Mechanism of Substrate and Inhibitor Binding of Rhodobacter capsulatus Xanthine Dehydrogenase*

Rhodobacter capsulatus xanthine dehydrogenase (XDH) is an (αβ)2 heterotetrameric cytoplasmic enzyme that resembles eukaryotic xanthine oxidoreductases in respect to both amino acid sequence and structural fold. To obtain a detailed understanding of the mechanism of substrate and inhibitor binding at the active site, we solved crystal structures of R. capsulatus XDH in the presence of its substrates hypoxanthine, xanthine, and the inhibitor pterin-6-aldehyde using either the inactive desulfo form of the enzyme or an active site mutant (EB232Q) to prevent substrate turnover. The hypoxanthine- and xanthine-bound structures reveal the orientation of both substrates at the active site and show the importance of residue GluB-232 for substrate positioning. The oxygen atom at the C-6 position of both substrates is oriented toward ArgB-310 in the active site. Thus the substrates bind in an orientation opposite to the one seen in the structure of the reduced enzyme with the inhibitor oxypurinol. The tightness of the substrates in the active site suggests that the intermediate products must exit the binding pocket to allow first the attack of the C-2, followed by oxidation of the C-8 atom to form the final product uric acid. Structural studies of pterin-6-aldehyde, a potent inhibitor of R. capsulatus XDH, contribute further to the understanding of the relative positioning of inhibitors and substrates in the binding pocket. Steady state kinetics reveal a competitive inhibition pattern with a Ki of 103.57 ± 18.96 nm for pterin-6-aldehyde.

eukaryotic xanthine oxidoreductases. Despite differences in subunit composition, the folds of bovine XDH and R. capsulatus XDH are very similar (1). The bacterial enzyme can be described as a butterfly-shaped (␣␤) 2 heterotetramer. Each (␣␤) dimer represents one half of the active molecule and is encoded by two separate gene products, termed XdhA and XdhB, unlike the (␣) 2 dimeric eukaryotic protein, which is derived from a single polypeptide chain (2). Each subunit of the (␣␤) heterodimer carries a specific set of cofactors, which are crucial for catalysis and electron transfer. The 50-kDa XdhA subunit harbors two [2Fe2S] clusters as well as a FAD cofactor; the 85-kDa XdhB subunit contains the molybdenum cofactor harboring a catalytically essential terminal sulfido ligand (1,2). This cofactor is part of the active site binding pocket and catalyzes the oxidative hydroxylation of hypoxanthine to xanthine and further to uric acid. Most XDHs, with the exception of R. capsulatus and avian XDH, can be converted to the oxidase form (XO) while losing their ability to use NAD ϩ as the electron acceptor (3,4).
The catalytic sequence of R. capsulatus XDH is initiated by abstraction of a proton from the Mo-OH group by the highly conserved active site residue Glu B -730 (where B indicates the XdhB subunit), followed by nucleophilic attack of the resulting Mo-O Ϫ on the carbon center of the substrate (C-2 in hypoxanthine and C-8 in xanthine) and concomitant hydride transfer to the MoϭS of the molybdenum center (3). Residue Glu B -232, on the other hand, is involved in both substrate binding and transition state stabilization (3,5). Mutation of Glu B -232 to alanine leads to a 12-fold increase in the K D for xanthine (3). It has been suggested that interaction of Arg B -310 with the C-6 carbonyl group of the substrate xanthine stabilizes negative charge accumulation on the heterocycle that accompanies nucleophilic attack at C-8, thus stabilizing the transition state and accelerating the reaction of substrate oxidation (6). However, oxypurinol and 2-hydroxy-6-methylpurine were shown to bind in the opposite orientation in the active site, with the C-4 of oxypurinol facing Glu B -232 in the R. capsulatus enzyme and C-2 of 2-hydroxy-6-methylpurine facing Arg-880 in bovine XO (7,8).
