Generation of quinolone antimalarials targeting the Plasmodium falciparum mitochondrial respiratory chain for the treatment and prophylaxis of malaria

Generation of Hits Against PfNDH2 and bc₁.

A high-throughput screen (HTS) against PfNDH2 was set up using recombinant PfNDH2 expressed in a heterologous expression system in Escherichia coli NADH dehydrogenase knockout strain ANN0222 (nuoB:: nptI-sacRB, ndh::tet) (7), eliminating background NADH:quinol oxidoreductase activity. An assay protocol suitable for HTS was optimized and validated for screening with the Z′ ranging from 0.66 to 0.9 and a signal-to-background ratio >10. A chemoinformatics strategy was initiated using HDQ (7), a dual inhibitor of PfNDH2 and bc₁ (but with poor pharmacokinetics and drug-like properties), as a query molecule to perform similarity and scaffold-hopping activities. A number of algorithms, including applied molecular fingerprints (8), turbo similarity (9), principal components analysis, Bayesian modeling (10), and bioisosteric replacement strategies (11), were used to select ∼17,000 compounds for screening. The focused library was selected from a commercial library of ∼750,000 compounds (BioFocus DPI), and compounds were predicted to possess favorable absorption, distribution, metabolism, excretion, and toxicity characteristics (12). Following a primary screen, 419 actives (>30% inhibition at 20 μM) were retested in triplicate, and from these, 150 compounds were progressed for potency determination (10-point concentration curves, 1:3 dilution). From the active compounds tested for potency, 22 compounds had IC₅₀ values falling between 11–40 μM, and 24 compounds had IC₅₀ values <10 μM and purity >70%. These hits were observed to occupy distinct areas of chemical space (Fig. 1 and Table S1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

3D plot of chemical space showing progression from HDQ via hits from HTS to SL-2–25. Chemical space representation of first three principal components is based on calculated physiochemical properties of 48 hit compounds that account for 88.5% of the variance. All other compounds projected into principal component space using this model. The cyan sphere represents HDQ; orange points represent ∼17,000 compounds selected via chemoinformatics methods and screened in HTS; green spheres represent 48 hits from HTS; and the back sphere represents SL-2–25. The three-component principal components analysis model was built using Knime (http://www.knime.org), and the 3D plot was produced using R using the Ruby Graph Library (http://cran.r-project.org).

Twelve distinct chemotypes were identified from the hits. Hit expansion and the evaluation of “near neighbors” confirmed the selection of the quinolone core as the main target for structure–activity relationship (SAR) development. Initial studies focused on compounds with monoaryl groups at position 2; however, it was impossible to drive activity below 500 nM against the 3D7 strain of P. falciparum. A progression toward the close HDQ analogs, where a longer biaryl/phenoxy biaryl replaced the metabolically vulnerable HDQ side chain, improved both antimalarial and PfNDH2 activity as demonstrated by RKA073 (Fig. 2). A further structural alteration introduced a methyl substituent at position 3. This manipulation twists the 2-aryl side chain, altering the torsion angle (presumably leading to a reduction in aggregation), and resulting in better overall solubility and greatly enhanced activity. This medicinal chemistry strategy generated more than 60 compounds, as exemplified by CK-2–68 with activity of 31 nM against the P. falciparum 3D7 strain and 16 nM against PfNDH2. Investigations into the level of substitution in the A-ring of the quinolone core showed that F, Cl, or H at the 7 position were optimal for 3D7, W2, and PfNDH2 activity (Fig. 2, Fig. S1, and Tables S2 and S3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

(A) Medicinal chemistry strategy deployed in the discovery of SL-2–25 and SL-2–64. (B) Radar plot of physicochemical properties of SL-2–25 (red line) and SL-2–64 (blue line) and how these properties map onto the desired target (Lipinski-like) region. (C) Dose–response curve for lead quinolones versus P. berghei (four doses). (D) Cartoon representation of bc₁ complex with enlarged view of compound SL-2–25 docked in the Q_o site of yeast bc₁ (PDB code 3CX5). In the close-up view residues His181 (iron-sulfur protein), Glu272 (cytochrome b), and H-bond interactions are highlighted. Q_o site water molecule 7187 forms an H-bond bridge between Glu272 and the quinolone NH. Also shown in orange and yellow sphere representation is the Fe₂S₂ cluster (iron-sulfur protein). Interatomic distances for hydrogen-bonding interactions are given in Å.

Selectivity of Inhibitors for bc₁ and PfNDH2.

