Identification of a Subversive Substrate of Trichomonas vaginalis Purine Nucleoside Phosphorylase and the Crystal Structure of the Enzyme-Substrate Complex*

Trichomonas vaginalis is an anaerobic protozoan parasite that causes trichomoniasis, a common sexually transmitted disease with worldwide impact. One of the pivotal enzymes in its purine salvage pathway, purine nucleoside phosphorylase (PNP), shows physical prop-erties and substrate specificities similar to those of the high molecular mass bacterial PNPs but differing from those of human PNP. While carrying out studies to identify inhibitors of T. vaginalis PNP (TvPNP), we discov-ered that the nontoxic nucleoside analogue 2-fluoro-2 (cid:1) deoxyadenosine (F-dAdo) is a “subversive substrate.” Phosphorolysis by TvPNP of F-dAdo, which is not a substrate for human PNP, releases highly cytotoxic 2-flu-oroadenine (F-Ade). In vitro studies showed that both F-dAdo and F-Ade exert strong inhibition of T. vaginalis growth with estimated IC 50 values of 106 and 84 n M , respectively, suggesting that F-dAdo might be useful as a potential chemotherapeutic agent against T. vaginalis . To understand the basis of TvPNP specificity, the structures of TvPNP complexed with F-dAdo, 2-fluoro-adenosine, formycin A, adenosine,

Trichomonas vaginalis is an anaerobic protozoan parasite that causes one of the most common sexually transmitted infections in the world. In the United States alone there are an estimated 5 million new cases occurring annually. Infection with the organism often causes vaginitis in women and urethritis in men and is also related to many other diseases (1). Trichomoniasis has important medical and social implications and augments predisposition to HIV infection (2)(3)(4). Metronidazole is the only available drug for treating the disease. But it is teratogenic, carcinogenic in rodents, and has an Antabuse-like reaction with alcohol (5). Isolates of T. vaginalis resistant to metronidazole have been identified since 1989 (6). A new approach to chemotherapeutic control of this pathogen is urgently needed.
Parasitic protozoa lack de novo synthesis of purine nucleotides in general and must depend on purine salvage to replenish their purine nucleotide pools to survive (7). The concept of exploiting this particular metabolic deficiency among these organisms for anti-parasitic chemotherapeutic gain has been accepted and pursued by many research laboratories in the last 20 years (8). Some successful studies in targeting purine salvage processes in protozoan parasites for cell growth inhibition have been conducted in model organisms such as Tritrichomonas fetus (9 -12) and Giardia lamblia (13,14). The reduced expression or specific inhibition of the crucial enzyme in the purine salvage pathway indeed inhibits the growth of protozoan parasites. These findings indicate not only the feasibility of a "rational approach" to new drug discovery against particular protozoan parasites by purine salvage enzyme inhibitor searching and design, but also the likelihood of using a single enzyme inhibitor to control the growth of certain protozoan parasites due to the simplicity of their purine salvage pathways.
T. vaginalis is also a protozoan parasite with a relatively simple scheme for purine salvage (15), albeit one that is quite distinct from those in T. fetus and G. lamblia. It relies on the functions of two pivotal enzymes, purine nucleoside phosphorylase (PNP), 1 which catalyzes interconversion between purine bases and the corresponding purine nucleosides, and purine nucleoside kinase (PNK), which converts nucleosides to the nucleotides (16 -18). T. vaginalis does not incorporate hypoxanthine or inosine into purine nucleotides (19) and possesses no detectable activity of purine phosphoribosyltransferase (16,17). Thus, it apparently depends primarily on the sequential reactions catalyzed by PNP and purine nucleoside kinase to incorporate exogenous adenine and guanine into its purine nucleotide pool. Both enzymes are thus amenable to a specific inhibitor design that may lead to effective anti-T. vaginalis drugs if the parasitic enzymes are distinct from their mammalian counterparts in terms of substrate and inhibitor specificities.
