The 2-Å Crystal Structure of 6-Oxo Camphor Hydrolase

6-Oxo camphor hydrolase (OCH) is an enzyme of the crotonase superfamily that catalyzes carbon-carbon bond cleavage in bicyclic β-diketones via aretro-Claisen reaction (Grogan, G., Roberts, G. A., Bougioukou, D., Turner, N. J., and Flitsch, S. L. (2001)J. Biol. Chem. 276, 12565–12572). The native structure of OCH has been solved at 2.0-Å resolution with selenomethionine multiple wave anomalous dispersion and refined to a final Rfree of 19.0. The structure of OCH consists of a dimer of trimers that resembles the “parent” enzyme of the superfamily, enoyl-CoA hydratase. In contrast to enoyl-CoA hydratase, however, two octahedrally coordinated sodium atoms are found at the 3-fold axis of the hexamer of OCH, and the C-terminal helix of OCH does not form a discrete domain. Models of the substrate, 6-oxo camphor, and a proposed enolate intermediate in the putative active site suggest possible mechanistic roles for Glu-244, Asp-154, His-122, His-45, and His-145.

The microbial metabolism of the bicyclic monoterpene camphor 1 has been a topic of great interest for more than 40 years. Much of this work has been stimulated by the adoption of the hydroxylating enzyme cytochrome P450cam from Pseudomonas putida as a structural and mechanistic model for the activity of cytochromes P450 in general (1). In addition, microbial enzymes from P. putida involved in the degradation of camphor, including cytochrome P450cam, have been exploited as biocatalysts for asymmetric oxidation reactions in synthetic chemistry (2). The metabolism of (1R)-(ϩ)-camphor 1 (Fig. 1) by P. putida proceeds via hydroxylation in the 5-exo position to give hydroxyketone 2, followed by oxidation to diketone 3, which is oxidatively cleaved by an enzymatic Baeyer-Villiger reaction via lactone 4 (3). However, in Rhodococcus sp. NCIMB 9784, an alternative regio-distinct pathway is adopted whereby camphor is hydroxylated in the 6-endo position to give hydroxyketone 5, followed by oxidation to give a symmetrical diketone 6. 6 is not ring-opened oxidatively, but a carboncarbon is cleaved, possibly hydrolytically, by a "␤-diketone hydrolase" enzyme (4) termed 6-oxo camphor hydrolase (OCH), 1 to yield the optically active (2R,4S)-␣-campholinic acid 7 (Fig. 2). Our interests in preparative synthetic biocatalysis led us to isolate OCH and to apply it to the asymmetric cleavage of a number of native substrate analogues (5). In addition, we cloned and sequenced the gene camK, which encodes OCH (6). Analysis of the primary structure of OCH showed it to have significant homology to the crotonase superfamily of enzymes.
The crotonase superfamily of proteins is a group of low sequence homology enzymes that catalyze a wide range of biochemical reactions (7). The parent enzyme of the superfamily, enoyl-coenzyme A hydratase (ECH) catalyzes the stereospecific hydration of enoyl-coenzyme A, an integral step in the ␤-oxidation of fatty acids (8). Other members of the superfamily catalyze dehalogenation (9), carbon-carbon bond cleavage (10), and decarboxylation (11) (Fig. 3). In each case, the substrate for enzymatic reaction is an acyl-coenzyme A (CoA) thioester.
Each member of the crotonase superfamily, for which a mechanism has been proposed, shares mechanistic similarities, related in each case to the stabilization of a transition state enolate by a conserved oxyanion hole (12). Gerlt and others (7,12,15) have described how the crotonase superfamily serves as a paradigm for the study of enzymatic adaptive evolution, whereby fundamental enzymatic activity such as enoyl-CoA hydration is recruited to more atypical metabolic function such as the dehalogenation of chlorobenzoic acids for the derivation of metabolic energy.
