Nonaromatic Products from Anoxic Conversion of Benzoyl-CoA with Benzoyl-CoA Reductase and Cyclohexa-1,5-diene-1-carbonyl-CoA Hydratase*

The enzymes benzoyl-CoA reductase and cyclohex-1,5-diene-1-carbonyl-CoA hydratase catalyzing the first steps of benzoyl-CoA conversion under anoxic conditions were purified from the denitrifying bacterium, Thauera aromatica. Reaction products obtained with [ring-13C6]benzoyl-CoA and [ring-14C]benzoyl-CoA as substrates were analyzed by high pressure liquid chromatography and by NMR spectroscopy. The main product obtained with titanium(III) citrate or with reduced [8Fe-8S]-ferredoxin from T. aromatica as electron donors was identified as cyclohexa-1,5-diene-1-carbonyl-CoA. The cyclic diene was converted into 6-hydroxycyclohex-1-ene-1-carbonyl-CoA by the hydratase. Assay mixtures containing reductase, hydratase, and sodium dithionite or a mixture of sulfite and titanium(III) citrate as reducing agent afforded cyclohex-2-ene-1-carbonyl-CoA and 6-hydroxycylohex-2-ene-1-carbonyl-CoA. The potential required for the first electron transfer to the model compoundS-ethyl-thiobenzoate yielding a radical anion was determined by cyclic voltammetry as −1.9 V versus a standard hydrogen electrode. The energetics of enzymatic ring reduction of benzoyl-CoA are discussed.

Traditionally, the biodegradation of aromatic compounds has been considered as a domain of aerobic metabolism. More recently, an increasing number of bacteria has been shown to metabolize aromatic compounds under anoxic conditions. Benzoyl-CoA (Fig. 1, compound 1) has been identified as a central intermediate in anoxic degradation of many aromatic substrates (for review see Refs. [1][2][3][4]. Benzoyl-CoA reductase has been purified from the denitrifying bacterium Thauera aromatica (5). The 160-kDa protein consists of four different subunits specified by the bcrCBAD genes. The enzyme contains a minimum of two Fe-S clusters (5,6) and is only active under strictly anoxic conditions (5). The reduction of benzoyl-CoA is coupled to the hydrolysis of two molecules of ATP/molecule of benzoyl-CoA. The natural electron donor for the enzyme-catalyzed reduction of benzoyl-CoA is a [8Fe-8S]-ferredoxin with a midpoint potential of Ϫ450 mV (versus a standard hydrogen electrode) (7). Alternatively, titanium(III) citrate, dithionite, or reduced methyl viologen can serve as artificial reductants. Some CoA-thioesters of fluoro-, amino-, and hydroxybenzoate can be reduced by benzoyl-CoA reductase albeit at low rates. Moreover, low molecular mass compounds such as hydroxylamine and azide are also reduced. All these reactions strictly depend on the hydrolysis of ATP (5).
In analogy to the Birch reduction of aromatic compounds by metallic sodium in liquid ammonia, the enzyme-catalyzed reduction of benzoyl-CoA has been proposed to proceed by two one-electron transfer reactions, each of which is followed by a protonation step (8,9). The crucial step is the first electron transfer to the aromatic ring yielding a radical anion. EPR studies indicated that the hydrolysis of ATP is responsible for conformational changes in the vicinity of [4Fe-4S] clusters, most probably resulting in a lowered reduction potential of the clusters (6).
In contrast to the results obtained with T. aromatica, cyclohexa-2,5-diene-1-carboxylate (Fig. 2, compound 4) and cyclohexa-1,4-diene-1-carboxylate (compound 5) were reported as products of benzoic acid reduction in cell suspensions of the phototrophic bacterium Rhodopseudomonas palustris. The structure assignments were based on gas chromatography and mass spectrometry after hydrolytic treatment of the reaction mixture (11).
