Plasmodium falciparum Purine Nucleoside Phosphorylase Is Critical for Viability of Malaria Parasites*

Human malaria infections resulting from Plasmodium falciparum have become increasingly difficult to treat due to the emergence of drug-resistant parasites. The P. falciparum purine salvage enzyme purine nucleoside phosphorylase (PfPNP) is a potential drug target. Previous studies, in which PfPNP was targeted by transition state analogue inhibitors, found that those inhibiting human PNP and PfPNPs killed P. falciparum in vitro. However, many drugs have off-target interactions, and genetic evidence is required to demonstrate single target action for this class of potential drugs. We used targeted gene disruption in P. falciparum strain 3D7 to ablate PNP expression, yielding transgenic 3D7 parasites (Δpfpnp). Lysates of the Δpfpnp parasites showed no PNP activity, but activity of another purine salvage enzyme, adenosine deaminase (PfADA), was normal. When compared with wild-type 3D7, the Δpfpnp parasites showed a greater requirement for exogenous purines and a severe growth defect at physiological concentrations of hypoxanthine. Drug assays using immucillins, specific transition state inhibitors of PNP, were performed on wild-type and Δpfpnp parasites. The Δpfpnp parasites were more sensitive to PNP inhibitors that bound hPNP tighter and less sensitive to MT-ImmH, an inhibitor with 100-fold preference for PfPNP over hPNP. The results demonstrate the importance of purine salvage in P. falciparum and validate PfPNP as the target of immucillins.

Each year, Plasmodium species infect 300 to 500 million people and cause nearly two million deaths, mostly in children under the age of five in sub-Saharan Africa (1). Most deaths are due to infection with Plasmodium falciparum. Only AIDS and tuberculosis are more lethal human infectious diseases. No vac-cine is available, despite intensive international efforts. At present, control of malaria is dependent on prevention with bed nets, insecticides or chemoprophylaxis, and chemotherapeutic treatment of clinical cases. However, chemotherapy for malaria has been complicated by recent increases in mortality and morbidity due to the emergence of drug-resistant strains (2).
Because parasitic protozoa are unable to synthesize purines de novo, purine salvage has been proposed as a potential target for chemotherapy of protozoan parasite infections, including those caused by Plasmodium spp. Malaria parasites are obligate intracellular parasites that carry out their asexual cycle in erythrocytes. Unlike most mammalian cells, erythrocytes have no biochemical machinery for de novo purine synthesis, but act as a rich source of purine salvage enzymes, particularly purine nucleoside phosphorylase (PNP) 4 and adenosine deaminase (ADA). The purine salvage pathway of Plasmodium begins with the deamination of adenosine to inosine by ADA, followed by conversion of inosine to hypoxanthine by PNP. The final enzyme in the pathway is hypoxanthine-guanine-xanthine phosphoribosyltransferase (HGXPRT). Hypoxanthine is a precursor for all purines and is a central metabolite for nucleic acid synthesis in P. falciparum. Interestingly, P. falciparum is able to survive in both PNP-and ADA-deficient erythrocytes, suggesting that Plasmodium enzymes, PfADA and PfPNP, are sufficient for survival of parasites within the erythrocyte (3,4).
Prior studies have shown that PfPNP and PfADA have an additional specificity for 5Ј-methylthiopurines, and that salvage of 5Ј-methylthioadenosine (MTA), a dead-end molecule of polyamine synthesis, is through the malarial purine salvage enzymes (5). Although humans and other Apicomplexa, such as Toxoplasma and Eimeria, have a redundant pathway for purine salvage via adenosine kinase (6), PfPNP, PfADA, and PfHGX-PRT appear to be the only pathway by which purines are salvaged in Plasmodium. Because P. falciparum expresses fewer enzymes in the pathways for purine nucleoside and MTA salvage than its host, these pathways are attractive targets for antibiotic design. Disruption of PfPNP expression or activity may not only affect purine salvage, but could potentially perturb homeostasis of the polyamine pathway.
