Identifying inhibitors of the Leishmania inositol phosphorylceramide synthase with antiprotozoal activity using a yeast-based assay and ultra-high throughput screening platform

Leishmaniasis is a Neglected Tropical Disease caused by the insect-vector borne protozoan parasite, Leishmania species. Infection affects millions of the world’s poorest, however vaccines are absent and drug therapy limited. Recently, public-private partnerships have developed to identify new modes of controlling leishmaniasis. Drug discovery is a significant part of these efforts and here we describe the development and utilization of a novel assay to identify antiprotozoal inhibitors of the Leishmania enzyme, inositol phosphorylceramide (IPC) synthase. IPC synthase is a membrane-bound protein with multiple transmembrane domains, meaning that a conventional in vitro assay using purified protein in solution is highly challenging. Therefore, we utilized Saccharomyces cerevisiae as a vehicle to facilitate ultra-high throughput screening of 1.8 million compounds. Antileishmanial benzazepanes were identified and shown to inhibit the enzyme at nanomolar concentrations. Further chemistry produced a benzazepane that demonstrated potent and specific inhibition of IPC synthase in the Leishmania cell.

and resistance is increasingly apparent 10 . Therefore, the current elimination programme in South Asia relies upon liposomal amphotericin B administered as a single injection. However, whilst effective and less toxic than the non-liposomal formulation, treatment failure and post-Kala-Azar dermal leishmaniasis (PKDL) may preclude elimination 11 .
With these severe problems in the treatment and control of both CL and VL, it is recognized that new antileishmanial targets and drugs need to be identified. High throughput screening (HTS) is a vital component of drug discovery and can be carried out using either in vitro assays against validated targets or in phenotypic assays against the parasite itself 12,13 . Recent high content phenotypic screening across the pathogenic kinetoplastids gave a disappointingly low number of novel potent hits against Leishmania donovani when compared with the related parasites Trypanosoma brucei and T. cruzi 14 . This demonstrated the importance of target-based screening for antileishmanial discovery to access to novel chemical space and new modes of action which may overcome resistance and be compatible in combination therapies.
In contrast to mammalian cells, where the predominant complex sphingolipid is sphingomyelin (SM), fungi, plants and some protozoa synthesize inositol phosphorylceramide (IPC) 15,16 . The fungal IPC synthase (AUR1p) has long been established as a drug target for pathogenic fungi 17 . Similarly, the Leishmania orthologue, and those from the pathogenic Trypanosoma species, have been suggested to be ideal, non-mammalian, targets for the development of new, less toxic, antiprotozoals 15,[18][19][20] . Furthermore, systems biology studies have reinforced the status of the Leishmania enzyme as a putative target for drug discovery programmes 21 . However, as an integral membrane protein with 6 transmembrane domains, and lipid substrates (phosphatidylinositol and ceramide) and products (diacylglycerol and IPC), formatting the Leishmania IPC synthase into a conventional in vitro assay platform is challenging 22 . Therefore, utilizing the ability of the kinetoplastid enzyme to complement for the absence of the Saccharomyces cerevisiae orthologue AUR1p 23 , here we describe the development and formatting of a robust yeast-based ultra-HTS (uHTS) assay platform. This was then utilized, in the largest effort of its type, to screen a high content (1.8 M) compound library for specific Leishmania IPC synthase inhibitors. 500 potent and specific such compounds were identified; these were then reduced to 211 following clustering to remove structural replicates. Following screening against mammalian-stage axenic amastigote L. donovani, 25 of these compounds were selected on the basis of potency, selectivity and physicochemical properties. These hits were then reduced to five following further screening of L. donovani infected macrophages. From these, one pair of structurally related compounds, the benzazepanes, was selected for further analyses. Importantly, these compounds demonstrated sub micro-molar activity against the enzyme target in a secondary in vitro assay and selectivity for the enzyme in cellulo when using an available L. major sphingolipid mutant 24 . This work demonstrated the tractability of yeast-vehicles for uHTS, with the identification of the benzazepanes as potential antileishmanials with specific inhibitory activity against IPC synthase.

