Crystal Structures of Inhibitor Complexes Reveal an Alternate Binding Mode in Orotidine-5′-monophosphate Decarboxylase*

The crystal structures of the enzyme orotidine-5′-monophosphate decarboxylase from Methanobacterium thermoautotrophicum complexed with its product UMP and the inhibitors 6-hydroxyuridine 5′-phosphate (BMP), XMP, and CMP are reported. A mutant version of the protein, in which four residues of the flexible phosphate-binding loop180Gly–Gly190 were removed and Arg203 was replaced by alanine, was also analyzed. The XMP and CMP complexes reveal a ligand-binding mode that is distinct from the one identified previously with the aromatic rings located outside the binding pocket. A potential pathway for ligand binding is discussed.

Orotidine-5Ј-monophosphate decarboxylase (ODCase) 1 catalyzes the last step in the de novo pyrimidine biosynthesis pathway, converting OMP to UMP (Scheme I), which in turn serves as the source of all cellular pyrimidine nucleotides. ODCase continues to elicit strong interest not only because of its obvious importance in DNA and RNA synthesis, and thus in cell growth and proliferation, but even more so because of the enormous acceleration it conveys on the catalyzed reaction as well as the elusive nature of its reaction mechanism.
The decarboxylation of orotic acid in water of neutral pH is a very slow process (t1 ⁄2 ϳ 78 million years at 25°C). The enzyme ODCase, however, catalyzes the breaking of this stable carboncarbon bond at rates that are only 2 orders of magnitude below the limits set by diffusion. This astonishing feat qualifies ODCase as the most proficient enzyme known (1). Such catalytic power is even more remarkable if one considers that it is achieved without the help of any metal ions or cofactors and does not involve acid/base catalysis (2,3). Not too surprisingly, several laboratories have made the chemical mechanism of this enzyme an object of their studies (4 -8). Two strong inhibitors, 6-hydroxyuridine 5Ј-phosphate (barbituric acid ribosyl monophosphate; BMP) and 6-aza-UMP, are thought of as close mimics of the postulated carbanion intermediate thereby functioning as transition state analogues. Several weaker inhibitors include the product UMP and the purine nucleotides XMP and CMP (Scheme II). All of these compounds show a competitive inhibition pattern.
Recently published crystallographic studies of ODCases from four different organisms (9 -12) identified the close conservation of their three-dimensional structures but still did not lead to a generally accepted mechanistic proposal (13)(14)(15). A Lys 42 -Asp 70 -Lys 72 -Asp 75B arrangement creates a chain of alternating charges, which lies on one side of the binding pocket for the aromatic base. This cavity fits tightly around the bound product or inhibitor ligands (9 -12). The substrate OMP itself, however, only enters this site in crystals of the double mutant D70A/K72A. Without the removal of the two side chains, there was insufficient space for substrate molecules to bind, with Asp 70 not only overlapping in space but also exerting major electrostatic repulsion. In addition, there were no large conformational changes observed by either main chain or side chains when the wild-type enzyme and several active site mutants were crystallized in complex with substrate, product, and inhibitors (16). In light of this peculiar nature of the substratebinding site, we have put forward the idea that stress created by charge-charge repulsion in a small rigid and hydrophobic environment was a major factor in rate acceleration. This stress would be expected to build up as the C6-carboxylate approaches the side chain of Asp 70 leading to the release of CO 2 . The resulting carbanion would then be neutralized through proton transfer from the side chain of Lys 72 . The substantial binding energy generated by interactions of the phosphate and ribose parts of OMP with the enzyme matrix could provide a large part of the driving force for this process (9). This model, however, does not explain why the purine nucleotide XMP with its bulkier two-ring base is a much better inhibitor of ODCase (K i ϭ 4.1 ϫ 10 Ϫ7 M) than the product UMP (K i ϭ 2.0 ϫ 10 Ϫ4 M), which in its chemical structure more closely resembles the strong inhibitor 6-aza-UMP (K i ϭ 6.4 ϫ 10 Ϫ8 M) (4). Given the remarkably tight fit to the binding site found in all structurally characterized pyrimidine nucleotide complexes (9 -12, 16), one would have to postulate a major conformational change of the protein to accommodate the larger ring of XMP. The potential for such a reconfiguration of the active site would clearly weaken the argument for the involvement of electrostatic repulsion in the catalytic mechanism of the enzyme. We, therefore, set out to determine the binding mode of several well characterized inhibitor molecules, including XMP, to ODCase from the methanogenic archaeon Methanobacterium thermoautotrophicum.
