Closed Site Complexes of Adenine Phosphoribosyltransferase fromGiardia lamblia Reveal a Mechanism of Ribosyl Migration*

The adenine phosphoribosyltransferase (APRTase) from Giardia lamblia was co-crystallized with 9-deazaadenine and sulfate or with 9-deazaadenine and Mg-phosphoribosylpyrophosphate. The complexes were solved and refined to 1.85 and 1.95 Å resolution. Giardia APRTase is a symmetric homodimer with the monomers built around Rossman fold cores, an element common to all known purine phosphoribosyltransferases. The catalytic sites are capped with a small hood domain that is unique to the APRTases. These structures reveal several features relevant to the catalytic function of APRTase: 1) a non-proline cis peptide bond (Glu61–Ser62) is required to form the pyrophosphate binding site in the APRTase·9dA·MgPRPP complex but is a trans peptide bond in the absence of pyrophosphate group, as observed in the APRTase·9dA·SO4complex; 2) a catalytic site loop is closed and fully ordered in both complexes, with Glu100 from the catalytic loop acting as the acid/base for protonation/deprotonation of N-7 of the adenine ring; 3) the pyrophosphoryl charge is neutralized by a single Mg2+ ion and Arg63, in contrast to the hypoxanthine-guanine phosphoribosyltransferases, which use two Mg2+ ions; and 4) the nearest structural neighbors to APRTases are the orotate phosphoribosyltransferases, suggesting different paths of evolution for adenine relative to other purine PRTases. An overlap comparison of AMP and 9-deazaadenine plus Mg-PRPP at the catalytic sites of APRTases indicated that reaction coordinate motion involves a 2.1-Å excursion of the ribosyl anomeric carbon, whereas the adenine ring and the 5-phosphoryl group remained fixed.G. lamblia APRTase therefore provides another example of nucleophilic displacement by electrophile migration.

Adenine phosphoribosyltransferase (APRTase) 1 catalyzes the reversible Mg 2ϩ -dependent transformation of adenine and 5-phospho-␣-D-ribosyl-1-pyrophosphate (PRPP) to AMP and pyrophosphate ( Fig. 1; Ref. 1). The primary functions of APRTase, present in bacteria, yeast, plants, and mammals, are adenine salvage and adenine recycling (2)(3)(4). In humans, APRTase has the sole metabolic function of recycling adenine formed in the polyamine pathway. A deficiency of APRTase results in the inappropriate oxidation of adenine to 2,8-dihydroxyadenine, an insoluble metabolite that causes urolithiasis and is capable of causing kidney failure after several decades of accumulation (5). The symptoms are variable and are reversible upon removal of 2,8-dihydroxyadenine. The relatively mild effects of APRTase deficiency in humans make APRTase a possible target for the treatment of parasite infections by disrupting the pathways of adenine salvage and polyamine biosynthesis.
Giardia lamblia is a human and animal parasite that causes the most common protozoan infection in North America (6,7). Protozoan parasites, including G. lamblia, are deficient in de novo purine synthesis and require purine uptake from the host, as well as efficient purine salvage. The stringent purine economy makes APRTase especially important in these organisms. Giardia expresses two distinct purine ribosyltransferases (PRTases), APRTase and GPRTase, for purine base salvage rather than the broad-specificity HGXPRTase found in the Plasmodia (8). GPRTase from Giardia prefers guanine as the purine substrate; its catalytic mechanism has been well studied by enzyme kinetics, mutagenesis, and x-ray crystallography (9 -11). The deduced sequences of APRTases from genome sequences were summarized recently, but only two crystallographic studies have been reported, neither of which exhibit closed catalytic loops (12)(13)(14).