T. vaginalis PNP (TvPNP) was recently cloned and expressed (20). The recombinant TvPNP shows remarkable similarity to the bacterial PNPs and differs significantly from the human PNP in terms of primary and quaternary structures as well as substrate and inhibitor specificities. Bacterial PNPs and TvPNP recognize adenine/adenosine as the most efficient substrates, but the mammalian PNPs do not recognize these substrates at all (21)(22)(23). This distinction may help to identify some adenine/ adenosine analogues as specific inhibitors of TvPNP, as in the case of bacterial PNPs (24). For example, formycin A, an adenosine analogue and a useful antibiotic, is totally inactive toward the mammalian PNP but inhibits Escherichia coli and T. vaginalis PNPs by competing with adenosine with a K i of 5.3 M and 2.3 M, respectively (20,24). Such a selective enzyme inhibition indicates the feasibility of searching for specific inhibitors among adenosine analogues against TvPNP.
Two adenosine analogues, 2-fluoro-2Ј-deoxyadenosine (F-dAdo) and 9-␤-D-arabinofuranosyl-2-fluoroadenine (F-araA), are relatively innocuous substances in mammals because they cannot be phosphorolyzed by mammalian PNP (25). But these substances are well known "subversive substrates" for E. coli PNP because, when cleaved by this enzyme, highly cytotoxic 2-fluoroadenine (F-Ade) is released (26). This interesting observation was recently developed into a strategy of suicide gene therapy against cancer cells by first expressing E. coli PNP in the cancer cells and then treating the tumor-bearing mice with these subversive substrates (27,28). Excellent anti-tumor activities were demonstrated in these cases. This newly developed anticancer strategy announced again the discovery and design of subversive substrates for one of the parasitic enzymes as an alternative means to anti-parasitic chemotherapy (29). In consideration of the presence of multiple purine salvage enzymes in some protozoan parasite species, which will be difficult to block effectively and simultaneously, and also the possibility of multiple parallel pathways that will enable the parasitic pathogens to overcome the inhibition of a single purine salvage enzyme, the design of subversive substrates for only one of the purine salvage enzymes could be a powerful strategy for anti-parasitic drug discovery.
In the case of T. vaginalis, TvPNP apparently bears a structural and a functional as well as a pharmacological resemblance to E. coli PNP, but it differs from human PNP in all these aspects (20). This activity profile classifies TvPNP as a suitable target for specific inhibitor design as well as subversive substrate research for a potential chemotherapeutic effort. There is a good possibility that F-dAdo and F-araA could also be subversive substrates of TvPNP with a potential anti-trichomoniasis effect.
In the present investigation we identified specific inhibitors as well as subversive substrates of TvPNP and tested them on the in vitro growth of T. vaginalis for potential anti-trichomoniasis chemotherapy. We found that F-dAdo is indeed a subversive substrate of TvPNP because it can be cleaved by TvPNP, thus releasing cytotoxic F-Ade. More importantly, both F-dAdo and F-Ade present outstanding inhibition effects on T. vaginalis growth, suggesting that F-dAdo could be a potential chemotherapeutic agent against T. vaginalis. We also crystallized TvPNP complexed with F-dAdo and other substrates and inhibitors. The crystal structures of these complexes were determined by x-ray diffraction. The structural studies revealed a TvPNP homohexamer bearing striking structural similarities to that of the bacterial PNPs.
TvPNP Assays and Inhibition Studies-TvPNP activity was assayed at 37°C in a 100-l reaction mixture containing 50 mM HEPES, pH 7.2, 0.8 mM ribose 1-phosphate, 2.5 mM adenine, and 1 M (0.005 Ci) [8-14 C]adenine. TvPNP (23 nM) was added to start the reaction, which was subsequently stopped by adding 1% SDS to the reaction mixture. A sample (10 l) of each reaction mixture was spotted onto a polyethyleneimine-cellulose TLC plate, developed in 1.8 M ammonium formate and 1.5% ammonium hydroxide, and exposed to general phosphor storage screens for 22 h. The radiolabeled spots on TLC plates were imaged using a PhosphorImager (Storm 840; Amersham Biosciences) and quantified with ImageQuant software (version 5.2, Molecular Dynamics). Potential inhibitors were titrated in the TvPNP assays with a reaction time of 10 min to estimate their IC 50 values.
