α-Methylacyl-CoA Racemase from Mycobacterium tuberculosis

α-Methylacyl-CoA racemase (Amacr) catalyzes the racemization of α-methyl-branched CoA esters. Sequence comparisons have shown that this enzyme is a member of the family III CoA transferases. The mammalian Amacr is involved in bile acid synthesis and branched-chain fatty acid degradation. In human, mutated variants of Amacr have been shown to be associated with disease states. Amino acid sequence alignment of Amacrs and its homologues from various species revealed 26 conserved protic residues, assumed to be potential candidates as catalytic residues. Amacr from Mycobacterium tuberculosis (MCR) was taken as a representative of the racemases. To determine their importance for efficient catalysis, each of these 26 protic residues of MCR was mutated into an alanine, respectively, and the mutated variants were overexpressed in Escherichia coli. It was found that four variants (R91A, H126A, D156A, and E241A) were properly folded but had much decreased catalytic efficiency. Apparently, Arg91, His126, Asp156, and Glu241 are important catalytic residues of MCR. The importance of these residues for catalysis can be rationalized by the 1.8 Å resolution crystal structure of MCR, which shows that the catalytic site is at the interface between the large and small domain of two different subunits of the dimeric enzyme. This crystal structure is the first structure of a complete enzyme of the bile acid synthesis pathway. It shows that MCR has unique structural features, not seen in the structures of the sequence related formyl-CoA transferases, suggesting that the family III CoA transferases can be subdivided in at least two classes, being racemases and CoA transferases.

␣-Methylacyl-CoA racemase (Amacr) 1 catalyzes the racemization of (S)-and (R)-enantiomers of a wide spectrum of ␣-methyl-branched carboxyl coenzyme A thioesters (1). In mammals, such substrates are derived from (i) C 27 -bile acid intermediates or (ii) branched-chain fatty acids (pristanic acid), but also 2-arylpropionic acids (ibuprofen), used as non-steroidal anti-inflammatory drugs (Fig. 1), are substrates. It has been shown that Amacr is localized in both peroxisomes and mitochondria (2)(3)(4), where racemization of (␣R)-methylacyl-CoA esters is a condition for their subsequent degradation in ␤-oxidation (5,6). Consequently, Amacr is required for bile acid synthesis (7) and ␤-oxidation of ␣-methyl-branched fatty acids (8). In mouse, Amacr is an unmodified product of a single gene and it carries both N-and C-terminal targeting sequences for input into mitochondria and peroxisomes, respectively (2). Recently, mutated variants of human Amacr have been shown to be associated with various disease states. Amacr deficiency caused by S52P and L107P sequence changes in human Amacr results in adult-onset sensory motor neuropathy (9), vitamin K deficiency and retinopathy (10), and also in neonatal liver dysfunction (11). Using a knock-out mouse strain, it could be shown that Amacr is essential for detoxification of alimentary phytol, whereas alternative minor pathways for generation of bile acids suffice for survival of Amacr-deficient individuals (12). Elevated levels of Amacr have been found to correlate with human prostate cancer (13,14), kidney tumors (15), and colon carcinomas (16).
Amacr is a cofactor-independent racemase, which interconverts the R and S-chirality of the 2-methyl carbon atom of a 2-methyl-thioester molecule. Currently there is no structural information on this class of enzymes. Detailed structural enzymological studies of other cofactor independent racemases and epimerases have been reported, for example on mandelate racemase (23), L-Ala-D/L-Glu-epimerase (24), aspartate racemase (25), and D-ribulose-5-phosphate 3-epimerase (26), which are all proposed to work via the initial abstraction of a proton from the chiral carbon atom. Each of these enzymes catalyze an overall 1,1-proton transfer reaction and have two catalytic protic residues, which are required for (i) the initial proton abstraction from the substrate and for (ii) the subsequent proton donation to the intermediate. The protic residues in these examples are a lysine/histidine pair, two lysines, two cysteines, and two aspartates in, respectively, mandelate racemase, L-Ala-D/L-Glu epimerase, aspartate racemase, and the ribulose epimerase.
Interconversion of the ␣-methyl group and the ␣-hydrogen atom in the Amacr-catalyzed racemization is also presumed to occur via a two-base mechanism involving two protic residues (1). To test this presumption, we mapped here the conserved protic amino acid residues in Amacr and related proteins by a multiple amino acid sequence alignment. Subsequently, an Amacr homologue from the eubacteria Mycobacterium tuberculosis, referred to as MCR, was taken as a model protein.
