High-content screening of antimalarial drugs using metabolomic approaches

© 2016 Dr. Hwa Huat Charlie ChuaMalaria continues to have an appalling health and economic impact on many of the poorest countries in the world, with more than 214 million cases and 438,000 deaths a year. While considerable progress has been made in reducing the overall impact of malaria, the emergence of clinical resistance to current first-line antimalarial drugs threatens to roll back these advances. The reliance on drug treatments is further exacerbated by the limited efficacy of recent vaccine trials. Malaria is caused by mosquito-transmitted protists belonging to the genus Plasmodium. Of these, P. falciparum is the cause of the most serious form of malaria. In order to identify new lead compounds, a number of high-throughput live cell-based screens using P. falciparum have been undertaken with millions of compounds derived from the GlaxoSmithKline (GSK), St. Jude Children's Research Hospital, and Novartis compound libraries. These screens identified thousands of new drug-like compounds which selectively kill asexual intraerythrocytic stages of P. falciparum. In order to promote research on the modes of action of these compounds the Medicines for Malaria Venture (MMV), assembled a subset of 400 compounds into the Malaria Box which has been made freely available to the research community for further characterisation. The major goal of this study was to develop a generic metabolomic approach for investigating the mode of action of the Malaria Box compound library and other lead compounds identified in live-cell screens. A high content metabolomics screen was developed to allow measurement of several hundred intracellular metabolites using gas chromatography-mass spectrometry (GC-MS). GC-MS was chosen as it allows highly reproducible coverage of many metabolites in parasite central carbon metabolism. Initial studies demonstrated that GC-MS profiling could readily differentiate between uninfected red blood cell (uRBC) cultures and P. falciparum infected RBC (iRBC) cultures (at 10% parasitaemia) without fractionation of the infected cultures (i.e. separation of infected RBC from 90% uninfected RBC). A screen was subsequently developed in which synchronised P. falciparum-infected RBC cultures in a 24-well plate format was treated with drugs of interest (1 µM concentration) for 12 hours when they reached mid-trophozoite stage (32-hour post-invasion). Extensive method development resulted in an optimised protocol for rapid quenching of parasite/RBC metabolism and metabolite extraction. This screen was validated using metabolic inhibitors with known modes of action, as well as several front-line antimalarial drugs. Atovaquone, an antimalarial that is thought to primarily target the cytochrome bc1 complex in P. falciparum asexual erythrocytic stages was found to induce changes in a limited number of mitochondrial metabolites including fumarate and γ-aminobutyric acid (GABA). In contrast, drugs such as artemisinin and chloroquine that kills the parasites induced changes in a large number of metabolites across many metabolic pathways. This screen was subsequently used to profile the impact of the 80 most potent Malaria Box compounds on the metabolome of mid-trophozoite stage (1 µM concentration, 12 hours of exposure). Hierarchical cluster analysis (HCA) was used to triage these compounds into groups with putatively similar modes of action. Strikingly, this analysis allowed clustering to many of the 80 compounds into two major groups: In particular, 31 Malaria Box compounds, as well as chloroquine, resulted in characteristic changes in a range of metabolites, including a signature decrease (> 1.5-fold) in the metabolite pipecolate (a putative intermediate in lysine degradation). Another 22 Malaria Box compounds clustered together, and resembled atovaquone-induced metabolic phenotype, that included a > 1.5-fold accumulation of fumarate. Three additional metabolic phenotypes were also identified in this initial screen, highlighting its utility in broadly grouping drugs with known antimalarial activity and potentially similar modes of action. To further characterise the mode of action for the 22 Malaria Box compounds that induced atovaquone-like metabolic perturbations, the effect of these compounds on the mitochondrial electron transport chain and pyrimidine biosynthesis were investigated. These processes are linked as one of the key enzymes in pyrimidine biosynthesis, dihydroorotate dehydrogenase (DHODH), is dependent on the respiratory chain for activity. The impact of the Malaria Box compounds on pyrimidine biosynthesis was assessed by measuring changes in early intermediates in this pathway using