Background

In recent years, the fight against malaria has succeeded in a considerable decrease in clinical episodes, based on the use of insecticide-treated nets (ITNs), indoor residual spraying (IRS), and treatment with combination medicines [1]. Nevertheless, each year there are still hundreds of thousands of fatal outcomes from infection with Plasmodium [2]. As resistance against older anti-malarial drugs, such as chloroquine and pyrimethamine, is widespread [3], the World Health Organization (WHO) recommends that artemisinin-based combination therapy (ACT) should be the first-line treatment for Plasmodium falciparum malaria [4]. Despite this approach, artemisinin resistance is emerging in South East Asia [5,6,7,8,9]. Broad dissemination of artemisinin resistance to other malaria endemic regions of the world could lead to a dire situation, undoing much of the progress achieved in the last decade. As a consequence, the development of antiplasmodial drugs based on novel chemotypes and new mechanisms of action is urgently required to ensure efficacious treatment of malaria in the future [2].

With a view to identifying novel anti-malarial chemotypes that display novel mechanisms, a phenotypic screening of 32 compounds comprised of kinase inhibitor analogues and other structurally diverse molecules was performed. Viability assays on blood stages of transgenic P. falciparum parasites that constitutively express luciferase (PfNF54-luc), measuring their bioluminescence in a luciferase assay system were employed. Follow-up studies undertaken with a hit compound revealed structure–activity relationships in the chemical class, suggestive of a specific mechanism of action involving coenzyme A synthesis, which ultimately led to the generation of a new potent and drug-like antiplasmodial molecule.

Methods

Luciferase-based viability screening for anti-malarial activity

Asexual erythrocytic stages of transgenic NF54-luc P. falciparum were used for the luciferase-based viability assay. These parasites constitutively express high levels of luciferase. The parasites were cultured as described previously [10, 11]. Parasite culture with parasitaemia of 0.5–1% was dispensed in triplicate into white 96-well flat-bottom plates (each well contains 250 µL) (NUNC, Roskilde, Denmark) and incubated in the presence of 3 µM test compounds for 48 h (37 °C, 90% N2, 5% CO2, and 5% O2). 0.01% DMSO was included in the untreated infected RBC cultures as negative control, since the compounds stock initially was diluted in DMSO and each treatment well also contained 0.01% DMSO. Subsequently, 100 µL RPMI1640 media was removed from each well and a 100 µL volume of the Bright-Glo® substrate solution added to each well. The resultant cleavage product of the reaction, light, was measured using a FLUOROSKAN FL luminometer (Thermo), to ascertain viable parasites. Untreated cultures were used as negative controls and to calculate the inhibition rate (0% inhibition of parasite growth). Blasticidin (Sigma-Aldrich, St. Louis, MO, USA), used for selection of transfected parasites, was included as a positive control on each plate and gave >90% inhibition of parasite growth at concentration 2 µg/mL. Results are summarized in Table 1. Compounds with >50% inhibition of viability were rated as actives. Experiments were performed in triplicate and were repeated at least twice.

Table 1 Results of biological evaluation
Full size table

Selection and synthesis of additional test molecules

From the initial screening, compound 3a (Fig. 1) was identified, representing a chemotype unrelated to established antiplasmodial compounds. As additional test candidates, congeners of 3a were purchased (3b–3m, Fig. 1) from Enamine (Enamine Ltd, Kyiv, Ukraine) or synthesized (3n–3s, Fig. 1; 6a–6c, Fig. 2; 8a–8k, Figs. 3, 4). The syntheses mainly consisted of base-catalyzed nucleophilic aromatic substitution reactions, employing thioxo-substitued heterocycles as nucleophiles and electron-poor heterocycles as substrates, with chloro-substituents as leaving groups. The molecular structure of compound 3p was corroborated by an X-ray structure analysis (refer to Additional file 1). Compounds 3r and 3s were prepared from their precursors 3n and 3o, respectively, by cleaving off the Boc protection group using trifluoroacetic acid and precipitating the aminium hydrochlorides with hydrochloric acid from a solution in propan-2-ol and diethyl ether. The structural identity of all new compounds was confirmed by elemental analysis, IR, 1H NMR, 13C NMR and EI mass spectra. The purity of all synthesized products was >95%, determined by HPLC. Details of synthesis and characterization of all new products are described in Additional file 2. Exemplary 1H NMR spectra for compounds 3p, 6d, 8d and 8e are shown in Additional file 3.