In addition, we investigated the role played by the phosphate-binding loop, amino acids Gly 180 -Gly 190 , in ligand binding by analyzing the 6-aza-UMP complex of the deletion mutant ⌬R203A, in which amino acids 184 -187 were removed and the phosphate anchoring Arg 203 was changed to alanine. The crystal structures of the complexes of wild-type ODCase with the product UMP and the most tightly binding inhibitor 6-hydroxyuridine 5Ј-phosphate (BMP) were also determined to allow meaningful comparisons with the other complexes studied by us but also with the results described by other groups for the enzymes from yeast (10), Bacillus subtilis (11), and Escherichia coli (12).

EXPERIMENTAL PROCEDURES
Cloning, Protein Expression, and Purification-Wild-type and mutant M. thermoautotrophicum ODCases were expressed and purified as recently described (17). The ⌬R203A mutant was constructed by first introducing a single site mutation, R203A (QuikChange TM , Stratagene), followed by the removal of the four residues Ala 184 -Gln 185 -Gly 186 -Gly 187 in a second round of mutagenesis. The coding strand sequences of the respective primers are (mutating triplet in bold and underlined; the point of deletion is indicated by vertical line (͉): R203A, GATGCCATAATAGTTG-GAGCGTCCATCTACCTTGCAG and ⌬, CATTTCTCATATCCCCCGGT-GTGGGA͉GACCCAGGGGAGACCCTC.
Crystallization, Data Collection, and Processing-All crystals were grown in hanging drops using ϳ1.2 M trisodium citrate, pH 6.5-8.5, as the main precipitant. To obtain reasonably sized, single crystals of the UMP and BMP complexes, 3-7% dioxane had to be added to the crystallization medium. The resulting crystals belonged to space group C222 1 , with one monomer per asymmetric unit. Other additives were needed to generate single crystals of the XMP and CMP complexes. 10% (v/v) (Ϯ)-1,3-butanediol (Fluka) initiated crystal formation and sup-pressed twinning of the XMP crystals. A mixture of 1.5-2% (w/v) 1,2,3heptanetriol (Fluka) and 5% (v/v) 12-crown-4 (Fluka) was essential in growing diffraction quality crystals of the CMP complex (space group P2 1 , with one dimer per asymmetric unit) and to again suppress twinning of the crystals. To stabilize the 6-aza-UMP complex crystals of the deletion protein ⌬R203A, MgCl 2 was added to a final concentration of 25 mM, in addition to 5% dioxane (space group P1; four monomers per asymmetric unit). Nevertheless over time, the inhibitor molecules dissociated from the protein and crystals of the free protein appeared on the surface of the complex crystals.
Structure Determination-Both the structures of the XMP and ⌬R203A complexes were determined using the molecular replacement package EPMR (19). In the case of the XMP complex, the wild-type ODCase monomer coordinates were used as the search model for the two monomers in the asymmetric unit (an identical procedure was followed for the CMP complex), whereas a dimer was used to find the two dimers of ⌬R203A. Subsequent refinements were done with the program Crystallography NMR Software, version 0.9 (20). No noncrystallographic symmetry restraints were used for the final model. The refinement statistics are given in Table I. All images were prepared using SPOCK (21), MOLSCRIPT (22), and RASTER3D (23).