Testing of Potential Subversive Substrates of TvPNP-Potential subversive substrates were tested at 5 mM in a 40-l reaction mixture containing 50 mM HEPES, pH 7.2, 125 mM P i , pH 7.2, and 0.8 M TvPNP for varying durations of time. The reactions were stopped by adding 3% SDS to the mixtures. Samples from the latter were developed on polyethyleneimine-cellulose TLC plates, imaged under short wave UV light, and quantified as described previously.
In Vitro Cultivation of T. vaginalis-T. vaginalis strain 3001 trophozoites were cultivated at 37°C in a modified semi-defined medium (30), with cell growth monitored by daily cell counting using a hemacytometer. Potential inhibitors were added to the culture medium at time 0 at varying concentrations for IC 50 estimation.
Crystallization of TvPNP-Purified TvPNP in 2 mM dithiothreitol, 2 mM MgCl 2 , and 50 mM HEPES, pH 7.2, was concentrated to 10 mg/ml using a 10-kDa cutoff Microcon concentrator (Amicon). The protein solution was then subjected to a series of sparse matrix screens (Hampton Research, Emerald Biostructures) to determine initial crystallization conditions. Crystallization was performed using the hanging drop method at 18°C with drops containing 1.5 l of the protein solution and 1.5 l of the reservoir solution. Crystals were observed in a variety of morphologies, mostly from conditions containing polyethylene glycol with a broad range of molecular masses (400 -8,000 Da). Trigonal prismatic crystals belonging to space group P3 usually appear within a couple of days. However, these crystals tended to be highly twinned. The most suitable crystals grew as small cubes and took 2-3 months to appear. The crystallization conditions were 50% polyethylene glycol 400, 12% ethylene glycol, 200 mM MgCl 2 , and 100 mM Tris-HCl, pH 8.5-8.7. The crystals belong to space group P4 1 32 (a ϭ 136.8 Å), with one monomer per asymmetric unit and a Matthews coefficient of 4.3 Å 3 /Da, which corresponds to a solvent content of 70%.
Preparation of TvPNP-ligand Complexes-To obtain the structure of TvPNP complexed with ligands, native crystals were soaked for 1-2 h with a series of nucleosides at their saturated concentrations. A variety of nucleosides, including F-dAdo, 2-fluoroadenosine (F-Ado), Ado, Ino, and dIno, was used in the soaking experiments (Fig. 1). The stabilization solution used for soaking experiments is similar to the mother liquor, but with a 2-5% higher concentration of polyethylene glycol 400. The crystals were directly flash-frozen by plunging them into liquid nitrogen after the soaking experiments.
X-ray Intensity Measurement-The x-ray intensity data were collected at beam line 8BM at the Advanced Photon Source using a Quantum 315 detector and at beam line A1 at the Cornell High Energy Synchrotron Source using a Quantum 210 detector. Exposure times ranged from 20 to 40 s per frame with an oscillation range of 0.5°. A total of 18 -30°of data was collected for each crystal depending on the crystal orientation. The HKL 2000 suite of programs (31) was used for integration and scaling. The data processing statistics are summarized in Table I.
