Redox Centers of 4-Hydroxybenzoyl-CoA Reductase, a Member of the Xanthine Oxidase Family of Molybdenum-containing Enzymes*

4-Hydroxybenzoyl-CoA reductase (4-HBCR) is a key enzyme in the anaerobic metabolism of phenolic compounds. It catalyzes the reductive removal of the hydroxyl group from the aromatic ring yielding benzoyl-CoA and water. The subunit architecture, amino acid sequence, and the cofactor/metal content indicate that it belongs to the xanthine oxidase (XO) family of molybdenum cofactor-containing enzymes. 4-HBCR is an unusual XO family member as it catalyzes the irreversible reduction of a CoA-thioester substrate. A radical mechanism has been proposed for the enzymatic removal of phenolic hydroxyl groups. In this work we studied the spectroscopic and electrochemical properties of 4-HBCR by EPR and Mössbauer spectroscopy and identified the pterin cofactor as molybdopterin mononucleotide. In addition to two different [2Fe-2S] clusters, one FAD and one molybdenum species per monomer, we also identified a [4Fe-4S] cluster/monomer, which is unique among members of the XO family. The reduced [4Fe-4S] cluster interacted magnetically with the Mo(V) species, suggesting that the centers are in close proximity, (<15 Å apart). Additionally, reduction of the [4Fe-4S] cluster resulted in a loss of the EPR signals of the [2Fe-2S] clusters probably because of magnetic interactions between the Fe-S clusters as evidenced in power saturation studies. The Mo(V) EPR signals of 4-HBCR were typical for XO family members. Under steady-state conditions of substrate reduction, in the presence of excess dithionite, the [4Fe-4S] clusters were in the fully oxidized state while the [2Fe-2S] clusters remained reduced. The redox potentials of the redox cofactors were determined to be: [2Fe-2S]+1/+2 I, −205 mV; [2Fe-2S]+1/+2 II, −255 mV; FAD/FADH⋅/FADH, −250 mV/−470 mV; [4Fe-4S]+1/+2, −465 mV and Mo(VI)/(V)/(VI), −380 mV/−500 mV. A catalytic cycle is proposed that takes into account the common properties of molybdenum cofactor enzymes and the special one-electron chemistry of dehydroxylation of phenolic compounds.

Aromatic compounds comprise a large group of natural products many of which contain hydroxyl or methoxyl functionalities. Methoxyl groups are usually converted to hydroxyl groups and C1 units before further metabolism takes place. In recent years a growing number of anaerobic bacteria have been identified that use low molecular mass aromatic compounds as their sole sources of cell carbon and energy. In the presence of molecular oxygen, hydroxylation is catalyzed by mono-or dioxygenases, and oxidative aromatic ring cleavage is catalyzed by dioxygenases. However, in the absence of the highly reactive co-substrate oxygen, alternative strategies have developed in anaerobic bacteria (1,2). Many compounds are converted to the central intermediate benzoyl-CoA, which becomes dearomatized by benzoyl-CoA reductase (BCR) 1 (1). This enzyme couples the difficult ring reduction to a stoichiometric ATP hydrolysis (3). It is assumed that BCR also catalyzes the direct ring reduction of ortho-or meta-substituted hydroxy-, amino-or methyl-derivatives of benzoyl-CoA. 2 Examples are 3-hydroxyor 3-methyl-benzoyl-CoA (5). However, the aromatic ring of para-substituted derivatives, for example 4-hydroxybenzoyl-CoA (4-HBCoA) cannot be reduced by BCR or similar enzymes for mechanistic reasons. 2 Therefore, an additional group of enzymes should exist that reductively removes hydroxy or amino groups from the ring prior to dearomatization. The only enzyme of this group of which the biochemistry has been studied so far is 4-HBCR from the denitrifying bacterium Thauera aromatica (6). This enzyme plays a key role in the anaerobic metabolism of phenolic compounds (1). 4-HBCR catalyzes the reductive dehydroxylation of 4-HB-CoA to benzoyl-CoA as two electrons are transferred to the substrate (6) (Fig. 1). The natural electron donor is a reduced 2[4Fe-4S] ferredoxin (7), which also serves as the in vivo electron donor for the next enzyme in aromatic metabolism, benzoyl-CoA reductase (8). 4-HBCR has a molecular mass of 260 kDa and consists of three subunits of 75 (␣), 35 (␤), and 17 kDa (␥), suggesting an (␣␤␥) 2 composition (6). Purified 4-HBCR contained 15 mol of iron, 12.5 acid-labile sulfur, 1.3 mol of FAD and 1.9 mol of molybdenum per mol of dimeric enzyme (7). The enzyme was inactivated by cyanide giving initial evidence that it may belong to the xanthine oxidase (XO) family of molybdenum cofactor-containing enzymes (6).
The genes coding for the three subunits of 4-HBCR have been cloned and sequenced in T. aromatica (7) and in the phototrophic bacterium Rhodopseudomonas palustris (9). They are highly conserved between these two organisms and also show high similarity to several enzymes of the XO family. This was highest with carbon monoxide dehydrogenases (CODH) from Oligotropha carboxydovorans and Pseudomonas thermocarboxydovorans, with nicotine dehydrogenase from Arthrobacter nicotinovorans, and with eukaryotic xanthine dehydrogenases (XDH) (7).
The XO family of molybdenum-cofactor-containing enzymes comprises a large group of enzymes that usually catalyze hydroxy-or oxo-transfer to the substrate. They usually contain a molybdopterin-cofactor and [2Fe-2S] clusters; many also harbor an FAD cofactor on a special subunit (10 -12). Although 4-HBCR has high similarity to the above members of the XO family (30 -41% overall identity), some special features are unique: (i) The ␤-subunit harbors an additional domain consisting of ϳ40 amino acids, with five cysteine residues, suggesting the presence of an additional [4Fe-4S] cluster (7); this assumption agrees with the higher iron content (7.5 mol/mol monomer) and is inconsistent with the presence of only the usual two [2Fe-2S] clusters/monomer. (ii) 4-HBCR is the only member of this family whose function is to catalyze the reverse reaction, the reduction of the substrate; note that the 4-HBCR reaction is considered to be irreversible for mechanistic reasons (see below) (13). (iii) The CoA-ester substrate of 4-HBCR is exceptionally large suggesting the presence of an additional nucleotide binding site for the CoA-moiety.
