A Toxoplasma gondii Oxopurine Transporter Binds Nucleobases and Nucleosides Using Different Binding Modes

Toxoplasma gondii is unable to synthesize purines de novo, instead salvages them from its environment, inside the host cell, for which they need high affinity carriers. Here, we report the expression of a T. gondii Equilibrative Nucleoside Transporter, Tg244440, in a Trypanosoma brucei strain from which nucleobase transporters have been deleted. Tg244440 transported hypoxanthine and guanine with similar affinity (Km ~1 µM), while inosine and guanosine displayed Ki values of 4.05 and 3.30 µM, respectively. Low affinity was observed for adenosine, adenine, and pyrimidines, classifying Tg244440 as a high affinity oxopurine transporter. Purine analogues were used to probe the substrate-transporter binding interactions, culminating in quantitative models showing different binding modes for oxopurine bases, oxopurine nucleosides, and adenosine. Hypoxanthine and guanine interacted through protonated N1 and N9, and through unprotonated N3 and N7 of the purine ring, whereas inosine and guanosine mostly employed the ribose hydroxy groups for binding, in addition to N1H of the nucleobase. Conversely, the ribose moiety of adenosine barely made any contribution to binding. Tg244440 is the first gene identified to encode a high affinity oxopurine transporter in T. gondii and, to the best of our knowledge, the first purine transporter to employ different binding modes for nucleosides and nucleobases.


Introduction
Toxoplasma gondii is probably one of the most successful parasites in the world and it is thought to be potentially infective for all warm-blooded land and sea animals. The infection occurs predominantly by ingestion of: meat containing tissue cysts harboring bradyzoites; milk containing tachyzoites; or water or sward contaminated with sporulated oocysts. After ingestion, the parasites penetrate the intestinal tissue, differentiate into tachyzoites and multiply intracellularly by endodyogeny before disseminating to other tissues [1,2]. Infection by T. gondii can be lethal for some species of marsupials and sea mammals [3], but in most hosts it is effectively controlled by the immune response and the presence of the parasite is restrained to latent cysts formed in muscle and nervous tissues [1,2,4,5].
Toxoplasmosis highly impacts livestock production by causing abortions in sheep, goats, and pigs, as well as causing diarrhea and anorexia in pigs, which causes the loss of transporter that recognizes several nucleoside analogues, and is the only T. gondii purine transporter for which a substrate binding model has been reported, allowing the rational optimization of analogues with high transport efficiency on this carrier [29].
It is interesting to notice, thus, that although TgAT2 has higher affinity for adenosine than TgAT1 and seems to have broader substrate selectivity, resistance to Ara-A was acquired after loss of TgAT1 [28]. Therefore, we decided to investigate the mode of interaction of TgAT1/Tg244440 with its substrates. To do so, we used a gain-of-function approach utilizing T. brucei as the expression system, as previously done by us for ENT family transporters of T. congolense, T. cruzi, Leishmania spp. and Trichomonas vaginalis [30][31][32][33]. We found Tg244440 to have high affinity for oxopurine nucleobases and nucleosides and only low affinity for aminopurines and pyrimidines. The very high affinity for both oxopurine nucleosides and nucleobases is highly unusual and the binding modes were carefully investigated using a series of selected matched pair analogues. We conclude that the nucleosides and nucleobases bind in different orientations in the Tg244440 binding pocket.

Equilibrative Nucleoside Transporters from Coccidian Parasites form a Separate Phylogenetic Cluster within the Family
We have previously shown that it was possible to predict the substrate selectivity of Trypanosoma cruzi ENT transporters by analysing their phylogenetic alignment with ENTs from other kinetoplastids [32]. Figure 1A shows a phylogenetic tree of nucleoside and nucleobase transporters from a number of different gene families. Transporters belonging to Nucleoside-Cation Symporter 1 (NCS1) and Nucleobase-Ascorbate Transporter (NAT, also known as NCS2), both belonging to the Amino acid-Polyamine-Organocation (APC) superfamily were positioned in a separate branch, but forming two distinct clusters. Members of the Concentrative Nucleoside Transporter (CNT) family formed a separate branch close to the NCS1 transporters as did AzgA-like transporters, although recent analyses have proposed AzgA-like transporters are associated with the NAT family [34,35]. It is probable that the higher number of ENT sequences used for the construction of the phylogenetic tree caused the positioning of AzgA-like and CNT transporters close to the NCS1 cluster.
