Expansion of Substrate Specificity and Catalytic Mechanism of Azoreductase by X-ray Crystallography and Site-directed Mutagenesis*

AzoR is an FMN-dependent NADH-azoreductase isolated from Escherichia coli as a protein responsible for the degradation of azo compounds. We previously reported the crystal structure of the enzyme in the oxidized form. In the present study, different structures of AzoR were determined under several conditions to obtain clues to the reaction mechanism of the enzyme. AzoR in its reduced form revealed a twisted butterfly bend of the isoalloxazine ring of the FMN cofactor and a rearrangement of solvent molecules. The crystal structure of oxidized AzoR in a different space group and the structure of the enzyme in complex with the inhibitor dicoumarol were also determined. These structures indicate that the formation of a hydrophobic part around the isoalloxazine ring is important for substrate binding and an electrostatic interaction between Arg-59 and the carboxyl group of the azo compound causes a substrate preference for methyl red over p-methyl red. The substitution of Arg-59 with Ala enhanced the Vmax value for p-methyl red 27-fold with a 3.8-fold increase of the Km value. This result indicates that Arg-59 decides the substrate specificity of AzoR. The Vmax value for the p-methyl red reduction of the R59A mutant is comparable with that for the methyl red reduction of the wild-type enzyme, whereas the activity toward methyl red was retained. These findings indicate the expansion of AzoR substrate specificity by a single amino acid substitution. Furthermore, we built an authentic model of the AzoR-methyl red complex based on the results of the study.

We have previously reported the crystal structure of oxidized AzoR at 1.8 Å resolution (8). That was the first report on the structure of AzoR orthologues and of FMN-dependent NADHazoreductase. This structure has provided a wealth of information, including the overall fold, the nature of the dimer, and the interactions of the FMN cofactor. A structural similarity search revealed that the overall structure of AzoR resembles that of mammalian NQO1 2 (FAD-dependent NAD(P)H:quinone oxidoreductase 1, originally called DT-diaphorase) (9), ROO from Desulfovibrio gigas (FAD-dependent rubredoxin:oxygen oxidoreductase) (10), and yeast YLR011wp (FMN-dependent NAD(P)H:ferric iron oxidoreductase) (11) without the explicit overall amino acid sequence similarity. They are homodimeric flavodoxin-like proteins.
Many types of azoreductases have been isolated and extensively studied in the field of environmental biotechnology (12)(13)(14)(15)(16)(17)(18)(19)(20). However, those studies have focused mainly on the screening of microorganisms exhibiting azoreductase activity. Because their detailed molecular reaction mechanisms remain largely unknown, AzoR is an interesting case for azoreductase research. Structural and enzymatic insights into AzoR provide knowledge of the molecular mechanism underlying the reduction of azo compounds and would also provide important clues for substrate-specificity expansion techniques that will be essential for industrial use of the enzyme in the biodegradation process of azo compounds.
In this study, we have determined the structure of reduced AzoR. This work reveals the structural changes upon reduction of the enzyme and is the first report on the crystal structure of homodimeric flavodoxin-like proteins in the reduced state. In addition, we have determined the crystal structure of oxidized AzoR in a different space group as well as the crystal structure of the enzyme in complex with the inhibitor dicoumarol. Substrate-specificity analysis and site-directed mutagenesis were also performed. According to these analyses, we succeeded in expanding AzoR substrate specificity, and we have built an authentic model of an AzoR-methyl red complex. Finally, the mechanisms underlying the reductive cleavage of azo compounds are discussed.

EXPERIMENTAL PROCEDURES
Protein Preparation for Crystallization-The recombinant AzoR used for crystallization was expressed and purified as described previously (1,21).
Improvement of the Quality of the Tetragonal Crystals of Oxidized AzoR-The improved oxidized tetragonal crystals (P4 2 2 1 2) were obtained in the same crystallization method as reported previously (21), except that 0.1 mM warfarin (4-hydroxy-3-(3-oxo-1-phenylbutyl)coumarin) was added to the crystallization solution as an additive reagent for crystal growth (the electron density of warfarin was not observed). The crystals grew to full size (0.1 ϫ 0.1 ϫ 0.6 mm) within 1 week.
