Malaria parasites

In 2012, Jones et al. reported the first global study of S-palmitoylation in the asexual stage of the malaria parasite P. falciparum, by applying both MTCC with the well-established palmitate analogue 17-ODYA (4) (Fig. 2) and ABE (Jones et al. Reference Jones, Collins, Goulding, Choudhary and Rayner2012). Quantitative comparison between sample and control in both approaches was carried out using SILAC (stable isotope labelling of amino acids in culture) and the study identified >400 potential palmitoylated proteins. The authors combined ABE with the compound 2-bromopalmitate (2-BP) to analyse the degree to which specific palmitoylations are dynamic. A significant caveat to these results is that 2-BP is not a specific thioesterase inhibitor and has broad non-specific reactivity, including particularly on lipid metabolic pathways (Coleman et al. Reference Coleman, Rao, Fogelsong and Bardes1992; Davda et al. Reference Davda, El Azzouny, Tom, Hernandez, Majmudar, Kennedy and Martin2013; Zheng et al. Reference Zheng, DeRan, Li, Liao, Fukata and Wu2013); the continued use of this molecule despite its potent promiscuity is symptomatic of the lack of well-characterized specific inhibitors for palmitoyl transferases. Despite this issue, the study of Jones et al. was nevertheless a landmark application of ABE and MTCC in a protozoan parasite, and elegantly demonstrates the complementarity of the two techniques (Fig. 3A).

Fig. 3. (A) Comparison of P. falciparum proteins identified via ABE and MTCC (with 17-ODYA (4); workflows: Figs 1 and 2A) techniques in the study of Jones et al. (B) Comparison of T. gondii proteins identified in the studies of Caballero et al. and Foe et al. which used an ABE approach (Fig. 1) and a global analysis of 17-ODYA (4) tagged proteins in the presence and absence of HA, respectively. Total rather than high-confidence hits were used for the analysis shown. (C) Comparison of the 321 putative P. falciparum palmitoyl proteins (from Jones et al. Reference Jones, Collins, Goulding, Choudhary and Rayner2012) that have T. gondii orthologues with putative T. gondii palmitoyl proteins (Foe et al. Reference Foe, Child, Majmudar, Krishnamurthy, van der Linden, Ward, Martin and Bogyo2015; Caballero et al. Reference Caballero, Alonso, Deng, Attias, de Souza and Corvi2016). (D) Likely dual acylated proteins in P. falciparum (see also Supplementary Table S1). Numbers in the Venn diagrams may differ slightly from those reported in the primary literature due to revisions in sequence databases over time, ID mapping issues (e.g. between the two Tg species analysed in B), and, in diagram C, the manner in which the protein inference problem has been handled; most proteomic analyses group proteins when they cannot be distinguished by MS, but here each protein was treated independently.

Our laboratory and others have worked extensively on N-myristoylation in malaria and other biological systems, both in terms of inhibitor development (see ‘Inhibitors of Protein Lipidation in Protozoan Parasites’ section) and to globally identify myristoylated proteins. MTCC relies on the cellular machinery to take up fatty acid analogues, convert them into substrates for the acyl transferase (the acyl-CoA thioesters) and incorporate them enzymatically. Early work demonstrated that NMTs will accept azide- and alkyne-tagged myristate mimics in vitro and incorporate them into peptide substrates (Heal et al. Reference Heal, Wickramasinghe, Leatherbarrow and Tate2008). Furthermore, the binding mode of YnMyr-CoA crystallized in the active site of P. vivax NMT (PvNMT) (Fig. 4A) is nearly identical to the conformation adopted by Myr-CoA (Wright et al. Reference Wright, Clough, Rackham, Rangachari, Brannigan, Grainger, Moss, Bottrill, Heal, Broncel, Serwa, Brady, Mann, Leatherbarrow, Tewari, Wilkinson, Holder and Tate2014). MTCC with YnMyr (1) was applied to identify myristoylated proteins in asexual stage P. falciparum schizonts, revealing not only putative N-myristoylated proteins but also proteins known to be modified with a glycosylphosphatidylinositol (GPI) anchor (Fig. 4B). Jones et al. found that 17-ODYA (4) is also incorporated into Plasmodium GPI anchors (Jones et al. Reference Jones, Collins, Goulding, Choudhary and Rayner2012). These results are not surprizing, since Plasmodium GPI anchors are known to incorporate both fatty acids, and illustrate the versatility of the lipid analogues which are readily incorporated by the GPI biosynthetic machinery. N-myristoylation is thought to take place mostly on the N-terminal glycine of substrate proteins. To confirm that YnMyr (1) was attached to these sites, a cleavable azido-biotin reagent (Broncel et al. Reference Broncel, Serwa, Ciepla, Krause, Dallman, Magee and Tate2015) was used: after pull-down of tagged and biotin-labelled proteins, tryptic digest released both the unmodified and modified peptides (Fig. 2A) (Wright et al. Reference Wright, Clough, Rackham, Rangachari, Brannigan, Grainger, Moss, Bottrill, Heal, Broncel, Serwa, Brady, Mann, Leatherbarrow, Tewari, Wilkinson, Holder and Tate2014). Indeed, around 30 modified protein N-termini were detected in this case, providing conclusive evidence for the site-specific attachment of YnMyr (1) to these proteins.

Fig. 4. (A) YnMyr-CoA crystallized in the Myr-CoA binding pocket of PvNMT (PDB: 2YNC). (B) YnMyr (1) is incorporated into proteins via both amide (NaOH-insensitive) and ester (NaOH-sensitive) linkages. Proteomics revealed the base-sensitive incorporation to be on GPI-anchored proteins. (C) Dose–response of YnMyr (1) incorporation upon co-incubation with NMT inhibitor DDD85646 (19). In-gel fluorescence read-out (graph on the right: quantification of fluorescence intensity) following the workflow shown in Fig. 2A and including a basic hydrolysis step to remove GPI anchor labelling. Figure adapted from: (Wright et al. Reference Wright, Clough, Rackham, Rangachari, Brannigan, Grainger, Moss, Bottrill, Heal, Broncel, Serwa, Brady, Mann, Leatherbarrow, Tewari, Wilkinson, Holder and Tate2014).

