Structural and Functional Basis for (S)-Allantoin Formation in the Ureide Pathway*

The ureide pathway, which mediates the oxidative degradation of uric acid to (S)-allantoin, represents the late stage of purine catabolism in most organisms. The details of uric acid metabolism remained elusive until the complete pathway involving three enzymes was recently identified and characterized. However, the molecular details of the exclusive production of one enantiomer of allantoin in this pathway are still undefined. Here we report the crystal structure of 2-oxo-4-hydroxy-4-carboxy-5-ureidoimidazoline (OHCU) decarboxylase, which catalyzes the last reaction of the pathway, in a complex with the product, (S)-allantoin, at 2.5-Å resolution. The homodimeric helical protein represents a novel structural motif and reveals that the active site in each monomer contains no cofactors, distinguishing this enzyme mechanistically from other cofactor-dependent decarboxylases. On the basis of structural analysis, along with site-directed mutagenesis, a mechanism for the enzyme is proposed in which a decarboxylation reaction occurs directly, and the invariant histidine residue in the OHCU decarboxylase family plays an essential role in producing (S)-allantoin through a proton transfer from the hydroxyl group at C4 to C5 at the re-face of OHCU. These results provide molecular details that address a longstanding question of how living organisms selectively produce (S)-allantoin.

The ureide pathway, which mediates the oxidative degradation of uric acid to (S)-allantoin, represents the late stage of purine catabolism in most organisms. The details of uric acid metabolism remained elusive until the complete pathway involving three enzymes was recently identified and characterized. However, the molecular details of the exclusive production of one enantiomer of allantoin in this pathway are still undefined. Here we report the crystal structure of 2-oxo-4-hydroxy-4-carboxy-5-ureidoimidazoline (OHCU) decarboxylase, which catalyzes the last reaction of the pathway, in a complex with the product, (S)-allantoin, at 2.5-Å resolution. The homodimeric helical protein represents a novel structural motif and reveals that the active site in each monomer contains no cofactors, distinguishing this enzyme mechanistically from other cofactordependent decarboxylases. On the basis of structural analysis, along with site-directed mutagenesis, a mechanism for the enzyme is proposed in which a decarboxylation reaction occurs directly, and the invariant histidine residue in the OHCU decarboxylase family plays an essential role in producing (S)-allantoin through a proton transfer from the hydroxyl group at C4 to C5 at the re-face of OHCU. These results provide molecular details that address a longstanding question of how living organisms selectively produce (S)-allantoin.
The purine degradation pathway has an essential role in nitrogen metabolism in most organisms. In this catabolism pathway, inosine monophosphate, which is the final product of de novo purine biosynthesis, is degraded through sequential enzymatic steps into uric acid, which contains a high level of nitrogen (1). Most organisms, including some bacteria, plants, and animals, utilize a common pathway for uric acid metabolism and produce stereospecific (S)-allantoin as the final product (2,3). Subsequently, (S)-allantoin, which contains an even ratio of nitrogen to carbon, is used as a nitrogen source through further enzyme-dependent degradation. This metabolism of uric acid, a process named the ureide pathway, has a pivotal role in transforming the nitrogen that is fixed in leguminous plants (2,3) and also plays a crucial role in some bacteria under nitrogen-limited conditions (4). This pathway was once thought to be executed by a single enzyme, urate oxidase (5), but recent investigations have revealed two additional enzymes in the pathway (Scheme 1). The pathway is initiated by urate oxidase, producing the unstable 5-hydroxyisourate (HIU), 3 which has a half-life of ϳ30 min in aqueous solution, followed by hydrolysis by HIU hydrolase to produce OHCU, which also undergoes spontaneous degradation (6 -8). The third enzyme, OHCU decarboxylase, was identified through the phylogenetic analysis of a whole genome and has been shown to catalyze the decarboxylation of OHCU, producing the stereospecific (S)-allantoin (8). These observations are consistent with the previous identification of two chemically distinct labile intermediates produced by urate oxidase (9) but differ in that racemic allantoin is produced in the nonenzymatic decomposition of HIU (10).
