Acyclic nucleoside phosphonates with adenine nucleobase inhibit Trypanosoma brucei adenine phosphoribosyltransferase in vitro

All medically important unicellular protozoans cannot synthesize purines de novo and they entirely rely on the purine salvage pathway (PSP) for their nucleotide generation. Therefore, purine derivatives have been considered as a promising source of anti-parasitic compounds since they can act as inhibitors of the PSP enzymes or as toxic products upon their activation inside of the cell. Here, we characterized a Trypanosoma brucei enzyme involved in the salvage of adenine, the adenine phosphoribosyl transferase (APRT). We showed that its two isoforms (APRT1 and APRT2) localize partly in the cytosol and partly in the glycosomes of the bloodstream form (BSF) of the parasite. RNAi silencing of both APRT enzymes showed no major effect on the growth of BSF parasites unless grown in artificial medium with adenine as sole purine source. To add into the portfolio of inhibitors for various PSP enzymes, we designed three types of acyclic nucleotide analogs as potential APRT inhibitors. Out of fifteen inhibitors, four compounds inhibited the activity of the recombinant APRT1 with Ki in single µM values. The ANP phosphoramidate membrane-permeable prodrugs showed pronounced anti-trypanosomal activity in a cell-based assay, despite the fact that APRT enzymes are dispensable for T. brucei growth in vitro. While this suggests that the tested ANP prodrugs exert their toxicity by other means in T. brucei, the newly designed inhibitors can be further improved and explored to identify their actual target(s).

Trypanosomatid parasites (e.g. Trypanosoma spp., Leishmania spp.) are incapable of purine synthesis de novo and acquire purine derivates from their environment 1 . As a consequence of the complete dependency on an external purine source, these parasites have developed a complex purine salvage pathway (PSP) that enables them to uptake, metabolize and incorporate any naturally occurring purine nucleobases and nucleosides into their nucleotide pools. Therefore, purine derivatives have been considered as a promising source of anti-parasitic compounds. They can inhibit the PSP enzymes or they can become toxic (e.g. cordycepin, tubercidin) when activated by these enzymes [2][3][4][5] . Several PSP enzymes (i.e. GMP synthase, hypoxanthine-guanine-(xanthine) phosphoribosyltransferases HG(X)PRT) were recently validated experimentally as promising therapeutic targets 6,7 .
Acyclic nucleoside phosphonates (ANPs) represent a group of compounds whose biological activity is based on their structural resemblance to the natural nucleotides 8,9 . Their flexibility enables them to adopt a conformation suitable for the interaction with the active site of the nucleotide binding enzymes. Structurally, this type of nucleotide analog is characterized by a heterocyclic base linked to a phosphonate group by various acyclic chains mimicking sugar moiety. These nucleotide analogs are excellent templates for the drug design because of the absence of the labile glycosidic bond and the stability of the phosphonate moiety compared with the phosphate ester bond that can be easily enzymatically or chemically hydrolyzed 10 . The presence of the phosphonate group in the ANPs is responsible for their highly polar character and deprotonation at physiological pH. The prodrug approach has been developed to mask the charge of the phosphonate group and to facilitate the transport across the cell membranes independently from the nucleoside transporters, therefore, improving their

Results and discussion
In T. brucei BSF parasites, the APRT1 is localized in the cytosol and the APRT2 shows a partial distribution between the cytosolic compartment and glycosomes. In PCF T. brucei cells, the C-terminally tagged APRT1 and APRT2 were localized to the cytosol and glycosomes, respectively 22 . In order to decipher the subcellular localization of these two enzymes in the BSF cells, the genes were N-terminally tagged with a V5 tag not to interfere with the C-terminally localized glycosomal signal of the APRT2. Their ectopic expression was triggered by adding tetracycline into the medium. Digitonin was used to dissolve the plasma membrane allowing to enrich for intact organelles in the pelleted fraction. Western blot analysis using antibodies recognizing cytosolic enolase, glycosomal hexokinase and mitochondrial hsp70 protein revealed the purity of the cytosolic and organellar fractions ( Fig. 2A). The APRT1 enzyme was found in the cytosolic fraction and this localization was further confirmed by direct immunofluorescence of the