Phenol Hydroxylase from Bacillus thermoglucosidasius A7, a Two-protein Component Monooxygenase with a Dual Role for FAD*

A novel phenol hydroxylase (PheA) that catalyzes the first step in the degradation of phenol in Bacillus thermoglucosidasius A7 is described. The two-protein system, encoded by the pheA1 and pheA2 genes, consists of an oxygenase (PheA1) and a flavin reductase (PheA2) and is optimally active at 55 °C. PheA1 and PheA2 were separately expressed in recombinant Escherichia coli BL21(DE3) pLysS cells and purified to apparent homogeneity. The pheA1 gene codes for a protein of 504 amino acids with a predicted mass of 57.2 kDa. PheA1 exists as a homodimer in solution and has no enzyme activity on its own. PheA1 catalyzes the efficient ortho-hydroxylation of phenol to catechol when supplemented with PheA2 and FAD/NADH. The hydroxylase activity is strictly FAD-dependent, and neither FMN nor riboflavin can replace FAD in this reaction. The pheA2 gene codes for a protein of 161 amino acids with a predicted mass of 17.7 kDa. PheA2 is also a homodimer, with each subunit containing a highly fluorescent FAD prosthetic group. PheA2 catalyzes the NADH-dependent reduction of free flavins according to a Ping Pong Bi Bi mechanism. PheA2 is structurally related to ferric reductase, an NAD(P)H-dependent reductase from the hyperthermophilic Archaea Archaeoglobus fulgidus that catalyzes the flavin-mediated reduction of iron complexes. However, PheA2 displays no ferric reductase activity and is the first member of a newly recognized family of short-chain flavin reductases that use FAD both as a substrate and as a prosthetic group.

Phenolic compounds constitute one of the largest groups of natural products. They are predominantly found in plants, where they occur in a great variety of structures and functions. During the last century, the natural pool of phenolic compounds has been increased with products of industrial origin. Many of these synthetic compounds cause environmental pollution and human health problems as a result of their persistence, toxicity, and transformation into hazardous intermediates (1,2).
The aerobic mineralization of natural and xenobiotic phenolic compounds by mesophilic microorganisms has been intensively investigated, and numerous pathways are known (3). Almost invariably, phenols are first converted into more reac-tive dihydroxylated intermediates and then subjected to intraor extradiol ring cleavage by molecular oxygen. The initial hydroxylation of the phenolic ring usually is catalyzed by single-component NAD(P)H-dependent flavoprotein monooxygenases (4). These enzymes share a typical dinucleotide-binding fold for complexation of the FAD cofactor while lacking a common NAD(P)-binding fold (5,6). Because the regioselective hydroxylation of phenols is notoriously difficult to achieve by chemical methods, the mechanistic and structural features of single-component flavoprotein monooxygenases have received much attention (7)(8)(9)(10)(11). The reduced forms of these enzymes react with molecular oxygen to yield a transiently stable flavin C4a-hydroperoxide species that is involved in substrate oxygenation. For Pseudomonas p-hydroxybenzoate hydroxylase (12)(13)(14) and phenol hydroxylase from yeast (15), it was shown that the oxygenation reaction takes place in the inner part of the protein and that this aprotic environment is crucial for preventing uncoupling of flavin reduction from substrate hydroxylation. With 4-hydroxyphenylacetate 3-hydroxylase from Pseudomonas putida, the binding of a second protein component prevents the uncoupling of substrate hydroxylation (16 -18). This enzyme is an unusual example of a two-component flavoprotein hydroxylase in which flavin reduction and substrate oxygenation take place in the same protein.
Relatively little is known about the mineralization of phenolic compounds by thermophilic microorganisms (19). Several Bacillus species isolated from geographically distinct thermal sources degrade phenol at 65°C via the meta-cleavage pathway ( Fig. 1) (20 -22). The initial conversion of phenol to catechol in these microorganisms requires two protein components that are encoded by the pheA1 and pheA2 genes (23). Characterization of the Bacillus phenol hydroxylase system appeared to be severely hampered by the low yield and instability of the purified enzymes. This prompted us to clone the Bacillus pheA genes in an Escherichia coli expression system (24).
