EGFR Mutations and Resistance to Irreversible Pyrimidine-Based EGFR Inhibitors

EGFR tyrosine kinase inhibitors (TKI) are effective clinical therapies for EGFR-mutant advanced non–small cell lung cancer (NSCLC) patients (1–4). Several randomized clinical trials have demonstrated that EGFR TKIs are more effective, as measured by response rate (RR) and progression-free survival (PFS), than chemotherapy when used as initial systemic treatment for advanced EGFR-mutant NSCLC (1, 4–9). However, the vast majority of patients will develop disease progression following successful treatment with an EGFR TKI. The most common mechanism of acquired resistance, detected in 60% of patients, is a secondary mutation in EGFR at position T790 (T790M; ref. 10). This mutation leads to an increase in ATP affinity, thus making it more difficult for reversible EGFR TKIs gefitinib and erlotinib to bind the EGFR TKI domain (11).

Covalent EGFR inhibitors have emerged as strategies to inhibit EGFR T790M containing cancers. In preclinical models, afatinib, a covalent quinazoline-based EGFR inhibitor, is effective both in models harboring only an EGFR-activating mutation and in those with a concomitant T790M resistance mutation (12). However, in lung cancer patients, afatinib is only effective in EGFR TKI–naïve EGFR-mutant cancers and has a RR of < 10% in patients with NSCLC that have developed resistance to gefitinib or erlotinib (13). Afatinib is a potent inhibitor of both mutant and wild-type (WT) EGFR. Inhibition of WT EGFR leads to toxicities, including skin rash and diarrhea, which limits the ability to escalate afatinib doses in patients to those necessary to inhibit EGFR T790M. Irreversible pyrimidine EGFR inhibitors, including the tool compound WZ4002 and clinical compounds CO-1686 and AZD9291, overcome many of the limitations of afatinib (14–16). They are not only more potent on EGFR T790M, but also selectively inhibit mutant over WT EGFR and hence should lead to increased clinical efficacy and less toxicity compared with afatinib (14–16). In phase I studies to date, treatment with either CO-1686 or AZD9291, has resulted in a RR > 50% in EGFR-mutant EGFR T790M NSCLC patients that have developed resistance to gefitinib or erlotinib (17, 18). In addition, both agents are associated with substantially less skin toxicity than typically observed for EGFR TKIs (17, 18).

Despite the clinical efficacy of CO-1686 and AZD9291, it is fully anticipated that patients will ultimately develop acquired resistance to these agents. To date, little is known about the mechanisms of acquired resistance and whether cross resistance will occur to all irreversible pyrimidine based and to existing EGFR inhibitors. Understanding the mechanism(s) of resistance to this class of agents may both help predict the mechanism(s) that will occur in patients and allow the development of subsequent treatment strategies. Prior studies, using chronic exposure models, have identified reactivation of MAPK signaling and IGF1R signaling as mechanisms of acquired resistance to WZ4002 (19, 20). Secondary mutations in EGFR itself have not been described as a mechanism of acquired resistance.

In the current study, we used a mutagenesis strategy to identify secondary EGFR mutations that impart resistance to WZ4002 and to CO-1686 and AZD9291. We further evaluate how the secondary mutations cause resistance and evaluate their impact on cross resistance to other EGFR-targeted therapies.

WZ4002, AZD9291, and CO-1686 were synthesized using previously published methods (14–16). Gefitinib, afatinib, neratinib, and CL-387,785 were obtained from Selleck Chemicals. The synthesis of TX2-30 is described in Supplementary Materials and Methods. Stock solutions of all drugs were prepared in DMSO and stored at −80°C.