Identifying the mode of action of drugs and small molecules of interest in a live cell context is very challenging, particularly in protozoan parasites. We next exploited our robust and rapid MTCC approach to assess whether NMT inhibitors were acting on-target in the parasite, using both previously reported T. brucei inhibitors (Fig. 10A, compounds 19 and 20) (Frearson et al. Reference Frearson, Brand, McElroy, Cleghorn, Smid, Stojanovski, Price, Guther, Torrie, Robinson, Hallyburton, Mpamhanga, Brannigan, Wilkinson, Hodgkinson, Hui, Qiu, Raimi, van Aalten, Brenk, Gilbert, Read, Fairlamb, Ferguson, Smith and Wyatt2010; Brand et al. Reference Brand, Cleghorn, McElroy, Robinson, Smith, Hallyburton, Harrison, Norcross, Spinks, Bayliss, Norval, Stojanovski, Torrie, Frearson, Brenk, Fairlamb, Ferguson, Read, Wyatt and Gilbert2012) and a novel chemically distinct series developed in-house (Fig. 9A, compounds 15 and analogues) (Rackham et al. Reference Rackham, Brannigan, Rangachari, Meister, Wilkinson, Holder, Leatherbarrow and Tate2014). All five compounds specifically inhibited incorporation of YnMyr (1) into N-myristoylated proteins, and furthermore the in-cell dose–responses calculated from levels of YnMyr (1) incorporation correlated well with EC₅₀ for parasite growth inhibition (Fig. 4C). These experiments therefore demonstrated direct engagement of compounds with NMT in cells and linked parasite death to loss of substrate protein myristoylation. With validated tools in-hand, the phenotype of NMT inhibition in the malaria parasite could be characterized (Wright et al. Reference Wright, Clough, Rackham, Rangachari, Brannigan, Grainger, Moss, Bottrill, Heal, Broncel, Serwa, Brady, Mann, Leatherbarrow, Tewari, Wilkinson, Holder and Tate2014).

YnMyr (1) has proven a versatile tool in Plasmodium species, and this analogue has also been applied in the mouse malaria parasite P. berghei to detect myristoylation of specific proteins involved in sexual development: two inner-membrane complex proteins (Poulin et al. Reference Poulin, Patzewitz, Brady, Silvie, Wright, Ferguson, Wall, Whipple, Guttery, Tate, Wickstead, Holder and Tewari2013), and two protein phosphatases (Guttery et al. Reference Guttery, Poulin, Ramaprasad, Wall, Ferguson, Brady, Patzewitz, Whipple, Straschil, Wright, Mohamed, Radhakrishnan, Arold, Tate, Holder, Wickstead, Pain and Tewari2014).

Toxoplasma gondii

Palmitoylation has been implicated in the intracellular life cycle of the related apicomplexan parasite T. gondii. For example, a palmitoyl protein thioesterase was identified as a target for a small molecule enhancer of host cell invasion, suggesting that dynamic protein S-acylation may play an important regulatory role in this process (Child et al. Reference Child, Hall, Beck, Ofori, Albrow, Garland, Bowyer, Bradley, Powers, Boothroyd, Weerapana and Bogyo2013). Child et al. also used 17-ODYA (4) (Fig. 2B) to verify S-palmitoylation of specific proteins associated with the enhanced invasive phenotype. Building on this work, Foe et al. carried out a global analysis of 17-ODYA (4) tagged proteins in T. gondii extracellular invasive stages (Foe et al. Reference Foe, Child, Majmudar, Krishnamurthy, van der Linden, Ward, Martin and Bogyo2015). Recognizing that the lipid probe may be incorporated into multiple sites in addition to S-palmitoylated cysteines (such as GPI-anchored proteins), the authors also compared samples treated with or without HA. This treatment should selectively cleave thioesters, releasing only S-acylation sites. Quantitative label-free proteomics was used to generate hits within these two experimental set-ups. Comparing the results revealed a final list of ~280 high-confidence S-palmitoylated proteins in T. gondii. Follow-up analysis of one newly identified S-palmitoylated protein, AMA1, which is known to be associated with invasion, showed that one cysteine in particular is likely S-acylated. S-acylation did not appear to be related to AMA1 localization, but a subtle effect on the rate of secretion of microneme proteins was observed. Interestingly, the S-palmitoylated proteome of T. gondii showed limited overlap with palmitoylated orthologues in P. falciparum, perhaps reflecting differences in the life stages analysed, or indicating that most palmitoylation is organism specific (Foe et al. Reference Foe, Child, Majmudar, Krishnamurthy, van der Linden, Ward, Martin and Bogyo2015).

The complementary approach of ABE was also recently applied to T. gondii. Semi-quantitative comparisons of samples treated with and without HA resulted in >400 protein hits (Caballero et al. Reference Caballero, Alonso, Deng, Attias, de Souza and Corvi2016). Around half of these were also identified by Foe et al. using their complementary approach (Fig. 3B). Although only around 50 of the proteins identified by Caballero et al. were found in two biological replicates, suggesting that there is quite some variability in the ABE technique, again around half matched to high-confidence hits from Foe et al. (Supplementary Table S1). Together, these two studies have provided a wealth of data on potential palmitoylated proteins in T. gondii.

Comparisons across the Apicomplexa

Foe et al. compared their T. gondii palmitome dataset (17-ODYA (4)) with the P. falciparum datasets (17-ODYA (4) plus ABE) of Jones et al. by identifying orthologues in these two related species. They noted poor overlap. We added the data of Caballero et al. to this analysis, comparing the aggregate of all three studies: of ~320 putative palmitoylated proteins identified in P. falciparum that have T. gondii orthologues, 141 now have some evidence for S-acylation from either ABE or 17-ODYA (4) analyses in T. gondii (Fig. 3C). As has been noted by the studies described above, it is clear that MTCC and ABE identify not only overlapping but also distinct subsets of the palmitome (and therefore also distinct sets of false positives), and there is value to using both methodologies in combination. In addition, by comparing the P. falciparum data from our myristoylation study with the palmitoylation analyses, we identify 10 likely dually acylated proteins in the parasite (Fig. 3D). The Apicomplexa comparisons are given in Supplementary Table S1, although it should be noted that exact numbers in these comparisons are dependent on how the analysis is carried out: orthologues frequently do not map one-to-one and MS data often cannot distinguish between closely related proteins, which is a widely studied problem in protein inference (Li and Radivojac, Reference Li and Radivojac2012).