Results
Design and validation of a robust Leishmania IPC synthase assay for uHTS. HTS campaigns predominantly rely on two approaches: (i) in vitro target-directed screening using biochemical assays; (ii) cell-based phenotypic screening which takes no account of the target. Both of these approaches have significant limitations, such as the production of soluble protein and a lack of cellular context for biochemical assays, and the problems of process redundancy in cells used for phenotypic screening 13 . Furthermore, both can be difficult to miniaturize and expensive to utilize for uHTS. With these problems in mind, yeast-based systems have been utilized to provide cost-effective, target-directed, screening platforms for protein targets within a eukaryotic cellular context 13 . Recently, this approach has been adopted for antiprotozoal drug discovery 12,25,26 .
The transmembrane nature of the target Leishmania IPC synthase, and the hydrophobicity of the lipid substrates and products, rendered it challenging to develop a uHTS biochemical assay 18,19,22 . However, given that the enzyme has been shown to complement an auxotrophic Saccharomyces cerevisiae mutant 18 the development of yeast-based screening platform was considered tractable. To develop a robust assay suitable for uHTS, rather than the previously utilized auxotrophic mutant 18 , a strain of S. cerevisiae completely lacking the yeast IPC synthase, AUR1p, was selected as the vehicle (a kind gift from Teresa Dunn, Uniformed Services University) 27 . This strain is reliant on the expression of essential AUR1p from a uracil selectable expression plasmid (pRS316-URA-AUR1) 27 . Transforming these with the plasmid pESC-LEU (Agilent) harbouring the L. major IPC synthase (LmjFIPCS) or AUR1p under the control of a galactose promotor, allowed for curing the yeast of pRS316-URA-AUR1 using the pyrimidine analogue 5-fluoro-orotic acid (5-FOA) and the selection of a strain dependant on the galactose inducible expression of LmjFIPCS (Fig. 1). The yeast dependant on either LmjFIPCS or AUR1p expression were then used to format an LmjFIPCS assay suitable for uHTS.
Exploiting the fact that growing yeast secretes exo-β-glucanase, which, as well as modifying the cell wall, can also hydrolyse non-fluorescent fluorescein di-(β-D-glucopyranoside) (FDGlu) to release fluorescein, an assay was developed in which cell growth was monitored by fluorescent output 28 . The optimal concentrations of galactose (0.1%), starter culture (OD 600 0.063) and FDGlu (10 μM) were established in 96-and 384-well formats incubated at 30 °C for 24 hours. The assay was then miniaturized into a 1536-well format, the most appropriate positive control (cycloheximide) selected and the robustness of the platform indicated by Z' values >0.5 29 . Screening using a GSK standard set of (9766) compounds further validated the assay, established inhibition threshold values and indicated a hit-rate of 1.07%.

Identification of selective and potent inhibitors.
Taking the robust 1536-well format assay developed, a primary uHTS screening was conducted using the 1.8 million compound GSK library against the LmjFIPCS dependent yeast. Given the hit-rate of 1.07% with the standard set approximately 19000 hits were anticipated. Initially this expectation was not reached, however reducing the inhibition threshold to capture small (≤300 Da SCientifiC RepoRts | (2018) 8:3938 | DOI:10.1038/s41598-018-22063-9 molecular weight), polar (cLogP ≤ 3) 'drug-like' compounds with lower efficacy, led to the identification of 19669 hits. The mean Z' for the screen was 0.70, demonstrating the platform to be robust.
At this stage, the primary screen, it was unclear whether the hit compounds inhibited the LmjFIPCS enzyme or (an)other essential process(es) within the yeast vehicle. Therefore, a counter screen was performed against the yeast dependant on AUR1p expression at a single compound concentration of 10 μM. In parallel, a confirmatory screen against the LmjFIPCS dependent yeast was performed at the same concentration. After calculation of the inhibition thresholds for each strain, those compounds demonstrating activity above the LmjFIPCS threshold but below AUR1p threshold were selected. In addition, to avoid loss of highly potent compounds, those demonstrating >80% inhibition of the LmjFIPCS dependent yeast were retained whatever the response against AUR1p dependent yeast.
As a result of this process, 4166 compounds were progressed further for dose response analyses. Using the optimized assay conditions, these compounds were screened against the yeast dependant on either LmjFIPCS or AUR1p expression at 11 concentrations between 100 μM to 1.7 nM. From these data, the −logIC 50 (pIC50) molar unit values of each compound were calculated against each strain and the data plotted (Fig. 2). This allowed analyses of both potency and selectivity, with 500 compounds demonstrating potent activity against LmjFIPCS dependant yeast, pIC 50 ≥ 5 (IC 50 ≤ 10 μM), and good selectivity when compared to AUR1p dependant yeast,  These 500 compounds were subsequently clustered using a sphere exclusion algorithm 30 . Briefly, a fingerprint for each hit was compared to every other molecule in the set and compounds demonstrating significant similarity ( > 0.85 on a scale of 0-1) were sorted into clusters. Two representatives from each cluster plus all singletons, 211 compounds, were taken forward for further analyses.