Immucillins are powerful transition state inhibitors of PNP that kill P. falciparum in vitro by inducing purine-less death (7,8). Most immucillins tested in malaria cultures, such as ImmH, bind with a higher specificity to human PNP than PfPNP, and it could not be determined whether inhibition of PfPNP alone was capable of causing purine starvation in the parasite (7,8). The discovery of a novel purine-recycling pathway in malaria, with additional specificity for 5Ј-methylthiopurines, led to the development of a second generation of immucillins with 5Ј-methylthio modifications. MT-ImmH shows a 100-fold preference for PfPNP and kills P. falciparum in vitro with a similar IC 50 to ImmH (5). The additional specificity of PfPNP for 5Ј-methylthiopurines permits specific targeting of the malarial purine salvage pathway and perhaps the polyamine pathway by inhibiting a single enzyme.
Many drugs have off-target interactions, and genetic evidence is required to demonstrate the importance of the proposed target. To further explore the importance of PfPNP we used single crossover homologous recombination (9) to genetically disrupt pfpnp in P. falciparum. Immucillin drug sensitivity profiles of both wild-type (WT) parasites and transgenic parasites lacking PfPNP (⌬pfpnp) were altered, confirming the importance of inhibition of PfPNP for immucillin efficacy. The ⌬pfpnp lines have a greater requirement for exogenous purines and are unable to thrive at physiological concentrations of hypoxanthine. These results illustrate the importance of purine salvage enzymes for P. falciparum viability.

EXPERIMENTAL PROCEDURES
P. falciparum Culture-Human erythrocytes were collected from local volunteers under protocol CCI 99 -240 of the Albert Einstein College of Medicine. P. falciparum, strain 3D7, was grown in RPMI supplemented with 0.5% (w/v) Albumax II (Invitrogen Corporation) and 50 mg/liter of hypoxanthine (ϳ370 M) at 4% hematocrit (10).
Vector Construction and Parasite Transfection-The pfpnp allelic exchange fragment was PCR-amplified from 3D7 genomic DNA, using the primer combination p10/p11 (see supplementary materials). This yielded a 0.5-kb fragment that was cloned into the plasmid, pBSDmini, which contains the bsd gene that encodes resistance to blasticidin, flanked by promoter and terminator elements from cam (calmodulin) and hrpII (histidine-rich protein II), respectively (11,12). The resulting 5.1-kb transfection plasmid, pBSDmini/pfpnp, was electroporated into the 3D7 P. falciparum ring stage parasites according to established protocols (9,13,14). In short, plasmid-transfected parasites were selected by the addition of 2.5 nM blasticidin (InvivoGen, San Diego, CA) to the culture medium, starting 48 h post-transfection. Parasite clones were obtained by two rounds of limiting dilution using 500 M hypoxanthine in the culture media, and identified using the MALSTAT assay reagent specific for P. falciparum lactate dehydrogenase (15,16).
Nucleic Acid Analysis-Genomic DNA of P. falciparum was purified as previously described (17). Plasmid integration into the pnp locus was detected by PCR, using the primer combinations p1/p4 and p3/p6 (see supplemental materials and Fig. 1, A  and B). Integration into the cam and hrpII loci was assessed using primer combinations p7/p4 and p8/p9, respectively (see supplemental materials and data not shown). Amplification of non-recombinant, endogenous pfpnp sequence was assayed using primers p1/p2 (see supplemental material and Fig. 1, A and B). PCR conditions included a primer extension temperature of 62°C to account for the high A-T content in P. falciparum genomic DNA (18).
For Southern blot analysis, genomic DNA was digested with EcoRI or HindIII. It was electrophoresed overnight at 2.5 V/cm in 0.8% (w/v) agarose gels (1.5 mg of DNA/lane), alongside the digoxigenin-labeled DNA molecular mass ladder (Roche Applied Science) and 2 pg of transfection plasmid digested with the appropriate enzyme. The DNA was then transferred to Nylon membranes, and cross-linked by UV irradiation. An internal 0.5-kb region of pfpnp and a 0.3-kb region of bsd were PCR-amplified using primers p12/p13 and p9/p10, respectively (see supplementary materials), and labeled with digoxigenin (Roche Applied Science). Hybridizations were performed at 60°C, and membranes were washed at a maximum stringency of 0.3% SSC, 0.1% SDS (w/v) at 60°C, prior to CDP-Star detection, and autoradiography.