Although the reduction of 4-hydroxybenzoate to benzoate plus water by molecular hydrogen is highly exergonic (59 kJ/ mol, calculated from the free energy of formation for the individual compounds from the elements (14)), the reductive removal of an hydroxy group at the aromatic ring is difficult to achieve mechanistically (13). In particular, 4-hydroxybenzoate undergoes activation to 4-HBCoA by specific ATP-dependent CoA ligases in R. palustris and T. aromatica (15,16), which should facilitate the subsequent reductive elimination of water. Buckel and Keese have proposed that the CoA-thioester moiety plays an essential role in enzymatic dehydroxylation by stabilizing a ketyl radical intermediate (13). In chemical systems the removal of phenolic hydroxy groups usually proceeds in a Birch-like mechanism involving alternate one-electron and one-proton transfers to the aromatic ring (see also Fig. 11A). Kinetic and electrochemical data on the enzyme indicate two successive one-electron transfer steps to the aromatic ring, supporting a Birch-like mechanism (17). In addition, it was predicted that a one-electron redox center with a midpoint potential of Ϫ357 mV catalyzes the rate-limiting step in the reductase reaction (17). However, a radical mechanism has not yet been proposed for any other molybdenum-cofactor-containing enzyme.
In this work, we studied the redox centers of 4-HBCR by EPR and Mössbauer spectroscopy. We provide spectroscopic evidence that 4-HBCR contains the expected [2Fe-2S] clusters, FAD and molybdenum, plus an additional [4Fe-4S] cluster that is unique among enzymes of the XO family. We suggest that the [4Fe-4S] cluster is involved in electron transfer from ferredoxin to the substrate and provide evidence that it is located close to the molybdenum-cofactor, which is molybdopterin mononucleotide.
The Mo(V) EPR signals are typical of the XO family enzymes. Based on these results, a catalytic mechanism is proposed.

EXPERIMENTAL PROCEDURES
Growth of Bacterial Cells-T. aromatica (DSM 6984) was isolated in our Freiburg laboratory and has been deposited in the Deutsche Sammlung von Mikroorganismen (Braunschweig, Germany) (18). It was grown anoxically at 28°C in a mineral salt medium in a 200-liter fermenter with 4-OH-benzoate and nitrate in a molar ratio of 1:3.5 as sole sources of energy and cell carbon. Continuous feeding of the substrates, cell harvesting, storage, and preparation of cell extracts were carried out as described previously (19). 57 Fe-enrichment of 4-HBCR was achieved by adding 57 Fe as sole source of iron to the medium. For this purpose 175 mg of metallic 57 Fe (95.7% enriched, Advanced Materials and Technology Consulting, New York, NY) was dissolved in 2.2 ml of 12 M HCl overnight at 80°C. After the metal was completely dissolved, the solution was added drop-wise to 100 ml of 0.6 M nitrilotriacetic acid, pH 5.2, under continuous stirring. During this procedure the pH was held constant between 3 and 5 by adding 2 M NaOH. With this amount of 57 Fe T. aromatica was grown in a 100-liter batch culture with continuous feeding of 4-OH-benzoate and nitrate to A 578 ϳ 2.6 yielding 250 g of cells (wet mass). Purified 4-HBCR was estimated to be Ͼ90% enriched in 57 Fe (20).
Protein Purification, Enzyme Activity Assay, Purity Control, and Sample Storage-Purification of 4-HBCR from extracts of T. aromatica was performed under strictly anaerobic conditions in a glove box under a N 2 /H 2 atmosphere (95:5, by vol.) as described earlier (6). The procedure included three chromatographic steps using anion exchange chromatography on DEAE-Sepharose (Amersham Biosciences, Inc.), chromatography on Hi Load Q-Sepharose (Amersham Biosciences, Inc.), and affinity chromatography on Cibacron Blue-Agarose (Sigma). Concentration of the protein samples was achieved by centrifugation (8000 ϫ g) in Microsep Microconcentrators (exclusion limit 50 kDa). The concentration of the enzyme was ϳ150 mg ml Ϫ1 for Mössbauer studies or 15-30 mg ml Ϫ1 for EPR spectroscopy. The purity of these enzyme preparations was Ͼ90% as estimated by Coomassie staining of SDS gels. Enzyme activity was determined in a continuous spectrophotometric assay recording the 4-HBCoA-dependent oxidation of reduced methyl viologen at 730 nm at 37°C (⑀ 730 ϭ 2.4 mM Ϫ1 cm Ϫ1 ), (6). 25-35 mg of purified 4-HBCR was obtained from 200 g of cells (wet mass) with specific activities of 8 -12 mol 4-HBCoA reduced min Ϫ1 mg Ϫ1 . Concentrated protein samples were stored anaerobically in tubes sealed with gas-tight stoppers at Ϫ80°C for several months without loss of activity.
Sample Preparation for EPR and Mössbauer Spectroscopy-All 4-HBCR samples for EPR and Mössbauer spectroscopy were prepared in an anaerobic glove box under a 100% nitrogen atmosphere (Ͻ1.0 ppm O 2 ). Prior to sample preparation, excess dithionite and corresponding oxidation products were removed by passing the concentrated enzyme sample (0.1-1 mM; 0.5-1 ml) over a Biogel P-6 (Bio-Rad) desalting column (volume: 5 ml, diameter: 0.7 cm) equilibrated with either 100 mM Mops/KOH, pH 7.5, or 100 mM Hepes/HCl, pH 8.3, both containing 10 mM MgCl 2 . Unless stated otherwise, this dithionite free enzyme was the starting material for all sample preparations.