The ENT sequences formed a consistent branch, with separations between the sequences from trypanosomatids (T. brucei and Leishmania spp.), humans and apicomplexans ( Figure 1B). Among trypanosomatids, which are the best studied protozoans in terms of purine and pyrimidine transport, there was a clear separation between purine nucleoside (T. brucei P1 and P2 and Leishmania NT2 clusters) and nucleobase (T. brucei NT8 and Leishmania NT3 and NT4 clusters) transporters, with the purine/pyrimidine nucleoside Leishmania NT1 transporters forming an outside group within this branch ( Figure 1B). This consistent grouping of transporter sequences with their known activities indicates a high level of predictability in the tree. The apicomplexan ENTs formed a separate group within the ENT family, with a separation between sequences from P. falciparum and coccidian apicomplexans (T. gondii, Hammondia hammondi, Cystoisospora suis and Sarcocystis neurona), indicating apparent gene proliferation from a common ancestral ENT gene after divergence of Plasmodium and coccidian species, but T. gondii sequences ( Figure 1B; black stars) always clustered very closely to their homologues in H. hammondi, its closest relative [36]. Moreover, our phylogenetic analysis indicates that Tg244440 (and its homologue in H. hammondi) was the most distant of the ENT sequences found in coccidia ( Figure 1A; red star), being at the edge of the ENT clade. We also observed an extraordinary conservation of the Tg244440 protein sequence among T. gondii strains: of the sequences retrieved from 15 strains, only one amino acid change (Leu → Phe in position 304) in Tg244440 was observed and only in strain MAS (Supplementary Figure S1). This high a level of conservation was not observed for any other T. gondii ENT transporter. Full-length Tg288540 (696 a.a.) was present in 14/15 of these strains, with the allele in strain TgRUB lacking the N-terminal 272 a.a.; Tg233230 (531 a.a.) was found in all 15 strains and displayed four amino acid changes; Tg359630 was found in only 11/15 strains available at ToxoDB, and displayed six amino acid changes  Table S4. * highlights the position of the T. gondii ENT transporters.
The ENT sequences formed a consistent branch, with separations between the sequences from trypanosomatids (T. brucei and Leishmania spp.), humans and apicomplexans ( Figure 1B). Among trypanosomatids, which are the best studied protozoans in terms of purine and pyrimidine transport, there was a clear separation between purine nucleoside (T. brucei P1 and P2 and Leishmania NT2 clusters) and nucleobase (T. brucei NT8 and Leishmania NT3 and NT4 clusters) transporters, with the purine/pyrimidine nucleoside Leishmania NT1 transporters forming an outside group within this branch ( Figure 1B). This consistent grouping of transporter sequences with their known activities indicates a high level of predictability in the tree. The apicomplexan ENTs formed a separate group within the ENT family, with a separation between sequences from P. falciparum and coccidian apicomplexans (T. gondii, Hammondia hammondi, Cystoisospora suis and Sarcocystis neurona), indicating apparent gene proliferation from a common ancestral ENT gene after divergence of Plasmodium and coccidian species, but T. gondii sequences ( Figure 1B; black stars) always clustered very closely to their homologues in H. hammondi, its closest relative [36]. Moreover, our phylogenetic analysis indicates that Tg244440 (and its homologue in H. hammondi) was the most distant of the ENT sequences found in coccidia ( Figure 1A; red star), being at the edge of the ENT clade. We also observed an extraordinary conservation of the Tg244440 protein sequence among T. gondii strains: of the sequences retrieved from 15 strains, only one amino acid change (Leu  Phe in position 304) in Tg244440 was observed and only in strain MAS (Supplementary Figure 1). This high a level of conservation was not observed for any other T. gondii ENT transporter. Full-length Tg288540 (696 a.a.) was present in 14/15 of these strains, with the allele in strain TgRUB lacking the N-terminal 272 a.a.; Tg233230 (531 a.a.) was found in all 15  Supplemental Table S4. * highlights the position of the T. gondii ENT transporters.

Tg244440 Is a High-Affinity Oxopurine Transporter
To understand the role of Tg244440 in purine transport, we transfected this gene into procyclic forms of the T. brucei TbNB-KO strain that has a null background for the uptake of guanine (and a much-reduced background for hypoxanthine uptake), through the knockout of the locus containing the three nucleobase transporters Tb927.11.3610, Tb927.11.3620, Tb927.11.3630 [32,36]. T. brucei also have a near-null background for thymidine uptake at low concentrations [37,38]. Three clones transfected with Tg244440 (B11, F5 and G10) were randomly picked and used for a screen of substrate transport ability, using 0.  (Figure 2). All three TbNB-KO + Tg244440 clones showed highly significant increases in the levels of guanine and hypoxanthine uptake whereas uptake of thymidine was only slightly increased, and only in clone F5. No difference in the transport of adenosine was observed but TbNBT-KO procyclics retain the high-affinity adenosine/inosine/guanosine P1 transporters [39], which gives a high background in the assays using radiolabelled adenosine, as evidenced by the high uptake rate for adenosine compared to the other [ 3 H]-permeants ( Figure 2). Clone F5 was selected for all further experimentation. adenosine was observed but TbNBT-KO procyclics retain the high-affinity adenosine/inosine/guanosine P1 transporters [39], which gives a high background in the assays using radiolabelled adenosine, as evidenced by the high uptake rate for adenosine compared to the other [ 3 H]-permeants ( Figure 2). Clone F5 was selected for all further experimentation.  Figure  3A). In the presence of 25 μM unlabelled guanine there was no significant uptake of the  Figure 3A). In the presence of 25 µM unlabelled guanine there was no significant uptake of the label. In the control cells without the Tg244440 transporter (empty vector, EV), the uptake over 60 s was also not significantly non-zero (Table 1). Hypoxanthine uptake was very similar to that of guanine: linear over 60 s, saturable, and with a significantly non-zero rate ( Table 1). The slope of the EV control was also nonzero, as previously reported [32] but 86.2% lower than the +Tg244440 clone F5 (p < 0.0001) ( Figure 3B). We conclude that Tg244440 efficiently transports guanine and hypoxanthine but at best marginally transports thymidine, while no conclusion on adenosine uptake could be reached.