M. tuberculosis is the causative organism of tuberculosis. This disease is widespread, causing the death of approximately 2 million human beings every year (27,28). Further understanding at the atomic detail of its metabolism will contribute much toward discovering specific drugs for curing this disease. In this respect it is noteworthy that it has been shown that MCR plays a critical role in the ␤-oxidation of methylbranched alkanes in Mycobacterium (29).
MCR has a high sequence identity with mouse, rat, and human Amacr of 41, 44, and 43%, respectively. Preliminary studies (30)  In the biochemical studies reported here conserved protic amino acid residues in MCR were mutated in turn to alanine and the mutated variants were overexpressed and characterized in vitro. Four protic residues, which are important for efficient racemization, were identified. Complementary to these studies the structure determination of MCR at 1.8 Å resolution is reported. This is the first structure of a complete enzyme of the bile acid synthesis pathway. The importance of the four protic residues for binding and catalysis can be rationalized from this structure.

EXPERIMENTAL PROCEDURES
Sequence Comparisons-We took as a working hypothesis that the racemization is conducted via acid-base transfer reaction(s), catalyzed by protic residues in Amacrs. Sequence alignments were used to select potential protic catalytic residues. Altogether 28 sequences were collected from the Swiss-Prot data base using Blast and Fasta algorithms and aligned using the Clustal X program (32). This set includes active bacterial and mammalian Amacr sequences, as well as Amacr-related sequences. The obtained sequence alignment suggested the positions of conserved protic amino acid residues (histidine, lysine, arginine, aspartic acid, glutamic acid, tyrosine, serine, and cysteine) (see supplemental Fig. 1). These residues were regarded as potential catalytic residues and selected for mutagenesis.
Cloning and Expression-The MCR gene (from M. tuberculosis) was cloned into expression vector pET3␣ (Novagen, Madison, WI) and used as a template for PCR reactions for generating the mutants with MCRspecific primers (Amersham Biosciences, Buckinghamshire, UK) using a QuickChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA). The primers were designed so that the desired conserved protic residue would be changed to alanine (see supplemental Table I). Also the two variants corresponding to the known disease causing missense mutations were made. The plasmids, containing the MCR and mutated variants, were transformed into supercompetent E. coli cells, supplied with the mutagenesis kit. Bacteria were grown, plasmid DNAs were extracted, and introduced mutations were verified with automated sequencing with ABI PRISM 5700 (Amersham Biosciences). Extracted plasmids were transformed into BL21(DE3)pLysS cells and the mutated MCRs were overexpressed in Luria Broth medium containing 50 mg/liter ampicillin and 34 mg/liter chloramphenicol with 2% isopropyl ␤-D-thiogalactopyranoside induction, as described previously for the wild-type MCR (30).
Activity Measurements-The interconversion of ␣-methylacyl-CoA by MCR and rat Amacr was measured with (S)-2-methylmyristoyl-CoA as substrate, with the gas-liquid chromatography method as described previously (1). The enzyme assay for measuring enzyme activities of wild-type MCR and its variants was done by measuring the release of 3 H from [2-3 H]pristanoyl-CoA (20 M, 6.45 Ci/mol) as described (1). Briefly, the appropriately diluted enzyme sources were incubated with the labeled racemic substrate in 50 l of 50 mM Tris-HCl, pH 8.0, 0.1% octylglycoside and 10 g BSA at 37°C for 30 min. The reactions were terminated with 450 l of 1% trichloroacetic acid, and [ 3 H]H 2 O was separated from unreacted substrate on reverse-phase silica gel (RP-18) and quantified by liquid scintillation counting. For the determination of V max and K m the substrate concentration was 56.0, 75.6, 164.5, and 200 M respectively. The activity is expressed in units (mol/min) per mg of protein.
Western Blot Analysis-For the determination of the presence of wild-type MCR and its variants in the supernatants of lysed induced E. coli cells, samples of each supernatant were applied on SDS gels. Western blots were made using as primary antibody rabbit anti-mouse Amacr antibody, which cross-reacts with MCR. Racemase activities of the unpurified MCR mutants were obtained by activity measurements at a substrate concentration of 20 M, using diluted supernatants as enzyme source. The observed enzyme activities were normalized with respect to the amount of racemase in the supernatant by scanning of the Western blot gels and by quantification of the bands using the software package Quantity One (Bio-Rad).
