Structure-Based Design of Potent and Selective LeishmaniaN-Myristoyltransferase Inhibitors

Inhibitors of LeishmaniaN-myristoyltransferase (NMT), a potential target for the treatment of leishmaniasis, obtained from a high-throughput screen, were resynthesized to validate activity. Crystal structures bound to Leishmania major NMT were obtained, and the active diastereoisomer of one of the inhibitors was identified. On the basis of structural insights, enzyme inhibition was increased 40-fold through hybridization of two distinct binding modes, resulting in novel, highly potent Leishmania donovani NMT inhibitors with good selectivity over the human enzyme.


■ INTRODUCTION
The leishmaniases are a spectrum of infectious diseases caused by protozoan parasites of the genus Leishmania. Cutaneous leishmaniasis (CL), caused mainly by Leishmania major (Lm), can lead to permanent scarring and disfiguration, while visceral leishmaniasis (VL), caused mainly by Leishmania donovani (Ld), is often fatal due to failure of the host immune system. The leishmaniases are endemic in 88 countries, 72 of which are low-income, 1 and is a major health issue with an estimated 0.2− 0.4 million cases of VL, 0.7−1.2 million cases of CL, and a conservative estimate of 20000−40000 deaths per year. 2 Treatment of leishmaniasis has previously been dominated by the use of pentavalent antimonials which are toxic, painful to administer, and require long treatment regimens; 3 resistance has also developed to these antimonials in India. 4 Some progress has been made in the last 10 years in the development of safer, more easily applied therapeutics with the development of lipid formulations of amphotericin B, miltefosine, and paromomycin. However, side effects are common and resistance to these therapies may still be a problem, 5 thus the need for new antileishmanials remains high. 6,7 Despite these issues, development of new antileishmanial drugs is limited 8 and compounded by challenges of cell permeability. The amastigote form of the parasite most relevant to human disease resides within an acidic parasitophorous vacuole inside host cells, 9 and the parasite bears a glycoinositolphospholipid coat that could limit uptake of xenoantibiotics. 10 N-Myristoyltransferase (NMT), an enzyme ubiquitous in eukaryotes, catalyzes the transfer of myristate (a 14-carbon fatty acid) to the N-terminal glycine of target proteins, 11 either cotranslationally 12 or post-translationally. 13 Between 0.5% and 3% of the cellular proteome is predicted to be Nmyristoylated, 14 and this modification is vital for multiple regulatory processes, including protein−protein interactions and protein stability. 15−17 Inhibition of NMT therefore has pleiotropic effects on cellular function. NMT has been shown to be essential in a range of parasitic organisms including Leishmania, 18 and small-molecule cytotoxic inhibitors have been developed for NMTs in parasitic organisms including Trypanosoma brucei 19 and Plasmodium species. 20−22 Inhibition of Leishmania NMT therefore represents a rational drug target for development of new therapeutics for this neglected tropical disease. 14,23,24 The NMT enzyme operates via a Bi−Bi mechanism, with myristoyl CoA (MyrCoA) binding to the enzyme first and inducing a conformational change before binding of the peptide substrate. The myristate group is then transferred to the Nterminal glycine of the peptide before sequential release of the myristoyl peptide and reduced CoA products. 25,26 The structures of several parasitic NMTs have been reported 19,27,28 and show a conserved binding site for MyrCoA. The peptidebinding region is less conserved between different species and therefore presents a target for selective inhibition of NMTs from different species. 29 A recently published high-throughput screen (HTS) of a diverse subset of the Pfizer corporate collection against LdNMT, Plasmodium falciparum NMT, and the two human isoforms (HsNMT1 and HsNMT2) revealed four novel series of Leishmania-selective NMT inhibitors. 30 Here we report the development of highly potent LdNMT inhibitors based on structure-guided fusion of two of these series; piperidinylindoles, exemplified by PF-03393842 1, and aminoacylpyrrolidines, exemplified by PF-03402623 2 and PF-03402619 3 (Chart 1).