V5-tagged APRT1 (Fig. 2B), although this protein was detected in the glycosomal fraction by proteomics suggesting possible dual localization 23 . The BSF APRT2 was localized to both compartments, glycosomes as well as cytosol ( Fig. 2A,B), similarly to the reported distribution of the PCF APRT2 by TrypTag.org database 24 . Our data are consistent with a partial locali- The RNAi silencing of the APRT1 and the double-silencing of APRT1/2 has no major effect on the growth of the BSF cells but leads to the growth defect in the purine restricted medium. The activity of APRT enzymes was shown to be dispensable for the PCF T. brucei cells grown in vitro 22 . To investigate if the same is applicable for the T. brucei BSF cells, we generated two RNAi cell lines: a single knock-down (SKD) of the APRT1 and a double knock-down (DKD) of APRT1 and 2. The two proteins do not share any similarity at the level of genomic DNA, ensuring the specificity of the RNAi approach. Upon the RNAi induction in HMI-11 medium, no effect on the growth was observed for the SKD APRT1 while a DKD APRT1/2 cells displayed minor, but statistically significant growth phenotype (Fig. 4A). A Western blot using the anti-APRT1 antibody verified a silenced expression of the APRT1 in both cell lines. We noted a significant reduction of the APRT1 levels in the noninduced cells suggesting a leaky expression of the dsRNA that induces RNAi silencing in the absence of tetracycline. Upon the addition of tetracycline, APRT1 expression was fully silenced at day 3 (Fig. 4C). The leaky expression was more profound in the DKD APRT1/2 cell line with only 7% of the APRT1 enzyme expressed. This observed phenomena was confirmed by a qPCR analysis showing a reduction of APRT1 in the noninduced cells when compared to the APRT1 expression levels in BSF427 cells. The qPCR analysis also verified the reduction of the APRT2 transcript in the DKD APRT1/2 cell line (Fig. 4D). The lack of a major growth phenotype upon silencing of APRT1 and 2 suggests that when the medium contains Western blot analysis of the BSF cells overexpressing V5-tagged APRT1 and 2 which were treated with digitonin to obtain cytosolic and organellar fractions. The purified fractions were analysed with the following antibodies: anti-V5, antienolase (cytosol), anti-hexokinase (organellar fraction, glycosomes), and anti-mt hsp70 (organellar fraction, mitochondria). The protein marker is indicated on the left. (B) Immunofluorescence microscopy of the tetracycline induced (IND) V5-tagged APRT1 and 2. The tagged proteins were visualized using a monoclonal V5-antibody and an anti-mouse secondary antibody conjugated with fluorescein isothiocyanate (FITC). The MitoTracker Red was used to visualize mitochondria, while the DAPI was used to stain the DNA content [nucleus (n) and kinetoplast (k)] of the cell. WCL whole cell lysate, ORG organellar fraction, CYT cytosolic fraction. www.nature.com/scientificreports/ hypoxanthine, adenosine and other purine derivates, the BSF cells can fulfil their purine requirements by the interconnected 6-oxo and 6-aminopurine pathways (Fig. 1). Since the HMI-11 medium contains 1 mM hypoxanthine as well as small amounts of adenosine coming from the fetal bovine serum 26 , we tested our generated RNAi cell lines in the home-prepared medium that contains only adenine as the purine source. The medium was supplemented with dialyzed FBS serum to remove any traces of the serum-derived nucleosides (HMI-11 adenine ). The RNAi and the parental (single marker, SM) cell lines were adapted to these new conditions for at least 2 weeks. Upon the RNAi induction, we detected a strong phenotype for the SKD APRT1 RNAi-induced cells suggesting that the APRT2 enzyme is not able to fully compensate for the loss of APRT1. Even stronger growth effect was detected for the noninduced DKD APRT1/2 cells with a low APRT1 expression and for the RNAi induced cells (Fig. 4B) implying an additive negative effect on trypanosoma growth when expression of the APRT2 is silenced. Our results show that there are no other enzymes that can convert adenine to AMP at least with the sufficient capacity and efficacy to allow growth on adenine as a single purine source. The HMI-11 adenine medium contained adenine only at the 50 µM concentration because this purine is toxic to the BSF cells at elevated (millimolar) concentrations. The adenine toxicity is explained by its inhibitory effect on a methylthioadenosine phosphorylase, an enzyme that mediates protection against toxic levels of deoxyadenosine 27,28 . Synthesis of three types of acyclic nucleotide analogues and their prodrugs. Although the APRT