Purification and CD Spectroscopy-The overexpressed wild-type and mutated MCRs were purified from E. coli extracts as discussed (30). The resulting samples were further applied onto a size exclusion Superdex 200® HR 10/30 column (in equilibrium with 10 mM sodium phosphate, pH 8.8, containing 1 mM reduced dithiothreitol, 1 mM EDTA, and 1 mM NaN 3 ; the measuring buffer), and the fractions containing MCR were pooled and concentrated. Absorption at 280 nm was used for adjustment of the protein concentrations of the samples for CD spectroscopy, which was carried out using a Jasco J710 spectropolarimeter at 22°C. The far-UV spectra of the proteins (0.1 mg ml Ϫ1 ) were measured from 200 to 250 nm in measuring buffer, with the following instrument settings: response, 2 s; sensitivity, 100 millidegrees; speed, 50 nm/min; average of 30 scans. The molar ellepticities were calculated with Jasco software.
Crystallization and Data Collection-Well diffracting crystals of the wild-type MCR were grown at 22°C by the hanging-drop, vapor-diffusion method (reservoir 1.52 M ammonium phosphate and 10 mM BaCl 2 at pH 7.0) in drops of 2-l volume containing 1:1 (v/v) mixture of reservoir and protein solution (6 mg ml Ϫ1 in 10 mM potassium phosphate buffer, pH ϭ 8.8). These well diffracting crystals, which were only the minor crystal form in the initial crystallization experiments (30) corresponding to a solvent content of 55% (33). The heavy atom derivatives of methyl mercuricacetate (MMA, C 3 H 6 HgO 2 ) and monosodium salt of p-chloromercuriphenylsulfonic acid (PCMPS, C 6 H 4 ClHgO 3 SNa) were obtained by soaking native protein crystals in a solution of heavy atom compounds in mother liquor at a concentration of 1.0 mM for 30 -40 min.
All the data sets were collected under cryogenic conditions using a synchrotron radiation source (EMBL-outstation at DESY, Hamburg, Germany). Crystals were transferred from the crystallization drop to the cryoprotectant solution (mother liquor containing also 24% glycerol) using nylon loops and immediately flash-frozen in a cold nitrogen gas stream at 100 K just before data collection. For the data collection of the heavy atom derivatised crystals, the crystals were frozen using the same cryoprotectant solution, containing also 1.0 mM heavy atom derivative. The data collection was performed by the oscillation method using a CCD detector. The native data set was collected using beamline X11 ( ϭ 0.8122 Å) and the derivative data sets were collected using beamline X13 ( ϭ 0.8042 Å). The data sets were processed using program packages DENZO (34), XDS (35), and CCP4 (36).
Structure Determination and Structure Refinement-The initial phases were obtained with the MIRAS method. The program SOLVE (37) was used for finding and refining the heavy atom positions. 7 and 8 sites were used in the MIRAS calculations for the MMA and PCMPS data sets, respectively. The mean figure of merit of the refined phases was 0.50 at 2.5 Å resolution. Extensive fragments of each dimer were built in this map by the automatic model building procedures of RE-SOLVE (38). By superimposing these fragments on each other the matrices relating the four subunits were obtained and 4-fold averaging plus phase extension till 1.8 Å was carried out with DM (39). Then the improved electron density map at 1.8 Å resolution was used for automated model building with RESOLVE. The program could build about 60% of the side chains of all four subunits. Then iterative cycles of model refinement with REFMAC5 (40) and model building in electron density maps using program O (41) were carried out. Initially very tight NCS restraints were used for the refinement, but slowly the restraints were released as the model was getting more complete and at the end of the refinement only loose NCS restraints were used. The overall anisotropy was modeled with TLS parameters, using only two TLS groups, being the two dimers of the asymmetric unit. PRO-CHECK (42) was used to monitor the stereochemistry of the model. In the final model six residues (1 and 40 -44) are missing from all four subunits; these residues could not be built because of lack of features in the electron density map. All other residues are well defined in the map. There are no structural differences between the four monomers; for example the r.m.s difference for the 354 common C␣ atoms of the A-subunit of one of the dimers with each of the three other monomers is 0.2 Å. There are 1457 waters, 1 phosphate ion, and 15 glycerol molecules in the model. The final refinement statistics are presented in Table I.