Reduction and Oxidation of the Enzyme-Reduction of 4-HBCR was performed by adding sodium dithionite from a freshly prepared stock solution (100 mM in 100 mM Mops/KOH, pH 7.5) giving a final 10-fold excess of this reductant compared with the enzyme. The enzyme was quickly transferred into EPR tubes or Mössbauer sample holders and then frozen either inside the glove box on dry ice (Mössbauer-samples) or outside the glove box in a gas-tight sealed EPR tube in liquid nitrogen. Anaerobic oxidation of 4-HBCR was achieved by titrating the enzyme with an anerobically prepared filtered thionine stock solution (ϳ10 mM in the buffer described above). 4-HBCR was considered fully oxidized when the color of the enzyme solution remained blue for at least 5 min.
Redox Titration of 4-HBCR-Dye-mediated redox titration of 4-HBCR was performed in an anaerobic glove box under a nitrogen atmosphere (Ͻ1 ppm O 2 ). The enzyme/mediator mixture (2.5 ml) was in 100 mM Mops/KOH, pH 7. The concentration of 4-HBCR was 80 -100 M. The mediator consisted of methyl and benzyl viologens, neutral red, safranin O, phenosafranin, anthraquinone-2-sulfonate, 2-hydroxy-1,4naphthoquinone, indigo disulfonate, resorufin, methylene blue, phenazine methosulfate and NЈNЈNЈNЈ-tetramethyl-p-phenylenediamine at a final concentration of 30 M each. The redox potential was adjusted anaerobically with freshly prepared 10 -100 mM sodium dithionite and potassium ferricyanide solutions in the same buffer as the 4-HBCR. Potentials are reported with reference to standard hydrogen electrode and were obtained by using a potential for the saturated calomel elec- trode at 22°C of ϩ243 mV versus standard hydrogen electrode. Mediator/enzyme mixtures with a stable potential could be obtained from Ϫ573 mV to Ϫ100 mV. Stabilization (drift Ͻ1 mV/min) of the potentials usually required 1-5 min under continuous stirring. Samples with defined redox potentials were immediately frozen in anaerobic EPR tubes and stored in liquid nitrogen. Redox titrations were normally performed in the oxidative direction, but as a control for reversibility, some samples were prepared by re-reduction with dithionite.
EPR Spectroscopy-X-band EPR spectra were recorded on an updated Bruker 200D-SRC spectrometer. Low temperature measurements were made using an Oxford Instruments ESR 900 cryostat modified to take sample tubes of up to 4-mm internal diameter. Recording conditions are described in the legends to the individual figures. Spin concentrations of ground-state transition EPR signals were determined by comparison with a 1.00-mM copper sulfate sample in 11 mM sodium EDTA.
Mössbauer Spectroscopy-57 Fe Mössbauer spectra were recorded with a conventional constant acceleration spectrometer using a 57 Co source in a Rh matrix (1 GBq using the bath cryostat; 1.3 GBq using the magnet cryostat). Measurements at 4.2 and 77 K were performed with a bath cryostat (Oxford Instruments) and a permanent magnet mounted outside the cryostat producing a field of 20 mT Ќ␥. High-field measurements were performed with a cryostat equipped with a superconducting magnet (Oxford Instruments). The spectra were analyzed assuming Lorentzian line shape. Isomer shifts are quoted relative to ␣-Fe at room temperature.
Molybdopterin Cofactor Analysis-Molybdenum cofactor analysis was carried out based on the method of Johnson et al. (21). This procedure converts the molybdenum cofactor to its oxidized fluorescent degradation product (form A). The starting solution was 80 g of purified 4-HBCR in 4 ml of 100 mM Tris/HCl, pH 7.2. The sample was incubated with the oxidizing solution (1% I 2 , 2% KI in 1 M HCl) at 25°C overnight under shaking. The sample was subsequently treated with 560 l of 1% ascorbic acid. To remove the phosphate group from phospho-form A, the supernatant was mixed with 2 ml of 1 M Tris, pH 7.2, 200 l of 1 M NaOH, 130 l of 1 M MgCl 2 , and 8 units of alkaline phosphatase and left at room temperature for 2 h. The sample was then clarified by centrifugation and applied to an QAE-Sephadex (Amersham Biosciences, Inc. FPLC column (volume: 35 ml) that had been equilibrated with 1 M ammonium acetate. Once the extract was loaded onto the column, it was washed with four column volumes of water. To elute the dephospho-form A, 10 mM acetic acid was added to the column, and fractions collected. To elute any form A-nucleotide variant, 50 mM HCl was applied to the column. The pooled fractions were neutralized with NaOH and then concentrated by freeze-drying overnight. The samples were taken up in 2 ml of water and separated by reverse phase high pressure liquid chromatography with 3% methanol/50 mM ammonium acetate as the mobile phase. Peaks were collected and a 500-l aliquot was added to 100 ml of 20 mM Tris/HCl, pH 7.6, MgCl 2 to a final concentration of 2 mM and 1 unit of pyrophosphatase. The sample was left at room temperature for 1 h, and the fluorescence before and after pyrophosphatase treatment was analyzed as described (21). As standards, form A was produced from milk XO, form A-GMP was formed from Me 2 SO, reductase from Rhodobacter capsulatus, and form A-CMP was obtained from CODH of O. carboxydovorans. The standards were kindly provided by G. Schwarz, University of Braunschweig.
Other Methods-4-HBCoA was synthesized via the N-hydroxysuccinimidyl ester of the carboxylic acid according to Gross and Zenk (22). Protein concentrations were determined by the Bradford method using bovine serum albumin as standard (23). SDS-PAGE was performed as described by Laemmli (24). Protein was visualized by Coomassie Blue staining (25).

Nature of the Iron-Sulfur Clusters
The metal analysis of 4-HBCR revealed the presence of 15-16 mol of iron/mol of dimeric enzyme (␣ 2 ␤ 2 ␥ 2 ) indicating that the iron content was double of that of typical members of the XO family that have 8 mol of iron/mol of dimeric enzyme. Amino acid sequence analysis suggested the presence of an additional [4Fe-4S] cluster in the ␤-subunit of 4-HBCR in addition to the two [2Fe-2S] clusters that are usually found in the small ␥-subunit of XO-type enzymes. To elucidate the nature of the iron-sulfur clusters Mössbauer and EPR spectroscopic studies were performed.