EGFR-mutant NSCLC cell lines HCC827 (del E746_A750), H3255 (L858R), H3255GR (L858R/T790M), H3255DR (L858R/T790M amplified), HCC827EPR (del E746_A750/T790M), H1975 (L858R/T790M), PC9 (del E746_A750), PC9 GR (del E746_A750/T790M), PC9 DR (del E746_A750/T790M amplified), and SNU2315 (del E746_A750/T790M) were obtained from Dr. Adi Gazdar (UT Southwestern, Dallas, TX), ATCC or from the Korean Cell Line Bank (Seoul National University, Seoul, Korea) and have been previously characterized (14, 21–23). All cell lines were authenticated in September 2014 using the Promega Geneprint 16 cell ID system and authentication was performed at the Research Technology Support Facility at Michigan State University. All cell lines were maintained in RPMI1640 (Invitrogen) supplemented with 10% FBS 100 U/mL penicillin, 100 U/mL streptomycin, and 2 mmol/L glutamine. H3255, H3255GR, and H3255DR were maintained in ACL-4 media (Invitrogen) supplemented with 5% FBS, 100 U/mL penicillin, 100 U/mL streptomycin, and 2 mmol/L glutamine. The EGFR-mutant Ba/F3 cells and the NIH-3T3 cells have been previously characterized (14).

N-ethyl-N-nitrosourea (ENU) was purchased from Sigma Aldrich and mutagenesis was carried out as previously described (24). Briefly, L858R, L858R/T790M, DelE746_A750, and DelE746_A750 Ba/F3 cells (1 × 10⁶ cells/mL) were exposed to ENU (50 μg/mL) for 24 hours. Cells were then washed three times with RPMI, and expanded in growth media for 5 to 7 days. Cells were subsequently cultured in 96-well plates (1 × 10⁴ cells/well; total 5 × 10⁵ cells per inhibitor) in the presence of 100 nmol/L WZ4002, 1 μmol/L WZ4002. Wells were observed for growth by visual inspection and resistant wells were expanded in the presence of the corresponding inhibitor. Total RNA was isolated from drug resistant cell lines using TRIzol (Invitrogen) and purified using an RNeasy Minielute Cleanup kit (Qiagen). cDNA was transcribed from 2μg of total RNA with Superscript II Reverse Transcriptase (Invitrogen Life Technologies). The cDNA was used as template for subsequent sequencing of the EGFR tyrosine kinase domain (exons 18–21).

The EGFR L718Q, L844V, and C797S mutations were introduced via site-directed mutagenesis using the Quick Change Site-Directed Mutagenesis kit (Stratagene) according to the manufacturer's instructions. The L718Q mutation was generated using forward primer 5′-aaaaagatcaaagtgcagggctccggtgcgttc-3′ and reverse primer 5′ gaacgcaccggagccctgcactttgatcttttt-3′. The L844V mutation was generated using forward primer 5′-cctggcagccaggaacgtagtggtgaaaaca-3′ and reverse primer 5′-tgttttcaccactacgttcctggctgccagg-3′. The C797S mutation was generated using forward primer 5′-catgcccttcggctccctcctggacta-3′ and reverse primer 5′-tagtccaggagggagccgaagggcatg-3′.All constructs were confirmed by DNA sequencing. The constructs were shuttled into the retroviral vector JP1540 or lentiviral vector JP1698 using the BD Creator System (BD Biosciences). Ba/F3, NIH-3T3 cells were infected with retrovirus and, PC9, PC9GR4, and HCC827 EPR cells infected with lentivirus according to standard protocols, as described previously (25). Stable clones were obtained by selection in puromycin (2 μg/mL).

Growth and inhibition of growth was assessed by MTS assay and was performed according to previously established methods (14, 19). NSCLC or Ba/F3 cells were exposed to treatment for 72 hours and the number of cells used per experiment determined empirically and has been previously established (14, 19). All experimental points were set up in 6 wells and all experiments were repeated at least three times. The data was graphically displayed using GraphPad Prism version 5.0 for Windows, (GraphPad Software; www.graphpad.com). The curves were fitted using a nonlinear regression model with a sigmoidal dose response.

For clonogenic assays, 1,000 cells were seeded in a 6-well plate and allowed to adhere overnight then treated with 400 nmol/L of indicated drug. After 7 days, wells were fixed with 0.5% crystal violet solution. Number of colonies was quantified using Adobe Photoshop version CS4 Extended analysis tool. All experimental points were set up in three wells and all experiments were repeated at least twice. Data were graphically displayed using GraphPad Prism version 5.0 for Windows.