Chemical Reduction of the Tetragonal Crystals of Oxidized AzoR-Crystals of the enzyme in the reduced state were prepared by transferring the oxidized tetragonal crystals to a degassed reducing solution containing 200 mM MgCl 2 , 100 mM HEPES, pH 7.5, 30% (v/v) 2-propanol, 30% (v/v) ethylene glycol, and a saturated concentration of sodium dithionite at 4°C. After 10 min, the crystals were transferred to fresh reducing solution. This procedure was repeated two more times. The crystals lost their bright yellow color and became transparent over the course of the treatment. They were then immediately flash-cooled in a stream of nitrogen and stored in liquid nitrogen prior to data collection. It was verified that the crystal remained transparent after data collection.
Co-crystallization with Dicoumarol-The purified protein was dialyzed against a solution containing 10 mM Tris-HCl, pH 8.0, 0.1 mM FMN, 4% (v/v) pyridine, and 2 mM dicoumarol. Crystals of AzoR in complex with dicoumarol were obtained from a drop made by mixing 16 mg/ml of the protein solution mentioned above and an equal volume of reservoir solution containing 200 mM NaCl, 100 mM HEPES, pH 7.5, and 20% (w/v) polyethylene glycol 3000. The drop was equilibrated over the reservoir solution by the hanging-drop vapor diffusion method at 20°C. The crystals grew to full size (0.05 ϫ 0.5 ϫ 0.5 mm) within 1 week. These crystals belonged to the tetragonal space group P42 1 2.

Crystallization of the Orthorhombic Crystals of Oxidized
AzoR-The purified protein was dialyzed against a solution containing 10 mM Tris-HCl, pH 8.0, and 0.1 mM FMN. The orthorhombic (P2 1 2 1 2 1 ) crystals of oxidized AzoR were obtained from a drop made by equal volumes of three solutions: 8 mg/ml of the protein solution mentioned above, 100 mM NAD ϩ solution (the electron density of NAD ϩ was not observed), and a reservoir solution containing 200 mM NaOAc, 200 mM sodium cacodylate, pH 6.7, 15% (w/v) polyethylene glycol 8000, and 3% (v/v) dimethyl sulfoxide. The drop was equilibrated over the reservoir solution by the hanging-drop vapor diffusion method at 25°C. The crystals grew to full size (0.03 ϫ 0.05 ϫ 0.5 mm) within 2 weeks.
Data Collection and Processing-All diffraction data were collected under cryogenic conditions at 100 K. Prior to data collection, the crystals were soaked in a reservoir solution containing 30% (v/v) ethylene glycol or 25% (v/v) glycerol as a cryoprotectant. All diffraction data were collected at KEK (Tsukuba, Japan). The beamlines used are shown in Table 1. Data were reduced with MOSFLM, SCALA, and TRUNCATE from the CCP4 program suite (22).
Structure Determination-The initial structures of all the crystals were obtained by molecular replacement with MOL-REP (23) using the 1.8 Å resolution structure of oxidized AzoR as a search model (8). The solutions were then improved by ARP/wARP (24) followed by iterative manual model building with XtalView (25). All the stages of maximum likelihood refinement were carried out with REFMAC5 (26). For the refinement of the oxidized and reduced tetragonal crystal structures, the restraint of the planarity of the isoalloxazine ring was removed from the standard REFMAC5 library to allow the model to adopt the omit density map more precisely.
Model Analysis-The quality of the model was checked with PROCHECK (27). LSQKAB was used to superpose the molecules and to calculate the root mean square deviation between pairs of equivalent C␣ atoms and all atoms of the proteins (28). Structure figures were prepared with PyMOL (29).
