Design, Synthesis and Biological Evaluation of Trypanosoma brucei Trypanothione Synthetase Inhibitors

Trypanothione synthetase (TryS) is essential for the survival of the protozoan parasite Trypanosoma brucei, which causes human African trypanosomiasis. It is one of only a handful of chemically validated targets for T. brucei in vivo. To identify novel inhibitors of TbTryS we screened our in-house diverse compound library that contains 62 000 compounds. This resulted in the identification of six novel hit series of TbTryS inhibitors. Herein we describe the SAR exploration of these hit series, which gave rise to one common series with potency against the enzyme target. Cellular studies on these inhibitors confirmed on-target activity, and the compounds have proven to be very useful tools for further study of the trypanothione pathway in kinetoplastids.


Introduction
Human African trypanosomiasis (HAT), or sleeping sickness, is endemic in sub-Saharan Africa, claiming the lives of about 30 000 people every year and putting approximately 60 million people at risk of infection. [1] HAT is a progressive and fatal disease caused by the protozoan parasites Trypanosoma brucei gambiense and T. b. rhodesiense, which are transmitted to the human host by the bite of the tsetse fly. If left untreated the disease progresses to the central nervous system and is ultimately fatal. There is a clinical need for more effective drug therapies. Current therapies are toxic and have inappropriate treatment regimens for a rural African setting. There are also problems with treatment failures. [1][2][3] Differences in metabolic pathways have been discovered between parasite and host, which may be exploited for drug discovery programmes. An example of such a difference is found in thiol metabolism and the response of T. brucei to oxidative stress. [4][5][6][7][8] Studies have shown that trypanosomatid parasites are uniquely dependent on trypanothione (N 1 ,N 8 -bis(glutathionyl)spermidine) as their principal thiol, in contrast to most other organisms (including their mammalian hosts) that use glutathione (g-l-glutamyl-l-cysteinylglycine, GSH). [9] In T. brucei trypanothione is synthesised from GSH and spermidine (Spd) by an ATP-dependent CÀN ligase, trypanothione synthetase (TryS; EC 6.3.1.9), with N 1 -and N 8 -glutathionylspermidine as intermediates. [10,11] Selective inhibition of the trypanothione pathway with chemical agents (targeting trypanothione reductase, tryparedoxin, and tryparedoxin peroxidise) or classical gene knockout studies have shown a clear trypanocidal effect. [12,13] TbTryS has also been genetically validated as a drug target, with RNAi and gene knockout studies confirming that TbTryS is essential for T. brucei growth in both bloodstream and procyclic forms, and that there is no alternative bypass mechanism available to the parasite. [14,15] Before commencing a drug discovery programme, TbTryS was assessed for its suitability as a drug target using the traffic light scoring system that we have developed in house. [16] The assessment indicated TbTryS is an attractive target for drug development, especially as it is unlikely to have resistance or toxicity issues, as there is no obvious bypass metabolism or equivalent enzyme in humans. [10] The main concern was the potential druggability of the target. Because the active site of TbTryS is large enough to accommodate trypanothione and precursors, this may be an issue if the active site is a large featureless pocket, as is observed in T. brucei trypanothione reductase (TbTryR). [17] However, the structure of TryS from Leishmania major suggests this is not the case, [18] and the potential to cocrystallise ligands with the protein to inform a chemistry programme was a distinct advantage. Importantly, TbTryS is a bifunctional enzyme, which catalyzes the biosynthesis and hydrolysis of the GSH-Spd adduct trypanothione. The two catalytic domains are separate in Leishmania. The N-terminal domain is a cysteine-containing amidohydrolase/peptidase amidase site, with the C-terminal ATP grasp domain responsible for the synthetase activity of the enzyme. [18] Figure 1 shows the only previously disclosed inhibitor of TbTryS, compound 1. [19] Whilst a valuable tool molecule, the optimisation and development of this phosphinate inhibitor into a potential clinical candidate is limited due to the peptidic nature of such a compound, with a high polar surface area Trypanothione synthetase (TryS) is essential for the survival of the protozoan parasite Trypanosoma brucei, which causes human African trypanosomiasis. It is one of only a handful of chemically validated targets for T. brucei in vivo. To identify novel inhibitors of TbTryS we screened our in-house diverse compound library that contains 62 000 compounds. This resulted in the identification of six novel hit series of TbTryS inhibi-tors. Herein we describe the SAR exploration of these hit series, which gave rise to one common series with potency against the enzyme target. Cellular studies on these inhibitors confirmed on-target activity, and the compounds have proven to be very useful tools for further study of the trypanothione pathway in kinetoplastids.
(PSA), and charges at physiological pH (which are detrimental for cellular penetration, metabolic stability, bioavailability, and blood-brain barrier permeability).
Herein we describe a medicinal chemistry programme to develop inhibitors of TbTryS, which gave rise to some potent compounds. We recently reported the biological experiments to chemically validate TbTryS using five of these compounds: 9, 20, 71, 84, and 89. [15,20] Results and Discussions High-throughput screening As previously described, a high-throughput screening assay for TbTryS was developed and validated. [20] TbTryS was then screened against a 62 000 compound diversity set at singlepoint concentration (30 mm). This gave rise to > 720 hits (compounds with TbTryS percentage inhibition > 33 % at 30 mm). These hits were clustered and filtered down to 174 compounds that underwent potency testing (10-point, half-log dilution dose-response curves from which an accurate IC 50 could be calculated). This gave rise to the six putative hit series, plus a number of singletons. Appropriate hits were then re-purchased to confirm identity and activity. A round of purchasing (where available) and synthesis of analogues was initiated to validate the series and to investigate the SAR around these potential hit series to assess optimisation towards lead series. Table 1 shows the hit series identified from the HTS screen, and the ligand efficiencies of compounds from these series. [21] Although there are six distinct chemical series shown in Table 1, the series can be clustered into two distinct groups that share common pharmacophoric features. Attempts were made to co-crystallise hit ligands in the protein to obtain an Xray crystal structure showing ligands in the binding domain, but unfortunately these have been unsuccessful so far.
Initial hit exploration around series/group 1 Group 1 series were identified as having a common pharmacophore with two hydrogen bond acceptors (HBA) in 1,5 relationship, with one of the HBAs from a heterocyclic ring. The pharmacophore also includes regions of hydrophobicity, indicating a potential lipophilic pocket.