Colony formation was measured using the CytoSelect 96-well Cell Transformation Assay (Cell BIOLABS), according to manufacturer's protocol and as previously described (26). Briefly, 1 × 10³ cells per well (6 wells per condition) were treated with 1.0 μmol/L gefitinib, 400 nmol/L AZD9291, 200 nmol/L WZ4002, 200 nmol/L afatinib alone or in combination with 10 μg/mL, cetuximab for 7 days and growth assayed as described in ref. (26). The experiment was repeated three independent times.

Cells grown under the previously specified conditions were lysed in NP-40 buffer (Cell Signaling Technology). Western blot analyses were conducted after separation by SDS/PAGE electrophoresis and transfer to polyvinylidene difluoride-P membrane (Millipore). Immunoblotting was performed according to the antibody manufacturers' recommendations. Antibody binding was detected using an enhanced chemiluminescence system (PerkinElmer Inc.). Anti–phospho-Akt (Ser-473), anti-total Akt antibodies were obtained from Cell Signaling Technology. The phospho-specific EGFR (pY1068), total ERK1/2, phospho-ERK1/2 (pT185/pY187) antibodies were purchased from Invitrogen. Total EGFR antibody was purchased from Bethyl Laboratories. Tubulin antibody was purchased from Sigma.

For the synthesis of TX2-30, the acrylamide-modified pyrimidine of WZ4002 was synthesized as previously reported (14). The biotin tail was introduced following the procedure from Zhou et al. (27). Briefly, cell lysate was incubated with indicated concentrations of TX2-30 for 1 hour at 4°C. Excess compound was removed using DG-10 column (Bio-Rad). The protein was further denatured with 8 mol/L Urea solution. The addition of Streptavidin beads (Sigma) was followed by 1 hour incubation at room temperature. Beads were washed three times and biotin-labeled EGFR was released by heating beads to 95°C in SDS buffer.

Cells were treated with cetuximab (10 μg/mL) for 16 hours. Following treatment, cells were washed twice with cold PBS and incubated with 1.0 mmol/L bis(sulfosuccinimidy) suberate (Thermo Scientific) for 30 minutes at room temperature. The reactions were quenched with the addition of 20 mmol/L Tris, (pH 7.4). Cells were then washed twice with PBS and then lysed with NP40 buffer.

To identify EGFR mutations that may confer resistance to WZ4002, we performed an ENU mutagenesis screen as previously described (14). Following ENU exposure of Ba/F3 cells harboring the common EGFR-activating mutations (L858R and Del E746_A750) with or without the concomitant T790M mutation, we selected resistant clones in the presence of either 100 nmol/L or 1 μmol/L WZ4002 for 2 weeks. Three mutations, L718Q (codon 2153, T>A), L844V (codon 2530, C>G), and C797S (codon 2389, T>A) were identified with different frequencies in the 4 different Ba/F3 cell lines (Fig. 1A; Supplementary Fig. S1 and Supplementary Table S1). C797 is the site of covalent binding of WZ4002 and previously demonstrated to lead to drug resistance if mutated to serine (14). The L844V mutation has been previously observed in a NSCLC patient (28). Mutations at L718 resulting in L718P, L718V, and L718M mutations have been described but not EGFR L718Q (29–32). The clinical significance of these mutations is presently unknown. We next tested whether L718Q and L844V were oncogenic and caused drug resistance or only the latter. The EGFR T790M mutation is both oncogenic and leads to drug resistance (33, 34). Neither EGFR L718Q nor L844V mutations were sufficient to lead to IL3-independent cell growth in the background of wild-type (WT) EGFR when introduced into Ba/F3 cells (Fig. 1B). Intriguingly, while these mutations led to resistance to WZ4002, they remained sensitive to irreversible quinazoline EGFR inhibitors afatinib, neratinib, and CL-387,785 as demonstrated by growth inhibition using Ba/F3 cells and by the ability to inhibit EGFR phosphorylation (Fig. 1C; Supplementary Table S2). Gefitinib, a reversible quinazoline EGFR inhibitor, was moderately effective, with greater efficacy in the L844V model (Fig. 1C; Supplementary Table S2). We were not able to identify any secondary mutations using the Del E746_A750 Ba/F3 cells using our ENU mutagenesis assay (Fig. 1A; Supplementary Table S1). Engineered EGFR Del E746_A750/L718Q but not EGFR L858R/L718Q Ba/F3 cells demonstrated a growth disadvantage providing one explanation as to why these were not recovered in our original assay (Fig. 1D). In support of this, in the presence of the L718Q mutation, basal EGFR phosphorylation was reduced in EGFR Del E746_A750 cells (Supplementary Fig. S2).