Structure Analysis-The structure has been analyzed with the programs PROCHECK and O. The secondary structure assignment has been done with DSSP (43). Structure superpositions have been done with the LSQ option of program O, using the corresponding C␣ atoms. The images of Figs. 4, 6, 7, and 8 were made with Pymol (44).

Selection of the Conserved Protic Amacr
Residues-Alignment of amino acid sequences of 28 Amacrs and Amacr related sequences (see supplementary Fig. 1) suggested 26 conserved positions of protic amino acid residues. MCR was taken as a representative for Amacrs due to its stability and amenability to purification, because in contrast to the mammalian enzyme, it could easily be expressed in active form in E. coli. The racemase activity of MCR, measured as the interconversion of the (S) to the (R) form of 2-methylmyristoyl-CoA, when incubated with 8 mM (S)-2-methylmyristoyl-CoA is visualized in Fig. 2. Following the working hypothesis of acid base-based catalysis, and using the sequence alignment, 26 protic residues in MCR (listed in Table II) were mutated via site-directed mutagenesis to alanine and tested in vitro. Additionally, two other mutants (I56P, M111P, see Table II) were included in this analysis as they correspond to mutations in human patients (S52P, L107P) suffering from adult-onset sensory motor neuropathy and neonatal liver malfunction (9,11).
Properties of the Wild-type and Mutated MCRs-The wildtype and mutated MCRs were produced as recombinant proteins via expression from the appropriate plasmids in E. coli. The appearance of polypeptide bands in SDS-PAGE gels at 39 kDa, corresponding to the predicted size of MCR, indicated that the polypeptides were produced (see supplemental Fig. 2). This band was not present using cells transformed with empty plasmids. When measured from the soluble extract of overexpressing E. coli strains, nine of the MCR variants in which protic residues had been changed into alanine (R52A, E82A, R91A, H126A, D156A, D190A, E241A, C297A, and H312A) had Amacr activity less than 20%, as compared with the expression of wild-type MCR (Table II). These variants were selected for further characterization, to identify the catalytically critical amino acid residues. For these variants also the normalized activity in the cell lysate was determined, as described under "Experimental Procedures" and as listed in Table II, confirming that for each of these variants the activity is much lower as for wild type. For the disease-related variants I56P and M111P the normalized activities were found to be 76% and 1.6%, respectively.
When isolating the wild-type MCR, 12 mg of purified protein was obtained from an E. coli pellet of 11.5 g. The size exclusion chromatography yielded native molecular mass of 89 kDa for MCR, whereas SDS-PAGE showed the polypeptide to have the molecular mass of 39 kDa indicating that the recombinant protein is a dimer. Also dynamic light scattering measurements indicated that the purified MCR was monodisperse and a dimer (30). The MCR variants R91A, H126A, D156A, E241A, C297A, and H312A could be purified using the protocol applied for the wild-type MCR. The retention volumes of these six purified variants on the Superdex HR column indicated that they were also dimeric proteins. Although the variants R52A, E82A, M111P and D190A were present in the bacterial extracts, as revealed by the 39-kDa bands in SDS-PAGE and by the Western blotting analysis, their chromatographic purification failed despite several attempts (Table II). It was considered that their native conformations were not maintained due to the mutation, reflected also by the loss of enzymatic activity.
When the six purified MCR variants were subjected to CD spectroscopy, the far-UV region (200 -250 nm) spectra for wildtype, R91A, H126A, D156A, and E241A MCRs were practically  b The normalized specific activities in the bacterial lysate supernatants are racemase activities per quantified immuno detected protein bands (see "Experimental Procedures").
c The corresponding variants of human Amacr are associated with Amacr deficiency in human.

FIG. 3. CD spectroscopy of MCR and its variants.
