Development of Small-Molecule Trypanosoma brucei N-Myristoyltransferase Inhibitors: Discovery and Optimisation of a Novel Binding Mode

The enzyme N-myristoyltransferase (NMT) from Trypanosoma brucei has been validated both chemically and biologically as a potential drug target for human African trypanosomiasis. We previously reported the development of some very potent compounds based around a pyrazole sulfonamide series, derived from a high-throughput screen. Herein we describe work around thiazolidinone and benzomorpholine scaffolds that were also identified in the screen. An X-ray crystal structure of the thiazolidinone hit in Leishmania major NMT showed the compound bound in the previously reported active site, utilising a novel binding mode. This provides potential for further optimisation. The benzomorpholinone was also found to bind in a similar region. Using an X-ray crystallography/structure-based design approach, the benzomorpholinone series was further optimised, increasing activity against T. brucei NMT by >1000-fold. A series of trypanocidal compounds were identified with suitable in vitro DMPK properties, including CNS exposure for further development. Further work is required to increase selectivity over the human NMT isoform and activity against T. brucei.


Introduction
HumanA frican trypanosomiasis (HAT), or African sleeping sickness, is endemici ns ub-Saharan Africa, claiming the lives of about 10 000 peoplee very year. [1] The disease burden in this area is substantial, with approximately 60 million people at risk of infection.H AT is caused by two subspecies of the protozoan parasite Trypanosomab rucei gambiense and T. brucei rhodensiense,w hicha re transmitted to the humanh ost by the bite of an infected tsetse fly.I fl eft untreated the disease is often fatal. There is ac linicaln eed for more effective drug therapies, because current treatments are unsatisfactory due to toxicity, treatment failures,a nd inappropriate dosing regimensf or ar ural African setting. [1][2][3][4] The enzyme N-myristoyltransferase (NMT) is one of only af ew genetically and chemically validated drug targets in kinetoplastids, with the NMT inhibitorD DD85646 (Figure 1) being shownt oa ct on target and to cure the mouse model of stage-1 (non-CNS) T. brucei infection. [5][6][7][8] Functionally the NMT enzyme is ubiquitous and is responsible for catalysingt he cotranslational transfer of myristate from myristoyl-CoA to the Nterminal glycine residueo ft he target protein. Herein we describe the discoveryand optimisation of novel T. brucei NMT inhibitor scaffolds identifiedb yh igh-throughput screening, with appropriate physicochemical properties for oral bioavailability and CNS penetration.

Results and Discussion
Hit-to-lead chemistry In our initial programme to discover inhibitors of TbNMT, ah igh-throughput screen of our diversity set was carriedo ut.
The enzyme N-myristoyltransferase (NMT) from Trypanosoma brucei has been validated both chemically and biologically as ap otentiald rug target for human African trypanosomiasis. We previously reported the development of some very potent compounds based aroundapyrazoles ulfonamide series, derived from ah igh-throughputs creen. Herein we describe work aroundt hiazolidinonea nd benzomorpholine scaffolds that were also identified in the screen. An X-ray crystal structure of the thiazolidinone hit in Leishmania major NMT showedt he compound bound in the previously reported active site, utilis-ing an ovel binding mode. This provides potentialf or further optimisation. The benzomorpholinone was also found to bind in as imilarr egion. Using an X-ray crystallography/structurebased designa pproach, the benzomorpholinone series was furthero ptimised,i ncreasinga ctivity against T. brucei NMT by > 1000-fold. As eries of trypanocidalc ompounds were identified with suitable in vitro DMPK properties, including CNS exposure for further development. Further work is required to increase selectivity over the human NMT isoform and activity against T. brucei.
In addition to the pyrazole sulfonamide series previously reported, [8][9][10] severalo ther hits were identified. These were also investigated to determine if they offer opportunities to develop selective and blood-brain barrier penetrant TbNMT inhibitors. Twoh its around at hiazolidinone and benzomorpholinone were identified and validateda sh its ( Figure 2).
In parallel to ar e-evaluation of the originald iversity screen data, av irtuals creening exercise wasc arried out, the basis of which was known interactions betweent he pyrazoles ulfonamide DDD85646 and the enzyme (PDB code:2 WSA;F igure 3). Screeningo ur in-house database of commercially available compounds using this pharmacophore led to the discovery of three different compound classes (data not shown), including examples containing the thiazoldinone scaffold (Figure3B).

