Kinetic and Mechanistic Characterization of the Formyl-CoA Transferase from Oxalobacter formigenes*

Oxalobacter formigenes is an obligate anaerobe that colonizes the human gastrointestinal tract and employs oxalate breakdown to generate ATP in a novel process involving the interplay of two coupled enzymes and a membrane-bound oxalate:formate antiporter. Formyl-CoA transferase is a critical enzyme in oxalate-dependent ATP synthesis and is the first Class III CoA-transferase for which a high resolution, three-dimensional structure has been determined (Ricagno, S., Jonsson, S., Richards, N., and Lindqvist, Y. (2003) EMBO J. 22, 3210–3219). We now report the first detailed kinetic characterizations of recombinant, wild type formyl-CoA transferase and a number of site-specific mutants, which suggest that catalysis proceeds via a series of anhydride intermediates. Further evidence for this mechanistic proposal is provided by the x-ray crystallographic observation of an acylenzyme intermediate that is formed when formyl-CoA transferase is incubated with oxalyl-CoA. The catalytic mechanism of formyl-CoA transferase is therefore established and is almost certainly employed by all other members of the Class III CoA-transferase family.

Oxalobacter formigenes is a Gram-negative, obligate anaerobe (1, 2) that employs oxalate as a source of both energy and carbon for cellular biosynthesis (3). This microorganism, which colonizes the gastrointestinal tract of humans and other mammals, is of biological interest for two reasons. First, recent work has demonstrated an intriguing correlation between the absence of O. formigenes in humans and kidney stone formation because of elevated levels of oxalate in the blood (4,5), an observation that has led to the somewhat controversial hypothesis that O. formigenes plays a key role in mediating mammalian oxalate homeostasis (6). Second, ATP production appears to depend solely on the anaerobic conversion of oxalate to formate and CO 2 (7). Carbohydrates cannot be used to replace oxalate as a growth substrate, perhaps implying that this organism lacks a functional glycolytic pathway (8).
As reported for other bacteria (9 -11), cleavage of the oxalate C-C bond in O. formigenes is accomplished by the action of a thiamin-dependent, oxalyl-CoA decarboxylase (12) encoded by the oxc gene (13) (Fig. 1). ATP-dependent formation of oxalyl-CoA is avoided by coupling decarboxylation to an acyl transfer reaction in which formyl-CoA 1 and oxalate 2 are converted to oxalyl-CoA 3 and formate 4 by formyl-CoA transferase (FRC) 1 (14). The overall catalytic cycle therefore transforms oxalate into formate and CO 2 (Fig. 1). The metabolic importance of this coupled enzyme system in O. formigenes is suggested by the fact that oxalyl-CoA decarboxylase and FRC constitute ϳ20% of the total protein content in this microorganism (15). In contrast to aerobic bacteria, such as Pseudomonas oxalaticus (16), which employ formate dehydrogenase to oxidize formate to CO 2 with concomitant production of NADH (17), O. formigenes uses a membrane-bound, formate:oxalate antiporter (18 -20), encoded by the oxlT gene (21), to create the electrochemical gradient necessary for ATP synthesis (8,20). We also note that the genome sequence of Escherichia coli (22) reveals the existence of open reading frames encoding proteins that are homologous to both FRC and oxalyl-CoA decarboxylase. This is an intriguing observation given that the ability of this microorganism to metabolize oxalate has not been reported.
There are several aspects of FRC that are of considerable biochemical and mechanistic interest. First, primary sequence analysis has placed FRC within a new family (Class III) of CoA-transferases (23), which also includes enzymes that are involved in the anaerobic metabolism of carnitine (24), toluene catabolism (25,26), Strickland fermentation (27), and (putatively) bile acid transformation (28). All of these enzymes have similar masses and are active as homo-or heterodimers (23). The catalytic mechanism employed by the Class III CoA-transferases has yet not been experimentally defined, although it has been the subject of considerable speculation (23,29,30). Our recent x-ray crystallographic studies have shown that FRC exists as an interlocked dimer ( Fig. 2A) (29), raising significant questions concerning the folding pathways that might lead to such a structure. In this regard, we note that the protein encoded by the YfdW gene in E. coli, which is clearly homologous to FRC, has been shown independently to possess an identical interlocked structure (30,31). The ability of the YfdW gene product to catalyze coenzyme A transfer, however, has not yet been reported.
We now report the first detailed kinetic characterization of recombinant, wild type (WT) FRC, and a number of site-specific mutants, which suggest that catalysis proceeds via anhydride intermediates as proposed for Class I CoA-transferases (32). Further evidence for this mechanistic proposal is provided by the x-ray crystallographic observation of an acylenzyme inter-mediate that is formed when FRC is incubated with oxalyl-CoA. These studies therefore establish the catalytic mechanism of FRC, which is almost certainly employed by all other members of the Class III CoA-transferase family.

