High-throughput matrix screening identifies synergistic and antagonistic antimalarial drug combinations

Abstract

Drug resistance in Plasmodium parasites is a constant threat. Novel therapeutics, especially new drug combinations, must be identified at a faster rate. In response to the urgent need for new antimalarial drug combinations we screened a large collection of approved and investigational drugs, tested 13,910 drug pairs, and identified many promising antimalarial drug combinations. The activity of known antimalarial drug regimens was confirmed and a myriad of new classes of positively interacting drug pairings were discovered. Network and clustering analyses reinforced established mechanistic relationships for known drug combinations and identified several novel mechanistic hypotheses. From eleven screens comprising >4,600 combinations per parasite strain (including duplicates) we further investigated interactions between approved antimalarials, calcium homeostasis modulators, and inhibitors of phosphatidylinositide 3-kinases (PI3K) and the mammalian target of rapamycin (mTOR). These studies highlight important targets and pathways and provide promising leads for clinically actionable antimalarial therapy.

Figure 1. Single agent and combination analysis of a large collection of approved and investigational drugs for antimalarial activity.

(A) Examples of response profiles (10 × 10 plots) for Artesunate (AS) + Mefloquine (MFQ). Percent response values represent normalized growth, relative to controls based on SybrGreen fluorescence intensities. (B) Interaction plot of combinations passing a defined threshold (DBSumNeg <−3; reflecting synergy) against the Plasmodium falciparum 3D7 strain ( = Endoperoxides;  = HLF, LUM, MFQ, CQ, TFQ;  = hPI3K/mTOR;  = mitochondrial/DHODH;  = ion channel modulator;  = hybrid mechanism;  = other). Vector lengths do not reflect the strength or weakness of the interaction. Additional network plots reflecting antagonistic outcomes are found in the SI. Enlarged versions of this figure can be found online at http://tripod.nih.gov/pub/malaria-matrix/. (C) The interaction network formed from 2134 combinations of 111 single agents is colored grey. Overlaid on top (black edges) is the sub-network of 110 combinations involving all the single agents representing the most synergistic combinations (as measured by the sum of the DBSumNeg metric). ( = antimalarial drugs;  = growth inhibitor;  = hPI3K/mTOR;  = mitochondrial/DHODH;  = ion channel modulator;  = signaling/transporter inhibitor. Large symbols correspond to the high potency class and small symbols correspond to medium potency class). (D) Hierarchical clustering of combination profiles based upon 1) synergy assessment; 2) potency class; 3) mechanism of action (MOA) relationship. The number of clusters (six) was selected based on the largest number of clusters that led to zero or one cluster that was not enriched in any MOA combination (at the 0.01 level). (E) A summary of MOA combinations enriched in cluster 3 (green), relative to the entire dataset (see SI for full list of code definitions). Enrichment was tested using Fisher’s exact test with p-values adjusted using the Benjamini-Hochberg method, and plotted as –log₁₀ p-value. The blue and red dashed lines correspond to p = 0.05 and p = 0.01 respectively. While six MOA combinations were significant at the 0.01 level, we annotated the top three. See SI for full figure.

Figure 2. Disruption of calcium homeostasis and alteration of mitochondrial potential for selected drugs and drug pairs.

(A) Time lapsed capture of calcium dependent Fura −2 fluorescence for two live side-by-side intraerythrocytic strain Dd2 parasites showing rapid loss of digestive vacuole (D.V.) Ca²⁺ (bright green inner circle, top panels) upon perfusion with cytocidal (2 × LD₅₀) dose of CQ (see methods). (B) Examination of the combination responses of KN-62 and Artemether (ATM) in three parasite lines via an isobologram analysis of the mitochondrial membrane potential as judged by a combination JC1 assay (panel 1), heatmap analysis of the viability combination response (10 × 10 matrix) (panel 2), Delta Bliss analysis (panel 3), and isobolographic analysis of the viability combination response (panel 4).

Figure 3. Analysis of autophagosomal body puncta formation and trafficking in response to environmental and/or pharmacological stress.

Imaging and quantification of PfAtg8 containing puncta within parasite (the Plasmodium falciparum HB3 strain) infected RBC after 6 hour bolus exposure to Artemether (ATM) or ATM combinations: (A) ATM at the defined IC₅₀ value (23 nM). (B) ATM at the defined LD₅₀ value (80 nM). (C) ATM and Lumefantrine (LUM) at their respectively defined LD₅₀ values (80 nM and 323 nM, respectively). (D) ATM at the defined LD₅₀ value and GSK-2126458 at the defined LD₅₀ value (102 μM). (E) ATM at the defined LD₅₀ value and NVP-BGT226 at the defined LD₅₀ value (18 nM). Panel 1: transmittance image. Panel 2: anti-PfAtg8 peptide antibody imaging (ex: 450-490, em: 500 to 550). Panel 3: DAPI nuclear staining (ex: 340–380, em: 450 to 490). Panel 4: a merged image of all three views.

Figure 4. Comparative in vivo activities for Artemether (ATM), Lumefantrine (LUM) and NVP-BGT226 as single agents and in combination.

P. berghei infected BALB/c mice were treated with ATM (5 mg/kg), LUM (5 mg/kg), NVP-BGT226 (4 mg/kg), or NVP-BGT226 (8 mg/kg), or combinations thereof as indicated above. (A) Parasitemia values for ATM, LUM and NVP-BGT226 as either single agents or in combination. All treatment was started three days post infection, with doses administered by oral gavage on day 3, 4 and 5 (3 doses) or on days 3 and 4 (2 doses). (B) Survival plot for ATM, LUM and NVP-BGT226 as either single agents or in combination. All mice that survived until day 30 (5/5- ATM/LUM (3 doses); 5/5- LUM/NVP-BGT226 (3 doses); 2/5- ATM/NVP-BGT226 (2 doses); 5/5- LUM/NVP-BGT226 (2 doses)) remained negative for parasites through 60 day post infection follow-up.