Allopurinol (1-H-pyrazolo [3,4-d] pyrimidine-4-one), developed in 1963, is the current clinical treatment option for patients exhibiting symptoms of hyperuricemia, indicative of gout. The main drawback to allopurinol administration in humans is the possible onset of a toxicity syndrome manifested as eosinophilia, vasculitus, rash hepatitis, and progressive renal failure (9). This is most likely due to the inhibitory effect of allopurinol and its metabolites on other enzymes, such as purine nucleoside phosphorylase and orotidine-5Ј-monophosphate decarboxylase (10). Allopurinol is oxidized by XDH to oxypurinol (1,2-dihydropyrazolo [4,3-e] pyrimidine-4, 6-dione) ( Fig. 1), which commits suicide inhibition of XDH by replacing the hydroxyl ligand of the molybdenum ion and thereby inhibiting further catalysis (8).
Pterin-6-aldehyde ( Fig. 1) is a potent inhibitor of XO and is only found in the urine of cancer patients, a discovery that could play a key role in early cancer detection (11,12). Inhibition of XO by pterin-6-aldehyde is on the same order as that for oxypurinol, with a K i ϭ 10 Ϫ9 M (13)(14)(15)(16)(17). Because of its very slow rate of conversion to pterin-6-carboxylic acid, pterin-6-aldehyde is considered to be a very good inhibitor for XO (11,14,18).
Here we describe the first structures of the desulfo form and an active site variant E B 232Q of R. capsulatus XDH in the presence of its substrates xanthine and hypoxanthine as well as the pterin-6-aldehyde inhibited structure of the wild-type protein.
The structures reveal an intricate hydrogen-bonding network that can accommodate the substrates in the orientation required for catalysis and allow a tight approach of the inhibitor pterin-6-aldehyde toward the molybdenum cofactor. Furthermore, the comparison between the xanthine-and oxypurinolbound structures provides an explanation of why the inhibitor is bound in an opposite orientation relative to the xanthine molecule.

EXPERIMENTAL PROCEDURES
Expression and Purification of R. capsulatus XDH in Escherichia coli Cells-R. capsulatus wild-type XDH was expressed and purified as described previously (3). The mutation E B 232Q was introduced through PCR mutagenesis into R. capsulatus XDH. The generated variant in addition to the E B 232A variant was purified by nickel-nitrilotriacetic acid chromatography, ion exchange chromatography (Q-Sepharose; GE Healthcare), and size exclusion chromatography (Superose 12; GE Healthcare). To obtain the desulfo form of XDH, XdhAB were expressed from plasmid pMS3 (19) in the absence of XdhC in E. coli TP1000 cells (20). The cells were grown for 24 h at 210 rpm in LB medium supplemented with 1 mM Na 2 MoO 4 , 150 g/ml ampicillin, and 20 M isopropyl ␤-D-thiogalactopyranoside. Purification was achieved by nickel-nitrilotriacetic acid chromatography, and ion exchange chromatography using Q-Sepharose. To separate Moco-containing XDH from the enzyme lacking the cofactor, affinity chromatography on a Sepharose 4B/folate gel was used (3). Finally, the protein was purified by size exclusion chromatography. The different XDH variants in 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 200 mM NaCl, and 2.5 mM dithiothreitol were concentrated to 15 mg/ml. Crystallization trials followed immediately, and the remaining protein was stored for up to 5 days at Ϫ80°C.
Enzyme Assays-Enzyme assays were carried out at 25°C in 50 mM Tris and 1 mM EDTA, pH 7.5, in a final volume of 1 ml. Routine assay mixtures contained 1 mM NAD ϩ and various concentrations of xanthine or pterin-6-aldehyde. XDH activi- ties were determined by monitoring the absorbance changes at 340 nm for NAD ϩ reduction. The kinetic parameters were obtained by nonlinear fitting using the Microcal Origin 6.0 program. The data for enzyme inhibition were additionally obtained through a double reciprocal plot (1/v Ϸ 1/[S]).
