Coenzyme A-dependent Aerobic Metabolism of Benzoate via Epoxide Formation*

In the aerobic metabolism of aromatic substrates, oxygenases use molecular oxygen to hydroxylate and finally cleave the aromatic ring. In the case of the common intermediate benzoate, the ring cleavage substrates are either catechol (in bacteria) or 3,4-dihydroxybenzoate (protocatechuate, mainly in fungi). We have shown before that many bacteria, e.g. Azoarcus evansii, the organism studied here, use a completely different mechanism. This elaborate pathway requires formation of benzoyl-CoA, followed by an oxygenase reaction and a nonoxygenolytic ring cleavage. Benzoyl-CoA transformation is catalyzed by the iron-containing benzoyl-CoA oxygenase (BoxB) in conjunction with an FAD and iron-sulfur centers containing reductase (BoxA), which donates electrons from NADPH. Here we show that benzoyl-CoA oxygenase actually does not form the 2,3-dihydrodiol of benzoyl-CoA, as formerly postulated, but the 2,3-epoxide. An enoyl-CoA hydratase (BoxC) uses two molecules of water to first hydrolytically open the ring of 2,3-epoxybenzoyl-CoA, which may proceed via its tautomeric seven-membered oxepin ring form. Then ring C2 is hydrolyzed off as formic acid, yielding 3,4-dehydroadipyl-CoA semialdehyde. The semialdehyde is oxidized by a NADP+-dependent aldehyde dehydrogenase (BoxD) to 3,4-dehydroadipyl-CoA. Final products of the pathway are formic acid, acetyl-CoA, and succinyl-CoA. This overlooked pathway occurs in 4–5% of all bacteria whose genomes have been sequenced and represents an elegant strategy to cope with the high resonance energy of aromatic substrates by forming a nonaromatic epoxide.

Aromatic compounds like benzoate belong to the second most abundant class of organic growth substrates, next to carbohydrates, which nature provides as feedstock to animals (with limited aromatic metabolism), fungi, and bacteria. Normally benzoate is converted by dioxygenases to catechol (bacteria) or protocatechuate (mainly fungi), followed by ring cleavage catalyzed by central ring-cleaving dioxygenases. The well known ␤-ketoadipate pathway is the paradigm of this metabolic achievement (1,2). A new pathway for aerobic benzoate oxidation has been postulated for Azoarcus evansii and for a Bacillus stearothermophilus-like strain (3,4). In the meantime, this pathway or its characteristic genes were found in various other bacteria, either as the only pathway or as an additional benzoate metabolic pathway that comes into play under certain conditions (5,6).
The new principle differs in many aspects from the orthodox situation. First, all intermediates starting with benzoyl-CoA are processed as coenzyme A thioesters rather than as free acids (4). Second, ring cleavage is hydrolytic rather than oxygenolytic (7). Third, none of the classical enzymes is involved in any step, except ␤-ketothiolase, which cleaves ␤-ketoadipyl-CoA into succinyl-CoA and acetyl-CoA (8). Fifteen genes coding for the benzoate oxidation (box) pathway are clustered on the A. evansii chromosome (8). These box genes code for the following functions: a putative ATP-dependent benzoate transport system, a benzoate-CoA ligase, a benzoyl-CoA oxygenase and reductase, a ring-opening enzyme, enzymes for ␤-oxidation of CoA-activated intermediates, a thioesterase, and a lactone hydrolase, as well as completely unknown enzymes belonging to new protein families (8).
To elucidate the exciting mechanism of benzoyl-CoA oxidation catalyzed by BoxAB, we applied 18 O labeling studies and analyzed the products by mass spectrometry. It turned out that the product contained only one additional oxygen and could be derivatized by an epoxide trapping agent. In summary, the results indicate that BoxAB is a benzoyl-CoA epoxidase forming 2,3-epoxybenzoyl-CoA rather than a benzoyl-CoA dioxygenase/reductase forming the 2,3-dihydrodiol of benzoyl-CoA. BoxC acts on this epoxide or its oxepin tautomer by adding two molecules of water and thus eliminating ring C2 as formic acid. This new principle is widely distributed and has a counterpart in a novel pathway of phenylacetyl-CoA oxidation that also involves coenzyme A thioesters, epoxide, and oxepin intermediates. 2
Synthesis of CoA Esters-Benzoyl-CoA was prepared according to published procedures (4).