The diverse biochemical functions of the crotonase superfamily suggest that, from a genomic perspective, it is very difficult to delineate the role of any particular crotonase homologue in any genome from sequence alone. At this time, individual expression and assay of each gene and gene product would be required to begin to understand its contribution to metabolism. Even then, the mechanistic details of crotonase activity will be difficult to determine. Structural studies of enzymes of the crotonase superfamily are thus vital to the current understanding of the biochemical nature of this group. The structures of four members of the crotonase superfamily have been published to date: ECH (13), dienoyl-coA isomerase (DCI) (14), 4-chlorobenzoyl-coA dehalogenase (15), and methylmalonyl decarboxylase (MMD) (11). In addition, the structure of an enoyl-CoA hydratase homologue from humans has been published recently, which exhibits the ability to bind RNA (16).
In this paper, we present the crystal structure of OCH, into which we have modeled the native substrate, 6-oxo camphor, and a postulated enolate intermediate to (2R,4S)-␣-campholinic acid. Although the hexameric global organization of OCH is similar to enoyl-CoA hydratase, the structure reveals many differences, which suggest a wider spectrum of mechanistic and structural diversity for the crotonase superfamily than previously envisaged.

EXPERIMENTAL PROCEDURES
Chemicals-All chemicals were obtained from Sigma unless otherwise specified. 6-Oxo camphor was obtained as detailed in Ref. 6. Plasmid pET-26b(ϩ) was obtained from Novagen Ltd., Madison, WI. Plasmid pGEMT, restriction enzymes, and buffers were purchased from Promega.
Bacterial Strains, Plasmids, and Culture Conditions-Escherichia coli strains used in this study were BL21(DE3) and B8434(DE3). The plasmid vectors used were pGEMT (Promega) and pET-26b(ϩ) (Novagen). E. coli strains were cultured in Luria-Bertani broth with 30 g ml Ϫ1 kanamycin at 37°C. Cultures were routinely grown to an optical density of A 600 ϭ 0.6 and then induced for expression of OCH by the addition of 1 mM isopropyl-␤-D-thiogalactopyranoside and then grown for a further 3 h.
Cloning and Heterologous Expression-The gene encoding OCH, camK had been sequenced and cloned previously from the chromosomal DNA of Rhodococcus sp. NCIMB 9784 (6). A pGEMT construct consisting of a 4.3-kb EcoR1 restriction fragment from this study was used as the template for a PCR reaction designed to amplify the gene with NdeI and EcoR1 sites engineered at the 5Ј and 3Ј end of the camK gene, respectively. The primers were 5Ј GGAGGAAGTCATATGAAGCAAT-TGGCCACC 3Ј and 5Ј CGGAATTCTCACTGTTCGGACTCCATTCCC 3Ј. camK was amplified in a PCR reaction using conditions outlined previously. The amplified gene was cleaned and ligated into pET26b(ϩ) generating plasmid pGG3. Sequencing of this plasmid confirmed the DNA sequence of the amplified fragment.
Preparation of Selenomethionine Derivative of OCH-The methionine auxotroph E. coli B834 (DE3) transformed with pGG3 was cultured in 4 ϫ 500 ml of a liquid culture medium containing 2 mM MgSO 4 ; 37 mM NH 4 Cl; 44 mM KH 2 PO 4 ; 85 mM NaH 2 PO 4 ; 0.09 mM FeSO 4 ⅐7H 2 O; 22.2 mM glucose; 40 mg each of the 20 common amino acids (except methionine, which was substituted with D/L selenomethionine); 1 mg each of riboflavin, niacinamide, thiamine, and pyridoxine monohydrochloride; 30 g ml Ϫ1 kanamycin. Cultures were grown at 37°C in an orbital shaker at 200 rpm until an optical density of A 600 ϭ 0.4 was reached, at which point, 1 mM isopropyl-␤-D-thiogalactopyranoside was added as inducer. Cells were harvested after 3 h of further incubation at 37°C in an orbital shaker at 200 rpm.
Purification of OCH and Derivative-OCH and its selenomethionine derivative were purified as detailed in Ref. 6.
