Investigation of Trypanothione Reductase as a Drug Target in Trypanosoma brucei

There is an urgent need for new drugs for the treatment of tropical parasitic diseases such as human African trypanosomiasis, which is caused by Trypanosoma brucei. The enzyme trypanothione reductase (TryR) is a potential drug target within these organisms. Herein we report the screening of a 62000 compound library against T. brucei TryR. Further work was undertaken to optimise potency and selectivity of two novel-compound series arising from the enzymatic and whole parasite screens and mammalian cell counterscreens. Both of these series, containing either a quinoline or pyrimidinopyrazine scaffold, yielded low micromolar inhibitors of the enzyme and growth of the parasite. The challenges of inhibiting TryR with druglike molecules is discussed.


Introduction
Human African trypanosomiasis (HAT) is a serious health problem in sub-Saharan Africa, with an estimated 50 000 new cases each year. HAT, a fatal disease unless treated, is caused by the protozoan parasites Trypanosoma brucei gambiense and T. b. rhodesiense, which initially reside within the blood stream and then subsequently penetrate into the central nervous system giving rise to the classical symptoms of HAT. The current drugs to treat HAT are inadequate due to poor efficacy, side effects and the requirement for parenteral administration, which is not appropriate for a rural African setting. [1,2] Herein we report a target-based approach to the discovery of novel inhibitors of trypanothione reductase (TryR) as a potential therapy for HAT. TryR is an NADPH-dependent flavoprotein disulphide oxidoreductase unique to, and essential for growth of, trypanosomatid parasites, whose function is to convert trypanothione disulphide (N 1 ,N 8 -bis(glutathionyl)spermidine disulphide, T[S] 2 ; Figure 1) into the physiologically relevant reduced dithiol (T[SH] 2 ). [3] In these parasites, T[SH] 2 serves as a substitute for many of the metabolic and antioxidant functions ascribed to glutathione (GSH) in mammalian cells. [4] Mammalian glutathione reductase (GR) is the nearest homologue to TryR. However, the host and parasite enzymes have significant differences in their active site architecture resulting in a pronounced ability to discriminate between their respective disulphide substrates. These features make TryR an attractive target for selective drug design. [5] Before carrying out a drug discovery programme TryR was assessed for appropriateness using the traffic light scoring system that we recently described. [6] The process assesses a target in terms of: level of validation (genetic and chemical), assay feasibility, potential for toxicity and resistance, druggability, and level of structural information available allowing cross target comparisons and prioritisation for entry into hit discovery programmes. The results of this exercise are illustrated in Table 1. In brief, there was sufficient supporting information in the areas of genetic and chemical validation, assay feasibility and protein structure to progress the target into hit discovery. The main area for caution for this target was druggability since the active site is rather large and featureless.
There is an urgent need for new drugs for the treatment of tropical parasitic diseases such as human African trypanosomiasis, which is caused by Trypanosoma brucei. The enzyme trypanothione reductase (TryR) is a potential drug target within these organisms. Herein we report the screening of a 62 000 compound library against T. brucei TryR. Further work was undertaken to optimise potency and selectivity of two novel-compound series arising from the enzymatic and whole parasite screens and mammalian cell counterscreens. Both of these series, containing either a quinoline or pyrimidinopyrazine scaffold, yielded low micromolar inhibitors of the enzyme and growth of the parasite. The challenges of inhibiting TryR with druglike molecules is discussed. A number of different TryR inhibitors have been reported in the literature. These can be broadly classified as tricyclics, [7] polyamine analogues, [5f, h, 8] redox inhibitors, [9] substrate analogues, [10] and compounds identified through screening of a library of druglike compounds. [11] To our knowledge, none of these series have progressed beyond the early discovery phase.

