The Tautomeric Half-reaction of BphD, a C-C Bond Hydrolase

BphD of Burkholderia xenovorans LB400 catalyzes an unusual C-C bond hydrolysis of 2-hydroxy-6-oxo-6-phenylhexa-2,4-dienoic acid (HOPDA) to afford benzoic acid and 2-hydroxy-2,4-pentadienoic acid (HPD). An enol-keto tautomerization has been proposed to precede hydrolysis via a gem-diol intermediate. The role of the canonical catalytic triad (Ser-112, His-265, Asp-237) in mediating these two half-reactions remains unclear. We previously reported that the BphD-catalyzed hydrolysis of HOPDA (λmax is 434 nm for the free enolate) proceeds via an unidentified intermediate with a red-shifted absorption spectrum (λmax is 492 nm) (Horsman, G. P., Ke, J., Dai, S., Seah, S. Y. K., Bolin, J. T., and Eltis, L. D. (2006) Biochemistry 45, 11071-11086). Here we demonstrate that the S112A variant generates and traps a similar intermediate (λmax is 506 nm) with a similar rate, 1/τ ∼ 500 s-1. The crystal structure of the S112A:HOPDA complex at 1.8-Å resolution identified this intermediate as the keto tautomer, (E)-2,6-dioxo-6-phenyl-hex-3-enoate. This keto tautomer did not accumulate in either the H265A or the S112A/H265A double variants, indicating that His-265 catalyzes tautomerization. Consistent with this role, the wild type and S112A enzymes catalyzed tautomerization of the product HPD, whereas H265A variants did not. This study thus identifies a keto intermediate, and demonstrates that the catalytic triad histidine catalyzes the tautomerization half-reaction, expanding the role of this residue from its purely hydrolytic function in other serine hydrolases. Finally, the S112A:HOPDA crystal structure is more consistent with hydrolysis occurring via an acyl-enzyme intermediate than a gem-diol intermediate as solvent molecules have poor access to C6, and the closest ordered water is 7Å away.

Bacteria use meta-cleavage pathways to degrade a large variety of aromatic compounds, and some alicyclic compounds, such as steroids (1). Although each pathway has its own substrate specificity, they all employ the same underlying logic: vicinal dihydroxylation of an aromatic ring enables dioxygenase-catalyzed extradiol (or meta) ring opening. The resulting meta-cleavage product (MCP) 3 is degraded by an MCP hydrolase, which adds water across a carbon-carbon bond to generate a dienoate and a carboxylic acid (Fig. 1). Enzymes of the metacleavage pathway have been of interest because of their roles in biodegradation. For instance, the biphenyl (Bph) catabolic pathway transforms a number of polychlorinated biphenyls (PCBs). This process is limited by the inhibitory effects of PCBs or their chlorinated metabolites (2,3). More recently, it was discovered that Mycobacterium tuberculosis catabolizes cholesterol via a meta-cleavage pathway (4) that is essential for pathogen survival in the macrophage (5). A better understanding of the meta-cleavage pathway enzymes should accelerate the development of their potential for biocatalysis and biodegradation, as well as facilitate the design of novel therapeutics for the treatment of tuberculosis. MCP hydrolases, exemplified by BphD of the Bph pathway and HsaD of the cholesterol catabolic pathway, contain the canonical structural fold and Ser-His-Asp catalytic triad characteristic of the ␣/␤-hydrolase enzyme superfamily (6 -9). This triad is associated with the classical hydrolytic mechanism typified by the serine proteases, in which the Asp-His dyad first activates Ser to nucleophilic attack at the substrate carbonyl, and subsequently activates water to release the resulting acylenzyme covalent intermediate. Although the catalytic Ser and Asp residues vary within the nucleophile-His-acid framework, all known ␣/␤-hydrolase active sites contain the catalytic His and the oxyanion hole. The latter, which stabilizes the negative charge of the tetrahedral intermediate oxyanion, is created by the partial positive charges of backbone amide protons.
The proposed mechanism of MCP hydrolases differs from that of other ␣/␤-hydrolases in two important respects: (i) it invokes an enol-keto tautomerization prior to hydrolysis ( Fig.  1) (10,11) and (ii) it asserts hydrolysis via a gem-diol intermediate rather than an acyl-enzyme (12)(13)(14)(15)(16). The two best studied MCP hydrolases are BphD from Burkholderia xenovorans LB400, which hydrolyzes 2-hydroxy-6-oxo-6-phenylhexa-2,4dienoic acid (HOPDA) to benzoic acid and 2-hydroxy-2,4-pentadienoic acid (HPD), and MhpC involved in phenylpropionic acid degradation. Kinetic and structural studies of both enzymes have indicated that the catalytic His may mediate tautomerization, as well as help catalyze hydrolysis (13,17,18). For instance, stopped-flow studies showed that substitution of the catalytic histidine decreased the rate of a process proposed to represent tautomerization in MhpC (18) and a polyhistidinetagged BphD (Ht-BphD) (13). In the crystal structure of an MhpC:substrate analog binary complex, the catalytic His was 3.2 Å from the C2 carbonyl and 3.9 Å from C5 of the analog (17), consistent with histidine's role in abstracting a proton from the C2 hydroxyl and protonating C5. More convincingly, the crystal structure of the S112C variant of BphD incubated with HOPDA revealed a complex with the product HPD and an interaction between the catalytic His-265 and the 2-hydroxy/ oxo substituent of the dienoate, consistent with a general base role in substrate tautomerization (19). Protonation by His-265 at HOPDA C5 was deemed possible because of the observed conformational flexibility of the former, allowing us to predict a His-265-to-C5 distance as short as 2.9 Å.