Bacterial Cultures-A mutant of A. evansii KB740 (DSMZ6869) (13) with a modified chromosomal boxB gene coding for BoxB with a C-terminal Strep-tag (BoxB Strep , courtesy of J. Gescher) (11) was grown aerobically at 37°C with benzoate as the sole source of cell carbon and energy (14) in a 200-liter fermentor (air flow, 50 liters min Ϫ1 ; 200 rpm). Benzoate was added continuously when the initially added substrate (5 mM) was nearly consumed. The cells were harvested in the exponential growth phase at an optical density at 578 nm of 6, which corresponded to 1.6 g of cells (dry mass) liter Ϫ1 (4). The culture was cooled to 8°C, and the cells were harvested by continuous flow centrifugation. The yield was 200 g of cells (wet mass) mol Ϫ1 benzoate. The cells were frozen in liquid N 2 and stored at Ϫ70°C.
Cloning and Overexpression of the box His C Strep Gene-The gene was amplified from chromosomal DNA of A. evansii by colony PCR with Pfu polymerase, using forward primer KpnI-BoxC_fwd, 5Ј-GATCCTGGTACCCAAGCAGTCGCGAAC-AAGC-3Ј, and reverse primer rev_BoxCBamHI, 3Ј-CAAGC-TGACCTTGGCGCACCCTAGGATGAAG-5Ј, where bold type indicates the boxC gene and underlined type indicates restriction sites. The forward primer contained a KpnI restriction site instead of the start codon, and the reverse primer contained a BamHI restriction site instead of the stop codon. The amplified gene was purified using the PCR purification kit as described in the instruction manual (Qiagen). The amplified gene and the vector pASK-IBA43plus were cut with KpnI and BamHI, purified using the PCR purification kit, and ligated with T4 DNA ligase. The vector pASK-IBA43plus contains a N-terminal start codon followed by codons for a His 6 tag and C-terminal codons for a Strep-tag in front of the stop codon. The construct pASK-IBA43plus_Box His C Strep was transformed into Escherichia coli DH5␣ by electroporation, and the correct sequence was verified by restriction and sequence analysis. E. coli/pASK-IBA43plus_Box His C Strep was grown aerobically at 37°C in lysogeny broth medium with 50 g of ampicillin ml Ϫ1 in a 10-liter fermentor (300 rpm). At an optical density (light path, 1 cm) at 578 nm (A 578 nm ) of 0.5, 200 g of anhydrotetracycline liter Ϫ1 was added for induction. The cells were harvested in the exponential growth phase at an A 578 nm of 1.6. The culture was cooled down to 4°C, and the cells were harvested by centrifugation and stored at Ϫ70°C. The yield was 20 g of cells (wet mass).
Preparation of Cell Extracts-All of the steps were performed at 4°C under anaerobic conditions. Frozen cells were suspended in an equal volume of 20% (v/v) glycerol containing 0.05 mg of DNaseI ml Ϫ1 . The suspension was passed through a French pressure cell at 137 MPa and centrifuged (1 h, 100,000 ϫ g).
Enzyme Purification-All of the steps were performed at 4°C under anaerobic conditions. BoxA was purified and assayed according to Mohamed et al. (3). BoxB Strep was purified by affinity chromatography. Cell extract (20 -150 ml, 70 mg of protein ml Ϫ1 , 100,000 ϫ g supernatant) was applied to a column of Strep-Tactin Superflow (25 ml; IBA GmbH, Göttingen, Germany), which was equilibrated with 200 ml of 10 mM Tris/ HCl, pH 8.0 (buffer A), at a flow rate of 3 ml min Ϫ1 . The column was washed with 60 ml of buffer A, 180 ml of buffer A containing 250 mM KCl, and 60 ml of buffer A. Protein was eluted with 60 ml of buffer A containing 2.5 mM desthiobiotin. Eluted protein (30 ml, 20 mg) was concentrated to 1-5 ml (Amicon; 30 kDa). The enzymes were stored at Ϫ70°C with 10% (v/v) glycerol. Box His C Strep was purified similar to BoxB Strep ; the column was washed with 60 ml of buffer A and 180 ml of buffer A containing 100 mM KCl. BoxD Mal was purified as described (12).
Protein Analysis-Protein concentration was determined by the Bradford method (15) and the BC assay kit as described in the instruction manual (Uptima, Interchim, Montlucon Cedex, France) using bovine serum albumin as standard. SDS-polyacrylamide (11.5%) gel electrophoresis used the Laemmli method (15) and Coomassie Blue staining (16).