Crystallization-Crystals of the native and selenomethionyl enzymes were grown by the vapor diffusion hanging drop technique. In each case, a 10 mg/ml protein solution containing 50 mM Tris-HCl, pH 7.0, 1 mM dithiothreitol, 20 M phenylmethylsulfonyl fluoride was mixed in a 50:50 ratio with reservoir to form the hanging drops. For the native enzyme, the reservoir solution consisted of 0.1 M sodium acetate buffer, pH 4.5, 0.2 M ammonium sulfate, and 28% (v/v) polyethylene glycol 4,000 Da (adapted from Hampton screen II). For the selenomethionyl enzyme, the reservoir contained 0.1 M MES, pH 5.5, 0.2 M ammonium sulfate, 37.5% (v/v) 5,000 Da monomethyl ether, and 0.2% (w/v) n-octyl-␤-D-glucopyranoside. Both the native protein and the selenomethionine derivative crystallized in space group P2 1 , with cell dimensions a ϭ 78.95 Å, b ϭ 130.41 Å, c ϭ 81.32 Å, ␤ ϭ 114.16°.
Data Collection and Data Processing-A native data set extending to 2.0 Å was collected on a single crystal of the native enzyme, which had been flash-frozen previously at 120 K in a cryoprotectant solution consisting of 95% (v/v) reservoir and 5% (v/v) ethylene glycol. The data were collected on station 14.2 at the Synchrotron Radiation Facility, Daresbury, United Kingdom, using an ADSC Quantum-4 detector. The data were processed, scaled and merged using the HKL suite (19). Subsequently, a three wavelength multi-anomalous dispersion experiment was conducted on beamline BM14 at the European Synchrotron

FIG. 3. Reactions catalyzed by representative members of the crotonase superfamily.
Radiation Facility (ESRF), Grenoble, France, also using an MAR Research CCD detector. The three data sets were collected from a single crystal of the selenomethioyl enzyme, which was flash-frozen at 120 K using only the reservoir solution as a cryoprotectant. The data sets were processed as above. Data collection and processing statistics are given in Table I.
Structure Solution and Refinement-The three data sets from the MAD experiment were input into SOLVE (20) and scaled and merged internally by this program. Automated Patterson searching with data to 2.4 Å located 16 of the 18 expected Se sites, corresponding to six molecules of 6-oxo camphor hydrolase in the asymmetric unit. These positions were refined and then the phases were calculated using SOLVE. The native data set (extending to 2.0-Å resolution) was then scaled with the MAD data using SCALEIT in the ccp4 suite (21) and used in combination with the new phase to produce electron density maps. These were of good enough quality for model building, which was performed automatically using the warpNtrace option of ARP-WARP (22). The resulting model, consisting of a hexamer of 6-oxo camphor hydrolase, was then refined against the 2.0-Å data using REFMAC in the ccp4 suite and corrected manually using the X-AUTOFIT option in QUANTA (Accelrys Inc., San Diego CA). Solvent molecules were included in the model via the ARP facility in REFMAC. During this process, a randomly selected set of the data (5% of the total) was excluded from refinement for the purpose of R free cross-validation. Refinement statistics and details of the final protein model are given in Table I.
Molecular Modeling-Molecular modeling was used to construct models for two species along the reaction pathway with the CHARMM program and force field (23,24). Models for the substrates were built using QUANTA (Accelrys Inc., San Diego, CA), and the atom types and charges were assigned using standard CHARMM parameter sets (25). All crystallographic water molecules were retained except for those in the putative active site. Of the five water molecules that were located in the active site, two with B values of 32 and 35 Å 2 were removed as they clashed with the modeled substrate. The substrate itself was docked according to information gleaned from known crotonase mechanism and the stereochemical constraints of the reaction. Hydrogen atoms were built using the HBUILT functionality of CHARMM (26). Examination of the energetics of hydrogen bond formation by the His side chains led to a model where two histidines (His-45 and His-122), both of which form part of the putative active site, were protonated. The geometry and energetics of the docked substrates and protein were optimized by carrying out energy minimizations. These were performed with atoms within a 10-Å radius around the substrate subject to decreasing constraints whereas the atoms beyond this were restrained by a harmonic force of 10.0 kcal/mol/Å 2 . Minimizations were carried out using the Steepest Descent and Adopted Basis Newton Raphson algorithms (23) and were continued until the root mean square magnitude of the forces was lower than 10 Ϫ3 kcal/mol/Å.