Effect of Alcohol on the Kinetics of ODCase-Wild-type protein was expressed in the ODCase-deficient E. coli cell line SKP10 (24) and purified according to the standard protocol (17). A concentration of 0.12 M calculated for monomeric protein was used in each activity assay. The decarboxylation of OMP was monitored in an UV-visible spectrophotometer (Cintra 20, GBC) at 282 nm and 25°C in 50 mM Tris buffer, pH 7.5, and 1 mM dithiothreitol. For each concentration of OMP, the initial linear rate of decrease in absorption was taken as a measure of the velocity. The effect of (Ϯ)-1,3-butanediol on the capability of XMP to inhibit ODCase activity was tested with various concentrations of the alcohol in the constant presence of 200 M XMP. The data were fit to the Michaelis-Menten equation with the program KaleidaGraph 3.0 (Abelbeck software) to provide the values of V max and the apparent dissociation constant K m .

RESULTS
Both the CMP complex and XMP complex crystals were very difficult to grow. Modifications to the dielectric constants of the media caused by the addition of small organic compounds seemed to support the formation of crystals of ODCase complexes, but a large number of additives had to be screened before crystals of diffraction quality could be obtained reproducibly. For instance, it was mandatory to add (Ϯ)-1,3-butanediol to the crystallization mixture for the XMP complex crystals to suppress their very high tendency to twin.
The electron density maps for both the CMP and XMP complexes clearly show that neither of the nucleotide inhibitors has its base bound inside the cavity that accepts the aromatic rings of the substrate and product molecules. Also in both structures, the loop comprising amino acids Gly 180 -Gly 190 was disordered. This binding mode leaves the inhibitor molecules partially exposed to solvent (Fig. 1). In substrate or product complexes, this loop clamps down on the nucleotide. Gln 185 , the residue located in the center of the loop, interacts with both O-2 and an oxygen of the 5Ј-phosphate.
Purine Inhibitor XMP-The analysis of this complex was attempted to address why the purine nucleotide XMP with its significantly larger base is a much more effective competitive inhibitor of ODCases than the product UMP, especially given SCHEME I SCHEME II the narrowness of the site that was to accept the aromatic rings of these molecules.
Its limited size, together with its resistance to conformational change, prevents the base-binding cavity from accepting the xanthine group of XMP (Fig. 1B). Unlike 6-aza-UMP, UMP, BMP, and OMP, XMP adopts the lower energy structure also found in solution, with 3Ј-endo sugar puckering and anti-conformation of the base. Despite this change, Ser 127 still serves as an anchoring residue for the nucleotide base forming two hydrogen bonds to xanthine. The side chain hydroxyl accepts a proton from N-1 and the backbone amide donates a proton to O-2. When compared with the complex of 6-aza-UMP and wildtype enzyme, the interactions between the phosphate group and Arg 203 as well as those of the 3Ј-OH of ribose with Asp 20 and Lys 42 were conserved. However, the 2Ј-OH group was no longer in contact with its usual partners, Asp 75B and Thr 79B from the other monomer. Instead, it was now only 3.0 Å away from the amino group of Lys 42 and forms an additional bond to the C3-OH of (ϩ)-1,3-butanediol, a crystallization additive which, together with a water molecule W1, is bound in the active site of the XMP-ODCase complex ( Fig. 2A).
The molecule of butyl alcohol could be described to act like an adapter. Its hydrophobic chain fits nicely against the hydrophobic pocket of the active site, whereas its C1-OH binds to N-3 of xanthine, in addition to its C3-OH and the ribose 2Ј-OH interactions mentioned above. A superposition of the active sites of the XMP and 6-aza-UMP complexes (Fig. 3A) shows that water W1 assumes the regular binding spot of the ribose  2Ј-OH, forming hydrogen bonds to OG1 of Thr 79B and to the carboxylate of Asp 75B .
