Monitoring Drug Target Engagement in Cells and Tissues Using the Cellular Thermal Shift Assay

See allHide authors and affiliations

Science  05 Jul 2013:
Vol. 341, Issue 6141, pp. 84-87
DOI: 10.1126/science.1233606

Drug Targeting

Drug efficacy depends on the extent of binding to a cellular target (often a protein) with adverse effects caused by excessive target binding or by off-target binding. Engagement of a target protein inside cells is influenced by the effective drug concentration and by factors that regulate the protein conformation, making it difficult to predict efficacy based on in vitro affinity studies. Martinez Molina et al. (p. 84) took advantage of the shift in protein thermal stability caused by drug binding to directly monitor target protein-drug interactions in cells. The method was used to monitor drug target engagement in cancer cells and in mouse livers and kidneys.


The efficacy of therapeutics is dependent on a drug binding to its cognate target. Optimization of target engagement by drugs in cells is often challenging, because drug binding cannot be monitored inside cells. We have developed a method for evaluating drug binding to target proteins in cells and tissue samples. This cellular thermal shift assay (CETSA) is based on the biophysical principle of ligand-induced thermal stabilization of target proteins. Using this assay, we validated drug binding for a set of important clinical targets and monitored processes of drug transport and activation, off-target effects and drug resistance in cancer cell lines, as well as drug distribution in tissues. CETSA is likely to become a valuable tool for the validation and optimization of drug target engagement.

Drug development faces multiple challenges that lead to high costs and long development cycles for new therapeutics (13); meanwhile, insights into molecular, cellular and physiological processes have identified a large number of proteins as potential drug targets (4). Therefore, methods that promote accelerated drug development are urgently needed.

The therapeutic effect of most clinically available drugs is achieved through direct binding of the drug to one or a few target proteins. This binding typically occurs at a functional site of the protein and has either an activating or inhibitory effect. The resulting modulation of protein activity in the context of cells and tissues leads to the desired molecular, cellular, and physiological responses. The efficacy of a drug is critically dependent on the extent of its target engagement, and adverse effects are often due to excessive binding of the drug in toxicity-prone cells or its off-target binding to other proteins.

Target engagement by a drug in cells or tissues is determined by its local concentration and binding affinity. The effective drug concentration at the target depends on properties collectively referred to as ADME (absorption, distribution, metabolism and excretion), which dictate the pharmacokinetics and pharmacodynamics of the drug (5). Furthermore, the affinity of a ligand for the drug target can be modulated by changing the activation state of the target protein; for example, by phosphorylation or binding of regulatory proteins or allosteric effector ligands. Consistent with the complexity of protein regulation and ADME processes, residual target engagement can vary significantly during therapy and in individual patients, leading to inherent and acquired drug resistance (6, 7).

A problem in monitoring drug efficacy is that drug binding cannot be directly measured in cells and tissues; instead it is monitored indirectly by studying downstream cellular responses. Consequently, several drugs have failed in advanced clinical trials and later been shown to not act on the predicted drug target within cells (810). We have developed a method to directly monitor target engagement inside cells, based on ligand-induced stabilization of the target protein.

Protein thermal melting curves can be generated for purified proteins, in which the extent of unfolding is measured using different techniques (11, 12); this is exploited in thermal shift assays (TSAs). Thermal shifts at high compound concentrations have been shown to correlate with median inhibitory concentration (IC50) values and affinities measured by other methods (13, 14) and are widely used for characterization of ligand binding in structural biology and drug screening in a broad range of affinities. However, these methods have been applied only to purified proteins. We have developed a process in which multiple aliquots of cell lysate were heated to different temperatures. After cooling, the samples were centrifuged to separate soluble fractions from precipitated proteins. We then quantified the presence of the target protein in the soluble fraction by Western blotting (fig. S1).

