Purine Salvage Pathways in the Apicomplexan Parasite

We have exploited a variety of molecular genetic, biochemical, and genomic techniques to investigate the roles of purine salvage enzymes in the protozoan parasite Toxoplasma gondii . The ability to generate defined genetic knockouts and target transgenes to specific loci demonstrates that T. gondii uses two (and only two) pathways for purine salvage, defined by the enzymes hypoxanthine-xanthine-guanine phosphoribosyltransferase (HXGPRT) and adenosine kinase (AK). Both HXGPRT and AK are single-copy genes, and either one can be deleted, indicating that either one of these pathways is sufficient to meet parasite purine requirements. Fitness defects suggest both pathways are important for the parasite, however, and that the salvage of adenosine is more important than salvage of hypoxanthine and other purine nucleobases. HXGPRT and AK cannot be deleted simultaneously unless one of these enzymes is provided in trans, indicating that alternative routes of functionally significant purine salvage are lacking. Despite previous reports to the contrary, we found no evidence of adenine phosphoribosyltransferase (APRT) activity when parasites were propagated in APRT-deficient host cells, and no APRT ortholog is evident in

We have exploited a variety of molecular genetic, biochemical, and genomic techniques to investigate the roles of purine salvage enzymes in the protozoan parasite Toxoplasma gondii. The ability to generate defined genetic knockouts and target transgenes to specific loci demonstrates that T. gondii uses two (and only two) pathways for purine salvage, defined by the enzymes hypoxanthine-xanthine-guanine phosphoribosyltransferase (HXGPRT) and adenosine kinase (AK). Both HXGPRT and AK are single-copy genes, and either one can be deleted, indicating that either one of these pathways is sufficient to meet parasite purine requirements. Fitness defects suggest both pathways are important for the parasite, however, and that the salvage of adenosine is more important than salvage of hypoxanthine and other purine nucleobases. HXGPRT and AK cannot be deleted simultaneously unless one of these enzymes is provided in trans, indicating that alternative routes of functionally significant purine salvage are lacking. Despite previous reports to the contrary, we found no evidence of adenine phosphoribosyltransferase (APRT) activity when parasites were propagated in APRT-deficient host cells, and no APRT ortholog is evident in the T. gondii genome. Expression of Leishmania donovani APRT in transgenic T. gondii parasites yielded low levels of activity but did not permit genetic deletion of both HXGPRT and AK. A detailed comparative genomic study of the purine salvage pathway in various apicomplexan species highlights important differences among these parasites.
Like all parasitic protozoa, the obligate intracellular parasite Toxoplasma gondii lacks the ability to synthesize the purine ring de novo, and thus relies entirely on the salvage of purines from the host cell to meet its nutritional needs (1)(2)(3). This requirement, coupled with the shortcomings of conventional therapies for treating congenital toxoplasmosis and opportunistic infections associated with AIDS and other immunosup-pressive conditions (4 -8), makes purine salvage an attractive target for chemotherapy.
The purine metabolism of T. gondii has previously been examined biochemically, resulting in the identification of various activities capable of assimilating nucleosides and nucleobases from the host cell into the purine nucleotide pools of the parasite (2,3). (See "Discussion" for a model of the purine salvage pathway in Toxoplasma and other apicomplexan parasites.) Reported salvage activities include the phosphoribosylation of adenine, guanine, hypoxanthine, and xanthine, and the phosphorylation of adenosine. The latter seems to contribute most significantly to parasite purine economy, as adenosine is incorporated into nucleotide pools at a considerably higher rate than any purine nucleobase (2,3).
Most of the reported salvage activities can be accounted for by two enzymes: hypoxanthine-xanthine-guanine phosphoribosyltransferase (HXGPRT) 1 and adenosine kinase (AK). The genes for both have been cloned and expressed in bacterial systems, and the purified proteins have been examined biochemically and structurally (9 -13). Genetic studies indicate that neither enzyme is essential for parasite viability, suggesting that the purine salvage pathways of the parasite are functionally redundant. The incorporation of labeled inosine and hypoxanthine into adenine nucleotides, and of labeled adenosine into IMP, indicates that AMP and IMP are interconvertible. The conversion of AMP to IMP (via AMP deaminase), and of IMP to AMP (via adenylosuccinate synthetase/lyase) probably predominate, although adenine and adenosine deaminase activities (3) may also play a role. Despite the previously measured incorporation of adenine by Toxoplasma, no adenine phosphoribosyltransferase gene has been identified in this or any other apicomplexan parasite.