We next aimed to determine how EGFR L718Q and L844V may cause resistance to WZ4002. Examination of the crystal structure of WZ4002 in complex with EGFR T790M (14) reveals that both L718 and L844 are in direct contact with the inhibitor; L718 primarily with the aniline ring and L844 with the pyrimidine ring (Fig. 2A). The L718Q substitution likely sterically interferes with binding of the compound, while the L844V mutation can be expected to remove favorable contacts with the inhibitor (Fig. 2B). To experimentally determine whether WZ4002 binding was affected in the presence of these secondary mutations, we synthesized a biotinylated form of WZ4002, TX2-30 (Supplementary Fig. S3), and used it as an affinity reagent to label EGFR in cell lysates. We exposed NIH-3T3 cells expressing WT or mutant EGFR to TX2-30 and performed an affinity purification of the biotin followed by Western blotting for EGFR. TX2-30 labeled the mutant EGFR proteins more efficiently than WT, consistent with the “mutant” selectivity exhibited by the parental WZ4002 compound (Fig. 2C). However, in the presence of either the L718Q or the L844V mutation, we recovered less WT and mutant EGFR suggesting that these mutations impact drug binding (Fig. 2C).

We further evaluated the impact of the L718Q and L844V mutations when present with both an EGFR-activating mutation and the T790M mutation or only in the presence of an EGFR-activating mutation on drug efficacy (Supplementary Table S2). In the triple mutants (EGFR-activating mutation/T790M with L718Q or L844V), none of the quinazoline-based EGFR inhibitors tested showed any efficacy (Fig. 2D and Supplementary Fig. S4A and S4B). However, in the background of an EGFR-activating mutation alone, afatinib and CL-387,785, were particularly effective at inhibiting the growth of either L718Q or L844V-containing cells which was mirrored by inhibition of EGFR phosphorylation (Fig. 2E and Supplementary Fig. S4C). There were some differences in the efficacy of gefitinib in these models, the EGFR Del E746_A750/L844V and L858R/L844V being more sensitive than the corresponding L718Q models (Fig. 2E and Supplementary Fig. S4C). Given the differential effects of gefitinib and WZ4002 against Ba/F3 cells harboring EGFR L718Q/L844V and EGFR T790M mutations, we evaluated gefitinib alone, WZ4002 alone, or the combination using the ENU mutagenesis assay in EGFR L858R Ba/F3 cells. With gefitinib, we only recovered EGFR T790M–containing clones while with WZ4002 only L718Q-containing clones were recovered (Fig. 2F). However, no clones harboring EGFR secondary mutations were recovered with the combination. Furthermore, the combination was very effective in Ba/F3 cells harboring L858R or Del E746_A750 in conjunction with either L718Q or L844V (Fig. 2G and Supplementary Fig. S4D).

Given the WZ4002 is a tool compound not in clinical development, while CO-1686 and AZD9291 are in clinical studies, we compared the efficacy of all three covalent agents along with afatinib in Ba/F3 cell models and endogenous NSCLC cell lines harboring EGFR-activating mutations alone or in conjunction with EGFR T790M. All four drugs effectively inhibited EGFR Del E746_A750 and L858R-containing cells while WZ4002, CO-1686, and AZD9291 were significantly more effective in T790M-containing models than afatinib (Supplementary Table S3). None of the drugs were effective in triple-mutant models harboring C797S or the EGFR exon 20 insertion mutation A768_V769dupASV (Supplementary Table S3). Afatinib was a substantially better inhibitor of WT EGFR; AZD9291 inhibited WT EGFR at lower concentrations than WZ4002 or CO-1686 (Fig. 3A). Similar trends were observed using endogenous lung cancer cell lines (Supplementary Table S3). In T790M-containing endogenous lung cancer cell lines, including the PC9 GR4 (E746_A750/T790M) cells, AZD9291 was the most potent inhibitor (Fig. 3B; Supplementary Table S3). Analogously, EGFR phosphorylation and downstream signaling was inhibited at lower concentrations of AZD9291 than WZ4002 or CO-1686 in these cells (Fig. 3B).