Because the three enzymes in the pathway, produced from individual genes, are present in most organisms, the proteins belong to a large family of enzymes. However, Arabidopsis thaliana and Bacillus subtilis possess a bifunctional enzyme in which two of the three enzymes are fused into a single polypeptide (8). Structural and biochemical studies of urate oxidase (11) and HIU hydrolase (12-16) revealed a novel arrangement of the active site residues and possible mechanisms for these enzymes. However, the biochemical characteristics of OHCU decarboxylase are still unknown, except for the findings that the enzyme is unrelated in sequence to other known decarboxylases and has been implicated in regulating plant growth by interacting with a component of the brassinosteroid receptor (17), indicative of its unusual structural and functional properties. In this study, we identified novel structural and functional features of the A. thaliana and B. subtilis OHCU decarboxylases. Given the sequence similarities of OHCU decarboxylases from many organisms, other members of this family in prokaryotes and eukaryotes likely function similarly to the proposed mechanism.

EXPERIMENTAL PROCEDURES
Protein Purification and Crystallization-For the expression of His-tagged recombinant A. thaliana OHCU decarboxylase, a portion of the cDNA (483 bp; residues 1-161) encoding the A. thaliana bifunctional enzyme was amplified using two sequence-specific primers ( Table 1). The PCR product was inserted between the NdeI and BamHI sites of a version of pET28 (Novagen) that had been modified to contain a tobacco etch virus cleavage site between the N-terminal His tag and the multiple cloning site. Subsequently, Escherichia coli BL21-CodonPlus(DE3) harboring the plasmid for A. thaliana OHCU decarboxylase was grown at 37°C in LB and induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside for 16 h at 20°C. The cells were harvested and sonicated in buffer A (100 mM potassium phosphate, 10 mM ␤-mercaptoethanol, pH 7.6). The N-terminal His-tagged enzyme was purified using immobilized metal affinity chromatography with buffer B (buffer A plus 500 mM imidazole) followed by size-exclusion chromatography (Superdex-200, GE Healthcare) with buffer A, and subsequently concentrated to ϳ4 mg/ml. Selenomethionine-substituted protein was prepared as above except that the constructed plasmid was transformed into E. coli B834(DE3) methionine auxotroph cells (Novagen). Crystallization was performed using the hanging drop, vapor-diffusion method at 22°C. Crystals of OHCU decarboxylase were produced using the selenomethionine-substituted protein in the presence of 20 mM racemic allantoin with a crystallization buffer of 0.1 M HEPES (pH 7.5), 20 mM MgCl 2 , and 16% polyacrylic acid 5100.
Data Collection and Structure Determination-Multiwavelength anomalous and single-wavelength diffraction data to 2.7 and 2.5 Å, respectively, were collected on beamlines 4A and 6C at the Pohang Accelerator Laboratory (Pohang, Korea) using crystals of selenomethionine OHCU decarboxylase complexed with 20 mM allantoin. Data were collected at 100 K, and the crystals were cryoprotected by adding 20% glycerol to the crystallization solution. The collected data were processed using HKL2000 (18). The program SOLVE/RESOLVE (19,20) was used for phasing and density modification of the complex struc-ture using the 2.7-Å data. The identification of five selenium sites was sufficient to trace most of the residues, excluding residues 1-5 and 66 -78. Manual model building was performed using O (21), and refinement was performed using CNS against the 2.5-Å resolution data (22). The density for allantoin was clearly identified in an F o Ϫ F c map, and both (S)-and (R)allantoin were initially modeled, but the (S)-form corresponded to the density. The details of the data collection and refinement of the structure are presented in Table 2. No residues are in the disallowed region defined by the program PROCHECK (23). The figures were prepared using PyMOL, 4 and the structure was analyzed using the CCP4 suite (23).