In the present study, we have integrated genetic, biochemical, and genomic approaches to explore the contributions of HXGPRT and AK in maintaining a robust purine salvage pathway. The ability of T. gondii to grow in virtually any nucleated mammalian cell (14), combined with the availability of various mammalian somatic cell mutants and the ability to genetically manipulate the parasite (15), has allowed for a comprehensive analysis of the role played by individual parasite enzymes in overall purine nutrition. These studies provide formal proof that all purine salvage in T. gondii proceeds via HXGPRT or AK (there is no functional APRT activity), and fitness assays support the suggestion that AK is metabolically more important than HXGPRT. We have also exploited the large-scale genomic datasets now available for multiple apicomplexan species to create detailed maps of purine salvage pathways for several of these parasites, an important first step toward identifying potential drug targets for broad-spectrum anti-parasitic chemotherapy.

MATERIALS AND METHODS
Parasites, Cells, Chemicals, and Reagents-T. gondii strain RH (available through Ogden Bioservices Corporation, Rockville, MD) or mutants derived from this strain were used for all in vitro tissue culture experiments and were cultivated in primary human foreskin fibroblasts (HFF cells) or other fibroblast cells, as described previously (15,16). [  Molecular Genetic Manipulations-All molecular manipulations were performed according to standard protocols (15). Isolation of parasite genomic DNA was carried out by digestion with proteinase K (200 g/ml) in the presence of 1% sodium dodecyl sulfate, followed by phenol/ chloroform extraction and ethanol precipitation. Digested genomic DNA was blotted to Nytran filter paper and processed as described. The AK probe consisted of the open reading frame of the AK cDNA excised from the pBAceAK expression vector (11) by digestion with NdeI and XbaI. The HXGPRT probe was isolated from a similar expression vector (9) by digestion with NdeI. Probes were purified from agarose gels by Qiagen gel purification kit and 32 P labeled utilizing Stratagene Prime-It II TM random primer labeling kit.
Generation of Defined Knockouts-Plasmids for generating defined knockouts at the HXGPRT or AK loci (constructs HXGPRT-KO and AK-KO) have been described previously, as have the methods employed to isolate mutant parasites (9, 10). The HXGPRT::AK knockout construct was constructed by inserting an AK cDNA driven by the T. gondii dihydrofolate reductase promoter into the HXGPRT-KO vector. The construct pminAK (10) was digested with HindIII and NotI to remove the AK expression cassette (AK cDNA flanked by dihydrofolate reductase 5Ј and 3Ј sequences), and the resulting fragment was blunted and gel-purified using the Qiagen gel purification kit. The HXGPRT-KO vector was digested with SalI and blunted by removal of a 5.5-kb SalI fragment, which was replaced by ligation with the AK expression cassette. The (blunt) 5.5-kb SalI fragment was digested with SacI, and the resulting fragment was reintroduced by ligation into the vector polylinker (NotI [blunt]-SacI) downstream of the AK expression cassette. Targeting of this AK transgene to the HXGPRT locus was accomplished as described (9) by selection in 320 g/ml 6-thioxanthine (6TX).