A complicating factor in evaluating the catalytic role of His-265 is that the proposed keto intermediate has not been directly observed. Using single turnover ([E] Ͼ [S]) stopped-flow analysis, we recently identified an intermediate possessing an electronic absorption maximum that is red-shifted ( max is 492 nm) from that of the free HOPDA enolate ( max is 434 nm) (19). Two kinetic models were proposed, differing in their assignment of the red-shifted intermediate (E:S Red ) to either the enolate (E:S e ) or keto (E:S k ) tautomer of enzyme-bound HOPDA. A priori, it is more logical to assign the red-shifted species to E:S e as ketonization disrupts the conjugated system of the HOPDA enolate. However, electronic transition energies of small molecule ligands may be greatly perturbed in a protein environment (20 -22) and assignment of E:S Red to an E:S k species is more consistent with additional kinetic observations. First, under single turnover conditions, the decay of E:S Red was coupled to HPD formation, consistent with direct transformation of E:S k to HPD. Second, increased solvent viscosity slowed a relaxation assigned to tautomerization in the E:S e model, and to diffusive HPD release in the E:S k model. Because tautomerization is intramolecular and hence should not be affected by solvent viscosity, the E:S k model provides a more reasonable interpretation of the data. Finally, the apparently ordered product release could be more readily explained via the E:S k model. Identifying the E:S Red intermediate observed in BphD, therefore, represents an important step toward elucidating the mechanism of MCP hydrolases.
Herein, we describe kinetic and structural studies of alanine substitutions of the active site catalytic triad residues Ser-112 and His-265 of BphD. Stopped-flow spectrophotometry of the variants revealed the catalytic contribution of each residue based on the accumulation of different intermediates, such as E:S Red . In parallel, substrate complexes of the BphD variants were characterized at high resolution by x-ray crystallography. Monitoring the ability of enzyme variants to catalyze tautomerization of HPD provided further insight into the catalytic roles of the substituted residues in the tautomeric half-reaction. Implications for catalysis in MCP hydrolases are discussed.

MATERIALS AND METHODS
Chemicals-HOPDA was enzymatically generated from 2,3dihydroxybiphenyl (DHB) using DHB dioxygenase (DHBD) as previously described (19). The preparation of DHB has been described (23). HPD was generated together with benzoic acid by BphD LB400 -catalyzed hydrolysis of HOPDA. Upon reaction completion (monitored by absorbance at 434 nm), the solution was acidified to pH ϳ 3 with 2 N HCl, extracted 3 times with 0.3 volumes of ethyl acetate, dried over anhydrous MgSO 4 and rotary evaporated to dryness. All other chemicals were of analytical grade.
Mutagenesis, Protein Expression, and Purification-BphD from B. xenovorans LB400 was produced and purified as previously described (19,24). Ser-112 of BphD was substituted with alanine (S112A) using the Transformer site-directed mutagenesis method (Clontech Laboratories, Palo Alto, CA). Briefly, the S112A mutagenic primer (primer S112A: 5Ј-GCGC-CCCCCATGGCGTTGCCGACCAG-3Ј) and a second mutagenic selection primer to remove an EcoRI site (primer ODM: 5Ј-AGCTCGAATTGGTAATCATGG-3Ј) were mixed with pSS184, the pEMBL18 vector carrying bphD (24). After second strand synthesis and ligation using T4 DNA polymerase and T4 DNA ligase (GE Healthcare, Uppsala, Sweden), respectively, the DNA was digested with EcoRI to linearize wild-type plasmid, and transformed into Escherichia coli BMH 71-18 mutS. Plasmid isolated from this first transformation was subjected to a second round of EcoRI digestion and transformation, from which pSS184SA, carrying the mutated gene, was isolated. The mutated gene was cloned into pVLT31 using XbaI and HindIII restriction sites, and the resulting plasmid, pSS314SA, was used for protein production. Substitution of His-265 with alanine was performed using a 5Ј-phosphorylated primer (H265A: CTCCAAGTGCGGCGCTTGGGCGCAATGG-3Ј) and the QuikChange multi site-directed mutagenesis kit (Stratagene, La Jolla, CA). Genes encoding the single (H265A) and double (S112A/H265A) variants were generated using pSS184 and pSS184SA, respectively, yielding pSS184HA and pSS184SAHA. These constructs were used directly for protein production. The nucleotide sequences of variants were confirmed using an
Enzyme Activity Measurements-All enzyme kinetic experiments were performed using potassium phosphate buffer, I ϭ 0.1 M, pH 7.5, at 25°C. Steady-state enzyme activities were obtained by monitoring the decrease in absorbance at 434 nm of the HOPDA enolate versus time using a Varian Cary 5000 spectrophotometer (Varian Canada, Mississauga, ON, Canada). The latter was equipped with a thermostatted cuvette holder maintained at 25.0 Ϯ 0.5°C, controlled by Cary WinUV software version 3.00. To measure enzyme activity toward HPD, a small volume (Ͻ0.5% v/v) of HPD/benzoic acid in ethanol was added to a buffered solution, and absorbance at 270 nm was monitored before and after addition of enzyme.
The half-life of the S112A:HOPDA complex was determined by mixing 25 M S112A and 5 M HOPDA, and recording the absorption spectrum at 0.5-1-h intervals. The absorbance at 506 nm was plotted against time, and described by a first order decay using Excel (Microsoft, Redmond, WA).
Stopped-flow Spectrophotometry-Experiments were conducted using an SX.18MV stopped-flow reaction analyzer (Applied Photophysics Ltd., Leatherhead, UK) equipped with a photodiode array detector. The drive syringe chamber and optical cell were maintained at 25°C by a circulating water bath. Multiple wavelength data from the time courses of single turnover experiments were acquired using the Xscan software (Applied Photophysics Ltd.) and exported to Excel where replicate measurements from at least three shots were averaged. To obtain more data at earlier time points, a single wavelength was monitored using the SX18MV software. The data from at least three shots were averaged and equations for single or double exponentials were fit using the same software to obtain reciprocal relaxation times and amplitudes. Good fits were characterized by random variation in the fit residuals.