We thus proceeded to determine the affinity of Tg244440 for hypoxanthine and guanine, and found it to have equally high affinity for both oxopurine nucleobases, with K m values of 0.79 ± 0.06 µM and 1.26 ± 0.16 µM, respectively (n = 3, not significantly different, p = 0.099, unpaired t-test) ( Figure 4A,B). Tg244440 showed a clear preference for oxopurine nucleobases, as it interacted better with xanthine ( Figure 4A), with a K i of 42.4 ± 4.2 µM (n = 3), than with adenine, which was poorly recognised by the transporter (Figure 4B), with a K i of 425 ± 80 µM (n = 8) (  Hypoxanthine uptake was very similar to that of guanine: linear over 60 s, saturable, and with a significantly non-zero rate ( Table 1). The slope of the EV control was also non-zero, as previously reported [32] but 86.2% lower than the +Tg244440 clone F5 (p < 0.0001) ( Figure 3B). We conclude that Tg244440 efficiently transports guanine and hy- Interestingly, purine nucleosides were also recognised by the transporter and showed the same trend observed for their respective nucleobases: guanosine and inosine were strongly preferred over adenosine. Indeed, as for hypoxanthine and guanine, the transporter displayed statistically identical K i values for inosine and guanosine (4.05 ± 1.37 µM (n = 4) and 3.30 ± 0.3 µM (n = 7), respectively), only slightly higher than the K m for their corresponding nucleobases, whereas the affinity for adenosine was about 20-fold lower ( Figure 4C). In contrast, Tg244440 displayed 6-fold higher affinity for adenosine than for adenine ( Table 2).
Pyrimidine nucleobases and nucleosides showed very little affinity for Tg244440 as none of them was able to completely block the transport of 0.1 µM [ 3 H]-guanine at 1 mM and only uridine and thymine displayed any significant inhibition at this concentration (p < 0.05), 10,000-fold higher than that of the radiopermeant ( Figure 4D). We thus conclude Tg244440 is a high-affinity oxopurine transporter, with particular affinity for guanine and hypoxanthine, that poorly recognises aminopurines and shows no affinity for pyrimidines at physiologically relevant levels.
Frames A-C: Single experiments in triplicate, representative of three or more identical experiments; for number of replicate determinations for each inhibitor (Table 2). sine uptake could be reached.

Tg244440 Binds Purine Nucleobases and Nucleosides in Different Manners
In order to better understand the affinity of Tg244440 for oxopurines over aminopurines, and its intriguing higher affinity for adenosine than for adenine, we sought to investigate which functional groups of the substrates are involved in the interactions with the transporter. To do so, several commercial and newly synthesised analogues of purine nucleobases and nucleosides, with modifications at different positions, were used in competition to 50 nM of [ 3 H]-guanine to determine the Gibbs free energy (∆G 0 ) of each interaction between transporter and substrate, exactly as described for multiple other transporters [29,35,[40][41][42][43]. All K i and ∆G 0 values are listed in Table 2.

Binding of Purine Nucleobases
Given the difference in the high affinity of Tg244440 for guanine and hypoxanthine as opposed to the low affinity for adenine, we first investigated the participation of the keto group on C6 and of the protonation state of N1. We found that the presence of the 6-keto group was not important per se given that 6-mercaptopurine and 6-thioguanine (Table 2), which form double bonds with C6 that keep N1 protonated, showed statistically identical K i values as hypoxanthine (p = 0.17) and guanine (p = 0.35), respectively, although the sulphur is unable to make a strong hydrogen bond. In contrast, 6-chloropurine, in which N1 is unprotonated, was a poor inhibitor of Tg244440 compared to hypoxanthine (p = 5.8 × 10 −7 ) and displayed a large difference in Gibbs free energy (δ(∆G 0 = 9.6 kJ/mol)), consistent with the loss of an H-bond involving protonated N1. To verify this conclusion we determined the affinity of Tg244440 for 1-methylhypoxanthine and 1-deazahypoxanthine, and found a similar loss in Gibbs free energy (9 kJ/mol and 11 kJ/mol, respectively, relative to hypoxanthine, Table 2).
