The T790M “gatekeeper” mutation in EGFR mediates resistance to low concentrations of an irreversible EGFR inhibitor

Tyrosine kinase inhibitors (TKI) have emerged as important agents in the treatment of several different human cancers. For example, non–small cell lung cancers (NSCLC) harboring mutant epidermal growth factor receptor (EGFR) often respond well to treatment with the EGFR kinase inhibitors gefitinib and erlotinib (1–3). These TKIs compete with ATP in a reversible manner to bind the kinase domain of their targets. Although the initial response to treatment with these agents can be dramatic, most patients eventually relapse due to the acquisition of drug resistance. In approximately half of NSCLC cases that showed an initial response to EGFR TKIs and subsequently progressed, resistance is associated with the emergence of a single recurrent missense mutation T790M within the EGFR kinase domain (4–6). The bulkier methionine residue at position 790 sterically hinders the interaction with inhibitor, effectively preventing binding to the EGFR kinase domain while preserving catalytic activity. An analogous “gatekeeper” mutation (T315I) in the BCR-ABL fusion kinase in chronic myelogenous leukemia cancer cells renders these leukemias resistant to the ABL kinase inhibitors gleevec and dasatinib, suggesting a conserved mechanism of resistance to TKIs (7–9).

Cell-based studies have suggested that an alternative class of irreversible EGFR inhibitors, which bind covalently to EGFR, may be capable of overcoming the effects of the EGFR T790M resistance mutation, thereby potentially providing a second-line treatment to manage such resistance (5, 10). Several of these inhibitors are currently undergoing clinical testing in patients that initially responded to gefitinib or erlotinib and subsequently relapsed (11). Other irreversible inhibitors that target both EGFR and the related protein ErbB2 are also being tested in clinical trials of NSCLC as well as breast cancer. Although the clinical efficacy of this class of inhibitors has yet to be shown, previous experience with TKIs indicates that tumor cells will eventually acquire resistance to even the most effective TKIs. To anticipate molecular mechanisms that may underlie acquired resistance to irreversible TKIs, we modeled acquired resistance to the dual EGFR/HER2-irreversible kinase inhibitor HKI-272 in a NSCLC cell culture model.

To generate resistant subclones of PC-9 cells, these were treated with 400 μg/mL of ethyl methane sulfonate overnight. They were allowed to recover for several passages, and then seeded at a density of 6 × 10⁴ cells/10 cm² dish in 0.2 μmol/L of HKI-272. The relative resistance of these cells to HKI-272, compared with the reversible inhibitors gefitinib and erlotinib, was achieved by seeding 3,000 cells/well in a 96-well plate in the presence of varying concentrations of drugs, followed after 72 h by fixing cells with 4% formaldehyde, staining with Syto 60, and quantifying cell mass by using the Odyssey Infrared Imaging System (LI-COR Biosciences).

NIH3T3 cells were cultured in DMEM supplemented with 10% bovine serum, glutamine and penicillin/streptomycin. Epidermal growth factor (Sigma) was used at a final concentration of 100 ng/mL. Gefitinib and erlotinib were obtained from the Massachusetts General Hospital pharmacy. HKI-272 was provided by Wyeth Pharmaceuticals. EGFR wild-type, EGFR C797A, EGFR L858R, and EGFR C797A/L858R were subcloned into pUSEamp for transient transfection of 293T cells (LipofectAMINE 2000, Invitrogen). EGFR wild-type and EGFR T790M pRV-HygR constructs were previously described (12). EGFR L858R, EGFR Del746-750, EGFR C797A/L858R, and EGFR C797A/Del746-750 were subcloned into pBABE puro for infection of NIH3T3 cells. For retroviral infection studies, EGFR pRV-HygR or pBABE puro constructs were transfected into the 293T-derived BOSC-23 packaging cell lines. NIH3T3 cells were infected with the retroviral supernatant twice in the presence of polybrene. Infected cells were selected using hygromycin (700 μg/mL; Roche) or puromycin (2.5 μg/mL; Sigma).