Mössbauer Spectroscopy of Thionine-oxidized 4-HBCR-For
Mössbauer spectroscopy studies, BCR was purified from extracts of T. aromatica grown anaerobically with 57 Fe as the sole source of iron. High 57 Fe enrichment was achieved (Ͼ90%) as estimated from the resonance absorption effect of each subspectrum of the oxidized sample. In   3 shows the Mössbauer spectra of dithionite-reduced 4-HBCR taken at 120 K (A) and at 4.2 K in external fields of 20 mT Ќ␥ (B) and 7 T ʈ␥ (C). The central region of spectrum A (measured at 120 K) shows a strongly asymmetric doublet, the right part of which (in the interval 0.5-1.0 mm s Ϫ1 ) is not resolved as it is in Fig [2Fe-2S] 1ϩ clusters, in the simulations of the spectra of dithionite-reduced 4-HBCR (Fig. 3). The subspectra of [4Fe-4S] 2ϩ and of [2Fe-2S] 2ϩ at 4.2 K correspond to those shown in Fig. 2; however, at 120 K the parameters had to be slightly readjusted, i.e. ␦ ϭ 0.44 mm s Ϫ1 , ⌬E Q ϭ 0.88 mm s Ϫ1 , ⌫ ϭ 0.33 mm s Ϫ1 for [4Fe-4S] 2ϩ , and ␦ ϭ 0.29 mm s Ϫ1 , ⌬E Q ϭ 0.83 mm s Ϫ1 , ⌫ ϭ 0.36 mm s Ϫ1 for [2Fe-2S] 2ϩ . At 120 K (Fig. 3A) subspectrum 2 also represents the Fe 3ϩ sites of [2Fe-2S] 1ϩ clusters, while subspectrum 4 represents the Fe 2ϩ sites of these clusters with the parameters ␦ ϭ 0.60 mm s Ϫ1 , ⌬E Q ϭ 3.10 mm s Ϫ1 , ⌫ ϭ 0.38 mm s Ϫ1 . The remaining [4Fe-4S] 1ϩ clusters at 120 K are represented in Fig. 3A (subspectra 5, 6) by the parameters ␦ ϭ 0.54 mm s Ϫ1 ; ⌬E Q ϭ 1.04 mm s Ϫ1 , ⌫ ϭ 0.40 mm s Ϫ1 , which are characteristic of these clusters at elevated temperature (27). At 4.2 K (Fig. 3, B and C) the [4Fe-4S] 1ϩ clusters need to be simulated with the spin-Hamiltonian parameters of two different components, representing two antiferromagnetically coupled 2Fe subclusters with spins 9/2 and 4 respectively, yielding an overall cluster spin of 1/2 (28). Correspondingly at 4.2 K the [2Fe-2S] ϩ clusters were simulated with the spin-Hamiltonian parameters of the antiferromagnetically coupled Fe 3ϩ (S ϭ 5/2) and Fe 2ϩ (S ϭ 2) subsites (29), represented by subspectra 3 and 4. The relevant parameters of all subspectra, which were used to simulate the measured spectra of dithionite-reduced 4-HBCR at 4.2 K (Fig. 3, B and C), are summarized in Table l EPR Spectroscopy of the Fe-S Clusters at Different Redox Potentials-In the thionine-oxidized state of 4-HBCR (ϳ0 mV), no EPR signal indicative of an Fe-S cluster was observed. To follow the successive reduction of the individual redox active centers, samples of 4-HBCR were taken at potentials between Ϫ100 mV and Ϫ570 mV at ϳ30 mV steps. In these studies, we attempted to assign EPR signals to individual clusters and to elucidate their redox properties.
At potentials below Ϫ190 mV EPR signals appeared between 330 and 350 mT, which were almost fully developed at Ϫ350 mV. EPR signals of at least two different paramagnetic species were observed with characteristic features at g z1 ϭ 2.04 and g z2 ϭ 2.02 (using the convention g x Ͻ g y Ͻ g z ). Representative EPR spectra recorded at Ϫ190 mV and Ϫ284 mV are presented in Fig. 4, A and C. The Ϫ190 mV spectrum was greatly obscured by the radical EPR signal of the redox dyes. After subtraction of this signal, a rhombic S ϭ 1 ⁄2 EPR signal with g values at 2.04, 1.995, and 1.964 (g av ϭ 1.999) was obtained which we assign to an Fe-S cluster (Fig. 4B). This spectrum also included additional minor sharp features between g ϭ 1.96 and 1.98, which can probably be assigned to a resting Mo(V) signal (for a detailed characterization of the Mo(V) EPR signals, see below). The difference spectrum between the EPR spectra recorded at Ϫ284 mV and Ϫ190 mV (after normalized subtraction of the redox dye radical signal) is presented in Fig. 4D. It displays a slightly rhombic EPR signal, which we assign to a second Fe-S cluster with typical g values at 2.02, 1.977, and 1.961 (g av ϭ 1.986). The redox potential-dependent rise of both signals is shown in Fig. 5A. The g ϭ 2.04 EPR signal-fitted to a Nernst curve with EЈ 0 ϭ Ϫ205 mV and the g ϭ 2.02 signal to one with EЈ 0 ϭ Ϫ255 mV. To characterize these signals in more detail, a redox potential-dependent microwave power saturation study was performed. The signals differed considerably in their relaxation properties at 40 K (Fig. 5, B and C). The saturation of the g ϭ 2.02 EPR signal occurred only at microwave powers greater than 10 mW and was independent of the poised potential (Fig. 5C). In contrast, the g ϭ 2.04 EPR signal exhibited a strong dependence on the redox potential: at Ϫ190 mV (when the g ϭ 2.02 feature was only weakly present) power saturation was already observable at 0.2 mW, whereas at Ϫ284 mV (when the g ϭ 2.02 signal was strongly developed) the signal required microwave powers Ͼ2 mW for saturation. Obviously, the reduction of the "g ϭ 2.02-cluster" strongly influences the relaxation properties of the "g ϭ 2.04-cluster" indicating that these two clusters interact magnetically. Since [4Fe-4S] ϩ1 clusters usually relax much faster than this, we refer to the signals as [2Fe-2S] ϩ1 I (g av ϭ 1.999) and [2Fe-2S] ϩ1 II (g av ϭ 1.986). The maximum spin concentration was 1.5-1.8 spins/monomer of 4-HBCR. It is interesting to note that [2Fe-2S] ϩ1 I has an unusually slow relaxation rate (Fig. 5B) and was observable at temperatures up to 120 K.