We previously observed good in vitro and in vivo antiprotozoal potential of purine nucleobase analogues carrying a toxophore on C2 [44,45], and we thus decided to investigate whether Tg244440 could tolerate different C2 modifications. The 2-amine did not interact positively or negatively with the transporter as the affinities for guanine and hypoxanthine were almost identical (p = 0.10), as were the K i values of adenine and 2,6-diaminopurine (p = 0.17). Xanthine, featuring a 2-keto group, did display a significantly lower K i value than hypoxanthine (p = 7.8 × 10 −6 , δ(∆G 0 ) = 7.3 kJ/mol) but this could be the result of the protonation of N3 in xanthine. Indeed, 3-deazahypoxanthine displayed a similar loss in binding energy (δ(∆G 0 ) = 8.1 kJ/mol), which confirms the positive contribution of N3 as an H-bond acceptor. Furthermore, the observation that 3-methylxanthine displayed somewhat higher affinity than xanthine (δ(∆G 0 ) = 2.7 kJ/mol; p = 0.006) was also consistent with unprotonated N3 being the form contributing to binding. Thus, we conclude that the presence of small substitutions on C2 does not affect the interaction with Tg244440 as long as N3 remains unprotonated.
N7 was also found to be involved in the binding of substrates to Tg244440 as 7deazaguanine (p = 0.0028) and 7-deazahypoxanthine (p = 6.1 × 10 −5 ) showed significantly lower affinity for the transporter (δ(∆G 0 ) = 9.1 and 12.3 kJ/mol, respectively). Moreover, a Nitrogen atom shift from position seven of hypoxanthine to position eight (al-lopurinol) caused a significant drop in affinity (p = 1.2 × 10 −6 ), which resembles that of 7-deazaadenosine (p = 0.092), showing that a nitrogen at position eight is unable to make a similar H-bond as N7. Interestingly, the addition of a bromide on C7 of allopurinol (7-Brallopurinol) recovered 2.5 kJ/mol in Gibbs free energy, similar to what has been observed for the T. brucei P1 transporter in which the addition of 7-bromo substantially improves the affinity for 3 -deoxytubercidin [46]. The addition of an extra nitrogen on position eight (8-azahypoxanthine) seems to be highly deleterious to binding by the transporter, as 1 mM of 8-azahypoxanthine failed to inhibit even 50% of the transport of 50 nM [ 3 H]-guanine.
The 9-deaza analogues of guanine and hypoxanthine also displayed significantly lower affinity than their 9-aza counterparts (p = 0.012, 5.1 kJ/mol and p = 5.1 × 10 −5 , 3.5 kJ/mol, respectively). It should be noted that the protonation state of the purine ring nitrogens is part of tautomeric equilibria that are functions of pH and buffer composition, and may be somewhat different in guanine and hypoxanthine, and could contribute to the small differences in δ(∆G 0 ) in their 7-deaza and 9-deaza analogues. However, the sum of δ(∆G 0 ) for the 7-deaza and 9-deaza analogues was very similar for guanine and hypoxanthine at 14.2 and 15.8 kJ/mol, respectively. Figure 5A   7-deazaadenosine (p = 0.092), showing that a nitrogen at position eight is unable to make a similar H-bond as N7. Interestingly, the addition of a bromide on C7 of allopurinol (7-Br-allopurinol) recovered 2.5 kJ/mol in Gibbs free energy, similar to what has been observed for the T. brucei P1 transporter in which the addition of 7-bromo substantially improves the affinity for 3′-deoxytubercidin [46]. The addition of an extra nitrogen on position eight (8-azahypoxanthine) seems to be highly deleterious to binding by the transporter, as 1 mM of 8-azahypoxanthine failed to inhibit even 50% of the transport of 50 nM [ 3 H]-guanine. The 9-deaza analogues of guanine and hypoxanthine also displayed significantly lower affinity than their 9-aza counterparts (p = 0.012, 5.1 kJ/mol and p = 5.1 × 10 −5 , 3.5 kJ/mol, respectively). It should be noted that the protonation state of the purine ring nitrogens is part of tautomeric equilibria that are functions of pH and buffer composition, and may be somewhat different in guanine and hypoxanthine, and could contribute to the small differences in δ(ΔG 0 ) in their 7-deaza and 9-deaza analogues. However, the sum of δ(ΔG 0 ) for the 7-deaza and 9-deaza analogues was very similar for guanine and hypoxanthine at 14.2 and 15.8 kJ/mol, respectively. Figure 5A

Binding of Purine Nucleosides