EXPERIMENTAL PROCEDURES
Materials-All of the materials were of the highest purity available and, unless stated otherwise, were obtained from Fisher or Sigma-Aldrich. Protein concentrations were determined by the Lowry method (33) based on a standard curve constructed using known amounts of bovine serum albumin. DNA (34) and purified using anion exchange and affinity chromatography. Hence, the lysate supernatant from 4 -6 liters of culture was loaded on a 120-ml DEAE Fast Flow column equilibrated with buffer A (25 mM sodium phosphate, 1 mM dithiothreitol, pH 6.2), and FRC was obtained by stepping to a concentration of 35% buffer B (25 mM sodium phosphate, 1.0 M NaCl, 1 mM dithiothreitol, pH 6.2). Fractions containing FRC were then loaded on a 20-ml Blue-FF affinity column equilibrated with buffer A, and the column was washed with 1:1 buffer A/buffer B. Recombinant FRC was then obtained by eluting with buffer C (25 mM glycine, 1 mM dithiothreitol, 20% isopropanol, pH 9.0). After buffer exchange by passage through a 135-ml G-25 desalting column, equilibrated with buffer A, the solution containing FRC was injected on a 60-ml Q-Sepharose HP column equilibrated with buffer A. Purified FRC was then obtained by stepping to 20% buffer B followed by a linear increase in buffer B to a final concentration of 35%, with the desired enzyme eluting near the middle of this gradient. Glycerol was added to the combined FRC-containing fractions to a final concentration of 10%, given that the purified protein has a tendency to precipitate. The purity at each chromatographic step was verified by SDS-PAGE with Coomassie Blue staining. The flow rates used in these experiments were 5, 4, 10, and 5 ml/min for the DEAE, Blue-FF, G-25, and Q-Sepharose columns, respectively.
Expression and Purification of the D169A, D169E, and D169S FRC Mutants-Mutagenic primers were designed using Gene Runner version 3.05 (Hastings Software). The pET-9a plasmid containing the WT FRC sequence was purified using the Wizard Plus Minipreps® DNA purification system (Promega) and used as a template for PCR with mutagenetic primers using the QuikChange® site-directed mutagenesis kit (Stratagene). The desired point mutations were verified by DNA sequencing of the inserts in the resulting pET-9a plasmids isolated from transformed XL-1 or XL-10 Gold supercompetent cells (Stratagene). Subsequent transformation of BL21(DE3) competent cells (Novagen) permitted overexpression and purification of the D169A, D169E, and D169S FRC mutants as described for recombinant, WT FRC.
Synthesis of Formyl-CoA and Oxalyl-CoA-Formyl-CoA of very high purity was prepared by modification of literature procedures (35,36) (Fig. 3). Thus, formic acid (5.8 ml, 150 mmol) was added dropwise to acetic anhydride (7.1 ml, 75 mmol), and the resulting mixture was heated at 45°C for 2.5 h to give a solution of the mixed anhydride (formylating reagent) (37). After cooling to room temperature, pyridine (61 l, 0.75 mmol) was added to the solution immediately followed by thiophenol (5.1 ml, 50 mmol), and the resulting mixture stirred at room temperature for 24 h. Unreacted anhydrides, formic acid, and acetic acid were then removed by distillation under reduced pressure (20 mm Hg) at 50°C, until the total volume was ϳ6 ml. This oily material was then washed with cold, deionized water, and the organic layer was dried (MgSO 4 ). The desired formyl thioester 5 was then obtained by vacuum distillation as a clear oil: 3.1 g, 45%; boiling point 115-117°C, 23 mm Hg (literature valve 101°C, 15 mm Hg (36)). This material was then reacted with the sodium salt of coenzyme A using literature protocols to give formyl-CoA 1 (35).
Oxalyl-CoA 3 was prepared by reaction of coenzyme A with thiocresoxalic acid using literature procedures (38). Formyl-CoA 1 and oxalyl-CoA 3 were further purified by reverse-phase HPLC on a preparative C 18 column (Dynamax 60A C 18 , 250 ϫ 21.4 mm). In our standard procedure, the column was equilibrated with 90% mobile phase A (10 mM NaH 2 PO 4 , pH 4.5) and 10% mobile phase B (phase A containing 20% CH 3 CN) running at 8 ml/min. After injection of each sample, the amount of B was increased to 40% using a linear gradient over 20 min. The absorbance of the eluent was monitored at 260 nm, and fractions containing the pure CoA derivative were combined and lyophilized to give the desired compounds as white solids, which could be stored at Ϫ80°C without significant amounts of decomposition.
Enzymatic Assay-Wild type, recombinant FRC was assayed by

FIG. 2. Crystal structure of the interlocked FRC dimer complexed to Co-A.
A, for clarity, the protein monomers are colored red and green and represented by molecular ribbons. The bound co-factor is shown as a space-filling model, which is colored using the following scheme: carbon, black; nitrogen, blue; oxygen, red; sulfur, yellow; phosphorus, purple. B, active site residues located within 4 Å of the thiol moiety of bound CoA. The letter designation (A or B) in the numbering scheme indicates the FRC monomer in which the residue is located. This image was generated using VMD (58) and POV-Ray (Persistence of Vision Development Team).