Stoichiometry of BoxAB Reaction-NADPH oxidation at 30°C was measured spectrophotometrically at 377 nm (⑀[NADPH] 377 nm extrapolated 1,620 M Ϫ1 cm Ϫ1 ). Either limiting concentrations of benzoyl-CoA or oxygen were applied. Air-saturated (30°C) assay mixtures (500 l) with 100 mM Tris/ HCl buffer, pH 8.0, contained either 0.4, 0.1, or 0.05 mM benzoyl-CoA. They were mixed with 0.04 mg of BoxA ml Ϫ1 , 1.02 mg of BoxB Strep ml Ϫ1 , and 0.96 mg of Box His C Strep ml Ϫ1 . After the addition of 0.6 mM NADPH, the assay mixture was covered immediately with 200 l of paraffin oil to avoid further oxygen uptake. NADPH oxidation observed after the BoxABC (native enzyme BoxA in conjunction with recombinant enzyme BoxB Strep and recombinant enzyme Box His C Strep ) reaction caused by BoxA was subtracted. The slow endogenous, benzoyl-CoA-independent NADPH oxidation with O 2 , which is catalyzed by BoxA, was monitored as a constant slow absorption decrease at 377 nm, when benzoyl-CoA was consumed. This blank reaction was extrapolated back to the zero time point. This extrapolated A 377 nm value was subtracted from the initial A 377 nm value.
Enzyme Assays for BoxAB, BoxABC, and BoxABCD-Standard assay mixtures (0.04 -1 ml) containing 0.6 mM NADPH and 0.2 mM benzoyl-CoA in 10 mM Tris/HCl buffer, pH 8.0, were mixed at 4°C with 0.08 mg of BoxA ml Ϫ1 . Routinely, a NADPH regenerating system was included consisting of 3.3 mM MgCl 2 , 3.3 mM D-glucose 6-phosphate, and 2 units of glucose 6-phosphate dehydrogenase ml Ϫ1 . The reaction was started by the addition of 0.64 mg of BoxB Strep ml Ϫ1 . In some experiments, 0.16 mg of Box His C Strep ml Ϫ1 was added, and in some cases additionally 0.08 mg of BoxD Mal ml Ϫ1 . Labeling assays were performed in a closed tube (7.5 ml) with 50% 18 O 2 , 50% 16 The assay mixtures were stirred at 24°C for 30 min. The enzymatic reaction was stopped by adding a 5-fold volume of ethanol (Ϫ20°C). After incubation for 20 min at Ϫ20°C, the denatured protein was removed by centrifugation. The supernatant was evaporated under reduced pressure at 30°C. The residue was resolved in 100 l of H 2 O, and the products were purified by reverse phase HPLC. 3 BoxAB Reaction in the Presence of N,N-Diethyldithiocarbamate (DTC)-Standard assay mixtures (500 l) included 10 mM DTC. Aliquots (40 l) were taken at different points (t ϭ 0, 2, 4, 6, 8, 10, 15, 20, 40, and 60 min), and the samples were analyzed.
Purification of Products by Reverse Phase HPLC-Aliquots were applied to a column of Lichrospher 100 RP 18E, 5.0 m, 125 ϫ 4 mm (Wicom, Heppenheim, Germany), equilibrated with 40 mM ammonium acetate (NH 4 Ac), pH 6.8, containing 5% (v/v) acetonitrile (ACN) at a flow rate of 1 ml min Ϫ1 . An ACN gradient in the same buffer was used: 2 min to 5%, 1 min to 10%, 11 min to 30%, 1 min from NH 4 Ac at 30% to water at 30%, 3 min to 50%, 3 min back to 5%, 3 min from water plus 5% to NH 4 Ac plus 5%, and 6-min equilibration 5% in NH 4 Ac. Elution was monitored with an UV diode array detector routinely at 260 nm. The amount of the CoA thioesters was estimated based on the assumption that they exhibited identical absorption coefficients at 260 nm as benzoyl-CoA (⑀ 260 nm ϭ 21,100 M Ϫ1 cm Ϫ1 ). The retention times were as follows: 1.0 min for polar products; 3.5 min for product of BoxA, BoxB Strep , Box His C Strep , and Box-D Mal ; 6.4 min for product of BoxA, BoxB Strep , and Box His C Strep ; 7.0 min for product of BoxA and BoxB Strep ; 9.5 min for benzoyl-CoA; and 12.0 min for derivative product of BoxA and BoxB Strep with DTC. The fractions of 0.5 ml were collected and frozen at Ϫ20°C.
Analysis by Mass Spectrometry (MS)-Samples of HPLC fractions were transferred by a syringe pump or via nano-reverse phase HPLC into the nano-electrospray ionization source of a Finnigan LTQ-FT mass spectrometer (Thermo Electron Corporation, Waltham, MA) for online mass detection assembled from a linear ion trap and an ion cyclotron (7 Tesla magnet) with Fourier transform ion cyclotron resonance mass spectrometry.