Hit series 1 a: thiazole methylene sulfone
The synthetic route to prepare this series involved the condensation of a thioamide with an appropriate a-bromoketone (Table 2), based on the methodology of Dunn et al. [22] The corresponding a-bromoketones could be bought or synthesised from the corresponding acetyl compound (see Experimental Section below for more details). Activities of hit series 1 a compounds against TbTryS are listed in Table 2.
Exploration around the scaffold of 1 a (Table 2) shows that 2substitution with a small substituent, methyl (compound 4) and fluorine (compound 5), is tolerated without loss of potency. An electron-donating substituent at the 3-position, as in   [a] Yields: 60-87 %; the synthesis of compound 9 was previously described by Torrie et al. [20] [b] IC 50 values against TbTryS.
compound 6, gave a 10-fold loss in potency relative to unsubstituted compound 3; however, all potency was regained if the 3-position group was changed to an electron-withdrawing group, such as trifluoromethyl (compound 7). In addition, a slight improvement in potency was observed for the 3-F (9) and 3-Cl (8) derivatives (IC 50 : 0.4 mm). These analogues showed sub-micromolar potency is achievable within this hit series. Substitution at the 4-position was not tolerated with either an electron-withdrawing or -donating group. Both methyl (10) and trifluoromethyl (11) were found to be inactive. However, 3,4-disubstitution with an electron-withdrawing substituent at the 3-position restored potency (compounds 12 and 13), although ligand efficiency was decreased.

Hit series 1 b: tetrazole methylene carbonyl
In this series, the thiazole subunit is replaced by a tetrazole, and the sulfone by a carbonyl group. The tetrazoles were alkylated with a-bromoketones (Table 3). Amide analogue 23 was synthesised via an amide coupling of the corresponding acid, which in turn was obtained from saponification of the ethyl ester.
Subtle trends in SAR similar to series 1 a were observed in series 1 b (Table 3). In particular, there was a 25-fold gain in potency for the 3,5-dichloro analogue 20, relative to the unsubstituted scaffold 14. This could be due to the greater lipophilicity of this compound causing an increase in nonspecific binding, although the ligand efficiency improved (LE from 0.39 to 0.45), suggesting a favourable specific interaction. It was possible to modify the ketone moiety to an amide without loss of potency (compare amide 23 (IC 50 = 0.6 mm), with ketone 20 (IC 50 = 0.3 mm)). In going from methyl (21) to ethyl (22) to tertbutyl ketone (20), an improvement in potency was observed (IC 50 : 8.8 to 6.5 to 0.3 mm, respectively), indicating lipophilic bulk in this area is required. In contrast, the phenyl ketone 24 shows a marked loss in potency, with an IC 50 value of 45 mm.