Crystallization of BphD and Preparation of Substrate Complexes-A 1.6-Å resolution structure of BphD was previously determined using crystals grown from a solution containing 1.6 M ammonium sulfate (19,24). These crystals had space group P6 4 with a ϭ 135.0 Å, c ϭ 66.7 Å, and a sulfate ion bound in the active site. The failure of many attempts to prepare crystalline complexes by incubation of crystals with substrates suggested sulfate binding or the crystal form restricts formation of E:S complexes leading us to seek alternative crystallization procedures.
New protocols were established for both wild type and variant enzymes using both commercial (Hampton Research) and non-commercial sodium malonate grid screens and the vapor diffusion method. Sitting drops (1-l each of reservoir and protein solutions) were equilibrated at 20°C against 500-or 1000-l reservoirs. The protein sample contained 9 -28 mg/ml protein in 20 mM HEPES pH 7.5. Crystals or crystalline precipitates were obtained between 1.5 and 2.4 M sodium malonate within a pH range of 6.0 and 7.0.
Two crystal forms were obtained. Crystals of wild-type enzyme had space group P6 4 , a ϭ 135.2 Å, c ϭ 66.3 Å, whereas crystals of the S112A and the S112A/H265A variants had space group I4 1 22 with a ϭ 117.3 Å, c ϭ 87.3 Å. The hexagonal crystals are essentially isomorphous with the previously reported form obtained from ammonium sulfate and contain two protein monomers per asymmetric unit; the tetragonal form has one monomer in the asymmetric unit. Crystals of wild-type BphD grew over 12 weeks as hexagonal prisms in 2.0 M sodium malonate, pH 6.0 or 6.5. Crystals of S112A and S112A/H265A grew in 4 -6 days as tetragonal rods in 1.9 M sodium malonate. The best crystals of the S112A variant were obtained at pH 7.0, whereas the best S112A/H265A crystals grew at pH 6.5.
Crystals of HOPDA complexes were obtained by incubating crystals grown in the absence of HODPA in 30 -60 l of reservoir solution augmented with ϳ15 mM HOPDA for 30 -60 min at 20°C.
Diffraction Data Measurements and Processing-Crystals were prepared for flash-freezing by sequential transfer into solutions containing higher concentrations of sodium malonate. A mounting loop was used to transfer crystals from the growth drop into 60-l volumes of reservoir solution, then into similar solutions containing 3.4 M and, finally, 3.7 M sodium malonate. The pH was held at the growth value; the incubation time was 3-6 s per step. After the last transfer crystals were flash-frozen by immersion into liquid nitrogen. For enzymesubstrate complexes, each solution was supplemented with HODPA (ϳ5-10 mM).
Preliminary diffraction patterns were acquired with a typical laboratory instrument based on a rotating anode (Cu) generator equipped with focusing mirror or multilayer optics and an imaging plate detector (Rigaku/MSC). The diffraction data used for refinements of atomic models were acquired at SER-CAT beamline 22-ID-D at the Advanced Photon Source, Argonne National Laboratory. For the latter experiments, crystals were maintained at ϳ100 K, and diffraction images were recorded by a MarMosaic 300 CCD detector (Mar USA, Inc., Evanston, IL). For each crystal, ϳ100 frames were collected with a 1°rotation per frame; exposure times were 1-10 s per degree. All images were processed using DENZO, and then intensities were merged and scaled using SCALEPACK; both programs were from the HKL2000 program suite (25).
Structure Determination and Refinement-Programs from the CCP4 suite (26) were used for phasing and refinement. The crystal structure of wild-type BphD (PDB code 2OG1, Ref. 19,24) served as a search model for phasing by molecular replacement using the program MOLREP (27). Rigid body refinement was followed by iterative cycles of restrained atomic parameter refinement using REFMAC (28) and manual density fitting using the molecular graphics program O (29). PRODRG (30) was used to develop structures of substrates and malonate for density fitting and establishment of refinement restraints. The deviation of restrained torsion angles from their expected values was used to evaluate the compatibility of the x-ray data with different tautomers of HOPDA. In the final refinement cycles of HOPDA complexes, the torsion angles in the non-aromatic portion of HOPDA were not restrained; bond lengths and bond angles were restrained to values expected for the keto (S112A: HOPDA) or enol (S112A/H265A:HOPDA) forms. The stereochemical properties of the models and the hydrogen bonding were analyzed by programs PROCHECK (31) and REDUCE (32).
Optical Spectroscopy with Single Crystals-Visible light absorption spectroscopy was performed using crystals of the S112A and S112A/H265A variants exposed to HOPDA (ϳ15 mM in 30 -60 l of reservoir solution) for 30 min. The spectra were recorded at room temperature using a 4DX single crystal microspectrophotometer (4DX Systems) equipped with a MS125 TM 1/8m spectrograph (Thermo Oriel), DB401 CCD detector (Andor Technologies), and a CLX 500 xenon lamp (Zeiss). Spectra were generated from the integration of one hundred 19-ms exposures recorded over the wavelength range 300 -800 nm. Data acquisition and analysis was controlled by the Andor MCD software package supplied with the detector.
After recording the spectra, the crystals were flash-frozen and diffraction data were measured at 100 K using the Rigaku/ MSC equipment described above. The data were processed and electronic density maps were analyzed as described above.