Surprisingly, when we evaluated the thermal melt curve of four different clinical drug targets in lysates from cultured mammalian cells, all target proteins showed distinct melting curves.When drugs known to bind to these proteins were added to the cell lysates, obvious shifts in the melting curves were detected (Fig. 1, A to D). Because this method has conceptual similarities to the TSAs used for purified proteins, we named it the cellular thermal shift assay (CETSA). Further characterization of basic aspects of the method is found in the supplementary materials (figs. S2 and S3).

Fig. 1

CETSA melt curves and ITDRFCETSA in lysate. CETSA curves for four therapeutic targets, with their corresponding inhibitors measured in cell lysates; CDK2 (A) and CDK6 (B); v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) (C); and MetAP2 (D). Examples of ITDRFCETSA in lysates for MetAP2 with TNP-470 (E) and CDK2 with AZD-5438 (F) at fixed temperatures (72° and 52°C, respectively) are shown. Data are presented as means ± SEM, n ≥ 3 independent experiments.

To investigate drug concentration effects, we established an isothermal dose-response procedure in which lysate aliquots were exposed to different concentrations of drug while time of heating and temperature were kept constant. Unlike traditional dose-response experiments in which half-saturation points are related to affinities, the response here is typically reached at higher drug concentrations (see discussion in the supplementary materials). Nevertheless, this procedure yields a characteristic fingerprint of the target engagement along the drug concentration axis (Fig. 1, E and F). This isothermal dose-response fingerprint (ITDRFCETSA) was used to estimate relative differences in drug concentration required to establish a similar extent of target engagement.

Cancer cells actively import metabolites such as folates, nucleosides, and nucleobases that are critical for DNA synthesis (15). Folate and nucleoside analogs hitchhike on these import systems and accumulate in cancer cells (16). Dihydrofolate reductase (DHFR) and thymidylate synthase (TS) are targets for the antifolate cancer drugs methotrexate and raltitrexed (17, 18). DHFR and TS were used to determine whether CETSA could be used in intact cells as well as in lysates. Cells were exposed to either methotrexate or raltitrexed, washed, heated to different temperatures, cooled, and lysed. The cell lysates were cleared by centrifugation, and the levels of soluble target protein were measured, revealing large thermal shifts for DHFR and TS in treated cells as compared to controls (Fig. 2, A and B). To assess the relative binding in cells and lysates, an ITDRFCETSA experiment with TS was conducted, demonstrating that after 3 hours of exposure to the drug, the ITDRFCETSA of raltitrexed in lysate gave an at least three orders of magnitude higher value than in intact cells, suggesting a highly active transport of the drug into the cell (Fig. 2C). This transport results in dramatic accumulation of the drug in proliferating cells, which is clearly a key factor in the therapeutic efficacy. A similar shift of ITDRFCETSA was obtained for DHFR with methotrexate (fig. S4A). A time-course experiment with raltitrexed indicated that after 2 to 3 hours, the ITDRFCETSA of TS was saturated (Fig. 2D); this saturation was not due to depletion of the drug in the medium (fig. S4B). In contrast, a similar experiment using starved cells resulted in a limited accumulation of the drug (fig. S4C). In addition to providing information about the transport of this folate analog, these experiments prove that CETSA monitors target engagement in intact cells. This is further supported by dye exclusion experiments supporting the notion that the cell membranes remain intact during these experiments (fig. S2).

Fig. 2

Monitoring of antifolate drug transport and activation. CETSA curves in intact cells versus lysate for DHFR (A) and TS (B) with methotrexate and raltitrexed respectively. (C) ITDRFCETSA at 52°C in intact cells versus lysate for TS using raltitrexed (D) Time course for import of raltitrexed in intact cells. ITDRFCETSA at 52°C for DHFR (E) and TS (F) in intact cells after inhibition of methotrexate activation by blocking polyglutamate synthetase with suramin. (G) ITDRFCETSA at 52°C showing the effect of nucleoside transporter inhibitor NBMPR on import of 5-FU and target engagement for TS in intact cells. Data are presented as means ± SEM, n ≥ 3 independent experiments.