Trypanosoma brucei

MTCC was first applied in the sleeping sickness parasite T. brucei to validate the N-myristoylation of a particular protein of interest, ARL6 (Price et al. Reference Price, Hodgkinson, Wright, Tate, Smith, Carrington, Stark and Smith2012), which had a putative role in intracellular protein trafficking. Following metabolic tagging with YnMyr (1) (Fig. 2B), ARL6 was immunoprecipitated from lysate with a specific antibody. Subsequent labelling with a fluorescent azide reagent allowed detection of the modified ARL6 in-gel. Expanding this approach, a global profile of N-myristoylated proteins was performed comparing both bloodstream and insect stages of T. brucei (Fig. 5A) (Wright et al. Reference Wright, Paape, Price, Smith and Tate2016). Out of ~100 robustly enriched proteins in each life stage, roughly half possessed the canonical N-terminal glycine myristoylation motif. Others are known to be GPI anchored or S-acylated, consistent with the frequent observation that S-acylation is more permissive of fatty acid chain length in many eukaryotic systems – i.e. both myristate and palmitate (and their corresponding alkynyl analogues) can be incorporated onto cysteine side chains (Fig. 5B). Indeed, longer chain palmitate analogue YnPal (2) tagged a distinct but overlapping set of proteins in T. brucei (Wright et al. Reference Wright, Paape, Price, Smith and Tate2016). Comparison of the MTCC-derived dataset with the results of an earlier ABE experiment conducted in T. brucei by Emmer et al. (Reference Emmer, Nakayasu, Souther, Choi, Sobreira, Epting, Nesvizhskii, Almeida and Engman2011) revealed some overlap but also differences (Fig. 6A); these are likely the result of both biological (host vs insect form parasites) and technical (MTCC vs ABE) differences in the two studies. Interestingly, YnMyr (1) turned out to be toxic at extended incubation times to bloodstream but not insect forms of T. brucei (Wright et al. Reference Wright, Paape, Price, Smith and Tate2016). This observation is not without precedent (Doering et al. Reference Doering, Lu, Werbovetz, Gokel, Hart, Gordon and Englund1994) and is likely related to effects on the GPI anchor pathway, which is highly dependent on myristate incorporation and crucial for T. brucei host stages (Ferguson et al. Reference Ferguson, Low and Cross1985).

Fig. 5. (A) Comparison of MTCC (workflow: Fig. 2A) with palmitate analogue YnPal (2) and myristate analogue YnMyr (1) in BSF T. brucei. (B) In-gel fluorescence read-out of the effect of NMT inhibition with DDD85646 (19) on YnMyr (1) labelling in BSF parasites. YnMyr (1) incorporation into GPI-anchored proteins, such as the VSG (Variant Surface Glycoprotein; indicated by arrow), is unaffected but incorporation into N-myristoylated proteins drops. Figure adapted from (Wright et al. Reference Wright, Paape, Price, Smith and Tate2016).

Fig. 6. (A) Comparison of T. brucei proteins identified via ABE and MTCC (with YnMyr (1); workflows: Figs 1 and 2A) techniques in the studies of Emmer et al. and Wright et al. (B) Comparison of high-confidence T. brucei myristoyl proteins that have T. cruzi orthologues with high-confidence T. cruzi myristoyl proteins. (C) Comparison of high-confidence T. brucei myristoyl proteins that have L. donovani orthologues with high-confidence L. donovani myristoyl proteins. Numbers in the Venn diagrams may differ slightly from those reported in the primary literature due to revision of sequence databases over time and, in B and C, the manner in which the protein inference problem has been handled (see Fig. 3). (D) Candidate dual acylated proteins in T. brucei (high-confidence NMT substrates also identified by both palmitoyl proteome studies (Emmer et al. Reference Emmer, Nakayasu, Souther, Choi, Sobreira, Epting, Nesvizhskii, Almeida and Engman2011; Wright et al. Reference Wright, Paape, Price, Smith and Tate2016)). (E) High-confidence N-myristoyl proteins conserved across all three organisms (Supplementary Table S2). etc: further protein IDs - see Table S2.

Unlike for S-acylation where multiple, possibly redundant, PAT enzymes exist, NMT appears to be the sole enzyme responsible for protein N-myristoylation. The enzyme has quite narrow substrate specificity for myristoyl-CoA and closely related analogues (Wright et al. Reference Wright, Heal, Mann and Tate2009). Furthermore, whilst there are no specific chemical tools for the inhibition of PATs, for NMT there are several well-characterized molecules available (see ‘Inhibitors of Protein Lipidation in Protozoan Parasite’ section). The T. brucei NMT (TbNMT) inhibitors reported by Frearson et al. (Frearson et al. Reference Frearson, Brand, McElroy, Cleghorn, Smid, Stojanovski, Price, Guther, Torrie, Robinson, Hallyburton, Mpamhanga, Brannigan, Wilkinson, Hodgkinson, Hui, Qiu, Raimi, van Aalten, Brenk, Gilbert, Read, Fairlamb, Ferguson, Smith and Wyatt2010; Brand et al. Reference Brand, Cleghorn, McElroy, Robinson, Smith, Hallyburton, Harrison, Norcross, Spinks, Bayliss, Norval, Stojanovski, Torrie, Frearson, Brenk, Fairlamb, Ferguson, Read, Wyatt and Gilbert2012) specifically reduced YnMyr incorporation into N-myristoylated, but not GPI-anchored proteins in bloodstream form (BSF) parasites (Fig. 5B), demonstrating target engagement in cells. We therefore applied these compounds to simplify the interpretation of the complex YnMyr (1) tagged proteome data, and determine which proteins were true NMT substrates. Parasites were treated with different concentrations of two inhibitors with very different potency, and then proteins tagged with YnMyr (1). After enrichment, quantitative label-free proteomics was used to assess which proteins had reduced YnMyr (1) incorporation in response to inhibition. This analysis revealed ~50 high-confidence NMT substrates; for many of these the YnMyr (1)-modified N-terminal glycine-containing peptide was also identified via cleavable reagents.