The first APRTase crystal structures were the enzyme from Leishmania donovani in complex with adenine, AMP, and citrate-sulfate ions (13). APRTase possesses a Rossman fold core resembling other purine PRTases (15). It also contains a long catalytic loop that is proposed to close on the active site during catalysis. The catalytic site loop was found in the open conformation and has been well defined as a consequence of crystal packing forces in the L. donovani APRTase structure. The L. donovani enzyme is classified as a long-APRTase that has Nand C-terminal extensions of over 50 amino acids relative to the amino acid sequences of the short-APRTases, which include the human, yeast, Escherichia coli, and Giardia enzymes. The structure of the first short-APRTase structure was recently reported using the enzyme expressed by Saccharomyces cerevisiae (12). Giardia APRTase is also a short-APRTase and shares a 35% amino acid sequence identity with the enzyme from yeast. The goal of the present work was to establish the crystal structures of an APRTase with a closed catalytic site loop. In other PRTases, this loop makes contact with substrates in the catalytic sites and is required for catalysis. Resolution of the loop contacts in the APRTases is required to understand the molecular basis for adenine specificity, the atomic excursions during catalysis, as well as to assist in inhibitor design.
Giardia APRTase crystallized in the presence of 9-deazaadenine (9dA) or 9dA/PRPP yielded APRTase⅐9dA⅐SO 4 and APRTase⅐9dA⅐MgPRPP complexes. Both complexes demonstrated ordered catalytic loops closed over the active site. One Mg 2ϩ ion is bound to the pyrophosphate group of PRPP in contrast to the two Mg 2ϩ ions observed in other purine PRTases including human, Trypanosoma cruzi and Toxoplasma gondii HGPRTases and Plasmodium falciparum HGX-PRTase (16 -19). The structures described here are the first APRTase structures to be reported with closed catalytic loops.

MATERIALS AND METHODS
Crystallization-Cloning, expression, and purification of Giardia APRTase is described in Sarver et al. (14). For crystallization of the Giardia APRTase⅐9dA⅐SO 4 complex, protein in 50 mM Hepes, pH 6.0, 8 mM MgCl 2 , and 1 mM DTT was mixed with 1:2 molar ratio of 9dA and iminoribitol, and 1 mM sodium pyrophosphate. The mixture was incubated on ice for 45 min before preparing crystallization drops. The crystals were obtained by mixing 2 l of 10 mg/ml protein with an equal volume of mother liquid containing 100 mM sodium acetate (pH 4.6), 24% polyethylene glycol 4000, 0.2 M ammonium sulfate, and 0.05 M urea followed by equilibration against 1.0 ml of the mother liquid at 18°C. For crystallization of the APRTase⅐9dA⅐MgPRPP complex, protein was mixed with a 1:2 molar ratio of 9dA and 10 mM PRPP prior to the crystallization. The precipitant for the APRTase⅐9dA⅐MgPRPP crystals was similar to that for the APRTase⅐9dA⅐SO 4 crystals except that ammonium acetate replaced ammonium sulfate. Single rod-shaped crystals with hexagonal faces appeared overnight and grew to a maximum size of 0.05 ϫ 0.05 ϫ 0.3 mm 3 . These crystals exhibited diffraction consistent with the space group P3 1 21 with a ϭ 54.3 Å, b ϭ 54.3 Å, and c ϭ 108.8 Å. The crystals contain a monomer in the asymmetric unit, corresponding to V m ϭ 2.2 Å 3 /Da and a solvent content of 44%.
Data Collection and Processing-APRTase crystals were soaked in crystallization buffer containing 20% glycerol for 2 min and rapidly frozen in cold nitrogen stream at Ϫ178°C. X-ray diffraction data were collected from single crystals of APRTase using a Quantum CCD detector at beam line X9B ( ϭ 0.98 Å, NSLS, Brookhaven National Light Source). Indexing, integration, and scaling of the data were performed with the HKL package (20). The data set for the APRTase⅐9dA⅐SO 4 complex has an overall completeness of 98.8% to 1.85 Å resolution with an R sym of 2.7%. The APRTase⅐9dA⅐MgPRPP complex data set was 99.8% complete to 1.85 Å resolution with an R sym of 3.7%. Data collection statistics are shown in Table I.