In this study, we describe the overexpression, purification, and characterization of the recombinant phenol hydroxylase of Bacillus thermoglucosidasius A7. We show that this two-protein enzyme belongs to a newly recognized family of flavin-dependent monooxygenases that carry out the reductive and oxidative half-reactions on separate polypeptide chains. Members of this family are involved in various biological processes, including the biosynthesis of antibiotics (25)(26)(27), the desulfurization of fossil fuels (28), the degradation of chelating agents (29), and the oxidation of aromatic compounds (30). In contrast to most other family members, the reductase component of phenol hydroxylase from B. thermoglucosidasius A7 harbors a tightly bound FAD. Evidence is provided that the FAD cofactor is directly involved in the reduction of free flavins.
Enzyme Purification-All purification steps were carried out at room temperature. PheA1 and PheA2 were purified from transformed E. coli cells harboring the pheA1 and pheA2 genes (24), respectively. Recombinant cells were grown for 16 h in medium containing 10 g/liter Tryptone, 5 g/liter yeast, and 5 g/liter NaCl (pH 7.4) at 30°C, followed by induction with 1 mM isopropyl-␤-D-thiogalactopyranoside for 3 h.
Purification of PheA1 (Oxygenase Component)-Washed recombinant E. coli cells (15 g) were suspended in 15 ml of 50 mM sodium phosphate (pH 7.0) containing 0.5 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 mg of DNase and passed two times through a precooled French press operating at 10,000 p.s.i. Following centrifugation for 30 min at 15,000 ϫ g to remove cellular debris, the supernatant was made 0.5% in protamine sulfate from a 2% stock solution. The protamine sulfate aggregates were removed by centrifugation for 15 min at 15,000 ϫ g, and the resulting supernatant was applied to a Q-Sepharose Fast Flow column (1.6 ϫ 10 cm) equilibrated with 50 mM sodium phosphate (pH 7.0) containing 0.5 mM EDTA (starting buffer). After washing with starting buffer, the bound protein was eluted with a linear gradient of 0 -0.6 M NaCl in 100 ml of starting buffer. The PheA1 protein eluting between 0.3 and 0.5 M NaCl was concentrated by ultrafiltration (Amicon YM-30 membrane) and dialyzed against 5 mM sodium phosphate (pH 7.0). The PheA1 protein was then loaded onto a hydroxylapatite column (1.6 ϫ 10 cm) pre-equilibrated with 5 mM sodium phosphate (pH 7.0). After washing, PheA1 was eluted with a linear gradient of 5-500 mM sodium phosphate (pH 7.0) in 100 ml of water. Pooled fractions (80 -200 mM) were concentrated by ultrafiltration and stored at 5 mg/ml in 50 mM sodium phosphate (pH 7.0) at Ϫ70°C.
Purification of PheA2 (Reductase Component)-The preparation of cell extract from recombinant E. coli cells, the protamine sulfate precipitation, and the ion exchange chromatography step were carried out under the same conditions as described above. HiLoad Q-Sepharose fractions containing FAD reductase activity (eluting between 0.1 and 0.5 M NaCl) were pooled and concentrated by ultrafiltration (Amicon YM-30 membrane). After the addition of pulverized ammonium sulfate to a final concentration of 1.4 M, the protein solution was loaded onto a phenyl-Sepharose column (1.6 ϫ 10 cm) equilibrated with 50 mM potassium phosphate (pH 7.0) containing 1.4 M ammonium sulfate. After washing with starting buffer, bound protein was eluted with a linear descending gradient of ammonium sulfate (1. For estimation of the type of mechanism, the NADH:flavin reductase activity was determined as a function of FAD concentration at several constant levels of NADH and as a function of NADH at several constant levels of FAD. Reciprocal initial velocities were plotted against reciprocal substrate concentrations and fitted with a straight line determined by a linear regression program. Inhibition constants for NAD ϩ inhibition were determined according to Fromm (32).