Trypanosoma cruzi

NMT is also under investigation as a potential drug target in T. cruzi. Although NMT inhibitors developed against T. brucei were significantly less efficacious in this organism, Roberts et al. showed that the compounds inhibited parasite growth and reduced incorporation of azido-myristate mimic AzMyr (3) (Fig. 2B) in a dose-dependent manner (as read-out by in-gel fluorescence), suggesting that NMT is druggable in this system (Roberts et al. Reference Roberts, Torrie, Wyllie and Fairlamb2014). The authors recently followed this with a study applying AzMyr (3) to identify N-myristoylated proteins in T. cruzi (Roberts and Fairlamb, Reference Roberts and Fairlamb2016). They used both label-free and SILAC quantification and focused on N-myristoylation by treating samples with HA to cleave S-acylation sites. Additionally, they applied two concentrations of their well-characterized NMT inhibitor. This analysis identified ~50 high-confidence N-myristoylated proteins in the parasite. More than half of these had homologues identified as NMT substrates in the T. brucei YnMyr study (Wright et al. Reference Wright, Paape, Price, Smith and Tate2016) (Fig. 6B). Related compounds were also recently shown to act on-target in T. cruzi using a gel-based fluorescent read-out after AzMyr (3) tagging (Herrera et al. Reference Herrera, Brand, Santos, Nohara, Harrison, Norcross, Thompson, Smith, Lema, Varela-Ramirez, Gilbert, Almeida and Maldonado2016).

Leishmania donovani

The extent of protein lipidation in Leishmania species was similarly poorly characterized until recently. We applied YnMyr (1)-based MTCC with label-free quantitative proteomics to assess the potential of NMT as a drug target in L. donovani (Wright et al. Reference Wright, Paape, Storck, Serwa, Smith and Tate2015). As in T. brucei, YnMyr (1) was incorporated into likely N-myristoylated, S-acylated and GPI-anchored proteins, as well as into surface glycolipids that are prevalent in trypanosomatids. A quantitative chemical proteomics-based comparison of YnMyr (1) tagged proteins revealed an overlap of 67% between insect (promastigote) and mammalian host (amastigote) stages of Leishmania parasites, a reflection of their distinct metabolism, and proteome profiles. In addition to enabling study of the different life stages ex vivo, the high sensitivity of MTCC even allowed detection of YnMyr (1) incorporation into native levels of the well-studied N-myristoylated protein HASPB in macrophages infected with amastigotes.

Taking a similar approach to T. brucei, the effects of NMT inhibition on YnMyr (1) modification of each protein were assessed using inhibitor 19 and its N-methylated analogue (19a). These two compounds have nm potency against L. donovani NMT (LdNMT) but dramatically different potencies in cells, with the latter (19a) nearly 50-fold more active against amastigotes than the former (19). Whilst this discrepancy could be due to compound uptake, metabolism or efflux, it also raised the question of whether both compounds were truly acting on-target. YnMyr (1) tagging was performed in the presence of the two compounds at approximately their EC₅₀ values, and revealed the same loss of labelling of specific bands (Fig. 7A). Further quantitative chemical proteomics confirmed this result: YnMyr (1) tagging of the same group of proteins was sensitive to NMT inhibition, demonstrating that both inhibitors engage NMT in live cells. As in other systems, combining chemical inhibition and MTCC proved a powerful approach for dissecting the complex lipidation patterns and ~30 proteins were identified as high-confidence NMT substrates (Fig. 7B). In addition to proteins involved in trafficking, protein phosphorylation, Golgi function and proteasomal degradation, just over half of the hits are completely uncharacterized. Again, there was good overlap with high-confidence T. brucei myristoylated orthologues (Fig. 6C).

Fig. 7. Studying protein N-myristoylation in Leishmania parasites. YnMyr (1) was incubated with parasites, in some cases in the presence of NMT inhibitors 19 or 19a. Following incorporation into proteins, click chemistry was used to append a fluorophore or affinity handle; in-gel fluorescence analysis (A) provided a simple read-out for inhibitor activity in cells, and quantitative chemical proteomics (B) enabled the identification of proteins and lipidation sites. This study also compared different life stages of the parasite: promastigotes (Pro.) and amastigotes (Am.). MG = protein contains an N-terminal glycine (possible NMT substrate). Figure adapted from (Wright et al. Reference Wright, Paape, Storck, Serwa, Smith and Tate2015).

Based on this study, it is clear that on-target NMT inhibitors selectively reduce YnMyr (1) incorporation into specific proteins, but do not affect tagging of others (GPI anchored, S-acylated). Indeed, we observed this across Plasmodia, Trypanosoma and Leishmania parasites, as described above. Since YnMyr (1) incorporation can be assessed on-gel, MTCC provides a rapid method to screen for on-target activity of promising NMT inhibitors in these organisms.

Comparisons across the trypanosomatids

The three studies analysing N-myristoylation using MTCC in T. brucei, T. cruzi, and L. donovani, all applied well-characterized inhibitors from the series reported by Frearson et al. in quantitative proteomics experiments to define NMT substrates. This is important because it enables one to distinguish between proteins that incorporate Yn/AzMyr (1/3) at S-palmitoyl or other sites, from those truly N-terminally myristoylated by NMT. With this piece of information and the T. brucei S-palmitoyl proteomics studies of Wright et al. (YnPal MTCC) and Emmer et al. (ABE), candidate proteins for dual acylation in this organism can be identified (Fig. 6D and Supplementary Table S2). This analysis confirms well-validated examples, such as dual acylation of metacaspase 4 and the family of flagellar calcium-binding proteins (Godsel, Reference Godsel1999; Proto et al. Reference Proto, Castanys-Munoz, Black, Tetley, Moss, Juliano, Coombs and Mottram2011), and reveals further avenues of study. For example, the data suggest dual acylation of an ADP-ribosylation factor, two protein phosphatases and numerous proteins involved in fatty acid metabolism, although whether the latter are acylated or bind lipids as part of their catalytic activity remains to be fully explored.