Analysis by Direct Coupling of HPLC, MS, and MS/MS-Benzoyl-CoA and the product of the enzymatic reaction were injected with a FAMOS autosampler (Dionex) and desalted by transfer with an Agilent HPLC 1100 to a reverse phase trap column (Zorbax Eclipse XDB-C18, 5 m; 0.1 ϫ 15 mm). The sample was eluted at 200 nl min Ϫ1 with an ACN gradient from a quaternary HPLC pump (Ultimate, Dionex) and separated on a fused silica emitter of 0.075 ϫ 105 mm (Proxeon) packed with Pro C18, 3 m (YMC). Elution started with 100% A (H 2 O, 3% ACN, 0.1% formic acid) and 0% B (H 2 O, 80% ACN, 0.1% formic acid) for 11 min, followed by successive linear gradients in 4 min to 5% B, 20 min to 30% B, and 21 min to 70% B and terminated by 2 min at 70% B. The fused silica emitter connected to Ϫ2 kV was mounted in the nano-electrospray interface of the LTQ-FT. The mass spectrometer was operated in the data-dependent mode to automatically switch between MS and MS/MS acquisition. Survey MS spectra (from m/z 250 to 1800) were acquired in the FT-ICR with a resolution of 25,000. The most intense ion was isolated for high resolution (50,000) measurement in the FT-ICR with a 10-Da mass range. These ions were then fragmented in the linear ion trap using collisioninduced dissociation and recorded at low resolution (MS/MS scan). The latter ions were dynamically excluded for the following 30 s. The total cycle time was ϳ0.3 s.

Conversion of Benzoyl-CoA with NADPH and Oxygen and
Characterization of the Product-Benzoyl-CoA was transformed by the recombinant enzyme BoxB Strep in conjunction with the native enzyme BoxA (here referred to as BoxAB) with O 2 and NADPH as electron donor. The stoichiometry of the oxygen-dependent reaction was 1 NADPH oxidized and 1 O 2 consumed per 1 benzoyl-CoA transformed (Table 1). This ratio could be interpreted as the result of a dioxygenase/reductase reaction leading to the nonaromatic cis-2,3-dihydrodiol of benzoyl-CoA. The product was purified by HPLC and analyzed by electrospray ionization mass spectrometry (Fig. 1). However, no mass of benzoyl-CoA plus two oxygen atoms and two hydrogen atoms (expected MH ϩ ϭ 906) was found. The observed value of m/z ϭ 888.144 agrees well with the monoisotopic mass of benzoyl-CoA plus one oxygen atom (expected MH ϩ ϭ 888.1436). The peak at 910.127 represents the same product but associated with Na ϩ instead of H ϩ . Because not even traces of a product carrying two additional oxygen atoms could be observed, the product as isolated is not the expected benzoyl-CoA dihydrodiol. It is rather a hydroxylated benzoyl-CoA or an epoxide of benzoyl-CoA.
Still, the possibility exists that the actual product of the reaction was a labile dihydrodiol, which during isolation becomes stabilized by rearomatization through water elimination, thus forming 2-or 3-hydroxybenzoyl-CoA as a dead-end product. We therefore used direct coupling of HPLC, MS, and MS/MS in a fast experiment to minimize possible side reactions and to detect possible dihydroxylated intermediates. Direct injection is gentle compared with freeze-drying of the HPLC sample and dissolving it again. The fragments of CoA served as an indicator that the detected parent mass even of trace unknown products is indeed a derivative of CoA. Fragmentation of CoA in a mass spectrometer has been reported, but only the mass of 428 was attributed to adenosine 3Ј,5Ј-bisphosphate (17). We observed this mass and additional masses representing fragments of acyl-CoA thioesters. The fragments contain either the aromatic ring or the CoA adenylate moiety (Fig. 2). The method proved valid when tested with benzoyl-CoA ( Fig. 2A). We observed in addition a fragment with a mass of 365.2, which is likely caused by an additional loss of phosphoric acid (97.97) (463.13 Ϫ 97.97 ϭ 365.15). A related elimination of phosphate is known for phosphoserine leading to dehydroalanine (18). The product of BoxAB showed a minor 479.1 fragment and a major fragment of 381.2, which is likely formed from the minor fragment by an additional loss of phosphoric acid as in the case of benzoate (Fig. 2B).
These results indicate that the product detected under these gentle conditions was also a benzoyl-CoA derivative carrying only one additional oxygen atom. A monohydroxylated, aromatic derivative like 2-hydroxybenzoyl-CoA or 3-hydroxybenzoyl-CoA is an unlikely candidate for the following nonoxygenolytic cleavage of the ring, and the benzoate oxidation gene cluster does not code for a ring-cleaving dioxygenase. In case of an epoxide, one may expect a hydratase yielding the trans-2,3dihydrodiol of benzoyl-CoA, but a corresponding hydratase gene is not found in the gene cluster.   (Fig. 3). Again, no molecule species carrying two oxygen atoms were detected.