Structure Determination and Refinement-The structure of TvPNP was determined by molecular replacement using the CNS software package (32). A monomer of the E. coli PNP structure (Protein Data Bank code 1PK9) was used as the search model. The starting R-factors after molecular replacement and rigid body refinement ranged from 42 to 47%. Most of the protein backbone required only minor changes. The side chains were manually adjusted using the interactive graphics program O (33). The structures were refined by alternating cycles of simulated annealing and B-factor refinement using CNS and manually rebuilding using O. The structure of TvPNP complexed with F-dAdo was refined first at 2.4 Å resolution. The loop region from residues 211 to 215 was disordered in all of the ligand complex structures and, therefore, was not built. This region was built in the unliganded structure at a lower level (0.8) and showed high B-factors. The F-dAdo complex model was used as the starting point for all of the other data sets. Strong electron density for the ligand was observed in the substrate-binding site of each ligand-complexed structure. The ligand model was built directly into the corresponding electron density. Water molecules were included after refinement of the protein plus ligand model. The final electron density maps were clear and reasonable for the stated resolution (Fig. 2). All seven TvPNP structures have relative high B-factors, which is probably related to crystal quality. The refinement statistics are summarized in Table II.  F-dAdo Is a Subversive Substrate of TvPNP-Because F-dAdo is a known subversive substrate of E. coli PNP and cytotoxic F-Ade is the product from the enzyme-catalyzed reaction, the finding that both F-dAdo and F-Ade were inhibitors of TvPNP suggests that F-dAdo could be also a subversive substrate of TvPNP. A test of F-dAdo in the enzyme assay indicated that F-dAdo was indeed most likely phosphorolyzed by TvPNP to produce F-Ade (Fig. 3). F-araA was also tested in the enzyme assay but showed little conversion to the expected product, F-Ade (data not shown).

F-dAdo and F-Ade Are Equally Potent Inhibitors of in Vitro T. vaginalis Growth-
The biochemical studies encouraged us to test the potential anti-T. vaginalis activity of F-dAdo and F-Ade. The effects of these two compounds on the in vitro growth of T. vaginalis were monitored. The results, shown in Fig. 4, indicate that both compounds are similarly potent inhibitors of T. vaginalis growth with estimated IC 50 values of 106 nM for F-dAdo and 84 nM for F-Ade. These similar toxic effects on T. vaginalis from the innocuous compound F-dAdo and the well known cytotoxic agent F-Ade suggest that the toxicity of F-dAdo is probably derived from its conversion to F-Ade by TvPNP. A comparison of the IC 50 values against the enzyme with those against the cell growth demonstrated a 4,000-fold enhanced potency of F-dAdo and an 8,600-fold increased activity of F-dAdo against cell growth.
The remarkably enhanced toxicities of F-dAdo against T. vaginalis can be attributed either to direct inhibition of TvPNP by the nucleoside analogue or to the cytotoxicity of the cleavage product F-Ade. We favor the latter for several reasons. The in vitro activity of F-dAdo against T. vaginalis compares favorably with the 5.8 M IC 50 value of metronidazole against the in vitro growth of T. vaginalis (6). Furthermore, the demonstrated efficacy of F-dAdo in treating the tumor-bearing mice having cancer cells expressing E. coli PNP also indicates the bioavailability of this compound in mammals (27,28). All of the evidence thus supports the conclusion that F-dAdo is an attractive lead compound for a new approach to anti-trichomoniasis chemotherapy.
Crystal Structure of TvPNP-To gain more in-depth understanding of the interactions between TvPNP and F-dAdo, the crystal structures of the TvPNP⅐F-dAdo complex and five other TvPNP-ligand complexes were determined by x-ray crystallography. These studies showed that TvPNP is a homohexamer with the approximate dimensions of 60 ϫ 100 ϫ 100 Å (Fig. 5). The overall structure is similar to that of E. coli PNP (38). The hexamer has 32-point group symmetry where monomers alternate in an up/down fashion, forming six active sites. Each of the six active sites utilizes residues from a pair of 2-fold related monomers, between which ϳ3,000 Å 2 of surface area is buried in the interface. The structures of ligand-complexed TvPNP and the unliganded structure showed no global differences in ternary structure and subunit orientation.