To identify conserved N-myristoylated proteins, the set of high-confidence TbNMT hits were analysed for orthologues in T. cruzi and L. donovani and cross-compared with N-myristoyl datasets from those organisms. There was indeed good overlap between datasets (Fig. 6B and C, Supplementary Table S2), perhaps due to similarity in experimental protocols, as well as the close relatedness of the organisms. The 10 proteins identified across all three organisms (Fig. 6E) are likely only a snapshot of the conserved N-myristoylated proteome, but proteins with functions as diverse as protein degradation (the proteasome subunit), phosphorylation (two protein phosphatases), trafficking (two ARFs), and metabolic regulation (5′AMP-activated protein kinase subunit) can be identified.

Plasmodium and Leishmania

In order to identify P. falciparum NMT (PfNMT) and PvNMT inhibitor scaffolds, two high-throughput screening (HTS) approaches were performed. For the first HTS a 150 000 compound library was screened with our collaborators at Pfizer in a radioactive scintillation proximity assay (SPA) against PfNMT and LdNMT (Bell et al. Reference Bell, Mills, Williams, Brannigan, Wilkinson, Parkinson, Leatherbarrow, Tate, Holder and Smith2012), while a second HTS was performed in collaboration with MRC Technology (MRCT) and used a 60 000 compound library in a fluorescence-based assay against PvNMT (Goncalves et al. Reference Goncalves, Brannigan, Whalley, Ansell, Saxty, Holder, Wilkinson, Tate and Leatherbarrow2012b ). The fluorescence-based assay format exploits the reaction between the thiol reactive 7-diethylamino-3-(4-maleimido-phenyl)-4-methylcoumarin (CPM) and the free CoA liberated during the enzymatic acyl-transfer (Goncalves et al. Reference Goncalves, Brannigan, Thinon, Olaleye, Serwa, Lanzarone, Wilkinson, Tate and Leatherbarrow2012a ). Both high-throughput screens identified hits in a variety of structural series. SPA-based HTS resulted in the identification of excellent PfNMT inhibitor scaffolds that have been further progressed over the last 4 years. The most promising hit compound of the MRCT HTS was further progressed in a compound series based on a quinoline scaffold (Fig. 8A, compound 5) characterized by micromolar IC₅₀ against PvNMT and moderate selectivity over human NMT (HsNMT) (Goncalves et al. Reference Goncalves, Brannigan, Whalley, Ansell, Saxty, Holder, Wilkinson, Tate and Leatherbarrow2012b ). The HTS of the Pfizer library against LdNMT successfully identified four series (Fig. 8B aminoacylpyrrolidines – compound 7, piperidinylindoles – compound 8, thienopyrimidines and biphenyl derivatives) with good-to-excellent selectivity over all other NMTs tested (Bell et al. Reference Bell, Mills, Williams, Brannigan, Wilkinson, Parkinson, Leatherbarrow, Tate, Holder and Smith2012). The binding mode of all four inhibitor classes was subsequently determined by co-crystallization with LdNMT and/or L. major NMT (LmNMT) (Fig. 8B; Brannigan et al. Reference Brannigan, Roberts, Bell, Hutton, Hodgkinson, Tate, Leatherbarrow, Smith and Wilkinson2014). All inhibitors, apart from the aminoacylpyrrolidines interact via a basic centre with the C-terminal carboxylate of the enzyme. In the case of 7 and 9, the corresponding interaction is mediated by the hydroxyl substituent. Moreover, all compounds show significant interactions with aromatic side chains of Phe90, Tyr217 and Tyr345, and exhibit an additional set of individual interactions. These structural insights were used in a subsequent structure-guided fusion of scaffolds 7 and 8 (Hutton et al. Reference Hutton, Goncalves, Brannigan, Paape, Wright, Waugh, Roberts, Bell, Wilkinson, Smith, Leatherbarrow and Tate2014). The product (Fig. 8B, compound 9) of the inhibitor hybridization exhibits a 40-fold increased potency with good selectivity over HsNMT. However, one major issue with all three inhibitors (7, 8, and 9) is the lack of cell-based activity, which MTCC analysis suggests is due to poor cellular uptake limiting access to NMT in cells (Hutton et al. Reference Hutton, Goncalves, Brannigan, Paape, Wright, Waugh, Roberts, Bell, Wilkinson, Smith, Leatherbarrow and Tate2014; Paape et al. Reference Paape, Bell, Heal, Hutton, Leatherbarrow, Tate and Smith2014).

Fig. 8. (A) Superposition of the crystal structures of the quinoline (5) and the 1,2,4-oxadiazole (16) based inhibitors in complex with PvNMT (PDB code: 4A95, compound 5, orange; 4B14, compound 16, blue) and biological activity of the quinoline compounds 5 and 6 against PvNMT, HsNMT1, and HsNMT2 (Yu et al. Reference Yu, Brannigan, Rangachari, Heal, Wilkinson, Holder, Leatherbarrow and Tate2015). (B) Superposition of the crystal structures of aminoacylpyrrolidine 7, piperidinylindole 8, and the corresponding hybridization product 9 in complex with LdNMT (PDB code: 4cgl, compound 7, green; 4cgn, compound 8, blue; 4cyo, compound 9, red) and biological activity of 7, 8, and 9 against LdNMT, HsNMT1 and antiparasitic activity against extracellular amastigotes of L. donovani (Hutton et al. Reference Hutton, Goncalves, Brannigan, Paape, Wright, Waugh, Roberts, Bell, Wilkinson, Smith, Leatherbarrow and Tate2014). The piperidinylindole 8 and the hybridization product 9 show an interaction of a basic centre with the C-terminal carboxylate of NMT. Additionally, all compounds show interactions with a set of aromatic amino acids.