Enzyme Assays-After lysis of the host cell monolayer, parasites were isolated from residual host cell debris by filtration through 3-m Nucleopore filters. Parasites were pelleted by centrifugation at 2500 rpm for 20 min at 4°C and washed twice in phosphate-buffered saline before resuspension in 500 l of TMD buffer (100 mM Tris, pH 7.5, 5 mM MgCl 2 , 2 mM dithiothreitol) and lysis by sonication on ice or the addition of 1% Triton X-100. Protein concentrations were determined by using the Bio-Rad protein assay and normalized for all enzymatic determinations. All enzyme assays were conducted according to the radiometric methods described previously (17,18). AK reactions were performed in a 50-l volume at 37°C in TMD buffer with 1 mM ATP, 5 mM NaF (to inhibit phosphatase activity), 25 Ci of [ 3 H]adenosine (48 Ci mmol Ϫ1 ), and appropriate volumes of crude parasite lysate. Reactions were terminated by spotting 10 l of the reaction mix onto DEAE anion-exchange filters. Filters were washed 3ϫ in water and once in ethanol, dried completely at 60°C; radioactivity was assessed by liquid scintillation counting. Measurements were taken over a 120-min time course. Xanthine phosphoribosyltransferase (XPRT) and adenine phosphoribosyltransferase (APRT) reactions were performed similarly, using PRT reaction mixes containing TMD buffer with 1 mM phosphoribosyl pyrophosphate, 0.3 Ci of [ 14 C]xanthine (57 Ci mol Ϫ1 ) or 1.3 Ci of [ 14 C]adenine (56 Ci mol Ϫ1 ), and appropriate volumes of crude lysate.
Sensitivity of T. gondii to Toxic Adenine Analogs-The growth of intracellular T. gondii was measured in confluent 24-well plates of either wild-type human foreskin fibroblasts or APRT-deficient mouse fibroblasts by the incorporation of [ 3 H]uracil into acid-precipitable material. Parasites were grown in the presence of either 2-fluoroadenine (2-FA) (2 ng/ml to 200 g/ml) or 8-azaadenine (8-azaA) (10 ng/ml to 100 g/ml). After 48 h growth, parasites were subjected to a 4-h pulse of 5 Ci of [ 3 H]uracil (20 Ci mmol Ϫ1 ), and plates were processed as described (15).
In Vitro Uptake of [ 14 C]Adenine-Parasites were infected at a density of 5 ϫ 10 5 cells per well into 24-well plates containing confluent monolayers of APRT-deficient mammalian host cells. After 24 h, 1.3 Ci of [ 14 C]adenine (56 Ci mol Ϫ1 ) was added to each well, and adenine incorporation was allowed to proceed for 4 h before plates were processed as described for measurement of [ 14 C]adenine incorporation (15). Acid precipitable material was measured by liquid scintillation counting.
Expression of Leishmania donovani APRT in T. gondii-The open reading frame of the L. donovani APRT (19) was PCR-amplified using Advantage cDNA polymerase (Clontech), utilizing a 5Ј primer, GGAa-gatctATGGCCTTCAAGGAAGTCAG, and a 3Ј primer, AGTAgctagcGT-GCGGGTGCTCGGCCAG (restriction sites in lowercase). The 5Ј primer converted the second amino acid of LdAPRT from Pro to Ala, for optimized expression in Toxoplasma (20). The amplified fragment was digested with BglII and NheI, gel-purified, and ligated into appropriately digested ptubACP-YFP-HA vector, 2 replacing the acyl carrier protein-green fluorescent protein open reading frame with that of the LdAPRT. This vector also contains a chloramphenicol acetyl transferase selectable marker under the control of the SAG1 promoter (21). The resulting ptubLdAPRT-HA fusion construct was electroporated into wild-type, ⌬AK, and ⌬HXGPRT parasites as described (15). Transfected parasites were allowed to infect confluent fibroblast monolayers in 24-well plates, and after 24 h of growth, were processed for immunofluorescence detection of APRT-HA fusion protein. Samples were fixed in paraformaldehyde, permeabilized in Triton X-100, and assayed with Alexa Fluor 488-conjugated anti-hemagglutinin (HA) antibody (Molecular Probes, 1:1500 dilution). Transfected parasites were also subjected to chloramphenicol selection (6 g/ml); parasites surviving three rounds of passage in drug were cloned by limiting dilution in drug (15) and assayed for APRT activity.