For stage-specific reverse transcriptase-PCR assays, cultures were doubly synchronized by sorbitol treatment. Total RNA was prepared using TRIzol (Invitrogen). Cultures with Ͼ95% trophozoites were utilized for RNA extraction. First strand cDNA was synthesized from 5 g of total RNA, using the power shot pre-amplification system (BD Biosciences) with random hexamer priming.
Western Blots and PfPNP and PfADA Antibody Production-The manufacturer's protocols were followed to raise mouse polyclonal antibodies, using the adjuvant TiterMax Gold (TiterMax, Norcross, Georgia). Briefly, BALB/c mice were injected with 100 g of purified recombinant His 6 -tagged PfPNP or PfADA emulsified in adjuvant. Mouse serum was collected 4, 6, and 8 weeks post-immunization. Western blots were performed on 0.1% (w/v) saponin-lysed parasite pellets that were washed with phosphate-buffered saline and resuspended in SDS-PAGE loading buffer. Blots were probed with 1:1000 dilutions of each primary antibody and detected using Renaissance Western blot Chemiluminescence Reagents (PerkinElmer Life Science).
Drug Assays-In vitro drug responses were calculated using 72-h [ 3 H]ethanolamine incorporation assays (19,20). Levels of incorporation were measured using a 1450 Microbeta Liquid Scintillation and Luminescence Counter (PerkinElmer Life Science). MT-ImmH, ImmH, 2Ј-deoxy ImmH, and 2Ј-deoxy ImmG were synthesized as previously described (7,21,22). Culture media for studies with various immucillins contained no hypoxanthine supplement. Immucillins were dissolved in water to make stock solutions, and diluted with media prior to addition to cultures. Following incubation with inhibitor for 18 h, 200-l cultures in 96-well plates were supplemented with 1 Ci of [ 3 H]ethanolamine (Amersham Biosciences; 25 Ci/mmol). After 54 h, cultures were frozen overnight and thawed to lyse the cells. The lysates were harvested on glass-fiber filters and washed with 1.2 ml of H 2 O. Filters were dried and counted in a Winspectral 1414 Scintillation Counter. Experiments were carried out multiple times with six replicate wells for each drug concentration. Drug inhibition concentrations were calculated by regression analysis of dose-response curves. S.D. Ϯ mean was calculated for each IC 50 value using KaleidaGraph version 4.03 (Synergy Software, Reading, PA). Individual data points more than two mean Ϯ S.D. were discarded. The IC 50 values Ϯ the mean Ϯ S.E. were calculated from the IC 50 values from 3 to 12 independent experiments, and for each drug tested a oneway analysis of variance analysis in combination with a Bonferroni post hoc test was performed to determine the p values of the various group pairs using KaleidaGraph version 4.03 (Synergy Software, Reading, PA). Drug dilution series using the immucillins were carried out to expand the data set between concentrations of 1 ϫ 10 Ϫ8 and 1 ϫ 10 Ϫ6 M. These data are included in the determination of the IC 50 values. For some experiments, parasitemias were counted on Giemsa-stained smears of cultures treated in parallel.
Enzymatic Assays-Parasite lysates were obtained from parasites harvested from 60 ml of culture. The cell pellet was lysed using 0.1% (w/v) saponin and washed twice with STE, followed by two washes with 50 mM potassium phosphate pH 7.4. The pellets were resuspended in 300 l of 50 mM potassium phosphate pH 7.4, then frozen and thawed five times. Insoluble material was removed by centrifugation and the protein concentration of cell lysates determined using the Bradford method. 50 g of protein (30 -50 l lysate) was used to assay for enzyme activity. Assays were performed in triplicate. Hypoxanthine produced by PfPNP was assayed by coupling to the xanthine oxidase reaction and measuring uric acid formation at 293 nm over time (7). PfADA action on adenosine or MTA was measured by the absorbance decrease at 265 nm. Reactions were initiated by the addition of substrate (5). The activity and S.E. of the parameters was determined using the Levenberg-Marquardt algorithm in KaleidaGraph version 4.03 (Synergy Software, Reading, PA).