Site-directed Mutagenesis and Activity Assays-Functional analysis was carried out using HIU hydrolase and OHCU decarboxylase from B. subtilis. Urate oxidase from Candida sp. was purchased from Sigma. A gene encoding the mutation site in OHCU decarboxylase (residues 1-170) was generated by PCR using specifically designed primers (Table 1), and the sequence was verified by DNA sequencing. The proteins were expressed as N-terminal His-tagged proteins and purified as described above for the wild-type and mutant OHCU decarboxylases and as described previously for HIU hydrolase (14). The purified enzyme was concentrated to ϳ1.0 mg/ml in 50 mM Tris-HCl (pH 8.0) for the wild-type and mutant OHCU decarboxylases, and ϳ1.0 mg/ml in 20 mM Tris-HCl (pH 8.0) and 200 mM NaCl for HIU hydrolase.
CD measurements were performed at 30°C in a 10-mm path-length cuvette with a Jasco J-810 spectropolarimeter. Uric acid (50 M) in 1 ml of 50 mM potassium phosphate buffer, pH 7.6, was incubated in the cuvette with urate oxidase (1 unit) and HIU hydrolase (4.0 g), and once OHCU reached its maximal concentration, OHCU decarboxylase (4.0 g) was added. Rate constants for the decay of OHCU were obtained by fitting curves at 257 nm to a model of exponential decay by a firstorder reaction (SigmaPlot). The formation of allantoin in the presence of the three enzymes of the ureide pathway was measured in the range of 200 -360 nm at specified incubation times.

RESULTS
Overall Structure of OHCU Decarboxylase-The OHCU decarboxylase of A. thaliana consists of the N-terminal 161 residues of a bifunctional enzyme fused with HIU hydrolase (8, 17) (Fig. 1). The enzyme was crystallized in the presence of racemic allantoin as a P3 1 21 space group with one monomer in an asymmetric unit. In the crystal, dimerization was observed between two crystallographic symmetry-related monomers, consistent with size-exclusion chromatography results, which predicted that the enzyme exists as a homodimer. Fig. 2A shows the structure of the dimer, in which the crystallographic 2-fold axis coincides with the molecular 2-fold axis running perpendicularly to the plane of the figure. Except for highly disordered regions, including residues 1-5 and 66 -78, monomeric OHCU decarboxylase folds largely into ␣-helices, representing an unprecedented structure based on DALI searches (the highest Z score of 2.3) (24). The L-shaped monomer contains two distinct domains, composed of two layers of ␣-helices and oriented at ϳ90°rela-tive to each other. The N-terminal domain consists of helices 1-4 and the C-terminal region of helix 8. Four geometrically adjacent helices, numbered 1-3 and 8, are arranged in an antiparallel manner, such that the overall architecture of the four helices is largely reminiscent of a four-helix bundle (Fig. 2B). Hydrophobic residues, including Trp-10, Phe-19, Ala-20, Ile-35, Ala-38, Ile-41, Trp-42, Ala-142, Ala-143, and Phe-158, create a hydrophobic core in the center of the bundle, consistent with other known four-helix bundles. Helix 4, which is perpendicular to the four-helix bundle, further stabilizes this arrangement by providing the additional hydrophobic residues Val-46, Trp-51, and Phe-55 in the hydrophobic core. The residues that FIGURE 1. Multiple sequence alignment of OHCU decarboxylases. OHCU decarboxylase from A. thaliana (GenBank TM accession number NP_200630) and orthologous proteins from rice (AAR06356), mouse (NP_001034767), human (EAX08423), bowfin (ABF51667), mosquito (EAT32724), and B. subtilis (NP_391125) are compared. Residues in red enclosed in a blue box are highly conserved, and those with a red background are identical. The three residues that may be involved in catalysis are indicated by filled triangles, and the dimer interface residues are marked with asterisks. The secondary structure elements identified from the A. thaliana enzyme are shown near the corresponding sequences. The figure was prepared using ESPript (31).
