Allele-specific inhibitors inactivate mutant KRAS G12C by a trapping mechanism

Abstract

It is thought that KRAS oncoproteins are constitutively active because their guanosine triphosphatase (GTPase) activity is disabled. Consequently, drugs targeting the inactive or guanosine 5′-diphosphate–bound conformation are not expected to be effective. We describe a mechanism that enables such drugs to inhibit KRAS^G12C signaling and cancer cell growth. Inhibition requires intact GTPase activity and occurs because drug-bound KRAS^G12C is insusceptible to nucleotide exchange factors and thus trapped in its inactive state. Indeed, mutants completely lacking GTPase activity and those promoting exchange reduced the potency of the drug. Suppressing nucleotide exchange activity downstream of various tyrosine kinases enhanced KRAS^G12C inhibition, whereas its potentiation had the opposite effect. These findings reveal that KRAS^G12C undergoes nucleotide cycling in cancer cells and provide a basis for developing effective therapies to treat KRAS^G12C-driven cancers.

Wild-type RAS guanosine triphosphatases (GTPases) cycle between an active, guanosine 5′-triphosphate (GTP)–bound, and an inactive, guanosine 5′-diphosphate (GDP)–bound, state (1, 2). This is mediated by nucleotide exchange factors, which catalyze the exchange of GDP for GTP, and GTPase-activating proteins, which potentiate a weak intrinsic GTPase activity (3). Cancer-causing mutations impair the GTPase activity of RAS, causing it to accumulate in the activated state (4–6). Despite the prevalence of these mutations, no therapies that directly target this oncoprotein are currently available in the clinic (7–9). A recently identified binding pocket in KRAS^G12C (10) now enables the discovery of compounds that potently inhibit KRAS-GTP or effector signaling by this mutant.

Here we characterize a novel compound, ARS853, designed to bind KRAS^G12C with high affinity (11). The structures of ARS853 and previously reported (10) compounds (cmpds) 6 and 12 are shown in fig. S1A. Treatment of KRAS^G12C-mutant lung cancer cells with ARS853 reduced the level of GTP-bound KRAS by more than 95% (Fig. 1A, 10 μM). This caused decreased phosphorylation of CRAF, ERK (extracellular signal–regulated kinase), and AKT. In contrast, even at the highest concentration tested, cmpd 6 or 12 had only a minimal effect on pCRAF and pERK, without affecting KRAS-GTP levels (Fig. 1A and fig. S1B). ARS853 inhibited proliferation with an inhibitory concentration 50% (IC₅₀) of 2.5 μM, which was similar to its IC₅₀ for target inhibition (Fig. 1, A and B). ARS853 (10 μM) inhibited effector signaling (Fig. 1C and fig. S1C) and cell proliferation (Fig. 1D and fig. S2) to varying degrees in six KRAS^G12C mutant lung cancer cell lines, but not in non-KRAS^G12C models (Fig. 1E and fig. S1, C and D). Similarly, it completely suppressed the effects of exogenous KRAS^G12C expression on KRAS-GTP levels, KRAS-BRAF interaction, and ERK signaling (fig. S1E). Inhibitor treatment also induced apoptosis in four KRAS^G12C mutant cell lines (Fig. 1, F to H). Thus, ARS853 selectively reduces KRAS-GTP levels and RAS-effector signaling in KRAS^G12C-mutant cells, while inhibiting their proliferation and inducing cell death.

Fig. 1. Selective inhibition of KRAS^G12C signaling and cancer cell growth by ARS853.