RESULTS
Quaternary Structure of OCH-The quaternary structure of OCH is superficially similar to the parent enzyme of the superfamily, enoyl-CoA hydratase (13), dienoyl-CoA isomerase (14), and methylmalonyl decarboxylase (11), being a homogeneous hexamer composed of two stacked trimers (Fig. 4). Homology between the six monomers of the OCH hexamer is high (root mean square deviation ϭ 0.15 to 0.22 Å calculated on C␣). The hexamer is a flattened sphere with approximate overall dimensions of 80 ϫ 76 ϫ 71 Å. An octahedrally coordinated Na ϩ ion is present at the 3-fold axis of each trimer, coordinated via water molecules to residues Arg-168 and Asp-186 from each constituent monomer (Fig. 5). The conclusion that this coordinating species is a sodium ion rather than a water molecule or isoelectronic Mg 2ϩ was made because of the very distinctive octahedral coordination geometry and metal/water distances of 2.4 -2.5 Å, both of which correlate with sodium ion binding (17,18). Trimer stabilization appears to be mediated by a mixture of polar and hydrophobic interactions between ␣-5 (residues 167-174) and ␣-8 Monomer Structure-The OCH monomer consists of nine ␣-helices, ranging in length from four to 24 residues, and five well defined ␤-strands. The secondary structural elements, in common with other crotonase superfamily members, form four turns of a ␤␣␤ superhelix, resulting in an inverted prism whereof the top layer of twisted ␤-sheet is formed from the five well defined ␤-strands, the most N-terminal being antiparallel, the remainder parallel. One face of the inverted prism consists of a layer formed from five near-flat coil regions, which exhibit classical ␤-sheet bonding only in the central region of the sheet. Of these strands, all are parallel except the most C-terminal, which is antiparallel. The other face of the prism is formed by three ␣-helices, ␣-2, (residues 42-57), ␣-3 (residues 87-106), and ␣-8 (residues 214 -226), the last being the N-terminal region of the large, kinked C-terminal helix (␣-8/␣-9), the distal region of which (␣-9) loops around the core monomer structure. This is in contrast with the constituent monomers of other hexameric members of the crotonase superfamily, wherein the C-terminal helix of e.g. ECH forms a discrete, second domain (Fig. 6). ␣-9 is also involved in hexamer formation, making contact with one monomer of the adjacent trimer. Active Site Geometry-Although the location of the OCH active site has yet to be established from a ligand-bound structural complex, we have identified a hydrophobic pocket near the center of the inverted prism apex, which bears many characteristics suggestive of a putative active site. This pocket is wholly enclosed within one monomer, in common with MMD (11), but in contrast to ECH, DCI, and CBD, for which the active sites are observed at the trimer interface. Entrance to the pocket from the solvent appears to be gated by Phe-79. The pocket is bound on one side by helices ␣-2, ␣-3, and ␣-9 and is largely hydrophobic, but the protrusion of five polar residues, Glu-244, Asp-154, His-145, His-122, and His-45, into the cavity is noteworthy.
The distinctive (G/A)G(G/A) turn motif in the majority of other crotonase sequences (7) e.g. ECH (residues 140 -142) is absent in OCH, being replaced by an Asn-121 HP motif (residues 121-123). His-122 adopts unfavorable and angles of ϩ42.8 and ϩ58.1°. The dihedral angles for most other amino acid residues lie within permitted regions of the Ramachandran plot. The other notable exception is His-145, which adopts and angles of Ϫ97.6 and Ϫ105.1°, respectively.