Fitness Assays-Competition assays between the parental wild-type and ⌬HXGPRT and ⌬AK parasites were performed as described (22). For detection of viable wild-type or knockout parasites at various timepoints, plaque assays were performed in 6-well plates in the presence of suitable selective drugs (15). Mycophenolic acid or 6-thioxanthine was used as selection for ⌬HXGPRT versus wild-type competition: the former selects for wild-type parasites, whereas the latter selects for ⌬HXGPRT parasites. Similarly, adenine arabinoside was used to select for ⌬AK mutants in competitions against wild-type parasites.
Genomic Analysis-A BLAST searchable database of all publicly available Apicomplexan genomic and expressed sequence tag sequences (constructed by Dr. Martin Fraunholz, University of Pennsylvania, Philadelphia, PA) was used for this study. This database contains complete or nearly complete genomic sequences for Plasmodium falciparum and Plasmodium yoelii (5X), T. gondii (10X), Eimeria tenella (5X) and Theileria annulata (8X), Cryptsporidium parvum (7X), Cryptosporidium hominis (12X), Tetrahymena thermophila, and additional genomic and expressed sequence tag sequences from P. chabaudi, P. vivax, P. knowlesi, T. gondii, E. tenella, and Theileria parva. 3 Putative homologs of purine salvage transporters and enzymes were identified by using TBLASTN (WU-BLAST), with a cutoff of p Ͻ 10 Ϫ9 . Previously published and/or annotated sequences for the Apicomplexan purine salvage genes were used as query sequences whenever possible. A complete list of query sequences and matches is provided as supplemental material.

Defined Genetic Knockouts of Purine Salvage Enzymes-Suc-
cessful genetic knock-outs of either the AK or HXGPRT loci have been described previously and shown to be deficient in the corresponding purine salvage activities (9,10). The ease of generating such mutants prompted efforts to delete both loci in a single parasite. A summary of the results from numerous such experiments is provided in Table I. Deletion of either individual locus was readily achieved, using AK-KO or HXG-PRT-KO constructs (see "Materials and Methods"). In contrast, in three independent experiments, we were unable to delete the AK locus in the ⌬HXGPRT background (based on adenine arabinoside selection of parasites transfected with the AK-KO construct), or to delete the HXGPRT locus in the ⌬AK background (based on 6TX selection of parasites transfected with the HXGPRT-KO construct). Parasites resistant to both adenine arabinoside and 6TX were occasionally isolated in the latter experiment (using 6TX selection), but DNA hybridization analysis revealed no obvious lesion at the HXGPRT locus, and Western blotting showed normal levels of HXGPRT protein.
These parasites are presumed to harbor substrate specificity mutations at the HXGPRT locus or mutations at other loci that affect 6TX toxicity.
The inability to delete both HXGPRT and AK suggests that the double-knockout phenotype may be lethal for T. gondii. To further address this hypothesis, we attempted to genetically delete both loci, while providing AK activity in trans (producing a double knock-out genotype that is nevertheless AK ϩ ). The HXGPRT-KO construct was modified to contain the coding region of T. gondii AK (driven by the T. gondii dihydrofolate reductase promoter), generating construct HXGPRT::AK. 6TXresistant parasites were readily obtained after transfection of ⌬AK parasites, and the resulting clones were tested by Southern analysis to assess their genotype and enzyme assays to assess purine salvage activities, as shown in Fig. 1.
As expected, probing with AK cDNA demonstrates genomic EcoRI fragments of 2.1 and 3.8 kb in wild-type and ⌬HXGPRT parasites and deletion of the larger fragment in the ⌬AK mutant (10). This same pattern was observed in the HXGPRT::AK mutant, demonstrating maintenance of the ⌬AK genotype. In addition, a larger band was also observed of the size expected from integration of the HXGPRT::AK plasmid at the HXGPRT locus (Fig. 1, arrowhead). The same size band was observed with an HXGPRT probe, which also shows that the endogenous HXGPRT locus (observed in wild-type parasites and ⌬AK mutants) was successfully disrupted in the HXGPRT::AK mutant. The smaller size of the HXGPRT locus in this mutant relative to the ⌬HXGPRT clone is attributable to molecular manipulations associated with construction of the HXGPRT::AK vector (see "Materials and Methods").
