Malaria is caused by a protozoan parasite of the genus Plasmodium and is the most deadly vector-borne disease in the world. The World Health Organization estimates that there are 300–500 million clinical cases annually, resulting in approximately 1.5–2.7 million deaths5. Because there is no effective vaccine, and drug-resistant parasites are widespread, development of alternative chemotherapeutic targets and new vaccine strategies is crucial.

Unlike their mammalian hosts, malaria parasites cannot synthesize purines de novo and depend exclusively on purine salvage and recycling for RNA and DNA synthesis. The purine salvage pathway is streamlined in Plasmodium, composed of only adenosine deaminase (ADA), PNP and hypoxanthine-guanine-xanthine phosphoribosyltransferase1,2,3,4. Hypoxanthine is a precursor for all purines and a central metabolite for nucleic acid synthesis in P. falciparum6. The dual action of P. falciparum ADA and PNP for purines and methylthiopurines allows the parasite to form hypoxanthine from host purine pools and to recycle hypoxanthine from methylthioadenosine, a product of polyamine synthesis. P. falciparum ADA and PNP function at the intersection of the polyamine metabolism7,8,9 and purine salvage pathways, essential pathways for viability of Plasmodium in its human host.

In this study, we use genetic approaches to investigate the importance of Plasmodium PNP in vivo in the rodent malaria P. yoelii (lethal strain YM). P. yoelii is a favored model for vaccine development because of its high infectivity of laboratory mice, particularly in sporozoite infection of hepatocytes10,11,12,13,14. P. yoelii YM, like P. falciparum, infects both reticulocytes and mature erythrocytes, whereas nonlethal P. yoelii 17XNL and another rodent malaria, Plasmodium berghei, infect only reticulocytes. Because rodent and human erythrocytes have ample PNP and purine salvage activity, erythrocytes may provide sufficient purines to sustain Plasmodium parasites in infected animals.

The PNP genes from all Plasmodium species are highly conserved, with P. falciparum PNP being 78% identical at the amino acid level to P. yoelii PNP (PyPNP; Supplementary Fig. 1 and Supplementary Table 1 online). Detailed biochemical analysis showed that recombinant PyPNP has substrate specificity and kinetic properties similar to P. falciparum PNP (Table 1).

Table 1 Kinetic constants of Plasmodium PNP

The gene encoding PyPNP was disrupted by a single-crossover insertion strategy15,16 (Supplementary Fig. 2 and Supplementary Methods online). The viability of PNP-disruptant P. yoelii clones (PyPNP-INT) was compared to a recombinant parasite control with insertion into the PNP-encoding locus without disruption (PyPNP-WT). When PyPNP-INT parasite clones were passaged in mice without drug selection, wild-type revertants developed and rapidly overgrew the PyPNP-INT population (data not shown), providing initial genetic evidence that parasites lacking PNP were less fit than wild-type parasites. To prevent reversion, we adopted a PCR approach17 to generate PyPNP-knockout clones with the genes encoding the selectable markers human dihydrofolate reductase (hDHFR; ΔPyPNP clones) or hDHFR-gfp, a fusion protein of hDHFR and green fluorescence protein18PyPNP-GFP clones) replacing PyPNP by double crossover homologous recombination (Supplementary Fig. 3 and Supplementary Methods online). In parallel, we created a control strain lacking the gene encoding the circumsporozoite protein or CSP (ΔPyCSP), a gene that is not essential in erythrocytic stages19.