Thiazolidinone series
The initial hit expansion focusedo no ptimising the substituents around the thiazolidinoner ing. The synthetic route is shown in Scheme1and consisted of condensation of the aldehyde with an amine and thioacetic acid to give analogues 1-13 ( Table 1) in yields of 33-74 %. Work was initially carried out withoutX -ray co-crystal structures;w hen this structurali nformation was established using the Leishmania major enzyme, key early-stage molecules were also co-crystallised with the enzyme. As we discussed in ap revious publication, [10] the L. major NMT shows high sequence homologyt ob oth the T. brucei and humanN MTs. This has been an excellent system to improvet he activity of inhibitors, but given the similarities and lack of high-resolution structures of the T. brucei enzyme, it has much less use in predictings electivities.
The benzyl ring at R 2 could be replaced with af uranyl methyl group (compound 2). However, replacement with an onaromaticg roup (R 2 = ethyl, 3)l ed to ac omplete loss of activity.S imilarly,adirectly attached pyridine moiety also led to al oss of activity.T his loss is probablyd ue to ac lash of the pyridine moiety with the protein and suboptimal filling of the hydrophobic pocket with the methyl group. Ap rotein-ligand structure of 6 bound to L. major NMT (LmNMT) later confirmed this methylene-aromatic group occupies ah ydrophobic pocket in the active site (see binding mode T1, Figure4A,B). We were keen to replacet he 2-methoxyphenol group,d ue to the potentialf or de-methylation and oxidation to aq uinone. Removal of the phenolg roup from 2,a si nc ompound 5,r esulted in loss of potency,i ndicating the importance of this group for binding. The phenol hydrogenb ond donor of compounds 1 and 2 could be replaced by the indazole isostere, as in compound 6,w ithoutl oss of potency. The complexo f6 bound to LmNMT revealed binding mode T1 of the thiazolidinones eries, (Figure 4). The thiazolidinone ligand occupies as imilara rea of the peptideb inding pocket of NMT as the pyrazoles ulfonamide ligands [8] (Figure 4B). In bindingm ode T1, the thiazolidinone core of 6 packs against the side chain of Phe90, and the carbonyl group formsakey hydrogen bondingi nteraction with the side chain of Ser330. The benzyl moiety of 6 forms p-stacking interactions with the side chains of Phe88 and Phe232 (parallel and edge-face, respectively), witht he pocket enclosed by the side chains of Leu341 and Val374. The indazole N2 lone pair forms ah ydrogen bond to the backbonea mide of Asp396, ak ey interaction identified in the pyrazoles ulfonamide series. Despite the ligand being prepared withoutc hiral resolution, only the R enantiomer was observed bound in the crystal structure. Compounds 1 and 2 were assumed to have asimilarbinding mode. Scheme 1. Thiazolidinone synthesis. Reagents and conditions:a)1.aldehyde (2 equiv), amine (1 equiv), THF,08C, 5min;2.thioacetic acid (3 equiv), 0 8C, 10 min;b )polymer-supported carbodiimide (1.3 equiv), 20 8C, 16 h, 33-74 %. Figure 3. A) Structure-based pharmacophore based on the crystal structure of DDD85646 bound to LmNMT.The key pharmacophoric interactions used were:1)ahydrogen bond from the pyrazole Na tom to Ser330, showni nb lue;2)ap-stacking interaction between the pyrazole and Phe90,shown in green;3)ap-stacking interaction between the pyridine and Tyr217, shown in green;4)ahydrogen bond between the sulfonamide Oa tom and His219 and Asp396. Excludedv olumes are shown in purple. B) Virtual screening hit compound 9.
ChemMedChem 2015, 10,1821 -1836 www.chemmedchem.org  [11] [c] The assumed bindingm ode of each analogue is classified into eithero ft he two specific binding modes identifiedb yX -ray crystallography (see Figure4); this assumption was supported by the observed SAR data and by modelling thesea nalogues in PyMOL. ChemMedChem 2015ChemMedChem , 10,1821ChemMedChem -1836 www.chemmedchem.org Simultaneousr eplacement of the R 1 3-phenol-4-methoxy groups of 1 with a2 -pyridyl unit, andt runcation of the R 2 benzylgroup to ad irectly linked phenyl, resulted in compound 7 and an unexpected inversion of the binding mode from that observedf or 6,g iving rise to bindingm ode T2 ( Figure 4C). Compounds that adopted binding mode T2 show the R 1 2-pyridyl subunit forming ah ydrogen bondingi nteraction with the side chain of Ser330, andt he thiazolidinone carbonyl group forming ah ydrogen bonding interaction with the side chain of Asn376. The X-ray crystal structure also revealed the R 2 substituentt ob el ocated in the hydrophobic peptide binding groove, lying in as imilarp lane to the aryl group in the pyrazole sulfonamide series ( Figure 4). In bindingm ode T2, and in contrastt o6,t he S enantiomer was bound in the actives ite. It is unclear why compound 7 displayed selectivity for TbNMT over hNMT (Table 1). However,t his observed selectivity was lost upon removal of the (R 2 -phenyl)-3-chloro substituent, seen with 9,a lthough this is partly due to the overall diminished potencyo bserved for this compound. Substitution at the ortho (11)o rmeta (12)p ositions of the R 2 phenyl group appeared to be preferred over para substitution (10)w hich may be due to ac lash of this substituent with the side chain of Tyr217.