liquid chromatography-mass spectrometry (LC-MS) and by metabolic labelling studies with H13CO3- which is incorporated into these intermediates. Eight of the 22 Malaria Box compounds inhibited pyrimidine biosynthesis in the same way as atovaquone, leading to the accumulation of N-carbamoyl aspartate and depletion of downstream metabolite uridine monophosphate (UMP). Further differences in the mode of action of these Malaria Box compounds and atovaquone were indicated by the finding that episomal-expression of the yeast DHODH, which by-passes the need for a mitochondrial respiratory activity in asexual erythrocytic stages, conferred resistance to fourteen of these compounds. In contrast, expression of yeast DHODH leads to effective resistant to atovaquone. Finally, none of the Malaria Box compounds exhibited synergistic activity with proguanil, indicating that they either target different sites on the cytochrome bc1 complex or other proteins altogether. Studies were also undertaken to investigate the function of pipecolate, which appears to be part of a parasite specific metabolic pathway, as it was only found in P. falciparum infected RBC and was selectively depleted when infected RBC were treated with several first-line antimalarials (artemisinin and chloroquine) and 31 Malaria Box compounds. Several lines of evidence suggested that pipecolate is generated during catabolism of lysine, which in turn is produced during haemoglobin degradation in the parasite food vacuole. In particular, inhibition of haemoglobin degradation led to concomitant decreases in both lysine and pipecolate, while metabolic labelling studies with 13C-lysine demonstrated conversion to 13C-pipecolate in vivo. Further, in vitro studies using positionally labelled lysine suggested the operation of an unanticipated lysine catabolic pathway involving the initial conversion of lysine to saccharopine. Pipecolate, but not lysine, is actively secreted by P. falciparum-infected RBC, and its synthesis may constitute an important mechanism for removing and/or detoxifying excess lysine generated by constitutive haemoglobin degradation. Further studies are needed to determine whether lysine degradation is essential in P. falciparum, which could be further explored as a potential novel drug target. Finally, the 24-well plate screen was further extended to measure macromolecule biosynthesis using heavy water, 2H2O labelling. Incubation of cells in the presence of low concentrations of 2H2O leads to the incorporation of deuterium (2H) into a wide range of metabolic intermediates that are subsequently used for macromolecule biosynthesis. A workflow was developed for measuring the turnover of DNA, RNA, proteins and lipids in P. falciparum-infected RBC following labelling of cells in 5% 2H2O and GC-MS analysis of constituent components of these macromolecules (i.e. deoxyribose, ribose, amino acids and fatty acids). Incorporation of 2H enrichment in DNA (deoxyribose) and RNA (ribose) at different stages of intraerythrocytic parasite development indicated stage-specific expression of DNA and RNA throughout the 48-hour cycle. As expected, incorporation into these molecules was inhibited by aphidicolin (DNA synthesis inhibitor) or actinomycin (RNA synthesis inhibitor), respectively. Interestingly 2H-enrichment in protein and lipids was minimal or not detected indicating that precursors for these molecules are largely derived from the host cell. Screening of late ring stage parasites (18-hour post-invasion) with 240 of the Malaria Box compounds (1 µM concentration, 12 hours of exposure) led to the identification of 47 compounds that inhibited DNA biosynthesis by > 50% DNA synthesis, with concomitant partial or strong inhibition of RNA biosynthesis. Detailed stage-specific analysis of these compounds provided further insights into different modes of DNA/RNA biosynthesis. Collectively, these studies have led to the development of two screens for investigating the mode of action of compounds that have been shown to have antimalarial activity in live-cell screens. The high content GC-MS based metabolomic screen demonstrated that many of the Malaria Box compounds can be grouped based on the metabolic responses that they generated. This raises the prospect that many of the top hits identified in these screens may target relatively few processes in the parasite (or the host cell). It will be of interest to extend these studies to the next tier of compounds which may exhibit lower malarial activity but allow identification of other drug targets. These studies have also highlighted unanticipated metabolic pathways in P. falciparum, such as lysine degradation, which may be important for virulence and potential drug targets