Purine salvage is an essential function for all obligate parasitic protozoa studied to date and most are also capable of efficient uptake of preformed pyrimidines. Much progress has been made in the identification and characterisation of protozoan purine and pyrimidine transporters. While the genes encoding protozoan or metazoan pyrimidine transporters have yet to be identified, numerous purine transporters have now been cloned. All protozoan purine transporter-encoding genes characterised to date have been of the Equilibrative Nucleoside Transporter family conserved in a great variety of eukaryote organisms. However, these protozoan transporters have been shown to be sufficiently different from mammalian transporters to mediate selective uptake of therapeutic agents. Recent studies are increasingly addressing the structure and substrate recognition mechanisms of these vital transport proteins.
Protozoan purine transporters have attracted much interest in recent years. As a result, purine transporters have been identified in many of the most important pathogenic protozoa. Until very recently, all this work was performed on intact parasites, by measuring transport of radiolabelled purines and pyrimidines. Possibly the first such study was by Tracy and Sherman, who reported the uptake of radiolabelled purines by the avian malaria parasite Plasmodium lophurae in 1972 , though other researchers had already started to study protozoan purine and pyrimidine metabolism (e.g. ) The studies on purine metabolism revealed that, unlike most mammalian cells, protozoan parasites are unable to synthesise the purine ring de novo and rely solely on salvage mechanisms for these essential nutrients [3,4]. This dependence potentially makes protozoa vulnerable to inhibitors of the purine salvage pathways [5,6]. In contrast, parasitic protozoa are fully capable of synthesising the pyrimidine ring de novo, with the exception of Giardia lamblia, Tritrichomonas foetus and Trichomonas vaginalis[4,7], yet are also capable of salvaging pyrimidines such as thymidine or uracil (e.g. [8–11]).
Using biochemical techniques and live parasites, numerous purine and pyrimidine transport activities have been identified in a variety of protozoan species [12,13]. One important early observation was that many protozoan species, including Giardia intestinalis[14,15], Leishmania donovani[9,16], Crithidia luciliae, and Trypanosoma brucei possess at least two nucleoside transport activities in some of their life cycle stages, identified by their kinetic parameters and selectivity profile. In addition, they may express one or more nucleobase transporters [8,19–24]. The overlap in specificity between various purine and/or pyrimidine transporters often makes it hard to study them in situ, as specific inhibitors or permeants are often lacking.
However, starting in 1998, the cloning and heterologous expression of a number of protozoan nucleoside transporters from L. donovani[25,26], T. brucei[27,28], Toxoplasma gondii and Plasmodium falciparum[30,31] were reported in quick succession. These breakthroughs allowed the characterisation of protozoan nucleoside transporters in isolation for the first time, and generally confirmed the earlier experiments with intact parasites. Heterologous expression in a specially selected and well-characterised system also greatly facilitates the study of transporters that are otherwise not easily available for biochemical analysis, due to, for example, expression in life cycle forms that cannot be obtained in sufficient numbers or expression at an intracellular location. Yet, it needs to be stressed that in situ and heterologous approaches are complementary, as the observation of nucleoside transport in, for instance, oocytes of Xenopus laevis or a yeast cell, does not necessarily yield the same kinetic parameters as in the original cell, nor does it reveal much in itself about the physiological role of the transporter in the parasite's biochemistry and physiology. Nonetheless, several specific nucleoside transporter genes have now been unambiguously coupled to a particular transport activity in the parasite [25–29]. The genes encoding nucleobase transporters, on the other hand, have proved more elusive, with the first cloning of genuine nucleobase transporter genes reported only in 2003 [32,33]. These first nucleobase carriers were cloned from T. brucei brucei and turned out to be of the Equilibrative Nucleoside Transporter family (ENT), as are all the protozoan nucleoside transporters cloned to date. It can be expected that, with the completion of more protozoan genomes, the current momentum of discovery will further increase, and the cloning of a first nucleobase transporter from Leishmania was reported in early 2004 . Even so, it is not yet clear whether all protozoan nucleobase and nucleoside transporters are encoded by ENT family genes.
Beyond the undoubted importance of protozoan nucleoside/nucleobase transporters for the salvage of essential nutrients, these proteins can also mediate the uptake of chemotherapeutic agents by the parasite. This can add a layer of selectivity over and above any selectivity on the level of the drug target. Indeed, in some cases selective uptake may be the only or main level of selective action, as in the case of melaminophenyl arsenicals that are efficiently accumulated through the T. b. brucei TbAT1/P2 transporter [18,34]. In this case, the melamine ring provides the haptophore, which can be coupled to various toxophores to selectively introduce cytotoxic agents into the trypanosome [35–38]. More generally, purine transporters can be used to selectively mediate the uptake of purine or pyrimidine antimetabolites [11,39,40]. Such antimetabolites are used widely as antiviral and anticancer drugs, but limited effort has been made to develop similar strategies for antiprotozoal agents. Even though various nucleoside and nucleobase analogues display good activity in vitro and/or in vivo (e.g. [41–45]) only the pyrimidine antifolate pyrimethamine is widely used clinically, particularly in combination with sulphadoxine against malaria , and also against toxoplasmosis [47,48]. In addition the purine nucleobase allopurinol is used in combinations against leishmaniasis [49,50].
As both the efficacy and the selectivity of nucleoside antimetabolites may be dependent on the mechanisms of uptake by the parasite, a rational approach to drug design must address the structural determinants that allow efficient transport. Ideally, antimetabolites would be taken up by more than one transporter, as drug resistance in protozoa is often associated with the loss of a single transport activity [51,52]. An intimate knowledge of the numbers of transporter (sub)-types, expression levels in the relevant life cycle stages and substrate specificity should therefore be considered as part of drug development strategies. Techniques to model the interactions between substrate and transporter binding site that determine affinity and selectivity have been developed and shown to have predictive value [39,53]. They further highlight strong functional conservation between transporters in different protozoan species, which cannot be readily determined from primary sequence data [8,22]. The following review will catalogue the protozoan purine and pyrimidine transport activities identified to date and categorise those transporter genes that have been functionally confirmed or that are apparent in current genome databases. The uses and limitations of quantitative structure-activity (QSAR) modelling of the interactions between these proteins and their substrates, in relation to protozoan drug uptake and resistance, will be discussed.
2Purine transport activities in various protozoan species
Nucleoside transport has been extensively studied in the trypanosomatidae, particularly in T. b. brucei and L. donovani, but to a much lesser extent also in Trypanosoma evansi[54–56], Trypanosoma equiperdum, Trypanosoma congolense, Trypanosoma vivax and Trypanosoma cruzi.
An early report by Sanchez et al. in 1976  showed the utilisation of [3H]adenosine nucleotides, particularly AMP, by T. b. brucei and T. congolense. However, the incubations over 15 min did not convincingly exclude the possibility that the nucleotides were being dephosphorylated prior to transport by the trypanosomal 5′-nucleotidase. In addition, nucleoside transporters are not commonly believed to transport nucleotides. However, it should be noted that an L. donovani nucleoside transporter, NT1.1, appears capable of AMP transport, generating substrate-dependent currents when expressed in Xenopus oocytes with an apparent Km of 9.1 ± 3.2 μM, though, as before, it is difficult to exclude the possibility of a contribution from a putative endonucleotidase on the surface of the oocyte.
James and Born  first demonstrated the uptake of radiolabelled adenosine, inosine, guanosine, hypoxanthine and adenine in T. b. brucei and T. congolense. Though they measured uptake over relatively short times and showed that the rate of uptake of adenosine was greater than for the other purines, the conclusions from their inhibition experiments were flawed as they did not measure initial rates of transport. Their conclusion of a low-affinity adenosine transporter in addition to a high affinity carrier remains to be confirmed, though recent unpublished observations do indicate the presence of such a transporter in T. b. brucei and T. equiperdum, which is sensitive to adenine and insensitive to hypoxanthine (M.P. Barrett and M.L. Stewart, personal communication).
2.1.1Characterisation of the P1/P2 system
Interest in protozoan nucleoside transporters increased dramatically after the seminal observation in 1993 by Carter and Fairlamb  that a strain of T. b. brucei adapted to melaminophenyl arsenicals (one of the main classes of clinical trypanocides ) ‘lacks an unusual adenosine transporter’. They identified two high-affinity adenosine transporters, designated P1 and P2, which could be specifically inhibited by inosine and adenine, respectively . Their experiments clearly implicated P2 in the transport of melaminophenyl arsenicals , through interaction with the melamine pharmacophore [53,63], and further demonstrated that this transporter could also mediate the uptake of pentamidine , another widely used trypanocide. The involvement of this transporter in drug uptake will be discussed in detail in Section 4.2. The P1/P2 system has since been studied extensively. P1 is a broad specificity purine nucleoside transporter [18,27,53,65] (see Table 1) with low affinity for uridine . It has very low affinity for purine nucleobases such as adenine and hypoxanthine [27,65]. In contrast, P2 has no measurable affinity for oxopurines (guanosine, inosine, hypoxanthine, guanine and allopurinol), but displays submicromolar affinity for aminopurines, notably adenine and adenosine (Table 1) [18,53]. The P1/P2 system has since also been characterised in T. equiperdum and T. evansi[54–56,66] and displayed very similar characteristics as in T. b. brucei.
Table 1. Kinetic parameters of T. b. brucei purine and pyrimidine transporters
Values in bold type represent Km values; all others are Ki values. When several values are available from the literature, as a rule those obtained using intact T. brucei rather than heterologous expression are given, and from the first report, if multiple exist. Numbers in table are references. ADO, adenosine; INO, inosine; GUO, guanosine; URD, uridine; CTD, cytidine; THD, thymidine; HYP, hypoxanthine; ADE, adenine; GUA, guanine; XAN, xanthine; HPP, allopurinol; URA, uracil; CYT, cytosine; THY, thymine.
NE, no effect on permeant uptake at a concentration of a, 1 mM; b, 400 μM; c, 250 μM; d, 100 μM; e, 25 μM. HA, high affinity.
P1 was also the first protozoan nucleoside carrier shown to be a proton symporter . While this study was performed before the cloning of the P1 transporter and was therefore unable to directly demonstrate and characterise a proton flux by electrophysiology, strong evidence for proton coupling was presented. The rate of [3H]adenosine uptake was pH-dependent, and inhibited by ionophores such as CCCP, nigericin and gramicidin as well as inhibitors of the T. b. brucei plasma membrane P-type H+-ATPase. Dose-dependent effects of all these conditions were also measured on the plasma membrane potential (Vm) and the intracellular pH (pHi), from which the proton-motive force (PMF) was calculated using Eq. (1):
in which R is the gas constant, F is Faraday's constant, T is the absolute temperature and pHo is the extracellular pH. A linear correlation between PMF and the rate of adenosine uptake was established (r2= 0.93 over 11 data points) . Furthermore, an adenosine-induced cytosolic acidification was demonstrated after ‘base-loading’ with NH4Cl and the electrogenic nature of adenosine and 2-Cl-adenosine transport was demonstrated by measuring their effect on plasma membrane potential after inhibition of the proton pump .
Very similar approaches established also that T. b. brucei nucleobase transporters in both the procyclic (insect-borne) and long-slender bloodstream forms are also proton symporters [19,20]. To date four hypoxanthine transporters have been characterised: H1 and H4 in procyclics [19,32] and H2 and H3 in bloodstream forms . In addition, a high-affinity uracil transporter [11,40] and a cytosine transporter (De Koning, unpublished observation) have been identified in procyclics. The hypoxanthine carriers are all broad specificity purine nucleobase transporters with, as a rule, little or no affinity for pyrimidines or nucleosides (Table 1). Their Km value for hypoxanthine varies from 123 nM to 9.3 μM and they are clearly distinguishable on this basis and on their inhibition profile. For instance, H2, but not H3 or H1, is inhibited by uracil and guanosine [19,20]. H4 is inhibited by both, with similar Ki values as H2 , but also displays equally high affinity for guanine, 3-deazaguanine and 7-deazaguanine, whereas H2 displays much lower affinity for these analogues than for the natural nucleobase .