The present study was initiated to establish the product of benzoyl-CoA reductase from T. aromatica. For this purpose, benzoyl-CoA labeled with 13 C or 14 C was incubated with purified benzoyl-CoA reductase using various electron donors. Enyzme products were analyzed by HPLC 1 and two-dimensional homonuclear and heteronuclear NMR spectroscopy. Furthermore, the effect of a thioester residue on the reduction potential of the aromatic moiety was determined by cyclic voltammetry using S-ethyl-thiobenzoate as a model compound. Bacterial Strains-T. aromatica strain K172 (DSM 6984) was isolated in our laboratory (12,13) and has been deposited in Deutsche Sammlung von Mikroorganismen (Braunschweig, Germany). Clostridium pasteurianum (ATCC 6013) was obtained from the American Type Culture Collection.
Growth of Bacterial Cells and Preparation of Cell Extract-T. aromatica was grown anoxically at 28°C in a mineral salt medium containing a mixture of 4-hydroxybenzoate and nitrate (ratio 1:3.5) as sole source of carbon and energy (12). Hydroxybenzoate and nitrate were continuously fed at a molar ratio of 1:3.5. Cells were harvested by centrifugation under anaerobic conditions and were stored in liquid nitrogen. Cell extract was prepared as described (17).
Purification of Ferredoxin-Ferredoxin was purified from T. aromatica (wet cell mass, 200 g) as described earlier (7). The yield was 60 mg of pure protein.
Assay for Benzoyl-CoA Reductase-Benzoyl-CoA reductase was assayed spectrophotometrically using reduced methyl viologen as electron donor (5) or by HPLC as described earlier (6). The reaction mixture for HPLC analysis contained 150 mM MOPS/KOH, pH 7.3, 10 mM MgCl 2 , 7.5 mM ATP, 10 mM phosphoenolpyruvate, 0.2 mM [ring-14 C]benzoyl-CoA, 80 nkat of pyruvate kinase, and purified benzoyl-CoA reductase from T. aromatica (0.1-0.2 unit) in a total volume of 1 ml. Titanium(III) citrate or sodium dithionite were added to a final concentration of 5 mM. Assays using reduced ferredoxin (37 M) from T. aromatica as reductant were performed under an atmosphere of 100% hydrogen. Reduction of T. aromatica ferredoxin was catalyzed by a ferredoxin-free crude preparation of hydrogenase from C. pasteurianum (1.5 units mg Ϫ1 ) that had been prepared as described previously (7). When ferredoxin was used, other reductants were omitted. Unless otherwise stated, the reaction was started by adding 50 l of a purified benzoyl-CoA reductase solution.
Sample Preparation for NMR Analysis-Reaction mixtures (5 ml) contained 150 mM MOPS/KOH, pH 7.3, 10 mM MgCl 2 , 7.5 mM ATP, 10 mM phosphoenol pyruvate, and 400 nkat pyruvate kinase, and 10 mM titanium(III) citrate or 10 mM sodium dithionite. The reaction was started by the addition of a mixture of [ring-14 C]benzoyl-CoA (44.4 kBq) and 5 mg of [ring-13 C 6 ]benzoyl-CoA. The mixtures were incubated under anoxic conditions at 37°C for 20 -40 min. The reaction was stopped by stirring for 15 min at 0°C under an aerobic atmosphere. The solution was lyophilized, and the powder was dissolved in 1 ml of 20 mM ammonium formate, pH 5. The solution was applied to a Lichrosorb RP-C18 column (25 ϫ 2 cm; Merck; flow rate, 8 ml min Ϫ1 ), which had been equilibrated with 20 mM ammonium formate, pH 5. After washing the column with two bed volumes of 10% methanol (v/v) in the same buffer, the CoA thioesters were eluted with 98% methanol (v/v) in the same buffer. Methanol was evaporated under reduced pressure at 30°C, and the residue was lyophilized. Product recovery was determined by liquid scintillation counting and was 60 -80% of the initial amount of 14 C added to the assay.
