4.1. Cellular thermal shift assay

A recently adapted method of antimalarial target identification and validation is the cellular thermal shift assay (CETSA). ²¹⁵ CETSA establishes the target engagement of small molecules in cells or tissues by leveraging an increase in protein thermal stability when bound to a ligand. ²¹⁶ At elevated temperatures, proteins begin to unfold, exposing hydrophobic residues, and precipitate out of solution. However, ligand binding leads to an increase in protein stability and remain in solution at higher temperatures. ²¹⁷ CETSA has traditionally been used as a target validation technique where stabilized proteins are detected via western blot (CETSA‐WB). ²¹⁸ When CETSA is combined with mass spectrometry (CETSA‐MS), it becomes applicable for unbiased, proteome‐wide target identification. ²¹⁹ A major advantage of CETSA over other techniques is the ability to verify and quantify the binding of high‐affinity targets in live or lysed cells.

The workflow for CETSA‐MS has several key steps (Figure 22). First, the melting behavior of the proteome must be characterized to identify a suitable temperature range for testing. Samples of whole cells, tissues, or lysate are exposed to a gradient of either temperature in the melt curve method or drug concentration in the isothermal drug response (ITDR) method. ²²⁰ Following the thermal challenge, whole cells and tissues are lysed and the soluble protein fraction is isolated. ²²⁰ Proteins from this fraction can be labeled for quantitative determination and then digested into peptide fragments for analysis by tandem mass spectrometry. ²¹⁹

Experimental workflow for CETSA‐MS. Thermal challenge is applied to the samples of interest, modifying either temperature or drug concentration between samples. The soluble protein fraction is isolated, digested, and analyzed by MS/MS or western blot. [Color figure can be viewed at wileyonlinelibrary.com]

While CETSA is a robust method for a wide range of protein targets, not all proteins are amenable to the technique. In general soluble cellular proteins can be easily evaluated through CETSA, however, thermodynamic stabilization is less significant in transmembrane proteins. ²²⁰ Approximately 30% of the plasmodium genome is predicted to have at least one transmembrane domain (PlasmoDB). ²²¹ Examples with membrane proteins have been reported but require treatment with detergent to first liberate the proteins. ^{217 , 222} Additionally, the nature of the protein‐ligand interaction can influence a lack of stabilization. Should a ligand bind to a domain that is not significantly affected by denaturation or exert its effects by modulating interaction with a secondary protein, a stabilization effect will not be seen. ²²⁰ Increases in thermal stability may not always be a result of direct binding. Proteins involved with complex metabolic pathways can be stabilized by increases in physiological ligands or proteins as a result of drug treatment. ²¹⁶ Comparison of CETSA performed with whole‐cells and lysate can be used to control for this factor. ²²³ CETSA, such as AfBBPs and ABBPs, are prone to false positives. The use of high and nonphysiological concentrations of the compound can result in the detection of false‐positive binding proteins that may not be involved in the antimalarial mechanism of action. Careful selection of the compound concentration and the use of a structurally similar inactive control compound is helpful in decreasing the number of proteins detected and excluding false‐positive or nonphysiologically relevant proteins.

Unbiased CETSA‐MS has recently been adapted to the field of antimalarial target deconvolution. The first example was using quinine and its derivative mefloquine (Figure 23). ²²³ In this methodology, blood stage P. falciparum parasites and lysate samples were subjected to thermal melt or ITDR conditions; in all testing four separate experiment types. As CETSA had not been previously applied to P. falciparum, the melting properties of the proteome at trophozoite stage were characterized between 37°C and 73°C. T _m values could then be calculated for 80% (1821 proteins) of the trophozoite proteome, representing 65% of the overall blood‐stage proteome. ²²³ Interestingly, proteins in infected RBCs had comparatively less thermal stability than their counterparts found in the lysate. ²²³ Only 362 human erythrocyte proteins were characterized by this process, due to the disproportionate presence of hemoglobin which complicates the detection of peptides by MS. ²²⁴ For the ITDR method, thermal challenge temperature was performed at 51°C to represent the average T _m for the proteome and 57°C for the fraction of the proteome that had greater thermal stability. ²²³

Chemical structures of antimalarials assessed by CETSA. Pyrimethamine, quinine, and mefloquine were used as examples to develop and validate CETSA‐MS as a target deconvolution method in P. falciparum. As expected, CETSA‐MS identified the target engagement of dihydrofolate reductase‐thymidylate synthase (PfDHFR‐TS) as the target for pyrimethamine whose target was known. CETSA‐MS identified purine nucleoside phosphorylase (PfPNP) as a probable target for quinine and a potential weak target for mefloquine.