Results and Discussion
Hit discovery In order to discover novel lead compounds against TryR, a high-throughput screen of an in-house designed diverse compound library [12] (62 000 compounds) was performed.
A nonenzymatically coupled colourimetric method for detecting TryR activity as developed by Fairlamb et al. [13] was employed. In this assay, the activity of TryR is coupled to the reduction of DTNB (5,5'-dithiobis[2-nitrobenzoic acid]) by T(SH) 2 to form the yellow thionitrobenzoate anion (TNB À ) (see Figure 1). Unlike the direct assay, this method increases assay sensitivity, and allows the assay to proceed in a linear fashion for extended time periods with T[S] 2 concentrations at or below K m .
The TryR screening assay was miniaturised to a 384-well plate format and optimised to the standard required to support a large-scale screening campaign. The assay was assessed for robustness in an automated environment yielding the following typical performance statistics: Z' = 0.84 AE 0.001; % CV (plate) = 3.65 AE 0.4; S:B = 10 AE 0.25; IC 50 (clomipramine) = 12.4 AE 0.14 mm. Clomipramine was used as a standard throughout the screening process.
TryR was initially screened in single-point mode at 30 mm and the percentage inhibition (PI) value calculated for each compound ( Figure 2). The distribution of hits deviated from normality due to a shoulder of activity between 20 and 50 PI, indicating a capacity of TryR to bind with a low affinity to a range of diverse structures. A statistical assessment of the screen relative to the error around the full signal controls (5 standard deviations) indicated that PI values ! 42 % should be considered as highly significant. All compounds yielding PI ! 50 % were therefore selected for re-testing in duplicate (30 mm). A total of 767 compounds were confirmed as hits giving a confirmation rate (primary to re-test screen) of 84 % and an overall confirmed hit rate of 1.24 %.
A preliminary chemistry assessment of the confirmed hits identified a selection of 120 compounds for potency determinations. Each compound was subjected to a ten-point half-log titration in duplicate ranging from 30 mm to 1 nm. In all cases, residual material from the potency compound plates was subjected to LCMS analysis for structure and purity confirmation. Figure 3 illustrates the ability of this assay to report reliable potency values for test compounds, yielding a correlation coeffi- Table 1. Antiparasitic drug target evaluation for trypanothione reductase.

Criteria Status Comments
Assay feasibility green Recombinant enzyme available in large quantities. 96 well assay in place Genetic/chemical validation green Target essential in trypanosomes as evidenced by conditional knock-out. [3] Some chemical validation from inhibitors such as trivalent antimonials and melaminophenyl arsenicals. Druggability amber Several published low micromolar low molecular weight inhibitors. The active site of TryR is large and therefore the identification of potent low molecular weight inhibitors could be an issue. Potential for resistance amber TryR is a single copy gene in T. brucei and no reported gene amplification in response to inhibitors has been reported.
Reversal of competitive inhibitors by accumulation of T[S] 2 is a potential liability Potential for toxicity green Significant differences in substrate specificity and structure of TryR to corresponding human homologue, glutathione reductase. Protein structural information green Crystal structures of recombinant T.cruzi TryR, alone and in combination with substrates and a covalently linked inhibitor  Analysis of the potency data for the best hits revealed multiple putative hit series covering a range potencies from 0.7 to 30 mm. Key compounds were resynthesised or resupplied to further confirm their activity and identity. From the five series initially identified, two were deemed of sufficient interest to progress into formal hit validation. The aim of this phase was to rapidly assess the existence of defined structure-activity relationships (SAR), whilst securing a significant increase (> 10fold) in activity over the initial hit compound.

Hit series 1
Series 1 was based on a quinoline scaffold, which afforded a number of possible points of variation: the 5-methylfuran in the 2-position, the amide in the 4-position, and the 6-bromo substituent. To expand the SAR, we prepared or purchased 84 analogues. Table 2 provides data on the most potent and significant inhibitors.
The general synthetic route is shown in Scheme 1. In brief, the indoline-2,3-dione was reacted with an appropriate ketone to give the quinoline. The carboxylic acid could then be condensed with appropriate amines to form the required amides.
A summary of the SAR is given in Figure 4. Essentially, any variation of the 5-methylfuran at the 2-position led to a decrease in activity (1 and 17, 20 and 22; 8 and 13; 12 and 18); in the 4-position, amides with a basic substituent, such as dimethylaminoethyl were more potent than neutral aryl or alkylaryl amides (15,23,24,25); and the 6-bromo group could be moved around the ring without too much effect on the activity (1,9,10). Although replacing the 6-bromo substituent with a chloro atom led to a small effect on activity (1 and 8, 16), replacement of the 6-bromo with hydrogen or fluorine led to loss in activity (1 and 12; 4 and 14, 1 and 26). Particularly with substituents on the 4-position, increases in MW and clog P led to small increases in potency, suggestive of a general surface contact between inhibitors and protein. There may be a similar effect with the 6-bromo group, as changes in location to the 7or 8-position of the quinoline ring system, or even replacing it with a chlorine, led to only small effects on activity. It was difficult to deduce which changes, if any, would increase potency at these positions. It is possible that the 5-methylfuran at the 2-position is making a very specific hydrogen bonding and/or p-stacking interaction, which accounts for the requirement for this group at this position.