Modeling-We have so far been unable to obtain a crystal structure of OCH bound with either the natural substrate, 6-oxo camphor (this would be unlikely as turnover is very rapid), or the product, (2R,4S)-␣-campholinic acid. However, the location of a putative active site, in conjunction with known crotonase mechanisms and the stereochemical constraints of the reaction, have allowed us to model both substrate and putative intermediate into the proposed active site region (Fig. 7).
We have suggested previously (5) that cleavage of the carbon-carbon bond may proceed via nucleophilic attack at one prochiral carbonyl in 6-oxo camphor by a water molecule activated by an acidic residue in the enzyme active site, analogous to the hydrolyase mechanism of water addition observed in ECH (27). Such attack would have to occur from the left hand side of the molecule as drawn in Fig. 8 (pro-S) to result in the observed product stereochemistry (S) at the acetate-substituted center. When the substrate is modeled into the active site (Fig. 7, top), the geminal dimethyl group of the substrate is shown to be oriented toward a hydrophobic region of the active site making interactions with dominant hydrophobic residues such as Phe-82 (5.1 Å), Ile-93 (4.0 Å), Trp-90 (3.9 Å), and Leu-84 (3.8 Å). In this orientation, either the prominent Glu-244 or Asp-154 residue is well placed for the activation of a water molecule for attack at the (pro-S) carbonyl side of the substrate. The modeled orientation also allows for the (pro-R) carbonyl of the substrate to be stabilized by hydrogen bonding to His-122. Ring opening of the substrate may proceed via the enolate intermediate 8 shown in Fig. 8. In other crotonase homologues, intermediate enolates, thought too active to be kinetically competent (7), are thought to be stabilized by an oxyanion hole formed between two peptidic N-H groups from residue positions that are conserved throughout the previously determined superfamily structures (e.g. Ala-98 and Gly-141 in ECH). In OCH, the replacement of the GG 141 G motif with the N 121 HP motif suggests that this oxyanion hole is absent or constructed differently. Any proposed enolate must be stabilized by some other factor. Modeling of the enolate intermediate (Fig. 7, bottom) suggests that the enolate may be H-bonded to His-45 (2.6 Å), the carboxylate moiety of the intermediate being H-bonded both to His-122 (2.7 Å) and His-145 (2.9 Å), which are both observed to be held in unusual conformations (see above).

DISCUSSION
The structure of 6-oxo camphor hydrolase suggests that OCH is more removed from a possible progenitor of the crotonase superfamily of proteins than members described previously. Although the hexameric global structure is superficially similar to ECH, for example, the presence of sodium ions at the 3-fold axis of each constituent trimer of OCH was not observed in ECH, even though a similar coordination environment was observed in the latter protein. A nickel ion was reported to be coordinated by equivalent histidines along the 3-fold trimer axis of MMD, but this was thought to be an artifact of nickel/ His tag-dependent purification (11). In the case of OCH, sodium may arise from the purification or crystallization buffer. There does not appear to be an absolute dependence on any metal ion for OCH activity as evidenced by previous studies, however. Indeed, an increase in NaCl concentration was not observed to affect enzymatic activity (6).
Significant divergence from other members of the crotonase superfamily is also observed in the structure of the constituent monomer of OCH. Comparison of residues 13-192 of the ECH monomer and residues 31-210 of OCH (Fig. 6) reveals that the N-terminal catalytic domain of ECH is, for the most part, structurally conserved in OCH (root mean square deviation value of 4.82 Å). The C-terminal helix is positioned very differently, however, not forming a discrete domain as in ECH, but rather looping around the monomer structure of OCH to form the wall of a hydrophobic pocket, which may constitute the active site. The active site of OCH appears to have been recruited from a hydrophobic well bound by helices in the interior of the monomer, which already exists partially in ECH. In the case of OCH, this active site has been completed by recruitment of the C-terminal trimerization domain helix as a further wall of the active site seemingly at no cost to trimer stabilization, which is satisfied by alternative regions in OCH than ECH. It may be that trimer interactions centered around a sodium ion have evolved as a mechanism of stabilization to offset any loss of stabilization as a result of the recruitment of the C-terminal helix in ECH. It is also possible that the recruited C-terminal helix is positioned to provide the primary catalytic residue (Glu-244) for water activation leading to carbon-carbon bond cleavage. The C-terminal divergence among crotonase homologs is reflected additionally in the sequence alignment of a series of members of the crotonase superfamily ( Fig. 9), illustrating comparatively poor conservation of sequence in this region.