Because resistance to folate and nucleoside analogs often occurs through changes in transport or activation events, we investigated the effect of inhibitors of such pathways. Folate analogs are activated through polyglutamation by polyglutamate synthetase; this helps to retain these compounds in the cell by minimizing their export (19). Monitoring the TS and DHFR ITDRFCETSA for methotrexate in the presence of suramin, an inhibitor of polyglutamate synthetase, demonstrated that this value differs >30-fold between treated and untreated cells for both TS and DHFR. This implies that the activation of methotrexate is severely compromised by suramin (Fig. 2, E and F). Nitrobenzylmercaptopurine riboside (NBMPR) is an inhibitor of the import of fluorouracil (5-FU), which in its activated form, FUMP, is a potent inhibitor of TS (19). The ITDRFCETSA shift for TS in cells treated with 5-FU suggests that its import is inhibited by the addition of NBMPR (Fig. 2G). Thus, CETSA can monitor complex processes known to be involved in drug resistance.

To determine whether CETSA could yield information on target engagement in animals, we used the methionine aminopeptidase-2 (MetAP2) inhibitor TNP-470, an antiangiogenic compound currently in clinical development for use in treatment of solid tumors (20). In a double-blind study, mice were systemically treated with TNP-470 and compared to mice in a control group. To ensure reproducibility and facilitate sample handling in the subsequent experiments, the harvested organs were frozen and lysates were prepared before heating. Because the compound TNP-470 is a covalent inhibitor, this is likely to reflect target engagement in the cell. The corresponding thermal melt curves of MetAP2 from mouse liver samples clearly distinguished the treated mice from the nontreated mice (Fig. 3A).

Fig. 3 Monitoring of TNP-470 target engagement in tissue samples from mice.

(A) CETSA curves of MetAP2 in mouse liver lysates from untreated mice and mice treated with TNP-470 at 20 mg per kilogram of body weight. (B) ITDRFCETSA at 72°C of MetAP2 in liver and kidney at six different TNP-470 dosage levels. Data in (B) are presented as means ± SEM, n = 3 mice per data point.

To investigate dose dependence, we administered six different doses of TNP-470 to mice. Comparison of in situ CETSA response curves for MetAP2 in mice liver and kidney lysates demonstrated a 50-fold difference in ITDRFCETSA (Fig. 3B). The lower effective concentration of TNP-470 in the liver than in the kidneys could be explained by the large amount of catabolic enzymes present in this organ as well as renal excretion of active metabolites (21).

Target specificity is important for optimizing drug efficacy and minimizing their adverse effects (22). Cyclin-dependent kinases (CDKs) are major regulators of cell division and differentiation, and inhibitors of these are developed for cancer and inflammation therapy (23, 24). PD0332991, currently in clinical development for cancer treatment, was used to investigate target specificity in cells using CETSA. The ITDRFCETSA values show that it is selective for CDK4 and CDK6; no significant target engagement is seen for either CDK2 or CDK9, even at very high concentrations of the inhibitor in intact cells (Fig. 4A). This confirms results from activity assays, which have shown the selectivity of PD0332991 for CDK4 and CDK6 over CDK2 (25).

Fig. 4 Monitoring of drug and target specificity.

(A) ITDRFCETSA at 45°C for CDKs in intact cells showing specificity for PD0332991 to CDK4 and CDK6 over CDK2 and CDK9. (B) ITDRFCETSA at 56°C for the BRAF inhibitor PLX4032 in lysate versus intact cells for wild-type and V600E BRAF and (C) the corresponding CETSA curves. (D) CETSA curves in lysates with the PARP-1 inhibitors olaparib and iniparib and (E) the corresponding ITDRFCETSA at 50°C. Data are presented as means ± SEM, n ≥ 3 independent experiments.