Purine Requirements-Sorbitol-synchronized parasites were used, and each experiment was seeded with 1% parasitemia, with Ͼ95% rings and at 1% hemocrit. Assays were carried out in 200-l volumes in 96-well plates. Stock solutions of 2 mM purines (hypoxanthine, adenosine, and inosine) were used for subsequent serial dilutions. Before beginning the experiment, the parasites were washed three times in purine-free culture media, resuspended in purine-free media, and added to the plates, which already contained a 2ϫ concentration of each purine in an equal volume of media (i.e. 1ϫ final concentration). Parasites were allowed to grow for 48 h with the corresponding purine at each concentration. After 48 h, 100 l of media was removed from each well, and 100 l of media with the appropriate purine concentration and 1 Ci of [ 3 H]ethanolamine (GE Healthcare) was added. After 24 h of incubation with the radiolabel, cell cultures were frozen overnight and thawed to disrupt cells. The mixtures were then harvested on glass-fiber filters and washed with 1.2 ml of H 2 O. Filters were dried and counted in a Winspectral 1414 Scintillation Counter. The experiment was performed with six replicate wells for each purine concentration. Mean Ϯ S.D. was determined, and twotailed unpaired t tests were carried out to determine the statistical significance for each disrupted line as compared with the wild-type control line using KaleidaGraph version 4.03 (Synergy Software). These results were also confirmed using the lactate dehydrogenase assay, as described in Ref. 23.
Doubling Time-Highly synchronized parasites were used, and each experiment was seeded with 1% parasitemia, with Ͼ95% rings and at 4% hemocrit. All parasites were washed three times in purine-free medium and resuspended in 5, 10, or 100 M hypoxanthine. Giemsa-stained thin smears were taken every 24 h. The medium was changed every 24 h while maintaining the appropriate hypoxanthine concentration. Cultures were cut back to 1% when they reached 3-6% parasitemia (usually every 48 h). Smear identities were blinded and parasitemias counted in triplicate. Parasite -fold increases were calculated by correcting the parasitemias for the -fold dilution. Data were plotted using KaleidaGraph version 4.03, and curve fit analysis was carried out using the equation y ϭ N o e (kx) (where N o ϭ initial population, and k ϭ growth constant). Growth constants (k) were used to determine doubling times.

RESULTS
Production of PNP "Knock-out" Clones-A transfection plasmid, pBSDmini/pfpnp, was designed to disrupt full-length pfpnp in the chromosomal locus (9,11). A truncated portion of the pnp open reading frame was cloned into pBSDmini (12). Integration of pBSDmini/pfpnp into the parasite genome via homologous recombination and single site crossover was predicted to generate a pnp locus that was disrupted by the insertion of the plasmid, producing two non-functional fragments of pnp (Fig. 1A). The pBSDmini/pfpnp plasmid lacks the sequences for 10 amino acids at the N terminus and 53 amino acids at the C terminus. The 5Ј-non-functional fragment would lack Asp 206 , and the 3Ј-non-functional fragment would lack His 7 , both of which are critical for catalytic activity of PfPNP (21).