Mutant Preparations-For enzymatic analyses, mutations and C-terminal His tag were introduced into the AzoR gene by two rounds of PCR with pETacpD as a template (1). The NADH-methyl red reductase activity of the His-tagged wildtype AzoR was very similar to that of the non-tagged wild-type AzoR.
To mutate Arg-59 to Ala, the following pairs of oligonucleotide primers were used for the first PCR: 5Ј-terminal sense primer for AzoR (5Ј-GGGAATTCCATATGAGCAAGGTAT-TAGTTCTTAAATCCAGC-3Ј containing an NdeI site) and R59A mutation antisense primer (5Ј-CGGCGCATCGCTCG-GAGCCAGAGCGCCAACCAGTTC-3Ј) were used to amplify the 5Ј-terminal part of the AzoR gene. R59A mutation sense primer (5Ј-GAACTGGTTGGCGCTCTGGCTCCGAGCGA-TGCGCCG-3Ј) and 3Ј-terminal antisense primer for AzoR (5Ј-AAACCGCTCGAGTTAGTGATGGTGATGGTGGTGTG-CAGAAACAATGCTGTCGATGGC-3Ј containing an XhoI site and His 6 tag sequence) were used to amplify the 3Ј-terminal part of the AzoR gene. PCR products were used as templates for the second PCR, which used the 5Ј-terminal sense primer and 3Ј-terminal antisense primer for AzoR. To generate other mutants, the following primers were used in place of R59A mutation sense and antisense primers: Y120A mutation sense primer (5Ј-GCAGGCGTTACTTTCCGCG-CTACCGAGAACGGTCCG-3Ј) and Y120A mutation antisense primer (5Ј-CGGACCGTTCTCGGTAGCGCGGAA-AGTAACGCCTGC-3Ј) for Tyr-120 to Ala mutant; F162A mutation sense primer (5Ј-CCACGTTCCTCGGCGCTATC-GGCATTACCGATG-3Ј) and F162A mutation antisense primer (5Ј-CATCGGTAATGCCGATAGCGCCGAGGAAC-GTGG-3Ј) for Phe-162 to Ala mutant. To generate the Histagged wild-type AzoR gene as a control that is catalytically very similar to AzoR, the 5Ј-terminal sense primer and the 3Ј-terminal antisense primer for AzoR described above were used for PCR with pETacpD as a template. The PCR product of each mutated AzoR was inserted between the NdeI and the XhoI sites of the expression vector pET-22b (Novagen). The entire DNA sequence was confirmed by DNA sequencing.
His-tagged wild-type and mutant AzoR proteins were expressed in E. coli BL21(DE3) at 37°C in Luria Bertani medium containing 100 g/ml ampicillin. Protein expression was induced by adding 1 mM isopropyl-␤-D-thiogalactopyranoside to early exponential phase cultures (A 600 ϳ 0.5) for 3 h. Bacteria were lysed in a solution containing 50 mM Tris-HCl, pH 7.5, 2 mM 2-mercaptoethanol, 500 mM NaCl, 20 mM imidazole, and 10 mg/ml lysozyme by sonication. The lysate was centrifuged, and the His-tagged proteins were purified by gravity-flow chromatography using nickel-nitrilotriacetic acid-agarose (Qiagen) according to the manufacturer's instructions. Eluted proteins were stored at Ϫ80°C in a solution containing 20 mM Tris-HCl, pH 7.5, 2 mM 2-mercaptoethanol, 200 mM NaCl, and 0.1 mM FMN. Protein samples were analyzed on SDS-PAGE and were more than 95% pure.
Enzyme Assays-The NADH-methyl red reductase activity for each mutant was determined spectrophotometrically by a method described previously (1). The initial reaction rates were fitted to the Equation 1 1 where v is the initial reaction rate, V is the maximum reaction rate at infinite substrate concentrations, A and B are the concentrations of methyl red and NADH, respectively, and K A and K B are their corresponding Michaelis constants. The initial reaction rate of p-methyl red reduction was determined in the same manner as methyl red, except that a 460-nm wavelength was used to monitor the decrease in absorbance of p-methyl red and a molar absorption coefficient of 17310 M Ϫ1 cm Ϫ1 was used. K B was applied to obtain Michaelis constants for p-methyl red because K B does not depend on the kind of azo compound. All of these assays were performed in triplicate.