Other Assays-Phenol monooxygenase activity was determined by measuring phenol consumption colorimetrically (33). The reaction mixture (6 ml) contained 0.1 mM phenol, 0.5 mM NADH, 10 M FAD, and 1.0 nM PheA2 in 50 mM potassium phosphate (pH 7.0). After equilibration at 50°C and the addition of the desired amount of PheA1 (10 -200 nM), samples of 1 ml were taken for 10 min at 1-min intervals. The reaction was stopped by the addition of 12 l of 2% 4-aminoantipyrine followed by 40 l of 2 N ammonium hydroxide and 40 l of 2% potassium ferricyanide, and the final volume was adjusted to 2 ml with water. After 15 min of incubation at room temperature, the absorbance at 510 nm was read and compared with those of phenol standards. 1 unit is defined as the amount of enzyme that catalyzes the conversion of 1 mol of substrate/min under the assay conditions. Oxygen consumption was determined polarographically at 50°C in a closed reaction vessel fitted with a Clark-type oxygen electrode. Reaction mixtures contained 50 mM sodium phosphate (pH 7.0), 0.1 mM phenol, 5-10 M FAD, and 0.5 nM PheA2 in the absence or presence of 100 nM PheA1. NADH was added to a final concentration of 0.1 mM to initiate the reaction. When oxygen consumption ceased, 90 units of catalase were added to determine the degree of uncoupling of hydroxylation.
Thermostability-Studies on the thermostability of PheA1 and PheA2 were performed by incubating the enzymes (46 M PheA1 or 48 M PheA2) in closed vials at different temperatures (50, 60, 70, and 85°C) in 25 mM phosphate (pH 7.0) for 2 h. At timed intervals, aliquots were withdrawn from the incubation mixtures and assayed for residual phenol hydroxylase (PheA1) and NADH:flavin reductase (PheA2) activities.
determined by the method of Bradford (34) with bovine serum albumin as a standard. SDS-PAGE was carried out with 15% slab gels (35) or with 16.5% Tricine gels (36). The gels were scanned and analyzed with a computing densitometer system (Scanjet ADF, Hewlett-Packard Intelligent Scanning Technology). Analytical gel filtration was performed on an Amersham Biosciences Åkta FPLC system equipped with a Superdex 200 HR10/30 column running in 50 mM potassium phosphate (pH 7.0) containing 0.15 M NaCl. Isoelectric focusing was performed with a PhastSystem (Amersham Biosciences) using PhastGel 3-9 isoelectric focusing precast gels. Gels were stained with Coomassie Blue R-250.
Spectral Analysis-Absorption spectra were recorded at 25°C using a Hewlett-Packard 8453 diode array spectrophotometer or an Aminco DW-2000 double-beam spectrophotometer. For anaerobic reduction experiments, enzyme and substrate solutions were made anaerobic by alternated flushing with deoxygenated argon and evacuating. Anaerobic reduction of PheA2 was performed by adding aliquots of oxygen-free NADH or dithionite to the enzyme solution, followed by spectral recording. Photoreduction in the presence of EDTA and 5-deazaflavin as catalyst was performed essentially as described (37). Circular dichroism spectra were measured at 25°C on a Jasco J-715 spectropolarimeter. Fluorescence emission spectra were recorded at 25°C on an SLM-AMINCO SPF500C spectrofluorometer.
Cofactor Analysis-The flavin prosthetic group of PheA2 was isolated by boiling an enzyme sample for 3 min. After removal of the protein precipitate, the nature of the extracted flavin was identified by fluorescence analysis (38) and reverse-phase HPLC (39).
Cysteine Determination-The estimation of sulfhydryl groups of native or unfolded PheA2 was carried out by the procedure of Ellman (40) employing the modifications of Habeeb (41). All thiol determinations were carried out on enzyme preparations that were freshly incubated with 1 mM dithiothreitol, 50 M FAD, and 0.5 mM EDTA for 15 min and subsequently passed through a Bio-Gel P-6DG column equilibrated with 50 mM potassium phosphate (pH 7.0) containing 0.5 mM EDTA. The enzyme was diluted in 100 mM Tris sulfate (pH 8.0) immediately before assaying the time-dependent release of 5-thio-2-nitrobenzoate at 412 nm (final pH 7.9).
Sequence Analysis-N-terminal protein sequence analysis was performed by Edman degradation with an Applied Biosystems Model 473A pulsed-liquid sequencer fitted with an on-line phenylthiohydantoin analytical system following the procedures suggested by the manufacturer. Protein sequence similarity searches were performed with PSI-BLAST (42) using the facilities offered by the National Center for Biotechnology Information (NCBI). Multiple sequence alignments were made with ClustalX (43) using the PAM350 matrix.
