached 0.8; the culture was continued overnight at 30°C with shaking (125 rpm). The cells were harvested by centrifugation, and the membrane fraction was isolated and solubilized as described, omitting the dithiothreitol and 4-methylpyrazole (15). Wild-type P-450 reductase and mutants were purified by metal affinity chromatography using TALON IMAC metal affinity resin (Clontech). The proteins were eluted in 300 mM NaCl, 50 mM NaPO 4 , 1% Triton X-100, 10% glycerol, and 150 mM imidazole, pH 8.0, and stored in the same buffer at Ϫ20°C. The purified proteins were largely homogeneous and migrated with a molecular mass of 78.5 kDa, as did wild-type reductase (Fig. 2); immunoblotting with antibody to P-450 reductase indicated that the lower molecular mass contaminants were predominantly reductase degradation products (data not shown). Reductase was quantified by boiling to release the flavins and measuring the absorbance at 450 nm using an extinction coefficient of 23.5 mM Ϫ1 (17). Recombinant human squalene monooxygenase was purified as described (18).
Steady State Kinetics with Cytochrome c-K m and k cat determinations were carried out with 60 M cytochrome c from horse heart (Sigma) in 100 mM Tris, pH 8.0, with an ionic strength of 0.75 M with KCl. The reactions were started by the addition of nucleotide cofactor (NADPH or NADH) and were monitored for 40 s at room temperature on an HP 8453 spectrophotometer (Hewlett-Packard), following the absorbance change at 550 nm. An extinction coefficient of 21.1 mM Ϫ1 was used to quantitate cytochrome c reduction (17). For the determination of inhibition by NADP ϩ the nucleotide cofactor concentration was set at its K m for each protein, as follows: wild type, 2 M NADPH; W677A and R597M/W677A, 100 M NADH; and R597M/K602W/ W677A, 175 M NADH. The catalytic constants were estimated by nonlinear regression using Prism (GraphPad Prism 3).
Enzyme Assays-Squalene monooxygenase assays were carried out as described (18). The reactions were started by the addition of 8 nmol of 14 C-labeled squalene and were carried out for 30 min at 37°C. Product was quantified by thin layer chromatography followed by autoradiography on a Packard Instant Imager. Purified human P-450 1A2 was obtained from Oxford Biomedical, and activity was determined by measuring deethylation of ethoxyresorufin (Sigma) to resorufin, as described (19). Incubation mixtures (2 ml) contained 2 M ethoxyresorufin, 42 pmol of P-450 1A2, 40 pmol reductase, 1 mM NADPH or NADH, and 0.1 g/l of phosphatidylcholine in 100 mM potassium phosphate buffer, pH 7.7. Following a 5-min preincubation at 37°C, the reactions were started by the addition of cofactor and resorufin fluorescence was recorded for 15 min with an excitation wavelength of 510 nm and an emission wavelength of 586 nm. Turnover rates were calculated from linear time points.
Recombinant Ames Test-The activation of dimethylnitrosamine in Salmonella typhimurium strain YG7108 was carried out as described (20) with pIN3ER plasmids expressing human cytochrome P-450 2E1 and either wild-type rat cytochrome P-450 reductase or the W677A, R597M/W677A, or R597M/K602W/W677A mutants. Single colonies obtained from electroporation were used to inoculate overnight cultures in LB medium with isopropyl ␤-D-1-thiogalactopyranoside induction of protein expression. The cells were mixed with different concentrations of dimethylnitrosamine in top agar and plated onto histidine-deficient Vogel-Bonner plates. After incubation for 48 h at 37°C, the visible colonies were counted.

RESULTS
We have generated mutants of cytochrome P-450 reductase that have decreased specificity for the physiological cofactor NADPH. These mutants fall into three categories: those containing mutations in the 2Ј-phosphate-binding region of NADPH; a single mutant in which tryptophan 677, which is positioned directly opposite of the FAD isoalloxazine ring, is replaced with alanine; and double and triple combinations of these two. As expected, single and double substitutions in the 2Ј-phosphate-binding region yielded proteins with greatly reduced catalytic efficiency (k cat /K m ) with NADPH (Table I). The R597M, S596D, K602W, and R597M/K602W proteins showed a 40 -13,000-fold increase in the K m for NADPH as compared with wild-type with cytochrome c as the electron acceptor, with the K602W substitution having the least effect on K m and the S596D mutation producing the highest K m . Of these mutants, the S596D showed the most profound decrease in the k cat with both NADPH and NADH, indicating an overall detrimental effect on electron transfer; all of the mutants containing S596D had little or no activity with either cofactor. Proteins with the R597M substitution (R597M and R597M/K602W) maintained turnover numbers with NADPH and NADH that were comparable with that of the wild-type enzyme, whereas the K602W mutation decreased turnover with both cofactors in the absence of the R597M substitution. As an alternative to individual substitutions, the segment from Ser 596 to Val 603 (8 amino acids) was replaced with the corresponding segment from NADHcytochrome b 5 reductase, a closely related NADH-dependent flavoprotein (12). However, the resulting mutant protein bound flavin but had no observable cytochrome c reductase activity.