Fig. 1
figure 1

Structures and syntheses of 1,3,4-oxadiazole-containing dihetarylthioethers 3. Compounds 3a–m were commercially available. Reagents and conditions for 3n–q: i triethylamine, DMF or propan-2-ol, 90–120 °C. 3r and 3s were synthesized from 3n and 3o respectively by cleavage of the protecting group with trifluoroacetic acid in dichloromethane

Full size image
Fig. 2
figure 2

Synthesis of dihetarylthioethers 6a–c. Reagents and conditions: i triethylamine, DMF, 100–120 °C

Full size image
Fig. 3
figure 3

Exemplified synthesis of dihetarylthioether 8a. Reagents and conditions: i triethylamine, DMF, 120 °C

Full size image
Fig. 4
figure 4

Dihetarylthioethers 8b–k that were synthesized in analogous manner as described for 8a in Fig. 3. For details of synthesis and characterization, refer to Additional file 2

Full size image

Prediction of physicochemical parameters

The prediction of physicochemical parameters relevant for drug likeness was performed by computational methods [12]. In the Swiss ADME program package, the topological polar surface area (TPSA) is calculated according to Ertl et al. [13]. Furthermore, the lipophilicity was predicted according to Wildman et al. [14], the solubility was predicted according to Ali et al. [15], and violations of the Pfizer filter for oral bioavailability were assessed according to Lipinski et al. [16]. Results are indicated in Table 2.

Table 2 Physicochemical properties of selected test compounds in comparison to Amb180780
Full size table

Cytotoxicity assay: human THP-1 cell line

The cytotoxicity against THP-1 cells was determined following a protocol described previously [17] with serial compound dilutions ranging from 200 to 0.1 μM for calculating the IC50 values. THP-1 cells were passed into fresh THP-1 culture medium [18]. Cells were harvested by centrifugation (250×g, 10′, 4 °C), washed three times with RPMI 1640 medium, and diluted to 8 × 105 cells/mL. The cells were aliquoted in triplicate (125 μL/well) into 96-well flat-bottom plates containing compounds diluted in complete medium (125 μL/well, containing 0.01% DMSO). THP-1 cells in wells containing medium plus 0.01% DMSO served as negative controls. Cells were incubated with test compounds at 37 °C and 5% CO2 for 48 h prior to the addition of 10% Alamar Blue (25 μL/well). After an additional 4 h the fluorescence was read (λex = 544 nm, λem = 590 nm) using a microplate reader (Fluoroskan Ascent FL, Thermo Fisher Scientific, Waltham, MA, USA). Results were analysed using Prism 4 (GraphPad Software Inc., La Jolla, CA, USA). Experiments were performed in triplicate and were repeated at least twice.

Cytotoxicity assay: human HEK-293 cell line

Cytotoxicity against human embryonic kidney (HEK-293) cells was assessed as previously described [19]. Briefly, cells were maintained in DMEM medium supplemented with 10% fetal bovine serum at 37 °C and 5% CO2, and routinely split to keep log growth.

Cells were detached using Accutase treatment, then seeded into black clear-bottom 384-wells fluorescence microplates (Greiner), at a cell density of 44,000 cells/mL in 45 µL/well, using a Benchtop liquid dispenser (Thermo Fisher Multidrop). After seeding, the cells were brought back to the incubator and allowed to settle and reattach overnight.

The cells were then exposed to the experimental and reference compounds in full dose–response using three concentrations per log dose (14 points, final concentration range 40 μM–2 nM) and 0.4% final DMSO concentration. Each plate included 5 µM puromycin and 0.4% DMSO as in-plate positive and negative controls, respectively. Plates were incubated at the standard condition above for 72 h. At the end of the incubation the medium in each well was replaced with 1 mM resazurin, and the plates were read using a multilabel microplate reader (PerkinElmer Envision) at 530 nm excitation and 595 nm emission wavelengths after a further 6 h incubation.

Inhibition data was generated using Office Excel 2013 (Microsoft). IC50 values were calculated by fitting a four-parameter nonlinear regression model using Prism v.6.0 (Graphpad Software Inc., La Jolla, CA, USA).