Enzyme assays demonstrate no AK activity in the ⌬AK parent but show activity comparable with that of wild-type parasites (and ⌬HXGPRT mutants) when AK was targeted to the HXGPRT locus in the ⌬AK background. XPRT assays show normal activity in the ⌬AK parent but none in HXGPRT:AK parasites, similar to the ⌬HXGPRT mutants characterized previously (9). Taken together, these results demonstrate that it is possible to disrupt both the AK and HXGPRT genetic loci simultaneously, but only if one of these activities is provided in trans (in this case, by expression of AK at the disrupted HXG-PRT locus).
APRT Activity and Parasite Sensitivity to Subversive Substrates of APRT-The inability to generate ⌬HXGPRT, ⌬AK double-knock-out parasites is somewhat surprising, given the previous identification of APRT activity in T. gondii (3), which could provide for an alternative route of purine assimilation for the parasite. Consistent with prior reports, RH-strain T. gondii cultivated in normal human foreskin fibroblasts were susceptible to both 2-FA (IC 50 ϳ 75 ng/ml) and 8-azaA (IC 50 ϳ 1.5 g/ml), as shown in Fig. 2 (white symbols). These adenine analogs are known to function as subversive substrates of APRT, and both have been shown to produce APRT-dependent toxicity in other systems (23)(24)(25). When parasites were grown in APRT-deficient cells (16), however, IC 50 s for both drugs increased to Ͼ100 g/ml, a decrease in sensitivity of ϳ2.7 ϫ 10 3 and 67-fold for 2-FA and 8-azaA, respectively (Fig. 2, black  symbols). These results suggest that toxicity of 2-FA and 8-azaA is dependent upon host, rather than parasite, APRT activity.
Furthermore, although variable levels of APRT activity were observed in crude cytosolic extracts of parasites cultivated in wild-type human foreskin fibroblasts (data not shown), we were completely unable to detect phosphoribosylation of radiolabeled adenine when parasites were grown in APRT-deficient fibroblasts (Table II), supporting the suggestion that parasites lack an endogenous APRT activity. In addition, T. gondii tachyzoites grown in APRT-deficient host cells incorporate only very low levels of [ 4 C]adenine (see Fig. 3B).
Expression of L. donovani APRT in T. gondii-To further probe purine salvage pathways in T. gondii, we attempted to FIG. 1. Genomic organization and enzymatic activity of knockout parasites. Wild-type parasite, ⌬AK knock-out mutants, ⌬HXG-PRT knock-out mutants, and ⌬AK parasites in which an AK transgene was used to disrupt the HXGPRT locus (⌬AK, HXGPRT::AK) were harvested, subjected to Southern analysis of EcoRI-digested genomic DNA using AK or HXGPRT cDNA probes (A and C, respectively), and subjected to assays for AK or HXGPRT activity in crude parasite lysates (B and D, respectively), as described under "Materials and Methods." The arrowhead indicates the AK transgene introduced at the HXGPRT locus in ⌬AK and HXGPRT::AK mutants. This insertion restored AK activity while eliminating HXGPRT activity (enzyme activity presented as a percentage of values determined for control parasites). gondii. An LdAPRT-HA fusion construct (see "Materials and Methods") was expressed in the parasite cytosol (Fig. 3A), and the resultant protein possesses APRT activity both in vivo (Fig.  3B) and in vitro ( Fig. 3C and Table II). Wild-type parasites incorporate very little [ 14 C]adenine into acid-precipitable material when grown in APRT-deficient host cells, in contrast to transgenic parasites expressing L. donovani APRT (Fig. 3B). APRT activity was virtually undetectable in cell-free extracts of wild-type RH-strain parasites (grown in APRT-deficient host cells), but significant levels of activity were seen in RH:LdA-PRT, ⌬AK:LdAPRT, and ⌬HXGPRT:LdAPRT transgenics ( Fig.  3C and Table II). Analysis of other purine salvage activities in parasites expressing high levels of LdAPRT indicates that HXGPRT and AK activities are unaffected by the presence of the transgenic APRT activity (Table II). Interestingly, the relative values of these salvage activities suggest that adenosine utilization in T. gondii is significantly more efficient than for other purines, whereas the activity of transgenic APRT is detectable but considerably lower than either AK or HXGPRT.