When reduced ferredoxin was used as reductant, the reaction was performed in a 100-ml stoppered bottle containing 9.6 ml of the standard assay mixture. The bottle was flushed with hydrogen for 20 min under continuous stirring. A ferredoxin-free hydrogenase preparation (H 2 reduction activity: 5-6 mol/min) from C. pasteurianum (7), [ring-13 C 6 ]benzoyl-CoA (7 mg), 85 kBq [ring-14 C]benzoyl-CoA, and 3.5 mg ferredoxin from T. aromatica (final concentration, 37.5 M) were added. The mixture was flushed with hydrogen for 20 min under continuous stirring. The assay was started by the addition of benzoyl-CoA reductase (0.8 unit) and was terminated after 15 min by adding 250 l of 1% formic acid (final pH, 3-4). The mixture was then lyophilized.
NMR Spectroscopy-The lyophilized samples were dissolved in D 2 O at pH 6 (uncorrected glass electrode reading). 1 H and 13 C NMR spectra were measured at 17°C using a four-channel Bruker DRX500 spectrometer equipped with pulsed field gradient accessory, a lock switch unit, and an ASPECT station. One-dimensional and two-dimensional HMQC, HMQC-TOCSY, and INADEQUATE experiments were per-  formed as described earlier (8). The duration of the 1 H spin-lock was 60 ms in the HMQC-TOCSY experiment.
Cyclic Voltammetry-The midpoint potential of S-ethyl-thiobenzoate was determined at 25°C by cyclic voltammetry. All measurements were carried out in super-dry acetonitrile with tetrabutyl ammonium hexafluorophosphate as supporting electrolyte. A three-electrode cell compartment was used throughout. For the electrochemical measurements, a potentiostat model 173 and a programmer model 175 (EG&G, Burlington, Vermont) were used. All potentials were calibrated against a ferrocene/ferrocenium couple.
Miscellaneous Methods-Polyacrylamide gel electrophoresis and SDS-polyacrylamide gel electrophoresis were performed as described by Schä gger and von Jagow (18). Proteins were visualized by Coomassie Blue staining (19). Protein was determined by the methods of Bradford (20) or Lowry et al. (21) using bovine serum albumin as standard.

RESULTS
Benzoyl-CoA reductase was purified from T. aromatica as described earlier (5). The enzyme was enriched 40-fold by four chromatographic steps with a yield of 25% to a specific activity of 0.48 mol min Ϫ1 mg Ϫ1 (measured with reduced methyl viologen as electron donor (5)) and Ͼ95% purity as estimated by SDS gel electrophoresis. This protein fraction contained benzoyl-CoA reductase (7 mol min Ϫ1 ) and cyclohexa-1,5-diene-1carbonyl-CoA hydratase (86 mol min Ϫ1 ) activity. Further purification of benzoyl-CoA reductase was achieved by affinity chromatography on Cibachron Blue (Amersham Pharmacia Biotech). The effluent of the column contained benzoyl-CoA reductase activity (4.5 mol min Ϫ1 ) and 7% of the initial hydratase activity (6.0 mol min Ϫ1 ). Using buffer containing 0.5 M KCl, 70% of the initial dienoyl-CoA hydratase activity could be eluted from the column.
A mixture of benzoyl-CoA reductase (7 mol min Ϫ1 ) and 1,5-dienoyl-CoA hydratase (86 mol min Ϫ1 ) was incubated with [ring-14 C]benzoyl-CoA in the presence of reduced ferredoxin or titanium(III) citrate. Reversed phase HPLC of the reaction mixture showed four peaks (1, 2, 3, and F) reflecting at least three radioactive products (Fig. 3A). The compound reflecting peak 3 was apparently more polar than benzoyl-CoA (peak 1, compound 1), product 2 was less polar than benzoyl-CoA, and product F had the same retention volume as free aromatic carboxylic acids and was probably formed by hydrolytic cleavage of the thioester.