Cells expressing mutated forms of oncogenic proteins may (or may not) display altered drug susceptibility. BRAF is a member of the MAPK/ERK pathway that controls proliferation and apoptosis; it is mutated in 7% of human cancers; for example, in more than 50% of melanomas (26). Vemurafenib (PLX4032) is highly effective against melanomas harboring the frequent BRAF V600E mutation (27). Studies with purified BRAF kinase domains show a modest increase in the PLX4032 IC50 for V600E BRAF as compared to the wild type (27, 28). When comparing target engagement of PLX4032 in intact cells with the V600E substitution and wild-type BRAF, the ITDRFCETSA values demonstrate similar target engagement levels for the two protein variants (Fig. 4B), and their corresponding CETSA curves show comparable stabilization at high compound concentration (Fig. 4C).

CETSA may help to validate clinical drug candidates before significant investments in clinical trials are made. For example, the proposed PARP-1 inhibitor iniparib reached phase III clinical trials (10), where it showed no efficacy, and was subsequently shown to lack activity against PARP-1 in living cells (10, 29). We used CETSA to compare the target engagement of PARP-1 for iniparib and olaparib, a well-established PARP-1 inhibitor in clinical development (30). Although binding of olaparib induced a large thermal shift of PARP-1, iniparib failed to induce a shift (Fig. 4D). This is also apparent when studying the ITDRFCETSA results for both compounds (Fig. 4E). Apparently, the mechanism of action of iniparib is not via physical binding to PARP-1; instead, iniparib may kill cancer cells by unspecific modification of cysteine residues (29).

In this work we have shown that CETSA can be used to monitor target specificity, drug transport and activation, as well as dose-dependent target engagement in animals. The relative ease of establishing CETSA for the targets used in this study suggests that this method could be applicable for a wide range of soluble intracellular and extracellular drug targets. However, the method is not likely to work for highly inhomogenous proteins or for proteins in which unfolding of the ligand-binding domain does not promote aggregation.

There are no direct methods for monitoring the physical binding of a drug to its target in intact cells (3133). For cell lysates, physical target engagement can be assessed using drug-induced proteolysis protection (34) or protection of oxidation of specific amino acids (35). Apart from identifying drug targets of orphan ligands in cell lysates (31), the applicability of these methods for monitoring dynamic events in intact cells remains to be proven. When a selective and high-affinity ligand is available, in situ competition experiments can be performed for extracellularly accessible proteins or receptors to monitor target engagement (36).

In addition to the applications in preclinical drug development discussed above, CETSA has potential for monitoring drug efficacy at the target level in patients. For example, CETSA could assist in determining appropriate drug usage and dosage, as well as monitor when acquired drug resistance at the target engagement level has developed; for example, during cancer therapy. The versatility of CETSA is likely to make it a valuable tool for drug research and development.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Table S1

References (3742)

References and Notes

  1. Acknowledgments: We thank K. Johansson [Karolinska Institute (KI), Department of Medical Biochemistry and Biophysics (MBB)] for assistance with cell culturing, H. Schüler KI/MBB for PARP-1 constructs, and T. Lundbäck KI/Chemical Biology Consortium Sweden for valuable discussions. This work was supported by grants from the KI (Distinguished Professor Award), the Swedish Research Council (Vetenskapsrådet), and the Swedish Cancer Society (Cancerfonden) to P.N. and Y.C., as well as a Strategic Centre grant from Nanyang Technological University (Singapore) to P.N. and grants from the Karolinska Institute Foundation, the Torsten Söderbergs Foundation, and the European Research Council advanced grant ANGIOFAT (project no. 250021) to Y.H.P.N. is the inventor on a pending patent application covering aspects of the CETSA method, to be assigned to Pelago Bioscience AB P.N. and D.M.M. are cofounders of Pelago Bioscience AB, which will develop CETSA for commercial applications.
View Abstract

Stay Connected to Science

Navigate This Article