Electroporation of pBSDmini/pfpnp into 3D7 parasites generated episomally transformed lines that were maintained by selection with blasticidin (InvivoGen). Passaged lines were screened by PCR for homologous recombination and plasmid integration into the pnp, hrpII, or cam loci (data not shown). Bulk cultures from two independent transfection events tested positive for integration into the pfpnp locus at day 20 postelectroporation, and clones were immediately generated from this line by two rounds of limiting dilution (16). Initially, periodic cycling on and off blasticidin selection was performed before cloning; however, pnp disruptants were lost from cultures, suggestive of a growth disadvantage for pnp disruptants. Immediate cloning and maintenance of the parasites selected with blasticidin at high concentrations of hypoxanthine allowed us to successfully identify three pnp integrants from two independent transfections, as well as several hrpII integrants. The hrpII integrants were used as wild-type controls, because they had integrated the transfection plasmid but did not have a defect in PfPNP and could be maintained under similar blasticidin drug selection as the ⌬pfpnp clones. Prior studies have shown that integration into the hrpII locus is not deleterious (15,16).
The ⌬pfpnp clones were confirmed by PCR analysis to be recombinant solely for pnp integration, and were PCR-negative for wild-type pnp (Fig. 1B). The ⌬pfpnp clones were PCR positive for intact plasmid, which would be expected if multiple copies of the plasmid were integrated into the locus. It is not uncommon for plasmids to form concatemers before integrating into wild-type loci (24). DNA isolated from the cloned parasite lines was transformed into Escherichia coli, and no colonies were observed, consistent with only integrated copies of the transfection plasmid present in the clones (data not shown).
Southern blot hybridization confirmed that the endogenous pnp locus had been disrupted (Fig. 1C). Using a probe for pnp, hybridizing bands of ϳ26, 5, and 7 kb, respectively, were detected upon EcoRI and HindIII digestion of genomic DNA from the ⌬pfpnp clones. These bands contrasted with the 2.3and 8.5-kb bands generated by EcoRI and HindIII digests of untransfected 3D7 genomic DNA. The EcoRI restriction pattern was consistent with integration of multiple plasmids into the pnp locus. Integration of a single copy of the plasmid would generate a 7.4 kb band upon EcoRI digestion; however, a band at ϳ26 kb is indicative of at least four concatameric copies of the plasmid. HindIII cuts the plasmid once, linearizing it. The 7-kb band in the HindIII digests is in accordance with the expected plasmid fragments connected to the 5Ј-and 3Ј-truncated gene and promoter regions in the locus. Integration was also con- firmed using a probe for bsd, which encodes the blasticidin resistance cassette (data not shown).
Expression Levels in Knock-out Mutants-To verify disruption of the pfpnp locus abrogated pfpnp expression, we performed reverse transcriptase-PCR analyses on synchronized trophozoite-stage RNA preparations from the ⌬pfpnp clones and the 3D7 wild-type parent line. All ⌬pfpnp clones lacked pnp transcripts, but expressed normal levels of ada transcripts ( Fig.  2A). To compare PfPNP protein expression levels between parental and recombinant lines, Western blots were performed on lysates of synchronized parasite cultures, using anti-PfPNP and anti-PfADA antibodies (Fig. 2B). PfPNP protein was absent in the ⌬pfpnp clones, whereas a band of ϳ26 kDa was observed in the WT line. With longer exposures a cross-reacting band of ϳ30 kDa was visible in all lines (as seen in ⌬pfpnp 1 in Fig. 2B). All lines possessed equivalent PfADA protein levels indicated by the band migrating at ϳ42 kDa. To confirm the absence of enzymatically active PfPNP, protein lysates of wild-type (3D7 and hrpII integrants) and ⌬pfpnp derived from saponin-lysed infected erythrocytes were tested for PNP and ADA activity. All three ⌬pfpnp lines had ADA activity, but only the wild-type and hrpII integrant lines had detectable PNP activity (Table 1).
Purine Requirements of the ⌬pfpnp Lines-The growth requirement of the ⌬pfpnp lines for exogenous purines was analyzed by measuring [ 3 H]ethanolamine incorporation over time (19). Ethanolamine incorporation has been shown to be proportional to parasite number, and its incorporation should not be affected by perturbation of purine metabolism. Hypoxanthine is the major purine in human serum with concentrations reported in the low micromolar range (0.4 -6 M) (25,26). The ⌬pfpnp lines exhibited a statistically significant reduction in incorporation of [ 3 H]ethanolamine compared with the WT control at physiological concentrations of purine (12.5 M purine and below) (Fig. 3, A-C). The requirement for higher levels of exogenous purines in the ⌬pfpnp lines was also confirmed by using the lactate dehydrogenase assay (Fig. 3D), in which parasite lactate dehydrogenase activity is used as a surrogate of growth. Therefore, the ⌬pfpnp lines have a greater need for exogenous purines than the WT control. There was a slight preference for hypoxanthine over inosine, and for inosine over adenosine as a source of exogenous purine; however, the difference was minimal (supplemental Fig. S1). Under these conditions erythrocyte metabolism can convert adenosine and inosine to hypoxanthine.