■ RESULTS AND DISCUSSION
Synthesis and Validation of Hits. To validate the HTS results, synthesis of both piperidinylindole 1 and the most potent aminoacylpyrrolidine 2 was carried out. Synthesis of 1 was achieved in four steps from 5-nitro indole (Scheme 1). Condensation of 5-nitro indole 4 with N-Boc-4-piperidone, followed by concurrent reduction of the resulting double bond and nitro group, yielded amine 6. 31 Reaction with parafluorophenylacetyl chloride followed by Boc deprotection gave piperidinylindole 1.
Compound 2 was synthesized as a mixture of two diastereoisomers, as it was unclear from the original report whether the stereochemistry at the pyrrolidine ring in the HTS hit was relative or absolute. It was envisaged that the preferred stereochemistry could be identified by cocrystallization of the mixture with Leishmania NMT. The pyrrolidine core was accessed as a racemic mixture by cycloaddition of benzyl-(methyoxymethyl)[(trimethylsilyl)methyl]amine 9 and transmethyl 4-chlorocinnamate 8 in the presence of catalytic TFA to give trans-pyrrolidine 10 (Scheme 2). 32,33 The benzyl protecting group was removed using 1-chloroethyl chloroformate 34 to avoid the use of reductive methods in the presence of the aromatic chlorine prior to amide coupling with acid 12 to obtain amide 13. Reduction of the ester followed by Boc deprotection proceeded smoothly to afford 2 as a mixture of two diastereoisomers.
Testing of compounds 1 and 2 against LdNMT and HsNMT1 in our previously reported CPM assay 35 confirmed the results of the HTS, with both compounds showing selectivity for LdNMT over HsNMT1 ( Table 1). The hits were also tested against related LmNMT, which has been shown to be more amenable to X-ray crystallography 19,36 and were shown to have comparable activity to LdNMT.   Compounds were also tested against extracellular amastigotes of Leishmania donovani and against bone marrow derived macrophages to determine toxicity (Table 1). 37 Compound 1 displayed no cell activity up to 30 μM, although no toxicity was observed. Compound 2 showed an EC 50 between 10 and 30 μM, however, the compound was also toxic to macrophages at this concentration.
X-ray Crystallography. Our first strategy to optimize these NMT inhibitors was to drive down enzyme potency using structure-guided design. To elucidate the binding mode of the HTS hits and the preferred stereochemistry of 2, crystal structures of ternary complexes of LmNMT (97% sequence homology with LdNMT) and myristoyl-CoA cofactor were obtained for both resynthesized hits, as recently reported. 36 Both inhibitors were shown to bind in the peptide binding region. The structure of compound 1 bound to LmNMT revealed a direct interaction between the basic piperidine nitrogen and the C-terminal carboxylate of the enzyme (Leu421) (Figure 1). This type of charge−charge interaction has previously been observed with other NMT inhibitors in Plasmodium NMT 20,21 and via a bridging water molecule in LmNMT. 19 The indole adopts an equatorial position off the piperidine ring in a hydrophobic pocket, and the amide carbonyl is orthogonal to the indole ring, forming hydrogen bonds to Tyr345 and Asn376.
The cocrystal structure of 2 bound to LmNMT displays a unique binding mode compared to previously reported NMT inhibitors; the conformation of the inhibitor appears to be governed by a hydrophobic collapse 38 that folds the aromatic rings into a hairpin conformation about the flexible linker, with the chlorophenyl substituent of the pyrrolidine ring sandwiched between the edge of Tyr345 below and Tyr217 above. The inhibitor takes up a compact conformation in which its surface area is almost completely buried by the protein and MyrCoA. Interestingly, the key charge−charge interaction between the basic amine and Leu421 is not seen (Figure 2). Instead, the primary amine is adjacent to the thioester of MyrCoA and makes bridging contacts with the backbone carbonyl of Thr203 and the side chain of Asn167. The hydroxyl group is actually closest to the C-terminal leucine carboxylate (2.6 Å), and there is a potential hydrogen bond between the amide carbonyl and Thr203 ( Figure 2).
As expected, the crystal structure shows a single diastereoisomer (2a, Scheme 3) bound to the enzyme. To confirm that 2a is the most active isomer, both diastereoisomers (2a and 2b) were synthesized separately using enantiopure oxazolidinone 17 for the cycloaddition reaction (Scheme 3).
Cycloaddition yielded diasteroisomers 18a and 18b, which could be separated by column chromatography and were assigned by comparison with reported 1 H NMR data 39 (Scheme 3). Removal of the oxazolidinone gave esters 10a and 10b from which 2a and 2b were synthesized, respectively, using the route detailed in Scheme 2.
Enzyme inhibition assays confirmed that diastereoisomer 2a was more active with an IC 50 of 25 nM against LdNMT, with 2b exhibiting 60-fold lower potency and 4-fold lower selectivity for LdNMT over HsNMT1 (Table 2).
A crystal structure was also obtained for 2b bound to LmNMT. The structure shows a similar hydrophobic collapse of the ligand and that the key functional groups (the primary   amine and alcohol) of both diastereoisomers are superimposed in the active site (circled, Figure 3). However, as a result of maintaining these interactions, the scaffold is twisted such that the amide carbonyl no longer forms the hydrogen bond with Thr203 seen in the structure of 2a.
Hybridization of Binding Modes. Comparison of the distinct binding modes of hits 1 and 2 showed that the benzoring of the indole in 1 and the aromatic substituent of the pyrrolidine in 2 bind in the same region (Figure 4a). For this reason, it was hypothesized that addition of a para-fluorophenyl acetamide ortho-to the chlorine atom in this ring in compound 2 may significantly improve potency (Figure 4b). This could potentially introduce hydrogen bonding between the acetamide carbonyl and Tyr345 and Asn376 and allow 2 to extend into the same hydrophobic pocket as 1. To determine whether an acyl group would be sufficient to increase potency by addition of the hydrogen bond, or whether the hydrophobic bulk of the para-fluorophenyl group would be required, both the acyl and para-fluorophenyl acetamide derivatives of 2 were synthesized (19 and 20, respectively, Figure 4b).
To allow the addition of an acetamide in the correct position, synthesis of the pyrrolidine core was carried out using 3-nitro, 4-chloro cinnamate 21 (Scheme 4). Selective reduction of the nitro group with tin(II) chloride then provided an aniline moiety for formation of the required amide bond. Both the acyl derivative 19 and para-fluorophenyl derivative 20 were synthesized using the route shown in Scheme 4.
Testing of the hybrid compound 19 showed that the acyl group was tolerated but did not improve potency. However, addition of the para-fluorophenyl group in compound 20 led to an IC 50 at the lower measurement limit of the biochemical assay. In this case, converting to K i using the Cheng−Prusoff equation for tight binders is more informative 40,41 (Table 3). Introduction of the para-fluorophenyl group led to a 40-fold decrease in K i against LdNMT (K i = 1.6 nM). Alignment of the X-ray crystal structures of 20, 1, and 2 bound to LmNMT demonstrates that 20 binds as designed and that all interactions with the enzyme are conserved ( Figure 5 and Figure S1, Supporting Information).
Cell Testing. Despite the greater enzyme inhibition achieved by hybrid compound 20, no improvement in cell-      based activity was seen (Table 3). This highlights that simply driving down enzyme potency in this compound series is insufficient to increase cellular activity against this challenging target organism. The diastereoisomers of 2 were also tested separately, and both displayed an EC 50 of 10−30 μM and an LD 50 of 12−24 μM, demonstrating that both activity and toxicity are unrelated to NMT inhibition for these compounds. We hypothesized that the lack of cell-based activity for this series of compounds is due to lack of cellular uptake and thus, insufficient target engagement. Compounds 1 and 2 and derivatives synthesized here all contain a basic center (pK a 10.0 and 8.9 respectively) which would be charged at physiological pH, with a potentially adverse effect on membrane permeability.

