N-myristoyltransferase inhibitors as new leads to treat sleeping sickness.

African sleeping sickness or human African trypanosomiasis, caused by Trypanosoma brucei spp., is responsible for approximately 30,000 deaths each year. Available treatments for this disease are poor, with unacceptable efficacy and safety profiles, particularly in the late stage of the disease when the parasite has infected the central nervous system. Here we report the validation of a molecular target and the discovery of associated lead compounds with the potential to address this lack of suitable treatments. Inhibition of this target-T. brucei N-myristoyltransferase-leads to rapid killing of trypanosomes both in vitro and in vivo and cures trypanosomiasis in mice. These high-affinity inhibitors bind into the peptide substrate pocket of the enzyme and inhibit protein N-myristoylation in trypanosomes. The compounds identified have promising pharmaceutical properties and represent an opportunity to develop oral drugs to treat this devastating disease. Our studies validate T. brucei N-myristoyltransferase as a promising therapeutic target for human African trypanosomiasis.

Protein N-myristoylation is a ubiquitous eukaryotic co- and post-translational modification and is required for membrane targeting and biological activity of many important proteins 1,2. The N-myristoylation reaction, i.e., the transfer of C14:0 myristic acid from myristoyl-coenzyme A (CoA) to the amino group of N-terminal glycine residues within specific sequence contexts 3, is catalysed by the enzyme myristoyl-CoA: protein N-myristoyltransferase (NMT; EC 2.3.1.97) 4. In T. brucei, NMT activity is encoded by a single gene, which has been shown to be essential for parasite growth using RNA interference 5. The effects of NMT knockdown on T. brucei are likely to be complex since more than 60 proteins are predicted to be N-myristoylated in this organism 6. Experimentally validated targets for NMT include ADP ribosylation factors (Arf) 7, ADP ribosylation-like factors (Arl) 8, a calpain-type protease (CAP5.5, TbCALP1) 9 and, in the related Leishmania major and T. cruzi parasites, hydrophilic acylated surface proteins (HASPs) 10 and flagellar calcium-binding protein 11, respectively. The predicted pleiotropic effects of NMT inhibition on trypanosome physiology make it an attractive target for therapeutic intervention. NMT has also been considered as an anti-cancer 12, fungal 13 and viral 14 target. Fungal NMT orthologues have been shown to be druggable, although broad-spectrum activity has not been achieved. Nevertheless, since there is good evidence from these programmes that selectivity versus human NMT is possible, NMT has been proposed as a target for the treatment of human African trypanosomiasis (HAT) and other parasitic diseases 15,16. There are 2 human isozymes sharing 77% identity (huNMT1 and 2) 17 of which huNMT2 is the closest human homologue to TbNMT, with overall 55% identity and 69% similarity. Based upon 31 residues that are within 5Å of DDD85646 in the active site, this rises to 83% identity and 90% similarity.

Pyrazole sulfonamide inhibitors of TbNMT

To date, no drug-like, potent inhibitors of TbNMT have been reported 18. Screening of a 62,000 diversity-based compound library 19 against TbNMT identified a number of “lead-like” hits, including a chemically tractable series with moderate potency (2 μM) based on a pyrazole sulfonamide scaffold (DDD64558). Optimisation of the screening hit, involving the design and synthesis of over 200 compounds, identified highly potent inhibitors of TbNMT with single digit nanomolar IC50 values and levels of selectivity over human NMT enzymes in the 1 to >100-fold range (Fig.1a). The relative lack of activity of the piperidine analogue (DDD85635) of DDD85602 indicated the terminal basic nitrogen of DDD85602 was crucial for activity. The series was optimised by rigidifying the flexible linker to the amine moiety of DDD85602 and by adding the chlorines observed to give an increase in activity in the unelaborated template (DDD73234). TbNMT inhibitors so obtained (e.g. DDD85646) inhibited the proliferation of bloodstream form (BSF) T. brucei in culture with the best compounds yielding EC50 values between 0.8 and 3 nM and clear windows of selectivity (>200-fold) with respect to proliferation of a prototypical mammalian cell type (MRC5). We attribute the increased selectivity at the cellular level to differences in cell biology between host and parasite, although differential cellular pharmacokinetic behaviour has not been definitively ruled out. Critically, a tight correlation (R2 = 0.875) was observed between IC50 and EC50 values for TbNMT and T. brucei proliferation, respectively, over a 10,000-fold potency range (Fig. 1b); indicating that inhibition of TbNMT was driving the observed anti-parasitic effect of these compounds. The poorer correlation for the most potent compounds (see lower left quadrant of Fig. 1b) is most likely due to the limit of the enzyme assay to provide accurate IC50 determinations for such highly active, tight-binding inhibitors (see Supplementary Information, Figure 1). As a result of its impressive potency in inhibiting both TbNMT and T. brucei proliferation in vitro, together with its promising physicochemical properties, DDD85646 was assessed for efficacy in animal models of trypanosomiasis.