Crystallization-Wild-type XDH, the E B 232A and E B 232Q variants as well as the desulfo XDH were crystallized via the hanging drop vapor diffusion method equilibrating equal volumes of the wild-type protein with the reservoir solution containing 6 -8 mM BaCl 2 , 6 -8% polyethylene glycol 8000, 100 mM Tris-HCl, pH 8.3, 5-25 mM dithiothreitol, and 3-4% isopropanol, whereas the E B 232Q variant was mixed in a 1:2 ratio with the same reservoir solution. Prior to crystallization set-ups desulfo XDH was rebuffered into 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 2.5 mM dithiothreitol. The structures of the inhibitor-bound wild type, the hypoxanthine-bound E B 232Q crystals, as well as the hypoxanthine-bound desulfo XDH were obtained through soaking experiments. The xanthine-bound desulfo XDH was obtained by cocrystallization with a saturated xanthine solution. The inhibitor experiment was performed in the dark because of its light sensitivity. Because of its low solubility, the inhibitor was first suspended in 60 mM NaOH and then mixed in a 1:50 ratio with the reservoir solution. The final concentration is unknown because pterin-6-aldehyde failed to dissolve completely. The E B 232Q and desulfo XDH crystals were soaked in a solution containing 60 mM NaOH and 10 mM hypoxanthine. After 12 h of soaking, the crystals were transferred into a cryoprotectant containing 30% glycerol and the mother liquor components and subsequently cryo-cooled in liquid nitrogen.
Data Collection and Structure Solution-Data of the E B 232Q, the E B 232A mutant, and the wild-type protein were collected at beamline X26C at the National Synchrotron Light Source (Brookhaven National Laboratory) using an ADSC Quantum 4 charged coupled device. The diffraction data were indexed, integrated, and scaled using the HKL suite (21). Data for the desulfo XDH crystals were collected at the microfocus beamline ID23-2 at the European Synchrotron Radiation Facility using a MarCCD 225 detector. The diffraction data were indexed and scaled using Mosflm and Scala from the CCP4 suite (22). All of the crystals belonged to space group P1 and contained two (␣␤) 2 heterotetramers in the asymmetric unit. Representative unit cell dimensions, taken from the E B 232Qhypoxanthine structure, are a ϭ 92.7 Å, b ϭ 140.6 Å, c ϭ 157.6 Å, ␣ ϭ 109.5°, ␤ ϭ 106.1°, and ␥ ϭ 101.1°. Difference Fourier methods were sufficient to solve the new structures because all R. capsulatus crystals belong to the same space group as the previously solved apo structure (Protein Data Bank code 1JRO) (8). Model building was initiated through rigid body refinement of the four monomers, with the XdhA and XdhB subunits of each monomer considered as separate domains. Non-crystallographic symmetry restraints between the XdhA domains and XdhB domains of each monomer were subsequently applied. The E B 232A and E B 232Q variant structures were further refined using XPLOR (23) followed by REFMAC (TLS refinement) (22), separating the monomer further into three domains: the [2Fe2S] domain, the FAD domain, and the molybdenum cofactor domain, to aid in modeling anisotropic motion of the individual components. The structures of the desulfo form were refined with REFMAC without the utilization of TLS. The models were analyzed in O (24) and Coot (25) to check for difference density at the location of the cofactors and the proposed location of the substrates or inhibitor. Cofactors were added at this point, and restrained refinement continued, employing non-crystallographic symmetry averaging or/and TLS refinement, to obtain the best nonmodel biased difference density and facilitate substrate or inhibitor insertion. After model completion, the different complexes were refined using REFMAC (see Table 1) using medium weighted positional and thermal non-crystallographic symmetry restraints for the protein moiety excluding ligands. To support the refinement of the alternate xanthine conformations, xanthine was included in the non-crystallographic symmetry restraints using the medium weighting scheme. The B factors for all structures were refined isotropically. Outliers of the Ramchandran plot being in proximity to the active site are the residues Ser B -458, Arg B -342, and Ser B -233 (and all corresponding copies) and are clearly defined in the electron density. Because they were already observed in 1JRO, it is unlikely that these are of functional significance related to substrate binding. Calculation of omit maps was performed omitting all cofactors, substrates, and the active site residues Gln B -232, Arg B -310, Thr B -460, Glu B -730, Phe B -344, and Phe B -459.