Protein Homology Modeling-Model building of PheA2 was performed with MODELLER (44) using the CVFF force field (45). The structure of ferric reductase (FeR) from Archaeoglobus fulgidus with NADP ϩ bound (Protein Data Bank code 1IOS) (46) was used as a template. The model was verified after several rounds of energy minimization. The stereochemical quality of the homology model was verified by PROCHECK (47), and the protein folding was assessed with PROSAII (48), which evaluates the compatibility of each residue with its environment independently. The FMN part of FAD was placed in the model structure at the same position and orientation as the FMN in the template structure. The AMP part of FAD was placed in the groove continuing from the binding pocket of FMN. NADP ϩ was oriented in the same position as found in the crystal structure of FeR.

RESULTS
Overexpression of pheA Genes-Both protein components PheA1 and PheA2 from B. thermoglucosidasius A7 are required for phenol hydroxylase activity (24). However, E. coli cells containing a plasmid with the genes for PheA1 and PheA2 in tandem did not express PheA2, as only a clear protein band for PheA1 (but not for PheA2) could be detected by SDS-PAGE analysis of cell extracts (24). A likely explanation is that, in this construct, PheA2 expression must use its own B. thermoglucosidasius A7 ribosome-binding site, which might not be effective in E. coli. When PheA2 was expressed separately, expression was under the control of the highly efficient phage T7 promoter, and a clear protein band of PheA2 was visible upon SDS-PAGE analysis of cell extracts (24). Therefore, PheA1 and PheA2 were separately purified from E. coli cells containing the appropriate plasmid with the gene for PheA1 or PheA2.
Purification of PheA1-Extracts of E. coli cells containing the plasmid with only the pheA1 gene exhibited phenol hydroxylase activity. Because these cells do not contain PheA2, this indicates that a flavin reductase activity present in the E. coli host substitutes for PheA2 to give substrate conversion. A similar observation was made for 4-hydroxyphenylacetate 3-monooxygenase from E. coli W (30). During purification of PheA1, phenol hydroxylase activity was determined with and without complementation with PheA2. The phenol hydroxylase activity determined in the absence of PheA2 decreased after every purification step (Table I). The specific activity of purified PheA1 determined in the presence of PheA2 was 0.32 units/mg. SDS-PAGE analysis of the purified PheA1 protein revealed the presence of a single polypeptide chain corresponding to a molecular mass of ϳ57 kDa (Fig. 2). This value is in good agreement with the molecular mass predicted from gene sequence analysis (24).
Purified PheA1 did not contain any chromophore as judged by absorption spectral analysis and eluted from an analytical Superdex 200 HR10/30 column in a single symmetrical peak with an apparent molecular mass of 120 Ϯ 5 kDa. This suggests that the PheA1 protein is a dimer composed of identical subunits.
Isoelectric focusing of PheA1 revealed a main protein band with an isoelectric point of 5.2 Ϯ 0.1. This value is considerably lower than the theoretical value of 6.29 calculated from the amino acid sequence. Purified PheA1 was stable for 2 h at 60°C, as no significant decrease in phenol hydroxylase activity was observed in complementation experiments with PheA2. In contrast, at 70°C, inactivation of PheA1 was complete after 10 min of incubation. The purified PheA1 protein was not very stable when stored at Ϫ70°C because it formed aggregates after thawing. Therefore, PheA1 was stored as a protein precipitate in 80% ammonium sulfate at 4°C.