During our exploration of the structure-activity relationship (SAR) around the thiazolidinone scaffold, the 2-pyridylmethylene subunit of 12 was identified as the most ligand efficient R 2 substituent (LE = 0.39;T able 1). Modelling of 12 into the binding sites of 6 and 7 could explain the efficiency of binding. Assuming 12 adopted binding mode T1, there was no clear ligand-protein interaction with the His219 residue;h owever, we postulated ah ydrogen bonding interaction between the ligand and residue Asn376, with the R 2 2-pyridyl nitrogen atom as the hydrogen bond acceptor.C ompound 13 was synthesised;i ti sahybrid of compounds 12 and 1,w ith the addition of 4-hydro-3-methoxy to create an additional hydrogen bond with His219, seeking to afford as ignificant improvement in potency.X -ray crystallographyc onfirmed 13 as achieving this in-teraction and as ignificant improvement in potency( 20-fold, IC 50 :0 .27 mm)a nd as light improvement in ligand efficiency to 0.42 (Figure 5and Table 1).
The thiazolidinones eries, in particularc ompound 13,p resents ag ood startingp ointf or further optimisation.I ts hould be possible to gain additional affinity by accessing an additional local binding interaction or filling al ocal hydrophobic pocket by using as tructure-basedd esign approach. Replacement of the phenol subunit with ab ioisostere has proved achievable;t his substituent could be further optimised. [12] Most compounds were not sufficiently potent against the enzyme to give significant activity in the parasite assays. Compound 13, the most potent compound,s howed a5 0-fold decreasei na ctivity in going from enzyme to cell.H owever,f urtherw ork is required to derive compoundsw ith greater enzyme and parasite activity. . Binding mode of thiazolidinone-based ligands to LmNMT.A)Compound 6 (C atomsg old) binding to LmNMT(Ca toms grey), adoptingb inding mode T1. Hydrogen bonds are shown as dashed lines, and key residues are labelled.B)Overlay of the binding mode of 6 with ap yrazole sulfonamide ligand (C atomsc yan, PDB code:4A30). [9] C) Compound 7 (C atoms gold)b indingt oLmNMT(Ca toms grey), adopting bindingm ode T2. The image was prepared with PyMOL.  (Table 2a nd Figure 6). The initial optimisation focus for the benzomorpholinone series was to identify replacements for the potentially hydrolysable ester functional group of 14.

Synthetic route to the benzomorpholinone series
Chloroacetylchloride was reactedw ith the appropriately substituted 2-aminophenol tog ive the NH benzomorpholinone intermediate (Scheme 2). This was then alkylated with potassium carbonate as the base, with heatinginDMF.Workup and purification yieldedc ompounds 14-38.
Compound 14 was successfully co-crystallised with LmNMT. The complex shows the ligand bound in aa no rientation similar to that of the thiazolidinonel igand 13,c reating av ery similar interaction pattern ( Figure 6). The benzomorpholinone packs againstt he sidec hain of Phe90w ith the carbonyl forming ah ydrogen bond to the side chain of Ser330 ( Figure 6A). The carbonylg roup of the ester moiety forms ah ydrogen bond interaction with the side chain of Asn376, occupying as imilarp osition to the pyridyl group of compound 13 (Figure 6B).
On the fragment level, ao ne-(C1, 14)o rt wo-(C2, 16) carbon linker to the ester was equally well tolerated, with TbNMT IC 50 potencies aroundt he 10 mm level. The labile ester functional group was replaced by ap yrrolidine amide (compound 17)w ithout loss of potency,a lthought he dimethylamino amide (compound 18)showedafivefold decrease in potency.I tw as also found that C1-linked amides 19, 20,a nd 21 lost all potency. Although the reason for theses ignificantp otency losses is unclear, the decrease in lipophilicity for these particular compounds may be partly responsible for the lower potencies observed.