The far-UV spectrum of wild-type MCR is compared with the CD spectra of (a) the D156A variant, (b) the C297A variant, and (c) the H312A variant. The spectra of R91A, H126A and E241A variants are completely overlapping with the spectra as shown in a. identical. However, two of the variants, C297A and H312A, had different CD spectra compared with wild-type MCR (Fig. 3), demonstrating changes in their secondary structure elements, which correlate with the decrease in enzymatic activity (Table  II). At this point it was concluded that the variants R91A, H126A, D156A, and E241A were properly folded indicating that the lower catalytic efficiency was due to the importance of these side chains for binding of the substrate or for catalysis or for both. Subsequently, the kinetic constants of these four purified variants were determined. For wild type the V max and K m values are 214 units/mg and 41 M, respectively. The V max values for the variants H126A, D156A, and E241A are very low (less than 0.1 unit per mg), but the estimated K m values for these variants are similar as for wild type. For the variant R91A the only effect is a 9-fold increase of the K m , as compared with wild type.
Overall Structure of Wild-type MCR-The crystal structure of apo MCR has been determined at 1.8 Å resolution. The overall fold of the monomer and its secondary structure are depicted in Fig. 4. MCR has a similar fold as FRC from Oxalobacter formigenes (19,20) and the E. coli YfdW protein (21,22). Fig. 5 shows the sequence and secondary structure comparison with FRC. Each monomer of MCR consists essentially of two domains, the N-terminal large domain and the small FIG. 6. The structure of the interlocked dimer. A, stereo picture of the MCR dimer. The subunits are colored green (subunit A) and orange (subunit B). The local 2-fold NCS runs vertically, near the active site helices ␣7 of both subunits. The active site is near the N termini of helices ␣5 and ␣7, which are explicitly labeled for subunit A (green) at their N termini. In this view it can be seen that the C-terminal residues of subunit A (following after its small domain) meander back to the large domain, covering helices ␣5, ␣6, and ␣8 of subunit B (orange). ␤8 labels the middle ␤-strand of the small domain of subunit B. B, superposition of the MCR dimer on the FRC-dimer (slightly rotated view with respect to the view in a). MCR is in green. FRC is colored purple and blue (for the long insertions). The labels S and L mark the small (after ␤2) and the large (in the first linker region, after ␣8) insertions of FRC. N, C, and * (near the N terminus of ␣5) label the N terminus, C terminus, and active site of subunit A (green) of a. domain, two linker regions (␣8 and ␣12-␤11, respectively) that connect the two domains (Fig. 4) and the C-terminal region (␣14). The C-terminal region (residues 336 -360, ␣14) is associated with the large domain (Fig. 4). The small domain consists of residues 224 -300, ␤7-␤9. The core of the large domain has an open ␣/␤-sheet structure with a Rossmann-like fold consisting of a central six-stranded parallel ␤-sheet (␤1-␤6) with helices packed to both sides of it. The loop after ␤2 is disordered. ␤5 and ␤6 are connected by three helices: ␣5, ␣6, and ␣7. Helices ␣5 and ␣6 are buried between the large domain and the small domain. After ␣6, there is a long loop consisting of 13 residues leading back to the active site helix, ␣7, which belongs to the core of the large domain. Subsequently, ␤6 completes the ␣/␤ structure of the large domain. After ␤6 a kinked helix follows, helix ␣8, which is the first linker region between the large domain and the small domain. The core of the small domain is a three-stranded antiparallel ␤-sheet which starts and ends with ␤-strands ␤7 and ␤9, respectively. On the bulk-solvent side this ␤-meander is covered by three helices (␣9, ␣10, and ␣11), which are inserted between the second and third strand. After ␤9, the second linker region (␣12-␤11) connects the small domain back to the large domain, such that a large hole is created by the folded polypeptide chain of one subunit (Fig. 4). The connecting, kinked helix ␣8 of the second subunit, as well as the protruding helices ␣5 and ␣6 of the second subunit, are passing through this hole, making a very tight, interlocked dimer, in which the two subunits are related by a local 2-fold axis (Fig. 6).