Hypothesis building for antiplasmodial mode of action

Taking into account the structure–activity relationships obtained from the entire series of compounds, the thieno[2,3-d]pyrimidine and its thioether linkage to a second, five-membered heteroaromatic system was assumed as a pharmacophore. This substructure resembles that contained in Amb180780 (Fig. 5), an anti-malarial compound that was previously reported to interfere with plasmodial coenzyme A (CoA) synthesis [19]. A further substructure search in the CAS database of compounds for this pharmacophore revealed an additional 1667 related compounds, however, no other antiplasmodial analogue with an assumed or confirmed mode of action was found.

Fig. 5
figure 5

Structures of published antiplasmodial compounds structurally related to series 8: antiplasmodial compound Amb180780, which interferes with the CoA metabolism; compound 33 from Edlin et al. [26]; TCMDC 133557 from the Tres Cantos Antimalarial Set (TCAMS) [27, 28]

Full size image

Coenzyme A rescue assays

Plasmodium falciparum parasites (3D7 strain) were grown in RPMI 1640 supplemented with 25 mM HEPES, 5 mg/mL Albumax II, and 0.37 mM hypoxanthine. Parasites were subjected to two rounds of sorbitol synchronization before undergoing compound treatment. Ring stage parasites were exposed to the experimental compounds alone, or in combination with 0.8 mM coenzyme A (CoA) in 384-wells imaging microplates (CellCarrier, PerkinElmer), as previously described [19]. Chloroquine, artemisinin, puromycin, and also the CoA pathway inhibitors panthenol and Amb180780 were used as reference compounds. Puromycin 5 µM and untreated solvent were used as in-plate positive and negative controls, respectively. Samples were dissolved in DMSO to a final assay DMSO concentration of 0.4%. CoA, chloroquine and panthenol were dissolved in water. All the compounds were tested in full dose–response using three concentrations per log dose (14 points, final concentration range 40 μM–2 nM for the experimental compounds and puromycin; 10 mM–500 nM for panthenol; 10 µM–0.5 nM for artemisinin and chloroquine). All sample and control wells contained the same final amounts of solvents. Plates were incubated for 72 h at 90% N2, 5% CO2, 5% O2, then parasites were stained with 2-(4-amidinophenyl)-1H-indole-6-carboxamidine (DAPI), and imaged using an Opera QEHS micro-plate confocal imaging system (PerkinElmer) with 405 nm excitation and 450/50 emission filter, using a 20× water immersion objective. Images were analysed as previously described [20], to obtain normalized % inhibition data, which were then used to calculate IC50 values, through a four parameter logistic curve fitting in GraphPad Prism v.6. The experiment was carried out in two independent biological replicates, each consisting of two technical replicates. Dose–response curves are illustrated in the Additional file 4.

Determination of aqueous solubility

The aqueous solubility was determined using a miniaturized shake-flask-method. In brief, the compound (0.25–1.0 mg) was incubated in a Whatman Mini-Uniprep vial with aqueous phosphate buffer, pH 7.4, (300 µL) in an incubation shaker (IKA® KS 3000 ic control, Staufen, Germany) at 25 °C, 400 rpm. Presence of undissolved compound was checked at 24 and 48 h. After 24, 48, or 72 h of shaking, the filter plunger was punched into the vial and the concentration in the supernatant was determined by HPLC (AUC method, isocratic HPLC, at specific λmax as indicated in the Additional file 2). For the calibration, at least three different dilutions of the compounds from DMSO stock solutions diluted with ACN were quantified. Preparation of buffer pH 7.4: to a solution of Na2HPO4∙2 H2O (290 mg), K2HPO4 (20 mg) and NaCl (808 mg) in water (ad 100.0 mL), 12% aqueous HCl was added dropwise until the pH reached 7.4. In the case of 3a the concentration in the supernatant was below the limit of quantification. Therefore, the lowest concentration (1 µM) used for the calibration was indicated as upper limit of solubility in Table 2. Further details of the solubility testing methods are described in Additional file 2, including calibration concentration ranges, composition of chromatographic eluents, and wavelength for measurement of absorbance.