C1-linked heterocyclic isosteres for the ester compound 14 were tolerated, although there was ar equirement for ah eteroatom in the a position, presumably to act as ah ydrogen bond acceptor (HBA) with the side chain of Asn376. The vector and relative hydrogen bond potentialo ft he lone pair on the HBA atom, coupled with the orientation of the alkyl substituent appeared critical drivers for potency. The heterocyclic isostere with the highest potency was the methyl-isoxazole 23 (TbNMT IC 50 :2 .9 mm). This compound displayed af ourfold improvement in TbNMT potency over 14.T he thiazole (24), pyrazole (32), and oxadiazole (27)h eterocyclic analogues lost significant potency relative to compound 23.I nt he case of 24 and 32,w eh ypothesised from our binding models that the methyl group is suboptimally positioned, andf or oxadiazole 27 the oxygen is ar elativelyw eak HBA moiety.T his explanation was further supported by the recovery in potency observedf or the isomeric oxadiazole 28 (TbNMT IC 50 :7 .0 mm). In this oxadiazole isomer 28,t he nitrogen atom presumably acts as the HBA motif, and is predicted to have as imilarH BA potential as the isoxazole nitrogen atom in 23. [13,14] Therew as no potencyg ain observed upon adding furtherl ipophilic bulk to the methyl group in the b position of the heterocycle, for example homologation to isobutyl analogue in 29 or 30.F urthermore, for both compounds this substitution decreased the LE significantly.
The 2-pyridyl analogue 25 retained the potencyo ft he ester 14,b ut was found to be nearly 10-foldl ess potent than 23. This was not predicted by the HBA potential. The reason for Figure 6. Binding mode of benzomorpholinone ligands.A)Binding mode of 14 (C atomsg old) to LmNMT. B) Similarity in core binding of the benzomorpholinone of 14 (C atoms gold) andthe thiazolidinone of 13 (C atoms blue). C) Binding mode of 14 (C atoms gold) compared with the pyrazole sulfonamide ligand DDD85646 (C atomsc yan, PDB code:2 WSA). [8] The image was prepared using PyMOL.
Benzomorpholinone hit-to-lead strategy Compound 23 showeds imilarp otency and LE as those of the initial hit 14,a nd lacks the metabolically and chemically labile ester functional group. This scaffold was developed further in ah it-to-lead campaign to investigate other substitution vectors using ar ational structure-based design approachb ased on X-ray crystallography data. This approach used the binding mode overlayo ft he pyrazole sulfonamide inhibitorD DD85646 with 14 ( Figure 6). From our previous work on the pyrazoles ulfonamide series,a ppending ab asic moiety to reach the C-terminal carboxylic acid gave an approximate 1000-fold increase in potency. [8][9][10] Using this hypothesis, cross-over compounds were designed to target nanomolar efficacy in the parasite growth assay,g ood pharmacokinetic parameters, and selectivity for parasite efficacy (relative to hNMT enzymei nhibition) to identify compoundss uitablef or evaluation in an animal model of HAT. The selectivity ratio of T. brucei cell versus hNMT enzymep otencyw as used as ap rediction for the in vivo therapeutic window in the pyrazole sulfonamide series, and formed the basis of ranking and selecting ab ack-up series for lead optimisation (Table 3). [10] Hit-to-lead discussion Investigation of simple halogen and methyl group substitution around the core benzomorpholinone scaffold indicated substitution with ah alogen (Cl or Br), as in 36 and 37,i st olerated at the 7-position ( Table 3). The 8-bromoa nalogue 38 showeds imilar potencyt o7 -bromo compound 37.T his implies that the 7-or 8-positions are suitable vectors for extension of 23 toward the Cterminus, in agreement with the X-ray crystal structure ( Figure 6).