The Comparison of MCR with FRC and YfdW-The pairwise sequence identity between MCR and FRC and YfdW is 24 and 25%, respectively. FRC and YfdW, which have a 61% sequence identity, are close homologues, and both are ϳ60 residues longer than MCR. The sequence and structure comparisons (Figs. 5 and 6) show that FRC has two long loops, which are not present in Amacr: (i) 13 residues after ␤2 and (ii) 27 residues after ␣8 (the kinked linker helix); in addition FRC has an extra C-terminal helix (␣20). For the superposition, as shown in Fig.  6 the ␤-strands of the large and small domain, as well as the ␣-helices ␣1, ␣2, ␣5, and ␣7 of the large domain, of both subunits were used. The r.m.s. difference for the 198 corresponding C␣ atoms is 1.4 Å; the r.m.s. differences for just the large or the small domain superpositions are 0.8 Å (for 87 C␣ atoms) and 0.7 Å (for 12 C␣ atoms), respectively. As shown in Fig. 6, the two extra loops of FRC are near the active site region of this enzyme. Inspection of Fig. 6 also shows that the quaternary structure of the interlocked dimer is very well preserved between MCR and FRC. The tertiary structure of the large domain is also well conserved; however, in the small domain the helical part is different (Fig. 6). The sequences of the large domain are also much more conserved than the small domain sequences, having sequence identities of 31 and 13% for the large domain (residues 1-187) and the small domain (residues 224 -300), respectively.

DISCUSSION
The mutagenesis studies of MCR have identified four protic residues which are important for efficient catalysis: Arg 91 , His 126 , Asp 156 , and Glu 241 . Mutating these residues to alanines does not interfere with proper folding, but the catalytic efficiency of the respective alanine variants is much decreased. Fig. 7 shows the location of these residues in the structure of MCR. Arg 91 and Asp 156 are conserved across the family III CoA transferases; however, His 126 and Glu 241 are unique for the racemases (see supplemental Fig. 1). Arg 91 and Asp 156 of MCR are equivalent to Arg 104 and Asp 169 of FRC, respectively. From the structure of the FRC-CoA complex it has been reported that Arg 104 stabilizes the binding of the CoA moiety, and Asp 169 is the catalytic aspartate (19,20), suggesting that in MCR Arg 91 and Asp 156 are important for binding and catalysis, respectively. This is in good agreement with the kinetic data of its R91A and D156A variants. The H126A and E241A variants show much reduced rates but similar K m values, indicating that also His 126 and Glu 241 , like Asp 156 , are important catalytic residues in MCR.
The mode of binding of CoA to FRC and YfdW is similar, and when using the structure of YfdW complexed with acetyl-CoA (21) to map the predicted mode of binding of CoA to MCR (Fig.  7), it is seen that important features of the FRC/YfdW-CoA binding pocket are preserved in MCR. For this superpositioning the C␣ atoms of the ␤-strands of the large domain were used, as well as the ␣-helices ␣1, ␣2, ␣5, and ␣7 of the large domain, as outlined under "Results." Several residues described to be involved in the binding of CoA to FRC (19) are seen to be conserved. For example, in FRC a stacking interaction is observed between Arg 38 and the adenine moiety. The equivalent residue in MCR is Arg 38  for efficient catalysis in MCR is indeed confirmed by the mutagenesis studies. Interestingly, the residues interacting with the pantetheine moiety are less well conserved, for example Met 44 and Tyr 139 (interacting with the pantetheine dimethyl group) are not conserved in MCR, and Tyr 139 (interacting with pantetheine hydroxyl group) is a histidine in MCR. From the FRC-MCR comparison it is also predicted that Asp 156 will be a catalytic residue in MCR, which nicely agrees with the mutagenesis data. The mutagenesis data also points to His 126 and Glu 241 as being important for efficient catalysis. His 126 is in the large domain, at the N terminus of helix ␣5, while Glu 241 is in the small domain, just after ␤8. Apparently, the catalytic site is at the interface between the large and the small domain of two different subunits of the dimer, near the N terminus of helix ␣5 (Fig. 7).
The dimeric assembly of the two subunits results in an interlocked mode of interaction between the two polypeptide chains, because the C-terminal tail of one subunit folds back on its own N-terminal large domain, after a long excursion in which the small domain is formed (Fig. 6). The structure of MCR also allows us to discuss how the point mutations S52P and L107P of human Amacr could be related to deficiencies. The equivalent residues in MCR are Ile 56 and Met 111 . Ile 56 is in the large domain in the outer ␤-strand (␤3) of its parallel ␤-sheet, which interacts with the C-terminal tail (Fig. 8). The I56P variant has similar activity as wild type. Further studies are required to understand the deficiency caused by the S52P mutation of human racemase. Met 111 is at the C-terminal end of ␤5 pointing to Pro 18 of the ␤1-␣1 loop, very close to the active site and rather inside the folded protein (Fig. 8). The positioning of Met 111 near the catalytic site correlates very well with the observation that the corresponding L107P mutation in human Amacr results in a deficient racemase. Fig. 8 also visualizes the position of the other protic residues, which, when mutated into alanine, caused such structural changes that either the purification is affected (Arg 52 , Glu 82 , Asp 190 ) or the CD spectra are different (Cys 297 , His 312 ). Cys 297 , His 312 , and Arg 52 stabilize the interactions between the C-terminal tail and the rest of the protein, whereas Glu 82 and Asp 190 are rather buried polar residues stabilizing the fold of the N-terminal domain.