Results

Testing of the compound library revealed structure 3a as a hit, which, at a concentration of 3 µM, inhibited the viability of erythrocytic asexual stage P. falciparum NF54-luc parasites by 80%. Prediction of ADME data by the SwissADME interface predicted 3a to be only moderately soluble. To generate structure–activity relationships around 3a, analogues 3b–3q were tested in the same assay. Two congeners with ionizable side chains (3r, 3s) were included in this test run as compounds with predicted improved solubility. All analogues 3 showed, if any, only negligible viability reduction when tested on PfNF54-luc parasites at 3 µM (Table 1). Analogues 3b and 3c were tested in 30 µM concentration, but even at this high concentration only inhibited the parasites to an extent below 50%, the threshold level for further evaluation. While analogues 6a–6c were also completely devoid of antiplasmodial activity, members of series 8a–8k displayed a broad spectrum of potency. The congeners 8b, 8h, 8i, 8j were assessed as inactive. Compound 8a, the analogue most closely related to 3a, exhibited a medium inhibition of viability (33%). However, the other structures in the series (8c–8g, 8k) performed well, producing >95% inhibition of viability in 3 µM concentration. Amb180780 (Fig. 5) is a related structure that was recently identified as an inhibitor of the plasmodial CoA synthesis pathway (Fig. 6) [19]. Considering the structural similarity to Amb180780, the hit structure 3a, its closely related analogue 8a, and the most potent compounds 8c–8g, 8k were evaluated on 3D7 parasites in absence and presence of coenzyme A supplementation. Although 8b had shown only minor activity against the PfNF54-luc strain, the compound was also evaluated in the CoA rescue experiments because of its close structural similarity to the active 8c. The data from the initial screening on erythrocytes infected with PfNF54-luc strain were nicely congruent with the results with 3D7 parasites (Table 1). Test compounds showing >80% inhibition in the initial screening produced submicromolar IC50 values against the 3D7 parasites, with 8c as the only exception (IC50 = 1.26 µM). With double-digit nanomolar inhibitory activity, compounds 8d–8g stood out as potent molecules of interest. Supplementation with 0.8 mM CoA clearly rescued the parasites from the inhibitory effects of most of the experimental compounds. A clear reduction in the potency of most compounds was observed in the CoA-supplemented cultures, compared to the solvent controls. With the exception of 8a and 8c, all the experimental compounds showed IC50 ratios (IC50 CoA/IC50 solvent) exceeding a threshold of 5×, considered to be significant [21]. The most potent compound was 8d, with an IC50 value of 0.039 ± 0.001 µM in the absence of CoA. Under CoA rescue conditions its IC50 shifted to 2.6 ± 0.4 µM (67-fold). KuWei173 (8e) was the third most potent compound against P. falciparum 3D7, showing IC50 values of 0.075 ± 0.002 and 2.6 ± 0.4 µM (34-fold) without and with CoA supplementation, respectively. The largest IC50 shift was observed for 8f, which displayed non-rescue and rescue IC50 values of 0.096 ± 0.006 and >11 µM, respectively (117-fold). 8a and 8c showed only minor shifts in their IC50 values in rescue conditions of ~1.5 and 1.8 respectively. These compounds were not very potent, having non-rescue IC50s of 10.5 and 1.3 µM, respectively. All the control compounds showed the expected behaviour in rescue and non-rescue conditions. Chloroquine, artemisinin and puromycin displayed the expected IC50 values and were not rescued by coenzyme A supplementation. On the other hand, panthenol and Amb180780 were both rescued by coenzyme A with IC50 shifts of 41- and >13-fold, respectively.

Fig. 6
figure 6

Pathway of CoA biosynthesis. Although the enzymatic steps are conserved, the homology of the plasmodial and human enzymes involved in the pathway is not very high [19, 24]

Full size image

Tests of the six most potent antiplasmodial compounds against human cell lines revealed a low degree of cytotoxicity. For example, KuWei173 (8e) which displayed an IC50 = 75 nM against the 3D7 parasites, showed an IC50 of 33.6 µM towards the THP-1 cell line and above 10 µM against HEK-293 cells, corresponding to 450- and >133-fold selectivity, respectively (Table 1).