The refined B-factor values for XMP (average B ϭ 20.8 Å 2 ) indicate a high occupancy rate and low mobility of this molecule in the crystalline complex. For comparison, the average B-factor for the very rigid active site residues is 16.2 Å 2 . Water W1 and the butyl alcohol molecule show significantly higher values of 34.9 and 43.6 Å 2 , respectively, most probably indicative of increased mobility. It is interesting to note that both hydroxyls of the additive as well as the water W1 take up positions that are occupied by first-hydration shell water molecules in the free-enzyme structure.
Kinetic Studies-We performed kinetic measurements to establish whether the presence of (Ϯ)-1,3-butanediol had any major influence on catalysis. Increasing concentrations of alcohol do lower the apparent dissociation constant K m from that measured with wild-type enzyme; V max , however, stays constant (Table II). As the dielectric constant (⑀) of the medium was decreased, ODCase binds the substrate more tightly. Ionic interactions and hydrogen bonds increase in strength with decreasing ⑀; therefore our results corroborate the idea that a large part of the substrate binding energy was generated from the interactions of the enzyme with the phosphate and the ribose groups of the nucleotide. Lowering ⑀ will also make it easier to protonate the C6-carboxylate of OMP, minimizing potential electrostatic repulsion. Our results align well with the general finding that decarboxylation reactions are accelerated in desolvating environments (8).
At a fixed concentration of 200 M XMP, increasing amounts of (Ϯ)-1,3-butanediol (5-20%, v/v) do not significantly change the values of V max and K m (Table II) leading us to believe that the alcohol does not drastically change the binding mode of XMP. The high B-factor values of the alcohol molecules also imply that their interactions with amino acids and the xanthine ring were not very specific and not very strong. Another argument for the general relevance of the orientation of the aromatic ring of XMP observed in crystals was the fact that the base of bound CMP adopts an almost identical position in the absence of any mitigating molecules (see below).
Pyrimidine Inhibitor CMP-The potential substrate 6-carboxy-CMP binds only very weakly to ODCase and its decarboxylation was almost undetectable (25); yet, there was no strong structural reason for the absence of catalysis. When comparing the interactions possible between the enzyme and orotate on the one hand and a cytosine ring on the other, only one hydrogen bond might be lost in the latter (9). Most of the hydrogen bonds formed (a water molecule linking the backbone carbonyl group of Gln 125 with O-4 of orotate and the hydrogen bond of the latter to the backbone amide of Ser 127 ) could have the partners act equally well as hydrogen donors or acceptors. In complexes involving orotidine or uracil, the side chain hydroxyl of Ser 127 receives a bond from N3H of the base, and at the same time, donates one to the oxygen of the amide side chain of Gln 185 . However, it could only act as a donor to either one of these positions in a potential CMP complex. To shed some light on the structural basis of the rejection of CMP-carboxylate as a substrate, we determined the crystal structure of the CMP bound to M. thermoautotrophicum ODCase.
The crystal structure of the CMP complex reveals that the pyrimidine nucleotide not only adopts a position very similar to the one found for XMP but also engages in most of the interactions displayed by the purine nucleotide. Another feature  common to both complexes was the disordered state of the loop comprising residues Gly 180 -Gly 190 , the phosphate-binding loop (Fig. 1C).
Both active sites of the functional dimer were occupied by CMP molecules as well as by waters of increased mobility (Fig.  2B). CMP assumes its solution conformation, with the ribose ring in the 3Ј-OH endo position and cytosine oriented anti to the sugar ring (Fig. 3B), again mirroring XMP in its complex with ODCase. Both CMP molecules have relatively high Bfactor values, 34.2 and 42.3 Å 2 for the A and B monomers of the homodimer, respectively, in contrast to 19.4 and 23.3 Å 2 for the conserved enzyme residues surrounding them. The binding modes of the two cytosine bases were not completely identical; the one in the A monomer assumes a position slightly closer to that of the 6-azauracil ring in the 6-aza-UMP complex than does its counterpart in the B monomer. The increased mobility of CMP was accompanied by the disordering of the guanidinium head group of Arg 203 in the active site of monomer A. While the ribose 3Ј-OH was still fixed by hydrogen bonds to Asp 20 (2.7 Å) and Lys 42 (2.8 Å), the 2Ј-OH group was not making its proper contacts with the enzyme but formed a long hydrogen bond with the amino group of Lys 42 (3.1 Å) instead.