Molecular Replacement and Structural Refinement-The Giardia APRTase⅐9dA⅐SO 4 complex was solved by molecular replacement with AMoRe implemented in CCP4 (21) using the yeast APRTase structure as the search model (12). All non-identical side chains were trimmed to alanine. A single solution with a correlation coefficient of 29.3% was generated using 8.0 -4.0 Å data. Rigid body refinement using the 8.0 -4.0 Å data resulted in R cryst and R free values of 51.6 and 53.2%, respec-tively. Program O (22) was used to view the structure and build the missing side chains into the electron densities in the 2F o Ϫ F c map. The final structure contains residues 2-180, two histidine residues from the C-terminal His-tag, a single 9dA, three sulfate ions, and a total of 90 solvent molecules. The final values for R cryst is 21.9% and for R free 25.1% (Table I).
The Giardia APRTase⅐9dA⅐MgPRPP complex structure was refined using the structure of APRTase⅐9dA⅐SO 4 complex as the starting model. The refinement was carried out to 1.95 Å resolution with R cryst and R free values of 21.8 and 26.2%, respectively, using the program CNS (Ref. 23; Table I). The final model includes residues 2-180, a single 9dA, one Mg 2ϩ ion, one PRPP, and a total of 80 solvent molecules. Both structures show excellent geometry with 99% of the residues in the most favored and additionally allowed region of the Ramachandran plot by PROCHECK (24). Peptide segment Glu 61 -Ser 62 adopts a trans peptide bond conformation in the APRTase⅐9dA⅐SO 4 complex but is in the cis conformation in the APRTase⅐9dA⅐MgPRPP complex. In both structures, these amino acids fit well in their corresponding electron densities in the 2F o Ϫ F c maps.
Quaternary Structure-Giardia APRTase is found as a homodimer in both crystal structures (Fig. 2b). The enzyme crystallized in the trigonal space group P3 1 21 with a monomer in the asymmetric unit. Application of the crystallographic 2-fold symmetry operator generates a dimer that buries a total surface area of 1400 Å 2 from each monomer (2800 Å 2 total). The dimer interface for Giardia APRTases includes ␣2Ϫ␣3 and ␣4 of the first monomer with ␤4 and ␣4 of the second monomer (Fig. 2b). These interactions consist of 4 salt bridges, 20 neutral hydrogen bonds, and 190 van der Waals interactions (Table II). Of the 12 residues involved in hydrogen bonds at the dimer interface, only Asp 27 and Arg 83 are completely conserved throughout the known short-APRTases (12). The dimer interfaces in Giardia and yeast APRTases are similar. In contrast, the long-form L. donovani APRTase has an extra 30 amino acid residues at the C terminus that are wrapped around the neighboring monomer to contribute an additional 1250 Å 2 buried surface area from each monomer (5300 Å 2 total) (13).
APRTase⅐9dA⅐SO 4 Complex-The adenine binding site is defined by the position of 9dA, which is bound in a cleft formed by the hood and by the loop connecting ␤8 and ␣6 in the large ␤-sheet central to the core (Fig. 4a). The 9dA ring is stacked between the hydrophobic side chains of Phe 25 from the hood and Val 126 from the core. The backbone carbonyl of Ala 24 and the ⑀-amino of Lys 26 from the hood form hydrogen bonds to the exocyclic N-6 amine and to N-1 of 9dA. A carboxyl oxygen of Glu 100 from the catalytic loop forms a 2.7 Å hydrogen bond with N-7 of 9dA, a major contact between the substrate and the catalytic loop in its closed conformation. The reactions catalyzed by PRTases have been proposed to involve oxacarbenium transition states in which the purine leaving group is activated   by protonation or strong hydrogen bond formation to N-7 (19,25). In human HGPRTase, Glu 137 is proposed to play the same role as Glu 100 in Giardia APRTase (16,26). However, the amino acid sequence alignment between APRTases and HGPRTases reveals insufficient homology to predict this role for Glu 100 in APRTase (12,14). In the sequence alignment of APRTase, this Glu is conserved in all enzymes with demonstrated APRTase activity (12,14). Mutagenesis of this conserved Glu to Leu in yeast APRTase reduced k cat /K m by a factor of 10 6 and mutation from Glu to Gln decreased k cat /K m by 10 3 , as a Gln residue at this position retains the ability for hydrogen bond formation but is not competent to act as an acid/base (12).