enzymes are dispensable for the T. brucei BSF cells (this work and www. tritr ypdb. org), they might become necessary when other branches of the PSP pathway are impaired. To add into the portfolio of inhibitors for various PSP enzymes, we designed three types of acyclic nucleotide analogs as potential APRT inhibitors: (a) ANPs with a sulfur-containing linker connecting adenine and the phosphonate group (thia-ANPs, Scheme 1); (b) ANPs with a nitrogen atom as a branching "hub" in the acyclic moiety (aza-ANPs, Scheme 2) and (c) ANPs using a carbon atom for the side chain attachment to the main linker (Scheme 3). Based on our previous experience with the inhibitors of HG(X)PRTs, the number of atoms between the base and phosphonate group in the linker is optimally five, while the position and nature of the heteroatom(s) in the acyclic moiety influence the flexibility and conformation of the chain(s) 29 . The side chains bearing functional groups enable further interactions in the binding site of the enzyme 14,30 .
The thia-linked nucleotide analogs (thia-ANPs, Scheme 1) were prepared analogously to a known procedure 31 . Mitsunobu reaction of adenine with alcohols 1a and 1b in hot dioxane afforded compounds 2a and 2b, respectively. The phosphonate esters were cleaved with the bromotrimethylsilane in dichloromethane to provide free phosphonic acids 3a and 3b, which were either transformed into the sodium salts for better solubility in the enzyme assays or directly oxidized with hydrogen peroxide to the corresponding sulfoxides 4a and 4b and subsequently converted into their respective sodium salts.
Inhibitors with a nitrogen atom as a branching "hub" in the acyclic moiety (aza-ANPs, Scheme 2) were prepared via three main pathways. The first method used a modification of the side chain of preformed phosphonate 5 32 , followed by the cleavage of ester groups to form phosphonic acid 9a. The second procedure was based on the   38 was applied using transformation to silyl esters in the first step followed by reaction with an ester of (l)-phenylalanine in the presence of 2,2′-dipyridyl disulfide (Aldrithiol ® ) and triphenylphosphine (Scheme 4).
The synthesized ANPs inhibit APRT1 in vitro. The ANPs studies have shown a high specificity of their biological effect with respect to the type of purine nucleobase. Previously, we observed that ANPs bearing guanine or hypoxanthine as a nucleobase inhibit the activity of the T. brucei 6-oxopurine PRTs 6 . To test if the newly synthesized ANPs with an adenine as a nucleobase can inhibit the activity of the APRT1 and APRT2, both enzymes were overexpressed in E. coli. The APRT1 was purified as an active recombinant enzyme from the soluble fraction. On the contrary, the recombinant APRT2 was found insoluble, and therefore was purified in the presence of detergent. Unfortunately, the dialyzed APRT2 enzyme did not exert any enzymatic suggesting that the remaining detergent in the APRT2 sample interfered with the enzyme's activity. Thus, the synthesized ANP-based inhibitors were tested only as inhibitors of the recombinant APRT1. First, we monitored the APRT1 enzyme activity by a continuous spectrophotometric assay measuring the conversion of adenine to AMP at 256 nm. From the steady-state analyses we determined K m and K cat values for adenine (K m = 8.7 ± 1.9 µM and K cat = 0.82 s −1 , respectively) and phosphoribosyl pyrophosphate (PRib-P) (K m = 162 ± 21.6 µM and K cat = 1.8 s −1 ). Compared to the Leishmania donovani APRT1 K m (2.3 ± 1.1 µM) and K cat (17.9 s −1 ) for adenine and K m (25.1 ± 5.9 µM) for PRib-PP 40 , the T. brucei enzyme displays decreased catalytic efficiency and higher K m values for adenine and PRib-PP. The activity of the recombinant APRT1 enzyme was then tested in the presence of synthesized ANPs. Out of the fifteen ANPs tested, seven compounds inhibited APRT1 with K i values ranging from 3.07 μM ± 0.248 (compound 9g) to 27.7 ± 4.46 μM, the remaining nucleotide analogs were inactive at the concentration of 30 μM (Table 1). A common structural feature of all tested ANPs is a five-atom linker connecting adenine and the phosphonate moiety, mimicking the 5-phosphate group of the natural nucleotide. All the ANPs contain one heteroatom (S, N or O) in this main linker. The heteroatom increases the flexibility of the chain and modulates its position in the active site (compare activity of 3a and 3b). While sulfur enables a change in the geometry via oxidation 31 , but results in the loss of the inhibitory activity (compare 3a and 4a), nitrogen makes a facile attachment of the side chain possible. Moreover, these aza-ANPs are prochiral inhibitors (9a-9g and 17), since nitrogen atom is protonated at physiological conditions. The side chain can bear a second phosphonate moiety mimicking pyrophosphate in the active site 30,37 (see 9c, 9d, 9g, 20, 22) or a hydroxyl group (9a) and thus further contribute to the binding. The appropriate length of the side chain also seems to be important (compare, for example, 9d, 9g with 9c), while the addition of the third phosphonate group did not improve the inhibition (derivative 17).