Steady State Kinetics of the XDH Variant E B 232Q-
The crystal structure of R. capsulatus XDH with oxypurinol bound at the active site revealed that the six-membered ring of the inhibitor interacts with Glu B -232, Arg B -310, and Glu B -730, suggesting that these amino acids are either involved in substrate binding or in substrate oxidation (8). Kinetic analyses clearly established Glu B -232 to be important for both substrate binding and transition state stabilization and Glu B -730 to be essential for catalysis (3). These results favored a reaction mechanism, in which Glu B -730 acts as an active site base that abstracts the proton from the Mo-OH group, which in turn undertakes a nucleophilic attack on the C-8 position of the substrate xanthine. We generated the E B 232Q XDH variant to reduce turnover of the substrate while maintaining the geometric environment of the active site. Analysis of the cyanolyzable sulfur of the purified protein showed that the E B 232Q variant was purified with a MoϭS content of about 56%. Steady state kinetics of the recombinant protein with varying concentrations of xanthine and NAD ϩ were obtained by monitoring absorbance increases at 340 nm for NADH production. The k cat (xanthine) and K m (xanthine) for the E B 232Q variant were found to be 1.16 Ϯ 0.08 s Ϫ1 and 55.21 Ϯ 3.10 M, respectively. This indicates considerably slower rates for substrate turnover of the E B 232Q mutant, because k cat was decreased by a factor of 90 from wild type (108 Ϯ 1.5) (3). However, the K m (xanthine) remained the same (64.4 Ϯ 0.9 for the wild type). In comparison with the E B 232A variant (3), the activity was further decreased almost 3-fold. Thus the E B 232Q variant seems to be better suitable to obtain structural information of a substrate-bound complex in contrast to the E B 232A variant.

Structural Features of the E B 232Q-
Hypoxanthine Complex-Crystals of the E B 232Q variant were incubated with its substrate hypoxanthine. The mutant crystallized in the same space group as the wild-type enzyme and displayed two (␣␤) 2 heterotetramers per asymmetric unit. The 3.4 Å structure was refined to an R factor of 0.22 and an R free of 0.27 (Table 1). A superposition of the E B 232Q variant with the native enzyme yielded a root mean square deviation of 0.5 Å for all C␣ atoms. The electron density for the substrate was best defined in the first monomer (chains A and B; Fig.  2A), and therefore the subsequent analysis is based on this molecule. The hydrogen bonds discussed for this structure have, because of the low resolution, a qualitative character and should be seen as possible hydrogen bonds. However, the clear electron density for the involved side chains and ligands allowed us to unambiguously identify the position and interacting residues of the ligand.
The six-membered ring of the substrate is sandwiched between two phenyalanine residues: about 3.4 Å from and parallel to Phe B -344 and perpendicular to Phe B -459 (Fig.  2B), forming van der Waals' interac-

TABLE 1 Data collection and refinement statistics
The values in parentheses refer to the highest resolution shell.
where F o and F c are the observed and calculated structure factors, respectively. I/ I indicates the average of the intensity divided by its average standard deviation. R free is the same as R value for the 5% of the data randomly omitted from refinement. Ramachandran statistics indicate the fraction of residues in the most favored, additionally allowed, generously allowed, and disallowed regions of the Ramachandran diagram, as defined by PROCHECK. The most important contact made is that between the C-2 of hypoxanthine and the Mo-O with a distance ranging from 2.4 to 3.3 Å in the four molecules in the asymmetric unit. The C-2 to oxygen interaction distance is closer than the 3.4 Å extent of the van der Waals' radii for the two atoms, indicating that the substrate is poised to abstract the O from the Mo-O. However, the reaction has not proceeded far enough to enable formation of a C-2ϭO bond, which would be indicated by a distance on the order of 1.2 Å. The substrate is thus strongly coordinated to residues within the active site and thereby ideally oriented for catalysis to form the product xanthine.