FIG. 3. Improved synthetic preparation of formyl-CoA.
measuring the initial rate of oxalyl-CoA formation. The assay mixture contained 60 mM potassium phosphate, pH 6.7, FRC (90 ng), appropriate concentrations of substrates (formyl-CoA and oxalate), and, in the case of the product inhibition experiments, oxalyl-CoA or formate (total volume, 200 l). The reaction was started by the addition of formyl-CoA after incubating the other components at 30°C for about 30 s. Aliquots of the reaction mixture (90 l) were typically taken after 60 and 90 s and quenched with 10% AcOH (10 l) before being analyzed by reversephase HPLC using a modification of previous procedures for analyzing CoA derivatives (39). Thus, aliquots (75 l) of the quenched reaction mixture were injected onto a C 18 analytical column (Dynamax Microsorb 60 -8 C18, 250 ϫ 4.6 mm) that had been equilibrated using 86% buffer A (25 mM NaOAc, pH 4.5) and 14% buffer B (buffer A containing 20% CH 3 CN) running at 1.0 ml/min. Immediately after injection the proportion of buffer B was increased to 30% over a 210-s time period, followed by a step to 100% buffer B that was continued for 90 s before re-equilibrating with 86:14 buffer A:buffer B. The eluant was monitored at 260 nm. Under these conditions, oxalate eluted close to the void volume of the column (2.6 min), oxalyl-CoA eluted after 6.3 min, and free CoA and formyl-CoA eluted last from the column (9.0 min). The amount of oxalyl-CoA in the aliquots was determined by integration of the oxalyl-CoA peaks in the HPLC chromatograms and comparison with known amounts of authentic material. These measurements were calibrated using independent determinations of formyl-CoA concentration using a hydroxylamine-based colorimetric assay (35) and oxalate concentration in hydrolyzed and nonhydrolyzed samples of oxalyl-CoA with a standard detection kit (Sigma). No formation of oxalyl-CoA was detected in control experiments when the enzyme or either substrate was omitted or boiled enzyme was used. The limit of detection of this HPLC-based assay is 0.05 M of oxalyl-CoA when 75-l aliquots are injected onto the column.
The specific activities of the D169A, D169E, and D169S FRC mutants were assayed using an identical procedure except that reaction mixtures were incubated for up to 60 min prior to quenching with AcOH. In addition, given the much lower activity of the FRC mutants, the amount of protein in each assay was increased to 2 g, and the initial concentrations of oxalate and formyl-CoA were 100 mM and 200 M, respectively.
Analysis of Kinetic Data-Steady state kinetic constants (K m and V max ) were determined by fitting the data from the initial rate studies using weighted hyperbolic regression analysis (Hyper, version 1.1). Data analysis was performed using standard equations (40).
Determination of the Equilibrium Constant for FRC-catalyzed CoA Transfer-The equilibrium constant for the FRC-catalyzed reaction was determined by incubating the recombinant, WT enzyme (18 g) with 73 M formyl-CoA, 50 M potassium oxalate in 60 mM potassium phosphate, pH 6.7 (total volume, 200 l). The solution also contained 13 M formate and 13 M free CoA, present in the initial sample of formyl-CoA used in the experiment. This mixture was then incubated at 22°C for 90 min. During this time, aliquots (45 l) were withdrawn after 10, 27, and 52 min and quenched with 10% AcOH (5 l). The concentration of oxalyl-CoA, free CoA (and therefore formate), and formyl-CoA in each sample was measured by reverse-phase HPLC as described above. Under these conditions, equilibrium was achieved after 27 min, giving K eq ϭ 32 Ϯ 3.
Size Exclusion Chromatography Measurements-A BIOSEP SEC-S2000 column (300 ϫ 7.8 mm with 75 ϫ 7.8-mm guard column) was calibrated using lysozyme (14.4 kDa), carbonic anhydrase (29.0 kDa), peroxidase (44.0 kDa), bovine serum albumin (66.0 kDa), alcohol dehydrogenase (150 kDa), and ␤-amylase (200 kDa). The void volume was measured by injecting blue dextran. A sample of recombinant, WT O. formigenes FRC (72 g) was prepared by filtration through a 50-kDa cut-off spin column, and the retentate resuspended in 100 mM potassium phosphate, pH 6.6 (total volume, 75 l). An aliquot of this solution was then injected onto the BIOSEP SEC-S2000 column, giving a single peak with a retention time corresponding to a molecular mass of 81 kDa.
Crystallization and Structure Determination of the D169A, D169E, and D169S FRC Mutants-All three Asp 169 mutants were dialyzed and subsequently crystallized as their complexes with coenzyme A using identical conditions to those described previously for WT FRC (41). Diffraction data for the Ala and Glu mutants were collected at beam line X11 (Deutsches Electronen Synchrotron, Hamburg, Germany) equipped with a Mar research 165-mm CCD detector; diffraction data for the Ser mutant were collected at I711 (MaxLab) equipped with a Mar research 165-mm CCD detector. All of the data were processed with MOSFLM (42) and scaled with SCALA (43). The crystals of the three mutants all belonged to space group I4 with the following unit cell parameters: D169E: a ϭ b ϭ 150.0 Å, c ϭ 99.6 Å; D169A: a ϭ b ϭ 150.0 Å, c ϭ 99.6 Å; D169S: a ϭ b ϭ 151.8 Å, c ϭ 100.2 Å. All of the asymmetric units for these complexes contain a dimer.
The structures were solved by difference Fourier using the FRC-CoA complex (29) as a model and were refined by REFMAC5 (44) using maximum likelihood residual, anisotropic scaling, bulk solvent correction, and atomic displacement parameter refinement as used in the "translation, libration, screw rotation" method with each monomer as a rigid group. 5% of the reflections were excluded for subsequent use in monitoring R free . Noncrystallographic symmetries were not employed during the refinement. Water molecules were added by ARP-WARP (45), and the O (46) software package was used for model building. The quality of the models was assessed by PROCHECK (47), and structure comparisons were made in TOP (48) using default parameters. Statistics for data collection and the refined models are listed in Table I.