2.1.4Regulation of purine transport activity
It therefore appears that T. b. brucei expresses at least three different purine transporters in its procyclic stage (P1, H1 and H4) and at least four distinct purine transport activities in its long-slender bloodstream stage (P1, P2, H2, H3), but the situation in other developmental stages is far less clear. Sanchez et al.  identified several additional adenosine transporters in a locus of 6 closely related genes that includes the NT2 gene that encodes P1  (see Section 3.1). mRNA for all six genes was detected in long-slender bloodstream forms . As trypanosomes can convert any of the natural purine nucleosides or nucleobases (excluding uric acid) to any other , and thus develop normally on any single purine source, it is not immediately clear why they should require this many purine transporters.
A partial explanation for this plethora of purine transport activities may lie in the adaptability of the purine salvage system. First studied in Crithidia spp. (see Section 2.3), the purine salvage enzymes and transporters of the trypanosomatids can be dramatically up- or downregulated according to growth stage and availability of a purine source [24,68–74]. In T. b. brucei, adenosine and hypoxanthine transporters in procyclics appear to be differentially controlled under these conditions: hypoxanthine transport rates increased to 450% of control within 24 h of purine stress, whereas adenosine transport capacity was only increased after 48 h . Indeed, of two hypoxanthine transport activities present in these cells only the higher affinity activity, at the time presumed to be H2 but now thought to be H4, was upregulated , showing an appropriate physiological response to low purine levels. Under purine-replete conditions, the high capacity, lower affinity H1 transporter is dominant [19,68]. A further factor in explaining the high number of purine transporter activities and genes may be in differential expression in the various stages of the Trypanosoma lifecycle. Some differences between procyclic and long-slender bloodstream forms were mentioned above. In addition, at least one of the purine transporters is expressed only in short-stumpy bloodstream forms .
Nucleoside transport in Leishmania species, particularly L. donovani, responsible for the most severe form of human leishmaniasis, has been as thoroughly studied as the equivalent transporters in T. b. brucei. The L. donovani nucleoside transporters were also the first protozoan purine or pyrimidine transporters cloned (see Section 3.2). However, studies on nucleobase transporter activities have hardly been reported until 2003 and almost all studies so far have dealt with the promastigote (insect) stage rather than the far more difficult to study amastigotes, which reside in phagolysosomes within mammalian macrophages.
Like other parasitic protozoa, Leishmania were found to be auxotrophic for purines, with adenosine and hypoxanthine being preferred for optimal growth. While adenine and inosine were being salvaged as well, they were first converted to hypoxanthine before incorporation into the nucleotide pool . In 1982, Hansen et al.  published a study of the transport of various radiolabelled purines in L. braziliensis panamensis and postulated three different ‘transport loci’: for inosine, for hypoxanthine/adenine and for adenosine. This is still the current model for purine transport by promastigotes. In addition, promastigotes, like procyclic T. b. brucei, are able to transport S-adenosylmethionine (AdoMet) , a metabolite used in polyamine synthesis and methylation reactions that can also be converted to adenosine and adenine .
2.2.2Identification of the L. donovani nucleoside transporters
Ullman and colleagues  created cell lines of L. donovani promastigotes resistant to either of the nucleoside antimetabolites tubercidin and Formycin B (analogues of adenosine and inosine, respectively). The resistance phenotype was associated with the loss of either inosine or adenosine and pyrimidine nucleoside uptake over 60 min. Uptake of the nucleobases hypoxanthine, adenine and uracil was not affected. This model was still consistent with the three loci model of Hansen and, using the mutant cell lines both nucleoside transporters were characterised with respect to kinetic parameters  and eventually cloned using functional complementation of the deficient phenotypes [25,26]. Two closely related genes were identified encoding adenosine/pyrimidine transporters, LdNT1.1 and LdNT1.2, with only LdNT1.1 expressed in promastigotes . Both transporters displayed submicromolar affinity for adenosine (Table 2) and somewhat lower affinity for uridine . LdNT1.1 is also believed to mediate thymidine and cytidine transport [9,16,80,81], whereas NT1.2 is highly specific for the oxopurine nucleosides, inosine and guanosine (Table 2), and is present as a single copy in the genome . The presence of two distinct adenosine transporter genes in the Leishmania genome, however, may explain the results of Ogbunude and colleagues [81,82] who found significant variations in adenosine transport activities, but not in inosine transport activity, between various strains of both L. donovani and Leishmania major. It would seem possible that expression of NT1.2 does occur in some strains or is induced during certain growth conditions. Under purine stress conditions, for instance, LdNT1-like activity, but not LdNT2, is upregulated 10-fold . Changes in apparent adenosine affinity in adenosine-depleted parasites, and different levels of upregulation for adenosine and uridine, suggest that perhaps a different transporter than LdNT1.1 is being expressed during purine stress. A similar pattern occurs in T. b. brucei (see Section 2.1.4).
Table 2. Overview of Leishmania purine and pyrimidine transporters
Nucleoside transport in L. donovani amastigotes, as in promastigotes, appears to be mediated by at least two transporters, T1 and T2 (Table 2). T1 was similar to the LdNT1 activity, in that it is a high-affinity adenosine transporter and sensitive to inhibition by pyrimidine nucleosides but not by inosine or guanosine. The T2 transport activity, unlike LdNT2, also transports adenosine in addition to the oxopurine nucleosides . It seems entirely possible that T1 is encoded by LdNT1.1 or LdNT1.2.
2.2.3Electrophysiology of nucleoside transport
The L. donovani nucleoside transporters expressed in Xenopus oocytes have also been studied with electrophysiological techniques . It was clearly shown that transport by all three cloned nucleoside transporters is electrogenic, in a proton-dependent, sodium-independent manner. However, nucleoside transport maybe only partially dependent on the proton-motive force as the study found no significant charge co-transport associated with adenosine uptake by LdNT1.1 and only a very minor adenosine:proton ratio with LdNT1.2 (0.0035 charges per molecule of adenosine transported). This ratio was higher when uridine was the substrate (0.13 and 0.45 charges per molecule transported by NT1.1 and NT1.2, respectively). This is in contrast to a 1:1 ratio for the Leishmania proton/myo-inositol symporter . However, Stein et al.  argue that the charge:substrate translocation ratios may be underestimated by the presence of a constitutive proton leak, blocked more efficiently by adenosine than by uridine. In any event, it is clear that the L. donovani and T. b. brucei[19,20,65] nucleoside and nucleobase transporters can exploit the large proton-motive force across their plasma membranes to transport purines and pyrimidines with high affinity and, if necessary, against a concentration gradient.
The nucleoside transporters, and probably the nucleobase transporters, are members of the Equilibrative Nucleoside Transporter family (see Section 3.2), which facilitates the equilibrative exchange of substrate across plasma membranes in vertebrates . Concentrative transport of nucleosides in mammals is dependent on the sodium rather than the proton gradient and mediated by the CNT family , whereas concentrative nucleobase transport is mediated by sodium-dependent transporters of an as yet unidentified gene family . Interestingly, the first nucleoside transporter cloned from a plant, Arabidopsis thaliana, also appears to be a proton symporter of the ENT family  as are CNT family members from Caenorhabditis elegans and Candida albicans.
2.2.4Nucleobase transport in Leishmania
The first systematic study of purine nucleobase transport in Leishmania spp. after the initial report by Hansen  showed a single high-affinity transporter with broad specificity for purine nucleobases in L. major promastigotes, designated LmaNBT1 . Highest affinity was for hypoxanthine, with a Km value of 0.71 μM, followed by guanine, adenine and xanthine (Table 2). Interestingly, the antileishmanial hypoxanthine analogue allopurinol was also taken up by LmaNBT1, with a Km value of 54.3 μM. Strong kinetic evidence, based on mutual inhibition profiles, was presented that LmaNBT1 was the only transport activity for [3H]adenine, [3H]hypoxanthine and [3H]allopurinol in these cells . The LmaNBT1 transport activity was very similar to the H2 activity in T. b. brucei (see Section 5.4). The TbU1  and LmU1 uracil transporters  were equally similar in their substrate selectivity profile, with high affinity for uracil only (Tables 1 and 2).
Recently, the group of Landfear reported the cloning and heterologous characterisation of an L. major transporter, LmaNT3, which displays high affinity for purine nucleobases only . Though the Km values obtained in Xenopus oocytes are rather higher than reported for LmaNBT1 , with a hypoxanthine Km of 16.1 μM (Table 2), it seems possible that this gene encodes LmaNBT1 or the equivalent transport activity in amastigotes. This group also reports the cloning of a further ENT member from L. major, LmaNT4, which is currently being characterised .
Crithidia are parasites of insects, and do not require intermediate (mammalian) hosts as do other kinetoplastids discussed here. Due to the close evolutionary relationship between these parasites and, in particular, Trypanosoma and Leishmania spp., and the ease of their culture in vitro, they have been used quite extensively as model organisms for the kinetoplastidae.
Nucleoside and nucleobase transport in the two Crithidia species, Crithidia fasciculata and Cr. luciliae, is very similar (Table 3). The first report of purine transport in this genus was in 1978 by Kidder et al. , reporting both high- and low-affinity uptake of each of the bases hypoxanthine, adenine and guanine (Table 3). They further claimed that their experiments showed that each of the bases is transported by a separate carrier, though this seems unlikely as the uptake of each base was shown to be inhibited by each of the others . Nucleobase uptake in Cr. luciliae has been studied in much greater detail and was shown to be similar to purine transporters of the other kinetoplastidae: high affinity and broad specificity for purine nucleobases and low if any affinity for the corresponding nucleosides and for pyrimidine nucleobases (Table 3) .
Table 3. Nucleoside and nucleobase transporters of Crithidia
Nucleoside transport in these species is very similar to the situation in L. donovani: an NT1 transporter with high affinity for adenosine and pyrimidine nucleosides, and NT2 with high affinity for inosine and guanosine (Table 3). The one exception to this consensus seems to be the reported inability of cytidine and uridine to inhibit either NT1 or NT2 of Cr. luciliae, but it was not clear at which concentrations . Overall, purine/pyrimidine transport in Crithidia does appear to be very similar to that of other kinetoplastids, particularly to Leishmania promastigotes, which share a similar environment. Very recently, the group of Buddy Ullman reported the cloning (see Section 3.5) and heterologous expression of two Cr. fasciculata nucleoside transporters, an NT1-type and the probable NT2 transporter .
Groundbreaking work has been performed, using Cr. luciliae and Cr. fasciculata, on the regulation of purine salvage in response to different environmental conditions, particularly purine availability [24,68,70,71,74]. This situation was later also reported for T. b. brucei procyclics  and Leishmania promastigotes . In Crithidia, it has been shown that purine starvation leads to a concerted response with upregulation of extracellular 3′-nucleotidase [70,72,93], purine nucleoside and nucleobase transporters [70,71,74], hypoxanthine–guanine phosphoribosyltransferase  and adenine deaminase – all proteins involved in the salvage of purines. This response seems to be specific since transport of amino acids and glucose in Cr. luciliae was unaffected by purine-starvation , as was uracil transport in T. b. brucei, while hypoxanthine and adenosine transport was strongly upregulated . Conversely, depletion of different essential nutrients, arginine and haemin, did not affect 3′-nucleotidase activity in Cr. luciliae. The upregulation of purine salvage enzymes appears to require new protein synthesis, as the response is sensitive to cycloheximide [24,68,70]. It would now appear that this is required for the de novo synthesis of regulatory proteins rather than the purine salvage enzymes themselves, as Liu et al.  did not detect significantly increased levels of CfNT2 mRNA during purine depletion, though NT2 activity was upregulated >10-fold and regulation must therefore be post-transcriptional. The regulation of purine salvage in Crithidia has been extensively reviewed elsewhere [73,74].
Plasmodium parasites, the etiological agents of malaria, are obligate intracellular parasites with a complicated lifecycle. During the course of infection, sporozoites first infect liver cells, where they multiply and re-emerge as metacryptozoites that infect erythrocytes. It is the cyclic development in the erythrocyte that is ultimately responsible for the clinical manifestations of malaria, and it is on the intraerythrocytic stages that most research and drug development has focussed.