Extended analogues, postulated to achieve the interaction with the C-terminal residuei nt he active site, were designed and synthesised.D ata are listed in Table 3. These analogues weredesigned to interact directly,o rt hrough aw ater-mediated hydrogen bond, with the C-terminal carboxylate group to obtain potency gains as observed for the previously reported pyrazolesulfonamide series. [8,9] Crystal structure analysis of 44 in complex with LmNMT confirmed that this interaction was achieved ( Figure 7). The benzomorpholinone core of 44 retained the binding mode of the original hit 14,w ith the methyl isoxazole packing against Phe232 and the nitrogen lone pair forming ah ydrogen bond to Asn376. The substituent at the 7-position of the ben- [b] Ligande fficiency (LE), determined for compounds with T. brucei NMTp otency < 50 mm,was calculated as 0.6 ln(IC 50 )/(heavya tom count). [11] ChemMedChem 2015, 10,1821 -1836 www.chemmedchem.org zomorpholinone extended down the peptideb inding groove presenting the methyl-piperidine group towardt he C-terminal carboxylate. The piperidine nitrogen did not interact directly with the protein chain, but hydrogen bonded to ah ighly coordinated water molecule, an interaction observed in previous series of NMT inhibitors. [8,9] The potencies of the extended analogues 39-44 all indicatet his interaction has been achieved, with all six analogues showing > 100-fold improvement in potency at TbNMT. Notably,p otencies of < 0.002 mm were at the limit of the enzymea ssay'ss ensitivity; compounds at this end of the potencyr ange are probably subnanomolar inhibitors. In previous works [8][9][10] we reported that there is ag ood correlation between activity against the parasite and inhibition of TbNMT,s o we reported as electivity measure using the EC 50 value against the parasite in predicting selectivity. [10] Relative selectivities calculated as (hNMT IC 50 )/(T. brucei cell assay EC 50 )w ere used to comparec ompounds. [10] Whilst it is unknown if inhibition of mammalian NMT is dose limiting or whether the compounds have off-target effects, this "selectivity" value turned out to be better at predicting the "therapeutic ratio" in mousem odelso fi nfection than as imple ratio of cellular activities. For example, 44 was found to be > 10-fold more potent in the parasite growth assay than the other five extended compounds. Unfortunately, the hNMTp otency of 44 was correspondingly increased as well (hNMT IC 50 :0 .04 mm), so althought he cellular selectivity (MRC5c ell/T. brucei cell) was promising at 1400-fold, the more informative hNMT IC 50 /T. brucei cell assay EC 50 (in vivo predictive) selectivity index was only sixfold. Compound 40 had the highest hNMT IC 50 /T. brucei cell assay EC 50 selectivity index of 13-fold, but unfortunately this compoundh ad ap oorer T. brucei cellular efficacy (EC 50 :8 0nm)a nd was not sufficiently potent to progress into an in vivo efficacy study.  [11] [c] Selectivity ratio: hNMT (EC 50 )/T. brucei cell (EC 50 ). The lead compounds showedagood balance of physicochemical properties (Table 4), which were within acceptable lead optimisation limits.L ipophilicity was clearlyh igher for the Nmethyl analogues,a lthough the clogD values were stilla cceptable. The mouseh epatic microsomal intrinsic clearance was high for N-methyl analogues, mostp robably driven by N-demethylation to the secondary amines.T he secondary amines (e.g., 39)w ere inherently more metabolically stable and still showede nzyme and T. brucei cellular activity in the nanomolar range (Table 3). Interestingly,w eo bserved that the vector of substitution from the benzomorpholinone core had as ubstantial effect on the plasma protein binding (PPB) of these compounds and their ability to cross the blood-brain barrieri ni nv ivos tudies, where the 8-substitutedc ompound had am uch higher brainto-blood ratio (B/B = 27) than the 7-substituted compound (B/ B = 0.12). The latter property is essential for progression into lead optimisation and for candidates suitable for clinical trials, as the parasite has already invaded the CNS before mostp atients present for treatment.

Conclusions
Compounds around two different cores, thiazolidinones and benzomorpholinones,w ere validated as hits for TbNMT.T he benzomorpholinones originated from HTS, and the thiazolidinones were discovered by HTS and virtualscreening in parallel. Interestingly,t he thiazolidinone compounds displayed two distinct binding modes with different enantiomersb ound ( Figure 4). One of the binding modesw as very similar to that adopted by the benzomorpholinones ( Figures6and 7). All binding modes of the described compounds differ from the binding modeso ft he previously described pyrazole sulfonamide compounds ( Figures 4B and 7B), or an umber of structures of Leishmania NMT with various ligands that were published subsequent to the work reported herein. [15,16] This enabled us to explore new vectors andi nteractions when attemptingt oo ptimise the affinitieso ft he hit compounds.