Docking studies of ANP-based inhibitors.
To assess the probable binding modes of the most potent inhibitors, docking calculations were performed. Since T. brucei APRT1 has been slightly explored so far, the only experimental structure that is available for this enzyme is APRT1 in complex with adenine and ribose-5-phosphate, pyrophosphate and Mg 2+ (PDB ID: 5VN4) 41 . We therefore docked the six most potent inhibitors Scheme 1. Synthesis of thia-ANPs. The scheme was drawn using ChemDraw 18.2 (PerkinElmer).  Table 1) into this model. The results showed that all of these compounds can fit neatly into the active site with the adenine base adopting an identical position in all docked structures and this is in agreement with its location in the crystal structure. The side-chain of E120 and the carbonyl oxygen of R41 form key hydrogen bonds to the 6-amino group of the purine base, hence accounting for its specificity as an APRT (Fig. 5), in preference to a 6-oxopurine PRT. Another important feature is the presence of F42, which provides pi-stacking interactions with the base. Thus, the base is held in place by a high level of surface complementarity as well by a strong hydrogen bonding network. Since ribose-5-phosphate and pyrophosphate are present in the Scheme 2. Synthesis of aza-ANPs. The scheme was drawn using ChemDraw 18.2 (PerkinElmer). www.nature.com/scientificreports/ crystal structure, the pockets that house these sites are in an expanded conformation and are representative of the enzyme under catalysis. It is plausible and likely that in the absence of these ligands, this enzyme adopts different conformations as occurs in other APRTs 42,43 and closely related 6-oxopurine PRTs 44 . In pre-catalytic structures/conformations the enzyme may not be able to recognize the phosphate/phosphonate moieties or, indeed the adenine base. The highest docking score (78.5) for 3a places its phosphonate group in the pocket occupied by pyrophosphate (Fig. 5). However, we cannot completely rule out the possibility that it could prefer to bind in the 5′-phosphate binding pocket especially given that the APRTs have flexible structures. The remaining six compounds all have two (or three, 17) functional groups attached and thus can potentially fill both the 5-phosphate and pyrophosphate binding sites. The docking scores for 9a, 9d, 9g, 17, and 22 were 94.17, 112.25, 121.05, 79.9 and 103.85, respectively. The four compounds with the highest docking scores also had the lowest K i values and 17, which has a significantly higher K i value 21.3 μM (Table 1), also had the lowest docking score of 79.9. For 22, the 5-phosphate site is perfectly filled with one of the phosphonates, while the second phosphonate cannot fully reach to the pyrophosphate binding site. A similar docking result occurs for 9d and 9g, except the second phosphonate makes a closer approach to the pyrophosphate binding pocket, but again is not optimally placed compared to one of the phosphates when pyrophosphate binds. Compound 9a is able to extend into both pockets but the hydroxyl group may not be bulky enough and lacks negative charges for optimal binding. For 17, neither the 5-phosphate site nor the pyrophosphate are fully utilized for binding, consistent with the lower docking scores. Given the good correlation between K i values, expected binding modes (based on how the substrates and products bind) and binding modes produced by the docking, the continued use of docking with the GOLD program for further inhibitor design appears as an appropriate strategy. Figure 6 shows a surface representation of the crystal structure of the enzyme with the docking results of 3a, 9g and 22 superimposed. The image confirms the similar docking poses achieved by the three distinct chemical classes (thia-ANP, aza-ANP and C-branched ANP). It also highlights the fact that in this complex, the location where the adenine base and 5-phosphate bind are occluded from the solvent. It is therefore necessary that conformational changes would have been required to allow adenine and ribose-5-phosphate to bind to the enzyme. In the resting state, the enzyme likely adopts a more open conformation.