Hit series 1 c: tetrahydroindazole methylene amide
The pyrazole core for series 1 c was made by condensation of ethyl hydrazinoacetate with the commercially available dione. The ester was then hydrolysed, allowing amide couplings using standard methodology (Table 4). [23,24] Small changes to optimise the amide group in this series generated a "flat" SAR plateau (25-28; Table 4). Aliphatic amides (29-31) were tolerated, albeit at weaker potency, but bulky rigid substitution with the phenyl 32 and cyclohexyl 33 moieties was not tolerated, with complete loss in potency observed. Tertiary amide 34 with two bulky benzyl groups also lost all potency, but the less bulky diethyl amide 35 retained some inhibitory activity.

Initial hit exploration around series/group 2
Following the HTS campaign we also identified three hit series based around a second pharmacophore. This second group of hit series described a slightly different pharmacophore, with the heteroatom of the heterocycle in a 1,6 relationship with the HBA motif of the sulfonamide. These hit series had higher molecular weight and generally lower potency, with lower ligand efficiency (see Table 1). To explore the SAR around these series, compounds were purchased and screened. Data for these are listed in Tables 5, 6, and 7.
None of the group 2 series offered any benefits in terms of potency over the compound series from group 1. In addition, the DMPK data indicated less favourable properties of group 2  series compounds, relative to data for compounds from group 1 (see Table 12 below). Therefore, further optimisation work was focussed on the group 1 series.

Hit to lead optimisation strategy
Hit series 1 a-c were successfully validated and shared an overlapping pharmacophore for TbTryS activity (Figure 2), demonstrating clear SAR and a visible potential for further optimisation. The pharmacophoric features of the three group 1 series scaffolds were hybridised into one new core scaffold, which was based on an indazole ( Figure 3). One of the indazole nitro-gen atoms becomes the HBA, and the other is used for attachment of the other HBA. This indazole series is also predicted to have reasonable physicochemical properties: low molecular weight, clogP < 5, and low PSA. For example, compound 60 (Table 8) has M r = 292 Da, clogP = 4.0, and PSA = 35 2 .

Validation of the indazole series
The indazole scaffold was prepared as shown in Scheme 1. Indazole was first iodinated (at the 3-position) using standard conditions. [25] A Suzuki reaction afforded the 3-aryl intermediate. Alkylation of the N1 nitrogen atom with the required achloroketone (or arylmethylchloride) gave final products 60, 71, and 75-79. Alternatively for compounds 61-70, the alkylation of the 3-iodoindazole was completed first and the Suzuki reaction, to add the aryl substituent, was employed as a final step. Table 8 shows the data for key compounds from the initial SAR study, describing changes to the aromatic group at position 3

3-Indazole substitution
tBu 37 3-fluorophenyl tBu 0.09 0.42 [a] Ligand efficiency, calculated as À0.6 ln(IC 50 )/(heavy atom count). [21] [b] Compound 71 reported previously. [20]   98 www.chemmedchem.org of the indazole (depicted R in the structure). The initial compound 60 (R = phenyl), had an IC 50 value of 150 nm in the TbTryS enzyme assay (LE calculated as 0.43) and was the most potent compound to date. Para substituents led to loss in activity (e.g. 61, 63, 64, and 69). This is possible evidence for the presence of a tight binding pocket into which the aromatic subunit sits. A variety of heterocycles were tolerated, although some gave a 10-fold loss in activity, such as the 2-furanyl compound 66 and 3-pyridyl compound 67. The 3-fluorophenyl motif (71, IC 50 = 90 nm, LE= 0.42) was more potent than 3chlorophenyl (70, IC 50 = 354 nm), and was used as the standard template for exploration of substituent SAR around the R 1 position (see Tables 9-11 below). Having established the indazole as a potent heterocyclic core with good ligand efficiency, a number of SAR studies were carried out to explore chemical optimisation around this scaffold.

Investigation of alternative HBA moieties
We were concerned that the ketone group could react with nucleophiles, so a number of alternative HBAs were investigated. The initial sulfone analogue 74 (TbTryS IC 50 = 120 nm) was found to be equipotent to the ketone 60 ( Table 9). The posi-tion from which the pendant HBA is attached to the core scaffold was also investigated. In compounds 72 and 73 the sulfone HBA is appended from the 3-position of the indazole core (Table 9). This modification had little effect on activity, which is probably due to the symmetry of the molecule. For ease of synthesis, further HBA modifications were investigated at the N1 position of the indazole.
Approaches were made to use a heterocycle as the second HBA motif (Table 10). Pyridine and oxazole were investigated, but showed lower activity than the ketone or sulfone groups.