To explore whether APRT-expressing parasites might be amenable to genetic deletion of both the AK and HXGPRT loci, ⌬AK:LdAPRT and ⌬HXGPRT:LdAPRT parasites were transfected with HXGPRT-KO or AK-KO vectors (respectively) and subjected to selection for deletion of either locus. As indicated in Table I, no double knock-outs have been identified to date.
Fitness Assays for ⌬AK and ⌬HXGPRT Parasites-Previous studies on ⌬AK and ⌬HXGPRT mutants showed no dramatic differences in growth rates as compared with the wild-type parasites from which they were derived (9,10). More recently, however, we have found that sensitive competition assays enable the detection of subtle fitness defects in other mutants (22). As shown in Fig. 4, both the ⌬AK and ⌬HXGPRT knockout lines showed significant fitness impairment, with ⌬AK parasites displaying a fitness defect of 7.6% per generation (Fig. 4A) and ⌬HXGPRT exhibiting a defect of 3.7% per generation (Fig. 4B). These findings indicate that both AK and HXGPRT play an important role in parasite metabolism and fitness, although the presence of neither gene is essential for viability, as noted above. The larger defect observed in ⌬AK supports the suggestion that this enzyme plays the more important role in purine salvage.
Comparative Genomic Analysis of Purine Salvage Pathways in Related Species-Taking advantage of the large-scale genomic datasets that have recently become available for various apicomplexan species (26 -28), we have constructed a detailed model for purine salvage pathway in these parasites, as shown in Fig. 5. The ciliate T. thermophila was also included in this analysis, as ciliates, dinoflagellates, and apicomplexans are thought to be sister taxa, grouped in the superphylum Alveolata (29). No genes encoding de novo purine synthetic enzymes are evident in any of these protozoa, suggesting that all are incapable of purine biosynthesis. This is true even for the free-living ciliate T. thermophila, which supports previous studies on the incorporation of [ 14 C]glycine and formate (30). Many of these genes are evolutionarily well conserved, and redundant synthetic pathways have not been described, making it unlikely that purine synthetic pathways were missed in this analysis.
Purine salvage requires the ability to transport nucleosides and/or nucleobases, and equilibrative-type transporters

TABLE II Specific activities for purine salvage enzymes in wild-type and
LdAPRT-expressing parasites Determination of enzyme activities was conducted as described under "Materials and Methods." All values are given in nmol min Ϫ1 g Ϫ1 and represent the mean Ϯ S.D. of three independent measurements. (ENTs) have previously been reported for both T. gondii (31) and P. falciparum (32). TgAT1 transports adenosine, inosine, and guanosine nucleosides (31), and there is biochemical data for additional adenosine, inosine, and nucleobase transporters (33,34). The parasite genome provides evidence for three additional ENTs. PfNT1 transports both purine and pyrimidine nucleosides, and the genome also shows evidence for at least one additional ENT. A single ENT is evident in the Cryptosporidium genome. No obvious nucleoside/base transporter was found in the Theileria or Tetrahymena genomes.