Growth and Doubling Times of ⌬pfpnp Lines in Various Concentrations of Hypoxanthine-At physiological concentrations of hypoxanthine (5-10 M), the ⌬pfpnp lines displayed a growth defect relative to the hrpII integrant WT control as monitored by counting parasitemias of infected erythrocytes (supplemental Fig. S2). There was no significant difference in the -fold increase of the WT control and the ⌬pfpnp lines at supra-physiological concentrations of hypoxanthine. Doubling times were derived from the plots of parasite growth and are represented in Table 2. The WT control exhibited a small difference in doubling times at 10 and 100 M hypoxanthine: 9.1 and 7.0 h, respectively. In contrast, doubling times of the ⌬pfpnp lines at the same concentrations were ϳ24 and 7.5 h (Table 2). Thus, the WT parasites showed little impairment of growth at 10 M purine, whereas the growth of ⌬pfpnp lines was significantly compromised.
Response of ⌬pfpnp Lines to Inhibitors of PNP, Immucillins-ImmH is a powerful inhibitor of host and parasite PNP and MT-ImmH is more specific for PfPNP than human PNP (5). Both immucillins kill P. falciparum in vitro and have IC 50 values determined to be in the 50 nM range (5). Unlike host PNP, PfPNP has a preference for ribonucleosides over deoxyribonucleosides. Thus, 2Ј-deoxy ImmH and 2Ј-deoxy ImmG are potent inhibitors of hPNP, but not PfPNP with IC 50 values in the 500 nM range (7). Prior studies also suggested that inhibition of PfPNP was essential for anti-parasitic efficacy of immucillins (5,7).
The ⌬pfpnp lines are more sensitive to ImmH, a more potent inhibitor of hPNP than PfPNP (Fig. 4A), and less sensitive to MT-ImmH (Fig. 4B), a PNP inhibitor with ϳ100-fold tighter binding to PfPNP (5). Drug assays and IC 50 determinations using these immucillins showed that the ⌬pfpnp lines were ϳ11 times less sensitive to MT-ImmH than wild-type 3D7 and the hrpII integrant control, ϳ6 times more sensitive to ImmH, ϳ42 times more sensitive to 2Ј-deoxy ImmH, and ϳ88 times more sensitive to 2Ј-deoxy ImmG (Table 3). Growth curves are shown relative to total ethanolamine incorporation of

TABLE 1 Specific activities of PfPNP and PfADA in whole cell lysates of wild type and ⌬pfpnp lines
Whole cell parasite lysates were obtained from infected red blood cells that were saponin lysed before freeze-thawing. The substrates used were inosine, 5Ј-methylthioinosine (MTI), adenosine, and 5Ј-methylthioadenosine (MTA). When comparing the ADA specific activity of the WT parasite lines to the ⌬pfpnp lines the Student's t test showed no statistical difference in specific activity between the lines with both substrates (adenosine and MTA). The limit of detection for inosine was 1.1 nmol mg Ϫ1 min Ϫ1 and 0.8 nmol mg Ϫ1 min Ϫ1 for MTI, which represents ϳ20 fold increase above the background for the specific activity of PNP detected in the 3D7 and hrpII integrant lines for both substrates.  DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51 untreated controls. As illustrated in Fig. 4C (and Fig. 3), ⌬pfpnp lines grow poorly at low concentrations of hypoxanthine with less overall ethanolamine incorporation than wild-type parasites. Drug assays were performed without hypoxanthine supplementation. The ⌬pfpnp lines showed a hierarchy similar to the disassociation constant (Fig. 4D) of each inhibitor for human PNP. When treated with MT-ImmH the wild type kill curves show a more shallow drop in ethanolamine incorporation as compared with the steep drop in incorporation seen with ImmH. The ⌬pfpnp lines treated with MT-ImmH were more resistant to inhibitor suggesting that in wild-type parasites the initial inhibitory effects of MT-ImmH are due to inhibition of PfPNP with later responses due to inhibition of hPNP ( Fig. 4B). Consistent with this, as seen in Fig. 4C, the inhibition curves of wild-type and ⌬pfpnp lines overlap at higher concentrations of MT-ImmH.