Subsequent experiments were performed with [ring-13 C 6 ]benzoyl-CoA as substrate in an attempt to optimize the sensitivity and selectivity of NMR detection. The reaction mixture was desalted and was analyzed by NMR spectroscopy. Carbon atoms derived from the 13 C-enriched phenyl ring of benzoyl-CoA substrate gave 12 intense 13 C NMR signals (Fig.  4) that were attributed to two products (compounds 2 and 3) by two-dimensional NMR experiments. More specifically, two well separated spin systems were gleaned from a HMQC-TOCSY experiment, establishing the connectivities between 13 C and 1 H atoms as well as between 1 H atoms of the 13 C-labeled ring moieties ( Fig. 5 and Table I).
Each of the 12 13 C signals described above was split into a double doublet or a pseudo-triplet by 13 C 13 C coupling constants of 30 -70 Hz. Apparently, each of the observed 13 C atoms was directly bonded to two adjacent 13 C atoms. This result suggests that both products retained the connectivity of the ring carbon atoms of the benzoyl substrate. Some of the signals showed additional fine splittings (coupling constants of 1-8 Hz; Table  I) as a result of 13 C 13 C couplings via two or three bonds.
Four of the six intense 13 C NMR signals of compound 2 had chemical shifts values (121.9, 131.6, 137.4, and 140.6 ppm; cf. Fig. 4 and Table I) consistent with sp 2 -hybridization of the corresponding carbon atoms. Two signals were detected at 27.7 and 22.8 ppm reflecting two sp 3 -hybridized carbon atoms. This 13 C NMR signature in conjunction with the detected 13 C 13 C coupling pattern was consistent with a cyclic diene structure. Two-dimensional INADEQUATE, HMQC, and HMQC-TOCSY experiments (Table I and Fig. 5) afforded the entire network of carbon-carbon, carbon-proton, and proton-proton connectivities (28 detected homo-and heterocorrelations between 1 H and 13 C of the carbocyclus). Specifically, HMQC spectroscopy revealed correlations of the 13 C NMR signals at 121.9, 131.6, and 140.6 ppm to 1 H NMR signals at 6.22, 6.02, and 6.97 ppm, respectively. The 13 C signal at 137.4 ppm did not correlate to any 1 H NMR signal in the HMQC experiment. This finding suggested a cyclohexadiene-1-carbonyl derivative. Carbon-carbon coupling constants (68.8 and 53.0 Hz) detected for the NMR signal at 137.4 ppm (C-1) as well as INADEQUATE spectroscopy showed that C-1 was adjacent to two sp 2 -hybridized carbon atoms with 13 C NMR signals at 140.6 ppm (C-2) and 121.9 ppm (C-6). The signal at 121.9 ppm (C-6) gave an additional carboncarbon correlation to the signal at 131.6 ppm (C-5), indicating a cyclohexa-1,5-diene-1-carbonyl system. The presence of a CoA residue was revealed by the 1 H NMR data (not shown), thus establishing compound 2 as cyclohexa-1,5-diene-1-carbonyl-CoA.
The second product in the mixture (compound 3) was identified as 6-hydroxycyclohex-1-ene-1-carbonyl-CoA by the same  ). C, the assay was started with Ti(III) citrate (5 mM) as electron donor, after 1 min sodium dithionite (5 mM) was added to the assay. Products formed after 1 min (solid line) and 5 min (dotted line) of incubation with dithionite. Radioactivity was monitored in A and B, and ultraviolet absorption at 260 nm was monitored in C. The peaks with numbers were isolated by preparative HPLC and identified by NMR spectroscopy. F, free aromatic carbonic acids that do not bind to the HPLC column. approach, as described above for compound 2 (Table I). Compound 3 had been identified earlier as the reaction product of 1,5-dienoyl-CoA hydratase from T. aromatica (10). Cyclohexa-2,5-diene-1-carbonyl-CoA and cyclohexa-1,4-diene-1-carbonyl-CoA, which had been reported as products of benzoyl-CoA reduction catalyzed by benzoyl-CoA reductase in R. palustris (11), were not found in our reaction mixtures.