The (δ(ΔG 0 )) values for hypoxanthine and guanine are also close to the values of the corresponding nucleosides inosine and guanosine, in which, however, N9 would be unable to make any contribution through hydrogen bonds. In T. brucei, the P2/TbAT1 aminopurine transporter binds adenosine only through the adenine moiety, tolerating rather than requiring the ribose half of the molecule [40,47]. For TbAT1, the ribose hydroxyl groups and ring oxygen do not make a positive contribution to nucleoside binding-indeed 2′-deoxy and 3′-deoxyadenosine display somewhat higher affinity than adenosine, virtually identical to adenine [40,48]. To test whether the nucleoside binding mode of Tg244440 is similar to that situation, with at most a minor contribution to bind-

Binding of Purine Nucleosides
The Σ(δ(∆G 0 )) values for hypoxanthine and guanine are also close to the values of the corresponding nucleosides inosine and guanosine, in which, however, N9 would be unable to make any contribution through hydrogen bonds. In T. brucei, the P2/TbAT1 aminopurine transporter binds adenosine only through the adenine moiety, tolerating rather than requiring the ribose half of the molecule [40,47]. For TbAT1, the ribose hydroxyl groups and ring oxygen do not make a positive contribution to nucleoside binding-indeed 2 -deoxy and 3 -deoxyadenosine display somewhat higher affinity than adenosine, virtually identical to adenine [40,48]. To test whether the nucleoside binding mode of Tg244440 is similar to that situation, with at most a minor contribution to binding from the ribose moiety, we systematically determined the affinity of matched pair deoxy-inosine and deoxy-guanosine analogues.
Interestingly, inosine analogues lacking 2 -OH, 3 -OH or 5 -OH showed significantly lower affinity for Tg244440, with losses of 8.3 kJ/mol (p = 0.0049), 8.0 kJ/mol (p = 3.0 × 10 −6 ) and 5.4 kJ/mol in Gibbs free energy (p = 3.2 × 10 −4 ), respectively, compared to inosine. Similar results were shown by 2 -Deoxyguanosine and 3 -deoxyguanosine with loss of 11 kJ/mol (p = 2.9 × 10 −12 ) and 9.4 kJ/mol in Gibbs free energy (p = 1.9 × 10 −8 ), respectively, relative to guanosine ( Table 2). These results show that most likely all three hydroxy groups of oxopurine nucleosides interact with the transporter. The removal of both 2 -OH and 3 -OH (i.e., 2 ,3 -dideoxyinosine, ddI) caused a loss of Gibbs free energy of 12.8 kJ/mol compared to inosine (p = 4.9 × 10 −5 ), which is less than the sum of the energy lost in 2deoxyinosine and 3 -deoxyinosine (16.3 kJ/mol). The likely explanation for this is that both hydroxy groups interact with the same amino acid(s) in the binding pocket and the absence of either allows a higher energy interaction with the other. Figure 5B therefore shows a contribution of −12.8 kJ/mol for the 2 and 3 -hydroxys to the binding of inosine, for which a total ∆G 0 of −30.3 kJ/mol was determined ( Table 2). The difference, 17.5 kJ/mol, is likely the result of one or more interactions with the hypoxanthine base group.
Given the low affinity of Tg244440 for adenosine, we first investigated the -sixth position. Substituting the keto group by a sulphur on C6, creating 6-thioinosine, had no significant effect on affinity (p = 0.18) but the Gibbs free energy of interaction for nebularine (9-(β-D-ribofuranosyl)purine) was much lower than for inosine (11.9 kJ/mol, p = 0.0014). The importance of N1(H) for binding of inosine was confirmed by the ∆G 0 for 1-deazainosine, which was virtually identical to that of nebularine (18.7 versus 18.9 kJ/mol, p = 0.78). In contrast to the binding of the oxopurine nucleobases, position N7 is certainly not positively involved in the binding of inosine as the K i values for 7-deaza analogues of inosine (p = 0.29), of 2 -deoxyinosine (p = 0.019) and of 3 -deoxyinosine (p = 0.25) were not significantly different for Tg244440 than those of their matched pairs (Table 2). Moreover, 7-Cl and 7-Br analogues of 7-deazainosine still presented binding energies highly similar to inosine (p = 0.16 and 0.088, respectively). The further modification of O-methylation and O-ethylation of 7-Cl,7-deazainosine did, however, cause a significant reduction in binding energy, as this changed the protonation state of N1 (δ(∆G 0 ) = 5.7 and 9.2 kJ/mol, p = 1.4 × 10 −4 and 2.0 × 10 −5 , respectively). However, the existence of an (energetically unfavourable) tautomeric state resulting in N1 being a hydrogen bond acceptor (as opposed to a hydrogen bond donor in inosine) in the O-alkylated analogues gives these ligands a higher binding energy than 1-deazainosine (δ(∆G 0 ) = 6.4 and 2.8 kJ/mol, p = 0.0018 and 0.0060, respectively).