of the reaction catalyzed by BoxAB
To remove the reactive epoxide product of BoxAB, BoxC was included in the assay. The benzoyl-CoA-and oxygen dependent oxidation of NADPH at 30°C was measured spectrophotometrically at 377 nm (⑀(NADPH) 377 nm ϭ 1,620 M Ϫ1 cm Ϫ1 ). The assay (0.5 ml) contained 21 g of BoxA, 510 g of BoxB Strep , 280 g of Box His C Strep , and 0.6 mM NADPH. Either dissolved oxygen (air saturated water at 30°C contains 7.53 mg of O 2 liter Ϫ1 ; 0.235 mM O 2 ) or benzoyl-CoA limited the reaction. The assay mixture in the cuvette (optical path, 1 cm) was immediately covered with 200 l of paraffin oil to avoid further oxygen diffusion. NADPH oxidation by BoxA alone in the presence of oxygen was subtracted.

O or O 2 and Subsequent Transformation by Enoyl-CoA Hydratase BoxC and Aldehyde
Dehydrogenase BoxD-The products of the transformation of benzoyl-CoA with the three enzymes BoxABC are considered established: 3,4-dehydroadipyl-CoA semialdehyde and formic acid derived from C2 of the aromatic ring (7). Formic acid can-not be detected because of its small size (m/z Ͻ 50 cannot be detected), nor can a derivative of higher mass be obtained without the loss of one oxygen atom of formate. BoxD catalyzes the oxidation of the CoA-linked C 6 semialdehyde to the dicarboxylic acid under incorporation of water (12). Transformations were conducted with 16  O. When BoxAB or BoxABC were added, solely the mass peaks of unlabeled products were observed ( Table 2). The mass peaks corresponded to benzoyl-CoA carrying one additional oxygen atom and 3,4-dehydroadipyl-CoA semialdehyde, respectively. This finding seems to exclude the possibility that oxygen from water is incorporated into the C 6 product formed by BoxABC. Correspondingly, the assay with BoxABCD (native enzyme BoxA in conjunction with recombinant enzyme BoxB Strep , recombinant enzyme Box His C Strep and recombinant enzyme BoxD Mal ) yielded one product exhibiting two mass peaks with equal intensity corresponding to unlabeled and singly 18 O-labeled 3,4-dehydroadipyl-CoA ( Table 2). The single 18 O label therefore comes from water in the course of the aldehyde oxidation by BoxD. No double-labeled product was found.
When benzoyl-CoA was transformed in H 2 16 O in the presence of 50% 18 O 2 , 50% 16 O 2 , the product of BoxABC unexpectedly showed only one mass peak of 878 (Table 2), which corresponds to the unlabeled 3,4-dehydroadipyl-CoA semialdehyde. In contrast, the product of the assay with BoxABCD exhibited two mass peaks of equal amplitude ( Table 2). Mass peak 894  The product was isolated by reverse phase HPLC and analyzed by mass spectrometry (conditions as in Fig. 1). The peak at 889.155 originates from benzoyl-CoA with one 13 C. The mass difference of the benzoyl-CoA derivative with two 13 C of low abundance could not be resolved from the peak with one 18 O. The two peaks at 891.159 and 892.161 are benzoyl-CoA with one or two 13 C, respectively, plus one 18 O. corresponds to unlabeled 3,4-dehydroadipyl-CoA, and mass peak 896 corresponds to single 18 O-labeled 3,4-dehydroadipyl-CoA. Because the 18 O-labeled adipyl C6 carboxyl group is derived from C3 of the 3,4-dehydroadipyl-CoA semialdehyde (see Fig. 6), C3 of the semialdehyde must have been linked to 18 O. However, the observed missing labeling of the semialdehyde seemingly is contradictory to this conclusion because 50% of this product should also contain 18 O. This inconsistency can be explained if the free carbonyl oxygen of the aldehyde (product of BoxABC) rapidly exchanges 18 O with 16 O from water via the aldehyde hydrate, when the sample was prepared and the product was isolated in unlabeled water. This exchange reaction is well known (19). 3,4-Dehydroadipyl-CoA semialdehyde hydrate was observed before in NMR studies of the reaction mechanism of BoxC (7).