At potentials below Ϫ400 mV the EPR spectra of 4-HBCR became highly complex, with representative spectra shown in Fig. 6A; new features at g ϭ 2.06 and below g ϭ 1.95 developed. Both features showed substantially faster relaxation than [2Fe-2S] ϩ1 I and II (no saturation occurred at 20 K and 2 mW). Thus, we assign these EPR signal features to the [4Fe-4S] ϩ1 cluster of 4-HBCR. A difference spectrum taken from reduced 4-HBCR at Ϫ573 mV at 20 mW and 20 K minus one taken at 2 mW and 40 K is shown in Fig. 6B. This spectrum could be simulated as a rhombic S ϭ 1/2 species with g values at 2.063, 1.954, and 1.934, which can be interpreted as the EPR signal of a single non-interacting [4Fe-4S] ϩ1 cluster. However, due to the complexity of the spectra, presumably because of multiple magnetic interactions, this difference spectrum has to be taken cautiously and may only represent the part of the [4Fe-4S] ϩ1 clusters in a non-interacting form. As described above, this [4Fe-4S] cluster was also identified by Mössbauer spectroscopy; with this technique its redox potential has already been estimated to be substantially more negative than those of the  Table I).
[2Fe-2S] clusters. The gradual rise of the [4Fe-4S] ϩ1 signal features was accompanied by a loss of intensity of all features assigned to [2Fe-2S] ϩ1 clusters, e.g. those at g ϭ 2.04, 2.02, and 1.97 (Figs. 4 and 6A). To demonstrate this more clearly, features at g ϭ 2.04 (assigned to [2Fe-2S] I), g ϭ 1.97 (assigned to [2Fe-2S] ϩ1 I and II), g ϭ 2.06, and g ϭ 1.935 (assigned to [4Fe-4S] ϩ1 ) were plotted versus the redox potential. Both the rise of the EPR signal of the [4Fe-4S] ϩ1 cluster and the disappearance of the features assigned to the [2Fe-2S] clusters followed a Nernst-like curve with EЈ 0 ϳ Ϫ465 mV (Fig. 7) giving further evidence that reduction of the [4Fe-4S] cluster was associated with the disappearance of the EPR signals of the [2Fe-2S] ϩ1 clusters. The results obtained from Mössbauer spectroscopy of dithionite-reduced 4-HBCR neither indicated the reduction of the [2Fe-2S] ϩ1 clusters to the fully reduced [2Fe-2S] 0 state nor the formation of a high-spin system. Thus, we suggest that the complete loss of the low-spin [2Fe-2S] ϩ1 EPR signals results from strong magnetic interactions with each other and/or other paramagnets. Such interactions might result in a broadening and/or a splitting of EPR signals making a clear assignment highly complex. Moreover, the maximal total spin concentration determined at 40 K and 2 mW was 2-2.5 spins/monomeric 4-HBCR at Ϫ470 mV. At this potential the [4Fe-4S] cluster was only ϳ50% reduced. When the potential was lowered to Ϫ573 mV, the spin concentration/monomer decreased again to 1.7 Ϯ 0.3 spins at Ϫ573 mV. This suggests that spins were escaping detection due to broadening effects resulting from magnetic interactions among clusters.

EPR Spectroscopy of the Mo(V) State and the Flavin Semiquinone
Mo(V) EPR Signal of Reduced 4-HBCR As Isolated-Depending on the amount of dithionite added and the incubation time of enzyme with excess dithionite different types of Mo(V) signals are usually obtained in the XO family; according to Bray's nomenclature these are termed "slow" and "rapid" (30). Isolated, dithionite-reduced 4-HBCR exhibited an axial Mo(V)  Fig. 3, B and C Subspectra 3-6 were simulated using a spin-Hamiltonian yielding spin expectation values that are used in the nuclear Hamiltonian to describe the magnetic hyperfine field, i.e. H hf ϭ ͗S ͘ Ā Ī, where I is the nuclear spin and Ā is the effective magnetic hyperfine coupling tensor, the main components of which are given in the  EPR signal with g ϭ 1.990, 1.965, and 1.965, similar to rapidtype signals observed in other members of the XO family (Fig.  8). At temperatures above 120 K this EPR signal was not obscured by those of paramagnetic Fe-S clusters. However, we did not observe a time and/or dithionite concentration-depend-ent change in the shape and intensity of the Mo(V) EPR signal. Thus, we exclude the possibility that a significant amount of a slow-type EPR signal can be obtained from the dithionitereduced enzyme. The split features of the Mo(V) EPR spectrum of 4-HBCR are typical for a hyperfine interaction of a Mo(V) species with one or more protons, such as that bound to the terminal sulfur ligand of molybdenum, as reported for several other members of the XO family (10 -12).