2.4.1Purine and pyrimidine salvage by Plasmodium
It has long been known that Plasmodium species, like other protozoan parasites, are incapable of synthesising purines [94–96], but do synthesise pyrimidines de novo [95,97]. Unlike most other parasites they cannot rely on the host cell to provide these essential nutrients: erythrocytes equally lack purine and pyrimidine biosynthetic pathways and, being anucleate, have a limited requirement for nucleotides . Yet, the rapid development and division of the parasite is accompanied by a dramatic increase in purine and pyrimidine utilisation and it is has long been known that Plasmodium can utilise preformed purines for nucleic acid synthesis [2,98]. In contrast, Plasmodium spp. lack the ability to salvage preformed pyrimidines [98–100] except orotate . Consequently, inhibitors of pyrimidine biosynthesis such as pyrazofurin  and the hydroxynaphtoquinone BW58C have strong anti-Plasmodium effects . Pyrazofurin is an inhibitor of orotate phosphoribosyltransferase and orotidine-5′-phosphate decarboxylase and BW58C inhibits dihydroorate dehydrogenase.
Whereas the non-infected erythrocyte salvages nucleosides through the equilibrative hENT1 transporter  and purine nucleobases through the facilitative nucleobase transporter hFNT1 [39,104], the plasma membrane of the infected erythrocyte contains an additional uptake mechanism for nucleosides and many other nutrients, which is sensitive to furosemide [105–107]. While the nature of this unique New Permeation Pathway (NPP) is still a matter of intense debate, it is clear that its role is to import the nutrients into the erythrocyte cytosol rather than into the Plasmodium parasite. The porous nature of the membrane of the parasitophorous vacuole (PVM) in which the parasite resides [108,109] makes it likely that the parasite experiences the same purine concentrations as in the erythrocyte cytoplasm. An additional mechanism, however, is required for the transport of the nucleosides and/or nucleobases across the parasite plasma membrane.
2.4.2Central role for hypoxanthine uptake
To study purine transport by Plasmodium spp., parasites were released from the infected erythrocytes by treatment with saponin [1,110,111]. The first such report, using the avian malaria parasite P. lophurae, described rapid and saturable uptake of 14C-labelled adenosine, inosine, hypoxanthine and guanine . The apparent mutual inhibition of uptake of each of these purines led to the assumption of a single purine nucleoside/nucleobase transporter in this species, though it is noteworthy that purine uptake was not inhibited by adenine . They also did not exclude a separate transporter for oxopurines, as they reasoned that adenosine might inhibit an inosine transporter after deamination. Similar results were obtained by Manandhar and Van Dyke using Plasmodium berghei, who reported that adenosine is not only rapidly deaminated to inosine, but subsequently deribosylated to hypoxanthine prior to uptake by the free parasite . Hansen and colleagues  further confirmed this model by showing inhibition of [3H]adenosine uptake in saponin-freed P. berghei parasites with 2′-deoxycoformycin, a powerful adenosine deaminase inhibitor.
The consensus of the early studies of purine salvage in model malaria parasites was thus that purines are overwhelmingly taken up as hypoxanthine by Plasmodium parasites, as represented in Fig. 1(a). This view was entirely consistent with the observation that hypoxanthine is by far the preferred substrate for purine biosynthesis in Plasmodium. Indeed, it was shown that P. falciparum growth was up to 90% inhibited when xanthine oxidase was used to deplete infected erythrocytes of hypoxanthine .
2.4.3P. falciparum adenosine transporters
More recently, efforts have begun to clone and study purine transporters of P. falciparum, the agent responsible for the most serious forms of human malaria, by expression in oocytes of X. laevis. mRNA isolated from infected human erythrocytes and injected into oocytes induced the expression of transport activities for adenosine and hypoxanthine. Mutual inhibition again appeared to indicate a single transporter for both purines being expressed, but since Km and Ki values were not determined, this assumption remained unproven . A more thorough characterisation of a P. falciparum nucleoside transporter became possible with the almost simultaneous cloning of PfNT1 and PfENT by the groups of Ullman  and Baldwin , respectively. Despite both groups utilising the X. laevis expression system and the two genes differing by just a single amino acid (Phe or Leu at position 385), the conclusions of the two groups differed markedly. For instance, PfENT1 reportedly has a 25-fold lower affinity for adenosine compared with PfNT1. Furthermore, PfNT1 was reported to transport l-isomers of adenosine and thymidine whereas the P. falciparum adenosine transporter had previously been reported to be stereoselective [106,114]. Nor is there agreement about whether nucleobases are [31,114] or are not  substrates of the nucleoside transporter, or whether it is sensitive to inhibition by dipyridamole. Clearly, additional studies with heterologously expressed transporters and with isolated parasites will be required to reconcile these studies. It does appear certain, however, that Pf(E)NT is expressed during the intraerythrocytic stages of the lifecycle [30,31,115] and is present on the parasite plasma membrane rather than on the plasma membrane of the infected erythrocyte . This is consistent with Pf(E)NT's insensitivity to furosemide [30,114], which blocks nucleoside uptake into infected erythrocytes [106,107].
The presence of an adenosine transporter on the P. falciparum plasma membrane changed the model proposed in Fig. 1(a). It is also now clear that adenosine deaminase and inosine phosphorylase are present both inside and outside the parasite  (Fig. 1(b)). Both studies of the P. falciparum adenosine transporter report that intraerythrocytic stages of Plasmodium should be able to take up at least some pyrimidines, notably thymidine. The parasite may not, however, have the metabolic pathways to utilise them. For instance, a probable uracil phosphoribosyltransferase can readily be identified in the T. brucei genome (Tb04.2H8.180; http://www.genedb.org) but not in any Plasmodium genome available to date (http://www.plasmodb.org/). The enzyme activities of the pyrimidine salvage pathway have not been identified biochemically either [7,112,116].
The contribution of the adenosine transporter to purine salvage in Plasmodium species needs to be reassessed in the light of the earlier studies indicating a more important role for hypoxanthine transport and further studies of purine transport by Plasmodium spp. are urgently required. Purine transport across the normal and infected erythrocyte plasma membrane is discussed in detail elsewhere  and will not be addressed here.
2.5.1Source of purines for salvage
As in the related Apicomplexan species, P. falciparum, the questions of purine salvage have largely centred on what the preferred purine source is and how the intracellular parasite, within the parasitophorous vacuole, could have access to sufficient levels of it. Although there are many parallels between Plasmodium and Toxoplasma purine salvage, one major difference is the type of host cell the parasite inhabits. While Plasmodium infects the metabolically incomplete, anucleate erythrocyte (no purine biosynthesis), Toxoplasma invades nucleated cells with the entire metabolic machinery intact. Unlike the malaria parasite, Toxoplasma is not believed to induce the formation of additional permeation pathways in the host cell and therefore relies on free purines in the host cell cytoplasm, which diffuse freely through pores in the parasitophorous vacuole membrane . The purine transporters need to be able to salvage the presumably very low concentrations of free nucleosides and nucleobases, as the parasite is unable to take up nucleotides directly . Much speculation has centred on whether efficient purine salvage includes hydrolysis of host cell ATP to adenosine by a Toxoplasma-encoded nucleoside triphosphate hydrolase (NTPase) present in the parasitophorous vacuole [119–121]. The characterisation  and cloning  of a low affinity, high capacity adenosine transporter, TgNT1, seemed to fit well in this model of massive generation of adenosine from host ATP. However, it was subsequently shown that less than 5% of NTPase in the vacuole is active – and that activation leads to rapid depletion of cellular ATP and exit of the parasite from the host cell . Furthermore, the NTPase hydrolyses ATP only to 5′-AMP or ADP [119,124] and it was speculated that a T. gondii ecto-5′-nucleotidase would complete the hydrolysis to adenosine. In a comprehensive study, Ngô et al.  were unable to demonstrate the existence of such an enzyme activity in the parasitophorous vacuole. The conclusion of this study, that host cell nucleotides are not the major source of purines for T. gondii, and the earlier observation that the parasite can incorporate hypoxanthine, adenine, guanine, xanthine, guanosine and inosine as well as adenosine [125,126], triggered a reinvestigation of purine transport by T. gondii and high affinity transporters, one for hypoxanthine and one with broad specificity for nucleosides, have now been reported .
2.5.2Characterisation of nucleoside and nucleobase transporters
All purine and pyrimidine transport studies with Toxoplasma have been performed with isolated tachyzoites, the replicating stage infective to nucleated mammalian cells. The first comprehensive investigation of purine uptake in these cells was performed by Schwab et al. . They reported a low affinity, equilibrative adenosine transporter (apparent Km= 230 μM). These experiments were performed in an adenosine kinase-deficient strain, so as eliminate any influence of adenosine metabolism on the apparent rates of transport. Slightly unusual for a protozoan nucleoside transporter, adenosine transport was inhibited by the mammalian nucleoside transport inhibitor, dipyridamole, with an IC50 of 0.7 μM. In contrast, dipyridamole inhibited transport of inosine, hypoxanthine and adenine only marginally at 10 μM, suggesting the presence of additional transporters. This adenosine transporter, TgAT, was subsequently cloned by insertional mutagenesis and selection on a cytotoxic adenosine analogue, adenine arabinoside . TgAT was a member of the ENT family and mediated the uptake of adenosine when expressed in oocytes, with an apparent Km value of 114 ± 37 μM, and was ∼60% inhibited by 50-fold excess inosine and to a lesser extent by guanosine, hypoxanthine and guanine . While adenosine transport in the insertional mutant was much reduced, it retained salvage capabilities for inosine and purine nucleobases and the authors concluded that TgAT encodes the sole adenosine transporter of T. gondii. However, on studies with intact tachyzoites we found that both adenosine and inosine are transported by an additional high-affinity nucleoside transporter (Km values of 0.49, 0.12 and 0.77 ± 0.20 μM, respectively) . This transporter, TgAT2, also displayed high affinity for guanosine, uridine and thymidine. In the same study, we reported the characterisation of a high-affinity T. gondii hypoxanthine/guanine transporter, TgNBT1 (Km (hypoxanthine) = 0.91 ± 0.19 μM). Adenine uptake was not saturable up to 1 mM and apparently depended on simple diffusion . The presence of TgAT2 contradicts the conclusion of Chiang et al.  that TgAT must be the sole adenosine transporter in T. gondii, though there are several possible explanations for this apparent discrepancy. These include the possibility that both transporters are the product of the same gene (alternative splicing, post-translational modification, etc.) and the possibility that additional mutations occurred during selection of the insertional mutant with adenine arabinoside. Several additional ENT genes can readily be identified in the T. gondii genomic database (ToxoDB.org) and have been named TgNT1, TgNT2 and TgNT3 .
2.5.3Localisation of T. gondii adenosine transporters
A recent study by the group of El Kouni  reports that intracellular tachyzoites are able to transport β-l-adenosine, nitrobenzylthioinosine (NBMPR) as well as other β-l-purine nucleosides. This ability was much reduced in tachyzoites of the ΔTgAT strain. Surprisingly, Toxoplasma-infected human fibroblasts similarly acquired the ability to transport these non-natural nucleosides, leading to the conclusion that TgAT must be present both on the parasite plasma membrane and the host cell plasma membrane. This would be in contrast to the situation in Plasmodium-infected erythrocytes, where PfNT1 localises exclusively to the parasite plasma membrane and could not be demonstrated with immunoelectron microscopy on the host cell membrane .
2.5.4Pyrimidine salvage in T. gondii
Toxoplasma tachyzoites are capable of de novo pyrimidine biosynthesis and this seems a vital pathway for virulence . Disruption of this pathway by deletion of the gene encoding carbamoyl phosphate synthetase II created a uracil auxotrophic cell line, which was able to invade host cells but needed supplementation of >20 μM uracil in order to replicate. This study shows that T. gondii express a uracil transporter, but do not ordinarily depend on it. In addition, TgAT2 appears to transport pyrimidine nucleosides, as well purine nucleosides .