Determination of the Rate of Oxygen Exchange between the Carbonyl Oxygen of 3,4-Dehydroadipyl-CoA Semialdehyde and
Oxygen from Water-To estimate the rate of oxygen exchange between carbonyl 18 O of the semialdehyde and 16 O from water, the exchange was measured in the opposite direction. The reaction was analyzed with enzymatically synthesized, HPLC-purified 16 O-semialdehyde in 50% H 2 18 O, 50% H 2 16 O and analyzed by MS as a function of time. Indeed, the formation of the 18 Osemialdehyde species was observed (Fig. 4). In our previous preparation (see above) the semialdehyde was in contact with unlabeled water for several hours, allowing the complete exchange between the carbonyl 18 O of the semialdehyde and 16 O from water.
Evidence for an Epoxide as the Actual Product of BoxAB-The possibility that a dihydrodiol intermediate rearomatizes under elimination of water can be excluded, because the product of BoxAB is cleaved by BoxC nonoxygenolytically, which would not be possible in the case of an aromatic ring. The remaining option of how benzoyl-CoA may be activated by BoxAB using 1 O 2 plus 1 NADPH is the formation of an epoxide most likely between C2 and C3 of the ring. The second oxygen atom is released as H 2 O. We set out to obtain a derivative with DTC, which reacts with epoxides even under gentle enzyme assay conditions under opening of the epoxide ring (20,21). DTC did not disturb the enzymatic reaction. The 2,3-epoxide of the benzene ring of benzoyl-CoA, 2,3-epoxybenzoyl-CoA, is expected to yield two possible DTC adducts depending on the steric hindrance of the products: a main product (little steric hindrance) and a side product (more steric hindrance) (Fig. 5A). When benzoyl-CoA was transformed at 24°C with BoxAB, the reaction did not go to completion because of inactivation of the enzymes by its product (Fig. 5B) (11). This alone indicates that the BoxAB product is chemically reactive. When DTC was added in excess, the reaction went to completion, and nearly stoichiometric amounts of a new product appeared (Fig. 5B), which migrated in reversed phase HPLC at 12-14 min. The derivative apparently did not inactivate BoxAB anymore. The new product peak was collected and analyzed by UV-visible spectroscopy (Fig. 5C) and MS (Fig. 5D).
The UV-visible spectrum of the BoxAB product showed an absorption maximum at 260 nm (⑀ 260 nm Ϸ 14,000 M Ϫ1 cm Ϫ1 at pH 6.8) and another characteristic maximum at 310 nm (⑀ 260 nm Ϸ 7,000 M Ϫ1 cm Ϫ1 at pH 6.8) compared with benzoyl-CoA (⑀ 260 nm ϭ 21,100 M Ϫ1 cm Ϫ1 at pH 6.8) (11) (Fig. 5C). The DTC derivative had nearly lost the absorption at 310 nm, which would be consistent with an opening of the epoxide (Fig. 5C). The MS spectrum showed a main mass peak at 1037.177 that agrees well with the theoretical mass of the DTC adduct of 2,3-epoxybenzoyl-CoA (expected mass 1037.1769) (Fig. 5D). The same derivative was obtained when the product of BoxAB was isolated and then treated with DTC. This indicates that the Benzoyl-CoA was converted by BoxABC, and the 3,4-dehydroadipyl-CoA semialdehyde was isolated by reverse phase HPLC (conditions as in Fig. 1). Samples were collected after different incubation times and immediately injected within 14 min into the electrospray ionization source of a Finnigan LTQ-FT mass spectrometer. O by the combined action of BoxAB, BoxABC, or BoxABCD. The products were purified by reverse phase HPLC (cf. Fig. 1) and analyzed by mass spectrometry (cf. Fig. 3).
The abundance of the monoisotopic mass with two 13 C, which was deduced from the amplitude of the monoisotopic peak at 888.14, was subtracted from the amplitude of the mass detected at 890.15 attributed predominantly to that with one 18 O. The standard deviation of the percentage values is Ϯ 5%. b Below limit of detection. epoxide is not a transiently formed intermediate in the BoxAB catalyzed reaction, but it is a true product. Benzene epoxide is known to be in equilibrium with its tautomeric oxepin form (22). 2,3-Epoxybenzoyl-CoA might undergo an epoxide-oxepin valence tautomerism (Figs. 5E and 6). An oxepin is an unsaturated seven-membered heterocycle having six carbon atoms, one oxygen atom, and three double bonds. This oxepin is not aromatic because it does not obey the Hückel rule. We conclude that benzoyl-CoA is oxidized by BoxAB to an epoxide, which is nonaromatic, thus allowing a nonoxygenolytic ring cleavage catalyzed by the following enzyme BoxC. BoxAB therefore is not a dioxygenase/reductase but an epoxide-forming monooxygenase, an epoxidase.