DISCUSSION
The research presented herein indicates that blocking the production of hypoxanthine, via targeted genetic disruption or biochemical block at PfPNP, causes a growth defect in P. falciparum. This is likely to occur before schizogony, as expression data indicate that transcripts of purine salvage enzymes are increased during the trophozoite stage, as the parasite prepares to replicate its genome and produce merozoites (27)(28)(29). Examination of blood smears of immucillin-treated parasites shows a prolongation of the replication cycle as evidenced by accumulation of trophozoite forms and delayed appearance of rings. Thus, absence of DNA precursors prolongs the cell cycle and eventually results in purine-less death of the parasite. These data collectively and definitely demonstrate the importance of PfPNP to P. falciparum viability. This growth defect may also be affected by a perturbation in the levels of polyamines (5), however, this mechanism has yet to be established.
Disruption of essential genes in P. falciparum is not possible, due to the haploid chromosome state of the blood-stage parasite. However, we were able to achieve a conditional disruption of PfPNP by hypoxanthine supplementation of media for our ⌬pfpnp clones. In vitro P. falciparum culture media usually contains 50 mg/liter of hypoxanthine (ϳ370 M), and maintenance of the ⌬pfpnp clones at such high, non-physiological levels restored normal growth rates. However, hypoxanthine is reported to be present in human serum at much lower concentrations (ranging from 0.4 to 6 M) (25,26). Culture of ⌬pfpnp parasites at these physiological concentrations resulted in severe growth retardation and eventual death of the ⌬pfpnp parasites.
The ⌬pfpnp lines required higher levels of exogenous purines, although no specific purine (i.e. hypoxanthine, inosine, or adenosine) appears to be of particular importance. Adenosine and inosine can be converted into hypoxanthine by ADA and PNP in the human erythrocyte cytosol and then used by the parasite to circumvent the block at PNP. In addition, ATP is present at millimolar concentrations in the erythrocyte cytosol and may be a purine source for parasites. Depending upon the concentrations of ribosyl phosphates within the erythrocyte cytosol, hypoxanthine may be converted to inosine by PNP. At  lower concentrations of adenosine, erythrocyte adenosine kinase activity will predominate to generate AMP (K m 50 nM versus 70 -90 M for hADA (30,31)). A parasite line with a knock-out in the nucleoside transporter, PfNT1, has been reported and has a similar impairment of growth at physiological purine levels as ⌬pfpnp (23). For both ⌬pfpnp and ⌬pfnt1 parasite lines, growth is rescued by high concentrations of purines. The exact attributes of the PfNT1 transporter have been somewhat controversial (32)(33)(34)(35)(36), but it appears that PfNT1 (also known as PfENT1) is the major purine transporter of the parasite plasma membrane and is capable of transport of hypoxanthine, adenosine, inosine as well as other purines and pyrimidines (37).
El Bissati et al. (23,37) have postulated that hypoxanthine is the major purine transported by PfNT1 and that PfPNP is unlikely to be important for parasite viability. Barring the unlikely possibility that PfPNP is primarily active in the erythrocyte cytosol (see Table 1 where PfPNP was measured on saponin-lysed-infected erythrocytes depleted of host cytosol), it seems likely that other purines are taken up by PfNT1 and that the activity of PfPNP is of similar importance to PfNT1.