ANPs based prodrugs exert a cytotoxic activity on T. brucei bloodstream through mechanism unrelated to APRT. The 6-oxopurine-based ANPs displayed cytotoxic effects on T. brucei BSF cells 6 . To assess the effect of aminopurine-based ANPs in parasites, some ANPs were tested in the form of their standardly used phosphoramidate prodrugs (Table 1, compounds 24-28, Scheme 4) that facilitate transport across the plasma membrane and subsequently are cleaved to free phosphonates inside of the cell 39,45 . All the tested prodrugs showed an effect in the single-digit μM range against T. brucei in vitro (Table 1) while the respective freephosphonates did not exert any cytotoxic effects most likely because of the polar character of the phosphonate group that can interfere with their uptake to the cell (Fig. 7). The most potent was compound 28 (prodrug of the parent inhibitor 17) with an EC 50 value of < 1 µM, although this compound at 10 µM is also toxic in Normal human dermal fibroblasts (NHDF) ( Table 1). All other tested prodrugs 24-27 had no effect on the viability of human NHDF and HeLa S3 cell lines at 10 µM (Table 1).
The inhibition of the growth of T. brucei BSF in HMI-11 medium in single micromolar EC 50 values (Table 1) was somewhat surprising considering the dispensability of APRT enzymes for these cells. This observation suggests that while ANPs can inhibit the activity of APRT1 in vitro, this enzyme might not be the primary target of the synthesized inhibitors in vivo. To get deeper insights into the in vivo action of ANPs on APRT enzyme we tested their cytotoxicity on wild type cells grown in HMI-11 adenine medium. In this medium, the cells rely on the www.nature.com/scientificreports/ activity of the APRT enzyme to produce AMP and other nucleosides. We did not detect any significant changes in EC 50 values between the cells growing in HMI-11 and HMI-11 adenine . To corroborate this observation, we also tested cells with induced ectopic expression of V5-tagged APRT1 in both types of media and in the presence of the selected ANPs. Again, no difference was found in the measured EC 50 values (Table 1), further suggesting that APRT enzyme is not the primary target of these compounds in cells. The direct microscopic observation of T. brucei cells that were exposed to the inhibitors 24a, 24b, 27 and 28 revealed problems in the cell cycle.   www.nature.com/scientificreports/ The changes in the distribution of different cell cycle stages within a population were further assessed by flow cytometry assaying the DNA content by staining the cells with propidium iodide (Fig. 8A). The most obvious effect was observed for treatment with compounds 24a and 24b, which caused an increase in the number of cells in G1-phase, while fewer cells were detected to be in S-and G2-phase (Fig. 8B). Our results suggests that in BSF T. brucei the tested ANP prodrugs exert their toxicity by other means than inhibition of the APRT enzymes.