RESULTS
Kinetic Analysis of Variant Enzymes-Catalytic triad residues His-265 and Ser-112 were substituted to construct three variants: S112A, H265A, and S112A/H265A. The extremely low activity of the S112A and S112A/H265A variants prevented steady state kinetic measurements. Transformation of HOPDA by H265A as measured by the decay of absorbance at 434 nm could only be detected using large quantities of enzyme (ϳ1 M), and the progress curve was biphasic. In an experiment in which 4 M HOPDA was mixed with 1.3 M H265A, the first phase could be described by a single exponential decay with a rate constant of 5.8 (Ϯ0.4) ϫ 10 Ϫ3 s Ϫ1 . In Table 1 this is presented as the third rate of decay, or reciprocal relaxation time (1/ 3 ), because it is preceded by two events observed by stopped-flow spectrophotometry (see below). The value of 1/ 3 is similar to that of HOPDA tautomerization in solution, as observed by deuterium exchange NMR (19). The slope of the linear second phase approximately doubled upon increasing the H265A concentration to 2.6 M. By contrast, doubling enzyme concentration did not affect the value of 1/ 3 . The apparent burst is consistent with first order decay of an E:S complex followed by steady-state turnover. Curiously, the amplitude of the first phase corresponded to only 9 (Ϯ3) % of the total enzyme added, suggesting only this fraction of the active sites were functional. Similar behavior was observed in stopped-flow experiments (see below). Correcting for the active fraction of enzyme provides a rate of 0.0009 (Ϯ0.0002) s Ϫ1 for the steady state phase, which is about half of the k cat measured for the H265A variant of Ht-BphD (13). This is in reasonable agreement considering the current experiments were performed using a substrate concentration below the K m of 37 M measured for Ht-BphD.
To better characterize the catalytic impairment of the variants, stopped-flow spectrophotometry was employed under single turnover conditions (E ϭ 8 M, S ϭ 4 M) at 25°C. The S112A variant rapidly generated an intermediate with similar kinetics as the E:S Red intermediate transiently observed in wild type BphD (1/ 1 ϳ 500 s Ϫ1 , Ref. 19). The spectrum of E:S Red was more red-shifted and more intense in S112A ( max ϭ 506 nm; Table 1, Fig. 2A) than in the wild type ( max ϭ 492 nm, Fig. 2C). Moreover, E:S Red decayed extremely slowly in S112A, and was effectively trapped as an orange-colored complex. The half-life of this complex (S112A ϭ 25 M, HOPDA ϭ 5 M), determined by monitoring its absorbance at 506 nm, was 4.4 h at 25°C.
Neither H265A nor S112A/H265A accumulated E:S Red , instead yielding a species possessing a more intense, slightly blue-shifted spectrum with respect to the free HOPDA enolate ( max is 432 nm; Fig. 2B). This species was provisionally identified as the fully deprotonated enolate because, in S112A/ H265A, its molar absorptivity matched that of fully deprotonated HOPDA in solution. 4 In contrast, H265A only partially deprotonated HOPDA under these conditions. Indeed, the amplitude of the absorbance increase at 434 nm was only 25% of the same signal in the experiment using S112A/H265A. Increasing the concentration of H265A to 32 M (4 M HOPDA) resulted in the same spectrum as observed in S112A/ H265A, suggesting that only ϳ12% of the active sites in H265A supported deprotonation, as observed in the steady-state experiments described above. This phenomenon was observed in two different preparations of H265A. Attempts to rescue E:S Red formation in the double variant with imidazole (1-5 mM) were unsuccessful. In summary, E:S Red formation is His-265dependent, and its decay requires Ser-112.
Reciprocal relaxation times (1/) and amplitudes for each phase observed in the single turnover stopped-flow experiments are summarized in Table 1. The first reciprocal relaxation time (1/ 1 ), corresponding to an increase in absorbance at 434 nm and assigned to HOPDA enolate formation, was 220 Ϯ 40 s Ϫ1 in S112A/H265A. In H265A, this same relaxation was 78 Ϯ 15 s Ϫ1 . This deprotonation step in both H265A variants was significantly slower than E:S Red formation in wild type and S112A (1/ 1 ϳ 500 s Ϫ1 ). In all three variants, slower relaxations of smaller amplitude followed the initial, relatively rapid relaxation ( Table 1).
Tautomerization of HPD by Variant Enzymes-To further probe the ability of the BphD variants to catalyze the tautomerization half-reaction, we investigated their respective abilities to catalyze the tautomerization of HPD to (E)-2-oxo-3-pentenoate (19). The tautomerization of HPD is observed as decay in absorbance at 270 nm and occurs non-enzymatically in aqueous solution at a slow rate. Wild-type and S112A catalyzed the tautomerization of HPD (ϳ14 M) with specific activities of Stopped-flow data at 434 nm are reported as reciprocal relaxation times for each of the 3 phases (1/ n ), except H265A, in which only the first two phases were studied by stopped-flow. Values in parentheses for stopped-flow data represent the percent of the total absorbance change (amplitude) in either the positive (absorbance increase) or negative (absorbance decrease) direction. Errors are no greater than 20%. a Measured using a Cary 5000 spectrophotometer as an exponential decay in absorbance at 434 nm.

Tautomerization in a C-C Bond Hydrolase
0.47 Ϯ 0.01 units/mg and 0.082 Ϯ 0.013 units/mg respectively. The slower S112A reaction was easily detectable (0.0067 mol/ min) over background HPD decay (0.0011 mol/min). By contrast, neither H265A nor S112A/H265A detectably catalyzed this tautomerization. Thus, His-265 is necessary for catalyzing tautomerization of HPD. Primary Crystallographic Results-Crystal structures of BphD, its S112A and S112A/H265A variants, and complexes of the variants with HOPDA were determined and refined at resolutions between 1.57 and 2.30 Å. Determination of high reso-lution structures for HOPDA complexes of the variants was facilitated by a new tetragonal crystal form (see "Materials and Methods"). Table 2 summarizes the properties of the crystals and the diffraction data, which are generally of high quality. Table 3 characterizes the refined models. The model for wild type BphD lacks the N-terminal methionyl residue in chain A (as expected from mass spectral analysis) and three N-terminal residues in chain B, whereas the models for the variants lacked the three N-terminal residues in all cases. Each model includes several residues in multiple conformations. For all models, Ͼ98% of the residues are in favored or allowed regions of Ramachandran plots as defined by PROCHECK. For the outliers, which include the key active site residue S/A112, the backbone conformations are stabilized by hydrogen bonds and are similar in all independent coordinate sets.