E232Q-hypoxanthine
We also solved the structure of the E B 232A variant in the presence of xanthine (R sym ϭ 0.13, Resolution ϭ 2.7Å Completeness ϭ 98.7%, R cryst /R free ϭ [0.242/0.286]). The structure revealed that the size of the amino acid is also a critical factor for positioning the substrate correctly. Multiple orientations of the bound xanthine molecule were observed in the active site, which could not be modeled correctly (data not shown). This, however, underlines the importance of this side chain for substrate binding and positioning in the active site.
Structure of Desulfo XDH in the Presence of Hypoxanthine-For further analysis of substrate binding and to exclude changes in substrate binding caused by the E B 232Q amino acid exchange, we pursued the structural characterization of the R. capsulatus desulfo XDH in the presence of hypoxanthine (Fig. 3). The 2.9 Å structure was refined to an R factor of 0.24 and an R free of 0.28 (Table 1). In all of the active sites in the asymmetric unit, the hypoxanthine is positioned with its C-2 toward the molybdenum and forms van der Waals' interactions with Phe B -344 and Phe B -459 and hydrogen bond interactions with Arg B -310, Thr B -460, and Glu B -232, with similar distances as observed in the E B 232Q variant. The observed hydrogen bonding pattern supports our observation from the low resolution E B 232Q structure in complex with hypoxanthine and underlines the significance of the structural data. A superposition of the E B 232Q and the desulfo hypoxanthine containing XDH structures reveals that the substrate orientation is similar in the two structures (Fig.  3C). In both structures the C-2 of the hypoxanthine points toward the molybdenum; however, the C-2 to Mo-O distance is increased approximately 0.6 Å in the desulfo form. Nevertheless both structures clearly reveal the orientation of the hypoxanthine in the active site with O-6 pointing toward Arg B -310 and C-2 being oriented toward the molybdenum cofactor.
Structure of Desulfo XDH in the Presence of Xanthine-The 2.6 Å structure of desulfo XDH in the presence of xanthine was refined to an R factor of 0.23 and an R free of 0.28 (Table 1). During the refinement process the electron density maps indicated two possible orientations of the xanthine molecule. The first orientation represents the substrate complex in which xanthine is positioned with its C-8 toward the molybdenum cofactor. The second orientation provides a view of the product complex after hypoxanthine has been oxidized to xanthine, i.e. the O-2 of xanthine is positioned toward the molybdenum cofactor. The resolution of 2.6 Å does not support the full refinement of alternate conformations. However, because only these two conformations are supported by the electron density maps, we opted to refine them with an occupancy of 0.5 each. Both orientations reveal again an intricate hydrogen-bonding network with active site residues. In the substrate orientation the xanthine molecule forms interactions with Arg B -310, Thr B -460, and Glu B -232 (Fig. 4A), albeit with different atoms of the sub- strate because of the altered orientation compared with the hypoxanthine molecule. O-6 of xanthine forms hydrogen bonds to the main chain N of Thr B -460 as well as to the guanidinium group of Arg B -310. The opposite side of the substrate is held in place through a hydrogen bond formed between Glu B -232 to N-9. Further, Phe B -344 and Phe B -459 form van der Waals' interactions to the substrate as observed in the hypoxanthine structure. In all four active sites the C-8 atom of xanthine is positioned toward the molybdenum cofactor at a distance of 2.5-3.2 Å, indicating that the substrate is poised to abstract the O from the Mo-O, but the reaction cannot proceed because of the utilization of the desulfo form of XDH.
In the product-bound form, in which the xanthine molecule points with its O-2 atom toward the molybdenum cofactor, the molecule is held in place by the same interactions as described for the hypoxanthine structure, although the two molecules are slightly shifted with respect to each other so that the O-2 of xanthine occupies approximately the same position as the C-2 in hypoxanthine, thus providing a glance how the substrate hypoxanthine is repositioned once it is oxidized to xanthine and is subsequently released from the substrate-binding pocket (Fig. 4B). Because the more relevant structure is revealed by the substrate complex, all comparisons with other complexes and structures were carried out using the substrate structure with the C-8 positioned toward the Mo-O.