Conclusions
All medically important unicellular eukaryotic parasites are fully dependent on the purine salvage pathway to synthesize building blocks for their DNA and RNA. This dependency represents a potentially promising ground for the discovery of anti-parasitic compounds with a purine-based scaffold. Indeed, we showed that potent 6-oxopurine acyclic nucleoside phosphonates (ANPs) inhibit 6-oxopurine phosphoribosyl transferases in vitro as well as possess strong cytotoxic effects on T. brucei bloodstream form cells 6,17,18 . Using our experience with this system, we designed new selective adenine-based ANPs and evaluated their activity in vitro and in T. brucei culture. Out of the 15 synthesized ANPs, seven derivatives inhibited adenine phosphoribosyl transferase (APRT) with K i values ranging from 3 to 28 μM. Our docking studies using the T. brucei APRT1 crystal structure revealed www.nature.com/scientificreports/ that the synthesized compounds can fit neatly into the active site with the adenine base adopting an identical position in all docked structures and this is in agreement with its location in the crystal structure. Although APRT1 and APRT2 enzymes are dispensable for the growth of BSF T. brucei cells under normal conditions, the cell-permeable adenine-based ANP prodrugs displayed anti-trypanosomal activity in the single µM range. These prodrugs can be further subjected to the structure-activity relationship analyses as well as to studies to identify their cellular target(s).

Methods
General remarks and methods for synthesis of ANPs and their prodrugs. Unless otherwise stated, the general remarks and methodology A, B, and C was adapted from Ref. 31 . Briefly, solvents were evaporated at 40 °C/2 kPa and prepared compounds were dried at 25-30 °C at 2 kPa. Starting reagents and compounds were purchased from commercial suppliers (Acros Organics, Carbosynth, TCI, Fluorochem, Sigma-Aldrich) and used without further purification or were prepared according to the published procedures. Analytical TLCs were performed on silica gel pre-coated aluminium plates with fluorescent indicator (Merck 60 F254). Flash column chromatographies were carried out by Teledyne ISCO CombiFlash Rf200 with dual absorbance detector. Various types of columns were used: (a) Teledyne ISCO columns RediSepRf HP Silica GOLD in sizes 12 g, 40 g, 80 g and 120 g; (b) Teledyne ISCO columns RediSepRf HP C18 Aq GOLD in sizes 50 g and 100 g; (c) column Chromabond Flash DL 40, DL 80, DL 120 and DL 200, filled with FLUKA silica gel 60; (d) Interchim puriFlash C18 Aq in sizes F0040 and F0080. Eluents used were cyclohexane-ethyl acetate 6:4 mixture (A), ethyl acetate modified with 10% of methanol (B), chloroform (C), methanol (D) and water (E). Preparative HPLC purifications were performed on Waters Delta 600 chromatography system with columns packed with C18 reversed phase resin (Phenomenex Gemini 10 μm 21 × 250 mm, Phenomenex Gemini 5 μm 21 × 250 mm, Phenomenex Luna 10 μm 21 × 250 mm) using gradient H 2 O/MeOH as eluent. Dowex ® 50 W resin was turned to Na + cycle by treatment of Dowex 50 W resin in H + cycle with 1 M NaOH aq. solution, followed by water wash to neutral pH.
Mass spectra, UV absorbance and purity of compounds were measured on Waters UPLC-MS system consisted of Waters UPLC H-Class Core System (column Waters Acquity UPLC BEH C18 1.7 mm, 2.1 × 100 mm), Waters Acquity UPLC PDA detector and Mass spectrometer Waters SQD2. The universal LC method was used (eluent H 2 O/CH 3 CN, gradient 0-100%, run length 7 min) and MS method (ESI+ and/or ESI−, cone voltage = 30 V, mass detector range 100-1000 Da). Purity of the final compounds was > 95%. High-resolution mass spectra were measured on a LTQ Orbitrap XL spectrometer (Thermo Fisher Scientific). NMR spectra were recorded on Bruker Avance 400 or 500 spectrometers referenced to the residual solvent signal.