RESULTS AND DISCUSSION
Structures of Oxidized and Reduced AzoR in Tetragonal Crystals-The structure of oxidized AzoR in tetragonal crystal was previously determined at 1.8 Å resolution (8). We have now improved the crystallization conditions and increased the resolution to 1.4 Å. In addition, we have determined the structure of reduced AzoR at 1.7 Å resolution using chemically reduced tetragonal crystals. The results allow the stringent comparison of the two states, especially the conformation of the isoalloxazine ring of FMN. The structure of this ring was unambiguously determined without the restraint of planarity for both the oxidized and reduced enzymes (Fig. 1). The reduced crystals turned colorless, indicating that two-electron reduction of the flavin occurred. The data collection and the final refinement statistics are summarized in Table 1.
The AzoR structures are nearly identical between the oxidized and reduced states. The root mean square deviations between equivalent 193 C␣ atoms and all atoms of the amino acid residues, except hydrogens, in the two states are 0.282 and 0.582 Å, respectively. However, prominent structural differences are found in the active site. In the oxidized enzyme, the isoalloxazine ring of the flavin is nearly planar but shows a slight twist conformation (Fig. 1, A and C). Upon reduction of the enzyme, the isoalloxazine ring adopts a butterfly bend conformation along the N5-N10 axis caused by a shift of the dimethyl and pyrimidine rings toward the re-face by an angle of ϳ15° (  Fig. 1, B, D, and F). The slight twist conformation found in the oxidized state is retained in the reduced state, and N5 and N10 of the isoalloxazine ring move up toward the si-side. Therefore, the central ring of the isoalloxazine is distorted to a twist-boat conformation upon reduction. The movement of N10 is small relative to that of N5. The other structural changes associated with reduction are a significant movement of a water molecule and a loss of two spherical electron density peaks on FMN. The water molecule, which is hydrogen-bonded to O␦-1 of Asn-97 in the oxidized state, still forms a hydrogen bond in the reduced state (Fig. 1E). This water molecule, however, moves by 2.4 Å and makes an additional hydrogen bond with N5 of the isoalloxazine ring upon reduction. The protonation of the N5 of the reduced isoalloxazine ring has been proved biochemically (30). This hydrogen bond may help stabilize the upward movement of the N5 atom, which adopts sp 3 hybridization in the reduced state (31). The two spherical electron density peaks on FMN, which were also found in the oxidized 1.8 Å resolution structure as reported previously, are currently modeled by water molecules, although they are slightly larger than the electron density peaks of water molecules. The loss of two peaks upon reduction may be attributable to electronic restructuring of the reduced flavin.
The Conformational Change of the Isoalloxazine Ring and the Solvent Rearrangement upon Reduction-The butterfly bending direction in reduced AzoR is the opposite of the conformation predicted for reduced free flavin (32)(33)(34). This considerable disagreement suggests that the peptide moiety may greatly influence the equilibrium conformation of the isoalloxazine system in AzoR. On the other hand, although it is not clear why the central ring of the isoalloxazine adopts the twist-boat form in the reduced enzyme, which is energetically less favorable than a chair form, this conformation would affect the redox potential of AzoR. In addition, because free oxidized isoalloxazine is planar (35), the twisted form of the oxidized isoalloxazine of AzoR may favor reduction.
In the reduced state of the enzyme, the up-moved protonated N5 of the isoalloxazine ring is stabilized by the hydrogen bond with O␦-1 of Asn-97 through a bridging water molecule (Fig.  1E). The protonation of the N5 of the reduced isoalloxazine ring has been proved biochemically (30). If methyl red binds to reduced AzoR on top of the isoalloxazine ring as described below, this water molecule must be replaced, resulting in the disruption of the hydrogen bond network. This may cause destabilization of the protonated state of the N5 atom, and a hydride anion would then be released efficiently from the N5 position to an electron acceptor azo compound.