These observations show that the base parts of oxopurine nucleosides interact with Tg244440 through only a single H-bond, to the protonated N1 ( Figure 5B), giving a Σ(∆G 0 ) of −30.1 kJ/mol for inosine, very close to the experimental value of −30.8 kJ/mol. It is abundantly clear from the binding model presented in Figure 5 that the interaction of the oxopurine base with Tg244440 is very different for inosine and for hypoxanthine/guanine and that the oxopurine nucleosides are predominantly bound through interactions with the ribose hydroxy groups.
The aminopurines adenosine and adenine displayed only moderate to low affinity for Tg244440, with the K i for adenosine significantly lower than for adenine (p = 0.0083). Curiously, the removal of 2 -OH (p = 0.067), 3 -OH (p = 0.46) or 5 -OH (p = 0.024) of adenosine barely affected its binding to Tg244440 ( Table 2), suggesting that they are not involved in the binding of this nucleoside to the transporter. Although it could alternatively be speculated that 2 -and 3 -deoxyadenosine bind to Tg244440 in slightly different orientations and compensate for the absence of each other, it remains unclear how the ribose moiety contributes to adenosine binding, perhaps through the ring oxygen, but clearly the adenosine binding mode is different from the binding mode of inosine.

Discussion
Every living cell requires purines and pyrimidines as building blocks for their nucleic acids, as energy storage and as intra-and extracellular signalling molecules. T. gondii, like other parasitic protozoa, is unable to synthesise its own purines and relies on the uptake of these nutrients from the host, a property that has been explored for the development of antiparasitic purine analogues [12,16,18,49]. The ability of T. gondii to incorporate exogenous purines, specifically adenosine, has been known for almost four decades [27,50], but the molecular mechanisms of its salvage process are poorly understood.
In its intracellular milieu, T. gondii will have to compete with the host's enzymes for the available nutrients, and free nucleobases and unphosphorylated nucleosides are present at very low concentrations. Thus, intracellular parasites must have evolved nucleobase and nucleoside transporters with particularly high affinity for their substrates, as recently shown for Trypanosoma cruzi [32] and before for Leishmania spp. [42,51], and for the unrelated apicomplexan Plasmodium falciparum [14].
Schwab et al. [27] had already reported that T. gondii tachyzoites are able to rapidly transport inosine and hypoxanthine, and they described an inosine transporter with an apparent K m of 81 µM; however, the authors also noticed that inosine can be internalised via two different mechanisms. One inosine transport activity was sensitive and one was insensitive to inhibition by adenosine; the latter could be inhibited by high concentrations of formycin B or hypoxanthine.
Chiang et al. [28] then reported that the genes encoding adenosine kinase and TgAT1 (whose cloned sequence is the gene annotated as Tg244440 in ToxoDB.org) are essential for susceptibility to Ara-A, and that TgAT1 is the sole adenosine transporter of T. gondii tachyzoites. Upon expression of Tg244440 in Xenopus laevis oocytes, the authors determined an adenosine K m of 114 ± 37 µM, using 10 µM [ 3 H]-adenosine. This report has always been puzzling to us, as T. gondii tachyzoites would be expected not to rely on a very low affinity transporter for their purine salvage. Moreover, although the affinity of TgAT1 for Ara-A had never been measured directly, it is unlikely that the transporter would have higher affinity for this adenosine analogue than for adenosine itself, and therefore the K m of TgAT1 for Ara-A would have been (at least) around the hundreds of micromolar range, which is confirmed here. Thus, it would be expected that TgAT2, which has much higher adenosine transport capacity than TgAT1 and interacts with Ara-A with a K i of 2.9 µM [29], was even more important in the internalisation of this nucleoside analogue.
The use of high concentrations of radiolabel used in some of the previous studies [27,28] allows only for the detection and characterisation of low affinity transporters, as it saturates the higher affinity carriers. This is particularly a limitation of the 1995 study [27], which used 1 µM of [ 3 H]-adenosine, a concentration too high to reliably detect TgAT2 with its reported K m of 0.49 µM [29], leaving only the low affinity flux. Moreover, the concentration of 20 µM of inosine that was used to determine the K m of the inosine transporter(s) expressed by tachyzoites would have completely saturated a high-affinity transporter like TgAT2 (K m of 0.77 µM [29]) and make its identification impossible. Here, we start to address the confusing state of the (all too sparse) literature on this subject, with a thorough characterisation of the Tg244440 gene, this time expressed in a protozoan cell type with a null background for guanine uptake.
Using radiolabelled substrate concentrations of 50 and 100 nM for guanine and hypoxanthine, respectively, we determined that the transporter encoded by Tg244440 has K m values of approximately 1 µM for these purine nucleobases. Similarly, this transporter was inhibited, with almost as high affinity, by the oxopurine nucleosides inosine and guanosine but only by much higher concentrations of the aminopurines, adenosine and adenine. We therefore conclude that Tg244440 is an oxopurine-specific transporter, although the adenosine K m reported by Chiang et al. [28] in X. laevis oocytes is consistent with the value we report here.