Mo(V) EPR Signals at Different Redox Potentials-Many members of the XO family in the as-prepared, resting, oxidized state display a Mo(V) EPR signal that usually accounts for 1-10% of the total molybdenum and is assigned to a "Resting signal" (30). Traces of such a signal were found with 4-HBCR at positive potentials (Ͼ Ϫ200 mV) in a redox titration study (see Fig. 4, spectrum A). This signal was obscured significantly by the redox dye radical, and its spin concentration was estimated to be less than 0.1% of total molybdenum. We did not investigate this EPR signal further. In a redox titration experiment Mo(V) EPR signals occurred at potentials below Ϫ300 mV, which were identical to the one observed in the as-isolated, dithionite-reduced state (Fig. 8). Such Mo(V) signals were observed over an unusually broad potential range from Ϫ300 mV to Ϫ570 mV. In Fig. 9D the relative signal intensities of the Mo(V) species are plotted versus the poised potential. A fit with Nernst curves gave the following apparent redox transitions of the molybdenum site in 4-HBCR: Mo(VI)/Mo(V) ϳ Ϫ380 mV and Mo(V)/Mo(IV) ϳ Ϫ500 mV (Fig. 9D). The maximum total spin concentration of the Mo(V) species was 1.6/native dimeric enzyme, which matches the molybdenum content of 1.9 mol/ mol dimeric enzyme.
To detect magnetic interactions with other paramagnets a power saturation/temperature study of the Mo(V) signal was performed at different potentials. At Ϫ354 mV the Mo(V) EPR signal did not change significantly in shape and/or line width when the temperature was lowered stepwise from 200 K to 60 K (Fig. 9A). In contrast, at Ϫ482 mV, the shape of the Mo(V) EPR signal showed a strong dependence on the temperature (Fig. 9B). When the temperature was lowered stepwise from 200 K to 80 K, the hyperfine pattern of the bound protons was gradually lost, probably by broadening of the overall spectrum (Fig. 9B), and at temperatures below 60 K it appeared again. As shown above, at Ϫ354 mV both [2Fe-2S] clusters were in the paramagnetic state, but the [4Fe-4S] cluster was not. This indicates that the reduction of the [4Fe-4S] cluster, which is mainly in the paramagnetic state at Ϫ482 mV, induced the temperature-dependent changes in the Mo(V) signal so that an interaction exists between the [4Fe-4S] ϩ1 cluster and the Mo(V). To test this further we also determined the power saturation behavior of the Mo(V) signal at potentials where the [4Fe-4S] cluster was diamagnetic and paramagnetic. At Ϫ354 mV, the Mo(V) EPR signal became saturated at 80 K between 2 and 20 mW, and at 200 mW the signal intensity fell to 50% of the non-saturated value (Fig. 9C). However, at Ϫ482 mV the Mo(V) EPR signal did not even saturate at 80 K and 200 mW. This can be explained by a magnetic interaction of the Mo(V) site with the faster relaxing paramagnetic [4Fe-4S] ϩ1 cluster.
Flavin Semiquinone EPR Signals-FAD has been previously determined to be the flavin cofactor in 4-HBCR (7). We followed the rise of the isotropic flavin semiquinone radical signal in a redox titration study. At high temperatures (Ͼ100 K) an isotropic EPR signal at g ϭ 2.008 was observed (not shown). Its line width (1.9 mT at 200 K and 2 mW) clearly differs from that of the radical EPR signal of the mediator mixture (1.3-1.4 mT). It is typical for an EPR signal of a blue neutral semiquinone FADH ⅐ as has also been suggested for the radical species found in the closely related quinoline 2-oxidoreductase (32). An accurate quantitative redox titration study to determine the midpoint potentials of the redox transitions of the flavin was complicated by the high ratio of redox dyes (30 M) to that of the enzyme (80 M). Thus, in the redox titration we followed both the change in line width as well as the total spin concentration of the isotropic EPR signal. Taking both parameters into account we estimate EЈ 0 ϳ Ϫ250 mV Ϯ 40 mV for the FAD/FAD ⅐ redox couple and EЈ 0 ϳ Ϫ470 mV Ϯ 25 mV for the FADH ⅐ / FADH redox transition.

Effect of 4-HBCoA on the Fe-S Clusters
Mössbauer spectra of dithionite-reduced 4-HBCR (650 M) taken after rapid mixing with 5 mM 4-HBCoA at 4°C and subsequent freezing on dry ice are shown in Fig. 10. Spectra were recorded at 120 K (A) and at 4.2 K in external fields of 20 mT Ќ␥ (B) and 7 T ʈ␥ (C). From the specific activity of the enzyme under the conditions used, 4-HBCoA was not depleted. In principle, the subspectra could be simulated with the same parameters as those used for the dithionite-reduced sample (Fig. 3, Table I (Fig. 10) showing that all [4Fe-4S] clusters were oxidized, whereas 75% of the [2Fe-2S] clusters were reduced. This indicates that under steady-state The enzyme was incubated for 5 s at 4°C and subsequently frozen on dry ice. A, spectrum taken at 120 K. B, spectrum taken at 4.2 K in an applied field of 20 mT Ќ␥. C, spectrum taken at 4.2 K in an applied field of 7 T ʈ␥. The envelope lines through the data points are the sum of four subspectra (see Table I  conditions of 4-HBCoA reduction, either the [2Fe-2S] clusters are not oxidized at all or that reduction of these clusters (by excess of dithionite present in the sample) is much faster than their substrate-dependent oxidation. A similar experiment was also performed using EPR spectroscopy. Spectra taken at 40 K and 0.2 mW were very similar to those obtained at Ϫ350 mV in the redox titration (Fig. 6) indicating that in the steady-state of substrate reduction most but not all of the [2Fe-2S] clusters were in the reduced state, whereas all [4Fe-4S] clusters were in the oxidized state. This also confirms that the complex changes of the Fe-S EPR spectra observed between Ϫ400 mV and Ϫ500 mV are mainly due to the reduction of the [4Fe-4S] cluster.

Nature of the Molybdenum Cofactor of 4-HBCR
Once released from molybdoenzymes, the molybdenum cofactor is extremely labile and in the presence of oxygen is oxidized to a number of stable variants including form A, form B, and urothione (21). However, controlled oxidation in the presence of iodine converts molybdopterin to form A and the nucleotide variant cofactor to form A-nucleotide monophosphate. Subsequent treatment of form A-nucleotide monophosphate with pyrophosphatase and alkaline phosphatase results in the formation of form A and nucleotide monophosphate.