Each monomer of TvPNP is an ␣␤␣ three-layer sandwich consisting of a ␤-sheet core flanked by seven ␣-helices. The central ␤-sheet core can be divided into an eight-stranded mixed ␤-sheet and a five-stranded mixed ␤-sheet, which pack together to form a distorted ␤-barrel. The ␤-barrel is flanked by four ␣-helices on one side and three ␣-helices on the other side.
Active Site of TvPNP-The six identical active sites of TvPNP are located at the interface of each pair of 2-fold related monomers. All the ligand complexed structures showed clear density for the nucleosides at the active sites, whereas the phosphate-binding sites were occupied by water molecules. All of the nucleosides bind to the active site in the high synglycosidic bond conformation, and the sugar pucker is generally C-4Ј-endo. The active site for the nucleoside binding can be broken down to two regions, the purine-binding site and the sugar-binding site (Figs. 6 and 7). The ligands of the six present complexes bind to the active site in approximately the same orientation, with minor variations caused by the modifications on the purine base (Fig. 8).
The purine-binding site consists of four hydrophobic residues, Phe-159, Val-178, Met-180, and Ile-206. The purine base is stacked between Phe-159 and Val-178. Phe-159 lies between the purine base and the hydrophobic face of the sugar, and the  side chain of Phe-159 makes a ϳ60°angle with the plane of the purine base. Met-180 is also located between the purine base and the hydrophobic face of the sugar. In addition to the hydrophobic residues, a Thr-156 is located in the purine-binding site and causes slightly different binding of the purine base in the complexes with C-2 substituted nucleosides. In structures lacking a C-2 substituent, the C␥-atom of Thr-156 makes van der Waals contact with the C-2 atom of the purine base with a distance of ϳ4.1 Å. In the F-dAdo and F-Ado complexes, the purine base is pushed ϳ0.5 Å away from the Thr-156 C␥ atom because of steric hindrance. In those two structures, the distances between the Thr-156 C␥ atom and the 2-fluoro atom and between the Thr-156 C␥ atom and the C-2 atom are ϳ3.3 and ϳ4.6 Å, respectively. In the F-dAdo complex, the purine N-7 atom is ϳ3.6 Å from the hydroxyl group of Ser-203. This close contact is not observed in the other structures. In the complexes with ligands lacking a C-2 substitute, the N-1 atom accepts a hydrogen bond from a structurally conserved water molecule (Figs. 6B and 7B), which also makes hydrogen bonds to the carbonyl group of Phe-159 and a series of water molecules. This structurally conserved water molecule is observed in all TvPNP structures. However, in the case of F-dAdo⅐F-Ado complexes the N-1 atom is too far from this water molecule to form a hydrogen bond. Asp-204, the catalytically important residue that is proposed to stabilize the transition state by donating a hydrogen bond to the purine N-7 atom, is located in the vicinity of the purine base; however, the Asp-204 side chain points away from the purine base in the unliganded TvPNP structure and in all of the complexes, except the FMA complex. In the FMA complex, the side chain of Asp-204 points toward the purine base and makes hydrogen bonds to the purine N-6 and N-7 atoms with bond distances of 3.2 and 3.4 Å, respectively.
The sugar-binding site is primarily composed of two hydrophilic residues, Glu-181 and His-4* (residues from an adjacent monomer are designated by an asterisk throughout). The Glu-181 side chain accepts hydrogen bonds from the 2Ј-hydroxyl and 3Ј-hydroxyl groups of the ribose. His-4* makes a hydrogen bond to the 5Ј-hydroxyl group. In addition to His-4*, a water molecule was identified in the active site making a hydrogen bond with the 5Ј-hydroxyl group.