Method A: General procedure for Mitsunobu reaction at elevated temperature. In a 60 ml vial triphenylphosphine (2.00 g, 7.5 mmol), adenine (0.77 g, 5.0 mmol) and the corresponding alcohol (6.0 mmol) was suspended in 40 ml of dry dioxane under an argon atmosphere. The suspension was heated up to 80 °C and then DIAD (1.5 ml, 7.5 mmol) was added dropwise via syringe. The mixture got homogeneous and started to colour green in a few minutes. After 30 min, the reaction was quenched with 5 ml of water and was stirred for further 30 min and then the mixture was evaporated. The residue was purified by reverse phase flash chromatography (water/methanol) to yield the title compound as a colourless oil.  Method G: General procedure for the synthesis of prodrugs. The phosphonate ester (0.5 mmol) was dissolved in dry dichloromethane, acetonitrile or pyridine (5 ml) under argon atmosphere. BrSiMe 3 (0.5 ml, 7.8 eq.) was added and the reaction mixture was stirred at r.t. overnight. The solvent was evaporated under argon atmosphere. Isopropyl l-phenylalanine hydrochloride (for every phosphonate group: 0.37 g, 1.5 mmol), dry pyridine (5 ml) and dry triethylamine (for every phosphonate group: 1 ml) was added under argon atmosphere and the mixture was heated to 70 °C. A solution of Aldrithiol (for every phosphonate group: 0.33 g, 1.5 mmol,

9-[(N-Phosphonoethyl-N-phosphonoethoxyethyl)-2-aminoethyl]adenine (9 g), tetrasodium
salt. Prepared by Method F, starting from 14 (0.32 g, 0.55 mmol). The HPLC fraction containing product was passed through a short column of Dowex 50 in Na + cycle to obtain the product as soluble tetrasodium salt, yield 0.14 g (48%).         Docking calculations. Docking calculations were performed with the program GOLD 46 . Three-dimensional coordinated for the ligands were generated using eLBOW 47 . Coordinates used for the docking were polypeptide chain A of T. brucei APRT (PDB code 5VN4). All of the ligands and water molecules were deleted. All of the side chains were protonated according to the standard pK a values. The centre of the active site was defined as the location where the N9 nitrogen atom of adenine is located in the original coordinates. The search radius was 12 Å. Docking scores were calculated using ChemPLP within the GOLD program. ChemPLP is the default algorithm in GOLD for calculating docking scores and is dimensionless. It is calculated based on a hydrogen bonding term and multiple linear potentials to model van der Waals interactions and steric clashes 48 . Images were drawn using PyMOL 2.4 49 .

Tetra-(l-phenylalanine ethyl ester) prodrug of {[(2-[(adenin-9-yl)methyl]propane-1,3-diyl) bis(oxy)]bis(methylene)}diphosphonic acid (25). Prepared by
Trypanosoma culture and cell lines. The bloodstream form (BSF) T. b. brucei Lister 427 and single marker (SM) strains were cultivated in HMI-11 medium and 10% FBS at 37 °C in a humidified atmosphere at 5% CO 2 26 . HMI-11 medium containing dialyzed 10% FBS and only one defined source of the purine base, adenine, was used for selected experiments (50 μM, HMI-11 adenine ). The SM cell line constitutively expressing ectopic T7 RNA polymerase and tetracycline repressor was used for the tetracycline-inducible expression of dsRNA and V5-tagged proteins. To generate SKD APRT1 and DKD APRT1/2 RNAi cell lines, the 524 bp and 537 bp fragments of the aprt-I (Tb927.7.1780) and and aprt-II (Tb927. 7.1790) open reading frames, respectively, were PCR amplified from T. brucei BSF427 genomic DNA with the oligonucleotides (Supplementary File S1). The APRT1 Treatment with ANPs and cell cycle analysis by flow cytometry. BFS trypanosomes in log phase were resuspended in HMI-11 medium to a density of 2 × 10 5 cells ml −1 . Test compounds at concentrations ranging from 1 to 16 µM were added to the cell suspension followed by 12 h incubation at 37 °C. The cell cycle analysis was performed as described previously 52 . Briefly, approximately 2 × 10 6 cells were washed with PBS and fixed in 1 ml of 70% methanol in 1 × PBS and stored at 4 °C overnight. Following a PBS wash, samples were incubated with 10 µg/ml propidium iodide and 9.6 µg/ml of RNaseA at 37 °C for 45 min. Samples were analyzed on a FACS Canto II (BD) collecting 10,000 gated events.