Figure 1

Identification of NMT lead series inhibitors

a. Chemical evolution of DDD85646 from the initial high-throughput screening hit DDD64588. Combining the structure-activity-relationships from the two strategies led to the development of the potent compound DDD85646. Potencies were determined for all compounds synthesised against recombinant TbNMT and huNMT, as well as against BSF T. brucei and MRC5 proliferation in vitro.

b. Correlation between the inhibitions of recombinant TbNMT and BSF T. brucei proliferation for 175 members of the pyrazole sulfonamide series. Data shown are replicates of between 2 and 22 independent potency determinations using 10-point curves. Robustness of TbNMT and trypanosome proliferation assays are exemplified through routinely reported parameters of Z’ (0.703 ± 0.050, n=169 and 0.695 ± 0.095, n>1000 for TbNMT and trypanosome assays respectively) and reproducible potencies of standards (DDD73498 (TbNMT assay) pIC50 = 6.52 ± 0.14, n=276 and pentamidine (trypanosome assay) pEC50 = 8.37 ± 0.41, n=497).

TbNMT inhibitor cures acute HAT in vivo

DDD85646 is moderately bioavailable (ca. 20%) and demonstrates good exposure after oral dosing at 10 and 50 mg kg−1 to female NMRI mice, with the free drug level above the EC99 for T. b. brucei proliferation for over 6 and 10 h, respectively (Fig. 2a). Furthermore, this compound cured all animals in the T. b. brucei acute mouse model of HAT at a minimal oral dose of 12.5 mg kg−1 (b.i.d. for 4 days) (Fig 2b). Cure of all animals was also obtained with shorter oral dosing schedules: 100 mg kg−1 b.i.d. for one day and 25 mg kg−1 b.i.d. for 2 days. Importantly, DDD85646 also cured all animals at 50 mg kg−1 (b.i.d. for 2 days) in the more refractory, but clinically relevant T. b. rhodesiense model of HAT (see Supplementary Information, Figure 2). This reduced sensitivity in vivo is not due to reduced sensitivity to the compound in vitro (T. b. rhodesiense EC50 0.6nM), but maybe a result of the known precedent for this species to occupy privileged sites in vivo. Notably, the efficacy observed for DDD85646 was comparable with the responses observed for the clinically-used drugs pentamidine and melarsoprol in the T. b. brucei model (minimal full cure doses were 1 mg kg−1 and 0.5 mg kg−1 (IP) respectively). Furthermore, despite the minimal window of in vitro enzyme selectivity between TbNMT and mammalian (human) NMT (Fig. 1a and vide infra), this compound was well tolerated at efficacious doses.

Figure 2

TbNMT inhibitor cures acute trypanosomiasis in vivo

a. Mean total and free blood concentration time profiles following single oral administration of DDD85646 at 10 and 50 mg kg−1 free base to female NMRI mice (n=3 per dose group). EC99 is calculated from the average EC50 of 2.46 ± 1.8 nM and Hill slope of 4.84 ± 0.6 (n=5). Solid lines are total plasma concentrations and dashed lines are the predicted free plasma concentrations (fraction unbound in plasma = 0.063).

b. Kaplan Meier survival plot for female NMRI mice (n=5 per dose group) following infection with T. b. brucei strain 427 (variant 221) (inoculum 1 × 104 parasites). Oral treatment with DDD85646 commenced 3 days after infection at the indicated doses (all b.i.d for 4 days).