Purification of PheA2-The PheA2 protein expressed in E. coli cells containing the plasmid with the pheA2 gene was purified to apparent homogeneity using three chromatographic steps (Table II). The specific NADH:FAD reductase activity of the purified PheA2 protein at 53°C was 800 units/mg. Analysis by SDS-PAGE revealed the presence of a single band corresponding to a polypeptide chain molecular mass of ϳ18 kDa (Fig. 3). Again, this value is in good agreement with the mo-  (24). The purified PheA2 protein was bright yellow and eluted from an analytical Superdex 200 HR10/30 column in a single symmetrical peak with an apparent molecular mass of 35 Ϯ 3 kDa, indicative of a homodimeric structure. When PheA2 was mixed with PheA1 and analyzed by gel filtration, no interaction between both proteins was observed. The isoelectric point of PheA2 was determined by isoelectric focusing to be 5.2 Ϯ 0.1. This value agrees with the theoretical value (5.36) estimated from the amino acid sequence. PheA2 was stable in the frozen state because no significant loss of activity was observed during 4 months of storage at Ϫ70°C. When purified PheA2 was incubated for 2 h at 65°C, Ͼ65% of the NADH:FAD reductase activity was maintained. Time-dependent analysis showed that the partial loss of activity at 65°C occurred within the first 10 min of incubation. When the enzyme was incubated at 75 or 85°C, the activity decreased faster, but a similar behavior was observed. After 2 h of incubation at 85°C, still 15-20% of the NADH:FAD reductase activity remained.
PheA2 contains a single cysteine residue (Cys 112 ) per polypeptide chain (24). Absorption measurements at 412 nm revealed that this cysteine reacted rapidly with 5,5Ј-dithiobis(2-nitrobenzoate) when the protein was unfolded with 0.5% SDS. In the native enzyme, the side chain of Cys 112 is not accessible for solvent. This is concluded from the lack of reactivity of native PheA2 with Ellman's reagent.
Spectral Properties of PheA2-The optical spectrum of PheA2 was typical of a flavoprotein with maxima at 376 and 455 nm and characteristic shoulders around 355 and 485 nm (Fig. 4A). The A 270 /A 455 ratio for the purified enzyme was 4.1. This value is in good agreement with the low amount of aromatic amino acid residues (24) and suggests that each PheA2 subunit binds one molecule of flavin. Extraction of the cofactor revealed that the flavin was noncovalently bound. Moreover, the HPLC retention time and fluorescence properties of the extracted flavin were identical to those of authentic FAD. From the absorption spectra of PheA2 in the absence and presence of 0.5% SDS (Fig. 4A), a molar absorption coefficient of ⑀ 455 ϭ 11.8 mM Ϫ1 cm Ϫ1 for protein-bound FAD was estimated.
The fluorescence quantum yield of the protein-bound FAD of PheA2 was rather high. Upon excitation at 450 nm, a maximum fluorescence emission was observed around 525 nm (data not shown). Upon unfolding of PheA2 in 0.5% SDS, the fluorescence intensity of the flavin decreased by almost an order of magnitude. This indicates that the flavin fluorescence quantum yield of PheA2 is comparable to that of free FMN (38) and in the same range as that of lipoamide dehydrogenase, one of the most fluorescent flavoproteins (49). Fig. 4B shows the flavin circular dichroism spectrum of PheA2 in the visible region. This spectrum resembles those of flavin reductase from Streptomyces viridifaciens (27), ferredoxin:NADP ϩ oxidoreductase (50), and the enzyme-substrate complex of p-hydroxybenzoate hydroxylase (51) by displaying negative circular dichroism in the 450 nm region and positive optical activity in the near-UV band.
When PheA2 was (photo)chemically reduced under anaerobic conditions, the flavin hydroquinone was formed, and no thermodynamic stabilization of semiquinone species was observed. Upon titration with NAD ϩ , the two-electron reduced enzyme formed a complex with NAD ϩ , as evidenced by the formation of a long wavelength absorption band. Anaerobic reduction of PheA2 with NADH also led to the reduced enzyme-NAD ϩ complex (Fig. 5). At pH 8.0, the enzyme could not be fully reduced with NADH, in contrast to reduction at pH 6.0 (Fig. 5).
Catalytic Properties of PheA2-In addition to NADH:FAD reductase activity, PheA2 exhibited also NADH:cytochrome c reductase and NADH:dichlorophenolindophenol and NADH: AcPyAD ϩ transhydrogenase activities (Table III). However, PheA2 displayed no ferric reductase activity. PheA2 was also active with oxygen as an electron acceptor (Table III). In contrast to the aromatic substrate-stimulated NADH oxidase activity observed for the reductase component of 4-hydroxyphenylacetate 3-hydroxylase from Acinetobacter baumannii (52), phenol did not influence the rate of NADH oxidation or oxygen consumption. Moreover, the oxygen consumed by PheA2 in the presence and absence of phenol was halved in the presence of catalase, confirming that, in both cases, hydrogen peroxide was the only reduced form of molecular oxygen that was generated. The hydrogen peroxide formed results from the reaction of the reduced protein-bound flavin with oxygen, which presumably involves the intermediate formation of a flavin semiquinone-superoxide radical pair and a highly unstable flavin C4a-peroxide (7).