R free was calculated using 10% of the data excluded from refinement.
form the hydrophobic core are conserved for the hydrophobic properties in the OHCU decarboxylase family (Fig. 1), and their side chains are largely or completely buried in the core, as predicted on the basis of the accessible area calculated by the program AREAIMOL in CCP4 (23). Helices 5-7 and the N-terminal region of helix 8 are in the C-terminal domain. Two parallel helices, numbered 6 and 7, are also stabilized by the continuous hydrophobic interactions along the helices with the conserved residues Ala-92, Ile-96, Trp-99, Tyr-103, Phe-107, Phe-109, Phe-111, Leu-124, Leu-127, and Tyr-131 (Fig. 2C). Helix 5, which is in a position almost perpendicular to helices 6 and 7, caps the N-terminal regions of these two helices. The 24-residue helix 8 is intercalated between the two domains and stabilizes the perpendicular orientations of the domains, mainly by hydrophobic interactions. In the dimeric conformation, the dimer interface of 1500 Å 2 of buried surface area mainly consists of interactions between the side chains and the backbone atoms of the two domains of each monomer. Active Site-The active site in each monomer, which was identified by the binding of allantoin, is located at the cleft between the two domains. Although the enzyme was crystallized in the presence of racemic allantoin, the observed density for the bound allantoin corresponds to an (S)-form, consistent with the stereospecificity of OHCU decarboxylase.
The active site is enclosed by several structural segments from the two domains that contain the highly conserved residues of the OHCU decarboxylase family (Fig. 3A). Loops between helices 4 and 5, although partly missing in the current structure, and between helices 6 and 7 are primary elements of the top and bottom of the pocket, respectively, which is directly lined on both sides by the completely buried hydrophobic residues Ile-61, Ile-112, Ala-115, and Ile-149. Hydrophilic residues of the two domains, including His-58, Glu-80, and Arg-153, also contribute side chains to the active site. Therefore, the bound (S)-allantoin is embedded in a hydrophobic environ-ment but is covered with these hydrophilic residues on the surface of the pocket. The active site contains no indications of any electron density for cofactor(s) or metals, distinguishing this enzyme mechanistically from other cofactor-dependent decarboxylases (25). Possible hydrogen bonds between the bound (S)-allantoin and the active site residues are mainly mediated by atoms of the main chain of the enzyme (Fig. 3B). In particular, N6, O7, and N8 at the 5-ureido group of (S)-allantoin form hydrogen bonds exclusively with either the backbone carbonyl oxygen or nitrogen. In addition, N1 and O2 interact within 3.1 Å with the carbonyl oxygen of Ile-113 and the main-chain nitrogen of Ala-115, respectively. Four residues, including Ile-61, Ile-112, Ala-115, and Ile-149, also mediate extensive hydrophobic interactions within 4 Å of the bound (S)-allantoin.
In contrast to these nonspecific interactions, specific interactions are present that may be related to the stereospecificity of the reaction. The residues His-58, Glu-80, and Arg-153, the invariant residues in the family (Fig. 1), are the only amino acids whose side chains interact directly with the bound (S)-allantoin. These interactions are unusual in that all three side chains are localized 3.0 -4.2 Å from the O4 of (S)-allantoin, with the exception of a 2.9-Å hydrogen bond between N3 of (S)-allantoin and O ⑀1 of Glu-80 (Fig. 3B). In addition to interacting with (S)-allantoin, Glu-80 and Arg-153 perform other structural roles in the active site. The side-chain carboxylate of Glu-80 forms bidentate hydrogen bonds of 3.0 Å with the side chain of Arg-153, bridging the N-and C-terminal domains. The resulting extended conformations of these side chains serve as a lid to the active site, effectively sequestering (S)-allantoin from the solvent (Fig. 3A).
The production of (S)-allantoin by OHCU decarboxylase requires the removal of the carboxylate group from C4 and protonation at the re-face of C5 in OHCU (Scheme 1). Inspection of the active site revealed that His-58 can fulfill the stereochemical requirements for (S)-allantoin formation, in both distance and orientation. Unlike Glu-80 and Arg-153, His-58 is the only possible proton donor within 4.0 Å of C5, and moreover is located at the re-face of C5 in OHCU, based on the observed (S)-allantoin conformation (Fig. 3A). In particular, N ␦1 of His-58, which is 3.0 Å from O4 in (S)-allantoin, is not only proximal to C5 at a distance of 3.7 Å but also shows enhanced basicity due to the possible 2.7-Å hydrogen bond between the side-chain carbonyl oxygen of Gln-146 and N ⑀2 of His-58 (Fig. 3C).