Despite the differences of the two substrate complexes as well as the product complex, the same amino acids within the active site are utilized to form hydrogen bonds. The hypoxanthine and the xanthinebound structures thus clearly identify Glu B -232, Arg B -310, and Thr B -460 as important residues to position the substrates optimally for catalysis.
Steady State Kinetics and Structural Features of Wild-type XDH with Pterin-6-aldehyde-To analyze the inhibitory effects of pterin-6-aldehyde on R. capsulatus XDH, steady state kinetics were performed with varying concentrations of xanthine and pterin-6-aldehyde using NAD ϩ as the electron acceptor. A representative Lineweaver-Burk plot is shown in Fig. 5. As reported previously for XO, pterin-6-aldehyde exhibits a competitive inhibition pattern. Analysis of the kinetic results showed a K i value of 103.57 Ϯ 18.96 nM, which is 100-fold higher than that reported for XO (1.78 nM) (26). The 3.3 Å crystal structure of wild-type XDH in complex with pterin-6-aldehyde reveals that the two rings of the inhibitor are positioned in a slightly different orientation compared with hypoxanthine and xanthine; pterin-6-aldehyde has undergone a small lateral translation as well as an almost 30°rotation away from the cofactor, effectively blocking access to the binding pocket (Fig. 6). The active site maintains its architecture, and the overall superposition reveals a 0.4 Å root mean square deviation compared with the E B 232Q-hypoxanthine structure for all C␣ atoms and a root mean square deviation of 0.3 Å for the xanthine-bound structure. The inhibitor can be well placed within the electron density, which is best defined in the third monomer, and all distance calculations derive from this complex. Like the substrate-bound structures, the inhibitor is sandwiched between the two phenyl rings Phe B -344 and Phe B -459. The hydrogen bond network is increased compared with the xanthine and hypoxanthine-bound structures, leading to an involvement of Glu B -730, being in hydrogen bonding distance via its carboxylate group to the N-6 of pterin-6-aldehyde. Glu B -730 does not form any substrate interactions in the other complexes within our analysis. In addition the N-6 of pterin-6-aldehyde is likely to form a hydrogen bond to the Mo-O of the Moco. Unlike hypoxanthine and xanthine, which have no amino group jutting out from the ring structures, pterin-6-aldehyde can also form a hydrogen bond through this moiety to the Mo-O ligand, thereby locking the inhibitor tightly into place. N-4 can contact the other catalytically important glutamic acid, Glu B -232, via its carboxylate group. O-8, like O-6 of hypoxanthine and xanthine, is in hydrogen bonding distance to the guanidinium group of Arg B -310 and to the main chain N of Thr B -460. The aldehyde group only forms hydrophobic interactions with the neighboring Pro B -306 and Leu B -461. The high number of contacts indicates that the inhibitor is firmly situated within the active site, thereby inhibiting access of native substrates.

DISCUSSION
We cocrystallized the R. capsulatus XDH variant E B 232Q with hypoxanthine in addition to the desulfo enzyme with hypoxanthine and xanthine. These structures provide a thorough insight into the binding of both substrates in the active site. The structure of the desulfo enzyme in the presence of xanthine revealed two possible xanthine orientations. One of them being the substrate orientation the other one being xanthine in a product-bound state from the conversion of hypoxanthine to xanthine. Because the desulfo enzyme is catalytically incompetent, and the xanthine concentration in the experiment is a thousand times above the K D for xanthine, both orientations are likely to occur. This is further supported by the well known feature of substrate inhibition of eukaryotic XORs (27,28).