(A) KRAS^G12C mutant cells (H358) were treated with a novel (ARS853) or a previously described (cmpd 6) compound for 5 hours. The effect on the level of active, or GTP-bound, KRAS was determined by a RAS-binding domain pull-down (RBD:PD) assay and immunoblotting with a KRAS-specific antibody. The effect on ERK and AKT signaling was determined by immunoblotting with the indicated antibodies. A representative of at least two independent experiments for each compound is shown. (B) H358 cells were treated for 72 hours with increasing concentrations of the indicated inhibitors, followed by determination of viable cells by the ATP glow assay (n = 3 replicates). (C) The effect of ARS853 treatment on the inhibition of signaling intermediates in a panel of KRAS^G12C-mutant lung cancer cell lines. KRAS^WTA375 cells were used as a control. A representative of at least two independent experiments for each cell line is shown. (D and E) The cell lines were treated over time to determine the effect of ARS853 on the proliferation of G12C (D) or non-G12C (E) KRAS models (n = 3 replicates). (F) The effect of ARS853 treatment on the cleavage of apoptotic intermediates was determined by immunoblotting with the indicated antibodies. A representative of two independent experiments is shown. (G) Cell extracts from the indicated cell lines treated with ARS853 were subjected to a caspase activation assay with the Z-DEVD-AMC reporter substrate. The increase in fluorescence relative to untreated (0) is shown (n = 3 replicates). (H) Following treatment with ARS853 for 24 hours, the cells were stained with annexin V and evaluated by flow cytometry to determine the increase in annexin V–positive cells relative to untreated cells. Data in (B), (D), (E), and (G) are mean and SEM.

In contrast to the rapid inhibition of signaling by kinase inhibitors, inhibition of KRAS^G12C by ARS853 occurred slowly (Fig. 2A and fig. S3). In some cell lines, maximal inhibition of KRAS-GTP occurred in 6 hours; in others, in 48 to 72 hours. To understand this phenomenon, we examined the mechanism of KRAS^G12C inhibition in more detail. To determine whether ARS853 binds to the active or the inactive conformation of KRAS^G12C, we used differential scanning fluorimetry, which assays ligand-induced changes in protein thermal stability (12). Recombinant KRAS^G12C was loaded with either GTPγS or GDP (fig. S4A) and then incubated with ARS853. Samples were incubated at increasing temperatures in the presence of a fluorescent dye that binds to hydrophobic surfaces exposed during thermal denaturation. ARS853 increased the amplitude of the thermal denaturation curve of KRAS^G12C loaded with GDP but not with GTPγS (Fig. 2B and fig. S4B). ARS853 did not alter the denaturation curve of GDP-loaded KRAS^WT (fig. S4C). These data suggest that ARS853 preferentially interacts with inactive, or GDP-bound, KRAS^G12C.

Fig. 2. Inhibition of active KRAS levels despite an interaction with GDP-bound KRAS^G12C.

(A) KRAS^G12C-mutant lung cancer cell lines were treated with ARS853 (10 μM) over time, and cellular extracts were analyzed to determine the effect on KRAS-GTP as in Fig. 1A. (B) Recombinant KRAS^G12C was loaded with GDP or nonhydrolyzable GTPγS and then reacted with ARS853 at a 1:1 molar ratio for 1 hour, at room temperature. The samples were analyzed by differential scanning fluorimentry to determine the shift in thermal denaturation curve induced by ARS853 binding (n = 4 replicates for GDP and n = 3 replicates for GTPγS). Mean and SEM are shown.

KRAS mutants are thought to exist in a “constitutively” active (GTP-bound) state in cancer cells (13). Thus, inhibition of KRAS-GTP levels by a drug that preferentially interacts with GDP-bound mutant KRAS is puzzling. Codon 12 mutations disable the activation of RAS GTPase by GTPase-activating proteins (14–16). It is possible, however, that the basal GTPase activity of KRAS^G12C is sufficient to enable nucleotide cycling in cancer cells. Consequently, we hypothesized that binding of the inhibitor to KRAS^G12C traps it in an inactive (GDP-bound) conformation by reducing its susceptibility to exchange, which then results in the observed time-dependent reduction in cellular KRAS-GTP levels. For this to be the case, (i) inhibition by the drug should require KRAS^G12C GTPase activity. (ii) If KRAS^G12C GTPase activity is constant, the rate of RAS inhibition by the drug should depend on exchange factor activity. (iii) Regulating exchange factor activity should affect drug potency and the kinetics of inhibition.