Although phosphate is not bound in any of the TvPNP structures, the phosphate-binding site can be identified according to that of the E. coli PNP structure (38,39). Three hydrophilic or charged residues, Arg-87, Thr-90, and Arg-43*, are located at the active site where the phosphate is proposed to bind. In addition, Arg-24 is also in the vicinity but with its side chain pointing away from the active site. However, there is no steric hindrance to prevent the Arg-24 side chain from swinging closer to the phosphate-binding site. In the TvPNP structures, the phosphate-binding site is occupied by three water molecules forming a hydrogen bond network with each other and to Gly-20, Arg-87, Thr-90, and Arg-43*. These three water molecules appear to take the structural equivalent positions of the phosphate oxygen atoms in the E. coli PNP structure.
Comparison of TvPNP and Structural Homologs-TvPNP has high sequence homology with bacterial high molecular mass (hexameric) PNPs, including E. coli PNP with sequence identity of 59% (20). The high sequence identity suggested structural homology between the two PNPs, which is now confirmed by the structures reported here. In addition to E. coli PNP (38,40) and TvPNP, the structures of hexameric PNPs from Sulfolobus solfataricus (41), Plasmodium falciparum (42), Vibrio cholerae (Protein Data Bank code 1VHJ), Bacillus anthracis (Protein Data Bank code 1XE3), and Thermus thermophilus (Protein Data Bank code 1ODI) have recently been determined. These PNPs share the same monomer topology and hexameric subunit arrangement, as well as highly conserved substrate-binding site residues. This common subunit topology of hexameric PNPs, consisting of a central ␤-barrel formed by an eight-stranded mixed ␤-sheet and a small fivestranded ␤-sheet as well as seven flanking ␣-helices, is also the signature topology for all members of the nucleoside phosphorylase family I (43).
In addition to the hexameric PNPs, the nucleoside phosphorylase family I also includes mammalian PNPs (44 -46), 5Јdeoxy-5Ј-methylthioadenosine phosphorylases (41,47), uridine phosphorylases (48), 5Ј-methylthioadenosine/S-adenosylhomocysteine nucleosidase (49), and AMP nucleosidases (50). Members of this family all catalyze glycosidic bond cleavage reactions and share a similar substrate binding geometry. However, TvPNP has a much lower sequence identity with the mammalian PNPs and the nucleosidases despite the structural similarities and conserved substrate binding geometry. A structural comparison of TvPNP with other nucleoside phosphorylase family I members was performed using DALI (51) and is summarized in Table III. These results show that the structural homologs of TvPNP are the hexameric bacterial PNPs, whereas the most distant are the mammalian PNPs. This observation provides further encouragement for the development of subversive substrates that are specific for TvPNP.
Comparison of TvPNP and E. coli PNP Active Sites-The residues in the substrate-binding site are highly conserved among bacterial PNPs (20). For the purpose of comparison, we have chosen E. coli PNP, which is the most extensively studied among hexameric PNPs and has the highest sequence and structural similarity compared with TvPNP according to the results from BLAST and DALI searches. TvPNP and E. coli PNP share a conserved geometry of the substrate-binding site. Most of the substrate-binding site residues are also conserved between the two enzymes, except for Thr-90 and Thr-156, which are Ser-90 and Ala-156, respectively, in E. coli PNP. Thr-90 could play the same role as Ser-90 in E. coli PNP by making a hydrogen bond to the phosphate oxygen with its hydroxyl group. However, substituting threonine for alanine at position 156 causes both F-dAdo and F-Ado to tilt as compared with their positions in the structure of E. coli PNP (Protein Data Bank code 1PKE) (39). Because of the steric hindrance coming from the C␥ atom of Thr-156, the purine base of either F-dAdo or F-Ado is pushed ϳ0.5 Å away as compared with the binding scheme of F-dAdo in E. coli PNP. In the structures of E. coli PNP, 2-fluoro nucleosides bind in the same orientation as the nucleosides without modification at the C-2 position of the base. Consequently, the binding schemes of FMA, Ado, and dIno in the structures of the TvPNP complexes is very similar to the nucleoside binding in the structures of E. coli PNP.