Crystallization and Structure Determination of the FRC/Oxalyl-CoA Complex-FRC was dialyzed as described previously (29,41) and crystallized with 20 mM oxalyl-CoA in 100 mM sodium cacodylate, pH 6.5, 0.2 M Mg(OAc) 2 , and polyethylene glycol 8000 at 4°C using the hanging drop technique. Silicone oil was added to the crystallization drop as cryoprotectant, and crystals were flash frozen. Diffraction data were collected at 110 K at beamline I711 (MaxLab) on fresh crystals (11 days) to minimize uncatalyzed hydrolysis of oxalyl-CoA. All of the data were processed with MOSFLM (42) and scaled with SCALA (43). The crystals were tetragonal and belonged to the space group p43212 (a ϭ b ϭ 100.2 Å, c ϭ 196.9 Å).
Structure determination was achieved by difference Fourier using the structure of the Q17A FRC mutant 2 as a model. Refinement was carried out using REFMAC5 (44), as described above for the Asp 169 FRC mutants but using noncrystallographic symmetry, and water molecules were added by ARP-WARP (45). The software programs O (46) and Coot (49) were used for model building. As before, model quality was assessed by PROCHECK (47).

Expression and Purification of Recombinant, WT FRC-WT
FRC was overexpressed in E. coli (BL21(DE3)) following literature protocols (34) and was purified using a three step procedure that was substantially modified from that reported previously (14,34). Thus, after an initial purification using anion exchange chromatography, the resulting protein was subjected to affinity chromatography using a Blue Sepharose Fast Flow column. After buffer exchange, active FRC was obtained in highly pure form using an additional high performance, anion exchange chromatography step. We have observed that inclusion of the affinity step is important in obtaining material that can be used to obtain crystals of suitable quality for x-ray structure determination. This expression and purification procedure typically yielded 10 -15 mg of the purified enzyme/liter of culture.
Kinetic Characterization of Recombinant, WT FRC-Studies of the kinetic properties of FRC were performed using an HPLC-based assay rather than the coupled assay used in preliminary studies of the native enzyme isolated from O. formigenes (14). The rate of oxalyl-CoA production was therefore determined by direct measurement of the peak area, calibration being carried out using an authentic standard. Initial rates were measured as a function of oxalate concentration at different fixed levels of formyl-CoA. Although O. formigenes FRC exhibits activity over a broad pH range, all of the assays were carried out at pH 6.7 to minimize the extent of chemical hydrolysis of the formyl-CoA substrate. As observed for other Class III CoA-transferases, intersecting lines were observed in the double-reciprocal plots obtained from the steady state kinetic data (Fig. 4A), suggesting that the FRC-catalyzed reaction proceeded via a ternary complex. The kinetic mechanism of FRC is therefore different from that employed by Class I CoAtransferases for which ping-pong kinetics are observed (32,50). Standard curve fitting techniques and replots (40) gave values of 8.0 Ϯ 0.3 M and 3.9 Ϯ 0.3 mM for the apparent K m values of formyl-CoA and oxalate, respectively, and the turnover number of the enzyme (k cat ) as 4.3 Ϯ 0.1 s Ϫ1 . The apparent K m of formyl-CoA in our study is therefore 10 3 -fold lower than that reported for the native FRC (14), a difference that probably reflects the difficulty of obtaining highly purified formyl-CoA by the synthetic procedures used in the early studies of this enzyme. Having established these kinetic parameters, we then determined the effects of product inhibition. As before, these experiments were carried out at fixed initial concentrations of formyl-CoA, and oxalate was varied in the presence of either formate or oxalyl-CoA. The resulting double-reciprocal plots showed that formate is nearly a pure competitive inhibitor against oxalate (Fig. 4B), and oxalyl-CoA is a mixed type inhibitor with respect to oxalate (Fig. 4C). Given the relative magnitudes of the apparent K m values for formyl-CoA and oxalate and the crystallographic observation that an acylated enzyme can be formed upon the addition of oxalyl-CoA to WT FRC (see below), these results are consistent with an ordered bi-bi kinetic mechanism in which formyl-CoA binds initially to the free enzyme followed by oxalate to give a ternary complex. After CoA transfer to yield products, formate is then released followed by oxalyl-CoA. Analysis of the steady state kinetic data using this model gives K is (formate) ϭ 17 Ϯ 1 mM and K ii (formate) ϭ 380 Ϯ 40 mM, and K is (oxalyl-CoA) ϭ 150 Ϯ 50 M and K ii (oxalyl-CoA) ϭ 280 Ϯ 90 M.