A limited number of studies have been performed characterising purine transporters in other protozoa. The aerotolerant intestinal parasite G. lamblia (also known as G. intestinalis) causes a common form of waterborne diarrhoea. G. lamblia trophozoites (the flagellate forms that infect the duodenum) reportedly express at least three distinct transporters for nucleosides and/or nucleobases. ‘Type 1’ is described as a thymidine transporter, with a Km of 50 μM . This transporter seems to be selective for oxopyrimidines as it was inhibited with similar affinity by uracil, uridine, thymine and 2′-deoxyuridine but not by the aminopyrimidines cytosine and cytidine, nor by ribose. The transporter displayed slightly higher affinity for the bases than for the nucleosides . However, the authors concluded in a later paper  that the thymidine carrier does not actually transport thymine, despite the relatively high affinity, since 50 μM [3H]thymine uptake was inhibited by only 12 ± 3% by 2 mM thymidine. However, since the Vmax of thymine on the nucleobase transporter (‘Type 3’) was very much higher than the [3H]thymidine Vmax on ‘Type 1’ (70 versus 498 pmol min−1 (106 cells)−1[15,130]) it seems possible that Type 1 does transport [3H]thymine but that its contribution is a minor one.
‘Type 2’ is described as a broad-specificity nucleoside transporter  with moderate affinity for 2′-deoxycytidine, adenosine, guanosine, uridine and thymidine (Table 4). Affinity for 2′-deoxy and 5′-deoxynucleosides was slightly less than for the ribonucleosides and affinity for 3′-deoxyadenosine was much lower than for adenosine , showing the importance of the 3′-hydroxyl group in substrate–transporter interactions and explaining the low affinity for nucleobases of this transporter. In contrast, ‘Type 3’ appears to be a genuine nucleobase transporter, of very low affinity but very high capacity, with Km values of 1.4 ± 0.1 and 1.6 ± 0.4 mM for adenine and thymine, respectively . All these studies were performed at low temperatures to prevent parasite attachment to vessel walls and facilitate a rapid sampling technique. A further study at 24 °C broadly supported the above findings, but found that nucleoside transport was sensitive to the inhibitor dipyridamole , while the earlier study had not . This study also found a much greater difference in affinity between adenosine and 2′-deoxyadenosine . Both groups came to the conclusion, through different observations, that nucleoside and nucleobase transport in Giardia is non-energy dependent [130,131]. In summary, the transporters of G. lamblia appear to have much lower substrate affinity, especially for purines, than do the transporters of the Apicomplexa and Trypanosomatidae, despite Giardia being purine auxotrophs . This probably reflects the very nutrient-rich environment of the small intestine.
Table 4. Kinetic parameters of nucleoside and nucleobase transporters of Giardia lamblia
Numbers in bold type are apparent Km values, in normal type Ki values or IC50 values, all in μM. NS, not saturable, uptake probably by passive diffusion.
115 ± 5
45 ± 25
93 ± 13
89 ± 8
26 ± 5
93 ± 13
The purine salvage pathways of T. vaginalis, the etiological agent of sexually transmitted trichomonial vaginitis, are unique, consisting of a purine nucleoside kinase [133,134] and a bacterial-type purine nucleoside phosphorylase , while lacking phosphoribosyltransferases [133,134]. This simplified purine salvage system should therefore be eminently susceptible to a rationally designed chemotherapeutic strategy  and the adenosine analogue Formycin A, which inhibits both the PNP and the PNK activities, potently inhibited in vitro growth of T. vaginalis. There has also been a report of potent in vitro anti-trichomonal action of allopurinol in combination with dipyridamole , despite the reported lack of a hypoxanthine phosphoribosyltransferase and the inability to incorporate hypoxanthine or inosine into purine nucleotides . Pyrimidine salvage could be an equally good target for therapeutic intervention as T. vaginalis is auxotrophic for both purines  and pyrimidines .
Despite the clear potential for purine or pyrimidine-based chemotherapy of this important pathogen, almost nothing is currently known about nucleoside and nucleobase transporters in this parasite. T. vaginalis can incorporate adenine and guanine but their transport has not been studied. Two nucleoside transporters have been reported, however, with apparent Km values between 2 and 13 μM for adenosine, guanosine and uridine. Thymidine, cytidine and inosine also inhibited adenosine transport with Ki values < 100 μM, indicating broad specificity for nucleosides, but no inhibition was observed with nucleobases .
Purine salvage in T. foetus is manifestly different from that in T. vaginalis, relying almost exclusively on salvage through a hypoxanthine–guanine–xanthine phosphoribosyltransferase . This makes the HGXPRT an excellent target for chemotherapy and rationally designed inhibitors have been designed, which inhibit T. foetus growth in the low micromolar range [6,142]. This reliance on phosphoribosylation is reflected in the observation that T. foetus expresses two nucleobase transporters (a hypoxanthine/guanine transporter and xanthine transporter) but appears to lack nucleoside transporters and adenine uptake appeared to be by passive diffusion only . Both transporters were low affinity, with Km values of 0.7 ± 0.3 mM (hypoxanthine), 90 ± 20 μM (guanine) and 0.6 ± 0.2 mM (xanthine). The absence of nucleoside transporters is to our knowledge unique for protozoa. Interestingly, Hedstrom and Wang  also found that a T. foetus strain resistant to the IMP dehydrogenase inhibitor mycophenolic acid had lost the hypoxanthine/guanine transporter, apparently relying solely on a low affinity xanthine transporter and passive adenine uptake for its purine requirements . The same strain had reduced adenine deaminase, further contributing to low intracellular hypoxanthine concentrations and thus aiding the incorporation of xanthine into the nucleotide pool, bypassing IMP dehydrogenase .
Babesia bovis, the parasite that causes babesiosis in cattle, infects host red blood cells and reportedly introduces a broad specificity nucleoside/nucleobase transporter in the erythrocyte membrane [144,145]. However, as the transport experiments were performed with infected bovine erythrocytes, it is unclear whether the transporter, which is sensitive to common nucleoside transport inhibitors such as NBMPR, dilazep and dipyridamole , is located on the red cell membrane. Alternatively, the transporter could be located on the parasite plasma membrane, with the erythrocyte membrane fully permeabilised to nucleosides by a NPP-like entity such as Plasmodium induces in human red cells (see Section 2.4). A number of the adenosine analogues were shown to be selectively toxic to intraerythrocytic B. bovis, which is clearly dependent on the parasite-induced permeability for nucleosides, as normal bovine erythrocytes do not appreciably take up purine or pyrimidine nucleosides .
Surprisingly little is known about nucleoside or nucleobase transport in T. cruzi, and no new studies have come out for over a decade. Finley et al.  described a tubercidin-resistant epimastigote clone deficient in uptake of both [3H]thymidine and [3H]tubercidin, but not [3H]adenosine or [3H]inosine, leading once more to a two nucleoside transporter model – though it is certainly surprising that tubercidin (7-deazadenosine) would be taken up by the thymidine transporter (TcNT1) rather than the proposed transporter(s) for adenosine and/or inosine (TcNT2). Nonetheless, the model was underpinned by the observation that [3H]tubercidin uptake was inhibited similarly by thymidine, cytidine and uridine, but was only partially inhibited by 500 μM adenosine and not at all by other purine nucleosides . One possible explanation would be that the TcNT1, like LdNT1, also recognizes adenosine, but in this case with low affinity, thus explaining the small effect of adenosine on [3H]tubercidin transport. In such a model, TcNT2 must be a genuine purine nucleoside transporter with high affinity for adenosine, inosine and guanosine, explaining that adenosine transport was not significantly reduced in the tubercidin-resistant clone. The same group later published an additional report on a different tubercidin-resistant epimastigote clone , which displayed reduced transport only of [3H]uridine and [3H]tubercidin, but not of [3H]thymidine or [3H]adenosine. Possibly this clone expressed a mutated pyrimidine nucleoside transporter with altered substrate specificity. Epimastigotes were also able to transport [14C]allopurinol riboside, [3H]formycin A and [3H]formycin B, whereas in trypomastigotes lower adenosine uptake rates correlated with reduced transport rates of formycin A, but not of formycin B or allopurinol riboside, which each inhibited growth rates of some epimastigote lines . Allopurinol has shown promising activity against T. cruzi in vitro and in vivo  and even against chronic Chagas' disease [151,152], halving the cardiopathy associated with this infection . Yet, though the metabolism of allopurinol in T. cruzi has been well studied , and the compound must clearly be taken up by the parasite, nucleobase transport in this parasite has yet to be investigated.
3Molecular biology of protozoan nucleoside and nucleobase transporters
The existence of multiple purine transporters has been demonstrated for a number of protozoan parasites by functional assay on whole cells (Section 2), but the burgeoning genome sequences for protozoan pathogens reveal that the complexity of available purine transport systems is, in many cases, still greater than anticipated from functional assays. Moreover, characterisation of individual transporter isoforms by heterologous expression has led to some surprises that have relevance to purine uptake in other eukaryotic systems. Genome sequences are now available for a number of protozoa, mainly those whose parasitic lifestyle renders them pathogens of clinical or veterinary importance (http://www.sanger.ac.uk/Projects/Protozoa/). Many of these organisms encode multiple purine transporters, and more candidates are emerging as annotation of newly sequenced genomes proceeds. Given that most protozoan parasites transit between different hosts and environments, the existence of multiple membrane transporter isoforms is to be expected, and developmental regulation of transporter expression will permit the parasite to adapt to changing nutrient availability. Moreover, the subcellular compartmentalization that is prominent in many protozoa may require expression of transporters that are adapted to function at intracellular locations.
All of the protozoan purine transporters that have been cloned to date are members of the Equilibrative Nucleoside Transporter (ENT) family, originally identified in mammals but with members throughout the higher eukaryotes (see  for a recent review of the ENT family). A second distinct group of nucleoside transporters, the Concentrative Nucleoside Transporter (CNT) family has members in both prokaryotes and eukaryotes, but no CNT-like gene has yet been identified from protozoa. The first ENT proteins to be characterised were found to be nucleoside transport facilitators and were named as such because the group was distinct from the CNT family. However, some protozoan ENTs are, in fact, concentrative rather than equilibrative. Moreover, some ENT family members are also nucleobase transporters [18,156] and recent data shows that protozoan ENTs include dedicated nucleobase transporters [21,32,33]. This observation explains the apparent lack of any previously identified nucleobase transporter families  in the complete or emerging protozoan genome sequences. The broad range of substrate specificities exhibited by ENT members in protozoa may also have relevance in mammals, where no dedicated nucleobase transporter has yet been cloned. Our homology searches of current protozoan genome data reveal that, while ENT homologues are ubiquitous, there is no evidence for homologues of non-ENT family nucleoside or base transporters. In the discussion of nucleoside and nucleobase transporter genes below we will focus on a relatively small number of protozoan species of which the genome sequence is (almost) complete.
The genome project for this organism is nearing completion but, as of August 2004 (release 3) (http://www.genedb.org/genedb/tryp/index.jsp), the largest chromosomes, IX, X and XI are still made up of multiple contigs, with gaps of unknown length between them. As a consequence, it is not yet clear how many genes of the ENT nucleoside/nucleobase transporter family are encoded in the genome, and in some cases temporary names must still be used, awaiting final assembly of chromosomes. For example, the chromosomal position of TbAT1, the gene encoding the P2 transport activity and the first cloned T. b. brucei purine transporter, is still unknown. Its closest homologue (Tb03.6N20.700) is only 66% identical and 80% similar to the published sequence of TbAT1 (AF152369 ).
The TbAT1 gene was first isolated by functional complementation of Saccharomyces cerevisiae naturally deficient in purine nucleoside uptake, using a cDNA library of T. b. brucei STIB427  and since from numerous other strains  and field isolates  as well as from T. equiperdum (M.L. Stewart and M.P. Barrett, unpublished) and T. evansi. It does not appear in the latest release of the T. b. brucei genome and it is therefore possible that multiple ENT family sequences are not yet listed (see also below). The sequences of the T. equiperdum and T. evansi AT1 genes are >99% identical to the original TbAT1 sequence.