Journal of Medicinal Chemistry
Replacement of the Primary Amine. As the crystal structures of the aminoacylpyrrolidines show that the amine does not make the key charge−charge interaction with the Cterminal carboxylate observed previously in other series, we considered replacing the primary amine with a less basic moiety. It was envisaged that the amine could be replaced with an alcohol without loss of hydrogen bonding, potentially generating a potent, neutral NMT inhibitor. To synthesize this neutral compound, acid 35 was synthesized from 4chlorophenylacetyl chloride 32. Reaction with Meldrum's acid followed by hydrolysis gave ketone 33, which was subsequently reduced. Hydrolysis of the resulting ester gave the acid 35 (Scheme 5).
This acid was then used to synthesize the alcohol analogue of hybrid 20 (Scheme 6). As acid 35 was synthesized as a racemic mixture, pyrrolidine 39 was synthesized as a single enantiomer in order to give the hybrid alcohol 43 as a mixture of only two diastereoisomers (Scheme 6). The hybrid alcohol 43 was tested for its enzyme activity and showed reduced activity compared to the corresponding amine (Table 3). However, the activity for this neutral compound is comparable to the original primary amine hit 2, and selectivity over HsNMT1 is maintained.
The crystal structure of alcohol 43 bound to LmNMT appeared to overlay well with that of amine 20 ( Figure S2, Supporting Information, PDB code 4cyq). However, closer inspection revealed a slight difference in the position of the amine versus the alcohol in these two structures. The primary amine of 20 forms a hydrogen bond to the backbone carbonyl group of Thr203 at a distance of 2.9 Å. When this amine is replaced with an alcohol in compound 43, the corresponding oxygen is 3.5 Å away, reducing its potential to hydrogen bond.
Testing against extracellular amastigotes showed no improvement in activity for compound 43 compared with the original hits or the hybrid amine (Table 3). Metabolic chemical tagging 22 in Ld amastigotes ( Figure S3, Supporting Information) demonstrated that despite replacement of the primary amine, target engagement consistent with a K i of 59 nM was not achieved, supporting our hypothesis that the lack of cellbased activity for these compounds is due to poor cellular uptake.
Despite the potential of NMT as a drug target in Leishmania, these organisms are known to be difficult to target, due in part to their cell surface coats, a key component of Leishmania virulence and survival. 9,10 Advances in potency and particularly physicochemical properties will be required to progress this series of compounds and to chemically validate Leishmania NMT as a drug target in vivo.

■ CONCLUSION
Two Leishmania NMT-selective HTS hits have been resynthesized and their activities validated. Crystal structures of these inhibitors identified their binding modes and, in the case of compound 2, identified the active diastereoisomer. The crystal structures were used to increase enzyme affinity through hybridization of the two independent binding modes, and this led to the discovery of a highly potent inhibitor of LdNMT. The unusual binding mode of the aminoacylpyrrolidines allowed the replacement of the primary amine, leading to compound 43, a potent and neutral NMT inhibitor. Although poor uptake appears to lead to a lack of cell activity for these compounds, elucidation of the binding modes of these inhibitor series along with their hybridization provides a useful starting point for the development of LdNMT inhibitors with improved physicochemical properties.

■ EXPERIMENTAL SECTION
The purity of final compounds was determined by reversed-phase LC-MS on a Waters 2767 system and was ≥95% for all tested compounds.