Tertiary and Quaternary Structure and Crystal Packing-The monomer of BphD includes two domains: a core domain with the ␣/␤-hydrolase fold (residues 2-145 and 213-286) and a lid domain (residues 146 -212) occurring as an insertion in the core domain. The active site is located at the cleft between the two domains (Fig. 3).
BphD and its variants are tetramers in solution, and the crystal structures show the tetramer has 222 (D 2 ) point group symmetry. The 222 symmetry is crystallographic in the I4 1 22 crystals; thus, all monomers are equivalent. In the P6 4 crystals, the 222 symmetry is produced by a combination of crystallographic and non-crystallographic operations, and the asymmetric unit includes an A:B dimer. There is a significant difference in the environment of A and B (19), such that the average B factors are 34 and 65 Å 2 , for A and B, respectively. For these crystals we describe structural details only for molecule A because it is represented by better electron density.
Binding of Malonate: a Partial Mimic of the Substrate-The substrate binding site of BphD is conventionally partitioned in two subsites (33): the P subsite complements the substrate three polar groups, and the NP subsite binds the non-polar phenyl substituent. The active site Ser-112 and the oxyanion hole formed by the backbone NHs of Gly-42 and Met-113, are at the boundary between the P and NP subsites; the guanidinium group of Arg-190 lies at the opposite end of the P subsite.
Malonate (propanedioic acid) binds slightly differently in each of the BphD variants, but in each case mimics the binding of one or more of the HOPDA three polar groups. In wild-type BphD, the C1 carboxylate occupies a site near Arg-190 that is utilized by the HOPDA carboxylate in the substrate complex (see below). One oxygen atom from the malonate C3 carboxylate occupies the site of the HOPDA 2-oxo group near His-265, whereas the second oxygen atom hydrogen bonds with O␥ of Ser-112 and the peptide NH of Gly-42 near the oxyanion hole. The binding mode is similar in the S112A/H265A variant, although the deletion of the imidazole group from residue 265 results in minor adjustments in the position and orientation of the C3 carboxylate.
In the S112A variant, the malonate is ϳ2.5 Å further away from Arg-190 and closer to the oxyanion hole. One oxygen atom of the C3 carboxylate lies in the oxyanion hole and forms hydrogen bonds with the peptide NHs of Gly-42 and Met-113; the other oxygen hydrogen bonds with N⑀2 of His-265. One oxygen of the C1 carboxylate is near His-265 and the other forms a hydrogen bond with the peptide NH of Gly-43.
Binding of HOPDA to the S112A Variant-Our prior crystallographic analysis of the BphDS112C variant after incubation with HOPDA revealed clear electron density for all atoms from the C1 carboxylate through C5 (19). Because marginal electron density was observed for the phenyl group, the final interpretation was a complex of the enzyme with the product, HPD. The electron density maps obtained in the current study reveal the first images of the entire substrate bound to an MCP-hydrolase, albeit catalytically impaired variants missing the O␥ atom of the active site serine.
For the S112A variant, the electron density in unbiased (Fo-Fc) maps calculated before HOPDA was added to the model (Fig. 4) is reasonably compatible with multiple isomers of HOPDA: the monoanionic 2-keto form, (E)-2,6-dioxo-6-phenylhex-3-enoate; the enol form, (2Z,4Z)-2-hydroxy-6-oxo-6phenylhexa-2,4-dienoate; and a dianionic, 5,6-enolate variant of the 2-keto form, (3E,5Z)-2-oxo-6-oxido-6-phenylhexa-3,5dienoate. Although the unbiased maps did not exclude the presence of any of these forms in some fraction of the unit cells, the results of refinements suggest the monoanionic 2-keto form is most compatible with the x-ray data. In these experiments, the torsion angles about the double bonds characteristic of each species were tightly restrained to 0 or 180°, as appropriate. Following refinement, the deviations of the torsion angles from the expected values (Table 4) reflect the compatibility of the x-ray data with each isomer. The results suggest that the 2-oxo,6oxido isomer is the least compatible inasmuch as the refined torsion angle about the C5-C6 bond is 150°, a significant devi-

Tautomerization in a C-C Bond Hydrolase
ation from the expected value of 180°. By contrast, the 2,6-dioxo keto form refines without deviation from the expected value for the C3-C4 bond, and the 2-enol form refines with deviations of 5 and 12°for the C2-C3 and C4 -C5 bonds, respectively. Given that solution and single crystal absorption spectra indicate conversion of the enol form to another species, HOPDA was refined as the keto isomer, (E)-2,6-dioxo-6-phenylhex-3-enoate (Fig. 4). The observed conformation is consistent with a mechanism that requires access by a single protein residue (His-265) to the C2-hydroxo/oxo group and to C5. Moreover, the E conformation at C3-C4 and rotation about the C4 -C5 bond allow the substrate functional groups to simultaneously occupy the carboxylate and oxyanion binding sites: the distance between these sites is too short for a fully extended, coplanar conformation of HOPDA. Finally, the observed conformation of the keto isomer can be derived from an enol isomer sterically compatible with the active site with only minor changes in conformation.