3.1.1Purine nucleoside transporter genes
In contrast, the position of the NT2 gene, encoding a P1-type transporter, is very well defined. This gene is part of a cluster of 6 genes on chromosome 2, consisting of TbNT2, TbNT3, TbNT4, TbNT5, TbNT7 and TbNT6, in that order, with intergenic regions of ∼9 kb between the ORFs , each of which contains a copy of an ORF currently listed as a putative iron/ascorbate oxidoreductase family protein in GeneDB. NT2 was originally cloned from T. b. brucei strain EATRO 110 , which shares 447/463 (96%) identity at the amino acid level with the NT2 sequence of the 927 strain used for the genome project. Conservation within the NT2–NT7 cluster is also considerable (81–89% of amino acid sequence, relative to NT2). Despite this, two of the genes, TbNT3 and TbNT4, appear to encode proteins that either are non-functional or have a radically different function from the other 4 proteins. Whereas TbNT2, TbNT5, TbNT6 and TbNT7 appear to be high affinity P1-type adenosine/inosine transporters, no substrate was identified for TbNT3 and TbNT4 . It should be noted that TbNT5, TbNT6 and TbNT7 also displayed a modest ability to mediate hypoxanthine transport when expressed in X. laevis oocytes, with a Km value of 49 μM for TbNT5  (see Table 5).
Table 5. Overview of ENT family genes in the T. b. brucei genome
Additional T. brucei ENT genes, currently of unknown function, include NT11.1/AT-A (Tb09.244.2020) and NT11.2/AT-G (Tb09.218.0180) ( and De Koning, unpublished) as well as AT-E (Tb03.6N20.700) (De Koning, unpublished).
Recently, the group of Landfear  reported the cloning of another P1-like transporter, designated TbNT10, and showed that this was expressed exclusively in the non-dividing short-stumpy lifecycle form of the trypanosome. They demonstrated high-affinity uptake for adenosine, guanosine and inosine, but not for nucleobases, and these results are entirely in agreement with our own unpublished results with the same sequence, which we had provisionally named Adenosine Transporter-like B (AT-B). This gene is not part of the NT2 cluster and is located on chromosome 9. Another transporter, AT-D, is also a P1-type (De Koning et al., unpublished) and located on chromosome 6. In addition to mediating the transport of purine nucleosides it is capable of mediating hypoxanthine transport, but with very low affinity, as TbAT-D-mediated uptake of 1 μM [3H]hypoxanthine was not completely inhibited by 2 mM hypoxanthine but was fully inhibited by 2 mM adenosine (Fig. 2). These eight P1-type nucleoside transporter genes form a subgroup in a phylogenetic tree of all (known) T. brucei ENT family genes (Group II; Fig. 3). When ENT genes from additional protozoan species and the known human ENT genes are included, group II is shown also to include LdNT2 and CfNT2 (see below), with the other confirmed nucleoside transporters forming a separate group IV (Fig. 3).
3.1.2Nucleobase transporter genes
A third such group is formed by TbNBT1, the gene encoding the H4 purine nucleobase transporter , and a number of very closely related copies including the nucleobase transporter TbNT8.1 , which were cloned independently and concomitantly by our group and the Landfear group. Both genes, which differ in only three amino acids (see Table 6), were isolated from the ‘reference strain’ TREU 927. Both groups found evidence for multiple copies and Landfear et al.  sequenced an additional two copies of this gene, which were designated NT8.2 and NT8.3. Their Southern blots suggested the existence of a tandem repeat of related genes, similar to that described for the TbNT2 cluster. Considering the very high levels of sequence identify between the isoforms it is likely that each encodes a functional nucleobase transporter and it should be remembered that T. b. brucei displays at least four different purine nucleobase transporter activities and a least two distinct pyrimidine nucleobase transport activities (see Section 2.1). At the moment it is impossible to predict how many of the isoforms listed in Table 6 will be included in the final genome, and to what extent the polymorphisms listed have a functional relevance. Even with minimal differences in amino acid sequence, TbNBT1 expressed in S. cerevisiae appears to encode a higher affinity transporter than TbNT8.1 expressed in the same organism, particularly for inosine and guanosine, for which NT8.1 displayed no affinity [32,33].
Table 6. Polymorphisms of T. b. brucei nucleobase transporter genes
NBT1 and NT8.1 have been cloned and characterised in heterologous expression systems [21,32]. The NT8.2 and NT8.3 polymorphisms have been obtained by direct sequencing from T. b. brucei (strain TREU 927) genomic DNA .
Interestingly, the recently cloned nucleobase transporter LmaNT3 and the putative nucleobase transporter LmaNT4 (see Section 3.2) group together with the T. brucei nucleobase transporters (Group III, Fig. 3).
3.1.3TbAT1 and related genes: true nucleoside/nucleobase transporters?
TbAT1 was cloned by Mäser et al.  and was predicted to be a single copy gene on the basis of Southern blots. The construction of a TbAT1 null mutant by targeted gene deletion recently confirmed this, as the P2-type adenosine transport activity was not detectable in the Tbat1−/− strain .
As mentioned above, TbAT1 cannot as yet be located in the T. brucei genomic databases, but three related genes (see legend to Table 5), which we designated TbAT-A, E and G until a function can be assigned, have been identified. These genes have 58%, 66% and 58% identity with TbAT1 at amino acid level, respectively. With TbAT1, they form a fourth phylogenetic group of ENT family genes (Group I, Fig. 3). TbAT1 encodes a highly unusual transporter, P2, which has very high affinity for both adenine and adenosine, and transports both with similar efficiency (Table 7), but does not transport oxopurines or pyrimidines [18,53]. It could therefore be described as an aminopurine transporter rather than a nucleoside or nucleobase transporter. However, it also efficiently transports a number of non-purine trypanocides of clinical importance, including melaminophenyl arsenicals and diamidines, which share a structural recognition motif (see Section 4.2). While the Km value for adenosine and two diamidines (pentamidine and diminazene aceturate) are very similar, the maximal uptake rate (Vmax) for the trypanocides is much lower than for aminopurines (Table 7). One question now is the extent to which the related genes TbAT-A, E and G will prove to be aminopurine transporters or perhaps diamidine transporters. It has been well documented that trypanosomes express two pentamidine transport activities in addition to P2: the High Affinity Pentamidine Transporter (HAPT1) and the low affinity Pentamidine Transporter (LAPT1) [34,158–160]. Neither is inhibited by high concentrations of purines nor pyrimidines.
Table 7. Km and Vmax values for some purines and trypanocides in bloodstream forms of T. b. brucei
Vmax (pmol (107 cells)−1 s−1)
aVery similar values were published first by Carter and Fairlamb . For a consistent comparison of transport rates of adenosine, hypoxanthine and the diamidines, values were taken from studies by the same group.
bDe Koning et al., unpublished. Average of 3 experiments using 20 nM [3H]adenine in the presence of 100 μM hypoxanthine.
The first protozoan purine transporter gene to be characterised, LdNT1, was cloned  from a L. donovani cosmid library by functional complementation of a strain that had been rendered adenosine transport-deficient after selection with the cytotoxic adenosine analogue tubercidin. Two almost identical genes, LdNT1.1 and LdNT1.2, were identified by this approach, both clear homologues of the ENT genes that had recently been identified in mammals. Expression of LdNT1 in Xenopus oocytes and in adenosine transport-deficient L. donovani facilitated its characterisation as an adenosine/pyrimidine nucleoside transporter. A second nucleoside transporter, LdNT2, was cloned using a similar strategy, complementing an inosine/guanosine transport mutant. LdNT2 encodes a high-affinity inosine/guanosine transporter .
The recent completion of the first Leishmania genome project has brought to light homologues of the functionally identified transporters LdNT1 and LdNT2. A recent report  describes the functional characterisation of LmaNT3, an LdNT1 homologue that exhibits high affinity for a range of nucleobases. LmaNT3 has considerable identity with the trypanosome nucleobase transporters TbNBT1  and TbNT8.1 . Thus, in Leishmania as well as trypanosomes, ENT family members are, at least in part, responsible for nucleobase transport, in addition to nucleoside transport and it appears that ENTs that transport nucleobases with high-affinity cluster together in a phylogenetic tree of protozoan ENTs (Group III, Fig. 3). A further ENT homologue is apparent in the L. major genome (LmjF11.0550) and functional characterisation of this transporter, provisionally named LmaNT4, is reportedly in progress . This gene encodes a 550 amino acid protein, which appears to have a unique insertion of ∼180 amino acids. TMPRED, and our multiple alignments, suggest that this insertion is located in an intracellular hydrophilic loop between transmembrane domains 6 and 7 of an 11 TM topology (see Section 5.1). This uncharacterised transporter also aligns within the ‘nucleobase transporter domain’ of the phylogenetic tree (Fig. 3).
The genome sequence of P. falciparum was the first completed and has thus been the most extensively studied and most comprehensively annotated of the protozoan genomes. To date, a single P. falciparum ENT homologue has been identified. The gene was identified concurrently by two independent groups and named PfNT1  or PfENT1  (see Section 2.4). Functional characterisation by both these groups revealed a high-affinity nucleoside transporter of broad specificity, with moderate affinity for some nucleobases. The P. falciparum genome contains four different ENT genes (S.A. Baldwin, personal communication) and the characterisation of the additional transporters must be the highest priority for understanding purine salvage in this parasite.
The first T. gondii purine transporter, TgAT, was functionally identified by insertional mutagenesis, leading to resistance to a toxic adenosine analogue . Upon heterologous expression in Xenopus oocytes, TgAT was revealed as an adenosine transporter of moderate affinity. Inhibition studies suggested a broad specificity for nucleosides and nucleobases and the loss of adenosine transport capacity in the insertional mutant suggested that TgAT was the sole transporter responsible for adenosine uptake in T. gondii. However, studies of purine uptake by T. gondii tachyzoites suggested a multi-component uptake system  and several additional ENT genes can be readily identified in the now substantially complete T. gondii genomic database (ToxoDB.org). These genes have been named TgNT1, TgNT2 and TgNT3 .
Liu et al.  very recently cloned two Cr. fasciculata ENT transporters designated CfNT1 and CfNT2, by probing a Crithidia library with LdNT1 and LdNT2. While the two Crithidia genes were only 30% identical, they were 72% and 73% identical to their L. donovani counterparts, showing both functional and genetic conservation between the ENT transporters of these species. Southern blots revealed that, while CfNT2 exists as a single copy, multiple copies of CfNT1 could be detected. The observed pattern was not consistent with (only) a tandem array for CfNT1, and though it is not clear how many copies exist in the Cr. fasciculata genome, several more were identified and sequenced using PCR approaches . While the sequence of the Crithidia ENT genes may thus be closest to the Leishmania transporters, the organisation is much closer to that of T. brucei.
The translocation of purines and pyrimidines across the intracellular membranes that compartmentalise eukaryotic cells is poorly understood, and no data exist for any protozoa. Nevertheless, the high level of subcellular complexity exhibited by protozoa may partially account for the large number of structurally related transporters identified in the genomes of some parasites (notably that of T. brucei). Adenosine transport systems have been characterised in both mammalian lysosomes  and mitochondria , though the genes encoding the proteins that mediate these activities have not yet been identified. Transporters with discrete subcellular localisations are likely to be required for organellar purine/pyrimidine metabolism in protozoa as in other eukaryotes.
Mitochondria, for instance, require purines/pyrimidines for replication of, and transcription from, the mitochondrial genome. Rat testis mitochondria contain a high-affinity adenosine transporter distinct from the ATP/ADP exchanger . In the Apicomplexa, the membrane-delimited plastid may also require transporters for delivery of the purines and pyrimidines that are required for replication of the organellar genome . The glycosome, a modified microbody that appears to be unique to kinetoplastid parasites and which harbours some of the enzymes of the purine and pyrimidine metabolic pathways [164–168], is likely to require membrane transporters for the exchange of purines and pyrimidines from the cytosol.