For the crystallization of the CMP-ODCase complex, 1,2,3heptanetriol was used as an additive. As this 7-carbon alcohol is much larger than (ϩ)-1,3-butanediol, it cannot fit between the nucleotide and the active site wall. Instead, five water molecules fill the empty cavity designed to hold the base in the productive nucleotide-binding mode. One of them sits in the position that the 2Ј-OH would occupy in such a complex. It is equivalent to water W1 in the XMP complex and engages in the same interactions to Asp 75B and Thr 79B .
Impaired Phosphate Binding Mutant, ⌬R203A-Kinetic studies have made it quite clear that the binding energy contributed by the 5Ј-phosphate group is very important for ODCase catalysis. Mutants with impaired phosphate binding ability have drastically reduced catalytic rates (4). To investigate the structural basis of this property, especially to find out whether the loss of the majority of bonds between the phosphate group and the enzyme has any effects on the conformation of the residues surrounding the aromatic rings of the nucleotide bases, we mutated Arg 203 , the main phosphateinteracting residue, to alanine and deleted the four residues 184 AQGG 187 of the phosphate-binding loop. Following co-crystallization, the structure of this mutant protein was determined in its complex with the inhibitor 6-aza-UMP. The reduced ligand binding affinity correlates well with a loss in crystal stability but does not lead to any structural adaptation of the active site. First, crystals of the complex form appear but over time many of them dissolve and ligand-free crystals (space group P4 1 2 1 2) start to grow on their surfaces. The residual complex crystals start to bend. If complex crystals were harvested after nucleation of free enzyme crystals had started the resolution of their diffraction pattern dropped to about 5 Å. Both effects are consistent with the build up of stress in the crystal lattice. Given these observations, it came as no surprise to find that the refined overall B-factor value for this mutant complex is quite high (38.7 Å 2 averaged over the four monomers located in the asymmetric unit) with one of the monomers displaying an overall B-factor value of 44.6 Å 2 . The electron density corresponding to the nucleotide ligand in the deletion mutant is distinctly weaker and of lower quality than the density found for the XMP or CMP ligands in their respective complexes. Together with the problems encountered during the crystallization of the 6-aza-UMP⅐⌬R203A complex, this indicates an elevated tendency for ligand loss.
Although MgCl 2 had been added to the crystallization solu-tion for stabilization, no defined metal ion was observed in the resulting 1.9-Å electron density map. Superposition of active site residues in the wild-type enzyme with their counterparts in the ⌬R203A mutant (Fig. 3C) does not reveal any significant differences (root mean square deviations of 36 main chain and 9 side chain atoms ϭ 0.12 Å). Despite its reduced affinity for inhibitors, the ⌬R203A mutant binds 6-aza-UMP in exactly the same position as the wild-type enzyme (Fig. 1D). All interactions between the nucleoside part of the inhibitor molecule and active site residues were maintained, including the hydrogen bond between N-6 and Lys 72 (3.1-3.2 Å). This was obviously not the case for the phosphate group. In addition to missing the ionic interaction with Arg 203 , it has lost one of its first hydration shell waters that was held in place by the backbone amide of Gln 185 and the backbone carbonyl of Val 182 in wild-type ODCase.
One distinct difference between the structures of wild-type and deletion mutant enzymes was the presence of an additional, elongated electron density peak next to the azauracil ring in the latter (Fig. 2C). Whereas the average values of the B-factors for the active site residues and 6-aza-UMP range from 25 to 36 Å 2 , the B-factor for the water molecule modeled into the density has a value of 60 Å 2 . It was not clear whether this density represents alternate but overlapping binding positions of water molecules or whether it was caused by the low occupancy presence of an as yet unidentified low molecular weight compound.