TbNMT inhibitors are trypanocidal

Addition of DDD85646 resulted in rapid killing of trypanosomes both in vivo and in vitro (Fig. 3a, b). Parasite counts dropped to below detectable levels within 12 h of dosing mice at 50 mg kg−1 b.i.d. Addition of compound (50 nM) to BSF T. brucei cultures in vitro also resulted in rapid killing with numbers of motile cells reduced to below detectable levels between 24 and 48 h. The apparent differences in kinetics of death between the in vivo and in vitro systems are most likely a combination of the harsher in vivo environment for drug-damaged trypanosomes and the fact that compound exposure reached higher concentrations in vivo (up to ~1 μM), compared with 50 nM in vitro.

Figure 3

TbNMT inhibitors have rapid trypanocidal effects in vitro and in vivo

a. Parasitaemia in mice (n=3 per group) with (red) or without (black) DDD85646 treatment (50 mg kg−1, oral, b.i.d); for method see Figure 2b. Arrows represent dose administration times. Data: mean ± s.d.

b. T. b. brucei proliferation in culture determined by counting motile parasites in presence (red) or absence (black) of 50 nM DDD85646. Data: mean ± s.d. for 3 determinations.

c. Blood smears of infected mice and culture samples were stained by Giemsa and observed by light microscopy. Treated cells showed typical BigEye phenotype.

d. Scanning electron micrograph of T. b. brucei treated with 10 nM DDD85646 for 24 h. Inset shows an untreated control cell.

e. Transmission electron micrograph of sagittal section of flagellar pocket of T. b. brucei treated with 5 nM DDD85646 for 72 h. Inset shows a section of flagellar pocket of an untreated control cell.

Asterisks mark flagellar pockets. Dashed lines: cell detection limits. Scale bars: 500 nm.

The trypanocidal mechanism of compound action was confirmed by subjecting T. brucei treated in vitro with 50 nM compound to live/dead FACS analysis, which showed >95% cell death within 24 h of treatment (see Supplementary Information, Figure 3). Furthermore, wash-out experiments showed that death was irreversible after 48 h of exposure to 50 nM compound (data not shown). Microscopic examination of the trypanosomes treated with DDD85646 in vivo and in vitro revealed the same abnormal morphology, i.e., the development of a large vesicular structure (Figure 3c). A scanning electron micrograph of a compound-treated trypanosome clearly shows the rounded and swollen features of this phenotype compared to control (Fig. 3d). Interestingly, rapid cell killing with a similar morphological phenotype has been observed previously following treatment with myristate analogues, such as 10-(propoxy)decanoic acid 20. This morphology closely resembles the ‘BigEye’ phenotype observed in BSF T. brucei when endocytosis is disrupted through the knockdown of clathrin heavy chain, TbRab5 21,22 or TbArf1 7 leading to expansion of the flagellar pocket. Parasites with enlarged flagellar pockets are clearly visible in the DDD85646-treated trypanosome population (Fig. 3e). Additional studies are required to fully understand the cellular effects of NMT inhibition.

Inhibitor acts ‘on-target’

Incubation of BSF T. brucei with [3H]-myristic acid results in the biosynthetic radiolabelling of myristoylated substrates, particularly the highly abundant variant surface glycoprotein (VSG) 23. Using this method, but including a detergent lysis step that activates an endogenous phospholipase C to release [3H]-myristate label from the glycosylphosphatidylinositol (GPI) anchor of VSG 24, a number of putative N-[3H]-myristoylated proteins were visualised by SDS-PAGE and fluorography (Fig. 4a, lane 2). The labelling of most of these proteins was eliminated by prior treatment with DDD85646 (Fig. 4a, lane 1). To confirm that most of the proteins labelled with [3H]-myristic acid were indeed N-myristoylated, a duplicate gel was treated with 0.2 M NaOH in methanol prior to fluorography to remove base-labile hydroxy- or thio-ester linked [3H]-myristate (Fig. 4a, lanes 3 and 4). Three faint DDD85646-insensitive bands were removed (Fig. 4a, compare lanes 1 and 2 with lanes 3 and 4); these most likely include traces of residual GPI anchored VSG at 55 kDa 24 and thioester-myristoylated GPI-PLC at 42 kDa 25. In order to assess whether DDD85646 specifically inhibited N-[3H]-myristoylation, the same cells were labelled in parallel with [35S]-methionine. Pre-treatment of parasites with DD85646 had no effect on [35S]-methionine incorporation into proteins, showing that the compound has no effect on general protein synthesis (Fig. 4a, lanes 5 and 6).