We further evaluated whether there were differences in the efficacy of the clinical irreversible EGFR inhibitors in models harboring EGFR tertiary mutations. Afatinib was most effective in models harboring an EGFR-activating mutation (L858R or Del 19) along with either the L718Q or L844V mutations (Fig. 4A and B). Of the pyrimidine inhibitors, only AZD9291 inhibited Del/L718Q, Del/L844V and L858R/L844V cells, but was less effective against the L858R/L718Q cells (Fig. 4A). In the triple-mutant Ba/F3 cells, only AZD9291 was effective in models harboring L844V while having a minimal effect against those with the L718Q mutation (Fig. 4A). The efficacy paralleled inhibition of EGFR phosphorylation (Fig. 4C). We also repeated the ENU mutagenesis screen comparing all 3 irreversible pyrimidine EGFR inhibitors and selecting using either 100 nmol/L or 1 μmol/L of each drug (Fig. 4D; Supplementary Table S4). While multiple EGFR mutations were recovered at 100 nmol/L for both WZ4002 and CO-1686, >90% of the clones recovered using AZD9291 harbored a mutation in C797 mutation (Fig. 4D; Supplementary Table S4). At 1 μmol/L, only clones containing mutations in C797 (either C797S or C797G) were recovered for both WZ4002 and AZD9291 while roughly equal numbers of L718Q and C797S clones were recovered with CO-1686 (Fig. 4D). Cells containing the Q791R (codon 2372, A>G) mutation were relatively resistant to W4002 and CO-1686 but remained sensitive to AZD9291 and afatinib (Supplementary Fig. S5A–S5D).

We expressed EGFR del E746_A750/T790M/L718Q or EGFR del E746_A750/T790M/L844V in trans in NSCLC cell lines (PC9GR4 (Del E746_A750 or HCC827 EPR (Del E746_A750/T790M)) to determine whether they were sufficient to cause resistance to WZ4002 in the context of an endogenous NSCLC cell line. While expression of GFP was not associated with resistance, expression of either L718Q- or L844V-containing constructs resulted in resistance to WZ4002, CO-1686, and afatinib (Fig. 5A). The L844V-expressing cells remained sensitive to AZD9291. Similar results were obtained for HCC827 T790M–containing cells (Supplementary Fig. S6A and S6B). In a 7-day clonogenic assays, PC9GR/GFP cells remained sensitive to WZ4002, CO-1686, AZD9291, and afatinib (Supplementary Fig. S6C). However, the L718Q- and L844V-expressing PC9 GR4 cells were only sensitive to AZD9291 (Supplementary Figs. S5B and S6). Analogously, phosphorylation of EGFR and downstream signaling pathways were inhibited by AZD9291 but not WZ4002 or CO-1686 in PC9 GR cells expressing either EGFR del E746_A750/T790M/L718Q or EGFR del E746_A750/T790M/L844V (Fig. 5B). PC9 GR cells expressing EGFR del E746_A750/T790M/C797S were resistant to all tested agents (Fig. 5C and Supplementary Fig. S6D). Given the clinical efficacy of the combination of afatinib and cetuximab in gefitinib/erlotinib-resistant cancers, we evaluated this combination as well as WZ4002/cetuximab in PC9 GR cells expressing the tertiary mutations using a previously established system evaluating cetuximab-based combinations (26, 35). While afatinib/cetuximab had some effect on cell growth (Fig. 5C) and EGFR signaling (Fig. 5D) in cells expressing L718Q or L844V, in neither model was the effect as robust as observed with AZD9291 (Fig. 5C).