Functional Analysis of OHCU Decarboxylase-To investigate the roles of the proposed catalytic residue His-58 and the two nearby residues Glu-80 and Arg-153, we measured the enzyme activity of the wild-type OHCU decarboxylase and mutants of the enzyme that are altered at one of three residues. Because some of the A. thaliana enzymes exhibit solubility problems, in particular the bifunctional enzyme that has both OHCU decarboxylase and HIU hydrolase activities, the HIU hydrolase and OHCU decarboxylase from B. subtilis (7,14) were used in this experiment. The residues His-58, Glu-80, and Arg-153 in the A. thaliana enzyme are equivalent to His-68, Glu-84, and Arg-158 in the B. subtilis OHCU decarboxylase, respectively (Fig. 1).
In CD measurements, optically active HIU, OHCU, and (S)allantoin exhibit unique CD spectra in the region of 210 -360 nm (8). Given that the conversion of OHCU into (S)-allantoin is catalyzed by OHCU decarboxylase, we first selectively monitored the decay of OHCU in a time course by measuring CD signals at 257 nm, at which wavelength only OHCU shows noticeable ellipticity (8). Fig. 4 shows CD spectra of the decay of OHCU and the formation of (S)-allantoin over time. In the absence of OHCU decarboxylase, OHCU produced by the consecutive reactions of urate oxidase and HIU hydrolase underwent nonenzymatic degradation with a rate constant of 1.8 ϫ 10 Ϫ3 s Ϫ1 at 30°C, which is comparable to the previously published value of 1.2 ϫ 10 Ϫ3 s Ϫ1 at 25°C (8) (Fig. 4A). The presence of the wildtype OHCU decarboxylase, along with the other two enzymes, caused a rapid decomposition of OHCU, and thus there were no appreciable changes in the ellipticity at 257 nm. However, in a reaction containing the OHCU decarboxylase H68A mutant at 4 g/ml, OHCU slowly disappeared with a pseudo-first order rate constant of 5.8 ϫ 10 Ϫ3 s Ϫ1 . This rate of OHCU decay is only three times greater than that of the nonenzymatic spontaneous decay and those of the E84A and R158A mutants under identical experimental conditions (see below), indicating that a mutation at His-68 eliminates the enzyme activity. In the presence of the E84A and R158A mutant enzymes (Fig. 4A), the decomposition of OHCU, with rate constants of 2.0 ϫ 10 Ϫ3 s Ϫ1 and 1.8 ϫ 10 Ϫ3 s Ϫ1 , respectively, was almost identical to that in the absence of OHCU decarboxylase, suggesting that these two mutant enzymes are essentially inactive and the observed decay of OHCU is due to nonenzymatic decomposition.
The formation of (S)-(ϩ)-allantoin was also measured using the various OHCU decarboxylase forms. In the presence of all three enzymes, including the wild-type OHCU decarboxylase, the CD spectrum corresponding to (S)-(ϩ)-allantoin appeared within 3 min and remained stable (Fig. 4B). However, in the presence of the OHCU decarboxylase H68A mutant, the addition of uric acid produced a spectrum corresponding to OHCU within 3 min of the reaction, but after 8 min the spectrum disappeared and the (S)-(ϩ)-allantoin spectrum did not appear. This result indicates that through the completion of the reaction, no optically active substances were produced. Furthermore, in reactions with the E84A or R158A mutant enzymes, the OHCU spectrum was observed upon the addition of uric acid but gradually disappeared over time, and no optically active substances were observed after 25 min of reaction time (Fig. 4C). The absence of optically active substances in the presence of a mutant OHCU decarboxylase, even though the OHCU was apparently degraded, could be due to the formation of a racemic mixture of allantoin. In an uncatalyzed reaction in which the spontaneous decay of OHCU was monitored, a racemic mixture of allantoin was produced, with  (8). These data on the kinetics of OHCU decomposition and the formation of (S)-(ϩ)-allantoin suggest that a mutation at the proposed catalytic residue His-68 in the B. subtilis enzyme essentially inactivated the OHCU decarboxylase activity and caused a significant slowing in the rate of decay of OHCU while eliminating the stereospecificity of the enzyme. Other mutations at Glu-84 and Arg-158 also abolished the enzyme activity, possibly by impairing their catalytic roles or the binding affinity of the substrate to the enzyme.