In parallel an alternative strategy was exploited to identify new PfNMT inhibitors by testing the antimalarial potency of drug molecules that have already been evaluated by pharma companies as lead compounds for the treatment of other diseases. This so called ‘piggyback’ approach was based on a library of 43 inhibitors against NMT from Candida albicans (CaNMT). Although four hits successfully reduced parasitaemia in vitro, the compounds exhibit low ligand efficiency (LE) due to their high molecular weight and high lipophilicity relative to their low enzyme affinity (Bowyer et al. Reference Bowyer, Gunaratne, Grainger, Withers-Martinez, Wickramsinghe, Tate, Leatherbarrow, Brown, Holder and Smith2007, Reference Bowyer, Tate, Leatherbarrow, Holder, Smith and Brown2008). Screening a second small library of 25 inhibitors of CaNMT and TbNMT finally revealed ‘RO-09-4609’ (10) as a moderately selective hit compound. Further optimization resulted in the development of an inhibitor series with a benzo[b]furan scaffold (Fig. 9A, compound 11 + 12) that exhibits a 100-fold affinity improvement over the initial compound. Co-crystallization of inhibitor 12 with PvNMT revealed a competitive binding mode of the benzo[b]furan inhibitors with the peptide substrate (Yu et al. Reference Yu, Brannigan, Moss, Brzozowski, Wilkinson, Holder, Tate and Leatherbarrow2012). The corresponding inhibitors are characterized by moderate enzyme affinity, and the LE was still too low to consider the series to be favourable for further optimization. To overcome this issue, an inhibitor series based on a benzo[b]thiophene scaffold (Fig. 9A, compounds 13–15) was developed to mediate improved π-interactions with two tyrosine residues due to the increased aromatic character of the thiophene moiety. Crystallography of these novel inhibitors with PvNMT revealed an overlapping but distinct binding mode to the benzo[b]furans, with the benzo[b]thiophene moiety being buried deeper within a hydrophobic pocket (Rackham et al. Reference Rackham, Brannigan, Moss, Yu, Wilkinson, Holder, Tate and Leatherbarrow2013). The structure additionally showed that an appropriately positioned methoxyphenyl substituent should be able to interact with Phe105 and Ser319 of PvNMT, thereby increasing the affinity due to further π–π interactions. This hypothesis was validated by increasing the linker length between the benzo[b]thiophene core and the methoxyphenyl substituent. The resulting inhibitor 14 exhibits a 6-fold increased affinity against PfNMT (Fig. 9A; Rackham et al. Reference Rackham, Brannigan, Rangachari, Meister, Wilkinson, Holder, Leatherbarrow and Tate2014). However, one issue with the ester containing benzo[b]thiophenes is the high lipophilic LE (LLE) value of e.g. 13·2 for 13. Highly lipophilic compounds are more likely to partition from plasma to membranes and proteins and thereby exhibit increased promiscuity and toxicity. The LLE value considers lipophilicity, affinity and molecular size. Desirable leads exhibit an LLE of <10 (corresponding to LE > 0·3 and cLogP < 3) (Keserü and Makara, Reference Keserü and Makara2009). To significantly decrease the LLE and to further increase enzyme affinity, 1,3,4-oxadiazole was implemented as a bioisosteric replacement of the ester linker moiety (Rackham et al. Reference Rackham, Brannigan, Rangachari, Meister, Wilkinson, Holder, Leatherbarrow and Tate2014). The corresponding derivatives (Fig. 9A, compound 15) showed a 100-fold improved enzyme affinity and a 100-fold decreased lipophilicity while retaining the selectivity over HsNMT with respect to the first benzo[b]thiophene lead compound 13. In addition to its antiparasitic in vitro activity, compound 15 is potent against four parasite strains, including two drug-resistant ones, and shows promising activity against liver stage (EC₅₀ = 372 nm) parasites.

Fig. 9. (A) Biological activity against PfNMT and HsNMT1 and antiparasitic activity against P. falciparum 3D7 of NMT inhibitors derived from R0-09-4609 (10) by scaffold-hopping. The superposition of the crystal structures of the benzo[b]furan based derivative 12 and the scaffold simplified analogue 16 in complex with PvNMT indicates competitive binding of the inhibitors with the peptide substrate (PDB code: 4UFV, compound 12, orange; 4B14, compound 16, blue) (Yu et al. Reference Yu, Brannigan, Rangachari, Heal, Wilkinson, Holder, Leatherbarrow and Tate2015). (B) Crystal structure of the N-myristoylated inhibitor product in complex with PvNMT. The peptidomimetic inhibitor (18) is shown in orange and the myristic acid moiety in blue (Pdb code: 4c7h). Additionally, 20% of the electron density corresponds to the CoA by-product (in red), providing structural evidence for a product complex in NMT for the first time (Olaleye et al. Reference Olaleye, Brannigan, Roberts, Leatherbarrow, Wilkinson and Tate2014).

Scaffold simplification by substituting the bicyclic core with pyridyl (Fig. 5A, compound 16) resulted in the most recently reported oxadiazole-containing inhibitor series (Yu et al. Reference Yu, Brannigan, Rangachari, Heal, Wilkinson, Holder, Leatherbarrow and Tate2015). Remarkably, the scaffold-simplified inhibitors exhibit a similar binding mode in the PvNMT crystal structure as the benzo[b]furan derivatives (Fig. 9A). The 1,2,4-oxadiazole is sandwiched between Y334 and Y211, while the pyridyl nitrogen of 16 additionally stabilizes the enzyme-inhibitor complex via water-mediated hydrogen bonds with Y315. Strikingly, the 3-OMe phenyl moiety of compound 16 also overlays well with the quinoline scaffold of compound 5 (Fig. 8A). Exchanging the trimethylpyrazole with a quinolone moiety finally provided compound 17 (Fig. 9A) with an IC₅₀ of 1·7 nm against PfNMT and good cellular efficacy. The benzo[b]thiophene series was also routinely tested against LdNMT. Remarkably, the affinity spectrum changes significantly if the bicyclic system is truncated to a monocyclic thiophene scaffold (Rackham et al. Reference Rackham, Yu, Brannigan, Heal, Paape, Barker, Wilkinson, Smith, Leatherbarrow and Tate2015). Activity against human and Plasmodium NMTs decreases by almost two orders of magnitude while affinity against LdNMT increases 8-fold. However, since thiophene moieties have been associated with P450 inhibition, a 1,3,4-oxadiazole-containing 5-chlorophenyl derivative was obtained as an optimum scaffold that shows no macrophage toxicity. However, the compound failed to inhibit growth of axenic L. donovani amastigotes (leishmanial stage), likely due to difficulty accessing the target in this parasite. Therefore, further investigation of the physicochemical properties of this series is essential.