Substrate Stability and Equilibrium Constant Determination-The use of an oxalate:formate antiporter to maintain the pH and electrochemical gradients necessary for ATP synthesis in O. formigenes implies that the conversion of oxalate to formate and CO 2 must proceed without chemical hydrolysis of the thioester intermediates in the pathway. It has therefore been speculated that FRC might be complexed with oxalyl-CoA decarboxylase in the cellular environment and that this complex might be located in proximity to the membrane-bound oxalate: formate antiporter protein (29). Using our HPLC methods, we therefore determined the half-lives for formyl-CoA and oxalyl-CoA. In these experiments, the pseudo-first order rate for the uncatalyzed hydrolysis of these compounds at pH 6.7 and 30°C was determined by HPLC analysis of the relevant peaks. Standard fitting procedures gave an estimate of 150 min for the half-life of formyl-CoA under these conditions, which is in reasonable agreement with a literature value of 300 min in aqueous solution at room temperature and neutral pH (35). The measured half-life for oxalyl-CoA was about 10 days, which is again consistent with previous reports that solutions of oxalyl-CoA at pH 6.5 are stable for weeks when stored at Ϫ 15°C (38). The large difference in the rate of uncatalyzed hydrolysis of these two thioesters can likely be attributed to the presence of the negatively charged carboxylate group in oxalyl-CoA, which will destabilize the tetrahedral adduct formed by nucleophilic attack of water on the thioester carbonyl. The equilibrium constant for the FRC-catalyzed CoA transfer was determined to be 32 Ϯ 3, favoring oxalyl-CoA and formate to formyl-CoA and oxalate. This value is similar to the equilibrium constant determined for the reaction catalyzed by succinyl-CoA:acetoacetate transferase (32).
Alternate Substrate Studies-The substrate specificity of FRC was also examined, in part because of the x-ray crystallographic observation of a ternary complex involving E. coli YfdW, oxalate, and acetyl-CoA (30). In our initial series of experiments, no oxalyl-CoA formation was detected in the HPLC assay when FRC was incubated for up to 80 min in the presence of 50 mM oxalate and 200 M acetyl-CoA under our standard assay conditions. This is an interesting observation given that the steric bulk of acetyl-CoA is smaller than might be expected for oxalyl-CoA. The relevance of this observation to interpreting the functional roles of active site residues in YfdW remains unclear, however, in the absence of information upon the physiological reaction catalyzed by this gene product in E. coli.
Having established that acetyl-CoA was not a substrate, we next examined whether oxalate analogs could substitute for this substrate. Oxamate (H 2 NC(ϭO)CO 2 Ϫ ) was of particular interest given that it retains a carboxylate group capable of forming acylated enzyme intermediates. Recombinant, WT FRC was therefore incubated with 35 M formyl-CoA and 40 mM oxamate under our standard conditions. Aliquots of the reaction mixture taken after 60 and 105 min were analyzed by reverse-phase HPLC. Although small amounts of oxalyl-CoA were observed under these conditions, these were formed from contaminating amounts of oxalate in the oxamate samples. Control experiments also showed that oxamate did not impact the rate of formyl-CoA hydrolysis. 3 Gel Filtration Experiments-The remarkable interlocked dimer that is observed in the crystal structure of O. formigenes FRC ( Fig. 2A) appears inconsistent with previous claims that this enzyme exists as a monomer in solution (14). We therefore examined the oligomeric state of recombinant, WT FRC using size exclusion chromatography. After calibration of the column using a variety of molecular mass standards, we observed that FRC eluted as a single peak with a retention time corresponding to a calculated molecular mass of 81 kDa (Fig. 4D). Because the deduced primary structure of FRC gives a theoretical molecular mass of 47.2 kDa, we conclude that the recombinant enzyme is a dimer in solution with the deviation between measured and theoretical mass arising from the unusual, tightly interlocked dimer structure.
Expression and Characterization of Site-specific FRC Mutants-Recent crystallographic studies have shown that a conserved aspartic acid residue (Asp 169 ) is positioned in the active site of FRC close to the thiol group of bound coenzyme A (Fig.  2B) (29). This observation has given rise to a general consensus that this aspartate side chain must play a critical role in catalysis (Fig. 5) (29 -31), but no kinetic or structural evidence to support this hypothesis has yet been reported. We therefore expressed and purified a number of site-specific FRC mutants in which Asp 169 was replaced by alanine, serine, and glutamic acid to give the D169A, D169S, and D169E mutants, respectively. The catalytic activity of these proteins was then assayed following our usual procedures. In this regard, we note that a 3 S. Jonsson, unpublished results. mutant enzyme exhibiting a specific activity that is reduced by a factor of 3 ϫ 10 4 relative to WT FRC would lie at the detection limit of our HPLC assay. All three FRC mutants exhibited considerable decreases in specific activity under the optimum conditions for recombinant, WT FRC, even though their behavior during purification was very similar to the wild type enzyme. Unexpectedly, although both the D169E and D169S FRC mutants exhibited no activity above our detection limit, the specific activity of the D169A mutant was decreased only 1300fold. Control experiments using E. coli that had been transformed with plasmid lacking the O. formigenes frc gene established the absence of detectable FRC activity that might have arisen from the presence of the native enzyme encoded by the YfdW gene in the expression host (30), which might have copurified with our site-specific FRC mutants. Thus, no evidence for the production of oxalyl-CoA was obtained when formyl-CoA was incubated with the lysate of cells transformed with an expression vector lacking the FRC insert. We also showed that the decreased ability of these three FRC mutants to catalyze oxalyl-CoA formation was not associated with an increased rate of formyl-CoA hydrolysis because of changes in active site structure. Finally, so as to ensure that this loss of CoA-transferase activity was not associated with incorrect folding and dimerization of the FRC mutants, all three of these proteins were co-crystallized with coenzyme A. The three-dimensional structures of the D169A, D169S, and D169E FRC/CoA complexes were elucidated by x-ray crystallography to a resolution of 2.13, 2.3, and 2.10 Å, respectively (Table I). These studies showed that all three mutants were correctly folded and formed interlocked dimers. In addition, none of their complexes showed any significant difference in structure from that observed for the WT FRC/CoA structure, the superimposition of the 854 backbone C ␣ atoms of the FRC dimer and the D169A, D169S, and D169E FRC mutants giving root mean square deviations of 0.55, 0.41, and 0.35 Å, respectively. This is consistent with our original finding that the Asp 169 side chain forms only one hydrogen bond with the main chain amide of Glu 140 that is mediated by a water molecule (29). Given that all three FRC mutants bound coenzyme A in their active sites, we were able to investigate whether observed enzyme activities could be correlated with variations in the distal portion of the pantotheinyl moiety close to the thiol of the co-factor (Fig. 6).