Figure 4

Pyrazole sulfonamide series acts ‘on-target’ in the trypanosome

a. Fluorographs of SDS-PAGE gels loaded with lysates of BSF T. b. brucei cells labelled with either [3H]-myristic acid (lanes 1-4) or [35S]-methionine (lanes 5 and 6) after pre-incubation with (+) or without (−) 0.5 μM DDD85646 for 6 h. Gels were incubated with or without 0.2 M NaOH in methanol, as indicated, prior to fluorography.

b. Wild-type (“single marker”, SM) parasites and T. b. brucei over-expressing myc-tagged NMT were incubated with 0-100 nM DDD85646 for 64 h; motile cells were counted using a haemocytometer. Closed circles, T. brucei over-expressing NMT (n=3); open circles, wild-type cells (n=3). Levels of myc-tagged NMT expression were confirmed via western blotting.

c. T. b. brucei expressing ARF1Q71L (GTP-locked form of ARF1) under tetracycline control were treated with 10 nM DDD85646 for 6 h. Cells were then subjected to live/dead FACS analysis. Data shown represent mean ± s.d. from 2 independent experiments. Levels of myc-tagged ARF1 mutant expression were analysed via western blotting.

Further evidence that DDD85646 was acting on-target in the trypanosome was obtained by over-expressing TbNMT (5-fold) in a tetracycline-inducible manner; this resulted in an 8-fold reduction in DDD85646 potency against these cells (Fig. 4b).

Another independent approach was taken using one of the few known substrates of TbNMT in T. brucei, TbArf1. This protein plays a central role in endocytosis and Golgi-lysosome trafficking, where tetracycline-induced expression of a constitutively-active GTP-locked mutant (Q71L) causes rapid cell death in BSF T. brucei in an N-myristoylation-dependent manner 7. Short-term treatment (5 h) with 10 nM DDD85646 rescued these TbArf1 Q71L expressing cells from death (Fig. 4c), presumably by preventing N-myristoylation of newly produced TbArf1 Q71L protein. Although the Q71L mutant protein could not be detected by Western blotting, as reported previously 7, nonetheless the inducible expression of the related, but myristoylation blocked, G2A Q71L mutant in the presence of DDD85646 does show that its effect on cell survival was not due to interference with tetracycline-induction of these mutants.

Taken together, these data provide strong evidence that DDD85646 acts to inhibit TbNMT in BSF trypanosomes and that this is directly linked to inhibition of proliferation. These data also provide a potential link between inhibition of TbNMT and disruption of the function of the TbNMT substrate TbArf1, known to operate in protein trafficking and endocytic processes.

Inhibitor binds in TbNMT peptide pocket

The Target Product Profile for a new drug to combat HAT, as defined by Drugs for Neglected Diseases initiative, requires compounds that are safe and efficacious against both the stage 1 disease, when parasites are present in the blood, lymph and interstitial fluids, and the stage 2 disease, when parasites are also present in the CNS. The DDD85646 compound does not yet meet these criteria and will have to be optimised with respect to selectivity over human NMT and for diffusion into the CNS, while retaining excellent potency against TbNMT and BSF T. brucei cells. To achieve these objectives, a detailed understanding of the interaction between DDD85646 and TbNMT is required. Characterisation of the mode of inhibition of the early hits revealed competition with respect to peptide substrate as a likely mode of inhibition for the series. Thus, a shift in IC50 from 1 to 4.3 μM for an early hit was seen when the peptide substrate concentration in the assay was increased from 0.5 to 16 μM (Fig. 5a). Surface Plasmon Resonance (SPR) studies confirmed a 1:1 binding stoichiometry of DDD85646 with TbNMT (Fig. 5b) and showed that the binding affinity of the compound was increased in the presence of myristoyl-CoA from 33 nM to 1 nM (data not shown). Accurate determination of binding affinity via SPR also revealed a potential small window of selectivity between TbNMT and human NMT with binding constants of 1 and 14 nM, respectively. Due to tight binding of this compound (see Supplementary Information, Figure 1), this selectivity was less clear in the biochemical assay which recorded 2 and 4 nM (IC50), respectively. Furthermore, this emerging selectivity is clearly driven by differing off-rates and optimisation around this parameter will not only be important in improving selectivity, but also in sustaining a high residency time of binding to TbNMT. The structure of TbNMT has not been solved to date. However, using L. major NMT (LmNMT) as a model system, the binding of DDD85646 into the peptide substrate binding site has been confirmed by X-ray crystallography (Fig. 5c, for stereo view see Supplementary Information, Figure 4). LmNMT has 74% overall sequence identity with TbNMT and 94% identity within the peptide-binding site (see Supplementary Information, Figure 5). The binding mode shows the piperazine interacting with the C-terminal carboxylate of NMT, via a tightly coordinated water molecule as opposed to a direct H-bond. The sulfonamide makes water bridged interactions with the highly conserved residue His219 and to the backbone NH group of Gly397. The geometry of the sulfonamide creates a significant bend in the structure, allowing the pyrazole to fit into a lipophilic pocket where it acts as a hydrogen bond acceptor from Ser330. The binding mode is of particular interest because all but the latter interactions are through water-mediated hydrogen bonds; a mode not readily predicted using computational techniques. Overlaying the structure of Saccharomyces cerevisiae NMT in complex with substrate peptide (PDB 1IID) 26 with the LmNMT complex shows that DDD85646 occupies the peptide binding site with the basic piperazine moiety mimicking the N-terminus of the substrate (see Supplementary Information, Figure 6).