The wild-type enzyme shows substrate inhibition at xanthine concentrations Ͼ200 M (data not shown), which is well in line with previously measured xanthine concentrations for substrate inhibition of XORs from various organisms. Thus our structure describes a fea-  sible mechanism for the observed xanthine substrate inhibition of XORs. Furthermore, the structure of the wild-type enzyme was solved with the inhibitor pterin-6-aldehyde. Crystallization involving the R. capsulatus XDH E B 232A variant in the presence of xanthine revealed that the position of the substrate could not be clearly defined because of the openness of the active site resulting from the substitution of the glutamate by an alanine side chain (data not shown). This result indicates the importance of Glu B -232 for substrate binding and positioning at the active site. In contrast, the E B 232Q variant preserved the active site architecture and has lost the biochemical properties of a glutamate, and thus the variant maintained binding of hypoxanthine in a similar manner as the inactive desulfo enzyme.
The R. capsulatus XDH E B 232Q variant and the desulfo form allowed the first structural characterization of substrate-bound complexes and demonstrate that both molecules are ideally positioned for catalysis within the active site through a combination of stacking interactions and hydrogen bonds. Intriguingly both substrates utilize the same amino acids for the formation of hydrogen bonds despite their different orientation. Compared with pterin-6-aldehyde, however, both are more loosely held in place in terms of overall interactions within the active site, thereby facilitating product release and subsequent oxidation to uric acid. This suggests that the inhibitor is situated more tightly within the active site and thereby prevents the binding of native substrates.
The E B 232Q and the desulfo hypoxanthine complex both reveal that the C-2 atom approaches the Mo-O moiety at distances of 2.9 -3.9 Å, respectively (Figs. 2 and 3). The C-8 atom in the xanthine-bound structure approaches the Mo-O at a distance of ϳ3.0 Å (Fig. 4). In contrast, the previously determined 3.0 Å structure of R. capsulatus XDH with oxypurinol ( Fig. 7) showed that its N-2 atom coordinates directly to the molybdenum and obscures the hydroxyl moiety of the cofactor (oxypurinol numbering scheme taken from Massey (8,15)). All three molecules form contacts to residues Glu B -232 and Arg B -310 at similar distances, whereas oxypurinol also inter-acts with the catalytic Glu B -730. Most striking, however, is the direct comparison between the orientation of the bound substrate xanthine and the inhibitor oxypurinol. Although the O-6 of xanthine forms a hydrogen bond to Arg B -310, it is the O-6 of oxypurinol that interacts with the same residue, whereas O-4 of the inhibitor forms hydrogen bonds to Glu B -232. This orientation is stabilized through additional interactions formed between Glu B -730 with the N-3 and through a water-mediated hydrogen bond to N-7 of the inhibitor. These interactions are only possible through the tight approach of the inhibitor toward the molybdenum and thereby stabilize an orientation, which is opposite to the one observed in the xanthine-bound structure (Fig. 4C). Xanthine cannot form these interactions with a carbon at position 8 and therefore does not enter the active site as deeply to generate direct interactions to Glu B -730.
Prior to our structural studies it has been proposed that the O-6 of xanthine interacts with Arg B -310 in the active site (6). The orientation of xanthine was supported by kinetic studies utilizing a homologous series of purines as substrates in addition to mutant studies of Arg B -310 (6). Our analysis of the hypoxanthine and xanthine-bound structures and the previous study clearly support the importance of Arg B -310 for correct positioning of the substrates in the active site. Arg B -310 does not interact directly with the two carbon atoms that undergo the hydroxylation reaction but has been proposed to lower the activation energy for the reaction by stabilizing negative charge accumulation on the heterocycle through an electrostatic interaction with the C-6 carbonyl oxygen of the substrate. All of the structures clearly reveal hydrogen bonds between the O-6 of the substrate and the guanidinium group of Arg B -310, thus providing further support for the importance of this residue. Our results are, however, in contrast to a model proposed for the human xanthine oxidoreductase by Yamaguchi et al. (29), who suggested an orientation of the substrate similar to that observed for oxypurinol. Based on the structural identity of the active sites, however, and because the orientation of their substrate is hypothetical without supporting structural data, we believe that eukaryotic XORs bind the substrate in the same orientation as observed in our bacterial XDH complexes.