As reported earlier, dithionite can also act as an artificial electron donor for benzoyl-CoA reductase with rates comparable with other artificial electron donors (5). Surprisingly, HPLC analysis of the products formed from incubation of [ring-14 C]benzoyl-CoA with dithionite and a protein mixture contain-ing benzoyl-CoA reductase and 1,5-dienoyl-CoA hydratase revealed a product pattern (Fig. 3B) that was different from that obtained with titanium(III) citrate or reduced ferredoxin (Fig.  3A). Peaks reflecting the cyclic 1,5-diene (compound 2) or the cyclic 6-hydroxy-1-monoene (compound 3) were not observed with dithionite as artificial electron donor. On the other hand, two novel peaks reflecting compounds 8 and 9 were detected (Fig. 3B). Moreover, when dithionite (5 mM) was added to a benzoyl-CoA reductase assay that had been started with titanium(III) citrate as electron donor, the HPLC pattern shown in Fig. 3A changed to that shown in Fig. 3C.
Compound 8 and compound 9 formed by enzymatic conversion of [ring-13 C 6 ]benzoyl-CoA were analyzed by NMR spectroscopy (Table I). The detected 13 C NMR chemical shifts (120 -140 ppm) and coupling signatures (double doublets or triplets) were consistent with cyclic hexamonoene motifs for both compounds. Notably, each of the 13 C atoms correlated to directly bonded 1 H atoms in the HMQC experiment, thus indicating that the C-1 positions of both compounds carried a hydrogen atom and were not involved in a double bond. INADEQUATE and HMQC-TOCSY experiments afforded the entire network of carbon-carbon and carbon-proton bonds establishing compound 9 as [ring-13 C 6 ]cyclohex-2-ene-1-carbonyl-CoA and compound 8 as [ring-13 C 6 ]6-hydroxycyclohex-2-ene-1-carbonyl-CoA (Fig. 7). No cyclic diene was detected with dithionite as electron donor.
To clarify the effects of electron donors on the product pattern, we analyzed the effect of sodium dithionite on a benzoyl-CoA reductase free preparation of 1,5-dienoyl-CoA hydratase by HPLC. Under equilibrium conditions, similar amounts of cyclohexa-1,5-diene-1-carbonyl CoA (peak 2) and 6-hydroxycyclohex-1-ene-1-carbonyl-CoA (peak 3) were present in assays of the hydratase (Fig. 8A). After addition of sodium dithionite, both compounds were converted into 6-hydroxycyclohex-2-ene-1-carbonyl-CoA (Fig. 8, B and C, peaks 8). Fig. 8B shows that an additional intermediate apolar product (reflecting peak 10) is formed after 1 min of incubation. After 10 min of incubation this peak disappeared in the chromatogram (Fig. 8C). It is conceivable that peak 10 represents cyclohexa-2,5-diene-1-carbonyl CoA (Fig. 7, compound 10), which can be converted rapidly into compound 8 by the addition of water. When sodium dithionite was replaced by 1 mM sodium sulfite, the formation of compound 8 could not be detected, whereas a mixture containing 1 mM sodium sulfite and 1 mM titanium(III) citrate had the same effect as sodium dithionite (not shown). In the absence of 1,5-dienoyl-CoA hydratase, no spontaneous reaction of dithionite with 2 and 3 could be observed.
For the energetics of benzoyl-CoA reduction, the transfer of the first electron to the aromatic ring of benzoyl-CoA yielding a radical anion is considered as the crucial step in the reaction catalyzed by benzoyl-CoA reductase (9). A highly negative redox potential is expected for this reaction. The natural substrate benzoyl-CoA is not suitable for the direct determination of this potential because irreversible follow-up reactions occur in protic solvents. Therefore, the substrate analog S-ethyl- thiobenzoate, which is soluble in aprotic solvents, was synthesized and analyzed by cyclic voltammetry.