Several trends emerged from the analysis of the in vitro activity profiles (Table S2). Incorporation of a halogen at the 7 position increased the selectivity for PfNDH2 over that for bc₁, as best demonstrated when comparing the selectivity ratios (SRs: inhibitory activity against bc₁/inhibitory activity against PfNDH2) of CK-2–68 (SR = 31) vs. CK-2–67 (SR = 2.34) and SL-2–64 (SR = 6.4) vs. SL-2–25 (SR = 1.0). Removal of the methyl linker in the side chain also imparted greater selectivity; for example, CK-2–67 with the methylene linker has a selectivity ratio of 2.34, whereas SL-2–34 without the linker has a selectivity ratio of >38. Fig. S2 demonstrates the SAR effect seen when comparing the 2-aryl and 3-aryl series of compounds. 2-Aryl quinolones provide PfNDH2 inhibition levels of less than 20 nM, whereas the directly comparable 3-aryl quinolones have PfNDH2 inhibition levels greater than 200 nM. However, 3-aryl quinolones demonstrate high levels of bc₁ inhibition, which are not unexpected given the very close structural features of this series with the known GlaxoSmithKline pyridone bc₁ inhibitors (13).

Given the potent nanomolar inhibitory activity of Sl-2–25 against the bc₁ complex, a docking experiment was performed using the yeast bc₁ crystal structure to determine the nature of the key interactions at this target site (Fig. 2D). Although not identical to the parasite bc₁ complex, the yeast protein shares 40% homology, and the Q_o region is extremely well conserved between the two proteins: Yeast and P. falciparum cytochrome b show 71% sequence identity in the C-terminal region of helix C and the cd1 loop region of Q_o (20/28 residues) and 85% sequence identity in the ef loop region of Q_o (11/13 residues); the key hydrogen-bonding residues E272 and H181 (from the Rieske iron-sulfur protein) are completely conserved. Docking of SL-2–25 was performed, and results showed an average GoldScore of 53.7, consistent with this inhibitor displaying potent inhibitory activity. Inspection of the docking results shows that SL-2–25 occupies a position within the Q_o site of cytochrome b similar to that of stigmatellin, distal to low-potential heme. Hydrogen bonds are provided to the quinolone head group via the imidazole ring of Rieske protein residue His181 (1.6 Å), with a water-bridged hydrogen bond from the side chain of cytochrome b ‘PEWY’ (ef) loop residue Glu272 (2.5 and 2.3 Å) (Fig. S3). Within cytochrome b, the SL-2–25 binding site is formed from residues contained within the C-terminal region of transmembrane helix C (I122, I125), helix cd1 (I147, L150), the ef loop (L269, P271, E272, L275), and the F1–F2 linker region (F296, I299) (Fig. S3). Most interactions are hydrophobic and van der Waals in nature. The side chain of I122 is predicted to form a 58-Ų van der Waals interaction with the trifluoromethoxy moiety of SL-2–25, with significant hydrophobic interactions with the pyridinyl and quinolone groups of the inhibitor provided by the side chains of P271 and L275 (26 and 33 Ų, respectively).

Pharmacodynamics of bc₁/PfNDH2 Inhibitors.

From the series, CK-2–68 was used for validation of the target because of its selectivity for PfNDH2. CK-2–68 demonstrated potent activity against both native and recombinant PfNDH2, with IC₅₀ inhibition values of 23 ± 5 nM and 16 ± nM, respectively. Steady-state substrate competition assays indicated that CK-2–68 competes for quinone (Q₁) and not NADH-binding sites (Fig. S4 A and B), with an inhibition constant K_i value of 2.4 nM (14). CK-2–68 displayed relatively low activity (IC₅₀ 500 nM) against parasite bc₁ and no activity (IC₅₀ >100 μM) against parasite dihydroorotate dehydrogenase (DHODH). As with the other members of the series, CK-2–68 displayed potent activity against multidrug-resistant parasite lines including TM902CB, an atovaquone-resistant (atovaquone IC₅₀ >12 μM) Thai parasite isolate carrying the cytochrome b mutation Y268S (Table 1). We also generated the strain 3D7-yDHODH·GFP, a transgenic derivative of P. falciparum 3D7 containing yeast DHODH, through electroporation of purified pHHyDHOD-GFP plasmid (4). Transfection of malaria parasites with the yeast DHODH gene has been shown previously to confer resistance to the parasite against ETC inhibitors (4); consistent with this report, CK-2–68 showed a dramatic loss of activity against 3D7-yDHODH·GFP with an IC₅₀ of 4.6 μM. (The IC₅₀ for atovaquone in the transgenic 3D7-yDHODH·GFP was 5.8 μM; see Table 1). CK-2–68 prevented asexual development at the midtrophozoite stage (as does atovaquone) and demonstrated a time-to-kill profile similar to that of atovaquone (Fig. 3). Through repeated exposure (over the course of 4 mo) to sublethal concentrations of CK-2–68, we generated a P. falciparum (K1) line with a threefold-increased tolerance to CK-2–68 (a shift of IC₅₀ from 150 nM to 450 nM). Subsequent analysis of the PfNDH2 gene (PFI0735c) in this strain revealed a nonsynonymous nucleotide substitution at G609A replacing isoleucine with valine at residue 203. In the absence of an atomic structure for PfNDH2, it is not possible at this stage to determine the significance of the V203I mutation; however, there are no native single-nucleotide polymorphisms in plasmodial PfNDH2 (PlasmoDB.org).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Kill-rate profile of CK-2–68 (△), atovaquone (■), pyrimethamine (▲), chloroquine (□), and dihydroartemisinin (●) relative to drug-free control (○). All drugs were administered at IC₉₀ concentrations.