In contrast, the crystal structure of bovine XO in the course of the reaction with the poor substrate 2-hydroxy-6-methylpurine has been solved (7), in which the bound 2-hydroxy-6methylpurine molecule was clearly oriented with the C-2 rather than the C-6 position oriented toward Arg-880 (corresponding to Arg B -310 in R. capsulatus XDH). However, because 2-hydroxy-6-methylpurine contains a methyl group at the C-6 position, it was expected to bind in the opposite direction at the active site because of steric hindrance, as suggested previously by kinetic studies (6). The small 5 ϫ 3 ϫ 5 Å rectangular prismic nature of the active site-binding channel dictates that substrates, key inhibitors, and drugs are small planar molecules. All of the molecules evaluated here superimpose well in a small area sandwiched between two phenylalanine rings (Phe B -344 and Phe B -459 in R. capsulatus and Phe-913 and Phe-1008 in the bovine enzyme), which dictate the necessity for a planar molecule (Fig. 7). Glu B -730, Glu B -232, Arg B -310, and Thr B -460 in R. capsulatus and the corresponding residues Glu-1261, Glu-802, Arg-880, and Thr-1010 in the bovine enzyme define the other two walls of the binding pocket and provide the necessary atoms to form hydrogen bonds, which position the substrates for catalysis and as proposed for Arg B -310 (Arg-880 in the bovine enzyme) lower the activation energy of the reaction. One of the ring moieties of each of the six substrates and inhibitors therefore assume similar positions and stack on top of each other. The other ring is either in close proximity to the molybdenum cofactor, in the case of hypoxanthine, xanthine, and oxypurinol, or pushed back toward the entrance of the active site-binding channel, as observed for pterin-6-aldehyde.
Because all of the structures analyzed here contain four molecules in the asymmetric unit, a comparison of the substrates bound in the active site is possible. In both the hypoxanthine and xanthine structures, variability in the hydrogen bond distances was observed, thus indicating that the orientation of the substrate is well defined, but the position is not exactly maintained. This observation supports the hypothesis that a balance has to be achieved, which allows the substrates to be ideally positioned for catalysis but also held loosely enough in place to facilitate subsequent product release. The C-2 of the hypoxanthine is located in close proximity to an oxo ligand within the coordination sphere of the Mo. This would enable oxygen transfer from the Mo-O to the C-2 of the hypoxanthine, converting it to xanthine. Arg B -310 and Glu B -232 serve to orient the substrate in the proper manner for oxidation. pK a value calculations for Glu B -232 (calculated with PROPKA) (30) indicate a pK a of 8.1 for this residue, underlining its ability to form hydrogen bonds to the corresponding substrates. The surrounding residues anchor the substrate during catalysis while maintaining an unobstructed substrate channel, which facilitates substrate reemergence into the solvent. The space within the substrate-binding pocket is ideally designed to coordinate and position the substrates for catalysis. This tight binding pocket, however, obstructs reorientation of the substrate within the pocket to allow oxidation at the C-2 position, followed by oxidation at C-8. Our structures therefore suggest that after the first round of oxidation, xanthine has to retreat at least 16 Å from the Moco to acquire adequate space to rotate the C-8 into position close to the Mo atom for further oxidation. The xanthine-bound structure clearly shows that the C-8 is again located in close proximity to the planar oxo ligand of the molybdenum cofactor, allowing further oxidation to uric acid. Our crystal structures corroborate the canonical mechanism of substrate turnover within the bacterial XDH system. Hypoxanthine and xanthine are optimally positioned to coordinate their C-2 or C-8, respectively, to one of the oxo ligands within the molybdenum cofactor coordination sphere.
Pterin-6-aldehyde displays multiple binding contacts to key residues within the active site cavity. This double ringed moiety is firmly situated for enzyme inhibition, unlike the substrates hypoxanthine and xanthine, which are more loosely held to allow product release after oxidation.