Amide HBA analogues
Alkylation of the 3-arylindazole (synthesis shown in Scheme 2) with ethylbromoacetate, followed by saponification, yielded the indazole-N-acetic acid intermediate. This was coupled under standard amide coupling conditions to the appropriate amine to give access to amide analogues.
Encouragingly, the ketone subunit could be replaced by a simple amide without loss of potency or ligand efficiency (Table 11). Thus the diethyl and dimethyl amides (87 and 88) had IC 50 values similar to that of the ketone 71, with compound 88 having a ligand efficiency of 0.44. Potency was retained when fusing the dialkyl amide into an N-piperidine amide (89 TbTryS IC 50 = 135 nm), but activity was completely abolished with a larger alkyl aromatic substituent (e.g. 86 IC 50 > 100 mm).
Addition of an appended basic amine, to potentially pick up the TbTryS endogenous substrate Spd binding domain interactions and improve ligand potency, was investigated. This basic group could also improve the aqueous solubility of our compounds and lower the lipophilicity (logP). Although the pipera-Scheme 1. General synthesis of indazole analogues. Conditions: 1) KOH, I 2 , DMF, as in Edwards et al., [25] 89 % yield; 2) ArB(OH) 2 , Na 2 CO 3 , DME, EtOH, H 2 O, yields: 50-78 %; 3) NaH, DMF, a-haloketone (or other), yields: 37-62 %. Synthesis of compounds 71 and 89 were described previously. [20] [a] Ligand efficiency, calculated as À0.6 ln(IC 50 )/(heavy atom count). [21] Scheme 2. Synthesis of amide analogues. Conditions: 1) NaH, DMF, ethyl bromoacetate, as in Fujimura et al., [29]  zine-containing compounds 90 and 80 lost potency (ninefold relative to 71), and were less efficient binders (ligand efficiencies of 0.29 and 0.28 respectively), they were still sub-micromolar inhibitors of the TbTryS enzyme, with TbTryS IC 50 values of 0.86 and 0.83 mm, respectively. If the second basic amine centre was removed and the compounds were truncated to make the C2-and C3-linked dimethyl amine compounds 83 and 84, a significant improvement in potency (six-and 18-fold, respectively) over the analogous piperazines was observed. These compounds were also significantly more efficient binders than compounds 80 and 90, with respective ligand efficiencies of 0.38 and 0.39. The C3-linked dimethylamine compound 84 was observed to be the most potent compound to date, with a TbTryS IC 50 value of 45 nm. Compound 84 shows improved physicochemical properties over compound 60, especially in decreased lipophilicity, with a clogP value of 2.8 (1.3 units lower than that of 60), and M r = 354 Da and PSA = 50 2 .
As compound 84 shows similar potency to an analogue not containing an appended amine (71, IC 50 90 nm) it is unlikely that compound 84 has picked up the Spd binding domain. This conclusion is supported by competition binding studies which revealed that the compounds displayed mixed inhibition with respect to Spd, and did not show classical competitive binding kinetics. [20] Finally, capping the NH group of the amide of 84 with a methyl group (compound 85) resulted in a fourfold decrease in potency, and the C3-linked imidazole 81 was eightfold less potent than the dimethylamine.

Cell potency and DMPK parameters for key compounds
Following the discovery of promising novel TryS inhibitors, compounds were progressed into a trypanosomal cell proliferation assay and a human cell counter-screen. Selected compounds were also screened in an in vitro metabolic clearance assay (Table 12), to assess suitability for series progression. Metabolic stability studies using pooled human liver microsomes indicate a range of stabilities. Compounds from the group 2 series are highly metabolically unstable. However, compounds 8, 26, and 71 are reasonably metabolically stable, suggesting nothing fundamentally problematic with these scaffolds in terms of cytochrome P450-driven metabolism. Table 12 also shows cell data for key compounds from several of the various series of TbTryS inhibitors discovered. Although these compounds were not toxic to the MRC5 mammalian cell line, there was up to a 100-fold decrease in going from enzyme to trypanosomal cell efficacy, even with the lead compounds 71 and 84 (TbTryS IC 50 : 90 and 45 nm, respectively). While these cell potencies are equivalent to the drugs eflornithine (22 mm) and nifurtimox (2 mm) currently in clinical use for late-stage human African trypanosomiasis, they are much less potent than the alternative arsenic-containing drug melarsoprol (8 nm), although arsenic-based compounds do show significant toxicity in the clinic. [26] As this large potency shift between the enzyme IC 50 values and parasite EC 50 values was unexpected, further experiments were carried out to confirm whether hit compounds were entering the cell and acting on-target. As described fully elsewhere, exposing T. brucei parasites to the model TbTryS inhibitors 89 and 84 (2 EC 50 for 72 h) resulted in trypanothione levels dropping to < 10 % of wild-type levels. [15,20] In addition, there was a corresponding increase in the TbTryS substrate [a] Ligand efficiency, calculated as À0.6 ln(IC 50 )/(heavy atom count). [21] [b] Compounds 84 and 89 reported previously. [15,20]  [a] Not determined. 100 www.chemmedchem.org GSH, providing strong evidence that these compounds were acting on-target. As previously reported, the on-target effects of these hit compounds were further confirmed by generating TbTryS single knockout (SKO) and TbTryS overexpressing (OE) cell lines. Western blot analysis and densitometry demonstrated that TbTryS protein levels were decreased in the SKO cells and elevated in the OE cell line, relative to wild-type cells, and these cell lines showed the expected changes in potency to 89 (EC 50 values: 20.4, 6.9, and 44.5 mm for wild-type, SKO, and OE cell lines, respectively) and 84 (EC 50 values: 7.1, 1.2, and 23.3 mm for wild-type, SKO, and OE cell lines, respectively), confirming that TbTryS is the specific target of these compounds. [15,20]