Structure of AzoR in Complex with Dicoumarol-Dicoumarol inhibits the azo reduction activity of AzoR (1). To obtain clues to the substrate binding mode, we co-crystallized AzoR with dicoumarol and have determined the structure at 2.3 Å resolution. Although the average completeness of the data of the AzoR-dicoumarol complex is 81.1% (Table 1), there was not any uninterpretable electron density region in model building. The crystal structure shows that dicoumarol is bound to the space above the isoalloxazine ring mainly by hydrophobic interactions ( Fig. 2A). In their interactions, the two coumarin rings of dicoumarol are sandwiched as a result of ring-stacking interactions. The one ring is sandwiched between the phenyl group of Phe-162Ј and the isoalloxazine ring of FMN, 3

and the other
is the i-th intensity measurement of reflection hkl, including symmetry-related reflections, and ͗I(hkl)͘ is its average. b r ϭ 100 ΑʈF obs ͉ Ϫ ͉F calc ʈ/ Α͉F obs ͉, where F obs and F calc are the observed and calculated structure factors, respectively. c R free was calculated by using 5% of randomly selected reflections that were excluded from the refinement. ring is sandwiched between the phenolic groups of Tyr-120Ј and Tyr-178. The distances between these planes are within 4.0 Å and are stabilized by -stacking systems. The phenolic groups of Phe-98 and Phe-118Ј are also within 4.0 Å from and interact with dicoumarol, forming a hydrophobic pocket with the phenolic group of Phe-162Ј in the deep part of the active site cavity. In addition, the C atom of the carboxyl group of Gly-141, the C␣ of Gly-142, and the C␤ of Ala-177 of AzoR directly interact with dicoumarol in the range of 3.4 to 3.8 Å and contain no hydrogen bond. These observations suggest that the hydrophobic part around the isoalloxazine ring participates in the binding of hydrophobic substrates.
Mutational Analysis of AzoR-To elucidate the role of residues in the active site of AzoR, a site-directed mutagenesis study was performed. Tyr-120 and Phe-162 were chosen because they are predicted to participate in the substrate binding on top of the isoalloxazine ring, as observed in the AzoRdicoumarol structure. In addition, a structural comparison of the active site with that of NQO1 revealed the conservation of these amino acid residues, and they are thought to play a similar role in AzoR (Fig. 2, A-C) (8). That is, in NQO1 the side chain of Tyr-128Ј interacts with electron acceptor duroquinone (2,3,5,6-tetramethyl-quinone) by hydrogen bond and hydrophobic interactions (Fig. 2C) (9,36). Phe-178Ј of NQO1 stacks onto duroquinone, resulting in the sandwiching of its aromatic ring between the isoalloxazine ring and Phe-178Ј.
The K m values for NADH of Y120A and F162A mutants are ϳ10-fold larger than those of the wild-type enzyme, whereas the difference between the K m values for both the azo compounds of Y120A and F162A mutants are not significant compared with the case of NADH (Table 2). Although the V max value for methyl red of the Y120A mutant is moderately changed, the other V max values are comparable with those of the wild type. These results indicate that Tyr-120 and Phe-162 are mainly responsible for NADH binding rather than methyl red binding.
Substrate Specificity-To investigate the substrate recognition mechanism of AzoR, substrate specificity was analyzed by the steady-state kinetic procedures for methyl red and its analogue using the wild-type enzyme. Our previous data, from an experiment using ethyl red, showed that the size of the N,NЈdimethylaniline moiety of methyl red was not relevant to substrate specificity (Fig. 3) (1). Next, we focused on the carboxyl group of the other ring and selected p-methyl red as a substrate. As shown in Table 2, AzoR exhibits substrate specificity for methyl red over p-methyl red. This indicates that the position of the carboxyl group is the determinant for the substrate specificity of AzoR.