The initial crystallization of Giardia APRTase occurred in the presence of 0.2 M (NH 4 ) 2 SO 4 , resulting in a total of three sulfate ions in the APRTase⅐9dA⅐SO 4 structure. One sulfate ion is bound at the site typically occupied by the 5Ј-phosphate of AMP or PRPP molecules (Fig. 4, a and b). This sulfate ion binds to the turn formed by residues 128 ATGGT 132 , which is defined as the 5Ј-phosphate binding loop. The peptide backbone amide groups of these residues orient toward the sulfate and contribute four direct hydrogen bonds. The hydroxyl groups of Thr 129 and Thr 132 provide two additional hydrogen bonds to this sulfate ion. Amino acids corresponding to Ala 128 -Thr 132 in the 5Ј-phosphate binding loop are highly conserved in all APRTases, which is consistent with the requirement of a 5Ј-phosphate for substrate activity. The APRTase⅐9dA⅐SO 4 complex has Arg 63 located near the sulfate ion site. However, when the catalytic site is full, Arg 63 swings across the active site to bind pyrophosphate (Fig. 4, a and b; Fig. 5

).
A second sulfate ion is bound near the active site but does not occupy any site involved in substrate or product binding. It participates in hydrogen bonds with backbone and side-chain atoms of Glu 61 -Ser 62 . Residues Glu 61 , Ser 62 , and Arg 63 are highly conserved throughout all APRTases and are essential elements of the pyrophosphate binding site. In the APRTase⅐9dA⅐SO 4 complex, the peptide segment connecting Glu 61 and Ser 62 is in trans conformation in contrast to a cis conformation found in the APRTase⅐9dA⅐MgPRPP complex (Fig. 5). The equivalent peptide bond in related HGPRTases has been proposed to shift from trans to cis conformation to bind the ␤-phosphoryl moiety in phosphoribosyltransferase reactions; however, both trans and cis conformations have been observed in high resolution crystal structures of type I PRTases when the pyrophosphate binding site is empty (11,12,18). The sulfate ion in the Giardia APRTase⅐9dA⅐SO 4 complex is bound in a location opposite to the ␤-phosphoryl binding site, stabilizing the Glu 61 -Ser 62 peptide segment in a trans conformation and preventing formation of the pyrophosphate binding site.
The third sulfate ion is bound near the C terminus and forms hydrogen bonds with two of the histidine residues in the Histag. This sulfate ion is a result of the His tag expression system and is not relevant to catalysis; however, it may have facilitated the crystallization of the enzyme.
APRTase⅐9dA⅐MgPRPP Complex-Replacing (NH 4 ) 2 SO 4 in the crystallization buffer with NH 4 Ac allowed Mg 2ϩ and PRPP to fill the active site in the APRTase⅐9dA⅐MgPRPP complex (Fig. 3b). The overall structures of APRTase in the two complexes are very similar, with root-mean-square deviations of 0.35 Å for all C␣ atoms. Binding of 9dA is in a similar hydrophobic and hydrogen bond network as observed in the APRTase⅐9dA⅐SO 4 complex (Fig. 4, a and b). Beneath the 9dA, PRPP binds in the active site with C-1 of the ribose ring 3.4 Å away from C-9 of 9dA. The monophosphate group of PRPP is bound in the highly conserved 5-phosphate binding loop. The O-2 and O-3 hydroxyl oxygens of ribose are anchored by hydrogen bonds (2.6 and 2.7 Å) to carboxyl groups of Glu 124 and Asp 125 , the two highly conserved acidic residues found at the catalytic sites of all known purine PRTases. The Mg 2ϩ ion is octahedral-coordinated, making contact with O-2 and O-3 (both at 2.5 Å), O-2A and O-3B (2.3 and 2.2 Å, respectively) of PRPP, and two water molecules (both at 2.1 Å; Fig. 6). The Mg 2ϩ ion orients the ribose ring and neutralizes part of the negative charge from the pyrophosphate moiety. Both APRTase and OPRTase utilize a single Mg 2ϩ ion at the catalytic site, whereas other purine PRTases sandwich the pyrophosphate group between two Mg 2ϩ ions (Fig. 4, b and d). OPRTase and APRTase solve the problem of pyrophosphate charge stabilization by replacing the second Mg 2ϩ ion with Arg or Lys groups.