The thioester was reduced at a midpoint potential of Ϫ2.1 V versus an Ag/AgCl reference electrode. The reaction was reversible; minor waves appeared in the reverse scan at a potential of approximately Ϫ0.5 V versus Ag/AgCl indicating, a second slow process (data not shown). This follow-up product probably results from protonation of the radical anion yielding a free radical. Residual traces of water in the solvent could be the proton donor for such species.

DISCUSSION
The structures of six CoA thioesters obtained after incubation of [ring-13 C 6 ]benzoyl-CoA with purified benzoyl-CoA reductase (BcrCBAD protein) and 1,5-dienoyl CoA hydratase (Dch protein) from T. aromatica were established by NMR analysis. The use of multiply 13 C labeled substrate facilitated signal assignments by two-dimensional INADEQUATE, HMQC, and HMQC-TOCSY spectroscopy and enabled NMR analysis with product mixtures.
Specifically, the NMR data established cyclohexa-1,5-diene-1-carbonyl-CoA (compound 2) as the product of benzoyl-CoA reduction with titanium(III) or with reduced [8Fe-8S]-ferredoxin from T. aromatica as electron donors. This is in agreement with the results of Koch et al. (8), who identified this compound as an early product in experiments with crude cell extracts of T. aromatica. The formation of this cyclic hexadiene implicates a two-electron reduction of benzoyl-CoA and fits the previously determined stoichiometry (6). The identification of 6-hydroxy-cyclohex-1-ene-1-carbonyl-CoA (compound 3) as next intermediate in the benzoyl-CoA pathway in T. aromatica further supports the cyclic 1,5-diene structure, because the 6-hydroxy-1-ene can be formed by the enzymatic addition of water to the conjugated double bonds of the diene (10), either via a 1,4-or via a 1,2-addition mechanism.
The four genes of benzoyl-CoA reductase from T. aromatica and R. palustris show high similarity (78% sequence identity) (22,23). Interestingly, Gibson and Gibson (11) detected cyclohexa-2,5-diene-1-carboxylate (compound 4) and cyclohexa-1,4diene-1-carboxylate (compound 5) as products of benzoate re-  duction in cell suspension experiments with R. palustris. These compounds were identified by gas chromatography/mass spectrometry techniques after derivatization of the free acids to the corresponding methyl esters. The discrepancy to our results might have different reasons. Either benzoyl-CoA reduction proceeds via different mechanisms in R. palustris as discussed elsewhere (3,22), or the diene isomers detected by GC/MS were artificially formed during derivatization of the products.
In the so-called Birch reduction of benzoic acid with elemental sodium or lithium as reductants, 2,5-cyclohexadiene-1-carboxylate (compound 4) is formed under kinetic control (24). The electron withdrawing character of the acid group of benzoate stabilizes the anion radical at carbon 1 of the ring; the second electron adds in para position because of electrostatic repulsion. However, in the presence of a strong base, the thermodynamically more stable conjugated 1,5-diene can be obtained (25,26). This base-catalyzed rearrangement seems to occur in the catalytic cycle of T. aromatica.
The formation of the hex-1-ene compound (compound 6) at a slow rate (5) explained by a two-electron reduction of the 1,5diene (compound 2) catalyzed by benzoyl-CoA reductase. Although the hex-1-ene (compound 6) and the 2-hydroxyhexane compound (compound 7) may play an important role in the aromatic metabolism of R. palustris (3,22), their formation in T. aromatica appears as an in vitro effect without metabolic relevance.