The aqueous thermodynamic solubility of the compounds showing the highest potency was assessed by a shake flask method at pH 7.4 with HPLC quantification. The hit structure from the initial screening 3a and the benzimidazole 8g exhibited poor solubility below 1 µM. While 8d, 8f and 8k were assessed as moderately soluble (solubility between 1 and 100 µM), derivatives 8c and 8e (KuWei173) exhibited a remarkable solubility of >1 mM. In the cases of 8c and 8k considerable differences of more than one log value were found between predicted and experimentally determined solubility (Table 2).

Discussion

Given the current emerging resistances against artemisinin-based drugs, discovery and development of anti-malarials acting by hitherto unexploited mechanisms is urgently needed. Starting points for development campaigns in the field of anti-infectives typically are sourced from phenotypic screening. To initiate such a project, a small library of 32 structurally diverse compounds against asexual erythrocytic stages of PfNF54-Luc parasites was screened. Compound 3a was identified as a hit that showed no resemblance to established anti-malarial compounds and thus offered the possibility of acting via a novel mechanism. However, 3a displayed poor solubility, an unfavourable property for a drug development (Table 2). For a rational modification of the hit structure, information on structure–activity relationships was needed. The neutral analogues 3b–3q in the initial screening system were therefore tested. The ionizable amines 3r and 3s were also included in the assays with the intention to identify analogues with improved solubility (Fig. 1). Compounds 3b–3s all failed to show the desired level of antiplasmodial activity, thus it was speculated that the 2-furyl-1,3,4-oxadiazole system of 3a would be an important part of the antiplasmodial pharmacophore. This hypothesis was refuted by design, synthesis and testing of the analogues 6a–c, in which the 2-furyl-1,3,4-oxadiazole moiety of 3a was retained and the thieno[2,3-d]pyrimidine was replaced by other heterocycles (Fig. 2). Series 6 was completely inactive. Compound 8a, an analogue of 3a in which the oxadiazole oxygen is replaced by nitrogen, was then prepared (Fig. 3). Although 8a was inferior to 3a, it still showed at least some (33%) inhibition of the parasites at 3 µM. This result led to the conclusion that the thieno[2,3-d]pyrimidine is an indispensable part of the pharmacophore. Consequently, the congeners 8b–8k, in which the thieno[2,3-d]pyrimidine element is kept and connected to various five-membered heteroaromatic ring systems by a thioether linkage, were prepared (Fig. 4). Taking into account the results from tests on both P. falciparum strains, some conclusions were drawn regarding the contribution of this second heterocyclic element. First, most small heteroaromatic rings were tolerated. However, a methyl substituent adjacent to the thioether linkage led to decreased antiplasmodial activity (analogues 8b and 8c). Benzoannulation at the site opposite to the thioether linkage was tolerated and resulted in the very potent derivative 8g. Since 8g shows only submicromolar aqueous solubility, the aza analogue 8k was designed, which showed a slightly decreased antiplasmodial activity but remarkably improved solubility by more than two orders of magnitude (77 µM). The introduction of substituents at the five-position of the benzimidazole element completely shut down the antiplasmodial activity, as was exemplified by 8h, 8i, and 8j. It appears that in the binding pocket of an as yet unidentified target protein the available space around this position is restricted.

The benchmark for compounds suitable for drug discovery and clinical candidates is an aqueous solubility of >60 µg/mL [22], which translates to a 200 µM solution for molecules with a molecular mass of ~300 g mol−1. Of the compounds showing antiplasmodial activity in this study, compounds 8c and 8e (KuWei173) met this demand, exhibiting >1 mM solubility. Since KuWei173 (8e) shows 17-fold higher antiplasmodial activity it clearly represents the most promising result from the present study.

It has been shown that besides aqueous solubility, the topological polar surface area (TPSA) is a well-suited parameter for the oral bioavailability prediction of drugs. As a rule of thumb, molecules displaying less than 60 Å2 TPSA are absorbed to an extent of >90%, while compounds with more than 140 Å2 TPSA are absorbed by ≤10%. KuWei173 (8e) has a calculated TPSA of 105 Å2 and, therefore, appears more promising both in terms of solubility and permeability than the closely related thiazole analogues 3d and Amb180780, which both show TPSA values ~120 Å2 (Table 2).

For a rational structure-based drug development process with KuWei173 (8e) as the lead compound, the identification of a biological target in the parasite would be of great value. We previously reported that Amb180780 (Fig. 5), a compound structurally related to 8e, inhibits Pf3D7 parasites by interfering with the CoA synthesis pathway [19].