The pH and temperature optima of the NADH:FAD reductase activity of PheA2 were pH 6.7 and 55°C (Figs. 6 and 7). Determination of kinetic parameters revealed that NADH is a much better coenzyme for PheA2 than NADPH (Table IV). The FAD reductase activity exhibited a linear dependence on NADPH concentration over a range 10 -500 M, and no indication of saturation was observed. Thus, K m and k cat values for NADPH could not be determined under these conditions. However, from the slope of the kinetic plot, a second-order rate constant of kЈ cat /KЈ m ϭ 0.006 s Ϫ1 could be estimated. This value is at least two orders of magnitude lower than the corresponding value for NADH (Table IV).
When the NADH:FAD reductase activity of PheA2 was studied under anaerobic conditions, all free FAD was rapidly reduced, confirming that the free FAD acts as a true substrate. In addition to FAD, PheA2 was also active with FMN and riboflavin. The turnover rate of PheA2 with free flavins was strongly dependent on temperature. At 25°C, the activity with FMN and riboflavin was much higher than with FAD (Table III). However, when the temperature was raised to 53°C, the turnover rates with the different flavins became nearly identical (Table IV). On the other hand, the Michaelis constants for NADH, FAD, riboflavin, and FMN did not vary with temperature (Tables III and IV).
Initial velocity measurements in dependence of both FAD and NADH were performed to provide insight into the kinetic mechanism of PheA2. When the NADH concentration was varied at a fixed FAD concentration, a series of parallel lines was observed in a double-reciprocal plot (Fig. 8A). A similar pattern was observed when the FAD concentration was varied at a fixed NADH concentration (Fig. 8 B). This suggests that PheA2 acts according to a Ping Pong Bi Bi reaction mechanism in which NADH reduces the FAD cofactor, which in turn transfers electrons to the FAD substrate. A similar kinetic behavior was observed when the NADH:AcPyAD ϩ transhydrogenase activity of PheA2 was determined as a function of AcPyAD ϩ concentration at several constant levels of NADH and as a function of NADH at several constant levels of AcPyAD ϩ (data not shown). The NADH:FAD reductase activity of PheA2 was inhibited by NAD ϩ . In line with the above-described mechanism, the inhi-bition was noncompetitive to NADH and competitive to FAD (data not shown).
Phenol Hydroxylase Activity-Purified PheA1 showed almost no phenol hydroxylase activity when assayed at 50°C and pH 7.0 in the absence of PheA2. The PheA1-mediated conversion of phenol to catechol was strongly stimulated in the presence of catalytic amounts of PheA2. At relatively low concentrations of PheA1, phenol hydroxylase activity was not linear with time. The addition of catalase to the reaction mixture to avoid the  accumulation of hydrogen peroxide had no effect. At a constant concentration of PheA2 and increasing concentrations of PheA1 in the presence of phenol, the oxygen consumption increased, and the relative amount of hydrogen peroxide formed decreased. When the dependence of the hydroxylation reaction on the concentration of free FAD at constant concentrations of PheA1 and PheA2 (200:1 molar ratio) was studied, the oxygen consumption became faster with increasing concentrations of free FAD, but the amount of uncoupling increased (Table V). In contrast to the NADH:flavin reductase activity of PheA2, the hydroxylase activity of PheA was strictly FAD-dependent, and neither FMN nor riboflavin could replace FAD in this reaction.
Under optimal assay conditions with PheA1 in large excess over PheA2, maximum phenol hydroxylase activity was reached at 55°C and pH 7.0. HPLC analysis revealed that PheA from B. thermoglucosidasius A7 also catalyzed the conversion of 4-methylphenol, 4-chlorophenol, and 4-fluorophenol to the corresponding catechols. No monooxygenase activity was observed with 4-hydroxybenzoate, 4-hydroxyacetophenone, and 4-hydroxyphenylacetate.