The slightly lower affinity for the oxopurine nucleosides over their corresponding bases leads naturally to a hypothesis that Tg244440 is a nucleobase transporter that largely tolerates the N(9)-β-D-ribofuranose. However, the higher affinity of adenosine over adenine does not fit this hypothesis and competition experiments with 2 -, 3 -and 5 -deoxyinosine and/or -guanosine showed that the oxopurine nucleosides derive most of their binding energy from interactions through these hydroxy groups. Equally, systematic investigations of the purine ring interactions found that the oxopurine nucleobases interacted with Tg244440 through all four nitrogen positions (N1(H); N3; N7 and N9(H)), presumably with hydrogen bonds, while in the corresponding nucleosides probably only N1(H) contributed positively to binding. Furthermore, the binding mode for adenosine was different again, with little apparent involvement of the ribose hydroxy groups but binding contributions from N1, N3, N7 and 6-NH2. The 4.5 kJ/mol higher binding energy for adenosine over adenine might be attributable to an interaction with the ribose group but we were unable to ascertain the nature of that interaction. The transporter had at best very low affinity for pyrimidine nucleobases and nucleosides.
It is clear that Tg244440 can bind different classes of purine bases and nucleosides in different ways and is the first protozoan transporter documented to do so. Logically, the transporter either has multiple binding sites or allows the different substrates to orient differently in a single binding pocket. Of these possibilities the former seems unlikely as all the inhibition data were consistent with competitive inhibition and consistently displayed Hill slopes very close to −1. Moreover, current structural models of ENT family transporters [56] do not allow for multiple nucleoside and/or nucleobase binding sites but do show a rather large substrate-binding cavity deep inside the 11-TMD structure. Interestingly, this allows different binding orientations for the classical inhibitors S-(4nitrobenzyl)-6-thioinosine (NBMPR) and dilazep, using different but overlapping binding sites including the nucleoside-binding pocket [56]. In an example from the Nucleobase Ascorbate Transporter (NAT) family, different binding orientations for xanthine and allopurinol were demonstrated for the UapA transporter of Aspergillus nidulans [57]. Mutations in the binding pocket of this transporter were shown to alter both substrate specificity and substrate orientation [58]. Similarly, a recent study showed that a series of mutations applied to the Escherichia coli XanQ transporter in order to mimic the ancestral XanQ (An-cXanQ) altered it from a xanthine-specific carrier to a xanthine/guanine transporter [59]. Moreover, molecular docking simulations predicted two energetically favorable AncXanQguanine binding modes, both of which differed from the mode of interaction between AncXanQ and xanthine [59].
We conclude that Tg244440 is a broad-specificity, high affinity transporter for oxopurine bases and nucleosides and that this is likely enabled by a versatile architecture of its central binding pocket. As such, the transporter may serve as a conduit for therapeutic oxopurine analogues tailored to its unique selectivity profile and the binding models here presented. However, 2 ,3 -dideoxyinosine, which has been shown to have in vitro and in vivo activity against T. gondii [24][25][26], displayed a K i of almost 700 µM for Tg244440 and is therefore unlikely to be an efficient substrate of this carrier. Equally, the adenosine analogue Ara-A which is known to have moderate activity against T. gondii [23], also displayed low affinity for Tg244440. Nevertheless, much progress has recently been made in the identification of nucleoside analogues with highly promising antiprotozoal activities [12], including 7-deaza inosines [60], 3'-fluorinated 7-deazapurine nucleosides [61], 7-phenyl tubercidins [62] and pyrazolo [3,4-d]pyrimidine nucleosides [63]. Additionally, it is worth noting that 7-chloro-7-deazainosine and 7-bromo-7-deazainosine displayed high affinity for Tg244440. Our current activities include the characterisation of the other three T. gondii ENT transporters and the identification of purine nucleoside antimetabolites against toxoplasmosis.

mRNA Isolation and cDNA Synthesis
Parasites released from infected cells were pelleted by centrifugation and the RNA was extracted using the RNeasy Mini kit (Qiagen, Manchester, UK) following the manufacturer's recommendations. Any potential contaminant DNA was eliminated by using the TURBO DNA-free Kit (Thermo Fisher, Loughborough, UK). The RNA was precipitated overnight at −80 • C using 0.3 M ammonium acetate, washed with 70% ethanol, resuspended in nuclease-free water and used for cDNA synthesis utilizing the Invitrogen RETROscript Reverse Transcription Kit (Thermo Fisher), following the manufacturer's instructions, and a provided set of random primers.