Hit series 2
Series 2, containing the pyrimidopyridazines scaffold, gave five compounds with an inhibition of > 62 % in the initial screen. The general synthetic route is outlined in Scheme 2. The substituted 6-chlorouracil starting material was made by condensation of the appropriately substituted urea with malonic acid, followed by chlorination. The chloride was displaced with an appropriate hydrazine. The hy-  (35) was then condensed with aldehyde and cyclisation was achieved by treatment with sodium nitrite followed by dehydration through microwave heating in DMF with molecular sieves to give 42. [14] The free NH could be alkylated with various alkyl bromides to give the desired product. In total~30 compounds from this series were assayed. Table 3 provides data on the most potent and significant inhibitors. A summary of the SAR is given in Figure 5. Small alkyl groups were required in positions R 1 (methyl) and R 2 (methyl and ethyl) as replacement of R 2 or both R 1 and R 2 with hydrogen led to a loss in activity (42 and 43). The methyl at R 2 could be replaced by a small alkyl group; however, replacement with larger alkyl groups led to a decrease in activity. A number of different aromatic substituents were tolerated at R 3 , however, there was no clear SAR.

Selectivity and mode of action studies
The selectivity of these series with respect to nearest human homologue of TryR was assessed in glutathione reductase (GR) assays. Both TryR and GR assays were performed with substrate concentrations equal to K m allowing direct comparison of IC 50 values for selectivity assessment. All of the compounds tested demonstrated significant selectivity for TryR over GR (Table 4). This is not surprising if we assume our binding model is correct as the binding site in GR is smaller, less hydrophobic and less negatively charged than that of TryR. Hence, if our compounds are binding by a general hydrophobic interaction and interaction with negatively charged residues in the active site, they are unlikely to interact with GR.
Mode of inhibition studies were carried out on key representatives of series 1 and 2. The data indicated that neither compound series acted as simple linear competitive inhibitors with respect to T[S] 2 . The mixed inhibition of compounds in series 1 indicates the compounds can either bind to the free enzyme or to the enzyme-substrate complex, implying that the compounds are not binding in an orientation that excludes binding  of T[S] 2 . For series 2, the binding appears to be linear and uncompetitive, with compounds binding only to the enzymesubstrate complex. In enzymes such as TryR that obey a bi bi ping pong reaction mechanism, such an inhibition pattern can arise due to an inhibitor competing for the second independent substrate-binding site, NADPH. However, a similar uncompetitive inhibition pattern was observed when the concentration of the second substrate, NADPH, was varied in the presence of a fixed concentration of T[S] 2 . This finding indicates that compounds from series 2 bind at a site distinct from both substrate-binding sites. Since compounds of series 2 have some structural similarity to the FAD co-factor of TryR, we propose that disruption of the orientation or displacement of FAD could account for the uncompetitive inhibition pattern. In theory, both series present significant advantages over competitive inhibitors where build up of substrate trypanothione during inhibition in vivo could reduce the efficacy of the inhibitor. Indeed, substrate build-up in the case of uncompetitive inhibition may even drive binding of the inhibitor. The most potent compounds were also assessed for their ability to affect T. brucei and MRC-5 (prototypical mammalian cell line) proliferation in vitro (Table 4). Series 1 compounds showed weak inhibition of parasite growth. Whilst it is expected that cellular activity is likely to be lower than enzyme activity due to factors such as high intracellular substrate concentration, there was no clear correlation between enzyme inhibition and effect on trypanosomes. In the case of series 2, the cellular activity was more potent than would be predicted by the enzyme assay suggesting that these compounds are either selectively concentrated by the parasites or are acting off-target. However, the latter seems more likely, given the lack of selectivity apparent between the trypanosome and MRC-5 readouts.