Role of Asp-204 -The purine-binding site of TvPNP is mostly hydrophobic with the exception of Asp-204, which is proposed to be catalytically important. Based on studies on human PNP (52), the reaction is believed to proceed through an S n 1-type mechanism in which the glycosidic bond is cleaved, forming first an oxocarbenium-like intermediate followed by a nucleophilic attack at the anomeric carbon (52,53). During the transition state of the glycosidic bond cleavage reaction, the N-glycosidic bond elongates, and a negative charge accumulates at the N-7 atom of the purine base as a positive charge accumulates at the O-4Ј atom. The proposed catalytic mechanism suggests that Asp-204 is protonated and makes hydrogen bonds to the N-7 atom to offset the partial negative charge when the reactions occur. In addition, a hydrogen bond between the carboxyl group of Asp-204 and the N-6 atom of the base is also present in the structures of E. coli PNP. However, the hydrogen-bonding interactions between Asp-204 and the purine base were not observed in the structures of TvPNP, except in the FMA complex. The lack of these hydrogen-bonding interactions is probably an artifact resulting from the basic conditions (pH 8.5-8.7) under which TvPNP crystals were grown. Therefore, Asp-204 is mostly unprotonated and cannot donate a hydrogen bond to the N-7 atom. In the case of the FMA complex, the N-7 atom is always protonated and can act as the hydrogen bond donor, which is consistent with the observations from the structure.
Structural Basis of Function-The six complexes presented here generally show the same binding geometry for the nucle- Subversive Substrate of T. vaginalis PNP oside analogues; however, minor variations that affect some contacts made by the purine base are observed. Most significantly, substitutions at the C-2 position affect the binding of the purine base in the complexes of F-dAdo and F-Ado and shift the base by ϳ0.5 Å (Fig. 6 and 8). The environment for the 2-fluoro atom is hydrophobic, consisting of Thr-156, Phe-159, Val-178, and Met-180. With the 2-fluoronucleosides, the fluorine atom packs against the Thr-156 C␥ atom and the Met-180 C␥ and S␦ atoms. These van der Waals interactions, provided by the fluorine atom, are more favorable than those of the hydrogen atom of Ado/Ino. However, the shifted position of the purine base probably causes unfavorable interactions with Asp-204, which is proposed to donate a hydrogen bond to the N-7 atom at the optimum pH range. This may explain the inhibitory effects of these 2-fluoro analogues.
For the compounds without a 2-fluoro substituent, the overall binding geometry is nearly identical for both 6-oxo and 6-amino bases (Fig. 8), except that for FMA additional hydrogen bond interactions are observed between the base and Asp-204. Previous studies (20) showed that the enzyme has similar K m values for Ado (6.1 Ϯ 0.5 M) and Ino (31.5 Ϯ 4.3 M). Also, the enzyme cleaves Ado (k cat /K m ϭ 0.278 M Ϫ1 s Ϫ1 ) slightly more efficiently than it does Ino (k cat /K m ϭ 0.152 M Ϫ1 s Ϫ1 ). The similar binding scheme for the 6-oxo and 6-amino bases observed in the structures is consistent with the biochemical results.
Conclusion-We have, in the present investigation, identified a subversive substrate F-dAdo for TvPNP that has promising potential as a lead compound for anti-trichomoniasis chemotherapy. We have also determined the crystal structure of TvPNP and identified it as a homohexamer similar to that of bacterial PNPs but distinctly different from trimeric human PNP. The active site in TvPNP differs from that in E. coli PNP by the presence of a Thr-156 residue, which affects the binding of purine nucleosides with C-2 substituents, such as F-dAdo. This difference may help explain why the subversive substrate of E. coli PNP, F-araA, is not a substrate for TvPNP. It may also suggest an opportunity to search for more suitable subversive substrates for TvPNP among purine nucleosides without a C-2 substitution.