The binding of HOPDA to the S112A variant is illustrated in Fig. 4. The carboxylate group lies near Arg-190 and forms multiple hydrogen bonds with the side chains of Arg-190, Asn-51, and Trp-266, and the peptide NH of Gly-43. This is consistent with binding of other anionic groups in the same site in other crystal structures of BphD and its variants (sulfate and HPD (19), and malonate, this work) and homologues (2,6-diketonona-1,9-dioic acid complex with MhpC (17)). The 2-oxo group (see below) lies in a polar pocket formed by Trp-266, Asn-111, and His-265 and could potentially hydrogen bond with all or any subset of these residues; thus this site should readily promote the formation of an oxyanion during the course of the reaction. The 6-oxo group clearly forms a hydrogen bond (2.7 Å) with the peptide NH of Gly-42 and interacts with the peptide NH of Met-113 at longer distance (3.4 Å). The 6-phenyl substituent interacts with a number of non-polar side chains including those of Ile-153, Leu-213, Trp-216, and Val-240, but is not "buried" in the sense that the three most distal atoms of the ring, CB3, CB4, and CB5, are solvent accessible to a 1.4 Å radius probe.
Hydrogen bonding between His-265-N␦1 and the carboxylate of Asp-237 has been observed in all crystal structures of BphD and its S112A variant. In the structure of the S112A: HOPDA complex, His-265-N⑀2 is 3.1 Å from the C2-keto oxygen atom, 3.5 Å from C4, and 4.6 Å from C5. A small rotation about the C␣-C␤ bond would place His-265 in a position and orientation competent for abstraction of a proton from C2-OH.
Although His-265 is near C5 and closer to the predicted position of the C5 pro-S proton, its position and orientation (relative to HOPDA) suggest a significant change in the conformation/position of the HOPDA and/or of the His-265 side chain would be necessary for proton delivery to C5.
In the vicinity of the scissile bond, the C␤ atom of Ala-112 is in contact with both C5 (3.6 Å) and C6 (3.5 Å). On the opposite side of the binding pocket, side chain atoms of Ile-153 and Leu-156 are ϳ4.1 Å from C5 and C6, respectively, with the C5 protons directed toward Ile-153. When the two independent wildtype structures (sulfate-or malonate-bound) are superposed onto the S112A:HOPDA complex, they predict the same location (within 0.3 Å) for the O␥ of Ser-112, placing O␥ 1.8 Å (sulfate) or 2.1 Å (malonate) from C6. This placement is apparently stabilized by a hydrogen bond with the peptide NH of Met-113, which is common to both structures and positions O␥ away from His-265 and the plane defined by C5 and its protons. For the superposed BphD WT :malonate and S112A:HOPDA structures, the angle O␥-C6-O6 is 62°, which is not close to 109°, the Burgi-Dunitz angle for nucleophilic attack at a carbonyl. However, if the 1 torsion angle is adjusted from ϩ146°t o Ϫ130°so as to place O␥ in the plane of C5 and its protons (equidistant from C4 and C6), the angle O␥-C6-O6 changes to 103°and O␥ is 2.4 Å from C5. Moreover, in the S112A complex and both superposed wild-type structures, the C␤ atom of residue 112 is in contact with both C6 and N⑀2 of His-265 (all distances less than 3.8 Å). Although deletion of the O␥ atom and the flexibility and dynamics of the substrate and enzyme warrant cautious interpretation, the S112A:HOPDA crystal structure suggests there is inadequate room in the active site for a water molecule to approach C6 from either face. In this regard, it is worth noting that the substrate binding site includes no crystallographically ordered waters and no water-sized voids. The closest ordered water is 7 Å away, beyond a nonpolar wall formed by Ile-153, Val-240, and the HOPDA phenyl group.
As has been noted previously, occupation of the active site by different ligands results in the movement or reorientation of several residues. For example, comparison of the wild type structures with the S112A:HOPDA complex reveals coordinated displacements of the Phe-239, Phe-175, and Trp-266 side chains in the vicinity of carboxylate and C2-hydroxo/oxo groups. Similarly, interactions of the Arg-190 and Phe-175 side chains with the substrate appear to motivate two major helices of the lid domain to close on the active site. For example, a comparison of the S112A:malonate and S112A:HOPDA structures shows movements of backbone atoms as large as 0.8 Å.
Structure of BphD-S112A/H265A:HOPDA-Based on the solution spectrum of the S112A/H265A:HOPDA complex, HOPDA was modeled in the crystal structure as the enol form, (2E,4E)-2-hydroxy-6-oxo-6-phenylhexa-2,4-dienoate. The structural evidence for an enol tautomer is not definitive. We note that the C2-C3 (Ϫ4°) and C4 -C5 (178°) angles deviate from 0/180°by less than 5°, whereas the angle about the C3-C4 (Ϫ167°) deviates by 13°; these observations are consistent with expectations for the enol tautomer, but they are also reasonable for a (3E)-keto tautomer. As in the S112A complex, HOPDA is not in a fully extended conformation, in this case because of the conformation about C2-C3. The significant differences between the conformation and binding of HOPDA in the two crystal structures are most readily described in terms of a transition from the S112A complex to the state observed in the S112A/H265A complex (Fig.  5). Described as a transition, three torsion angles change dramatically: the C1-C2 angle by 135°, the C2-C3 angle by 180°, and the C4 -C5 angle by 120°. In addition, the overall orienta-FIGURE 5. Comparison of substrate binding and conformation in the BphD S112A:HOPDA and S112A/H265A:HOPDA complexes. Part A is a stereo view of the superposed models for S112A:HOPDA and S112A/H265A:HOPDA emphasizing the active site. N and O atoms are colored in blue and red, respectively. The C atoms and the covalent bonds are colored in wheat for S112A:HOPDA and in gray for S112A/H265A:HOPDA. Parts B and C illustrate the conformations of the bound HOPDA in the S112A and S112A/H265A complexes, respectively. In both cases, the view is perpendicular to the C5-C6 bond and nearly orthogonal to the view in part A. The atom names and the values of torsion angles (degrees) are indicated. JULY 6, 2007 • VOLUME 282 • NUMBER 27 tion of the HPD portion of HOPDA changes in a way that places the 2-hydroxo group in a completely different binding site. It is as if the HPD plane was rotated by 108°, moving the C2 oxygen substituent by 4.1 Å and also altering the binding interactions of the carboxylate group. In the S112A complex, the two carboxylate oxygens are hydrogen bonded to the N1 and N2 of Arg-190, but in the S112A/H265A complex, one carboxylate oxygen and the 2-hydroxo group hydrogen bond with Arg-190, and the second oxygen hydrogen bonds only with the peptide NH of Gly-43. Moreover, Phe-175 and Trp-266 assume significantly different conformations. Phe-175 must move to prevent a steric conflict with the new position of the C2-hydroxo oxygen and, in turn, Trp-266 must move to free space for Phe-175. The 6-oxo and 6-phenyl groups also change somewhat in position and orientation, but essentially remain in the same binding sites as in the S112A complex.