BMP, 6-Aza-UMP, and UMP Inhibitor Complexes-BMP, 6-aza-UMP, and UMP all have almost identical molecular structures except for positions 6 of the respective bases. Both BMP (K i ϭ 8.8 ϫ 10 Ϫ12 M) (26) and 6-aza-UMP (K i ϭ 6.4 ϫ 10 Ϫ8 M) (4) are strong inhibitors. This has been attributed to their similarity to the carbanion, which was postulated as a reaction intermediate stabilized by the positive charge of Lys 72 (26). This single favorable interaction alone, however, cannot adequately explain why the product UMP is so ineffective as an inhibitor (K i ϭ 2.0 ϫ 10 Ϫ4 M). The crystal structures of the three compounds complexed with various homologous ODCases were known (9 -12). We decided to reanalyze the corresponding complexes with M. thermoautotrophicum ODCase as the crystals of this enzyme diffract to significantly higher resolution than do those from other sources. At the same time, the results of such analyses provide a common frame of reference against which the significance of small changes between the various ligand complexes can be assessed.
When the structures of the different M. thermoautotrophicum complexes were compared, only minimal changes were identified. The overall root mean square deviation value for 418 C␣ atoms belonging to a dimer was less than 0.5 Å; the corresponding value was 0.2 Å for 50 backbone and C␤ atoms of the following conserved active site residues: Asp 20 , Lys 42 , Asp 70 , Lys 72 , Ser 127 , Gln 185 , Arg 203 , Asp 75B , Ile 76B , and Thr 79B . The similarities even extend to water molecules accompanying the inhibitors in the binding site. As described for the structures of the enzymes from E. coli (12) and yeast (10), the negatively charged atom O-6 of BMP interacts electrostatically with Lys 72 (2.8 Å) and at the same time via hydrogen bonding with a solvating water molecule and with OD1 of Asp 70 .
As part of the uracil ring of UMP, C-6 does not carry a negative charge and cannot act as a partner in ionic or hydrogen bond interactions. Consequently, the amino group of Lys 72 was located more than 1 Å away from its positions in the 6-aza-UMP⅐ODCase and BMP⅐ODCase complexes. The rest of the conserved side chains shows no changes in position. In addition to the loss of an energetically very favorable interaction with the side chain of Lys 72 , the UMP complex structure reveals another reason for the inefficiency of UMP to act as an inhibitor of ODCases. In the ligand-free structure, there was one well ordered solvent molecule for each of the charged residues in the active site. Binding of UMP creates a cavity between the inhibitor molecule and the enzyme surface because of desolvation. As the exact energy involved in such a process depends on the microenvironment it was extremely difficult to estimate. However, it will undoubtedly impose appreciable entropic cost on the formation of the UMP⅐ODCase complex, which in turn will lead to an increase in the inhibition constant. DISCUSSION All of the complex structures reported here should help to improve our understanding of the highly specialized and finely tuned active site of ODCase. There was only one direct interaction (the interaction with Lys 72 ) between the active site of the enzyme and UMP that generates binding energy and was lost when compared with the strong inhibitors BMP and 6-aza-UMP. The xanthine ring of XMP does not interact with Lys 72 either; nevertheless, this nucleotide is a rather good inhibitor of ODCase. We identified two entropic factors that could mitigate the effects of this missing interaction. Binding of XMP is not accompanied by the adoption of one fixed conformation by the loop Gly 180 -Gly 190 , and it does not lead to the stripping of water molecules from the charged residues in the active site. In addition, XMP itself retains the low energy conformation it assumes in solution.