The natural electron donor, reduced ferredoxin, and the artificial reductant titanium(III) citrate yielded the same prod-ucts. Surprisingly, cyclohex-2-ene-1-carbonyl-CoA (compound 9) and 6-hydroxycyclohex-2-ene-1-carbonyl-CoA (compound 8) were identified as products of benzoyl-CoA reductase and cyclohexa-1,5-diene-1-carbonyl-CoA hydratase when dithionite was used as an artificial electron donor. Our results strongly suggest that the artificial formation of the 6-hydroxy-2-ene (compound 8) was a product of dienoyl-CoA hydratase activity via a set of artificial isomerization and hydration reactions. In this case, cyclohexa-2,5-diene-1-carbonyl-CoA (compound 10) is suggested as an intermediate isomerization product (Fig. 7). Under anaerobic conditions in the presence of benzoyl-CoA reductase, MgATP, and an excess of dithionite, this hypothetical diene is further reduced to cyclohexa-2-ene-1-carbonyl-CoA (compound 9). The following observations support our suggestions: (i) When sodium dithionite was added to a benzoyl-CoA reductase/1,5-dienoyl-CoA hydratase assay that had been started with Ti(III) citrate as electron donor and benzoyl-CoA as substrate, the product pattern shifted immediately from the cyclohexa-1,5-diene/6-hydroxymonene mixture (reflecting peaks 2 and 3 in the HPLC chromatogram; cf.  carbonyl-CoA (compound 8) as the unique product from a cyclohexa-1,5-diene/6-hydroxymonene mixture. 6-Hydroxycyclohex-2-ene-1-carbonyl-CoA (compound 8) is considered to be a dead end product that is not in equilibrium with the corresponding diene 10 (Fig. 7). A mixture of sulfite and Ti(III) citrate could replace dithionite in the artificial product formation, whereas sulfite alone had no effect on dienoyl-CoA hydratase activity. We assume that a sulfoxide radical anion is formed from sulfite by reduction with Ti(III) citrate as an active species; it is known that anaerobic dithionite solutions contain the same radical anion (27).
For energetics of anaerobic enzymatic ring reduction, a Birch mechanism has been proposed for enzymatic reduction of the aromatic ring, which predicts alternating one-electron transfer and protonation steps (8,9). A ketyl radical anion at the carbonyl C atom of the thioester group rather than a benzene radical anion was suggested as the first intermediate. In case of benzene reduction, the potential for the first electron transfer step is Ϫ3.1 V (versus a normal hydrogen electrode) (28). A thioester group at the aromatic ring should lower this potential significantly (9). Indeed, using S-ethyl-thiobenzoate the potential is lowered to Ϫ1.9 V (versus a normal hydrogen electrode). This drastic effect illustrates the importance of the thioester group in enzymatic reduction of benzoyl-CoA.
However, the energetics of enzymatic ring reduction are still not clear. The two electron reduction of the aromatic ring is driven by the stoichiometric hydrolysis of 2 ATP to 2 ADP ϩ 2 P i . If the free hydrolysis energy of both ATP molecules (ϳϪ100 kJ) is coupled to a single electron transfer the redox potential can maximally be lowered by 1 V. Assuming this situation, the potential of the first electron transferred from the natural electron donor ferredoxin with EЈ°ϭ Ϫ0.45 (versus normal hydrogen electrode) could be lowered to Ϫ1.45 V in enzymatic aromatic ring reduction. In conjunction with the determined potential of the model compound S-ethyl-thiobenzoate (Ϫ1.9 V), the hydrolysis of ATP appears to be insufficient to enable ring reduction. However, partial protonation of the electron accepting carbonyl function by a protonated amino acid residue or a protonated Fe-S cluster could result in a more positive redox potential for the first electron transfer yielding the ketyl radical anions 11 and 12 (Fig. 9) as proposed by Buckel and Keese (9). The subsequent protonation of the radical anion seems to be highly exergonic, thus driving the endergonic first electron transfer step. The transfer of the second electron to the protonated free radical (compound 13) should require less energy than the formation of the radical anion, and protonation of compound 14 to the diene product (compound 2) is considered a highly exergonic reaction.