Thus it is probable that parasite uptake of both purine nucleosides and nucleobases is important for optimal viability of Plasmodium. Although hypoxanthine is the most abundant purine in serum, the low K m of PfHGXPRT (Ͻ1.0 M) (38) and the lower K m of PfPNP versus hPNP (5 versus 40 M (7)) enables the parasite to metabolically trap any available purines, including adenosine, inosine, or hypoxanthine (37). As erythrocyte and Plasmodium purine transporters are equilibrative, steady state concentrations of purines will be equivalent in serum, the erythrocyte cytosol, and parasite cytosol. Any available purines will be rapidly utilized by the parasite.
The activity of PfPNP on 5Ј-methylthioinosine may also be critical when purine levels are low. This activity allows the parasite to recycle purines from MTA derived from polyamine synthesis (5).
A unified model that integrates the phenotype of ⌬pfpnp and ⌬pfnt1 is presented in Fig. 5. In this model, the low affinity purine transporter PfNT1 is responsible for equilibrative transport of adenosine, inosine, and hypoxanthine (as well as guanine, guanosine, and xanthine). These purines are incorporated into parasite nucleotide pools via the action of PfADA, PfPNP, and PfHGXPRT. If PfNT1 is not functional, purines can be taken up by alternative transport pathways. At low purine concentrations, in ⌬pfnt1, purines are poorly transported, whereas The ⌬pfpnp lines are more sensitive to ImmH (A), a transition state inhibitor that binds tighter to hPNP, and less sensitive to MT-ImmH (B), an inhibitor that binds 100-fold more tightly to PfPNP. Data shown in panels A and B are normalized to untreated controls. IC 50 values Ϯ S.E. from multiple independent experiments were calculated and are represented in Table 3. Overall growth of ⌬pfpnp as represented by total ethanolamine incorporation is lower than wild-type controls, as can be seen when inhibition curves are represented as total incorporated ethanolamine counts (C; subtracting uninfected erythrocyte controls; see Fig. 3 also). The disassociation constants and structures for the immucillins are presented in panel D (5,7). All assays were performed without hypoxanthine supplementation of the culture media. in ⌬pfpnp, purines are transported but must be converted to hypoxanthine in the erythrocyte cytosol to be utilized by the parasite. Low levels of hypoxanthine are available from the erythrocyte cytosol purine pools so ⌬pfpnp are impaired but viable. The dramatic increase in doubling time and severe retardation of development observed in the ⌬pfpnp lines at physiological concentrations of hypoxanthine provide evidence of the vital role that purine salvage plays in the parasite life cycle. The difference in drug response of the ⌬pfpnp lines compared with wild type is also noteworthy. Uncoupling the host and parasite pathways caused a decrease in sensitivity to the parasite-specific PNP inhibitor (MT-ImmH), and an increase in sensitivity to the PNP inhibitors that target both host and parasite PNP (ImmH, 2Ј-deoxy ImmH, and 2Ј-deoxy ImmG). Together, these results demonstrate the importance of PfPNP, even when host PNP is abundant, and thus validate PNP as the specific target of the immucillins.
The in vitro culture system does not completely mimic conditions in an infected human, but these studies indicate PfPNP is likely to play a critical role in vivo in malaria-infected human hosts. In parallel studies, we have demonstrated that Plasmodium yoelii parasites lacking PyPNP are attenuated in mice (39), providing further genetic evidence for the importance of Plasmodium PNP in vivo.
The attenuated phenotype of the ⌬pfpnp parasites validates further studies to pharmacologically target purine pathways for malaria treatment. Immucillins have been extensively investi-gated in primate and human models; there is a wealth of information about their oral availability, toxicity, and efficacy as a treatment for certain types of cancer (40,41). Further studies in the human host or primate models are required to determine whether immucillins kill the parasite or attenuate the parasite sufficiently to allow the human host to successfully clear the infection. Thus our study supports further investigation of immucillins, or related transition-state inhibitors of the purine salvage pathway, as novel chemotherapeutic agents for the fight against malaria.