Plasmodium falciparum is capable of CoA de-novo synthesis and is thus independent of CoA synthesis by the host [23]. Although the CoA synthesis pathway is highly conserved among eukaryotes (Fig. 6), the homology of the five enzymes catalyzing the transformation of pantothenate to CoA between Plasmodium and mammalian species is not high [19, 24]. Consequently, the CoA synthesis pathway was suggested as a potential target for antiplasmodial drugs [19, 25]. Considering the structural similarity to Amb180780, a similar mode of action for members of the series 8 was postulated. Upon evaluation of 8a–8g, 8k and 3a for antiplasmodial activity in the absence and presence of coenzyme A, we noted that CoA supplementation rescued the parasites from inhibition by the compounds, with the exception of 8a and 8c. The results corroborated the hypothesis that inhibition of CoA synthesis was a mechanism associated with the observed antiplasmodial activity for this series. The two outliers 8a and 8c are less active compared to other molecules, and may act by other mechanisms than CoA synthesis inhibition. Low potency alone, however, does not explain lack of rescue by CoA, because 8b, the second least potent compound (IC50 = 6.4 ± 4.3 µM in non-rescue conditions) was significantly rescued by CoA supplementation to an IC50 of ~40 µM, corresponding to a ~sevenfold shift in activity.

In terms of solubility and antiplasmodial activity in vitro, KuWei173 (8e) appears as the most interesting compound resulting from the project reported here. Follow-up studies are now warranted with KuWei173 (8e) to assess parameters including metabolic stability and in vivo efficacy.

While searching for analogues of series 8 the two antiplasmodial thieno[2,3-d]pyrimidines compound 33 [26] and TCMDC-124514 [27] (Fig. 5) were also encountered. The latter was designated as a “quality chemical starting point” [27] among the 13,533 hit compounds of the Tres Cantos Anti-malarial Set, which resulted from the screening of nearly 2 million compounds from GSK [28]. Indeed, TCMDC-124514 differs only by one methyl group from the compound 8g presented here. To the best of our knowledge, neither a systematic structure optimization study nor a mechanism of antiplasmodial action has been published for TCMDC-124514. It is tempting to speculate that on the one hand TCMDC-124514 may also act by inhibition of CoA synthesis, and that on the other hand the methyl substituent at the ring nitrogen may negatively influence the antiplasmodial activity, considering the results with 8b and 8c. In contrast, compound 33, published by Edlin et al. [26], was the best compound in a series that was systematically developed from a HTS hit. In a Plasmodium berghei infected mouse model, compound 33 injected at 50 mg/kg subcutaneously over 4 days produced a 34% drop in parasitaemia, but failed to prolong the life-span of infected animals. The authors rated compound 33 as promising for further follow-up through optimization of ADME properties. In this respect KuWei173 (8e) can be considered as optimized analogue of compound 33 at least in terms of solubility and potency, although it has not yet been studied for metabolic stability in vitro or in vivo malaria models.

For further structure-guided optimization of this compound class, it will be necessary to identify which of the enzymes involved in CoA synthesis constitutes the eventual target. A recent study using the same rescue approach used here, but with additional CoA pathway intermediates found that addition of pantethine, dephospho CoA and CoA all rescued the activity of Amb180780 to more than 75%, pointing at the enzymes phosphopantothenoylcysteine synthetase or phosphopantothenoylcysteine decarboxylase as the putative targets for Amb180780 [29]. For similar studies aiming at corroborating the putative targets of the new compound class presented here, the readily soluble and potent inhibitor KuWei173 (8e) is now available.

Conclusion

A phenotypic screening campaign towards new antiplasmodial drugs was performed by testing a compound library on PfNF54-luc parasites. Modification of hit structure 3a led to the simple dihetarylthioether KuWei173 (8e) which shows high viability inhibition on asexual erythrocytic P. falciparum parasites and good water solubility. Studies with CoA supplementation showed that KuWei173 most probably acts by inhibition of the plasmodial CoA synthesis pathway. Cytotoxicity testing further revealed a considerable safety margin towards human cell lines. Taken together, KuWei173 (8e) appears to be a good lead structure for further development and a useful tool to study plasmodial CoA biosynthesis.