DISCUSSION
Product of BoxAB-In previous work the product of the oxygenand NADPH-dependent transformation of benzoyl-CoA by BoxAB was tentatively assigned to cis-2,3-dihydro-2,3-dihydroxybenzoyl-CoA, and the enzyme system BoxAB was named benzoyl-CoA, NADPH:oxygen oxidoreductase (2,3-hydroxylating) (11). Although the NMR data were consistent with a cis-diol configuration, they did not provide unequivocal evidence (11). In the conventional metabolism of aromatic compounds, a cis-dihydrodiol undergoes oxidation and rearomatization to a dihydroxy aromatic product, catalyzed by a diol dehydrogenase. This dihydroxylated aromatic central intermediate is then cleaved by ring-cleaving dioxygenases, yet a putative diol dehydrogenase gene could not be found in the benzoate oxidation gene cluster. In addition, the BoxB amino acid sequence shows no similarity to known oxygenase subunits of ring hydroxylating or ringcleaving dioxygenases. This study revealed that the BoxAB enzyme system forms an epoxide and should be renamed benzoyl-CoA, NADPH:oxygen oxidoreductase  (2,3-epoxide-forming). The suggested trivial name is benzoyl-CoA 2,3-epoxidase (EC 1.14.13).
In view of the new findings, the former NMR data were reevaluated ( Table 3). The product that was studied before and in this work had the same UV-visible spectrum and therefore appears to be identical. The true nature of the product may well be 2,3-epoxybenzoyl-CoA (7-oxabicyclo (4.1.0) hepta-2,4diene-2-carboxyl-CoA), as the comparison between data-based and rule-based chemical shifts indicates. Still, these NMR data do not prove the proposed structure. NMR spectroscopy in this instance does not allow discriminating unambiguously between a cis-dihydrodiol and an epoxide. This is because both compounds exhibit cis carbon-oxygen bonds at C2 and C3 of the ring. The tautomeric oxepin has no carbon-carbon bond between C2 and C3; thus it would have different chemical shifts and signal splitting (Table 3), which were not observed. Taking into account the DTC derivative and the spectral properties of the product, we consider the epoxide as the true product and not only as a transient intermediate. BoxB has a counterpart in phenylacetyl-CoA oxygenase, which also forms an epoxide of an aromatic CoA thioester that is converted to an oxepin form. 2 Thus this common unprecedented epoxide formation represents a new paradigm of aerobic aromatic metabolism.
Occurrence of the New Benzoate Pathway and Relation of Benzoyl-CoA Oxygenase (BoxB) to Other Enzymes-A BLAST search (BLASTP 2.2.22ϩ) (23,24) with BoxBC from A. evansii (NCBI accession numbers Q9AIX7 and Q84HH6) revealed that 4 -5% of all fully sequenced eubacterial genomes (mostly ␣and ␤-proteobacteria) harbor the two key genes of the CoA-dependent benzoate oxidation pathway, which is lacking in Archaea (supplemental Table S1). For comparison, 7% of the species harbor benzoate 1,2-dioxygenase benABC and cis-diol dehydrogenase benD genes characteristic for the classical benzoate pathway involving ring-cleaving dioxygenases. Some species contain even both options (1.3%) (supplemental Table S1), which might be required for high turnover of benzoate or if reduced oxygen tension is present, as has been suggested for Burkholderia xenovorans LB400 (6,25). These percentages indicate that the new pathway is not a minor route. The amino acid sequence similarity and identity for BoxB are between 97 and 92% and between 72 and 57%, respectively.
The active center of BoxB is a dinuclear iron center 4 13

C NMR data of the product of BoxAB versus simulation of 13 C NMR data of 2,3-dihydro-2,3-dihydroxybenzoyl-CoA and 2,3-epoxybenzoyl-CoA and its oxepin tautomer
The 13 C NMR spectroscopic features of the product of BoxAB originally were interpreted as indicating a dihydrodiol structure (11). The predicted chemical shifts for 2,3-dihydro-2,3-dihydroxybenzoyl-CoA and 2,3-epoxybenzoyl-CoA and its oxepin were recalculated based on known average data (data based) or on numerical rules (rule based) using SpecInfo. b, broad; d, doublet; dd, doublet of doublets, nd, not determined.