Conclusions
We have reported the identification of two novel compound series active against TryR in vitro from a high-throughput screening campaign. Both hit series were low molecular weight compounds with leadlike properties suitable for a medicinal chemistry optimisation programme. They are structurally very different to other TryR inhibitors reported in the literature and constitute novel chemical lead structures against TryR.
SAR studies were undertaken for both series. For series 1, there was some discernable SAR. Unfortunately, we were unable to significantly increase the potency of the compounds against the enzyme to a level likely to have therapeutic significance. The TryR active site contains both hydrophobic and acidic regions (for interaction with the spermidine moiety); it is likely that what we are observing is hydrophobic interactions between the hydrophobic regions of our inhibitors and the active site and electrostatic interactions between the positive charge on the inhibitors and the negatively charged region of the active site. In order to get a significant increase in potency, it will be necessary to build in some additional specific interactions between the inhibitor and the enzyme. This process would be greatly aided by a co-crystal structure of an inhibitor bound in the active site. Some cellular activity was observed, which implies that the compounds are able to penetrate into cells. However we believe that to get a significant correlation between enzyme and cellular activity will require enzyme inhibitors that are significantly more potent.
Data from the gene knockout studies indicated that it would be necessary to cause > 90 % loss in activity of TryR for cell death to occur. [3] Therefore, either a very potent inhibitor (low nanomolar) of the enzyme is required or some kind of irreversible inhibition. Furthermore, the TryR active site is large and relatively solvent exposed, which may make the discovery of low molecular weight, highly potent, competitive inhibitors problematic, especially in the absence of structural information from inhibitors bound into the active site.
In summary, we have identified two novel series of TryR inhibitors from a HTS screen and have carried out SAR studies. Further work is needed to increase the potency of these compounds to a level where the activity is sufficient to achieve significant cellular activity. This would be greatly facilitated by structural information.

Mode of inhibition studies
An assay mixture consisting of TryR, NADPH and DTNB was made up in 40 mm HEPES; 1 mm EDTA, pH 7.4. Assay mixture (180 mL) containing three different concentrations of test compound was added to three rows of a microtitre plate, a fourth row contained only the assay mixture. The test compound concentration ranged up to~1 IC 50 value. T[S] 2 was serially diluted across a fifth row of the plate to produce a 12-point range from 500 mm to 5.8 mm. The assay was initiated by transferring 20 mL of the T[S] 2 row to each of the assay rows. The final 200 mL assay contained 150 mm NADPH; 50 mm DTNB and 20 mU mL À1 U TryR.
Reaction rates were measured from the linear increase in absorbance at 412 nm using a Molecular Devices Thermomax plate reader. Individual data sets were plotted as 1/v versus 1/[S] using Grafit 5.0 (Erithacus software) and inspected for diagnostic inhibition patterns. Data sets were then globally fitted by nonlinear regression to the appropriate equation for competitive, mixed or uncompetitive inhibition.

Cell-based assays
Measurement of the ability of the compounds to inhibit trypanosome (bloodstream form T. brucei brucei single marker) and human (MRC5, human lung fibroblast cells) cell growth was performed using a modification of the cell viability assay previously described by Raz et al. [15] Trypanosome growth assay Cell culture plates were stamped with 1mL of an appropriate stock concentration of test compound, followed by addition of 200 mL trypanosome culture (HMI9-T, [16] 10 % FCS) to each well at a density of 2 000 cells per well (except column 12 receiving media only).

MRC-5 growth assay
MRC-5 cells (2 000 cells per well) were seeded in 200 mL DMEM, 10 % FCS and allowed to adhere overnight. One microlitre of test compound was added to each well on the day of assay initiation.
Trypanosome and MRC-5 plates were incubated at 37 8C in an atmosphere of 5 % CO 2 for 69 h, prior to the addition of 20 mL rezasurin (final concentration 50 mm). After a further 4 h, fluorescence was measured (excitation 528 nm; emission 590 nm) using a BioTek flx800 plate reader.
For cell assays, ten-point screening was conducted between 15 mm and 0.5 nm or between 50 mm and 2 nm, depending on stock concentrations. The final DMSO level was 0.5 % in all cases.