The sequence alignments of selected regions of OCH, ECH, MMD, CBD, and DCI ( Fig. 9) are also revealing in respect to conserved catalytic residues throughout crotonase homologs. In conjunction with sequence comparison and structural studies, site-directed mutagenesis experiments have revealed that although catalytic residues may be conserved between superfamily members, they may make no catalytic contribution. This appears to be the case with Glu-124 in OCH (equivalent to catalytic Glu-44 in ECH), which is Hbonded to His-122 and therefore presumably cannot have a primary catalytic role. Asp-154 (OCH), which protrudes into the putative active site, is near homologous to the catalytically active Asp-145 in CBD. The presence of Glu-244 in the putative active site is particularly noteworthy, as such an extremely C-terminal residue in the active site of a crotonase homolog would be unprecedented, this area usually being involved in trimerization.
Another striking divergence of the primary structure of OCH from that of other crotonase homologs is the absence of the (G/A)G(G/A) motif, which, in other members, provides the conserved glycine for formation of the oxyanion hole for enolate stabilization. In OCH, (G/A)G(G/A) appears to be replaced by an N 121 HP motif, His-122 being held in an unusual conformation, perhaps suggestive of a mechanistic role for this residue. In other homologues, the oxyanion hole is formed by peptidic backbone N-H groups (such as Ala-98 and Gly-141 in ECH), but equivalent residues in OCH, Ile-77, and His-122 are not suitably positioned for this role spatially.
A proposed mechanism must account for proton donation to the intermediate enolate. We have already shown that, as the abiotic acid-catalyzed ring opening of 5 proceeds with the same diastereoselectivity as the enzyme catalyzed reaction (5), that the proton donated to the intermediate is not necessarily enzyme-derived. Indeed, Bahnson et al. (28) have shown recently that, for enoyl-CoA hydratase, both hydroxy group and proton in the product of enoyl-CoA hydration are derived from the same water molecule. It is possible that a similar concerted mechanism is responsible for the formal hydration of the 6-oxo camphor molecule. If this were the case a more accurate designation of 6-oxo camphor-hydrolyase might be appropriate to OCH.
It is notable that among the range of activities that have been attributed to the crotonase superfamily, the activity of OCH is unusual in that the native substrate is not an acyl-CoA thioester. It is difficult at this stage, however, to assess whether the ability of OCH to bind coenzyme A has been lost, as although residues have been implicated for coenzyme A binding by other crotonase homologs (e.g. Lys-92 and Lys-101 in ECH (13)) these are characteristic of individual homologs, a general motif for coenzyme A binding not having been identified in this superfamily. The addition of 1 mM coenzyme A to OCH assay has no detectable effect on activity, however (results not shown), suggesting no competitive binding of coenzyme A over the native substrate, 6-oxo camphor.
The range of enzymatic processes catalyzed by crotonases identifies them both as a source of many new activities with potential biotechnological application and, additionally, as a range of metabolic activities for possible therapeutic targeting. However, the function of crotonase sequence homologs is difficult to determine without their isolation and assay. The structure of OCH reveals significant structural divergence from other superfamily members, illustrated by the recruitment of the C-terminal helix for active site formation and the possible provision of catalytic residues. Structural differences such as this, which hinge on short coil regions, are difficult to predict. Alternative substrate binding modes that do not require either substrate ligation to coenzyme A or oxyanion hole stabilization of enolate intermediates are also implied by the OCH structure. This may point to a wider range of undiscovered enzyme activities recruited from the crotonase fold than previously envisaged with further implications for the more cautious interpretation of enzyme function from sequence primary sequence data alone in functional/structural genomics projects.