Combining the mutagenesis and enzymological data together with the structural homology of MCR with the FRC/ YfdW structures it is predicted that the catalytic site of MCR is near Asp 156 and His 126 (Fig. 7). What could be the role of His 126 and Asp 156 ? The original assumption of this study was that possibly two protic residues would be involved in the catalysis, as proposed by analogy with other racemases. From the combined structural and enzymological data it can be postulated that Asp 156 functions as such a catalytic base. Binding studies with substrate or its analogues are required to establish such a mechanism and also if Asp 156 is the only base. In the latter mechanism this would concern removing the proton from one side of the C-2 carbon and adding it back at the other side after a structural rearrangement of the substrate (as has been discussed for other racemases (45,46)). Alternatively another side chain, for example the side chain of His 126 or possibly a water molecule is involved as the second protic residue to achieve the 1,1-proton exchange reaction.
The bound acetyl-CoA as seen in YfdW shows two modes of binding for the acetyl group (Fig. 7). In the MCR structure the corresponding binding modes can be described as the acetyl moiety pointing upwards (to the ␤1-␣1 loop) or downwards (to His 126 ) (Fig. 7). It is interesting to note that the upwards binding mode of the acetyl moiety, as seen in YfdW, is not favorable in MCR due to potential clashes with the Pro 18 side chain in the MCR ␤1-␣1 loop (which is a small side chain residue in FRC/YfdW). Interestingly, a small rearrangement of the downward orientation of the acetyl moiety brings the thioester oxygen in a potential oxyanion hole formed by ND1(His 126 ) and N(Asp 127 ) at the N terminus of helix ␣5. The possible importance of the corresponding thioester oxygen atom interaction in the YfdW complex with N(Glu 140 ) for the YfdW reaction mechanism has been discussed (21). In the current apostructure of MCR the oxygen atom of a glycerol molecule is bound in this putative oxyanion hole, formed by ND1(His 126 ) and N(Glu 127 ). Such oxyanion holes, facilitating thioester-dependent catalysis, are commonly found in CoA-dependent enzymes (47).
The importance of His 126 for the catalytic function is also nicely consistent with the importance of Glu 241 for catalysis, as the Glu 241 side chain (of subunit B) is hydrogen-bonded to His 126 (of subunit A) via a good hydrogen bond between OE1(Glu 241 ) and NE2(His 126 ) of 2.7 Å (Fig. 7). Apparently, the Glu 241 side chain enhances the catalytic properties of the histidine side chain. The His-Glu pair in MCR is not conserved in FRC, as the corresponding residues in FRC are Tyr 139 and Ala 285 , which are not functional homologues, in this respect. Also, in the liganded FRC structure the N terminus of helix ␣5 is covered by the flexible GGGG motif of residues 258 -261, while in MRC the corresponding loops are shorter (Fig. 6). The differences in active site properties agree with the differences in catalytic properties of MCR and FRC, being that MCR is a racemase and FRC is a CoA transferase. Also the substrate specificities are different. In MCR the binding pocket is much more extended and solvent exposed, due to the shorter loops near the active site (Fig. 6), which agrees with the large size of its substrate molecules, being steroid or pristanoyl moieties (Fig. 1).
These studies confirm that MCR belongs to the family III CoA transferases, despite very low sequence similarities, especially in the small domain. Apparently this family covers a wide range of substrate and catalytic specificities as MCR and FRC are very different in both respects. MCR is a CoA-racemase, which acts upon the ␣-methyl group of acyl moieties of CoA-fatty acids, which are large and hydrophobic (Fig. 1), while FRC is a CoA transferase, which transfers the CoA moiety between small polar acyl groups. Current studies on MCR are aimed at determining the precise mode of binding of the substrate to MCR as well as understanding its reaction mechanism.