Finally, we also reported development of a PvNMT and LmNMT peptidomimetic inhibitor based on an fungal NMT inhibitor (Olaleye et al. Reference Olaleye, Brannigan, Roberts, Leatherbarrow, Wilkinson and Tate2014). The structure of the peptide (Fig. 9B, compound 18) comprizes a Ser-Lys dipeptide, a C-terminal cyclohexyl moiety, and an aliphatic chain at the N-terminus. The resulting peptidomimetic is characterized by sub-micromolar potency against both enzymes and marginal selectivity over HsNMT. Interestingly, 20% of the electron density of the inhibitor-NMT complex structure corresponds to an N-myristoylated inhibitor product and the CoA by-product, providing the first direct structural evidence for a product complex in NMT (Fig. 9B). This complex is presumably formed in situ in the crystal, favoured by high inhibitor occupancy in the solid state.

Toxoplasma

S-palmitoylation is a dynamic PTM that requires a corresponding acyl-protein thioesterase (APT) for the cleavage of the lipid thioesters. In T. gondii, TgPPT1/TgASH1 was recently identified as an orthologue of human APT1. This serine hydrolase can be inhibited by substituted chloroisocoumarin-,β-lactone-, and triazole urea-based inhibitors (Child et al. Reference Child, Hall, Beck, Ofori, Albrow, Garland, Bowyer, Bradley, Powers, Boothroyd, Weerapana and Bogyo2013; Kemp et al. Reference Kemp, Rusch, Adibekian, Bullen, Graindorge, Freymond, Rottmann, Braun-Breton, Baumeister, Porfetye, Vetter, Hedberg and Soldati-Favre2013). Interestingly, these inhibitors enhance tachyzoite invasion. Although enhancers of invasion are not obvious therapeutic agents, Child et al. speculated that the increase in number of host cells infected by multiple parasites and the corresponding increase in the competition for resources within the infected cell might be an unconventional point of action for therapeutics, although further studies are required to test this hypothesis (Child et al. Reference Child, Hall, Beck, Ofori, Albrow, Garland, Bowyer, Bradley, Powers, Boothroyd, Weerapana and Bogyo2013).

2-BP is a small molecule compound that is often incorrectly considered a global inhibitor of palmitoylation. Treatment of T. gondii with 2-BP resulted in altered gliding motility patterns of the parasite and a significant reduction of the invasion process (Alonso et al. Reference Alonso, Coceres, De Napoli, Nieto Guil, Angel and Corvi2012), but as mentioned above these findings should be interpreted carefully, and in full appreciation of the very promiscuous activity and high non-specific toxicity of 2-BP (Davda et al. Reference Davda, El Azzouny, Tom, Hernandez, Majmudar, Kennedy and Martin2013).

Trypanosoma

Pyrazole sulphonamides are NMT inhibitors identified in an HTS against TbNMT by Wyatt et al. (Frearson et al. Reference Frearson, Brand, McElroy, Cleghorn, Smid, Stojanovski, Price, Guther, Torrie, Robinson, Hallyburton, Mpamhanga, Brannigan, Wilkinson, Hodgkinson, Hui, Qiu, Raimi, van Aalten, Brenk, Gilbert, Read, Fairlamb, Ferguson, Smith and Wyatt2010). Eight sulphonamide hits of this initial HTS were further investigated by Maldonado et al. using high content imaging and a metabolic labelling approach. The authors proved the on-target activity of three compounds in sub-micromolar concentrations also against T. cruzi NMT with very low cytotoxic side-effects (Herrera et al. Reference Herrera, Brand, Santos, Nohara, Harrison, Norcross, Thompson, Smith, Lema, Varela-Ramirez, Gilbert, Almeida and Maldonado2016). The lead compound of the HTS, DDD85646 (Fig. 10A, compound 19), shows excellent activity in mouse models during the hemolymphatic peripheral infection stage of T. brucei (stage 1) (Brand et al. Reference Brand, Cleghorn, McElroy, Robinson, Smith, Hallyburton, Harrison, Norcross, Spinks, Bayliss, Norval, Stojanovski, Torrie, Frearson, Brenk, Fairlamb, Ferguson, Read, Wyatt and Gilbert2012). However, due to a low blood–brain barrier (BBB) permeability, the inhibitor is not active in the second stage during which the parasite enters the CNS (central nervous system) thereby giving rise to the classic symptoms of sleeping sickness. Therefore, Read et al. prepared modified pyrazole sulphonamides with a reduced polar surface and a capped sulphonamide group, significantly increasing the BBB permeability (Brand et al. Reference Brand, Norcross, Thompson, Harrison, Smith, Robinson, Torrie, McElroy, Hallyburton, Norval, Scullion, Stojanovski, Simeons, van Aalten, Frearson, Brenk, Fairlamb, Ferguson, Wyatt, Gilbert and Read2014). Their new lead compound (Fig. 10A, compound 20) demonstrated partial efficacy in stage 2 mouse models. However, one issue is that the increased lipophilicity results in off-target effects and poor tolerability of the new lead compound.