To determine the molybdenum cofactor content of 4-HBCR, cofactor analysis was performed as described under "Experimental Procedures." High pressure liquid chromatography analysis of the pooled 10-mM acetic acid fractions from purified 4-HBCR revealed a major peak with the same retention time as standard form A (not shown). The fluorescence spectrum was examined and was shown to be identical to that of form A. No significant peaks appeared upon high pressure liquid chromatography analysis of the 50-mM HCL-pooled fractions from 4-HBCR. These observations indicate that the cofactor present in 4-HBCR is molybdopterin mononucleotide and not a molybdopterin dinucleotide derivative.

Redox Centers of 4-OH-Benzoyl-CoA Reductase
Comparison with Other Xanthine Oxidase Family Members-The nature and spectroscopic properties of the redox centers in 4-HBCR have many similar features to those of other molybdopterin-containing enzymes of the XO family. In common with all members of this family, monomeric 4-HBCR contains two different [2Fe-2S] centers, a molybdopterin cofactor and a FAD binding subunit that is completely lacking in aldehyde oxidases (10 -12). Despite high similarities with other members of the family in terms of molecular architecture, the redox cofactors of 4-HBCR have some interesting spectroscopic and electrochemical properties that are summarized in Table II and are compared with other members of the XO family in Table III.
[4Fe-4S] Cluster and Flavin-The most dramatic difference between 4-HBCR and all other XO family members is the presence of a low potential [4Fe-4S] ϩ1/ϩ2 cluster (EЈ 0 ϭ Ϫ465 mV). The function of this cluster can be assumed to be to mediate low potential electrons from the natural electron donor ferredoxin to the molybdenum site. As the redox potential of the two [4Fe-4S] ϩ1/ϩ2 clusters of this ferredoxin are Ϫ435 mV and Ϫ585 mV, respectively (31), a direct electron transfer from each of the ferredoxin clusters to the [4Fe-4S] cluster of 4-HBCR would be possible. The reduction of the [4Fe-4S] cluster markedly influenced the EPR properties of both the [2Fe-2S] clusters and the Mo(V) species, indicating magnetic interactions between these redox cofactors. The [4Fe-4S] cluster is located in the ␤-subunit of 4-HBCR that contains an extra loop with 35-40 amino acids including five cysteine residues of which four have been considered to be involved in cluster ligation (7). Ligation of this cluster by four cysteines agrees with our Mössbauer spectroscopy data. A similar loop, with four conserved cysteines, has also been deduced from the gene coding for the ␤-subunit of 4-HBCR in the anaerobic phototrophic bacterium R. palustris (9) but has not yet been described for any other member of the XO family. Notably, this amino acid insertion is located directly adjacent to FAD binding motifs, suggesting that the [4Fe-4S] cluster is close to the flavin (7). These motifs include two boxes that are involved in pyrophosphate binding and are typical for the FAD-binding domain of the vanillyl-alcohol oxidase family (33). Notably, the molybdenum containing CODH to which 4-HBCR shows high amino acid similarities also contains such a FAD-binding domain (34). The redox properties of FAD in 4-HBCR are very different from those of other XO family members (see Table III). The gap between the two redox transitions from FAD to FADH is unusually high (Ͼ200 mV) and the FAD ⅐ /FADH (EЈ 0 ϳ Ϫ470 mV) couple is much more negative when compared with other XO family members (Table III). A possible function of the FAD could be the mediation of electron transfer between the "low potential" ([4Fe-4S] cluster, Mo(IV/V) transition) and "highpotential" electron carriers (both [2Fe-2S] clusters).
The three-dimensional structures of molybdenum and flavin containing enzymes available so far, e.g. those of CODH (34), XDH, or XO (35), show a common arrangement of the redox cofactors with the flavin being spatially separated from the molybdenum cofactor. The distance is in all cases Ͼ20 Å making direct electron transfer between them unlikely. In contrast, the architecture of these enzymes suggests an electron transfer chain from the flavin via both [2Fe-2S] centers to the molybdenum. Our EPR spectroscopy data provide evidence for a magnetic interaction between the [4Fe-4S] cluster and the molybdenum, suggesting that an alternative electron transfer chain exists in 4-HBCR. This would also fit with the redox potentials determined for the cofactors. Full reduction of molybdenum to the Mo(IV) state (EЈ 0 ϭ Ϫ500 mV) by the [2Fe-2S] ϩ1 clusters (EЈ 0 ϭ Ϫ205 mV and Ϫ255 mV) is thermodynamically unlikely (Table II), and it is more likely that the fully  (11,12). It has generally been accepted that they can be distinguished and assigned by their typical EPR spectroscopic properties. Usually, the cluster referred to as [2Fe-2S] I is of the plant type ferredoxins with g av ϳ 1.96 being observable up to 100 K, whereas [2Fe-2S] II exhibits a wider range of g values and relaxes substantially faster (optimal temperature is generally below 20 K). Typical differences between the anisotropies of both clusters in several members of the XO family have been compared elsewhere (36). Although some variations and overlapping existed, the results suggested that in all members of the XO family the differences and main EPR characteristics of both [2Fe-2S] clusters are conserved. However, the EPR properties of the two [2Fe-2S] centers of 4-HBCR do not follow this classification. None of the clusters exhibits either an unusual fast relaxation behavior nor a wide spread of g values. In contrast to other XO family enzymes the cluster with a wider spread of g values, referred to as [2Fe-2S] I, relaxes much slower than cluster II. Both clusters interact magnetically in 4-HBCR presumably because of their close proximity as demonstrated in other enzymes of the family, e.g. Desulfovibrio gigas aldehyde oxidoreductase (closest distance: 13.5 Å) (37) or CODH of O. carboxydovorans (12.6 Å) (34). More unusually, the reduction of the [4Fe-4S] cluster strongly affected the EPR properties indicating a strong interaction between these clusters and produced the curious effect that the EPR signals of the [2Fe-2S] clusters gradually disappeared, presumably because of broadening and/or splitting effects; shifts to the fully reduced or high-spin states of the [2Fe-2S] clusters can be excluded. So far, such a situation has not been described for [2Fe-2S] clusters of XO family enzymes.