As demonstrated by Döhr et al. (13), substitution of the penultimate amino acid in P-450 reductase with alanine (W677A) produces a reductase with 1000-fold higher specificity for NADH than NADPH. We found that the W677A mutant retained only 1.2% of wild-type activity (k cat ) with NADPH, whereas the K m for NADPH remained similar to wild type when assayed with cytochrome c (Table I). In contrast, the W677A mutant had a k cat with NADH that was ϳ3-fold higher than that of wild type with NADH, whereas the K m of NADH decreased by ϳ250-fold. Despite this 700-fold increase in catalytic efficiency with NADH, the most striking feature of the W677A mutant is that it is highly inhibited by NADP ϩ (13 and Fig. 3).
The mutations in the phosphate-binding region noted above were combined with the W677A substitution to overcome the profound inhibition by NADP ϩ . We combined the S596D, R597M, and K602W mutants, as well as the R597M/K602W double mutant, with W677A. Consistent with the other S596D mutants, S596D/W677A produced a reductase with little or no activity with either cofactor. However, the R597M/W677A, K602W/W677A, and R597M/K602W/W677A proteins were active and displayed reduced catalytic efficiency with NADPH as compared with wild type and increased catalytic efficiency with NADH (Table I). Of these three proteins, the triple mutant (R597M/K602W/W677A) showed a 30,000-fold change in cofactor specificity. Inhibition of cytochrome c reduction by NADP ϩ was determined with these proteins and compared with that of wild-type reductase (Fig. 3). With the W677A protein, inhibition by NADP ϩ increased ϳ125-fold, from an IC 50 of 17.4 M (wild type) to an IC 50 of 138.3 nM. Addition of the R597M mutation to the W677A protein returned inhibition by NADP ϩ to a level comparable with wild-type reductase. Inhibition by NADP ϩ was greatly decreased in the R597M/K602W/W677A protein to an IC 50 of 870 M, 50-fold higher than the IC 50 for wild-type reductase.
We determined the ability of these three mutant proteins to support squalene monooxygenase activity, a physiological redox partner for P-450 reductase, with both NADPH and NADH (Fig. 4). Consistent with the results obtained with cytochrome c, the W677A protein showed limited activity with NADPH, having only ϳ13% of the activity of wild-type reductase. The addition of the R597M substitution to the W677A protein restored full activity with NADPH, whereas the further addition of the K602W substitution (R597M/K602W/W677A) reduced activity with NADPH to ϳ28% of wild-type activity. With NADH, the wild-type reductase supported squalene monooxygenase poorly, and the W677A protein was only slightly more active. However, the double and triple mutants were 4.4-and 3.7-fold more active with NADH than wild-type reductase.
Cytochrome P-450 1A2 catalyzes the O-deethylation of ethoxyresorufin, and the sensitivity of this assay made it ideal for further evaluation of reductase mutations. The results of these experiments are shown in Fig. 5. All of the mutants showed significantly lower deethylase activity with NADPH than wild type. With NADH all of the mutants showed similar activity, roughly twice the activity of wild type with NADH. To explore how these mutants might fare in a mixed cofactor environment, we determined ethoxyresorufin deethylase activity with equal concentrations of NADPH and NADH. The W677A reductase exhibited only 3.5% of wild-type activity, despite the presence of ample NADH, indicating a marked inhibition by NADP ϩ . In contrast, the R597M/W677A and R597M/K602W/W677A proteins showed activity levels approaching those seen with NADPH or NADH alone, indicating that inhibition by NADP ϩ was not a factor.
Based on the above studies, we anticipated that the W677A reductase would be inhibited by NADP ϩ and unable to efficiently couple with redox partners in vivo. To evaluate the effectiveness of the reductase mutants in vivo, each was coex-  pressed with cytochrome P-450 2E1 in S. typhimurium mutagenicity tester strains as described (20). The number of revertants generated with dimethylnitrosamine, a mutagenic substrate for P-450 2E1, was determined with each reductase mutant (Fig. 6). Expression of the W677A protein yielded few revertants, as expected if NADP ϩ was inhibitory in vivo. Surprisingly, the R597M/K602W/W677A protein was equally ineffective, indicating that activity in vitro does not necessarily predict performance in vivo. However, the R597M/W677A reductase yielded nearly twice as many revertants as wild type, suggesting that the improved ability to use NADH increased turnover and mutagen activation.