Cells were lysed with ice-cold buffer [150 mmol/L NaCl, 1% Triton X-100, 50 mmol/L Tris-HCl (pH 8.0), and 1 mmol/L EDTA] containing 1 mmol/L of sodium orthovanadate and 1× protease inhibitor cocktail (Roche). Debris was removed by centrifugation. Clarified lysates were boiled in gel-loading buffer and separated by 4% to 15% gradient SDS-PAGE. Proteins were electrotransferred to nitrocellulose and detected with specific antibodies. The phosphorylated EGFR Y992 antibody was obtained from Biosource, total EGFR antibody was from Santa Cruz Biotechnology, phosphorylated tyrosine antibody was from BD Transduction Laboratory, and β-tubulin antibody was from Abcam. This was followed by a horseradish peroxidase–conjugated secondary antibody. The bands were visualized with enhanced chemiluminescence (Perkin-Elmer) followed by autoradiography.

Multiple clones of exon 20 were sequenced to examine codon 790. Scorpion ARMS primers for the detection of E746_A750del and T790M was done with an EGFR Scorpion Kit (DxS Ltd.) as described by Kimura et al. (13). Briefly, all reactions were done in 25-μL volumes using 5 μL of template DNA, 16 μL of reaction buffer mix (including primers), 3.2 μL of water, and 0.8 μL of Taq polymerase. All reagents were included in this kit. Real-time PCR was carried out using Applied Biosystems 7500 under the following conditions: initial denaturation at 95°C for 10 min, 50 cycles of 95°C for 30 s, and 61°C for 60 s with fluorescence reading (set to FAM that allows optical excitation at 480 nm and measurement at 520 nm) at the end of each cycle. Data analysis was done with ABI 7500 Prism software (SDS version 1.3). The cycle threshold (Ct) was defined as the cycle at the highest peak of the second derivative curve, which represented the point of maximum curvature of the growth curve. Both Ct and Run versus Cycle were used for interpretation of the results.

To model the acquisition of resistance to an irreversible EGFR kinase inhibitor in the context of NSCLC treatment, we used the PC-9 NCSLC-derived cell line. These cells harbor an amplified EGFR allele containing an in-frame deletion in the kinase domain (delE746-A750), and they exhibit exquisite sensitivity to EGFR inhibitors (IC₅₀ ∼20 nmol/L). To establish drug-resistant PC-9 clones through mutational mechanisms, we exposed PC-9 cells to the mutagen ethyl methane sulfonate, and then cultured them continuously in HKI-272, an irreversible dual EGFR/HER2 kinase inhibitor that is currently being evaluated for clinical activity. The cells were maintained in 0.2 μmol/L of HKI-272, which approximates the pharmacologic concentration. Although the vast majority of PC-9 cells were rapidly killed by this treatment, approximately 1 in 2,000 cells gave rise to colonies after ∼4 to 6 weeks in the presence of the drug. Several individual colonies were expanded and they were all found to exhibit an average relative increase in HKI-272 resistance of ∼10-fold compared with the parental PC-9 cells, as measured by their IC₅₀ in a cell survival assay (Fig. 1A). When grown in the absence of drug for up to 6 weeks, all clones maintained their HKI-272–resistant phenotype, suggesting a stable mechanism of resistance (data not shown; Fig. 1B).

Although all of the resistant clones exhibit similar degrees of HKI-272 resistance, further biochemical analyses revealed two distinct classes of clones: those in which phosphorylated EGFR is efficiently suppressed by drug (clones 8, 19, and 23), and those that retain phosphorylated EGFR in the presence of drug (clones 37, 47, and 49; Fig. 2A and B). For simplicity, we refer to clones with drug-suppressible phosphorylated EGFR as class A, and those with nonsuppressible phosphorylated EGFR as class B. Notably, the suppression of phosphorylated EGFR by drugs in class A clones is not a permanent phenotype as EGFR is reactivated when cells are grown in the absence of drug (Fig. 2B). The absence of EGFR activity in class A clones suggests that these cells may have become dependent on an alternative survival pathway. It has been reported that MET gene amplification or activation of the IGF-IR pathway are possible mechanisms of acquired resistance to gefitinib in vitro (14, 15). We therefore tested whether either of these pathways was activated in the class A–resistant clones, but found that neither MET nor IGF-IR proteins exhibited increased phosphorylation levels in the HKI-272–resistant clones relative to the parental PC-9 cells (data not shown).