Molybdenum Cofactor-Molybdopterin mononucleotide is the common cofactor of eukaryotic molybdoenzymes but has only rarely been reported as the cofactor among prokaryotes, such as in XDH from R. capsulatus (39). The impact of the presence or absence of a second nucleotide moiety at the pterin cofactor on the catalytic properties of an enzyme is not clear. It is generally accepted that catalysis takes place at the molybdenum cofactor. Thus, the lack of the second nucleotide in the pterin mononucleotide cofactor might enable the spatial positioning of the CoA-ester substrate of 4-HBCR adjacent to the molybdenum cofactor.
The EPR properties of the Rapid Mo(V) species are very similar to those of other XO family members; at least one proton is coupled to the Mo(V) species, possibly derived from an ϪSH or ϪOH/H 2 O group (10 -12). No evidence was obtained for a Slow Mo(V) signal. The big difference of Ϫ120 mV between the two redox transitions Mo(VI)/Mo(V) (EЈ 0 ϭ Ϫ380mV) and Mo(V)/Mo(IV) (EЈ 0 ϭ Ϫ500 mV) and the extremely low redox potential of the latter redox couple are more unusual. Similar potentials have only been reported for aldehyde oxidoreductase but for the slow signal rather than the rapid as we have found in 4-HBCR (38). Most other reported Mo(VI)/Mo(V)/Mo(VI) couples are in a narrow range around Ϫ350 mV (Table III). Our proposed catalytic mechanism for 4-HBCR, as discussed below, requires low potential electrons, which fits well with the unusually negative redox potentials of the Mo-cofactor.

Mechanistic Aspects of 4-OH-Benzoyl-CoA Reductase
It is agreed that in general an essential role of enzymes containing a molybdenum-pterin cofactor is the catalysis of controlled two-electron transitions, or oxygen transfers, between a substrate and spatially separated one-electron carriers such as ferredoxins, flavodoxins, or cytochromes. In the mechanism of known XO family enzymes one can normally distinguish two half-cycles. First, the two-electron oxidation of the substrate occurs as Mo(VI) is reduced to Mo(IV). Secondly, Mo(VI) is oxidatively regenerated in two single electron transfer steps (10 -12). As discussed above, in 4-HBCoA reduction two-electron transfer chemistry is most unlikely, and a model of alternate one-electron and one-proton transfer steps to the aromatic ring has been suggested to be the most plausible mechanism. This is by analogy with the Birch reduction of organic chemistry for the dehydroxylation of an aromatic ring. Due to the electron-withdrawing character of the para-hydroxy group, a single electron transfer to 4-HBCoA is facilitated compared with electron transfer to non-substituted benzoyl-CoA. 2 A single electron transfer would yield a radical anion with the highest electron density at the para-position (Fig.  11A). Therefore, hydroxy substituents at this position largely render the redox potential for the radical anion more positive. Preliminary kinetic studies with 4-HBCR supported one electron transfer steps to the aromatic ring (17).
In Fig. 11B we propose a catalytic cycle of 4-HBCR involving a separation of two-electron and one-electron transfer steps, as usual for XO family members, as well as the special one electron redox chemistry required for the removal of the phenolic hydroxy group from the aromatic ring. It runs counterclockwise to the catalytic cycle of other XO family members: the oxidative half-cycle (with respect to the oxidation of the molybdenum) proceeds in two single electron transitions to the substrate, whereas the reductive half includes a two-electron reduction from Mo(VI) to Mo(IV). Due to the striking similarity of 4-HBCR and other members of the XO family with respect to the amino acid sequence and the spectroscopic properties of the molybdenum-center, we assume a typical XO-type coordination of the molybdenum as shown in Fig. 11B. The catalytic cycle as presented consists of five distinguishable steps termed I-V.
Step I: Binding of the Substrate-In contrast to other XO family members 4-HBCoA binds via the para-hydroxy group not to the oxidized Mo(VI) but to the reduced Mo(IV) species with concomitant displacement of the water ligand.
Step II: Charge Transfer from the Molybdenum Cofactor to the Substrate Generating a Radical Anion and a Mo(V) Species-Such a charge transfer seems plausible because of the low reduction potential of the Mo(V)/Mo(IV) couple. Note that the radical anion can be stabilized not only via the aromatic ring including the carbonyl of the thioester but also over the dithiolene group at the pterin cofactor. The binding of the substrate alone should result in such a charge transfer.
Step III-Protonation of the para-position of the 4-HBCoA radical anion yielding a free radical (proton transfer to/from the sulfur-ligand of molybdenum is considered a typical partial reaction in the catalytic cycles of XO family members (10 -12)). As mentioned above this protonation is facilitated due to the high electron density of the radical anion at the para-position. Electron transfer (Step II) and protonation (Step III) may well be concerted reactions as a parallel proton-assisted electron transfer, as has been suggested for the enzymatic benzoyl-CoA reduction. 2 The protonation of the radical anion will pull the charge transfer of Step II forward.
Step IV: Homolytic Cleavage of the Carbon-Oxygen Bond Yielding Benzoyl-CoA and a Mo(VI)-OH Site.-This cleavage will be largely driven by the re-aromatization of the radical yielding benzoyl-CoA. Alternatives to this homolytic cleavage, such as formation of a phenyl cation or a H 2 O ϩ species are extremely unfavorable thermodynamically. Interestingly, a rate-limiting electron transfer step has been predicted at Ϫ360 mV in earlier kinetic studies (17) fitting to the Mo(VI)/Mo(V) transition (Ϫ380 mV).
Step V: Reductive Two-electron Regeneration of the Mo(IV)-Two electrons and two protons are transferred to the molybdenum center in order to close the cycle.
The mechanism presented is consistent with known chemistry and the properties of this enzyme.