We have shown previously that, although PfNDH2 is not directly involved in proton pumping (5), inhibition of this enzyme leads to a depolarization of the mitochondrial membrane potential by virtue of starving the bc₁ of reducing equivalents (15). We therefore perfused trophozoite-stage parasites with CK-2–68 and, using dynamic single-cell measurements of parasite mitochondrial membrane potential (15, 16), demonstrated rapid depolarization at concentrations as low as 1 nM, 500-fold below the IC₅₀ against bc₁ (Fig. 4A). The mitochondrion of asexual parasites has been shown previously to provide orotate for pyrimidine biosynthesis through the activity of DHODH (3, 17). The link between mitochondrial function and pyrimidine biosynthesis was supported further by the generation of an atovaquone-resistant phenotype in transgenic P. falciparum parasites expressing Coenzyme Q-independent yeast DHODH (4). To confirm disruption of pyrimidine biosynthesis by CK-2–68, we undertook a targeted metabolomics approach in which we simultaneously monitored the response of 35 key parasite metabolites to CK-2–68 and atovaquone in both 3D7 and transgenic 3D7-yDHODH·GFP falciparum parasites. The metabolome dynamics following drug exposure confirms previous reports that atovaquone disrupts pyrimidine metabolism at the point of DHODH as indicated by the accumulation of dihydroorotate and reduction of downstream intermediates (Fig. 4 B and C and Fig. S4C). Significantly, no change in pyrimidine metabolite dynamics was observed upon atovaquone addition in the transgenic 3D7-yDHODH·GFP strain, confirming the hypothesis that in this strain the yeast DHODH rescues the parasite from bc₁-mediated inhibition of pyrimidine biosynthesis (4). CK-2–68 also was shown to disrupt pyrimidine biosynthesis, causing an increase in dihydroorotate and carbamoyl-l-aspartate indicative of a similar choke point at DHODH (Fig. 4 B and C). These data are consistent with CK-2–68 causing parasite death by inhibiting pyrimidine biosynthesis caused by the collapse of mitochondrial function. (A schematic of the ETC of P. falciparum is given in Fig. S5.)

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

(A) Time course of tetramethylrhodamine ethyl ester-dependent mitochondrial fluorescence after the addition of CK-2–68 to P. falciparum-infected erythrocytes. Data were normalized to 100% in untreated cells and to 0% in cells treated with 10 μM carbonyl cyanide m-chlorophenyl hydrazone. ♢, Untreated cells; ○, 1 nM CK-2–68; ●, 10 nM CK-2–68; □, 100 nM CK-2–68. Graph shows data derived from experiments performed independently (n ≥ 3). (B) Dihydroorotate and (C) carbamoyl-l-aspartate concentrations in P. falciparum 3D7 and 3D7-yDHODH·GFP parasites following addition of 100 nM CK-26–8 and atovaquone. △, untreated 3D7 parasites; ▲, atovaquone-treated 3D7 parasites; □, untreated 3D7-yDHODH·GFP parasites; ■, atovaquone-treated 3D7-yDHODH·GFP parasites; ○, untreated 3D7 parasites; ●, CK-2–68–treated 3D7 parasites.

As described, CK-2–68 was used as a proof-of-concept inhibitor. More drug-like leads displayed equipotent PfNDH2 and bc₁ inhibitory activity (Table 1 and Table S1), and these leads were tested further to determine their inhibitory activity against the liver and sexual stage of the parasites. Using a high-content imaging assay, the development of liver-stage Plasmodium cynomolgi parasites was monitored in the presence of lead compounds (Fig. 5A). P. cynomolgi, like its close relative Plasmodium vivax, forms dormant liver stages as well as immediately developing liver stages, described as “small” forms (hypnozoites) and “large” forms (schizonts), respectively, in 6-d-old in vitro liver-stage cultures (18). As expected, liver-stage development of P. cynomolgi schizonts and hypnozoites was inhibited by incubation with the positive control primaquine. Incubation with SL-2–99 (a morpholine prodrug of SL-2-25), SL-2–25, and CK-2–68 resulted in the arrest of liver-stage P. cynomolgi schizont development at pharmacologically relevant exposure concentrations (see below). Schizont development is arrested early in this assay, because schizonts remain small (or disappear from culture as total parasitemia goes down with higher levels of compound) and thus are counted as hypnozoites. Hypnozoites themselves appear unaffected by these compounds, similar to previous reports with atovaquone (18).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Activity of lead compounds against Plasmodium liver (also known as exoerythrocytic forms) and sexual stages. (A–D) Activity of lead compounds against P. cynomolgi large (liver schizonts) and small (hypnozoites) forms. (E) Activity of lead compounds against P. berghei ookinete formation.