The thiazolidinoneh it 1, which displayed relativelyw eak TbNMT potencyw ith no selectivity over hNMT,w as expanded into ah it series of compounds using ao ne-pot,t wo-stepr eaction. The indazole isostere 6 for the methoxyphenol was quickly identified as being equipotent and more selective( displaying no measurable inhibition of hNMT). The discoveryo fab inding mode switch for the thiazolidinone in compound 7 and the highly ligand efficient lead 13 provides ap otentials tartingp oint for lead optimisation chemistry,a lthough cellulare fficacy was ap otential issue for this series (best in class T. brucei EC 50 :6 .3 mm). Further optimisation of 13 should be possible with as tructure-based design approacht ot arget additional interactions in the binding site. Alternatively,f ollowingo nf rom SAR observations with the pyrazole sulfonamide and benzomorpholinones eries, ah ydrogen bonding/salt bridge ligand-protein interaction with the C-terminal carboxylate group couldb et argeted. This should lead to nanomolar TbNMTp otencies (as previously observed), which should afford potent parasite cell activities at T. brucei. [8,9] A suitable group could be designed, with correct vector and linker to achievet his interaction, as seen in the benzomorpholinone series hit-to-lead strategy.H owever,t his strategy is not withoutr isk, because data appeart os uggest that targeting the C-terminal residuef or al igand-protein interaction may also result in correspondingly potent hNMT inhibition, aso b- www.chemmedchem.org served with the benzomorpholinones eries (Table 3). This could lead to potential issues with selectivity and toxicity (through mammalian NMT inhibition). The thiazolidinone scaffold also offers the opportunity to explore previously unexplored vectors, with the possibility of identifying new protein-ligandi nteractions which could deliver selective potency at TbNMT (over hNMT).
On the benzomorpholinone series, we started with the lowmolecular-weight, ligand-efficient hit compound 14.W ei dentified several bioisosteric replacements of the chemically labile ester functionality,t wo of which additionally displayed improved TbNMT potencya nd good ligand efficiency.Astructure-based drug design approachw as instigated, using 23 as the parentt emplate for substitution. Vectors for attaining substitutiont oward the protein's Cterminus were identified by Xray crystallography and from initial SAR investigations around the core scaffold. Six compounds, 39-44,w ere described which appear to have achieved this additional ligand-protein interaction, and which showed the expected increase in potency at TbNMT. Within this series we developed TbNMT enzyme potencyf rom the 10 mm level to < 0.002 mm.T he enzyme and cellular potencies correlated directly,i ndicating an on-target effect of killing the parasites in culture. Due to high structural homology between the human and T. brucei NMT enzyme orthologues, improvements in potencyt oward the sub-10 nm levels at TbNMT were often accompanied by ac orresponding improvement in hNMTinhibition (Table 3). Compound 42 gave some indicationt hat TbNMT/hNMTselectivity may be achievable within this series,a lthough 42 itself had insufficient cellular potencyt oprogress furthera nd into in vivo efficacy studies.
In compounds from these two series there was a3 0-to 50fold decreasei na ctivity going from TbNMT enzyme to T. brucei cell efficacy,w hich furthern arrowed selectivity over mammalian NMT inhibition;this may have the potentialtodrive toxicity. Compounds from the benzomorpholinone series afforded potent antiparasiticc ell activities (T. brucei EC 50 :0 .007 mm for 44); they display promising in vitro DMPK profiles, showing this series has the potential to be optimised further to deliver orally activeantiparasiticcompounds. The 8-substitutedbenzomorpholinone series in particulars hows good potentialf or stage-2 in vivo efficacy,a lthough improvements in selectivity would be neededt oe xtend the therapeutic window and to allow higherd ose levels in order to maximise the chances of curing stage-2 infectedm ice. It is unclear at this time how to rationally achieve improved selectivity to the required levels, althought his could be possible by using currently undiscovered selectivity pockets and protein secondary structure differences between T. brucei and human NMT.
Startingf rom the singleton HTS hits 1 and 14,a nd the virtual screening hit 9,w eh ave described the development of two potent TbNMT enzymei nhibitor series, which provide good startingp oints for drug discoveryp rogrammes for the treatment of stage-1 and stage-2 HAT. These provide novel scaffolds and give opportunities forexploring new vectors.