We tested this hypothesis by using mutations that occur in cancer patients and affect the guanine nucleotide cycle of RAS (Fig. 3A and fig. S5A). A59G is a “transition state” mutant that abrogates GTPase activity (17, 18). When assayed for their GTPase activity (Fig. 3B), KRAS^WT and KRAS^G12C generated approximately three times as much hydrolysis product as the control reactions, whereas KRAS^A59G or KRAS^G12C/A59G had little activity. Thus, KRAS^G12C has basal GTPase activity that is suppressed by the A59G mutation (see also Fig. 3C, lane 5 versus lane 1). ARS853 reduced KRAS-GTP levels and ERK phosphorylation in human embryonic kidney 293 (HEK293) or H358 cells engineered to express KRAS^G12C but not in those expressing KRAS^G12C/A59G (Fig. 3C and fig. S5B). The crystal structure of HRAS^A59G shows that Q61, a critical residue for GTP hydrolysis, is displaced in this mutant (fig. S5C). Q61L impairs in vitro GTP hydrolysis by HRAS (19), and, when expressed in HEK293 cells, KRAS^G12C/Q61L was less sensitive to ARS853 than KRAS^G12C (fig. S5D). Y64A both reduces RAS GTPase activity and impairs the interaction of HRAS with the active site of the guanine nucleotide exchange factor (GEF) SOS (18, 20, 21) (fig. S6, A to C). As expected, ARS853 caused only minimal inhibition of GTP-bound KRAS^G12C/Y64A or signaling driven by this mutant (fig. S6D). Thus, KRAS^G12C GTPase activity is required for its inhibition by ARS853.

Fig. 3. Inhibition of KRAS^G12C requires GTPase activity and is attenuated by nucleotide exchange.

(A) Schematic of the mutations used to block the GTPase activity of KRAS. NTF, nucleotide free. (B) KRAS mutants, expressed and affinity purified from HEK293 cells, were subjected to a GTPase reaction. The orthophosphate product was measured by the Cytophos reagent and expressed as fold change over GTP alone (n = 3, mean and SEM). (C) HEK293 cells expressing HA-tagged KRAS constructs were treated with ARS853 for 5 hours. Cell extracts were evaluated by RBD:PD and immunoblotting with an anti-HA antibody, to determine the effect on GTP-bound mutant KRAS^G12C. A representative of three independent experiments is shown. (D) Extracts from cells expressing the KRAS constructs treated as shown were evaluated by mass spectrometry to determine the percentage of KRAS^G12C bound to the drug. KRAS^G12C-GTP levels in the same extracts were determined as in (C) and quantified by densitometry (n = 3, mean and SEM). (E) HEK293 cells expressing the constructs shown were treated with ARS853 for 5 hours. HA-tagged KRAS was immunoprecipitated (IP) and subjected to a binding assay with the catalytic domain of SOS (SOS^cat). (F) Schematic of the mutations used to biochemically increase (+) or decrease (−) the nucleotide exchange function of KRAS^G12C. (G) HEK293 cells expressing the indicated constructs were analyzed as in (C). (H) The level of active KRAS-GTP was quantified by densitometry. Each replicate is shown. A similar result was obtained in H358 cells (fig. S6F). The baseline level of KRAS^G12C/Y40A detected by the RBD:PD is low because the Y40 mutation impairs the interaction of RAS with RAF (33). See also the effects of N116H and A146V mutations, which increase nucleotide exchange without affecting RAS-RAF interaction (fig. S6G).