Whereas UMP was much less tightly bound than the analogues that carry a (full or partial) negative charge close to position C-6 of the base, its aromatic ring enters the basebinding cavity and adopts a position common to those chargecarrying analogues. Somewhat surprisingly, given its close similarity to UMP, the base of CMP assumes a position almost identical to that of its counterpart in XMP. As mentioned above, the removal of a hydrogen atom from N-3 of CMP makes it impossible for this atom to act as a hydrogen bond donor. This, in turn, would lead to the loss of one hydrogen bond emanating from the side chain hydroxyl of Ser 127 . Whether this reduction in binding energy was sufficient to prevent proper binding of the aromatic ring of CMP was not clear, especially as the phosphate and ribose binding pattern was unchanged. There exists, however, another property of the nucleotide bases that could form the basis for discrimination: their dipole moment. For cytosine, its value was about 1.5 times larger and its direction rotated 70°when compared with the corresponding values for uracil (27). Given the unusual distribution of charges in the active site of ODCase (9), the binding site of the enzyme could well be able to recognize such differences in dipole moments and use them for ligand selection. After all, the enzyme does have a higher affinity for UMP than for ribose 5Ј-phosphate alone (4).
Although it does drastically reduce the binding strength of the inhibitor, the major impairment of phosphate binding alone, caused by the removal of four amino acids and a ligandanchoring arginine side chain, did not shift the binding position of the 6-aza-UMP base if sufficiently high inhibitor concentrations were used (4) (Fig. 2C). Because there were no unfavorable steric clashes or electrostatic repulsions (including chargehydrophobic contacts), 6-azauracil fits right into the active site pocket, and the limited number of contacts made by the ribose and the base were sufficient to keep it in place.
For both the XMP and CMP complexes, the binding modes were the same despite the use of different additives and, more importantly, the quite dissimilar nature of their base components. Phosphate and ribose groups of the inhibitors superimpose very closely (Fig. 3B) although the location of the ribose ring was not identical to the one found in the high affinity complexes with BMP and 6-aza-UMP. In the latter ones, the place occupied by the 2Ј-OH group finds a water molecule in the XMP and CMP structures, indicating a minimum energy position for hydroxyl groups at this spot. Given the small number of identifiable direct bonds formed between ODCase and CMP, it was surprising to find the relatively well defined binding revealed in the electron density maps of the complex (Fig. 2B). For the cytosine ring, no direct contacts to the enzyme can be seen; two water molecules form a bridge between the side chain of Ser 127 and O 2 . The interactions of the phosphate group with Arg 203 in the two monomers in the asymmetric unit vary in their tightness. CMP seems to be mainly held in place by hydrogen bonds between the two ribose hydroxyls and Lys 42 . The xanthine ring of XMP was held in place more firmly by side chain and main chain bonds to Ser 127 .
The binding mode of the XMP and CMP inhibitors may reflect a transient conformation that the ligand passes through on its way to its fully docked position in the active site of ODCase, a position reflected in the 6-aza-UMP complex. Our results were consistent with the following scenario. The 5Јphosphate group serves as the first anchor point of binding. Then, while the Gly 180 -Gly 190 loop closes down on the ligand, the formation of bonds to the ribose hydroxyls pulls the molecule deeper into the binding pocket. The ring pucker of the ribose sugar changes as water molecules bound in the active site are replaced by atoms of the ligand. However, only when size and/or dipole moment of the base correspond closely to those that the enzyme has evolved to accept is the aromatic ring of the ligand allowed to assume the "productive" orientation. This explains the decrease in k cat /K m when the ribose 2Ј-OH interaction with the enzyme was damaged either by mutation on enzyme residues Thr 79B and Asp 20B or by chemical modification on the substrate (6). Even though it is true that the enzyme has the highest affinity for the transition state, the components contributing to that affinity, in particular 2Ј-OH interactions of the ribose 5Ј-phosphate part, do not reach their maximum possible strength until catalysis has progressed greatly along its reaction coordinate. The transient enzymesubstrate complex perhaps resembles the CMP⅐ODCase or XMP⅐ODCase complexes where contacts were initiated but not finalized. Thus, the binding energy of the enzyme for phosphoribose is utilized for catalysis. Although the results of the structural analyses described above are suggestive of a temporal sequence of ligand binding, they do not yet yield definite conclusions about the chemical mechanism employed by this remarkable enzyme. Obtaining clues toward solving this puzzle will depend on the application of more dynamically oriented experimental methods.