To test whether a nonspecific FAD reductase might substitute for PheA2 to give substrate conversion, the phenol hydroxylase activity of purified PheA1 was measured in the presence of E. coli NADH oxidoreductase, which is physiologically asso-ciated with the so-called hybrid cluster protein (53). At 50°C, this iron-sulfur and FAD-containing reductase initiated the formation of catechol. However, the E. coli reductase was not    very stable under these assay conditions, and several aliquots of enzyme were needed to consume all of the phenol. Sequence Homology-N-terminal amino acid sequence analysis of the purified recombinant PheA1 and PheA2 proteins revealed sequences Met-Lys-Asp-Met-Met and Met-Asp-Asp-Arg-Leu, respectively. These sequences are identical to those deduced from the nucleotide sequences of the pheA1 and pheA2 genes (24). Multiple sequence alignment revealed that PheA1 shares the most sequence identity (up to 54%) with the oxygenase component of (i) phenol hydroxylase from Bacillus thermoleovorans A2 (23), (ii) chlorophenol 4-monooxygenase from Ralstonia picketii (54) and Burkholderia cepacia (55), and (iii) several (putative) bacterial 4-hydroxyphenylacetate 3-hydroxylases (30,56,57).
PheA2 resembles a larger number of proteins than PheA1. Except for the reductase components of the above-mentioned two-component enzymes, PheA2 shares considerable sequence identity (up to 57%) with the reductase component of (i) styrene monooxygenase (58); (ii) dibenzothiophene monooxygenase (59); (iii) actinorhodin-polyketide dimerase (25); (iv) nitrilotriacetate monooxygenase (60) (62), for which the three-dimensional structure was recently solved (46). FeR reduces FAD and FMN, but not riboflavin, whereas PheA2 shows no ferric reductase activity. Another remarkable difference between PheA2 and FeR is that the latter enzyme was purified as an apoprotein (62). Furthermore, in the FeR crystal structure, only one subunit contains FMN, and NADP ϩ was found to bind only to the subunit that harbors FMN (46).
PheA2 shares 46% sequence homology (24% sequence identity) with FeR ( Fig. 9), which prompted us to build a model of the three-dimensional structure of PheA2 (Fig. 10). As found for FeR, each PheA2 subunit consists of a single domain that is organized around a six-stranded antiparallel ␤-barrel that is homologous to the FMN-binding protein from Desulfovibrio vulgaris (63). The FMN part of FAD was fit in the model structure at the same position and orientation as the FMN in FeR. The AMP part of FAD was docked into the PheA2 model in a groove elongating the FMN cofactor-binding site. However, because of steric constraints and the poor alignment with a flexible loop in the FeR structure comprising residues 85-100, the location of this part of the FAD molecule remained elusive. DISCUSSION PheA is a novel type of phenol hydroxylase that is induced in the thermophilic bacterium B. thermoglucosidasius A7 when the strain is grown on phenol as a source of carbon and energy. The phenol hydroxylase system consists of two homodimeric proteins with subunit molecular masses of 57 and 18 kDa, respectively. The large PheA1 component contains no cofactors and has no enzyme activity on its own. The small PheA2 component is an NADH-dependent FAD-containing reductase that supplies PheA1 with reduced FAD to catalyze the orthohydroxylation of simple phenols to the corresponding catechols.
During the past few years, several two-protein component aromatic hydroxylases relying on reduced flavin as a substrate  have been identified. These enzymes show a remarkable variety in protein size and flavin specificity (52). So far, no structural data of these enzymes are available, and their mechanisms of reduced flavin transfer and substrate hydroxylation remain to be elucidated. For 4-hydroxyphenylacetate 3-hydroxylase from E. coli (64) and A. baumannii (52), it was reported that the oxygenase component is able to use chemically reduced flavin for 4-hydroxyphenylacetate monooxygenation and that a physical interaction between the oxygenase and reductase protein components is not required for substrate hydroxylation (30,65). Here we have shown that the purified dimeric PheA1 and PheA2 proteins do not strongly interact, but that for the in vitro conversion of phenol to catechol, a large excess of PheA1 over PheA2 is needed to prevent the unproductive formation of hydrogen peroxide.