Cell-based screening of identified compounds. To establish the antileishmanial activity of the 211 selected hits, they were screened against L. donovani axenic amastigotes (the mammalian, pathogenic form) in a dose response assay with 11 compound concentrations between 50 μM and 0.85 nM. The data revealed 70 compounds with a calculated pIC 50 ≥ 5 (IC 50 ≤ 10 μM; Fig. 3A; SI Table).
Following a screen against HepG2 cells for mammalian cytotoxicity (11 compound concentrations between 100 μM and 1.7 nM), comparing these data with that for the 70 Leishmania hits allowed identification of selective antileishmanial compounds with pSI ≥ 1 (SI ≥ 10); 49 compounds fulfilled these criteria (Fig. 3B). Following a further screen based on the predicted physicochemical properties of these 49 compounds, those compounds showing a Property Forecast Index 31 (PFI) ≤8 and ≤4 aromatic rings were selected ( Fig. 3C; SI Table 1)). These 25 compounds have relatively low molecular weights and high levels of hydrophilicity indicating they are likely to be soluble and favourable for development, and therefore were progressed to screening against L. donovani infected THP-1 macrophages, the truest representation of the pathogenic state available in vitro. The 25 compounds were screened in a dose response assay with 11 compound concentrations between 50 μM and 0.85 nM. Given the additional membrane barriers the compounds must cross to reach the intra-macrophage amastigotes, the fact that only 7 (out of 25) were completely ineffective while 10 of the remaining 18 active hits demonstrated activity at pIC 50 ≥ 5 (IC 50 ≤ 10 μM) was promising. Of these 10, 4 had a pSI with respect to the activity against THP-1 macrophages of ≥1 ( Fig. 3D; SI Table 1), another had a pSI of 0.99 and was also retained. Of these 5 hits (Fig. 4), the benzazepane compounds (1 and 2) were both active in the low μM range against intra-macrophage amastigotes, pIC 50 5.5 and 5.7 respectively. The dithiophene compounds (3 and 4) exhibited similar levels of activity, pIC 50 5.7 and 5.9 respectively, however, the presence of the nitro group in this class was a concern given the link with toxicity issues 32 . Similarly, the singleton (5) possesses a nitro group. In light of the above, the benzazepanes were taken forward for further analyses in a secondary screening platform 19,22 . Verification of the benzazepane mode of action. Utilizing the availability of an in-house designed and validated biochemical assay for LmjFIPCS 19,22 , following re-synthesis in the free-base form (Aptuit, Verona, Italy) the two benzazepane hits were biochemically analyzed in a dose response assay (8 compound concentrations between 100 μM and 46 nM; Fig. 5). Whilst compound 2 was a modest inhibitor of LmjFIPCS (pIC 50 5.2), compound 1 was a potent, sub μM, inhibitor of the enzyme (pIC 50 6.8).   However, whilst compound 1 was antiprotozoal and a potent IPC synthase inhibitor, it was unclear whether the mode of action was on-target. To analyze this further, the in-house availability of a sphingolipid mutant L. major (Δlcb2) was exploited. This line 24 lacks serine palmitoyltransferase activity, and whilst LmjFIPCS is still present it is redundant in the absence of in situ substrate. Therefore, specific inhibitors of LmjFIPCS should be inactive against L. major Δlcb2. However, in a dose response assay, compound 1 demonstrated no significant difference in activity against both lines of L. major (pIC 50 5.5 for wild type and Δlcb2: SI Fig. 1)). Therefore, it is likely that off-target activity has a role in the antileishmanial activity of this benzazepane.