DISCUSSION
We have presented structural and functional evidence that OHCU decarboxylase, an enzyme of the ureide pathway, contains a novel helical structural fold and is a cofactor-independent enzyme. This is only the second demonstration of a direct decarboxylation by a decarboxylating enzyme.
Our structural and functional analyses of OHCU decarboxylase revealed that histidine, glutamate, and arginine in the hydrophobic active site (His-58, Glu-80, and Arg-153 in A. thaliana and His-68, Glu-84, and Arg-158 in B. subtilis in these studies) are the only residues that interact directly through their side chains with the product. In addition, these residues play essential roles in catalysis, specifically in the stereospecificity and/or the affinity of the ligand for the enzyme. These proposed functional roles are substantiated by the conservation of the residues in members of the OHCU decarboxylase family (Fig. 1) and by structural differences in the binding mode of (S)-allantoin in OHCU decarboxylase and urate oxidase. In particular, the binding environment of (S)-allantoin in OHCU decarboxylase sharply contrasts with that captured in the active site of urate oxidase, which was initially revealed by crystallization of the enzyme in the presence of uric acid, during which HIU produced in the crystallization process decomposed into allantoin in a nonenzymatic manner. Unexpectedly, (S)-allantoin was observed to bind selectively in the active site of urate oxidase (26). In the crystal structure of urate oxidase complexed with (S)-allantoin, the five-member ring of (S)-allantoin lacks specific interactions with the side chains of active site residues but instead shows nonspecific interactions with nearby water molecules, providing structural differentiation between the enzymatic (stereospecific) reaction in OHCU decarboxylase and the nonenzymatic (nonstereospecific) reaction in urate oxidase.
On the basis of the structural and functional analyses of OHCU decarboxylase, we propose a mechanism for (S)-allantoin formation (Scheme 2). The active sites of most decarboxylating enzymes bind metals or other cofactors, which function to activate decarboxylation and stabilize the carbanion upon the elimination of CO 2 from the substrate (25,27). The absence of cofactors or metals in the active site of OHCU decarboxylase leads us to postulate that the decarboxylation reaction occurs directly. This type of direct decarboxylation is unusual but has been extensively characterized in orotidine 5Ј-monophosphate (OMP) decarboxylase, the only enzyme of this class discovered to date, in which decarboxylation and protonation occur at the same C6 carbon of orotidine. In OMP decarboxylase, the substrate carboxylate group was suggested to form unfavorable electrostatic interactions with the side-chain carboxylate of an active site aspartate residue, thereby destabilizing the substrate carboxylate and facilitating a decarboxylation reaction. However, the binding environment of the substrate OMP to the enzyme indicates that the resulting negative charge cannot be delocalized into the pyrimidine ring (27). Decarboxylation and protonation in OMP decarboxylase is therefore a concerted event in which protonation is mediated by a nearby lysine residue (27,28). Further mutational studies have indicated that a mutation at this aspartate abolishes enzyme activity, consistent with the proposed catalytic role of this residue (29).