The virulence of PNP-deficient clones was compared to recombinant controls by infecting BALB/c mice and monitoring parasitemia for 30 days (Fig. 1). The growth rate of PyPNP-INT parasites was slower than that of the wild-type P. yoelii YM strain and of PyPNP-WT (Fig. 1a). All WT-infected and PyPNP-WT–infected mice developed high parasitemias and died within 10 d of challenge (Fig. 1a, b). Similarly, ΔPyPNP and ΔPyPNP-GFP parasites had no detectable impairment of intraerythrocytic development, but growth of these lines was impaired compared to that of WT or ΔPyCSP parasites (Fig. 1c, d). None of the parasite lines lacking PNP (PyPNP-INT, ΔPyPNP and ΔPyPNP-GFP) were lethal to mice, and infected mice recovered from infection with no detectable parasitemia (Fig. 1b, d). The PyPNP-INT, ΔPyPNP and ΔPyPNP-GFP clones were also attenuated in Swiss Webster mice (data not shown). Thus PNP is crucial for optimal replication of malaria parasites during the erythrocytic cycle, despite the considerable levels of PNP activity present in mammalian erythrocytes.

Figure 1: P.yoelii lacking PNP are attenuated.
figure 1

(ad) Parasitemia (percentage of infected erythrocytes ± s.d.; a,c) and survival curves (b,d) from groups of five female BALB/c mice (age 8–10 weeks) infected with the indicated parasite lines. Blood smears were made daily, stained with Giemsa and counted. For experiments with single crossover integrants (a,b), mice were infected with 2 × 105 infected erythrocytes from wild-type (wild-type YM), PyPNP-WT clone 1 (PNP control integrant), PyPNP-INT clone 1-3 (PNP disruptant) or PyPNP-INT clone 2-3 (PNP disruptant). For experiments with double-crossover knockout lines (c,d), mice were infected with 2 × 104 wild-type (wild-type YM), ΔPyCSP clone 2-2 (CSP knockout), ΔPyPNP (PNP knockouts made with hDHFR cassette; clones 3-4, 4-2 and 4-5), ΔPyPNP-GFP clone 8 (PNP knockout made with hDHFR-GFP cassette). All mice were infected with parasites on day 0 by tail intravenous injection. For each parasite line, parasitemia was checked daily for 30 d (at least 7 d after parasitemia was no longer detectable).

There are other points in the life cycle at which parasites have substantial purine requirements. The sexual-stage gametocytes are taken up by mosquitoes from mammalian blood. In the mosquito midgut, male gametocytes undergo exflagellation, a process that involves four rapid nuclear divisions and results in formation of eight microgametes within 15 min. Ookinetes develop from the fertilized zygote and migrate across the mosquito midgut, where they form oocysts. Thousands of sporozoites are made within each oocyst. These sporozoites invade mosquito salivary glands and are injected into the mammalian host. After sporozoite invasion of hepatocytes, the parasites develop into exoerythrocytic forms that undergo extensive multiplication, eventually releasing tens of thousands of merozoites into the circulation.

We analyzed oocyst and sporozoite development in Anopheles stephensi mosquitoes fed on mice infected with wild-type, PyPNP-WT, PyPNP-INT or ΔPyPNP parasite clones bearing similar numbers of mature gametocytes. At day 8 after feeding, PyPNP-INT–infected mosquitoes had markedly fewer numbers of oocysts compared to WT- or PyPNP-WT–infected mosquitoes, and we detected no ΔPyPNP oocysts (Table 2). On day 15 after feeding, PyPNP-INT or ΔPyPNP-infected mosquitoes had no detectable sporozoites (Table 2). The numbers of ΔPyCSP oocysts were normal, but infected mosquitoes had no sporozoites in their salivary glands, confirming that CSP is essential for development of P. yoelii sporozoites in mosquitoes, as previously described for P. berghei19. Thus PNP activity is key in blood stage development and crucial for mosquito stage development.

Table 2 Parasite development in mosquitoes

Because attenuated microorganisms are frequently used as vaccines, we tested whether PNP-knockout parasites protected against subsequent malaria infection (Table 3). Naive mice challenged with PyPNP-INT, ΔPyPNP or ΔPyPNP-GFP survived and remained parasite-free. (Table 3). These mice were rechallenged 6–8 weeks later with a lethal dose of wild-type P. yoelii YM parasites (2 × 105 infected erythrocytes). Immunized mice were completely protected, with none developing detectable parasitemia (Table 3).