Further synthetic efforts (Aptuit, Verona, Italy) led to the synthesis of compounds bearing the left-hand side of 1 and the right-hand side of 2 (compound 6) and vice versa (compound 7; Fig. 6A). Compound 6 lost activity against LmjFIPCS in the in vitro assay; however, compound 7 maintained a pIC 50 7.2; although it did not achieve greater than approximately 60% inhibition even at higher concentrations (Fig. 6B). In addition, and importantly, compound 7 demonstrated selectivity for L. major wild type over the Δlcb2 mutant (pIC 50 5.5 versus <5; Fig. 6C). The selectivity of compound 7 for the wild type parasite over a line with a redundant LmjFIPCS (Δlcb2), strongly supported a target-directed mode of action. Furthermore, compound 7 demonstrated considerably greater in cellulo inhibition of IPC synthase activity than compound 1, as determined by incorporation of BODIPY ® FL C 5 -ceramide into IPC in the presence of the compounds (Fig. 7). Unfortunately, activity against intramacrophage L. donovani was lost, possibly due to a lack of compound transport to the site of action or intramacrophage stability. However, these data demonstrated that the efficacy of compound 7 against L. major is on-target inhibition of IPC synthase, thus supporting the status of LmjFIPCS as a drug target like its orthologues in the related trypanosomatid pathogen, T. brucei 20 .

Discussion
The identification of leads for the discovery of new drugs for the Neglected Tropical Disease leishmaniasis is a recognized priority 2,6 . To achieve this aim the development of strong industry-academia partnerships, bridging chemistry and biology, is vital 23 . Much of such collaborative efforts have employed HTS to identify antiprotozoal compounds, however high content phenotypic screening has produced a poor return of hits against Leishmania spp when  The Leishmania IPC synthase has been previously identified as a putative drug target 18,21 , and more recent work has employed a modelling approach to identify coumarin derivatives as enzyme inhibitors with in vivo antileishmanial activity 33,34 . To search for new classes of Leishmania IPC synthase inhibitors, in a multidisciplinary public-private partnership, we developed a novel yeast-based assay to screen the 1.8 million entity GSK compound library against the protozoan enzyme. To the best of our knowledge, this work reports the largest yeast-based, target-directed assay undertaken to date against targets from any disease state. 5 non-toxic, antileishmanial compounds were identified (a hit rate of 0.00028%), 2 pairs and one singleton. On the basis of a lack of nitro groups the benzazepanes (compounds 1 and 2) were selected to take forward. Both of these demonstrated sub μM activity against axenic L. donovani amastigotes (pIC 50 , 7.6 and 6.4 for compounds 1 and 2 respectively; SI Table), and low μM activity against L. donovani infected macrophages (pIC 50 , 5.5 and 5.7 for compounds 1 and 2 respectively; SI Table). The activity of these compounds in cell based assays, when compared to other recently predicted Leishmania IPC synthase inhibitors 33,34 , was promising. These activity levels were similar to those seen for miltefosine, the only oral therapy for leishmaniasis but with an unclear mode of action 2,35 . However, they were one or 2 orders of magnitude lower than those for amphotericin B, which functions through selective sterol binding 2,35 .
Importantly, both benzazepanes demonstrated activity against the Leishmania IPC synthase in vitro, with compound 1 particularly potent (pIC 50 6.8). Despite these promising findings, this compound also demonstrated off-target activities in the parasite ( SI Fig.1). However, using a simple 'mix and match' approach a benzazepane was identified that had enhanced potency against the enzyme in vitro (pIC 50 7.2) and in cellulo (Fig. 7) and demonstrated more selective, on-target antileishmanial activity (Fig. 6).
In summary, this research demonstrated the tractability of yeast-based uHTS to identify inhibitors of challenging, membranous targets such as IPC synthase, which are on-target and specific. Furthermore, the benzazepanes identified represent a new class of antileishmanial compounds with a new mode of action. Further development of these may lead to novel therapies for the Neglected Tropical Disease, leishmaniasis.
In cellulo IPC synthase assay. The L. major Δlcb2 mutant 24 , maintained in Schneider's media (Sigma-Aldrich) with 15% sera (ThermoFisher), were washed and 200 µl, at 1 × 10 7 ml −1 , incubated in serum free media for 60 minutes at 26 °C before the addition of 10 µM of the compounds, in at least triplicate, and further incubation for 18 hours. Subsequently, the reaction was initiated by the addition of BODIPY FL C 5 -ceramide complexed to BSA (1.25 µl of 0.5 mM, ThermoFisher). Following further incubation at 26 °C for 1 hour, the lipids were extracted by phase separation in chloroform:methanol:water (10:10:3, 200 µl) and fractionated on HTPLC plates, as previously 18 . The fluorescence was read at Ex475/Em520 using a Fuji FLA-3000 plate reader and AIDA Image Analyser software (version 3.52).