Another important finding described in this study is the fact that PheA2 harbors a tightly bound FAD. The interaction between PheA2 and FAD is much stronger than found for other short-chain flavin reductases (30,64,66). From this and the Ping Pong Bi Bi kinetic mechanism observed, we propose that, in PheA2, the reducing equivalents are transferred from NADH via protein-bound FAD to exogenous free FAD. This proposal is supported by the inhibition of NAD ϩ that is competitive with FAD. After reduction by PheA2, the FADH 2 product is transferred to the PheA1 oxygenase component, in which the hydroxylation of phenol to catechol takes place (Scheme 1).
Our studies with PheA2 suggest a direct role for the proteinbound FAD in the mediation of electron transfer from NAD(P)H to exogenous free flavin. Such a role has also been proposed for the flavoprotein component of sulfite reductase (67) and for flavin reductase I, the major NAD(P)H:FMN oxidoreductase from Vibrio fisheri (68).
The kinetic mechanism and flavin-binding properties of PheA2 are clearly different from those of the related ActVB NADH:flavin reductase participating in the last step of actinorhodin synthesis in Streptomyces coelicolor (66). This homodimeric FMN-binding enzyme acts according to an ordered sequential mechanism and is not classified as a flavoprotein. The mechanism depicted in Scheme 1 also differs from the ternary complex mechanism proposed for Fre, the prototype of the class I flavin reductases (69 -71). In this monomeric enzyme, which does not contain any prosthetic group, flavins are recognized mainly through hydrophobic interactions with the isoalloxazine ring, providing an explanation of why Fre uses riboflavin as a substrate rather than as a cofactor (72,73). The mechanism depicted in Scheme 1 is more consistent with the mechanism proposed for the homodimeric FMN-containing flavin reductase P from Vibrio harveyi, which is involved in bioluminescence by providing reduced FMN to luciferase (74,75). With the latter enzyme, a shift in the kinetic mechanism in the luciferase-coupled assay was taken as evidence that the reduced FMN cofactor is transferred to luciferase via proteinprotein complex formation (76). Subsequent studies with the bioluminescence enzymes from V. fisherii (77) and V. harveyi (78,79) have indicated that the mechanism of reduced flavin transfer may depend on the constituent protein partners in the reductase-luciferase couple and that, under in vitro reconstitution conditions, direct flavin product transfer is feasible as well.
PheA2 belongs to a newly recognized family of short-chain flavin reductases (30). These enzymes are considerably smaller in size than members of the Fre family (71, 72) and the flavin reductases from luminous bacteria (68,74,75). Our studies show that PheA2 is structurally related to FeR, a ferric reductase from the hyperthermophilic Archaea A. fulgidus that uses its single domain to provide both the flavin-and NAD(P)Hbinding sites (46). However, PheA2 displays no ferric reductase activity, binds FAD instead of FMN, and is specific for NADH.
For FeR, it was proposed that the substrate binds in a pocket near the flavin isoalloxazine ring and that Thr 31 , Leu 35 , Cys 45 , and His 126 might serve as protein ligands for the metal ion. Except for the conserved histidine (His 123 ), these residues are replaced in PheA2, disfavoring iron ligandation. In FeR, the FMN ribityl phosphate interacts with Asp 51 , Thr 52 , Arg 82 , Ser 84 , and Lys 89 . These residues are not conserved or lacking in the PheA2 sequence. This suggests that the specificity for FAD binding is related to the shorter loop (region 84 -92, PheA2 numbering) (Fig. 9).
For FeR, it was found that the 2Ј-phosphate of NADP ϩ does not make strong interactions with the polypeptide chain. This might explain why FeR shows no preference for NADH or NADPH. From the three-dimensional model of PheA2, no indication was obtained as to why this enzyme is very specific for NADH. This suggests that FeR and PheA2 must differ in their mode of coenzyme binding.
Finally, we have shown here that the monooxygenase activity of PheA1 is strictly FAD-dependent. This is indicative of a specific interaction of PheA1 with the AMP moiety of FAD. Therefore and especially because both PheA1 and PheA2 have no signature sequences for FAD binding, it is of great interest to obtain more insight into the structural and mechanistic features of this novel type of two-protein component phenol hydroxylase in which FAD acts both as a substrate and as a prosthetic group.