Given the potential resistance imparted by the EGFR C797S mutation to WZ4002, CO-1686, and AZD9291, we next evaluated whether C797S-mediated resistance was dependent on the EGFR genomic context. Ba/F3 cells expressing an EGFR-activating mutation and C797S in the absence of EGFR T790M remained sensitive to afatinib (Fig. 4A; Supplementary Table S2). Similarly, while PC9 cells (EGFR E746_A750) expressing EGFR DelE746_A750/C797S were resistant to WZ4002, CO-1686, and AZD9291, they remained sensitive to both afatinib and gefitinib (Fig. 6A and B). Furthermore, in an ENU mutagenesis assay with EGFR L858R Ba/F3 cells followed by selection in gefitinib alone, in AZD9291 alone or in the combination, no resistant clones emerged following gefitinib and AZD9291 treatment (Fig. 6C).

Prior studies have demonstrated that some EGFR-mutant proteins, including L858R/T790M, do not require dimerization for oncogenic transformation (36). In contrast, other mutants, including EGFR L858R, require dimerization and are sensitive to the anti-EGFR antibody cetuximab which disrupts EGFR dimerization (36). Given these findings, we evaluated whether the triple EGFR-mutant proteins existed as monomers or dimers in transformed Ba/F3 cells. While L858R/T790M/L718Q and L858R/T790M/L844V existed predominately as monomers, the L858R/T790M/C797S was present in both monomers and dimers (Fig. 6D). Treatment with cetuximab disrupted dimerization, which resulted in inhibition of EGFR phosphorylation and growth inhibition of the cells (Fig. 6D and E). Unlike the L858R counterpart, Del 19/T790M/C797S exists predominately as a monomer with no effect on EGFR phosphorylation of cell growth following cetuximab treatment (Supplementary Fig. S7A and S7B).

Acquired drug resistance limits the long-term clinical success of targeted therapies including EGFR inhibitors in patients with EGFR-mutant NSCLC. The mechanistic understanding of drug resistance has led to the development of new and novel strategies that are currently being tested in clinical trials. These include the development of mutant-selective EGFR kinase inhibitors, including WZ4002, CO-1686, and AZD9291, which to date have demonstrated encouraging efficacy in both preclinical models and in EGFR-mutant NSCLC patients with EGFR T790M-mediated drug resistance (14–18). However, it is fully anticipated that acquired resistance will also develop to this class of EGFR inhibitors. Given that these EGFR inhibitors are structurally distinct from currently approved EGFR inhibitors, including gefitinib, erlotinib, and afatinib, it is important to understand whether mechanisms of acquired resistance would also lead to cross resistance to existing EGFR inhibitors. Furthermore, as both CO-1686 and AZD9291 are being tested in patients with EGFR T790M-mediated drug resistance, and in those that have an EGFR mutation but are EGFR TKI naïve, the mechanisms of acquired resistance may differ based on the presence or absence of EGFR T790M. These differences may have clinical implications for subsequent treatment with anti–EGFR-based strategies.

Our findings identify C797 (C797S and C797G) as the most common site of secondary mutations mediating resistance to WZ4002, C0-1686, and AZD9291 (Fig. 4D). Cysteine 797 is the site of covalent binding for all three of these agents (14–16). Covalent binding is required to overcome the increased ATP affinity mediated by T790M (11). In prior studies, introduction of C797S into the background of EGFR L858R/T790M or EGFR Del 19/T790M resulted in significant loss of cellular potency for WZ4002 (14). Analogously, a noncovalent analogue of WZ4002, WZ4003, was ineffective in in vitro models of EGFR T790M (14). Emerging clinical data reveal that the C797S mutation is detected in approximately 40% of EGFR-mutant NSCLC patients with T790M who develop acquired resistance to AZD9291 (37). Furthermore, recent studies have also identified mutations in C481 (C481S), the analogous cysteine residue to C797 in EGFR, in patients with chronic lymphocytic leukemia treated with the Bruton tyrosine kinase (BTK) inhibitor ibrutinib suggesting that mutations in this conserved residue may be a common mechanism of acquired resistance to covalent kinase inhibitors (38).