The conformation of co-enzyme A in the active site of the D169E FRC mutant was the most similar to that observed for the co-factor in the WT FRC/CoA complex. Because the glutamate side chain is bulkier than that of aspartate, however, the pantotheine group was displaced slightly (Fig. 6c), and some negative electron density appeared on the thiol group of the bound co-factor, which was therefore refined assuming an occupancy of 0.5. We interpret this to indicate that there is disorder in positioning the thiol because of the increased length of the glutamate side chain and that the lack of activity in the D169E FRC mutant is probably associated with problems in positioning the formyl-CoA correctly so as to permit anhydride formation by reaction of the thioester with the carboxylate moiety (Fig. 5).
In the complexes involving the D169A and D169S FRC mutants (Fig. 6), there was extra density connected with the sulfur of bound co-enzyme A, and so this moiety was modeled in its oxidized form. The oxidized thiol forms a hydrogen bond with the backbone amide of Glu 140 rather than to the cognate functional groups in Gln 17 and Ala 18 as observed in the WT FRC/ CoA complex (29). Oxidation of the thiolate in the D169S FRC/ CoA complex does not seem to be as complete (even if it is the predominant species in the crystal), however, because we observed residual density for CoA in the conformation seen in the WT FRC/CoA complex where the co-factor is reduced. The reason that oxidation is observed only in these two complexes is probably due to the fact that removal of the Asp 169 carboxyl group leaves the thiol group quite exposed. We conclude that the significant loss of transferase activity in these three mutants arises from their inability to form the key anhydride intermediates when incubated with formyl-CoA (Fig. 5).
The conformational properties of the mobile loop region comprising four contiguous glycine residues (Gly A258 -Gly 261 ) represented another interesting structural difference between the three FRC mutants and the wild type enzyme, which may also indicate the functional significance of this region of the protein.
In the apoenzyme dimer, we observed that this flexible loop adopted two different conformations (29). Hence, in one FRC monomer the loop was in an "open" conformation leaving a cavity in front of Asp 169 , whereas in the other monomer the loop was "closed." Upon co-factor binding, however, the loops in both monomers adopted the closed conformation. These two structural states appear to be correlated with the location of the Trp 48 side chain, which is packed against the loop residues in the closed conformation. Rotation of the tryptophan side chain through 90 o about the C ␣ -C ␤ bond allows the loop to adopt the open conformation. In the D169A and D169E FRC/ CoA complexes, both loops are present in the closed conformation. Conversely, the structure of this region in the D169S FRC/CoA complex resembles the apoenzyme in that the loop is open in one monomer and closed in the other. Although is it clear that mutating Asp 169 perturbs the loop conformational preferences, which accords with our belief that residues 258 -261 are somehow involved in catalytic function, firm conclusions are precluded by the fact that in all three structures there is some density consistent with a double conformation of the loop in monomer B. In any event, our working hypothesis is that the flexibility of this loop segment permits it to act either as a gate, opening and closing to let oxalate into the active site or as a moiety capable of stabilizing any anhydride or oxyanion intermediates formed during catalytic turnover. Crystallization and Structure Determination of an FRC/Oxalyl-CoA Complex-A key implication of the kinetic mechanism is that either formyl-CoA or oxalyl-CoA can bind to the free enzyme, raising the possibility that these thioesters form an acylated enzyme intermediate prior to binding of the second substrate (see below). Although our site-directed mutagenesis experiments had implicated Asp 169 as having catalytic function, we still sought more direct evidence that the carboxylate side chain was participating in anhydride formation. We therefore performed an extensive series of screening studies to establish conditions under which thioesters might be sufficiently stable for crystallization of an FRC/oxalyl-CoA complex. Considerable experimentation resulted in crystals of the FRC/oxalyl-CoA complex, which were of sufficient quality for the structure of the complex to be determined to 2.6 Å resolution. Once again, the enzyme dimer was of very similar structure to that seen in our other studies, with an root mean square deviation of 0.50 and 0.55 Å for superimposition of the 854 backbone C ␣ atoms of the FRC/oxalyl-CoA complex with the FRC apoenzyme and FRC/CoA complex, respectively. To our surprise, however, analysis of the electron density revealed that rather than trapping the enzyme/product complex, we had crystallized the product of reaction between WT FRC and oxalyl-CoA, i.e. the putative oxalyl-anhydride intermediate (Fig. 7A). Observation of this reactive intermediate was likely possible because the active site in monomer B of the dimeric structure is protected by another dimer molecule in the crystal, which results in very good electron density for the active site amino acids, bound CoA, and the modified side chain of Asp 169 . In contrast, the active site present in monomer A is exposed, and the bound co-factor exhibits poor density and very high B values. Some extra density, consistent with some remaining anhydride, is observed close to the Asp 169 side chain of monomer A; however, this was modeled as a water molecule. The following discussion of the anhydride structure and its interactions with the enzyme is therefore based solely on the active site in monomer B, and residue numbers are assigned a letter (A or B) indicating the monomer in which they are located.