Possible Mechanism of BoxAB and Catalysis of BoxC Revisited-BoxAB introduces one oxygen atom from molecular oxygen into benzoyl-CoA to form 2,3-epoxybenzoyl-CoA. The other oxygen atom is eliminated as water, which fuels the reaction and renders it irreversible. Oxygen becomes activated at the di-iron center, and the enzyme structure and properties of the active site are topics of current studies. 4 BoxC converts the product of BoxAB, 2,3-epoxybenzoyl-CoA, possibly via its oxepin form, to an aldehyde plus formic acid by integration of two water molecules (Fig. 6). BoxC was formerly termed benzoyl-CoA-dihydrodiol lyase and should be renamed 2,3-epoxybenzoyl-CoA dihydrolase. The quick removal of the reactive epoxide by BoxC is probably vital. In the absence of BoxC, BoxAB becomes inactivated in the course of the reaction; in vivo BoxC may even form a complex with BoxB (11).
The crystal structure of BoxC without substrate was solved, and benzoyl-CoA 2,3-dihydrodiol was modeled into the active center (44). This model needs revision in view of the true epoxide substrate. Water addition may be catalyzed consecutively by the same amino acids. Mainly two glutamate residues function as essential concerted acid/base catalysts (7,44). It is obvious from the 18 O 2 labeling experiment that BoxC opens the 18 O-2,3-epoxide by adding OH Ϫ regiospecifically at ring C2 of the epoxide or its oxepin tautomer, leading to a common sevenmembered ring (Fig. 6). This results in 18 O being linked to ring C3, which gives rise to the C6 carbonyl group of 3,4-dehydroadipyl-CoA semialdehyde formed by BoxC, in which 18 O was retained. The electron-withdrawing effect of the CoA activated carboxyl group facilitates ring opening of the epoxide by the addition of OH Ϫ to form a dialdehyde. The enolate anion intermediate is stabilized by an oxyanion hole characteristic for enoyl-CoA hydratases/isomerases (7,44). Protonation at ring C1 prepares the intermediate for the next addition of OH Ϫ at ring C2, which leads to elimination of the C2 atom as formic acid. Finally, a second protonation at ring C1 leads to the product 3,4-dehydroadipyl-CoA semialdehyde (7). The production of formic acid explains the odd finding that genes for enzymes of benzoate and formate metabolism are coinduced (5)(6)(7)45). N-terminal and C-terminal domains of BoxC are characteristic for the enoyl-CoA hydratase/isomerase (crotonase) protein family. This mechanism is consistent with the common mechanism of the enoyl-CoA hydratase/isomerase (crotonase) protein family, which catalyzes a variety of different reactions including enoyl-CoA hydration, enoyl-CoA isomerization, and C-C bond cleavage, all based on abstraction/addition of the ␣-proton of the carboxylic acid and the reversible syn-addition of water to enoyl-CoA-ester (46).
Summarizing the New Pathway-Benzoate appears to be transported into the bacterium by an ABC transporter system and immediately becomes activated by benzoate-CoA ligase forming benzoyl-CoA (Fig. 6) (8 -10). In the following reactions all of the intermediates are coenzyme A thioesters. This may be advantageous because the CoA activated group has an electronwithdrawing effect, which stabilizes negative charge and thus activates the aromatic ring. Furthermore, thioester formation allows an efficient trapping of aromatic acids within the cell, and CoA intermediates, especially the epoxide, might be less toxic. Because the energy-rich thioester bond is retained in the products, the energy initially spent is not lost. Benzoyl-CoA 2,3-epoxidase BoxAB catalyzes the introduction of one oxygen atom to form 2,3-epoxybenzoyl-CoA. The 2,3-epoxybenzoyl-CoA dihydrolase BoxC integrates two water molecules to form the open chain intermediate 3,4-dehydroadipyl-CoA semialdehyde; formic acid is split off. 3,4-Dehydroadipyl-CoA semialdehyde dehydrogenase BoxD oxidizes the semialdehyde to its corresponding acid, 3,4-dehydroadipyl-CoA (4,12). Modified ␤-oxidation leads to ␤-ketoadipyl-CoA, which is finally cleaved into acetyl-CoA and succinyl-CoA by ␤-ketoadipyl-CoA thiolase (8). The overall stoichiometry of aerobic benzoate degradation via CoA ligation follows the equation: Benzoate ϩ ATP ϩ 2 CoA ϩ O 2 ϩ 3 H 2 O ϩ NAD ϩ 3 Acetyl-CoA ϩ Succinyl-CoA ϩ Formic acid ϩ AMP ϩ PP i ϩ NADH ϩ H ϩ . Aerobic benzoate degradation via the ␤-ketoadipate pathway follows the equation: Benzoate ϩ CoA ϩ 2 O 2 ϩ H 2 O 3 Acetyl-CoA ϩ Succinate ϩ CO 2 . Hence, the new pathway uses less oxygen and produces reduced products, NADPH and formic acid.