Fig. 10. (A) Biological activity against TbNMT and HsNMT1, antiparasitic cell activity against BSF T. brucei parasites, and BBB permeability (B:B = brain-to-blood ratio, ND = not determined) of TbNMT lead inhibitors (Brand et al. Reference Brand, Norcross, Thompson, Harrison, Smith, Robinson, Torrie, McElroy, Hallyburton, Norval, Scullion, Stojanovski, Simeons, van Aalten, Frearson, Brenk, Fairlamb, Ferguson, Wyatt, Gilbert and Read2014; Spinks et al. Reference Spinks, Smith, Thompson, Robinson, Luksch, Smith, Torrie, McElroy, Stojanovski, Norval, Collie, Hallyburton, Rao, Brand, Brenk, Frearson, Read, Wyatt and Gilbert2015). (B) Superpositions of the crystal structures of DDD85646 (19) (PDB code: 2WSA, red) with, respectively, lead compounds of the thiazolidinone (21) (PDB code: 5AG6, green) and benzomorpholinone (22) (PDB code: 5AGE, blue) series in complex with the TbNMT surrogate LmNMT reveal that the two new inhibitor classes (21 and 22) exhibit a different binding mode than the sulphonamides (19) (Spinks et al. Reference Spinks, Smith, Thompson, Robinson, Luksch, Smith, Torrie, McElroy, Stojanovski, Norval, Collie, Hallyburton, Rao, Brand, Brenk, Frearson, Read, Wyatt and Gilbert2015).

Apart from the pyrazole sulphonamides, Gilbert et al. have progressed two further hits from the original HTS and developed a thiazolidinone (e.g. 21) and a benzomorpholinone (e.g. 22) series (Fig. 10A; Spinks et al. Reference Spinks, Smith, Thompson, Robinson, Luksch, Smith, Torrie, McElroy, Stojanovski, Norval, Collie, Hallyburton, Rao, Brand, Brenk, Frearson, Read, Wyatt and Gilbert2015). Like Read et al. the authors aimed at the development of BBB permeable NMT inhibitors. Due to a lack of high-resolution structures of TbNMT and the very high-sequence homology of the binding pockets of TbNMT and LmNMT, the authors used LmNMT as a surrogate. Co-crystallography of the two new TbNMT inhibitor classes with LmNMT revealed that they are characterized by a different binding mode than the sulphonamides (Fig. 10B).

The lead compound of the thiazolidine series (Fig. 10A, compound 21) is characterized by good selectivity over HsNMT, micromolar cellular efficacy, and a good LE value that indicates the potential of the series. The benzomorpholinone series (Fig. 10A, compound 22) contains potent antiparasitic compounds with cellular potencies in the nanomolar range that exhibit BBB permeable compounds. However, the selectivity of the series has to be further improved to enable higher dose levels, and thereby maximizing the chances of curing stage 2 infections.

Concluding remarks

Protein lipidation is an essential PTM for metabolic and cellular processes in protozoan parasites, and its modulation offers interesting opportunities for therapy. Therefore, an extensive investigation of the substrates of protozoan acyl transferases and the corresponding downstream effects of their inhibition is essential. In this context ABE and MTCC are powerful techniques that can be used to profile, image and identify previously unknown lipidated proteins in a data-driven manner and without the need for specific antibodies or protein overexpression. ABE-type approaches can be used on any lysate without the need to optimize incorporation of an analogue and without the risk that the analogue will perturb the system. However, ABE is limited to S-acylation and provides no information on the lipid. MTCC approaches, in contrast, are unbiased and wide in scope. Only proteins dynamically modified during the incubation time with the analogue will be tagged and therefore identified – whilst this can be a potential limitation, more importantly it offers the opportunity for profiling dynamic lipidation through pulse-chase approaches. These technologies have dramatically increased our appreciation of the extent of protein lipidation in parasites and demonstrate that targeting these modifications could have therapeutic value. The identification of small molecule inhibitor scaffolds that inhibit protozoan acyl transferases with high selectivity over the corresponding human enzymes is an important consideration for the development of therapies. In the case of Plasmodium and Leishmania, pyridyl, 1,3,4-oxadiazole containing 5-chlorophenyl, and peptidomimetic-based scaffolds show promising characteristics and good cellular efficacy, including against liver stage malaria parasites. In the case of Trypanosoma, pyrazole sulphonamides, thiazolidinone and benzomorpholinone inhibitors are potent antiparasitic compounds; however, BBB permeability and selectivity needs to be further improved to increase their efficacy against stage 2 of T. brucei infection. Bringing together lipid profiling technologies and medicinal chemistry efforts, MTCC platforms in particular have been successfully used to demonstrate the on-target mode of action of NMT inhibitors in live parasites.

Evidence from analytical tools has accumulated to the point where S-acylation must be considered a major regulatory pathway in all eukaryotes (Resh, Reference Resh2016). The enzymes involved in removal of this modification come from the superfamily of serine hydrolases, and selective small molecule inhibitors are available for some of these enzymes. Their inhibition can lead to interesting and unexpected phenotypes, as mentioned above, and further characterization of their apparently broad substrate scope and complex localization will be important in validating them as potential drug targets. In contrast, the diverse class of protein S-palmitoyl transferases (PATs), including >20 genes in humans, has yet to yield to small molecule inhibitor discovery, and the chemical tools available for PATs are effectively non-existent. Indeed, the continued use of 2-BP due to its commercial availability is to greatly compound the challenges of the field due to the exceptional promiscuity of this molecule, as noted above. Robust and widely applicable CRIPSR-Cas gene-knockout approaches will be an important enabling tool to unpick the roles of PATs in parasites and in the host, but the discovery of cell-active inhibitors selective for the class, or for members of the class, would be transformative for the field, and should be pursued as a high priority.

In contrast, the scope for NMT as a target in eukaryotic pathogens is very clear, and may be very broad, as recently demonstrated for helminths (Galvin et al. Reference Galvin, Li, Villemaine, Poole, Chapman, Pollastri, Wyatt and Carlow2014). The availability of multiple potent inhibitor series and powerful tools to analyse PTMs in living systems greatly enhances the opportunities for drug development against this target. With the exception of T. brucei, which is rapidly killed by NMT inhibition due to its exceptional reliance on myristoylation-dependent trafficking, NMT inhibition has quite an extended mode of action. This is hypothesized to be due to an indirect dependence on protein degradation: myristoylated proteins that were present prior to inhibition will typically need to undergo some degree of degradation in order for inhibition of co-translational myristoylation to impact viability. Careful consideration of compound uptake in cells, distribution/pharmacokinetics and pharmacodynamics (the dynamics of target engagement) will be required to realize the potential of NMT inhibitors as antiparasitic agents, and research towards this objective continues in our laboratories, in collaboration with other research groups.