In contrast, the features of the OHCU decarboxylase reaction differ from those of the OMP decarboxylase, in that decarboxylation at C4 and protonation at C5 are separated, and more importantly, that the carbanion can be stabilized by using the double bond between C5 and N1 as an electron sink. From a stereochemical perspective, these two events cannot occur in a concerted mode, but must occur in the order of decarboxylation and then protonation. On the basis of a bound (S)-allantoin, a substrate carboxylate was modeled on an sp 3 C4 of OHCU with two possible orientations, toward either His-58 or Gln-81. Given the cavity in the region between (S)-allantoin and Gln-81 (Fig. 3A), the substrate carboxylate fits better into the orientation toward Gln-81, suggesting that OHCU favors the (S)-configuration at C4, with the hydroxyl group at C4 locating toward His-58 (Scheme 1). In this modeled OHCU, Glu-80 is the only negatively charged residue in the vicinity (ϳ4 Å) of the modeled substrate carboxylate on OHCU, implying that Glu-80 could play a catalytic role for a decarboxylation, analogous to the aspartate in OMP decarboxylase. However, geometrically its side chain is not oriented toward a substrate carboxylate but instead points away from OHCU (Fig. 3A). In OMP decarboxylase, two carboxylate groups lie almost in the same plane, with a head-to-head orientation (27). Changes in torsion angle 1 of Glu-80 might place the side chain carboxylate in a more optimal orientation for a decarboxylation reaction, although this proposed movement would make the reaction intermediate vulnerable to solvent by opening the active site. Thus, although this mechanism cannot be ruled out, it seems unlikely that a mechanistic analogy to OMP decarboxylase can be extended to OHCU decarboxylase. An alternative proposal for direct decarboxylation by OHCU decarboxylase is based on the hydrophobic features of the active site, which may cause considerable destabilization of the carboxylate group on OHCU. This relatively simple postulate is supported by sequence conservation, in that the hydrophobic residues in the active site of the A. thaliana enzyme are highly conserved in the members of the OHCU decarboxylase family (Figs. 1 and 3A). These structural considerations, along with the identical functional results from the E84A and R158A mutants of the B. subtilis enzyme (Fig. 4), suggest that Glu-80 does not participate directly in a decarboxylation reaction but instead is involved in closing the active site in a concerted movement together with Arg-153 (Fig. 3A). Unlike the wild-type enzyme, in which these two residues are thought to sequester the bound OHCU from the solvent, the mutant enzymes may be unable to enclose the active site effectively, allowing the degradation of OHCU in a nonenzymatic manner. Isothermal titration calorimetric measurements have been made to determine whether the mutations have an effect on the binding of a ligand to OHCU decarboxylase, with a stable allantoin as the ligand. However, due to the extremely low binding affinity of allantoin to the enzymes, even with an excess molar ratio of ligand to enzyme (data not shown), no firm conclusion could be made. Therefore, decisive roles for Glu-80 and Arg-153 in direct decarboxylation require further investigation.
Upon the elimination of CO 2 from OHCU, the resulting negative charge on C4 could be delocalized through the conjugated electron system in the substrate and the protonation of N1; although the proton donor to N1 is not clear of the structure, the environment around N1 suggests that it is protonated (Fig.  3B). Subsequently, the unprotonated N ␦1 of His-58, with its enhanced basicity, may be involved in deprotonating the hydroxyl group at C4 from the re-face of C5 in OHCU. This stereochemistry allows the scissile bond to be perpendicular to the plane of the electron system of a possible transition state, gaining delocalization energy with a maximumoverlap. Finally, protonation at the sp 2 C5 on the same side is presumably achieved by the protonated N ␦1 of His-58, producing a stereospecific (S)-allantoin. This proposed proton transfer by His-58 is supported by the stereochemical considerations of OHCU, in which the hydroxyl group of the imidazoline ring appears to be on the re-face of C5. Proton transfer in enzyme catalysis has been well characterized in the pyridoxal phosphatedependent aminotransferase reaction, where the N of lysine carries out the 1,3-prototropic shift of a proton in the external aldimine, resulting in a ketimine intermediate (30).
This proposal is consistent with the structural and functional results obtained in the present study, as well as with the conservations of a histidine residue in the active site, e.g. His-58 in A. thaliana and His-68 in B. subtilis. This histidine residue is the only candidate for protonating the re-face of the prochiral trigonal center C5, and thus a mutation of this histidine residue abolishes the enzyme activity and stereospecificity of the allantoin that is produced.
We note the recent publication of Zanotti et al. (32) describing the crystal structure of this enzyme from zebrafish. Although the tertiary structures of the enzymes in their monomeric forms are essentially identical, the two OHCU decarboxylases differ with respect to quaternary structure and the stereoisomer of the bound allantoin. The relative orientations of the two monomers are different in these homodimeric structures, and (R)-allantoin is observed in the active site of the zebrafish enzyme, unlike the binding of (S)-allantoin in the active site of A. thaliana shown in the present study, which is consistent with the stereospecificity of the enzyme. Also a mechanism for direct decarboxylation that is analogous to that of OMP decarboxylase has been postulated.