Table 3 Protection of immunized mice against challenge with wild-type P. yoelii

The protection conferred by infection with attenuated PNP-deficient parasites was long lasting and also provided protection from the nonlethal 17XNL strain of P. yoelii. Twenty ΔPyPNP-immunized mice surviving a lethal challenge with YM (Experiment 3; Table 3) had no detectable parasitemia after challenge with P. yoelii 17XNL erythrocytic stages 5 months after initial exposure to ΔPyPNP. Age-matched control mice (five mice) developed parasitemias of up to 50% and then cleared the infection (data not shown).

Mice were also protected from 17XNL sporozoite challenge (Supplementary Fig. 4 online). Five BALB/c mice inoculated with 2.0 × 104 P. yoelii YM ΔPyPNP were exposed to the bites of 35 A. stephensi mosquitoes infected with P. yoelii 17XNL. Four of these mice developed low level parasitemia on days 3–5 that resolved within 3 d, with peak parasitemia of 0.05%. The naive mice all developed high parasitemias (up to 80%) but then cleared the infection (Supplementary Fig. 4).

Attenuated whole-parasite vaccines, particularly irradiated attenuated sporozoites, have been proposed as alternatives to subunit vaccines20,21,22. Whereas the implementation of such vaccines poses considerable logistical difficulties, better protection has been elicited with attenuated sporozoites than with candidate subunit vaccines22. More recently, genetically attenuated sporozoites have been developed23. Whole-parasite vaccines may circumvent the problems resulting from heterogeneous host responses to individual antigens by providing an array of conserved immunogenic proteins less subject to immune pressure than current subunit vaccine components24.

A small study in humans has suggested that low levels of malaria infection may be protective in humans25. This study was performed by drug-curing infected individuals and suggests that repeated low-level malaria infection may be protective25. Routine administration of chemotherapy is constrained by the widespread resistance to commonly used antimalarials and has the potential to enhance the rate of selection of drug-resistant organisms. An alternative approach might be vaccination with a highly attenuated asexual stage parasite that clears spontaneously.

Development of a live attenuated vaccine strain requires a balance between the level of attenuation and the immunogenicity of the vaccine to optimize safety and efficacy. Some level of parasite replication will probably be required to mount an effective immune response26,27. The parasitemias detected after infection with PNP-deficient parasites are higher than are likely to be tolerated in humans, but our studies illustrate that development of attenuated erythrocytic stage metabolic mutants is feasible. Although attenuated sporozoite vaccines are promising, their success is dependent upon complete protection, as a single successful sporozoite infection with development of merozoites within hepatocytes could lead to symptomatic malaria28. In contrast, even a partially effective erythrocytic stage malaria vaccine could have considerable clinical impact on morbidity and mortality by decreasing the incidence of severe disease, such as cerebral malaria.

Ideally, genetically engineered Plasmodium vaccine strains will not revert to wild-type virulence but will be able to establish a self-limited, subclinical and protective course of infection. Because disruption of the gene encoding P. falciparum PfNT1 (ref. 29), a purine nucleoside transporter, results in parasites that cannot survive at physiological concentrations of purines, parasite strains with multiple purine-salvage deletions will probably be more highly attenuated than single-deletion strains. Parasite strains lacking PNP can still make gametocytes and may have activity as transmission-blocking vaccines while not being transmissible. Although we favor nontransmissible vaccines as being safer, it is possible that attenuated vaccine strains that can infect mosquitoes will promote protective immunity beyond the vaccinee and therefore be more beneficial to the general population. Finally, our studies establish the importance of Plasmodium purine-salvage enzymes for in vivo survival and support continued exploration of purine salvage as a chemotherapeutic target.


Cloning and expression of P. yoelii PNP.