However, the challenges of obtaining intact organelles in sufficient quantities for the study of transport processes have so far proved insurmountable. Genome data have the potential to provide clues to subcellular localisation, but there is as yet no understanding of the targeting signals that must act to direct membrane proteins to discrete intracellular membranes. The use of fluorescent or epitope tags has the potential to reveal the subcellular localisation of membrane transporters and such approaches have been exploited (see below) despite the valid concern that introduction of a tag may disrupt the normal targeting.
Thus, for several reasons, the mechanisms by which integral membrane proteins are targeted to discrete subcellular membrane domains are poorly understood. The limited number of protozoan purine/pyrimidine transporters that have been localised to date are found to be homogeneously distributed in surface membranes (LdNT1 isoforms: ; LdNT2: ; TbNT8.1: ; PfNT1: ). Indeed the only example of differential subcellular localisation of membrane transporters amongst the protozoa is the discrete localisation of hexose transporters to the distinct pellicular and flagellar surface membranes of kinetoplastid parasites. Despite extensive studies, the targeting signal is still poorly understood [171–173], as is the possible function of the transporter in the flagellar membrane. Nevertheless, it is possible that some of the many ENT genes in the genomes of some kinetoplastids may encode isoforms with flagellar localisation, where they may contribute to acquisition of purines and pyrimidines.
Whereas T. brucei encodes a large number of ENT family transporter genes, some of which may localise to distinct intracellular locations, other parasites (notably P. falciparum and L. major) possess very few ENT family genes, apparently encoding proteins localised to the parasite plasma membrane. Intracellular purine/pyrimidine transport in these species may thus be mediated by transporters from an as yet unidentified gene family.
Characterisation of purine uptake by protozoa has proceeded over several decades, giving rise to a large body of data on a diverse range of (mainly pathogenic) protozoa. Parallel approaches with prokaryotes and higher eukaryotes have revealed that purine transport systems are ubiquitous amongst both auxotrophs and heterotrophs. In the past decade, many genome projects have been initiated and a number are now complete or nearing completion. Concurrently, functional cloning approaches have identified a number of genes that encode purine transporters. These transporters fall into a variety of families [85–87] and extensive functional characterisation and gene annotation will be required before the data on purine transport in whole organisms can be reconciled to the genetic data.
To date only ENT family members have been identified in protozoa: both function- and homology-based screens have failed to reveal purine or pyrimidine transporters from other families. For those parasites with complete genome sequences, it therefore appears that any additional nucleoside or nucleobase transport activity must be encoded by transporters of novel gene families. Yet, no such transporters have been identified to date. Functional screens for purine transporters have given rise only to ENT-like genes [25–32] and extensive in silico searches and hybridisation library screens have failed to identify homologues of non-ENT nucleoside or nucleobase transporters. At the time of writing, it seems possible that purine uptake in protozoa may be entirely mediated by members of the ENT family. It is not yet clear, however, whether the existing complements of ENT family members in each parasite species studied are sufficient to account for all the transport activities identified in the corresponding organism. For instance, no dedicated pyrimidine transporter genes have been identified, even though from several protozoa (nearly) all ENT members have been cloned and characterised. Therefore, some of these transport activities may be mediated by novel protein families, particularly since the structure of the binding pocket of the L. major and T. brucei uracil transporters appears to be remarkably different from the ENT purine transporters characterised to date .
It seems reasonable to postulate a requirement for multiple purine transporters to sustain a parasite through its life cycle, and indeed a battery of purine transport activities has been described in procyclic and bloodstream form trypanosomes. However, it is at present difficult to reconcile the presence of perhaps as many as 16 distinct ENT family members in the T. brucei genome while P. falciparum and Toxoplasma gondii, organisms with a no less complex life cycle, can apparently satisfy their purine requirements with far fewer ENT genes. However, all protozoan genomes that have been sufficiently investigated for the presence of nucleoside transporter genes have now been found to contain several ENT family members.
The recent advances have produced an abundance of evidence for significant molecular and functional divergence in purine salvage pathways between protozoan parasites and their mammalian hosts. This raises the possibility of specific chemotherapeutic intervention in parasite purine acquisition.
4Protozoan nucleoside and nucleobase transporters in chemotherapy
4.1Transport or diffusion?
In mammalian systems, nucleoside drugs may act on intracellular or extracellular targets, an example of the latter being purine receptors. In protozoa, all current targets for nucleoside or nucleobase antimetabolites are intracellular, which means that the drug needs to cross the parasite plasma membrane, and possibly further organellar membranes. In addition, when targeting an intracellular parasite, the (pro)drug must cross the host cell plasma membrane and any parasitophorous vacuole membrane. These requirements are obviously different for various protozoan species and need to be taken into consideration when designing any antiprotozoal agents.
The ability of drugs such as nucleosides to diffuse across biomembranes depends on their lipophilicity, usually quantified as the octanol–water partition coefficient log P or clog P for calculated log P. Both lipophilic and hydrophilic drugs have advantages with respect to targeting parasites. An obvious advantage of very lipophilic drugs is that they will cross any membrane and reach intracellular parasites and organellar targets perhaps easier than do hydrophilic compounds. Furthermore, they may cross the blood–brain barrier and thus have activity against parasitic infections of the CNS, e.g. late-stage African trypanosomiasis or cerebral toxoplasmosis, though lipophilicity is by no means the only factor for CNS penetration. A potential drawback is the loss of specificity that results from selective accumulation by a parasite, rather than host cells. Drugs that enter cells by diffusion will also not be accumulated to free intracellular concentrations higher than the extracellular drug concentration (blood, cerebrospinal fluid, etc.). While accumulation is still possible when driven by intracellular modification or high affinity binding to a target, active transport can accumulate hydrophilic compounds to free intracellular levels very much higher than therapeutic plasma concentrations, locally reaching concentrations not achievable or tolerated in blood. Furthermore, lipophilic compounds are more likely to be substrates of ABC-transporters that confer multidrug resistance  and clear ‘foreign substrates’ from the CNS , though some nucleoside and nucleotide analogues have been shown to be substrates for the human MDR4 and MDR5 transporters [174–176]. Most purines and pyrimidines do not diffuse through biomembranes at appreciable rates, but some nucleoside antimetabolites do, including desciclovir , abacavir  and NA-42 (2-cyclopentylamino, N6-cyclopentyladenosine), an experimental compound with submicromolar in vitro activity against African trypanosomes .
4.2The TbAT1/P2 transporter of T. brucei
The best-researched example of a protozoan transporter involved in chemotherapy is undoubtedly the P2/TbAT1 aminopurine transporter of African trypanosomes. There is abundant evidence of its ability to transport melaminophenyl arsenicals and diamidines, essential agents in the treatment of African trypanosomiasis. Though this has been extensively reviewed elsewhere [52,63,180,181], it is appropriate to discuss some of the key experiments here.
Evidence for P2 involvement in the uptake of melaminophenyl arsenicals followed from the observation by Carter and Fairlamb  that only substrates of this transporter (adenosine and adenine) were able to abrogate lysis induced by melarsen oxide in vitro. These purines were unable to affect lysis induced by the highly lipophilic arsenical phenylarsine oxide, showing the effect to be on the level of uptake. Furthermore, a melarsoprol-resistant line, cRU15, was shown to have lost the P2 transport activity . The expression of TbAT1 in yeast conferred some sensitivity to melarsen oxide  and P2-mediated adenosine transport is strongly and competitively inhibited by all melaminophenyl arsenicals [18,28,53]. Perhaps the definitive evidence was presented using a T. b. brucei clone in which both alleles of TbAT1 had been deleted . In tbat1-null trypanosomes, the rapid in vitro lysis induced by melarsen oxide or cymelarsan no longer occurred, but a much slower, delayed lysis was still evident and, in contrast to the rapid phase of lysis, not sensitive to inhibition by adenosine. Lysis with phenylarsine oxide was unchanged in tbat1-null cells. Consistent with these observations, resistance to melaminophenyl arsenicals, both in vitro and in vivo, was minimal . These experiments show that, even though P2 is the main transporter for melaminophenyl arsenical uptake, a secondary entry system exists. The additional uptake mechanism was shown to be highly sensitive to pentamidine and propamidine, with 50% effective concentrations similar to the pentamidine Km and propamidine Ki values for the HAPT1 pentamidine transporter, respectively . The implication is that loss of P2 function may be a necessary but not sufficient condition for melaminophenyl arsenical resistance and that concomitant loss of HAPT1 would be required for high levels of resistance.
Pentamidine transport by P2 was first demonstrated by Carter et al. , and accumulation of pentamidine was reduced in cRU15 , the melarsoprol-resistant line without a functional P2 transporter . Yet, this clone was not resistant to pentamidine , a finding that was later confirmed with the tbat1 null trypanosomes . As with the arsenicals, P2 is the main conduit for pentamidine, but additional transporters are sufficient for almost normal efficacy of the drug [34,159,160]. HAPT1 and LAPT1 have not yet been cloned and their physiological substrate and function remain unknown. It is certain, however, that these T. brucei pentamidine transporters are not purine or pyrimidine transporters [158,159]. Nor are they choline transporters (De Koning, unpublished) like the pentamidine transporter in P. falciparum. The P2 transporter clearly mediated the high-affinity transport of [3H]pentamidine in bloodstream trypanosomes [64,158,159] and tbat1 null trypanosomes lacked adenosine-sensitive pentamidine transport . However, adenosine transport by TbAT1 expressed in S. cerevisiae appeared not to be inhibited by pentamidine, leading to speculation about cofactors or modifications necessary for diamidine transport . However, a reinvestigation , using the same yeast strain and vector, found that diamidines, and in particular pentamidine, bind heavily to the yeast cell surface, possibly with higher affinity than to TbAT1, leading to very high backgrounds during transport studies and precluding the measurement of true initial rates of transport. Even so, accumulation of pentamidine over 1 h is clearly much higher in yeast expressing TbAT1 than in control cells .
Evidence for P2-mediated transport of the related diamidine diminazene aceturate (DA; Berenil) mostly parallels that for pentamidine. DA has high affinity for P2 [51,53,57,64] and a DA-resistant T. equiperdum line was shown to have lost normal P2 function . The tbat1 null clone was highly resistant to DA  and lost saturable [3H]DA transport . Furthermore, [3H]DA was clearly transported by TbAT1 expressed in yeast . Recently Witola et al.  reported that RNAi silencing of AT1 in T. evansi also confers resistance to diminazene.
The Km values for [3H]DA and [3H]pentamidine transport by P2 are similar to the Km for adenosine (Table 7), though the Vmax values for the diamidines appear very much lower. The different translocation rates, despite similar affinity, indicate different transport efficiencies for these compounds, expressed as Vmax /Km.
In addition, P2 displays high affinity for the veterinary phenanthridine trypanocide isometamidium (structurally a hybrid of diminazene and homidium (ethidium), see Fig. 4) [28,52], but at present there is no evidence that P2 is involved in isometamidium uptake other than a small but significant reduction in isometamidium sensitivity in T. evansi with induced RNAi for P2 .
It is not currently believed that P2 is the major route of entry for isometamidium. Yet, the fact that purine nucleobases, nucleosides, diamidines and phenanthridines could all have very high and almost identical affinity for the same ENT transporter that is otherwise very selective (no interaction with oxopurines or pyrimidines, for example, at 1 mM) is highly unusual. The key to this lies in the architecture of the P2 binding pocket and hence in the way it interacts with potential substrates such as adenosine. The main P2-adenosine interactions were independently elucidated from the ability of various compounds to antagonise melarsen oxide-induced lysis of T. b. brucei or through inhibition of P2-mediated [3H]adenosine transport . The presence of an accessible amidine motif NR1=CR2–NH2, an aromatic ring, and possibly a N or O residue in the position corresponding to N9 of the purine ring, are essential for high-affinity binding and all P2 substrates share this motif (Fig. 4). In addition, a similar structural configuration seems to be essential for high affinity binding to P2.
4.3Nucleobase transporters of Leishmania and Trypanosoma spp.