As noted above, Toxoplasma expresses a complex and robust pathway, including both AK and HXGPRT, but not APRT. The pathway in Eimeria tenella seems to be very similar. In contrast, Plasmodium, Cryptosporidium, and Theileria all exhibit a simplified, stripped down version of this pathway. Plasmodium relies solely on HXGPRT, whereas Cryptosporidium and Theileria rely solely on AK. All of these parasites lack APRT, but interestingly, the genome of Tetrahymena contains an APRT gene (as well as HGPRT, but not AK). Enzymes for interconverting AMP and IMP are present in all of these species except Cryptosporidium, which is unable to produce IMP except from AMP itself. These organisms also seem to differ in their ability to interconvert nucleosides and nucleobases. Toxoplasma and Plasmodium possess purine nucleoside phosphorylase (and probably adenosine deaminase). Tetrahymena uses a distinct (Family 2) purine nucleoside phosphorylase and a nucleoside hydrolase. Cryptosporidium and Theileria lack all of these genes. DISCUSSION This report describes an integrated genetic, biochemical, and genomic approach to assess the contribution of salvage path-ways to the purine economy of T. gondii and related organisms. T. gondii possesses two functionally redundant salvage enzymes, HXGPRT and AK, and previous results have shown that this parasite can survive elimination of either activity alone, i.e. neither HXGPRT nor AK is essential for T. gondii survival (9,10). Attempts to knock out both HXGPRT and AK in a single parasite were unsuccessful, providing circumstantial evidence that functional expression of at least one of these two enzymes is essential (Table I). Direct support for this hypothesis is provided by the ⌬AK and HXGPRT::AK parasites (Fig. 1), where expression of AK activity in trans permitted simultaneous disruption of the endogenous genomic loci for both AK and HXGPRT. These experiments show that AK and HXGPRT provide the only two physiologically relevant routes for purine acquisition in T. gondii. ⌬AK parasites exhibited a greater fitness defect than ⌬HXGPRT mutants (Fig. 4), arguing that flux through AK is probably greater than HXGPRT.
Previous biochemical studies have noted APRT activity in T. gondii (3), with parasite extracts exhibiting higher specific activity of adenine phosphoribosylation than that of guanine (but less than hypoxanthine or xanthine). In reviewing these data, however, we note that radiolabeled adenine was incorporated into nucleotide pools with a distribution virtually identical to that observed for hypoxanthine (3), suggesting that adenine may have been deaminated to hypoxanthine and incorporated via HXGPRT, rather than APRT (Fig. 5). Alternatively, the high levels of APRT present in mammalian host cells (35) may have resulted in inadvertent contamination of parasite extracts. In our hands, parasites grown in wild-type host cells were susceptible to toxic adenine analogs, but those grown in APRT-deficient host cells (16) were not (Fig. 2). Moreover, we found no APRT activity in wild-type RH-strain parasites or parasite extracts when APRT-deficient host cells were used for parasite cultivation (Fig. 3). Expression of a heterologous APRT allowed transgenic parasites to incorporate low levels of radiolabeled adenine (Fig. 3), although this activity was insufficient to permit inactivation of both HXGPRT and AK (Table I).
APRT activity has also been reported in Plasmodium (36,37), Eimeria (38), and Cryptosporidium (39). However, immucillin H (which inhibits purine nucleoside phosphorylase, and hence the acquisition of purines via HXGPRT) is lethal to P. falciparum (40), arguing against a biologically significant role for APRT. No APRT genes were evident in the genome of any apicomplexan parasite (Fig. 5), and it seems unlikely that any such genes were missed, as all known APRT genes are highly conserved in primary sequence (and no alternative enzymes are known to phosphoribosylate adenine). We conclude from the available genetic, biochemical, and genomic data that APRT is lacking from all of these apicomplexan parasites.
Interestingly, the T. thermophila genome encodes a putative APRT, suggesting that the last common ancestor of the Alveolata may have possessed three purine salvage pathways, utilizing APRT, AK, or HXGPRT. Phylogenetic analysis (not shown) provides no evidence for horizontal transfer of any of these genes, in contrast to Cryptosporidium IMPDH (41). Biochemical studies indicate that this ciliate also exhibits other differences from most of the apicomplexa, including a nucleoside hydrolase and the lack of XPRT and GMP synthetase activity (42); these observations are supported by the genome sequence. Genomic analysis also reveals a gene likely to encode GMP reductase, explaining the observation that Tetrahymena can survive on guanine or guanosine as a sole purine source (30). A probable GMP reductase was also observed in Theileria species, but this enzyme seems to have been lost in other apicomplexans. FIG. 4. Fitness of ⌬AK and ⌬HXGPRT mutants. ⌬AK or ⌬HXG-PRT parasites were mixed at a 1:1 ratio with the wild-type parasites from which they were originally derived and maintained by serial subculture (see "Materials and Methods"). Mutant:wild-type ratios were assayed every 6 days, as described previously (22); distinct symbols indicate three parallel replicates in each of two independent experiments (ϩ, lowest limit of detection in the plaque assays, ϳ1:128).