To validate the method, ITDR and melt curve assays were performed in the presence of pyrimethamine (Figure 23), a known inhibitor of P. falciparum dihydrofolate reductase‐thymidylate synthase (PfDHFR‐TS). ²²⁵ As expected, samples treated with pyrimethamine exhibited a temperature and dose‐dependent stabilization of PfDHFR‐TS. ²²³ However, no such stabilization could be detected in treated infected RBCs. ²²³ It was postulated that this could be the result of decreased affinity in a cellular context or due to the presence of a competing ligand such as folate. Validation of the infected RBC method was performed with the broad‐spectrum cysteine protease inhibitor, E64d. ^{226 , 227 , 228} In this study, it was found that E64d stabilized four proteins, three of which were cysteine proteases (falcipain 2A, falcipain 3, and dipeptidyl aminopeptidase), while one was unexpectedly not a cysteine protease, the DSK2 protein homolog (PF3D7_1113400). ²²³ The lack of thermal stabilization in cell lysate might also represent the necessity of the cellular environment for target engagement. This can include cellular drug activation, the availability of important cofactors, or the accumulation of the drug in a specific cellular compartment. ²²³ Therefore, it is recommended to perform experiments with both lysate and whole cells to give greater confidence in the data.

ITDR was performed on cell lysate treated with quinine and MFQ (Figure 23) and purine nucleoside phosphorylase (PfPNP) was the only protein that showed a significant dose‐dependent stabilization. ²²³ Ribosomal subunits and translation initiation factor 2 were also detected on treatment with MFQ, which is consistent with previous reports of its interaction with the ribosomal complex. ²²⁹ In infected RBCs, PfPNP was similarly stabilized by quinine, but interestingly not MFQ. ²²³ Instead, whole‐cell ITDR experiments with MFQ identified pyruvate kinase II (PfPyKII), although this may represent an increase in its abundance when cells are treated above 37°C and represent a downstream effect of drug binding or a stress response. ²²³ Hsp70 and a GrpE protein homolog Mge1, two mitochondrial proteins, were also shown to be stabilized by MFQ but only at the highest dose. ²²³ It was postulated that this may again be a result of an indirect effect on the mitochondrial membrane via reactive oxygen species formed by MFQ. ²²³ Target engagement of quinine to PfPNP was confirmed by CETSA‐WB where dose‐dependent stabilization was again seen. ²²³ Indeed, in vitro binding experiments by surface plasmon resonance (SPR) confirmed a K _d of 20 nM and 40 µM for quinine and MFQ, respectively. ²²³ The enzymatic conversion of inosine to hypoxanthine by PfPNP was also found to be inhibited by quinine (K _i 138 nM) and mefloquine (K _i 5.9 µM). ²²³ Overall, this data demonstrates PfPNP binds to quinine, but further investigation into the significance of PNP as the mechanism of action is required.

Alongside the chemical probe described earlier, Favuzza et al. demonstrated target engagement of their plasmepsin protease targeting compounds using CETSA‐WB. ¹⁵⁹ Resistance selection to the initial hit compound WM4 (Figure 24) indicated that PMX was the target. However, the potent tool compound WM382 appeared to have only a low level of cross‐resistance, indicating that it may have an additional target. ¹⁵⁹ It was hypothesized this additional target could be the closely related aspartyl protease, PMIX. As no recombinant PMIX was available at the time, CETSA‐WB was implemented to biochemically validate compound binding with HA‐tagged PMIX and PMX parasites. CETSA‐WB performed schizont purified parasite lysate successfully demonstrated that WM382 indeed stabilized both PMIX and X, while the initial hit compound WM4 stabilizes only PMX (Figure 24). ¹⁵⁹ The PMV inhibitor W601, used as a control did not induce the stabilization of either PMIX or PMX. ²³⁰

Structures of plasmepsin inhibitors and their specific targets. CETSA‐WB was used to confirm WM382 targets plasmepsin IX and X.