Tautomerization in a C-C Bond Hydrolase
Optical Spectroscopy with Single Crystals-Single crystal absorption spectra and x-ray data were obtained from multiple crystals to probe the relationship between the crystal structures and the complexes in solution. Although the x-ray data were of lower resolution (2.3-2.0 Å), the electron density maps confirmed binding of HOPDA to both mutants in conformations indistinguishable from the results presented above. S112A:HOPDA crystals had a similar spectrum to the solution complex, although it was red shifted by ϳ14 nm. Considering that a similar shift occurs between wild type and S112A in solution, this appears to be in reasonable agreement, thereby confirming that a E:S k gives rise to the Ͼ490 nm absorbance feature.
Curiously, the spectrum of the S112A/H265A:HOPDA crystal differed significantly from solution. Maximum absorbance occurred at 470 nm, with a large shoulder at 520 nm. This may reflect populations of both enol and keto tautomers in the crystalline complex, in contrast to the enolate form in solution. As noted above, the electron density is compatible with both tautomers. The long incubation (30 -60 min) may have provided sufficient time for the keto tautomer to accumulate.

DISCUSSION
Through a combination of mutagenesis, kinetic and structural approaches, the current study (i) reports the trapping of an E:S Red intermediate that we had previously observed (19), (ii) identifies it as the keto tautomer, E:S k , and (iii) demonstrates that the conserved His of the catalytic triad catalyzes its formation (Fig. 6). This study therefore describes the first direct observation of the long-proposed keto-intermediate on the reaction coordinate of an MCP hydrolase. Previous evidence for His-265-mediated tautomerization was based on its proximity to both proton donor and acceptor positions in the S112C:HPD crystal structure (19), and stopped-flow analysis of variant enzymes showing impaired rates of processes thought to represent ketonization (13,18). The role of His-265 in tautomerization is further established by the failure of S112A/ H265A and H265A variants to accumulate E:S k as well as their inability to catalyze HPD tautomerization.
The large (60 -70 nm) red shift in absorbance observed upon formation of E:S k is counterintuitive. According to the empirical rules developed by Woodward (34) the keto intermediate, 2,6-dioxo-6-phenyl-3-hexenoate, should be blue-shifted with respect to HOPDA ( max Յ300 nm). A reasonable explanation is that the protein environment perturbs the electronic properties of the bound HOPDA, as has been reported for several other ligands when bound to proteins. For example, the absorbance maximum of an ␣,␤-unsaturated thiol ester inhibitor was red-shifted 90 nm when bound to the active site of crotonase (20). Similarly, the spectrum of a substituted stilbene derivative was red-shifted 102 nm upon binding to an antibody (21). Recent theoretical calculations suggest that the spectrum of rhodopsin is primarily determined by the interaction of the positively charged chromophore with Glu-113, which may account for a shift as large as 157 nm (22). For BphD, it is possible that Arg-190, which interacts with the carboxylate of HOPDA, may similarly perturb the spectrum of the keto tautomer. Whatever the origin of the perturbation, it is interesting that the spectrum of the HOPDA enolate in the H265A variants is not similarly perturbed, and the crystal structure of the complex of the S112A/H265A:HOPDA complex reveals reduced interactions with Arg-190. Indeed, the similarity of the spectra of the HOPDA enolate bound to each of the H265A variants to that of the enolate in solution further supports the assignment of the red-shifted absorbance feature to the keto form of the substrate.
Assignment of E:S Red in the wild type enzyme relies upon linking it to the E:S k crystal structure (S112A:HOPDA) via the spectroscopic similarity of E:S Red in both enzymes. Although both spectra have similar peak shapes, E:S Red in S112A is redshifted 14 nm from that in wild type (Fig. 2C). If the large (ϳ200 nm) red shift of E:S k from the expected absorbance maximum of the keto tautomer in solution is dictated by active site interactions, it is reasonable to expect a relatively small shift upon substitution of the polar residue Ser-112 with non-polar alanine. Electrostatic interactions between the chromophore and polar amino acids are responsible for similar shifts in visual pigments. For example, the substitution of Ser-94 with Ala in a visual pigment from newt retina caused a 14-nm blue shift (35).
Identification of E:S k on the reaction coordinate enables us to refine our view of the catalytic mechanism of MCP hydrolases. In our previous stopped-flow study of wild type BphD, we presented two interpretations of the kinetic data, depending on assignment of the E:S Red intermediate (19). The scheme in which E:S Red was assigned to the enolate, E:S e , can now be ruled out in favor of the scheme involving two interconverting enzyme conformations, whereby only one active site per dimer can catalyze C-C cleavage/HPD release. Unfortunately, the present study cannot provide new insight into the two-conformation hypothesis itself, because the crystallographic symmetry forces all active sites to be identical.