Experimental Section
Chemistry Chemicals and solvents were purchased from Aldrich Chemical Co., Fluka, ABCR, VWR, Acros, Fisher Chemicals, and Alfa Aesar,a nd were used as received unless otherwise stated. Air-and moisturesensitive reactions were carried out under an inert atmosphere of argon in oven-dried glassware. Analytical thin-layer chromatography (TLC) was performed on pre-coated TLC plates (layer 0.20 mm silica gel 60 with fluorescent indicator UV254, Merck). Developed plates were air dried and analyzed under aU Vl amp (l 254/ 365 nm). Flash column chromatography was performed using prepacked silica gel cartridges (230-400 mesh, 40-63 mm, SiliCycle) (unless otherwise stated) using aT eledyne ISCO Combiflash Companion or Combiflash Retrieve. 1 Ha nd 13 CNMR spectra were recorded on aB ruker AvanceII5 00 spectrometer ( 1 Ha t5 00.1 MHz, 13 Ca t1 25.8 MHz), or aB ruker DPX300 spectrometer ( 1 Ha t 300.1 MHz). Chemical shifts (d)a re expressed in ppm recorded using the residual solvent as internal reference in all cases. Signal splitting patterns are described as singlet (s), doublet (d), triplet (t), quartet (q), pentet (p), multiplet (m), broad (br), or ac ombination thereof. Coupling constants (J)a re quoted to the nearest 0.1 Hz. LC-MS analyses were performed with either an Agilent HPLC 1100 series instrument connected to aB ruker Daltonics MicrOTOFo ra n Agilent Te chnologies 1200 series HPLC connected to an Agilent Technologies 6130 quadrupole spectrometer,w here both instruments were connected to an Agilent diode array detector.L C-MS chromatographic separations were conducted with aW aters Xbridge C 18 column, 50 mm 2.1 mm, 3.5 mm particle size;m obile phase:H 2 O/MeCN + 0.1 %H COOH, or H 2 O/MeCN + 0.1 %N H 3 ; linear gradient from 80:20 to 5:95 over 3.5 min and then held for 1.5 min;f low rate:0 .5 mL min À1 .A ll tested compounds had am easured purity of 95 %( by TLC and UV) as determined by this analytical LC-MS system. High-resolution electrospray MS measurements were performed on aB ruker Daltonics MicrOTOF mass spectrometer.M icrowave-assisted chemistry was performed using aB iotage initiator microwave synthesiser.
General procedure for the synthesis of thiazolidinones:T o ac ooled solution of aldehyde (2 mmol) in THF (5 mL) at 0 8C, was added amine (1 mmol) and the reaction was stirred for 5min. Thioacetic acid (3 mmol) was then added, and stirring continued for 10 min.
Polymer-supported carbodiimide (1.33 mmol, 1.33 mmol g À1 loading) was then added, the reaction warmed to RT,a nd stirring continued for 16 h. The reaction was filtered, the resin washed (THF,M eOH) and the filtrate concentrated in vacuo. The resultant crude residue was taken up in CH 2 Cl 2 ,w ashed (10 % citric acid (aq), water,s aturated sodium bicarbonate (aq), brine), dried (MgSO 4 )a nd concentrated in vacuo. The resultant crude residue was purified by column chromatography (0-100 %E tOAc/ hexane) to give the title compound.

Virtualscreening
Preparation of small-molecule database:Asubset of the Drug Discovery Unit Dundee in-house database containing 2.2 million commercially available compounds was created using the criteria: [18] -No unwanted groups; -No primary and secondary amides, or sulfonamides to avoid high polar surface area; -Number of heavy atoms between 10 and 2; -clogP in the range of 0-4; -Fewer than five ring systems; -Fewer than eight rotatable bonds; -Fewer than seven HBAs, fewer than four HBDs; -Sum of HBAs and HBDs > 1b ut < 9; -No fused ring systems. ChemMedChem 2015ChemMedChem , 10,1821ChemMedChem -1836 www.chemmedchem.org In total, 145 127 molecules were contained in this subset. Subsequently,p rotonation states and tautomers were calculated using in-house scripts based on OpenEye's python toolkit (OEChem, version 1.7.2.4, OpenEye Scientific Software Inc.,S anta Fe, NM (USA), 2009 (http://www.eyesopen.com), whereas for acidic groups with estimated pK a values between 5a nd 9, both the neutral and charged forms were stored. Corina (Molecular Networks, Germany) was used to generate three-dimensional coordinates from smiles.
Pharmacophore search:T he module UNITY in SYBYL 8.0 was applied to perform af lexible 3D pharmacophore search. The crystal structure of LmNMT in complex with ap yrazole sulfonamide (PDB code:4 A30) was used as receptor.A cceptor atoms, donor sites, aromatic features, and exclusion volumes were selected to create the UNITY query as described in the Results and Discussion section ( Figure 3). The UNITY database was created by uploading the mol2 file that was created by Corina. The import options "perceive chirality at carbons", "perceive chirality at nitrogen and phosphorous", and "perceive chirality at double bonds" were chosen;8 773 compounds passed the pharmacophore filter and were subsequently clustered using the Jarvis-Patrick algorithm of the Daylight clustering package (Daylight, Aliso Viejo, CA, USA). Representative hits per cluster were minimised in the binding site using the Moloc MAB force field. [19] The binding modes obtained were visually inspected, and al ist of 200 compounds was selected for purchase.