The hypoxanthine analogue allopurinol is in clinical use against various forms of leishmaniasis [49,50,183–186] and is taken up by hypoxanthine transporters of L. major promastigotes and L. mexicana amastigotes (see Section 2.2). In both cases a single transporter was responsible for the uptake of the drug, which should raise some concern about possible drug resistance, as the hypoxanthine transporters are unlikely to be essential given the presence of two high affinity nucleoside transporters in amastigotes of L. donovani. It has been reported that allopurinol resistance can be readily induced through in vitro exposure of promastigotes . In contrast, T. b. brucei express multiple nucleobase transporters, both in procyclic and in long-slender bloodstream forms, each with (relatively) high affinity for allopurinol [19,20,32]. If, as expected, the speed at which drug resistance develops in these parasites were partially dependent on the number of individual transporter activities mediating uptake of the drug, onset of allopurinol resistance would be much delayed in trypanosomes. This was tested in procyclic T. b. brucei, where allopurinol transport was clearly mediated by both the H1 and H4 hypoxanthine transporters. As predicted, attempts to induce transporter-related allopurinol resistance were unsuccessful after >12 months of exposure to 3 mM allopurinol . Culture conditions, using inosine as sole purine source, had been chosen to preclude resistance arising through loss of enzymes of the purine salvage pathways. This experiment showed that avoidance of at least one major mechanism of resistance to purine drugs is feasible if the drug is accumulated through multiple transporters.
Another cytotoxic nucleobase that is actively accumulated by both T. b. brucei and L. major insect forms is 5-fluorouracil. In both organisms, otherwise very exclusive uracil transporters (TbU1 and LmU1) displayed high affinity for this drug, with Km values of 3.0 ± 0.8 and 0.66 ± 0.14 μM, respectively [11,40]. Fluorouracil proved to be more effective against L. major promastigotes in vitro than either allopurinol or aminopurinol, with an ED50 of 5.1 ± 1.6 μM (only 3-fold higher than for pentamidine in the same assay ) or 4.5 ± 0.6 μM against L. amazonensis promastigotes . Crucially, the uracil analogue also displayed promising activity using L. major-infected macrophages . Flow cytometric analysis unambiguously shows a massive reduction in parasite burden after treatment with just 10 μM 5-fluorouracil (Fig. 5).
A major advantage of the use of purine nucleobase analogues against trypanosomes or Leishmania is the fact that substrate recognition by their nucleobase transporters is very different from that of their human counterpart, the facilitative nucleobase transporter (hFNT1) . This allows efficient uptake of antimetabolites by the parasite transporters, while excluding them from normal host cells. While this would seem to be a disadvantage when combating intracellular parasites, purine uptake systems are generally altered in parasitised host cells. This usually results in wider access of purine antimetabolites to the intracellular parasite than might be expected from purine transport by the pre-invasion cell (see below). Examples of selective recognition of purine antimetabolites by protozoan nucleobase transporters are 3-deazaguanine, 6-thioguanine and 6-thiopurine  as well as a range of thieno-separated tricyclic nucleobase analogues . The tricyclic purines retained very high affinity for the T. b. brucei H2 transporter (Ki values <2 μM) and the hypoxanthine-like tricyclic, TRI-B-002 displayed approximately 100-fold higher affinity for the trypanosome transporter than for hFNT1 . It needs to be recognised, of course, that inhibition of a transporter does not necessarily equate transport of the inhibitor at any significant rate. However, some of these tricyclic nucleobases show appreciable trypanocidal activity in vitro , indicating that they are internalised.
4.4Nucleoside antimetabolites and transport
A rather large number of nucleoside analogues have been reported to possess antiprotozoal activity, and it is certainly not within the scope of this review to give an exhaustive listing. The major concern with taking nucleosides into human trials has always been the potential for toxic or teratogenic side-effects (see also Ref. ) and, without doubt, many nucleosides with antiprotozoal activity are also acutely toxic to their hosts. Selectivity, however, is possible on the grounds of: (1) selective metabolism by the parasite, (2) selection at the drug target level and (3) selective accumulation. Examples of all three mechanisms can be readily identified.
An example of the first is 9-deazainosine, which is converted to 9-deazaadenine nucleotides by Leishmania and trypanosomes but not by mammalian cells [192,193], due to differences in their adenylosuccinate synthetase and/or lyase specificities [192,194]. As pointed out by Bhattacharya et al. , the amination appears to be the key to the selective activity of 9-deazainosine as 9-deazaadenosine is acutely cytotoxic .
The second category includes inhibitors of glycolytic enzymes in T. b. brucei, in particular glyceraldehyde-3-phosphate dehydrogenase (GAPDH), as T. brucei spp. lack a functional citric acid cycle and are totally dependent on glycolysis for their energy metabolism . Large numbers of 2,N6-disubstituted and 2′-N6 disubstituted adenosine analogues have been synthesised and optimised for GAPDH inhibition by structure-based design [43,198], some of which displayed low micromolar activity against T. b. brucei, while being much less toxic to mammalian cell lines.
Most protozoa express nucleoside transporters with much higher apparent substrate affinities than (most) mammalian nucleoside transporters. One example is the T. b. brucei P1 transporter, which displays high affinity for a whole range of nucleoside drugs including Formycin A, Formycin B, 2-chloroadenosine, ribavirin ). In contrast, ribavirin, for example, enters human erythrocytes through the NBMPR-sensitive nucleoside transporter with a Km of ∼0.5 mM . Other examples include the uptake of tubercidin and cordycepin by TbAT1 , Formycin A and tubercidin by L. donovani NT1 [26,80] and T1 , Formycin B by LdNT2  and T2  and l-nucleosides by P. falciparum (see below).
4.5Use of protective nucleoside transporter inhibitors
Apart from antimetabolite uptake by parasite transporters on the basis of different permeant selectivity, a further approach to targeted uptake of antimetabolites is possible by specific inhibition of the host nucleoside transporters. This approach was extensively reviewed by El Kouni recently , and allows the use of inherently toxic nucleosides, as their uptake into mammalian cells will be blocked. Again, this approach has been abundantly validated, but no clinical strategies have resulted. Most protozoan nucleoside transporters have been shown to be insensitive to the traditional inhibitors of mammalian nucleoside transport, dipyridamole, dilazep and NBMPR (reviewed in ). Co-administration of, in particular, NBMPR together with nucleoside antimetabolites has been shown to protect the host from their toxic effects and be efficacious in the treatment of several parasitic infections including T. b. gambiense, Plasmodium spp. [201,202] and Schistosoma spp. [203–206]. NBMPR itself displayed no toxic effects in mice at 25 or even 100 mg/kg [206,207].
It might be anticipated that this approach would not work with intracellular parasites, as the transport inhibitor would prevent uptake of the nucleoside drug across the host cell plasma membrane. However, the expression of the New Permeation Pathway in Plasmodium-infected erythrocytes (see Section 2.4) and, apparently, of T. gondii-encoded transporters in infected human fibroblasts , ensure NBMPR-insensitive uptake even in parasitised cells. In fact, these infected-cell-specific nucleoside uptake pathways appear capable of NBMPR uptake, whereas this drug is not internalised in non-infected mammalian cells . Together with phosphorylation by the parasite adenosine kinase [191,208], this leads to selective toxicity of NBMPR against T. gondii[191,208,209] and P. falciparum. Therefore, NBMPR not only protects host cells from toxic nucleoside analogues, it also is selectively toxic to the parasite itself. Similar observations have been made with other mammalian nucleoside transport inhibitors, in particular dilazep , which is already in clinical use as a vasodilator.
The absolute dependence of Plasmodium on an external purine source has led to the identification of purine, and in particular adenosine, analogues as anti-malarials . Many of these compounds do not exhibit acceptable parasite selectivity. However, targeting may be achieved following the observation that non-physiologically relevant l-nucleosides are transported into the parasite but not into mammalian cells . In this study, both l-and d-adenosine were not only transported into, but also metabolised in, merozoite and intraerythrocytic parasites. Intriguingly, l-adenosine uptake was not observed in saponin-freed intraerythrocytic parasites, although it was in infected erythrocytes treated with sendai-virus . This permeation pathway also remains enigmatic at the molecular level due to conflicting reports regarding PfENT1/PfNT1 substrate specificity (see Section 2.4). Nevertheless, the apparent lack of stereospecificity displayed by the parasite's purine salvage pathways provides a strong rationale for drug design: only the parasitic enzymes are capable of metabolising the l-nucleosides . The selective entry and metabolism strongly suggests the chemotherapeutic use of l-enantiomers of toxic nucleosides, and initial analysis confirms the strength of this approach . For example, l-isocoformycin (a structural analogue of the potent adenosine deaminase (ADA) inhibitor 2′-deoxy-d-coformycin), has a Ki of 7 pM (∼90% inhibition) for parasite ADA, and yet the mammalian homologue neither binds to the drug in vitro nor is exposed to the drug in vivo . A second approach is the utilisation of l-nucleosides as ‘carrier’ molecules to deliver pro-drugs or established anti-malarial compounds , in a piggy-back approach as has been envisaged for substrates of the T. b. brucei P2 transporter [35–38].
4.7Possible resistance mechanisms
The central role of transporters in therapy based on nucleoside and nucleobase antimetabolites means that loss of transporter activity may cause resistance to such drugs . This principle has been utilised to advantage in the cloning strategies of the Leishmania[25,26] and Toxoplasma nucleoside transporters. The loss of the T. brucei TbAT1/P2 transporter has been linked to resistance to several key trypanocides (see Section 4.2). Analysis of TbAT1 alleles from isolates of patients that relapsed following melarsoprol treatment, using RFLP analysis , SSCP (single strand conformation polymorphism) and direct sequencing has revealed a remarkably small number of TbAT1 polymorphisms that differ in their drug transport abilities, with a common set of nine mutations found in various geographical locations and in both human infective subspecies . However, a number of relapse patients retained the wild-type TbAT1 gene suggesting that additional factors may be involved . As discussed in Section 4.3, it is possible to avoid, or at least delay, transporter-associated resistance if the drug is taken up by several distinct carriers.
Resistance can also be associated with increased extrusion of the drug from the parasite. Many, if not all protozoa, are also known to encode ATP-Binding Cassette proteins (ABC proteins) including P-glycoproteins [217–220] and some of these have been shown to be involved in (multi-)drug resistance or the efflux of xenobiotics. Non-exhaustive examples include EhPgp1 and EhPgp5 of Entamoeba histolytica, PGPA of Leishmania spp. , verapamil- and cyclosporin A-sensitive P-glycoproteins in T. gondii, T. b. brucei MRPA  and P. falciparum mdr1/pgh1 [224,225]. Such ABC transporters certainly contribute to resistance to clinically important drugs such as quinine and mefloquine (malaria) and antimonials (leishmaniasis), but in other cases it is their ability to mediate drug efflux when experimentally overexpressed that allocates them a potential role in drug resistance. However, to our knowledge, little information exists about the ability of protozoan ABC transporters to confer resistance to therapeutic nucleoside analogues as is the case with human MRP4 and MRP5 . Expression of L. major PGPA, for instance, resulted in 10-fold resistance to arsenite and trivalent antimonials but not to puromycin . However, Katakura et al.  recently reported that overexpression of LaMDR2 in L. amazonensis promastigotes leads to 2.5-fold resistance specifically to 5-fluorouracil, and not to the nucleosides Formycin B, puromycin and tubercidin, arsenite, cadmium or regular antileishmanials. For recent reviews of protozoan ABC transporters in drug resistance, see Refs. [220,227].
A further mechanism of reducing the accumulation of therapeutic purines was reported in Leishmania spp. [228–230]. In L. mexicana amazonensis selected for resistance to tubercidin or inosine dialdehyde, uptake of purine nucleosides and nucleobases was reduced . This was linked to an extrachromosomal DNA of approximately 55 kb in size, which was not present in sensitive parent strains or revertants . From this DNA a TOxic nucleoside Resistance (TOR) gene (GenBank Accession No. AF016581) was identified, which encodes a 478 amino acid protein that contains a possible (but unconfirmed) trans-membrane domain (http://www.ch.embnet.org/software/TMPRED_form.html) but displays similarities to mammalian transcription factors , suggesting a role in the regulation of gene expression. The TOR locus was independently isolated from an L. major cosmid library under tubercidin selection and shown to confer high levels of resistance for allopurinol and inosine dialdehyde when expressed in L. major or L. mexicana.