Structures of Oxidized AzoR from the Orthorhombic Crystal-We have also determined the structure of oxidized AzoR from the orthorhombic crystal at 2.1 Å resolution. In this structure, Asp-62 of the crystallographically related neighboring molecule penetrates the active site and forms a salt bridge with Arg-59Ј on top of the isoalloxazine ring of FMN (Fig. 2D). This salt bridge fixes a loop that includes Arg-59Ј and permits the full equivalent. NQO1 is also a homodimeric enzyme in a similar fashion to AzoR (36). Thus, primes will be used in a similar way as with AzoR.  Fig. 1. B, human NQO1 in complex with dicoumarol (Protein Data Bank accession code 2F1O). C, rat NQO1 in complex with duroquinone (Protein Data Bank accession code 1QRD). B and C, FAD, dicoumarol, duroquinone, and the amino acid residues involved in dicoumarol or duroquinone binding and conserved in AzoR and NQO1 are shown in a similar manner to A. D, AzoR in the orthorhombic crystal. FMN and the side chain of Arg-59Ј are shown in a similar manner to A. The side chain of Asp-62 of the crystallographically related neighboring molecule is shown in a similar manner to Arg-59Ј except that the carbon atoms are colored in green. The loops containing Arg-59Ј and Asp-62 are colored in yellow and green, respectively.

TABLE 2 Kinetic parameters of wild-type and mutant AzoR
The activity is expressed as the mean Ϯ S.E. (n ϭ 3). The relative values to wild type (WT) are indicated in parentheses. MR, methyl red; PMR, p-methyl red. tracing of this loop, which is not observed in other crystals because of poor electron density.

Enzyme
As described under "Substrate Specificity," the position of the carboxyl group of the azo compounds is the AzoR substrate-specificity determinant. On the other hand, Arg-59 is the only positively charged residue that can be positioned over the isoalloxazine ring, as has been demonstrated in the orthorhombic crystal structure. These facts imply an electrostatic interaction between Arg-59 and the carboxyl group of the azo compounds, which govern the substrate specificity.
Expansion of Substrate Specificity-As described above, the electrostatic interaction between Arg-59 and the carboxyl group of the azo compounds was predicted to determine the substrate specificity. In this case, the mutation of this residue was expected to contribute to a conversion of the substrate specificity. To investigate these observations, the kinetic parameters were determined ( Table 2).
The R59A mutant resulted in a 27-fold increase in the V max value for p-methyl red reduction compared with that of the wild-type enzyme, which is comparable with the V max value for the methyl red reduction of the wild-type enzyme, with a 3.8fold increase of K m value. This result indicates that the R59A mutant is capable of reducing p-methyl red much more effectively than the wild-type enzyme. On the other hand, the activity toward methyl red was retained. These results indicate the acquisition of AzoR's p-methyl red reduction activity besides the methyl red reduction activity and demonstrate the interaction between Arg-59 and the carboxyl group of the azo compounds.
Azo Compound Binding-Although azoreductases have long been studied, there has been no report on the structure of the enzyme in complex with the azo compound nor has there been any spectroscopic study related to the azo compound binding mode. The very low solubility of azo compounds may make it difficult to prepare samples. However, the crystal structure of AzoR in complex with dicoumarol, the orthorhombic crystal structure, and the enzymatic analyses shown here provide the first view of this binding.