Drug metabolisma nd pharmacokinetics (DMPK)
In vitro intrinsic clearance assay (CL int (m)):T est compound (0.5 mm) was incubated with female CD1 mouse liver microsomes (Xenotech LLC;0 .5 mg (mL 50 mm potassium phosphate buffer) À1 ,p H7.4), and the reaction was started by the addition of excess NADPH (8 mg (mL 50 mm potassium phosphate buffer) À1 ,p H7.4). Immediately,a tt ime zero, then at 3,6,9,15 and 30 min, an aliquot (50 mL) of the incubation mixture was removed and mixed with acetonitrile (100 mL) to stop the reaction. Internal standard was added to all samples, the samples were centrifuged (10 min, 5 8C, 3270 g)t o sediment precipitated protein and the plates then sealed prior to UPLC-MS/MS analysis using aQ uattro Premier XE (Waters Corp., USA).
XLfit (IDBS, UK) was used to calculate the exponential decay and consequently the rate constant (k)f rom the ratio of peak area of test compound to internal standard at each time point. The rate of intrinsic clearance (CL int :[ mL min À1 (g liver) À1 ]) of each test compound was then calculated using Equation (1), in which V [mL (mg protein) À1 ]i st he incubation volume per mg protein added, and microsomal protein yield is taken as 52.5 mg protein per gram of liver.V erapamil (0.5 mm)w as used as ap ositive control to confirm acceptable assay performance.
In vitro plasma protein binding experiments:I nb rief, a9 6-well equilibrium dialysis apparatus was used to determine the free fraction in plasma for each compound (HT Dialysis LLC, Gales Ferry,C T, USA). Membranes (12)(13)(14) were conditioned in deionised water for 60 min, followed by conditioning in 80:20 deionised water/EtOH for 20 min, and then rinsed in isotonic buffer before use. Female CD1 mouse plasma was removed from the freezer and allowed to thaw on the day of experiment. Thawed plasma was then centrifuged (Allegra X12-R, Beckman Coulter,U SA, 10 min, 5 8C, 3270 g), spiked with test compound (10 mgg À1 ), and 150 mL aliquots (n = 6r eplicate determinations) loaded into the 96-well equilibrium dialysis plate. Dialysis versus isotonic buffer (150 mL) was carried out for 5h in at emperature-controlled incubator at 37 8C( Barworld Scientific Ltd.,U K) using an orbital microplate shaker at 125 rpm (Barworld Scientific). At the end of the incubation period, aliquots of plasma or buffer were transferred to micronic tubes (Micronic B.V., the Netherlands), and the composition in each tube was balanced with control fluid such that the volume of buffer to plasma was the same. Sample extraction was performed by the addition of 400 mLa cetonitrile containing an appropriate internal standard. Samples were allowed to mix for 1min and then centrifuged at 3000 rpm in 96-well blocks for 15 min (Allegra X12-R, Beckman Coulter,U SA). All samples were analysed by UPLC-MS/MS on aQ uattro Premier XE Mass Spectrometer (Waters Corp.,U SA). The unbound fraction was determined as the ratio of the peak area in buffer to that in plasma. Mouse brain penetration studies:C ompounds 41 and 44 were dosed as ab olus solution intravenously at 1mgf ree base per kg (dose volume:5mL kg À1 ;d ose vehicle:1 0% DMSO, 95 %s aline or 10 %D MSO, 40 %P EG400, 50 %s aline) to female NMRI mice (n = 6). At 5a nd 30 min (compound 41 only) following intravenous bolus injection of test compound, mice (n = 3p er time point) were placed under terminal anaesthesia with isofluorane. Ab lood sample was taken by cardiac puncture into two volumes of distilled water and the brain removed. After suitable sample preparation, the concentration of test compound in blood and brain was determined by UPLC-MS/MS using aQ uattro Premier XE (Waters, USA). For each mouse at each time point, the concentration in brain (ng g À1 )w as divided by the concentration in blood (ng mL À1 ) to give ab rain/blood ratio. The mean value obtained was quoted.
Ethics.A ll regulated procedures on living animals were carried out under the authority of ap roject license issued by the Home Office under the Animals (Scientific Procedures) Act 1986, as amended in 2012 (and in compliance with EU Directive EU/2010/63). License applications will have been approved by the University'sE thical Review Committee (ERC) before submission to the Home Office. The ERC has ag eneral remit to develop and oversee policy on all aspects of the use of animals on University premises and is as ubcommittee of the University Court, its highest governing body.