The oxalyl-aspartyl anhydride intermediate is stabilized by numerous interactions (Fig. 7A). The carbonyl oxygen is hydrogen-bonded to the side chain amide of Gln 17B , O-3 interacts with the backbone amide of Gln-17B, O-1 hydrogen bonds with the backbone amide of Glu-140B, and O-2 is stabilized by two hydrogen bonds to the amide of Gly A261 and to the backbone carbonyl of Gly A260 . Finally, although the oxalyl moiety can be modeled very well within the observed electron density (Fig.  7B), it appears that there is some continuous density connecting the oxalyl oxygen adjacent to the thiol group of the bound CoA. This finding, coupled with a 2.1 Å distance between sulfur and one oxygen atom of oxalate, suggests that a small fraction of oxalyl-CoA has not reacted with Asp 169 , which is consistent with the hypothesis that a mixture of active sites containing either intact oxalyl-CoA or free CoA and the anhydride is present in the crystal.

DISCUSSION
The catalytic mechanism of Class III CoA-transferases has been the subject of considerable speculation in light of the fact that these enzymes employ a different kinetic mechanism to that observed for Class I CoA-transferases (23). The situation is also exacerbated by the fact that Class III CoA-transferases differ widely in their substrate selectivity (23), thereby complicating the alignment of deduced primary structures and the identification of catalytically important residues (29). The xray crystallographic identification of a conserved aspartate positioned close to the thiol moiety of co-enzyme A bound in the putative active site of O. formigenes FRC has, however, given rise to the hypothesis that the catalytic mechanism of Class III CoA-transferases proceeds via intermediates in which Asp 169 is acylated (Fig. 5) (29,30). This proposal is consistent with isotopic exchange experiments that support the existence of similar anhydride intermediates in the catalytic mechanism of Class I CoA-transferases (51). On the other hand, the func-tional importance of Asp 169 in FRC was not experimentally established in previous studies of FRC (29), its E. coli homolog (30,31), or other Class III CoA-transferases (24 -28).
Previous steady state kinetic studies of O. formigenes FRC were limited by difficulties in preparing formyl-CoA in pure form and the need to employ a coupled assay rather than direct measurement of product oxalyl-CoA (14). We have resolved both of these problems by (i) developing a more efficient approach to preparing and purifying the thioester precursor of formyl-CoA and (ii) devising a sensitive HPLC-based assay for quantitating formyl-CoA, free CoASH and oxalyl-CoA. Replacing Asp 169 by alanine, glutamate, or serine residues gave three FRC mutants that exhibited significantly reduced or no detectable CoA-transferase activity in our HPLC-based assay. To ensure that Asp 169 was not merely playing a role in stabilizing the active site structure or unusual three-dimensional fold of the FRC dimer, we obtained high resolution x-ray structures for all three FRC mutants complexed with co-enzyme A. These structures showed that all three of the mutant enzymes were folded correctly and that there was no major disruption to the positions of other active site residues (Fig. 6). In the case of the D169E FRC mutant, the additional bulk of the glutamate side chain appears to correlate with increased disorder of the bound co-factor, suggesting that the reactive carbonyl group of formyl-CoA (or oxalyl-CoA) is positioned incorrectly for formation of the oxyanion leading to an acylated enzyme intermediate. The reasons underlying the lack of CoA-transferase activity for the D169S FRC mutant are less well defined because replacement of aspartate by serine increases the size of the active site cavity, and it is therefore likely that formyl-CoA can still bind in a reactive conformation. One explanation is that in the absence of an appropriate histidine residue, the serine hydroxyl is not sufficiently nucleophilic to attack the thioester to form an ester and free CoA. It is also possible, however, that the serine hydroxyl may not be positioned sufficiently close to the carbon atom for acyl transfer to take place. Third, in the event that acylation of the Ser 169 side chain gives a formate ester, the carbonyl group might then be incorrectly oriented or insufficiently reactive for subsequent reaction with oxalate in the ternary complex to give an anhydride intermediate capable of reacting with bound CoASH. We note that our efforts to detect the formylated serine residue using electrospray mass spectroscopy have so far proved unsuccessful. 4 Perhaps the most surprising observation, however, was the ability of the D169A FRC mutant to catalyze the synthesis of oxalyl-CoA (albeit at a 1300-fold reduced rate) despite lacking the critical active site carboxylate. Given that control experiments seem to rule out contamination by the YfdW gene product, the simplest explanation for this observation is that the transfer reaction proceeds by a different mechanism. One possibility is that oxalate can directly attack formyl-CoA in a ternary complex to give an oxalyl-formyl anhydride, which can then react with bound CoA to yield oxalyl-CoA and formate. No active site waters are observed in the active site of the D169A FRC mutant, and their presence as a result of the smaller side chain of Ala 169 would be expected to hydrolyze any bound thioester or reactive anhydride intermediate. In this regard, we do not observe increased formyl-CoA hydrolysis when this substrate is incubated with oxalate and the D169A FRC mutant. Moreover, in the absence of wild type FRC or the D169A FRC mutant, we observe no oxalyl-CoA formation when formyl-CoA and oxalate are incubated under our standard conditions, at least at levels above the detection limits of our HPLC-based assay. This finding is consistent with previous model studies on the rate of reaction of carboxylic acids with thioesters (52).