Results are plotted on a log 2 scale, enabling the relative fitness of mutant versus wild-type parasites to be calculated based upon the slope of the line (described in Ref. 22). ⌬HXGPRT parasites exhibited a fitness defect of 3.7% per generation. ⌬AK parasites displayed a 7.6% fitness defect per generation and were lost from the culture by day 18 (ϳ66 generations).
In sum, HXGPRT and/or AK seem to provide the sole routes for purine assimilation in apicomplexans, but different species may possess one, the other, or both of these activities. Toxoplasma and Eimeria parasites contain genes predicted to encode both enzymes, as well as many interconverting enzymes indicated by arrows in Fig. 5. Adenosine deaminase was not readily identifiable in these parasite genomes, but this gene is not highly conserved, and the observed conversion of adenosine into inosine (3,10,43) suggests that it is probably present. Biochemical studies also indicate an adenine deaminase in Eimeria (43), although this was not evident in the genome. Theileria and Cryptosporidium species contain AK, but seem to have lost HXGPRT. One report describes HXGPRT activity in Cryptosporidium (39), but this is not supported by the parasite genome sequence or more recent biochemical and pharmacological studies (44). We were unable to detect AMP deaminase in the T. parva or T. annulata genomes (see supplementary material), but presume that this activity must be present as the sole means for guanylate nucleotide production. Plasmodium expresses a highly active and well characterized HXGPRT activity (45) but lacks AK.
Several subversive purine analogs and inhibitors have shown efficacy against protozoan pathogens (46 -48), but functional redundancy in purine salvage pathways poses a potential impediment to rational drug design, as effective therapies may have to inhibit multiple activities to maximize potency and reduce the likelihood of resistance. The lack of an APRT activity in T. gondii indicates that only HXGPRT and AK would have to be blocked, and both of these enzymes have previously been characterized biochemically and structurally (9,11,13,49). The significant fitness effect of genetically deleting either locus (Fig. 4) suggests that this may be a viable drug target, even in T. gondii. The more highly reduced purine salvage pathways observed in other parasites makes these targets even more attractive. For example, inhibiting any of the five activities leading from host cell adenosine to GMP in the Cryptosporidium panel in Fig. 5 (adenosine transport, adenosine kinase, AMP deaminase, IMP dehydrogenase, GMP syn- FIG. 5. Comparative analysis of purine salvage pathways. The presence or absence of purine transport and salvage pathway enzymes in various apicomplexan parasites and the ciliate T. thermophila was determined based upon TBLASTN analysis of available genome sequences (see "Materials and Methods"), motif searches, and manual curation. HC, host cell cytoplasm; PV, parasitophorous vacuole; PC, protist cytoplasm; EC, extracellular medium. Transporters are indicated by ovals (the number of ovals does not necessarily reflect the number of predicted transporters; see "Discussion"). Purine salvage enzymes are indicated by arrows labeled AK, HXGPRT, APRT, or with numbers in the Toxoplasma panel (unless otherwise indicated): 1, adenosine deaminase; 2, purine nucleoside phosphorylase (2*, nucleoside hydrolase, in Tetrahymena only); 3, adenine deaminase; 4, AMP deaminase; 5, IMP dehydrogenase; 6, GMP synthetase; 7, adenylosuccinate synthetase; 8, adenylosuccinate lyase; 9, GMP reductase (in Theilera and Tetrahymena only). Dashed lines indicate activities not evident from the genome sequence, but for which there is strong supporting biochemical and/or pathway evidence. Larger arrows are intended to indicate dominant purine salvage pathways. See text for further discussion. thetase) would be expected to kill the parasite. The necessary framework in terms of structural and biochemical data for both the parasite and mammalian enzymes is available (9, 11-13, 45, 50 -54) and should expedite the development of antiparasitic drugs with high therapeutic potential.