Conclusions
In this work we successfully took HTS hits, clustered them into putative hit series, and rationalised their activities based on common pharmacophores. Initial investigation of SAR around the hit series confirmed an overlapping pharmacophore, and the optimisation potential of group 1 hit series in particular. Following the SAR on group 1 series, a hybridisation strategy and scaffold-hopping approach led us to discover the indazole lead series. Optimisation of this series for potency and improved DMPK properties led to compounds 71 and 84, which displayed in vitro enzyme potencies > 10-fold improved over the best HTS hits. Attempts so far to co-crystallise our inhibitors with the TbTryS enzyme have failed to produce robust data.
Although these indazoles inhibit TbTryS with IC 50 values of < 100 nm, they failed to show sub-micromolar potency in a trypanosome proliferation assay. This can be rationalised by the observation that parasites can survive with low levels of trypanothione beyond the timeframe of the standard whole-parasite proliferation assay. The extension of the time-course in screening assay format is prohibited by the need for repeated dilutions of samples to remain in log-phase growth, leading to unacceptable variability. The lead compounds do, however, show a robust biochemical effect in T. brucei, and are proven to act on-target, inhibiting TbTryS in cells. [15,20] The current lead compounds could also prove very useful in combination therapy with known trypanocides (such as melarsoprol), as studies have revealed TryS-depleted T. brucei procyclics are significantly more susceptible to trypanocides. [27] Our compounds are the most advanced, potent, and drug-like (as predicted by physicochemical and in vitro DMPK properties) inhibitors of TbTryS reported to date, and are extremely useful leads to further explore the trypanothione pathway in kinetoplastids.

Experimental section
Chemistry 1 H NMR spectra were recorded on either Bruker Avance DPX 500 or Bruker Avance 300 spectrometers. Chemical shifts (d) are expressed in ppm. Signal splitting patterns are described as singlet (s), broad singlet (bs), doublet (d), triplet (t), quartet (q), multiplet (m) or combination thereof. LC-MS analyses were performed with either an Agilent HPLC 1100 series instrument connected to a Bruker Daltonics MicrOTOF, or an Agilent Technologies 1200 series HPLC connected to an Agilent Technologies 6130 quadrupole LC-MS; both instruments were connected to an Agilent diode-array detector. LC-MS chromatographic separations were conducted with a Phenomenex Gemini C 18 column, 50 3.0 mm, 5 mm particle size; mobile phase, H 2 O/CH 3 CN + 0.1 % HCOOH 80:20!5:95 over 3.5 min, and then held for 1.5 min; flow rate: 0.5 mL min À1 . Highresolution electrospray MS measurements were performed on a Bruker Daltonics MicrOTOF mass spectrometer. Thin-layer chromatography (TLC) was carried out on Merck silica gel 60 F 254 plates using UV light and/or KMnO 4 for visualisation. TLC data are given as the R f value with the corresponding eluent system specified in brackets. Column chromatography was performed using RediSep 4 or 12 g silica pre-packed columns. LCMS chromatographic separations were conducted with a Waters Xbridge C 18 column, 50 mm 2.1 mm, 3.5 mm particle size; Method A: mobile phase, H 2 O/ CH 3 CN + 0.1 % NH 3 ; linear gradient 80:20!5:95 over 3.5 min, and then held for 1.5 min; flow rate 0.5 mL min À1 . All reactions were carried out under dry and inert conditions, unless otherwise stated.