Although this site-directed mutagenesis provided the first indirect evidence for the functional role of Asp 169 in the mechanism of FRC-catalyzed CoA transfer, the crystallization of WT FRC with oxalyl-CoA at low temperature allowed us to trap an intermediate in which Asp 169 had been acylated by the thioester to give an oxalyl-aspartyl anhydride (Fig. 7A). This appears to be the first example of such an enzyme-bound anhydride intermediate and is probably a result of (i) stabilization of the anhydride by five hydrogen bonds and (ii) burial within an active site pocket that prevents the access of solvent. More specifically, two of the critical hydrogen bonds are to backbone amides in a loop segment comprised of residues Gly A258 to Gly A261 , suggesting that this closed conformation corresponds to that present during catalytic turnover, whereas the cognate loop in the other active site is in the open conformation. The fact that only one of the two active sites in the FRC dimer is fully occupied and has the closed conformation raises the possibility of "half-site" reactivity, i.e. both active sites cannot catalyze turnover simultaneously, as observed for E. coli CoAtransferase (53). We also note that recent studies of succinyl-CoA:3-oxoacid CoA-transferase have demonstrated that only one subunit in the dimer supports catalysis (54). On the other hand, this observation may merely be a crystallization artifact given that in the FRC/CoA complex (29), both active sites are occupied, and the two monomers have very similar conformation, including the flexible loop Gly 258 -Gly 261 .
Direct observation of this oxalyl-aspartyl anhydride intermediate, coupled with our site-directed mutagenesis studies, strengthens our proposal for the catalytic mechanism of FRC (29), which is likely to be employed by all Class III CoA-transferases. This finding, however, raises a number of questions that need to be addressed. First, why do the Class I and Class III CoA-transferases exhibit different kinetic behavior if both employ the same chemical strategy of forming anhydride intermediates? The answer to this problem is likely associated with the active site location of the free coenzyme-A that is produced upon anhydride formation (Fig. 5). In the Class I enzymes, bound CoA reacts with the initial anhydride in the absence of the second substrate to yield an active site thiol ester and the first product (55). The latter must then leave the active site prior to binding of the other substrate, giving rise to the observed ping-pong kinetics (32,50). In FRC, on the other hand, bound CoA appears to be positioned too far from the relevant carbonyl group of the Asp 169 side chain precluding the formation of an enzyme-linked thiol ester. Thus, release of the first product from the acylated enzyme intermediate requires the presence of the second substrate within the active site, resulting in sequential steady state kinetics.
A second issue concerns the timing of anhydride formation. Our kinetic data are consistent with the formation of a ternary complex, implying that oxalate might be present in the active site prior to reaction of Asp 169 with formyl-CoA. Such a mechanism would therefore preclude the hydrolysis of the anhydride intermediate prior to oxalate binding by removal of water, presumably by a conformational change in the contiguous loop region defined by residues A258 -A261. The fact that we observe anhydride formation in the absence of formate, however, suggests that oxalate could bind after the reactive intermediate is produced. The importance of the loop segment from the other FRC monomer in stabilizing the anhydride intermediate is also consistent with the results of dilution experiments (data not shown). Hence, lowering the FRC concentration results in a loss of specific activity and yields protein that gives a very broad peak in size exclusion chromatography experiments, which is centered about the K D value estimated for the enzyme monomer. We interpret these results to mean that the dimer dissociates at low concentration to give unfolded monomers that cannot catalyze CoA transfer. To date, we have been unable to identify conditions under which concentration of these solutions yields correctly refolded dimer with catalytic activity.
The final aspect of this hypothesis that remains to be defined is how the enzyme might stabilize three tetrahedral intermediates that have been proposed as intermediates in the catalytic mechanism (Fig. 5). One solution is to imagine the use of a hydrogen bonding moiety located upon a flexible side chain, and we have noted that Tyr 59 in FRC appears to be conserved throughout the Class III CoA-transferase family of enzymes (29), although our crystal structure of the enzymebound anhydride suggests that the side chain of Gln 17 could play this role. It is also possible that attack of the Asp 169 carboxylate on formyl-CoA proceeds via a concerted mechanism in which an oxyanion intermediate is not formed. We note that heavy atom kinetic isotope effect measurements (56) and theoretical studies of acyl transfer from thioesters (57) provide evidence for such a proposal. The resolution of this question, however, will likely require the use of 13 C and 18 O isotope effect measurements.
In summary, the work described herein provides strong evidence for the involvement of the Asp 169 side chain in mediating FRC-catalyzed CoA transfer via a series of anhydride intermediates. Given that this aspartate residue appears to be conserved in known Class III CoA-transferases (29,30), it is likely that the catalytic mechanism of formyl-CoA transferase is almost certainly employed by all other members of this enzyme family.