We next determined that all of the HKI-272–resistant clones were cross-resistant to both gefitinib and erlotinib, suggesting that the acquired resistance does not specifically reflect a mechanism unique to irreversible inhibitors (data not shown; Fig. 3A). Moreover, the clones were significantly more resistant to gefitinib and erlotinib than they were to HKI-272. Thus, these clones exhibit, on average, 10-fold increased resistance to HKI-272, but 100-fold more resistance to gefitinib and erlotinib. Notably, the parental PC-9 cells were slightly more sensitive to gefitinib and erlotinib than they were to HKI-272.

Acquired resistance to gefitinib and erlotinib is frequently associated with a T790M substitution mutation of the EGFR kinase domain, yielding a protein that exhibits reduced binding to drug, while preserving catalytic activity (4–6). Therefore, we determined whether the HKI-272–resistant clones harbor a T790M mutation. Because the endogenous mutant EGFR alleles are amplified in PC-9 cells, we used a sensitive assay that combines quantitative PCR and allele-specific amplification. In this assay, the primers are designed such that only the allele-specific mutation is amplified and the amount of relevant PCR product is then quantified. The number of PCR cycles required to detect the mutation provides a measure of the target molecule present in the reaction. The ΔCt defines the difference between the cycle threshold (Ct) of the mutant reaction and that of a positive control. Thus, the lower the ΔCt, the greater the representation of the mutation within the tested sample. Conversely, the closer the ΔCt is to the cutoff point, the fewer mutations these samples contained.

We first confirmed that all our HKI-272–resistant clones retained the EGFR Del15 mutation found in the parental cell line, confirming that resistance did not arise within a contaminating subpopulation of distinct cells. Specifically, all of the HKI-272–resistant clones as well as the parental PC-9 cells exhibited a ΔCt value between 1 and 2, thereby confirming that the Del15 mutation is present at similarly high levels in all cells (Fig. 3B). When the presence of EGFR T790M was assayed, the resistant clones fell into two groups that corresponded precisely to the two groups defined biochemically. Resistant class B clones, which retain phosphorylated EGFR in the presence of HKI-272, exhibit high levels of T790M within their genomic DNA, as determined by the low ΔCt values. However, class A clones, in which EGFR phosphorylation is efficiently suppressed by drug, exhibit high ΔCt values. These values are near or beyond the cutoff point of 11, indicating either that these clones do not contain a T790M mutation or that this mutation is present at such low levels that they are below the sensitivity of the assay.

To confirm that T790M is present in only one of the two groups of resistant clones, we did a PCR reaction to specifically amplify exon 20, which contains codon 790, and we subsequently sequenced individual clones. Confirming the previous results, we found that T790M is present in a significant number of PCR subclones derived from clone 37 (13%; Fig. 3B). The parental PC-9 cells as well as the resistant clones of class A, however, only harbor wild-type sequences within exon 20. Thus, we conclude that the resistant clones which retain EGFR phosphorylation in the presence of the drug harbor the EGFR T790M mutation.

Uncovering EGFR T790M mutation as a mechanism of resistance to HKI-272 in the cultured cell model was unexpected, as it has previously been reported that HKI-272 could overcome EGFR T790M resistance to gefitinib and erlotinib in cell culture studies (5, 12). Additionally, a recently published report describing a mouse model of NSCLC associated with the expression of EGFR L858R in cis with T790M showed partial regression of adenocarcinomas after treatment with HKI-272 (16). The in vitro studies, however, were done using concentrations of 1 μmol/L or more, whereas dose-limiting toxicity, as established in a recently completed phase I clinical study of HKI-272 was ∼0.2 μmol/L. Thus, it is possible that HKI-272 can overcome T790M only at relatively high doses that are not possible to safely achieve in patients. To test this directly, a dose-response curve was done with HKI-272 in NIH3T3 cells stably expressing EGFR wild-type or EGFR T790M. Indeed, 0.2 μmol/L of HKI-272 was sufficient to inhibit EGFR phosphorylation in cells expressing wild-type EGFR, but not in cells expressing EGFR T790M (Fig. 3C). Notably, EGFR T790M protein exhibits a higher basal level of phosphorylation relative to wild-type EGFR, confirming previous published results (12, 17).