The percentage of KRAS^G12C bound to ARS853 in cells was evaluated by mass spectrometry (Fig. 3D, dotted line). GTPase-impairing mutations A59G or Y64A diminished the cellular interaction of ARS853 with KRAS^G12C by ~75 and ~50%, respectively. This effect was inversely proportional to their effect on KRAS^G12C-GTP inhibition (Fig. 3D, solid line), supporting the conclusions that ARS853 interacts preferentially with GDP-bound KRAS^G12C in cells and that GTPase activity is required for the drug to bind and inhibit KRAS^G12C.

ARS853 reduced the interaction of KRAS^G12C with the catalytic subunit of SOS (Fig. 3E) and attenuated SOS-mediated nucleotide exchange by KRAS^G12C (fig. S6E). These findings suggest that ARS853 traps KRAS^G12C in a GDP-bound conformation by lowering its affinity for nucleotide exchange factors. It is therefore possible that nucleotide exchange activity modulates the effect of ARS853. We tested whether RAS mutations that affect nucleotide exchange alter the sensitivity of KRAS^G12C to ARS853 (Fig. 3F). Y40A, N116H, and A146V increase, whereas Y32S decreases, nucleotide exchange (22–24). Mutations that increase exchange attenuated the effect of ARS853 on KRAS^G12C-GTP, whereas Y32S caused a small augmentation of its effect (Fig. 3, G and H, and fig. S6, F and G). These data imply that potentiation of nucleotide exchange reduces sensitivity to ARS853, by lowering the levels of GDP-bound KRAS^G12C available for drug binding.

If the inhibitor works by the trapping mechanism suggested by these data (Fig. 4A), then cellular parameters regulating nucleotide exchange ought to modulate the inhibition of KRAS^G12C. Exposure to epidermal growth factor (EGF), an activator of SOS (25), delayed the inhibition of KRAS^G12C by ARS853 by several hours compared to inhibition in serum-free medium (Fig. 4B and fig. S7A). Treatment with an EGF receptor (EGFR) inhibitor (fig. S7, B and C) or SOS-specific small interfering RNAs (fig. S7D) had the opposite effect. We determined if EGFR signaling affects the inhibition of the mutant KRAS^G12C allele, by expressing hemagglutinin (HA)–tagged KRAS^G12C in EGFR-mutant PC9 cells, in which nucleotide exchange depends on EGFR (26, 27). The specific effect on the mutant allele was determined by pulling down active KRAS and then immunoblotting for HA-tagged KRAS^G12C. ARS853 inhibited GTP-bound KRAS^G12C more potently in the presence of gefitinib, as compared to its absence (fig. S7E). Furthermore, EGFR inhibition enhanced the suppression of KRAS-GTP levels by ARS853 in both KRAS^G12C homozygous and heterozygous cells (fig. S7F). Receptor tyrosine kinase (RTK) signaling and nucleotide exchange are also subject to feedback inhibition by ERK or AKT (28). Mitogen-activated protein kinase kinase (MEK) or AKT inhibitors relieve this feedback and activated RTK signaling (29, 30). Inhibiting these pathways also attenuated the inhibitory effect of ARS853 (fig. S7, G and H). Taken together, these findings show that RTK-mediated nucleotide exchange regulates the kinetics and potency of KRAS^G12C inhibition in cancer cells.

Fig. 4. Tyrosine kinase activation of nucleotide exchange modulates KRAS^G12C inhibition in cancer cells.