Figure 5

Characterisation of pyrazole sulfonamide interactions with NMT

a, DDD64558 potency against TbNMT (IC50) determined in the presence of 0.5 (closed circles) and 16 μM (open circles) CAP5.5 peptide substrate. Each data point represents mean ± s.d. (n=4).

b, Kinetics of binding of DDD85646 to TbNMT and huNMT1 determined by SPR.

c, X-ray crystal structure of DDD85646 bound to LmNMT. Left panel shows the LmNMT binding site with protein backbone (pink ribbon), solvent accessible surface (grey), DDD85646 (stick representation, carbon atoms gold, nitrogen blue, chlorine green, oxygen red and sulphur yellow), myristoyl CoA (C atoms cyan) and an omit map (Fo-Fc, 3.0 sigma) around DDD85646 (blue). Right panel shows in stick representation DDD85646 and residues forming the active site (C atoms grey). Key residues are highlighted (C atoms yellow) as are water molecules (red spheres) and H-bonds (dashed lines).

Conclusions

We have presented evidence that our model TbNMT inhibitor, DDD85646, kills BSF T. brucei by acting on TbNMT in situ. The downstream consequences of TbNMT inhibition in the parasite are likely to be multiple, since the enzyme has over 60 putative substrates 6, and this no doubt explains the speed of killing and the dramatic morphological changes observed upon treatment with this compound. The emergence of the ‘BigEye’ phenotype, and the possible link to TbArf1 activity, suggests that disturbance of endocytosis is one mechanism by which TbNMT inhibitors act in BSF T. brucei. Interestingly, knockdown of TbNMT levels in BSF T. brucei by RNAi to 16% of wild-type levels, while fatal in vitro and in vivo, does not lead to the ‘BigEye’ phenotype but rather to the accumulation of vesicles close to the flagellar pocket 27.

It is notable that, despite the small window of selectivity between human NMT and TbNMT, DDD85646 shows promising selectivity at the cellular level. One might speculate that T. brucei cells are hyper-sensitive to NMT inhibition because of unique or unusual aspects of their biochemistry and/or cell biology. In this context, the extremely high endocytic rate of BSF T. brucei (some 9 times faster than fibroblasts and 2.6 times faster than macrophages), combined with the entire endocytic/exocytic process occurring in the flagellar pocket, is noteworthy 28. The parasite’s requirement for this high endocytic rate relates to its need to remove antibody from the cell surface and to recycle the protective VSG coat 29.

Finally, we may conclude that TbNMT is one of the few T. brucei proteins that have been comprehensively validated as a drug target for HAT. The TbNMT inhibitors described meet many requirements for a greatly needed new therapeutic for HAT. Further optimisation of this series towards improved CNS penetration and selectivity is currently underway. In the meantime DDD85646 will serve as an excellent chemical tool for investigation of the biology of protein N-myristoylation across a range of organisms.