Although the two-conformation issue is not addressed, these results advance our understanding of the enzyme-catalyzed tautomeric half-reaction. The E:S k structure combined with the critical role of His-265 in tautomerization are consistent with previous proposals (18,19) whereby His-265 abstracts a proton from the 2-OH of the substrate and then protonates at the C5 position, perhaps via the intermediacy of a C6 enolate stabilized in the oxyanion hole (Fig. 6). Protonation by His-265 would direct a proton to the pro-S position of C5. Because BphD (19), Ht-BphD (36) and MhpC (11) incorporate an H5 E proton into HPD, pro-S protonation at C5 must be followed by C5-C6 fragmentation onto the re face of the C3-C4 double bond (Fig. 6). The synclinal conformation we observe about the C4 -C5 bond of E:S k (Fig. 4) is fully consistent with this stereochemical course.
Our data do not explicitly rule out two variations on the proposed mechanism of tautomerization. First, it is possible that His-265 is protonated in the substrate-free enzyme and that the enzyme binds the enolate form of the substrate. However, a decrease in rate of the proposed tautomerization step occurred at pH 5 in MhpC, implying that histidine is not protonated at neutral pH (18). Second, it is possible that the S112A variant traps an E:S k 6e species and that Ser-112 protonates C5. Although the S112A:HOPDA crystal structure is not rigorously incompatible with the E:S k 6e form, the presence of E:S k 6k is more likely for two reasons. First, the refinements suggest rotation about C5-C6 is necessary to fit the x-ray data. Second, the nonplanarity of the conjugated C3-C4 and C5-C6 double bonds in E:S k 6e suggests that minimal stabilization energy would arise from orbital overlap; hence the bond energies of the C-6 keto tautomer (E:S k 6k ) are predicted to be ϳ18 kcal/mol more stable than the C-6 enol (E:S k 6e ) (37). We are currently performing experiments to distinguish possible C6 tautomers.
The crystal structures also provide insight into several other aspects of catalysis. First, the S112A:HOPDA structure shows that the phenyl ring of the E:S k intermediate is twisted out of planarity with respect to the C-6 carbonyl, perhaps enhancing the latter's susceptibility to nucleophilic attack in the subsequent hydrolytic half reaction by disrupting overlap between the ring and carbonyl -orbital systems. Second, the distance between the carboxylate and oxyanion binding sites is appropriate for the non-planar E:S k intermediate but not the fully extended, coplanar conformation of HOPDA that presumably predominates in solution (2). It is possible that the observed conformation has additional significance relative to the hypothesis that tautomerization is promoted by destabilizing the substrate in a twisted, non-planar binding mode (10). Notably, model building experiments suggest that the fully extended 2Z,4E dienol configuration could associate with the active site prior to ketonization, but it cannot do so and simultaneously occupy the carboxylate and oxyanion binding sites. Moreover, the enol tautomer in the S112A/H265A:HOPDA crystal structure is not fully extended (Fig. 5).
The 2E,4E conformation of HOPDA in the S112A/H265A complex (Fig. 5) probably does not reflect a catalytically relevant species. The absence of E:S k accumulation in both H265A

Tautomerization in a C-C Bond Hydrolase
variants suggests that tautomerization becomes rate-limiting. Instead, a relatively slow (ϳ50% of wild type or S112A tautomerization, Table 1) HOPDA deprotonation occurs en route to a species not spectroscopically observed in the wild type reaction. Although E:S k does not accumulate without His-265, the substrate may slowly tautomerize, as in solution, to generate the crystallographically observed 2E,4E isomer, which may be the more stable form within the active site. Interestingly, two additional transient kinetic phases after deprotonation are observed in formation of this complex (Table 1), and may reflect molecular rearrangement to generate the different isomers of HOPDA.
While His-265 is crucial for the tautomeric half-reaction, Ser-112 clearly catalyzes hydrolysis, because its replacement with alanine impairs this reaction by at least 10 5 -fold (Ͼ7 kcal/ mol) in wild-type BphD. Large losses of activity were also observed in Ht-BphD (13), MhpC (18), XylF (38), and CumD (39). Although Ser-112 is catalytically important, its precise role remains uncertain. Several studies have suggested that, in MCP hydrolases, the catalytic serine does not act as a nucleophile as in other serine hydrolases, but as a hydrogen bond donor to stabilize a gem-diolate formed after His-mediated attack of water at the substrate carbonyl (12)(13)(14)(15)(16)18). Nevertheless, the present crystal structures are more consistent with a nucleophilic role for Ser-112. For instance, no solvent molecule suitable for attack of the E:S k C6 carbonyl was observed in the active site of the S112A:HOPDA complex, despite the extra space made available by removal of the serine hydroxyl: the closest water is 7 Å away. By contrast, the O␥ of Ser-112 is well positioned to attack the C6 carbonyl when it is modeled into this structure. Moreover, the main chain amide protons of Met-113 and Gly-42, which constitute the canonical ␣/␤ hydrolase oxyanion hole, could readily accommodate any oxyanion intermediates by hydrogen bonding. Hence, the possibility that Ser-112 contributes an additional hydrogen bond to an oxyanion intermediate, as proposed in the gem-diol hypothesis, seems unnecessary. Because Ser-112 appears appropriately poised for nucleophilic catalysis employing the canonical oxyanion hole, it is unclear why, in the MCP hydrolases, it should adopt an alternate role. As discussed previously (19), the gem-diol mechanism also fails to explain the observed release of benzoate from the active site after HPD despite the apparently lower affinity of the enzyme for benzoate and the solvent accessibility of the benzoate-binding pocket.