It has been demonstrated that the locations of some amino acid residues around the isoalloxazine ring are conserved between AzoR and NQO1 (8). In NQO1, one coumarin ring of dicoumarol is sandwiched between the isoalloxazine ring and Phe-178Ј (Fig. 2B) (37), and this coumarin ring is in the same position as duroquinone (Fig. 2C) (9,36), an electron acceptor of NQO1. This Phe-178Ј of NQO1 exists in a similar position as Phe-162Ј in AzoR (Fig. 2, A-C). In addition, NQO1 can exploit methyl red as an electron acceptor as well as AzoR (38). Therefore, the coumarin ring of dicoumarol sandwiched between the isoalloxazine ring and Phe-162 in AzoR can be expected to mimic the binding of the ring structure of the electron acceptor, such as the aromatic ring of methyl red. Indeed, a sandwiching system of aromatic electron acceptors by means of the isoalloxazine ring and an aromatic side chain over the isoalloxazine ring is seen in several flavoproteins (39 -41). In addition to the sandwiching system, the second clue about the mechanism underlying azo binding is the interaction between Arg-59 and the carboxyl group of the azo compounds, as has been demonstrated by the substrate-specificity analysis and by the acquisition of the p-methyl red reduction activity of the R59A mutant. Fig. 4 shows the proposed model of the AzoR-methyl red complex using the orthorhombic crystal structure of AzoR. In this model, the carboxyl group of methyl red is superposed onto the side chain of Asp-62 of the neighboring molecule, thus forming a salt bridge with Arg-59Ј, and the N,NЈdimethylaniline moiety is sandwiched between Phe-162Ј and the isoalloxazine ring by consideration of -stacking interaction, as has been observed in the NQO1-duroquinone complex structure (9,36,42). As a result, the azo group is placed just above and ϳ3.5 Å from N5 of the isoalloxazine ring. This configuration satisfies conditions for an efficient direct hydride transfer (9,43). In addition, our proposed model can account for the substrate specificity of azo compounds and is consistent with our proposition that the interaction between the side chain of Arg-59 and the carboxyl  group of the azo compounds determines the substrate specificity; if the carboxyl group of p-methyl red forms a salt bridge with Arg-59, the azo group would be directed away from N5 of the isoalloxazine ring and would no longer be allowed to accept hydride anion. This unfavorable arrangement, however, is avoided by the elimination of the Arg-59 side chain. Consequently, the R59A mutant is thought to obtain the ability to reduce p-methyl red. On the other hand, the active site has enough space to accommodate the N,NЈ-diethylaniline moiety instead of the N,NЈ-dimethylaniline moiety of methyl red. This may be a reason why AzoR can reduce ethyl red as well as methyl red. In addition, the hydrophobic interactions may be enough to stabilize the AzoR-azo compound binding; therefore, the methyl red reduction activity of R59A mutant is thought to be retained. The small effects observed on the kinetic parameters for both methyl red and p-methyl red of the Y120A mutant are probably due to the absence of the direct involvement by, or to the minor effects of, Tyr-120 on electron acceptor reduction. Although F162A has unexpectedly little effect on activity, this can be explained by additional hydrophobic interactions between the substrates and amino acid residues surrounding Phe-162. That is, Phe-98, Ala-112Ј, Phe-118Ј, and Phe-159Ј form a hydrophobic pocket with Phe-162Ј and may compensate for the role of Phe-162Ј in the active site cavity of the F162A mutant (Fig. 4).
The Reductive Cleavage of Azo Compounds-Stoichiometry experiments have demonstrated that AzoR required 2 mol of NADH for the reductive cleavage of 1 mol of methyl red into 2-aminobenzoic acid and dimethyl-p-phenylenediamine (1). Furthermore, the flavin rings in the two identical catalytic sites are ϳ25 Å apart in AzoR, and there are not any possible pathways for intramolecular electron transfer between these sites. Thus, it seems that the reaction proceeds not by a simultaneous four-electron reduction but by two subsequent two-electron transfers via a hydrazo compound, namely, two cycles of a pingpong mechanism or one cycle of a ping-pong mechanism followed by a non-enzymatic reduction in the presence of NADH. In the case involving two cycles, the aromatic ring of the hydrazo compound formed from methyl red may also be stacked on Phe-162 using a -stacking interaction as well as in the case of methyl red, with its hydrazo group (-NH-NH-) laid almost on top of N5 of the isoalloxazine ring. Attack by a hydride anion then breaks the -NH-NH-bond and the respective two amines are produced. Further experiments are needed to elucidate the four-electron transfer mechanism.