DISCUSSION
Cytochrome P-450 reductase has a 45,000-fold preference for NADPH over NADH, although the two cofactors differ only in the presence of a 2Ј-phosphate on the adenine-ribose of NADPH. The three-dimensional structure of P-450 reductase (4) revealed three amino acids that interact with this phosphate group: serine 596, arginine 597, and lysine 602, and it is assumed that these amino acids dictate the specificity for NADPH. Consistent with this prediction, Sem and Kasper (21) showed that substitution of a methionine for the positively charged arginine at position 597 produces a reductase with impaired ability to differentiate between the two cofactors. The studies presented here support this role for arginine 597 and demonstrate that substitution of lysine 602 leads to a further loss in specificity for NADPH. Of the four amino acids examined in this study, these two have the greatest influence on NADPH specificity; when combined these substitutions increase the K m for NADPH by 2,400-fold. The effect of these two substitutions on the K m for NADH was less predictable; although both decreased the K m individually (R597M 2-fold and K602W 5-fold), when combined the decrease was only about 1.5-fold. Moreover, the K602W mutation greatly decreased turnover with NADH. It is unclear why the substitution of tryptophan at this position should have such a detrimental effect on NADH-dependent activity, because sequence alignments suggest that a tryptophan is found at this position in NADH-specific enzymes (Fig. 7). Given the ambiguity in the alignments of the NADH-specific proteins in this region, it is possible that substitutions other than tryptophan at this position might more readily accommodate NADH binding and turnover.
Work by Scrutton, Perham, and co-workers (22)(23)(24) on the conversion of glutathione reductase to an NADH-specific enzyme suggested that an acidic residue opposite the adenineribose hydroxyls should enhance the binding of NADH. This position is occupied by serine 596 in P-450 reductase (Fig. 7). Consistent with this hypothesis, several NADH-specific reductases, including cytochrome b 5 reductase, contain an aspartate at this position. We generated the corresponding mutant in P-450 reductase (S596D), but the resulting protein exhibited poor turnover with both NADH and NADPH. Although it was anticipated that a negatively charged amino acid at this position would strongly oppose NADPH binding by repelling the 2Ј-phosphate (and in fact, the K m for NADPH increased by over 20,000-fold), the marked decrease in k cat with NADH was not expected. Although the K m for NADH decreased by greater than half, it was not enough to overcome the 90% decrease in catalytic activity. The incorporation of additional mutations in the 2Ј-phosphate-binding site (R597M and K602W) or the W677A mutation did not restore activity with either cofactor, suggesting that the S596D substitution disrupted the structure of the cofactor-binding pocket. Thus, like the K602W mutation, the substitution of an "NADH-specific" amino acid at this position had the expected effect on activity with NADPH but did not enhance activity with NADH.
In contrast to native P-450 reductase, a stable enzyme-NADP ϩ complex can be formed with a reductase in which histidine (25) or glycine (14) is substituted for tryptophan 677. This tryptophan lies directly opposite the FAD isoalloxazine ring and is suggested to serve as a gating mechanism for cofactor binding (14). In support of this idea, movement of the tryptophan is seen during NADP(H) binding and release (26). Loss or substitution of this tryptophan should enhance the binding of both NADH and NADPH, and indeed a remarkable increase in turnover with NADH is seen in Trp 677 mutants (13). However, turnover with NADPH is greatly decreased, apparently because of formation of this stable enzyme-NADP ϩ complex. Second site mutations that decrease the affinity of the enzyme for NADP(H) would be expected to decrease the stability of the enzyme-NADP ϩ complex and decrease the inhibition by NADP ϩ . As expected and in contrast to the W677A protein, the R597M/W677A and R597M/K602W/W677A proteins were inhibited only at high NADP ϩ concentrations (Fig. 3) and retained good activity in the presence of both cofactors in an in vitro P-450 assay (Fig. 5).
Although we identified mutations that enhanced the ability of P-450 reductase to use NADH in the presence of NADPH, we did not create a reductase that is more efficient with NADH than NADPH. Even with the best mutants (R597M/W677A and R597M/K602W/W677A), the catalytic efficiency values (k cat / K m ) for NADPH and NADH were at best equivalent. Thus, it remains unclear how to engineer a P-450 reductase to preferentially use NADH. Notably, a number of NADH-utilizing enzymes exhibit a preference for NADH over NADPH; for example, turnover with NADPH in NADH-dependent E. coli dihydrolipoamide dehydrogenase is undetectable (24). Sequence comparisons of the cofactor-binding region of this protein to P-450 reductase do not reveal obvious differences that might explain the greater selectivity of the dehydrogenase for NADH. Further biochemical and perhaps biophysical studies on P-450 reductase will be needed to identify the additional changes needed to enhance selectivity for NADH.
The studies reported here do indicate that it is possible to engineer P-450 reductase for use in vivo by rational design and subsequent evaluation of enzymes in vitro. The R597M/W677A protein, which showed strong NADH-dependent activity in vitro, produced a more sensitive Ames test, presumably be-cause of a greater ability to utilize NADH for cytochrome P-450 2E1-mediated activation of dimethylnitrosamine. The reductases generated in this study are especially suited for biotechnological uses in growth conditions that deplete NADPH versus NADH, such as in situ bacterial bioremediation schemes. Moreover, the initial Ames tests presented here suggest that NADPH might be limiting under standard laboratory growth conditions, as suggested previously (6). In addition to providing a valuable tool for understanding the determinants of nucleotide-cofactor specificity, these mutants might also lend themselves to creation of bioremediation schemes less influenced by nutrient levels, as well as cost-effective reconstitution of enzyme systems that do not require the use of expensive reducing equivalents from NADPH.