Lead compounds also were tested for activity against sexual stages of Plasmodium. Neither inhibitor showed any activity against late stages IV and V (as is also the case with atovaquone). CK-2–68 displayed moderate inhibition of P. falciparum microgametocyte exflagellation (IC₅₀ ∼10 μM) similar to that seen with atovaquone (IC₅₀ ∼1 μM). Both CK-2–68 and SL-2–25 also showed potent activity against Plasmodium berghei ookinete production (IC₅₀ 73 and 154 nM, respectively) (Fig. 5B).

In vivo experiments using P. berghei-infected mice demonstrated that CK-2–68 when given orally at 20 mg/kg [using 5% (vol/vol) DMSO, 5% (vol/vol) ethanol in tetraglycol as vehicle] resulted in complete parasite clearance in the Peters’ standard 4-d suppressive test (19), Despite this finding, it was apparent that the partition coefficient (ClogP) needed to be reduced, and aqueous solubility needed to be enhanced to deliver a molecule with improved drug-like properties. Introduction of various heterocycles into the quinolone side chain led to the selection of a series of compounds containing a pyridine group within the side chain. Incorporation of a pyridine group reduces ClogP, improves aqueous solubility, and allows the possibility of salt formation because of the presence of the basic nitrogen (pK_a = 4.7). From more than 50 compounds SL-2–25 and SL-2–64 were selected as leads for further preclinical investigation in the murine model of malaria (Fig. 2).

Formulation of SL-2–25 (with in vitro activity of 54 nM against 3D7 and 14 nM against PfNDH2) provided the phosphate salt with ED₅₀/ED₉₀ values of 1.87/4.72 mg/kg and a morpholine prodrug (Fig. S6) with ED₅₀/ED₉₀ values of 10.4/15.2 mg/kg. Introduction of a fluorine at the 7 position gave SL-2–64 with comparable in vitro activity and ED₅₀/ED₉₀ values of 3.3/6.48 mg/kg (Fig. 2). Against the same strain, chloroquine had an ED₅₀/ED₉₀ of 3.3 mg/kg/4.6 mg/kg, and artemether was active (ED₅₀/ED₉₀ of 3.1 mg/kg/5.8 mg/kg), indicating that these molecules have potency similar to that of other clinically used antimalarials in this model. All compounds are synthesized in four to six high-yielding steps from inexpensive, commercially available starting materials (Figs. S1 and S6).

The pharmacokinetic features of SL-2–25 phosphate salt (5 mg/kg) were favorable following oral administration [maximum concentration, 3.1 μg/mL; time to reach maximum plasma drug concentration, 7.0 h; elimination half-life, 10.6 h; volume of distribution, 1,261.4 mL/kg; area under the curve from time zero to time t (AUC_0–t). 57.9 μg/h/mL; and a total body clearance of 82.1 mL⋅h⁻¹⋅kg⁻¹] (Fig. S7). These pharmacokinetic parameters are consistent with a target product profile of once-daily oral dosing for 3 d. However, an elimination half-life of 10.6 h in the mouse suggests the possibility of a single-dose cure with these molecules.

Previous antimalarial projects that have focused on the development of bc₁ inhibitors have had to be terminated because of safety concerns regarding cardiotoxicity (20). Therefore we established a beef-heart bc₁ counterscreen to assess possible mitochondrial toxicity. We have shown in the past that this screen is a better model than rodent screens to assess toxicity (16). As shown in Table 1, all lead compounds display favorable in vitro therapeutic indices relative to our comparator drug atovaquone.

In summary, we have generated potent inhibitors that are active against blood stages of P. falciparum malaria that have the capacity to inhibit two key enzymes in the respiratory pathway of P. falciparum. The use of a privileged scaffold with multitarget pharmacology, as developed here, should be an advantage both in potency across a wide range of malaria parasite strains and in protection against the rapid evolution of resistance. Initial in vivo studies confirm these inhibitors as drug-like, with properties consistent with a potential deployment role in malaria control and eradication. Further lead optimization studies are in progress to identify the most promising candidate molecules for full preclinical evaluation en route to Phase 1 clinical trials in humans.