Residues Glu 61 , Ser 62 , and Arg 63 provide the major amino acid contacts in the pyrophosphate binding site. Peptide Glu 61 -  Ser 62 exhibits the unusual cis conformation that orients the amide nitrogen atoms of Ser 62 and Arg 63 to participate in hydrogen bonds with the ␤-phosphoryl of PRPP (2.9 and 3.0 Å, respectively). The side chain of Arg 63 is also involved in hydrogen bonds with the ␤-phosphoryl and the N-3 of 9dA (3.0 and 2.8 Å, respectively). Arg 63 is completely conserved throughout all APRTases, and mutation of this Arg residue to Ala in yeast APRTase resulted in a 10 5 -fold decrease in k cat /K m (12). Tyr 101 from the catalytic loop hydrogen bonds to the ␣-phosphoryl group of PRPP and is also completely conserved throughout APRTases. Mutation of the corresponding Tyr to Phe in the yeast enzyme reduced the k cat /K m by 10-fold (12). The hydrogen bond network at the catalytic site of the Giardia APRTase⅐9dA⅐MgPRPP complex indicates at least one hydrogen bond or metal-chelate interaction at every donor/acceptor site found on the ligands, except for O-3A (Fig. 7).
Comparison of Giardia APRTase with Other APRTases-Giardia APRTase shares 33-35% sequence identity with the APRTases from S. cerevisiae and L. donovani. The overall structures are similar, with root-mean-square deviations of 1.3 and 1.7 Å, respectively, over the conserved 163 C␣ atoms with the catalytic loop excluded from the comparison. The present structure of Giardia APRTase departs from both the yeast and L. donovani structures reported previously by the closed catalytic site loop and from the L. donovani structure by the absence of the C-terminal extension (12,13). Despite the missing residues at the C terminus, Giardia APRTase still forms an extensive dimer interface.
The structure of L. donovani APRTase was solved with AMP and citrate ion (a component in the crystallization buffer) in the active site (Fig. 4c). The location of the adenine group of AMP is similar to that of 9dA observed in Giardia APRTase, which stacks Phe 27 and Val 126 corresponding to Phe 42 and Val 148 in the L. donovani APRTase. Hydrogen bond patterns for the purine rings in the two complexes are also similar except for N-7, which is hydrogen bonded to the protein only in the closed-loop Giardia enzyme. The 5Ј-phosphate binding pocket is highly conserved for all type I PRTases, and the complexes of Giardia ARTase⅐9dA⅐MgPRPP and L. donovani APRTase⅐AMP⅐cit provide no exception, except for the hydrogen bond from the amide Glu 100 (Glu 107 in L. donovani), which does not form in the L. donovani complex because of the open catalytic loop.
The ribose groups of AMP and PRPP in the APRTase complexes show substantial differences in sugar pucker near C-1. The C-1 ribosyl atoms are positioned ϳ2.1 Å apart in the superimposed structures, consistent with the enzymatic mechanism of nucleophilic substitution by electrophile migration that has been recognized for several sugar transferases (27)(28)(29). The PRTase reactions are nucleophilic substitutions in which both attacking and leaving group nucleophiles are held tightly in their respective binding pockets, and the C-1Ј carbon of AMP translocates between the relatively immobile nucleophiles during catalysis (27,28). This distance is also 2.1 Å in the HGPRTases, and the 2.1 Å distance observed between the C-1 atom of PRPP and C-1Ј of AMP provides supporting evidence that Giardia APRTase uses this same catalytic strategy (see below).
The electron density assigned as a Mg 2ϩ ion and reported in the L. donovani APRTase⅐AMP⅐cit complex is ϳ1.5 Å distant from that observed here for Giardia APRTase⅐9dA⅐MgPRPP. However, this atomic assignment is questionable for the L. donovani complex because the proposed Mg 2ϩ ion has only three ligands in the coordination sphere, and these are all at 2.7-2.9 Å (13). Magnesium ions in crystallographic complexes typically demonstrate six coordinated ligands at distances from 2.0 -2.5 Å (30), suggesting that the electron density assigned as a Mg 2ϩ ion in the L. donovani complex may represent a solvent molecule. Despite the molecular differences between citrate and pyrophosphate ions, the amino acids that form the pyrophosphate binding sites are in similar positions in the Giardia and L. donovani complexes with bound PPi and citrate, respectively.