Finally, mutations of the target enzyme or the enzymes of purine salvage pathways could render these compounds ineffective, especially if the purine salvage pathways have high levels of redundancy and the activation of the prodrug does not. An L. donovani line deficient in adenosine kinase was highly resistant to the adenosine analogues tubercidin and Formycin A but not to the inosine analogues Formycin B and allopurinol riboside or to 6-thioguanosine . Similarly, insertional mutagenesis of T. gondii followed by selection on the toxic nucleoside adenine arabinoside led to the cloning of the Toxoplasma adenosine kinase .
5Structure–activity relationships of purine transporters
The identification of transporter genes and gene families is finally opening the way for structural studies of both mammalian and protozoan ENT-family transporters. Algorithms for the prediction of transmembrane (TM) helices commonly predict 9–11 such domains. To our knowledge, the topology of only one ENT transporter has been experimentally verified . By using native and engineered N-glycosylation sites in combination with immunological approaches, the human ENT1 transporter was shown to contain 11 TM domains, with an intracellular amino terminus and extracellular carboxyl terminus. Hydrophilic loops between the helices were short except for a relatively large glycosylated loop between TM1 and 2, and an even larger intracellular loop between TM6 and TM7 . Extensive analysis of the predicted amino acid sequences of many eukaryote ENT family members has shown that this basic structure is likely to be highly conserved [233,234], though it should be emphasised that individual models do need to be experimentally confirmed. The presence of a large extracellular loop is less conserved than the central intracellular loop . The structural predictions of protozoan ENT transporters  have been entirely consistent with this general model, originally based on extensive analysis of the human, Drosophila melanogaster and C. elegans ENT members [233,234]. The topology of one protozoan ENT transporter, TbAT1, is depicted in Fig. 6(a).
5.2Domains involved in substrate recognition
The strong topological conservation makes it probable that the basic folding of the ENT transporters is likewise mostly retained and that domains identified in one transporter to be part of the substrate translocation channel, or involved in substrate binding, could be relevant to ENT transporters in general. This seems to be borne out by the identification of four conserved regions in an alignment of >30 ENT family genes from many different species , consisting of TM1 and adjacent amino acids (I), TM4/5 (II), TM8 (III) and TM9/10 including the last intracellular loop (IV). Very limited data exists on the protein folding or the function of each domain. Experiments using chimeras of rat ENT1 and ENT2 identified a fragment, stretching from between TM4 and TM5 to the end of TM6, associated with substrate selectivity  and it seems likely that TM5 and/or TM6 should therefore be part of the substrate translocation channel. This study followed an earlier report from the same groups, using chimeric constructs of rENT1 and hENT1, which identified a larger fragment, incorporating TM3-6, as involved in the recognition of the competitive nucleoside inhibitors dipyridamole and dilazep . Furthermore, SenGupta et al.  identified two conserved glycine residues in TM5, at positions 179 and 184 of hENT (Gly 161 and 166 of TbAT1 (Fig. 6)), and determined that G184 may have a critical structural function, since its replacement led to poor targeting to the plasma membrane as well as complete loss of function. Substitution of G179, on the other hand, which is completely conserved in an alignment of all known protozoan and human ENT transporters, lead to an almost complete loss of transport activity while retaining proper membrane localization . Yao and colleagues  also identified a cysteine residue in rENT2 in the outer half of TM4 that was accessible to the thiol reactive agent PCMBS, yet protected by the presence of uridine, a permeant for this transporter. All these studies strongly suggest that the region of TM4-6 forms part of the substrate translocation channel.
The importance of this domain has also been evident from observations with the Leishmania nucleoside transporters and TbAT1, lending further support for substantial structural conservation. Vasudevan et al.  isolated two different dysfunctional LdNT1.1 alleles from the tubercidin-resistant line TUBA5. One of these contained a single mutation, C337Y, in TM7 that conferred reduced Vmax. TbAT1 from drug-resistant field isolates contains a deletion mutation of a highly conserved phenylalanine in TM7 , supporting a role for TM7 in translocation of substrate. The other allele displayed a single point mutation to G183, corresponding to G184 of hENT1 and G166 of TbAT1. The mutation, G183D, is located in predicted transmembrane region 5 and led to drastically diminished Vmax. Intriguingly, the more conservative substitution of G183A, by site directed mutagenesis, led to a mutant LdNT1.1 protein that retained the ability to transport adenosine but no longer accepted uridine, strongly suggesting that this glycine residue plays a critical role in substrate recognition. This observation stimulated further analysis of the role of TM5 in LdNT1.1 function  by the substituted cysteine accessibility method, a scanning mutagenesis approach that permits identification of residues that are exposed to substrate . This recent study indicated that 3 of the 4 glycines in TM5, including G183, were required for transport activity. These glycines are all highly conserved within the ENT family and may play a role in helix packing [241,242]. Those cysteine mutants that were accessible to chemical modification were arranged on the same helical face as those with hydrophilic side chains and most of these residues were protected from modification by the presence of adenosine. These data indicate that TM5 is not only critical for transport activity but also contributes to the pathway of substrate translocation.
5.3A preliminary model for helix packing of TbAT1
To reveal residues and domains that may have a key function in substrate recognition by the T. brucei transporter TbAT1, we performed helical wheel analyses for each predicted hydrophobic transmembrane domain. Polar residues were highlighted, revealing that helices 5, 6 and 8 have substantial amphipathic faces (defined here as more than two stacked polar residues). The polar residues aspartic acid, asparagine and arginine are all stacked on one face of TM8 and are highly conserved across the ENT family. The hydrophilic face of TM6 is also highly conserved and a body of evidence (see above) implicates TM5 in substrate recognition. TM10 is one of the less well-conserved transmembrane domains and is generally amphipathic. Considering that hydrophobic domains are likely to be shielded from water and that the limited length of several of the predicted hydrophilic loops will impose packing constraints in the tertiary structure, we generated a model of transmembrane helix packing that defines a putative substrate permeation pathway for TbAT1 (Fig. 6(b)). In this model, the substrate permeation pathway is defined by helices 4–8, with the most significant contribution from the hydrophilic faces of TM5, TM6 and TM8. The majority of predicted TM helices in TbAT1 have substantial hydrophobic faces and these are arranged in the model so that they are oriented toward the lipid environment. TM10 is amphipathic, is rich in residues and motifs that are commonly involved in helix packing (glycine, alanine and threonine) and we have thus buried this helix in the core region. Relatively short hydrophilic loops connect TM9/10 and TM2/3. Therefore these helices have been placed adjacent to each other. This arrangement gives rise to an asymmetric positioning of the substrate permeation pathway, but makes no prediction regarding the structure or functional role of the larger hydrophilic loops that are likely to be exposed at the membrane surface.
Comprehensive alignments of ENT proteins reveal limited but significant sequence identity, and suggest a conserved topology. While very preliminary, the model is consistent with the experimental data that has been obtained with trypanosomal, leishmanial and mammalian ENTs reviewed in the previous section. Polar residues in helices that are predicted to delimit the substrate permeation pathway are good candidates for site-directed mutagenesis in a reverse genetics analysis, and this working model could direct mutagenesis studies and aid the interpretation of mutations that are revealed by forward genetic analysis.
5.4Models for transporter–permeant interactions
An additional approach to studying interactions between the transporter and the permeant is by studying the binding affinity of various (potential) substrates, determined as Km, the concentration at which 50% of transporters is occupied, for a range of structurally modified analogues. For instance, the P1 transporter of T. b. brucei bloodstream forms displays a Km of 0.41 ± 0.08 μM (n= 4) for [3H]adenosine but only 6.5 ± 1.8 μM (n= 3) for [3H]tubercidin (De Koning et al., unpublished). Since the only difference between adenosine and tubercidin (7-deazaadenosine) is a =N– or a =CH– at position 7 of the purine ring, it is reasonable to suppose that a substantial loss of binding energy would be associated with the concurrent loss of a strong H-bond acceptor. The Gibbs free energy of binding can be estimated from a derivation of the Nernst equation, ΔG0=−RTln(Km) , and the difference in binding energy between adenosine and tubercidin, δ(ΔG0), thus gives a value in kJ/mol for the H-bond between N7 of adenosine and one or more amino acid residues in the substrate translocation pathway of the P1 transporter (7.2 kJ/mol in the current example). Using this approach carefully with a sufficient number of well-chosen analogues allows the construction of a model for all the interactions between substrate and transporter, which complements the mutagenesis approach in understanding transporter function.
However, the determination of Km values requires radiolabelled substrates of high specific activity, which are almost never available for all the various structural analogues required for such a study. This limitation requires us to use the inhibition constant Ki instead, calculated from the Cheng–Prusoff equation: Ki= IC50/[1 + (L/Km)], in which IC50 is the inhibitor concentration causing 50% inhibition of the transport of a permeant, at permeant concentration L, as an approximation. This equation is only valid for competitive inhibition, which must therefore be verified, and the Ki value is not necessarily equal to the Km, particularly if the rates of translocation of permeant and inhibitor are substantially different . We recently reported one such example, where the Km of [3H]diminazene for TbAT1 was determined at 0.45 ± 0.11 μM , whereas the Ki had previously been determined at 2.4 ± 0.5 μM in T. b. brucei and 3.9 μM in T. equiperdum– a discrepancy likely to be the result of the very different translocation rates of adenosine and diminazene (see Table 7). This complication could lead to an overestimate of bond energy when using the Ki rather than the Km value in the Nernst equation given above.
While accepting the above limitations, semi-quantitative models for transporter–permeant interactions can be constructed using transport kinetics and structural analogues [8,22,39,40,53,245,246] and we have shown that such models have predictive value for Ki values of potential permeants [8,39,40], which is potentially of great importance for structure-based design of chemotherapeutics. We have already mentioned the structural motif recognised by TbAT1/P2 (Section 4.2; Fig. 4). Such a motif, the haptophore, can be coupled to an active substance, the toxophore, to improve translocation rates and/or specificity [35–38]. Moreover, such motifs can highlight functional conservation, where an amino acid sequence would at most show probable conservation of topology and conserved domains. For instance, we have characterised the main purine nucleobase transporters of T. b. brucei (H2) and L. major (LmaNBT1) in great detail and found that both interact in a very similar way with natural purine nucleobases . Fig. 7 shows that TbH2 and LmaNBT1 appear to make similar hydrogen bonds with hypoxanthine, through N(1)H, N3, N7 and N(9)H. However, the two protozoan nucleobase transporters form very different interactions with oxopurines than the human FNT1 nucleobase transporter, except for a similar H-bond at N3 (Fig. 7) , leading to different substrate selectivity such as a 50-fold higher affinity for 3-deazaguanine at the protozoan transporters [22,39]. The genes encoding LmaNBT1 and TbH2 have not been identified with certainty, but it is highly likely that both will be ENT family members and that identity at amino acid level will be around 50% (as it is between TbNBT1 and LmNT3, both very similar hypoxanthine transporters to H2 and LmaNBT1, respectively [21,32]). This is considered a high level of conservation, yet a single amino acid change can completely change a transporter's function or selectivity (see previous section) and it must be concluded that at the current state of knowledge primary sequence data is a poor guide to transporter function, beyond a rough classification as ‘probable nucleoside or nucleobase transporter’. The substrate binding model depicted in Fig. 7, however, does have predictive qualities and can be constructed even in the absence of genetic data. LmaNBT1 and TbH2 were termed ‘functional homologues’ to distinguish them from genetic homologues .
In the last decade, our understanding of protozoan nucleoside and nucleobase transporters has increased tremendously. We have moved from establishing the presence and number of such transporters in various organisms through characterisation of their kinetic properties to characterisation of their genes and finally towards understanding their structure and mechanism of action. We thus move that much closer to using this knowledge to selectively target drugs to these pathogens, or, in isolated cases, validating the transporters themselves as drug targets. The study of protozoan transporters is thus finally completing its voyage from biology to therapy. The next decade should see the fulfilment of the promise for a purine- or pyrimidine-based therapy for at least some protozoan infections.