Our findings suggest that the effectiveness seen with this drug at concentrations required in vitro may not be achievable in patients due to drug toxicity, and hence, HKI-272 might not effectively inhibit EGFR T790M at a clinically achievable concentration. Thus, there is clearly a need to develop a more potent irreversible EGFR inhibitor, with potentially reduced toxicity, to overcome the drug resistance associated with secondary T790M mutations. Furthermore, these findings suggest that the use of HKI-272 as first-line TKI therapy may yield EGFR T790M-associated disease progression similar to that seen with the reversible EGFR inhibitors.

Previous studies have shown efficient suppression of the EGFR T790M mutant signaling by relatively high concentrations of HKI-272, suggesting that a more potent irreversible EGFR inhibitor may be effective in the clinical setting of acquired TKI resistance. Notably, several such irreversible inhibitors are currently being tested in clinical studies. If any of these prove to be sufficiently potent to overcome the gefitinib/erlotinib resistance associated with T790M at pharmacologic concentrations, it will still be important to anticipate secondary resistance mechanisms that arise during treatment with those inhibitors. This phenomenon has been well exemplified by the experience with dasatanib as a second-line treatment for cases of chronic myelogenous leukemia with acquired imatinib resistance (18).

Therefore, to anticipate alternative mechanisms of acquired resistance to an irreversible EGFR kinase inhibitor, we considered the potential role of the cysteine residue within the EGFR kinase domain to which HKI-272 binds covalently. As with gefitinib and erlotinib, HKI-272 binds EGFR within its ATP-binding pocket. However, unlike gefitinib and erlotinib, HKI-272 forms a covalent bond with EGFR through a sulfhydryl group on cysteine 797 (19). We speculated that substituting this cysteine residue would result in the decreased ability of HKI-272 to inhibit EGFR, consequently leading to HKI-272 resistance. To test this possibility, we generated an EGFR variant in which cysteine 797 was changed to alanine (C797A) either in a wild-type EGFR background, a L858R background, or a Del15 (Del747–750) background. Whereas EGFR wild-type, L858R, or Del15 were effectively inhibited at HKI-272 concentrations as low as 0.02 μmol/L, EGFR variants C797A, C797A/L858R, and C797A/Del remained phosphorylated at concentrations of 1 μmol/L or higher (Fig. 4A and B). These findings, together with findings from a recent study, indicate that EGFR cysteine 797 is necessary for inhibition by irreversible inhibitors, and mutating this residue abolishes the activity of the irreversible inhibitors (20). Significantly, erlotinib can effectively inhibit EGFR C797A mutants, suggesting that this mutation is specifically resistant to irreversible inhibitors (Fig. 4C).

Our findings also indicate that substituting cysteine 797 with alanine does not detectably affect the EGFR kinase activity. Thus, EGFR C797A mutants are autophosphorylated to a similar extent as the EGFR C797 proteins under basal growth conditions. These findings suggest that treatment with an irreversible EGFR kinase inhibitor with sufficient potency to overcome the effect of the T790M mutation may still give rise to resistance conferred by the acquisition of a C797 substitution of EGFR. Notably, EGFR C797 is conserved in ErbB2/Her2 (C805). Because there are several irreversible TKIs currently undergoing clinical testing in patients with breast cancer harboring amplified ErbB2, it is possible that a mutation at ErbB2 cysteine 805 could contribute to acquired resistance to these drugs. If the EGFR C797 substitution does indeed emerge as a clinical mechanism of acquired resistance to irreversible inhibitors, a combination of gefitinib/erlotinib plus an irreversible inhibitor may prove beneficial, as that combination would overcome both the T790M and C797 resistance mutations.

We thank Wyeth Pharmaceuticals for providing HKI-272.