The closed-loop structure at the active site of Giardia APRTase⅐9dA⅐MgPRPP demonstrates closer contacts between the protein and catalytic site contents than those observed in the L. donovani APRTase⅐AMP⅐cit complex (Fig. 8). This loop motion moves the tip of the loop (Glu 100 ) by Ͼ10 Å to exclude solvent access to the otherwise exposed adenine in its binding site and also to cover a portion of the PRPP site. Motion of this loop is required for entry and release of components to the catalytic site. It is proposed that catalysis occurs only in the closed-loop complex because of the contacts to substrate. N-7 deprotonation/protonation of the adenine is an important feature of APRTase catalysis and the 2.7 Å carbonyl group of Glu100 interaction to N-7 of the 9dA is seen only in the closedloop complex. This interaction provides the general acid/base catalyst, and in the other purine PRTases, forms a novel downfield hydrogen bond in complexes with transition state ana-logue inhibitors (19,31). As discussed above, the E100L mutation in the yeast APRTase reduced catalytic efficiency to 10 Ϫ6 of the parental enzyme.
The dimer interfaces of APRTase and OPRTase are also similar but differ from those of other members in the type I PRTases. However, the catalytic site loops of APRTase and OPRTase differ in that the catalytic loop in Salmonella OPRTase closes onto the active site of the adjacent monomer, whereas the catalytic loop in APRTase closes onto a self-contained active site (32). Giardia APRTase crystallized to give a monomer in the asymmetric unit, thus the two active sites in the dimeric structure are related by a crystallographic 2-fold axis and are identical to each other. The recently reported Salmonella OPRTase⅐orotate⅐MgPRPP complex showed asymmetric properties between monomers in its homodimer (32). One active site is fully occupied and is closed with the catalytic loop from the adjacent monomer. The other active site is open and disordered, suggesting a single-site binding sequential catalysis mechanism for OPRTase. The catalytic sites of APRTase are identical and equally filled. The structural simi- larity between OPRTase and APRTase suggests that these enzymes share a common evolutionary progenitor, which is distinct from other PRTases.
Reaction Coordinate Motion in APRTases-The availability of APRTase structures with bound substrate and product analogues (compare Fig. 4, panels b and c), permits comparison of these molecules in relation to portions of the proteins that are constant through the reaction. This approach has been used previously for the N-ribosyltransferases: purine nucleoside phosphorylase and hypoxanthine-guanine phosphoribosyltransferase (27,28). When the AMP from the catalytic site of L. donovani is overlaid with 9dA and PRPP in Giardia, it is apparent that the major atomic motion involves translation of the C-1 portion of the ribosyl ring, whereas the purine base, 5-monophosphate, and the pyrophosphate remain relatively fixed (Fig. 9). The 2.1-Å excursion of C-1 of the ribosyl groups is within the range of 1.5 to 2.1 Å observed in the other Nribosyltransferases (28). Thus, the APRTases represent another example in which nucleophilic displacements occur with inversion of configuration by migration of the electrophile between two fixed nucleophiles.
Summary and Conclusions-The structures of Giardia APRTase⅐9dA⅐SO 4 and APRTase⅐9dA⅐MgPRPP are the first APRTase complexes with a closed flexible catalytic site loop. Glu 100 at the tip of the catalytic loop has been identified as an acid/base for catalysis. Similar to OPRTase, APRTase utilizes only one Mg 2ϩ ion in the active site with Arg 63 replacing the second Mg 2ϩ found in HGPRTases. A comparison of APRTases with substrate and product analogue complexes in the catalytic sites supports a reaction mechanism of ribosyl group migration between relatively fixed adenine and Mg 2ϩ -pyrophosphate groups. This mechanism has been recognized in other N-ribosyltransferases.