PyPNP was identified as a protein of 244 amino acids and 26.943 kDa by protein sequence homology to P. falciparum PNP (gene PY04622; ). We amplified the PyPNP open reading frame sequence by PCR from genomic DNA of P. yoelii strain YM with primers PyPNP-F (5′-GATGAGGAACAAAGACATATAAAGC-3′) and PyPNP-R (5′-ATATTTTTCTGATAATCTTGCACAA-3′). We ligated the 729–base pair PCR product into a pTrcHis2-TOPO vector (Invitrogen) and expressed it in E. coli strain Top 10. We induced expression of recombinant PyPNP with a C-terminal 6 × His tag and the c-MYC epitope with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at 37 °C for 10 h and purified recombinant protein by nickel chromatography. The purified PyPNP was >95% homogeneous, as indicated by denaturing PAGE and staining with Coomassie blue. We measured protein concentration by Bradford protein assay (Bio-Rad).

Enzymatic assays and determination of kinetic constants.

We performed all phosphorylase activity assays with purified enzyme in 50 mM KPO4 pH 7.4 as previously described1. Briefly, we measured phosphorylysis of inosine, 2′-deoxyinosine, and 5′-methylthioinosine in a coupled assay with 115 milliunits ml−1 xanthine oxidase to convert hypoxanthine into uric acid. We followed uric acid formation by spectrophotometric measurement of absorbance at 293 nm (ε293 = 12.9 mM−1 cm−1). We monitored phosphorylysis of guanosine to guanine by measuring the disappearance of guanosine at an absorbance of 258 nm (ε258 = 5.2 mM−1 cm−1). We measured adenosine and 5′-methylthioadenosine phosphorylase activities by following the disappearance of the substrate at an absorbance of 265 nm (ε256 = 1.9 mM−1 cm−1). We measured uridine phosphorylase activity by following the conversion of uridine to uracil at an absorbance of 272 nm (ε272 = 2.9 mM−1 cm−1). We determined the Michaelis constant Km as the substrate concentration at one-half the maximum velocity. We determined the catalytic constant kcat for each substrate by fits of substrate saturation data to the Michaelis-Menten equation. Values for kcat assume one catalytic site per subunit and are expressed as moles of product per second per mole of subunit.

Phenotypic analysis of parasite growth and infectivity in mice.

To determine whether the disruption of PyPNP had an effect on parasite growth in vivo, we injected parasite inocula (2 × 104 or 2 × 105 blood-stage parasites) intravenously into groups of BALB/c mice (female, 8 weeks of age). We prepared a thin blood smear daily from each infected mouse and stained it with Giemsa reagent (Sigma). For each sample, we examined at least 1,000 red blood cells by microscopy and calculated the percentage of infected erythrocytes (parasitemia). All experiments with mice were conducted in Association for Assessment and Accreditation of Laboratory Animal Care–approved facilities after review and approval of experimental protocols by the Institutional Animal Care and Use Committees of Albert Einstein College of Medicine and New York University School of Medicine.

Phenotypic analysis of parasite development in mosquitoes.

We infected A. stephensi mosquitoes with P. yoelii parasites by blood-feeding on two sequential days for 15 min on infected Swiss Webster mice, and we subsequently maintained them under 80% humidity at 24 °C. We monitored the presence of gametocytes on Giemsa-stained blood smears before mosquito blood meal. We dissected infected mosquitoes at days 8 and 15 after the first infectious blood meal to evaluate the development of parasites by counting the numbers of oocysts per midgut and the numbers of sporozoites per salivary gland. We determined the number of sporozoites per salivary gland by mixing the salivary glands of 20 infected mosquitoes in 200 μl of PBS and counting the numbers of sporozoites in duplicate in a hemocytometer.

Statistical analyses.

We analyzed statistical significance by analysis of variance with the GraphPad Prism software package.

Note: Supplementary information is available on the Nature Medicine website.