Tg244440 was amplified from T. gondii cDNA with primers HDK1328 (forward; TTCATCAAGCTTATGAGTACAATCGAAGAGAGAGCC; HindIII) and HDK1329 (reverse; ATGATAGGATCCCTAGTAAGCGAGAGCGAGGTAC; BamHI) using the high-fidelity Phusion DNA polymerase (New England Biolabs) and cloned into pHD1336 [64] after overnight digestion with HindIII and BamHI. The final plasmid was then sequenced and used for transfection into T. brucei.

T. brucei Axenic Culture and Genetic Modification
T. brucei procyclic form (PCF) cells of TbNBT-KO, a clonal line derived from Lister strain 427, from which a locus of three nucleobase transporter in tandem repeat was deleted [32], were cultivated in SDM-79 medium (Life Technologies), supplemented with 7.5 µg/mL of hemin and 10% of Fetal Bovine Serum (FBS, Biosera, Kansas City, MO, USA) in non-vented plastic bottles at 27 • C. Cells in logarithmic growth were harvested by centrifugation, resuspended in Cytomix buffer and transfected with NotI-linearized plasmid pHD1336 containing the amplified Tg244440 gene from T. gondii. Transfectant cells were selected exactly as reported previously [32].
Briefly, PCF cells of T. brucei in logarithmic growth were harvested by centrifugation at 1000× g for 10 min, washed twice in AB and then resuspended in fresh buffer at a final density of 10 8 cells/mL. Cell suspension in the amount of 100 µL was added to an equal volume of radiolabel solution for a predetermined time at room temperature, after which the reaction was stopped by the addition of 1 mL of a saturating concentration of ice-cold unlabeled substrate solution in AB (usually 1 mM hypoxanthine); this was followed by immediate centrifugation through an oil layer (7:1 (v/v) mix of di-n-butylphthalate and mineral oil (Sigma)). Immediately after finishing the assay, all tubes were flash-frozen in liquid nitrogen and the cell pellets were cut into scintillation vials and lysed with a solution of 2% SDS for one hour under agitation. Then, 3 mL of scintillation fluid (Scintilogic U, Lablogic) was added to each vial and left agitating overnight on a rocking platform. The next morning, the vials were vigorously shaken to ensure optimal mixing and the scintillation was measured in a 300SL (Hidex) scintillation counter. All assays were carried out in triplicate. All experimental data such as K m , K i and V max values are the average and SEM of at least three separate experiments, each in triplicate, unless otherwise indicated.
Background scintillation was measured in vials containing no cells or radiolabeled substrate, and counts associated with extracellular radiolabel were calculated from samples where cells were briefly incubated with the same concentration of radiolabel, but in the presence of saturating concentrations of non-radiolabeled substrate before centrifugation through oil. The average of the three control replicates was subtracted from each data point.

Calculation of Transport Kinetic Parameters
To determine the K m , cells were incubated with a low, non-saturating concentration of radiolabeled substrate and increasing concentrations of unlabeled substrate for 30 s, a time that falls very much within the linear phase of transport (see the Results section) during which the observed rate reflects the true initial rate of transport rather than intracellular accumulation of substrate, depletion of substrate from the assay buffer or transporter turnover. The resulting transport data was plotted to the Michaelis-Menten equation (V 0 = V max × [S]/(K m + [S]), with [S] being the substrate concentration, using GraphPad Prism 8.0. The inhibition constants (K i ) were calculated using the Cheng-Prusoff equation: K i = IC 50 /(1 + ([S]/K m )) [67]. IC 50 values were obtained from sigmoid dose-response plots with variable slope, using a constant permeant concentration several-fold below the K m and a variable inhibitor concentration. The Gibbs free energy (∆G 0 ) of the inhibitortransporter interaction was calculated using the equation ∆G 0 = −RTln(K i ), in which R is the gas constant and T is the absolute temperature of the reaction. It should be noted that these equations apply to competitive inhibitors, which is likely to be the case given (1) all inhibition curves presented Hill slopes consistently close to -1 and, most importantly, (2) no background for [ 3 H]-guanine uptake (the radiolabeled probe used for transport inhibition assays) was seen in TbNBT-KO cells, evidencing that the heterologous expression of Tg244440 created a single guanine transport mechanism in these cells.

Synthesis of Nucleoside and Nucleobase Analogues
Nucleosides, nucleobases and nucleoside analogues were obtained from commercial sources where available, itemized in Table S1 of the Supplemental Information. A number of nucleoside analogues were synthesized as described previously (Table S2 of the Supplemental Information) or synthesized through new routes described in Supplemental File S2.

Phylogenetic Analysis
Amino acid sequences of transporters belonging to several purine and pyrimidine transporter families [12] were obtained from genomic data bases, aligned with Clustal Omega [68] and then subjected to a Maximum Likelihood analysis in Mega-X [69] using the Jones-Taylor-Thornton substitution model with 500 bootstraps. All sequences used for the phylogenetic tree, as well as the transporter families they belong to and the organisms they come from, are listed in Supplemental Information Table S4.