(A) A schematic of the trapping mechanism by which ARS853 targets KRAS^G12C-dependent signaling and the role of RTK- or feedback-regulated nucleotide exchange in this process. Structures were adapted from Protein Data Bank accessions 4luc (inhibitor-bound KRAS^G12C-GDP), 4ldj (KRAS^G12C-GDP), and 4l9w (HRAS^G12C-GTP). (B) H358 cells were serum starved overnight followed by treatment with ARS853 (10 μM), with or without EGF (100 ng/ml) for the indicated times, to determine the effect on KRAS-GTP. A representative of two independent experiments is shown. (C) The indicated cell lines, selected because of their relative insensitivity to KRAS^G12C inhibition, were treated with inhibitors of RTK signaling, alone or in combination with ARS853 for 5 days, in order to determine the most effective combinations. The effect of treatment is shown relative to the proliferation of untreated cells. The dots indicate combination treatments resulting in significantly more pronounced inhibition, compared to either drug alone (n = 3 replicates, mean). DMSO, dimethyl sulfoxide. (D) The effect of inhibitors targeting tyrosine kinases on the antiproliferative effect of KRAS^G12C inhibition in KRAS^G12C mutant cell lines (n = 3 replicates, mean and SEM). The concentrations of the inhibitors used in (C) and (D) are noted in Materials and Methods.

The above results also suggest that pharmacologic inhibition of the exchange reaction, as achieved, for example, with inhibitors of RTK signaling, ought to potentiate the antiproliferative effect of KRAS^G12C inhibitors. To test this hypothesis, we evaluated the effects of 11 inhibitors targeting RTKs, or their downstream signaling pathways, in KRAS^G12C-mutant cell lines. In agreement with our model, inhibiting the tyrosine kinase activities of MET, SRC, or FGFR (fibroblast growth factor receptor) enhanced growth inhibition by ARS853 in H1972 and H2030 cells, whereas MEK, ERK, phosphatidylinositol 3-kinase (PI3K), or AKT inhibitor treatment did not have a significant effect (Fig. 4, C and D). In H2122 and Calu-1 cells, the effect of ARS853 was potentiated by concurrent inhibition of EGFR (Fig. 4D). Inhibition of EGFR also enhanced the induction of caspase activity by ARS853 (fig. S7I). These findings support another report showing that EGFR is required for KRAS-driven cancer formation (31) and suggest that the RTKs responsible for modulating the timing of KRAS^G12C inhibition in lung cancer vary from tumor to tumor. They also provide a rationale for selecting optimal combination treatments to achieve maximal KRAS^G12C inhibition in patients.

We thus report an inhibitor that suppresses KRAS^G12C signaling and tumor cell growth by binding to the GDP-bound form of the protein. This reduces the levels of KRAS^G12C-GTP, because the protein retains GTPase activity and undergoes nucleotide cycling in cells. Some RAS mutants have low levels of basal GTPase activity in vitro; yet, perhaps owing to an inability to directly test such activity in vivo, these oncoproteins are assumed to be constitutively GTP-bound in cancer cells. The data presented here, however, indicate that inhibition by ARS853 requires GTPase activity and that nucleotide exchange activity is inversely related to the kinetics and/or magnitude of inhibition. The model suggests that the RAS inhibitor and exchange factors compete for binding to KRAS^G12C-GDP. Drug-bound KRAS^G12C-GDP is not susceptible to exchange and exits the cycle, thus lowering the levels of KRAS^G12C-GTP.

This model also predicts that genetic alterations that completely disable KRAS^G12C GTPase activity, or those that activate nucleotide exchange function, will confer resistance to this class of drugs. Careful selection of combination therapies may in turn prevent or delay the emergence of resistance by enabling a more potent inhibition of KRAS^G12C. The mechanism of action elucidated here establishes a basis for such combination therapies. Inhibitors of receptors dominantly responsible for activation of RAS nucleotide exchange are predicted to enhance activity, whereas inhibitors of downstream pathways that reactivate these receptors are likely to blunt inhibition of mutant KRAS.

KRAS^G12C mutations occur in 16% of lung adenocarcinomas (32), making them one of the most frequent activating genetic alterations in lung cancer, a disease responsible for approximately 1.6 million annual deaths worldwide. No effective inhibitor of mutant KRAS has yet been developed for use in the clinic. Our results suggest that KRAS^G12C allele–specific inhibitors, such as the one described here, merit investigation as potential therapeutics for KRAS^G12C-lung cancers.