We also identify secondary mutations in EGFR, L718Q, and L844V, both of which occur more frequently in WZ4002- and CO-1686–resistant models (Fig. 4D). These mutations are not oncogenic but mediate drug resistance likely through steric hindrance and impact drug binding (Fig. 2A–C). AZD9291 retains activity against L844V and moderate activity in models harboring EGFR L718Q, although only in the background on an EGFR exon 19 deletion, while both mutations mediate resistance to CO-1686 (Figs. 4A and C and 5A). Our findings suggest that CO-1686 and AZD9291 are distinct from one another and that AZD9291 may still be clinically effective in some CO-1686–treated patients if they were to develop resistance mediated by L844V or L718Q. CO-1686 is structurally more closely related to WZ4002 than to AZD9291 and in all of the models tested CO-1686 is very similar in vitro to WZ4002 (14–16).

Remarkably, models harboring an EGFR-activating mutation alone that develop resistance through L718Q, L844V, or C797S retain sensitivity to quinazoline-based EGFR inhibitors including both gefitinib and afatinib (Fig. 4A and Supplementary Fig. S4C and S5A). This has immediate clinical implications as gefitinib and afatinib are approved EGFR inhibitors and could be clinically effective in EGFR-mutant EGFR TKI–naïve patients who develop these drug resistance mutations following initial treatment with either AZD9291 or CO-1686. Alternatively, combinations of mutant selective and quinazoline-based EGFR inhibitors could be used together as initial treatment for EGFR-mutant NSCLC, as the two classes of EGFR inhibitors are effective against noncross-resistant mutations and together prevent the emergence of resistant clones (Figs. 2F and 5C). Our findings suggest that the sequential and combination approaches should be tested in future clinical trials assuming these tertiary mutations will also arise in patients treated with CO-1686 and AZD9291.

In the presence of EGFR T790M, the tertiary mutations (L718Q, L844V, and C797S) cause resistance to all current EGFR inhibitors (Supplementary Table S2). In only the EGFR L858R/T790M/C797S model, cetuximab still retains some activity (Figs. 6D and E and Supplementary Fig. S7A and S7B). Additional studies are needed to determine whether cetuximab or cetuximab-based combinations are effective in vivo or clinically in EGFR L858R/T790M patients that develop the C797S mutation. Furthermore, additional strategies to inhibit EGFR in the presence of these tertiary mutations, and T790M, are needed and our findings can help guide such strategies. Given the increased ATP affinity imparted by the T790M, it may not be possible to identify a noncovalent EGFR inhibitor with sufficient affinity to inhibit EGFR in the presence mutations in C797 and hence an alternative approach may be necessary.

Our study is performed in vitro using model systems and hence has limitations. It is not known whether all of the mutations identified in our study will also occur in patients. However, ENU mutagenesis studies have been predictive of clinical drug resistance mutations in several prior studies (14, 24). Additional preclinical studies have also identified activation of MAPK signaling and IGF1R signaling as well as epithelial-to-mesenchymal transition as potential resistance mechanisms to WZ4002 and CO-1686 (15, 19, 20). However, as ENU exposure leads to point mutations, mechanisms of drug resistance that involve genomic gains or losses are unlikely to be identified using ENU mutagenesis. Studies of tumors from NSCLC patients that have responded and subsequently developed acquired resistance to CO-1686 and AZD9291 are needed to determine whether the predominant mechanism of resistance involves mutations in EGFR itself or activation of downstream or bypass signaling pathways. The preclinical studies are useful in describing the spectrum of potential mechanisms of acquired drug resistance thus allowing the development of potential treatment strategies that can implemented clinically.

M.J. Eck is a consultant/advisory board member for and reports receiving commercial research grants from Novartis Institutes for Biomedical Research. N.S. Gray has ownership interests (including patents) at Gatekeeper. P.A. Jänne has ownership interests (including patents) at Gatekeeper Pharmaceuticals; is a consultant/advisory board member for Astra Zeneca, Boehringer Ingelheim, Chugai, Clovis Oncology, Genentech, Merrimack Pharmaceuticals, Pfizer, and Sanofi; and reports receiving post-marketing royalties from DFCI-owned intellectual property on EGFR mutations that are licensed to Lab Corp. M.J. Eck, N.S. Gray, and P.A. Jänne have a pending patent related to EGFR inhibitors and methods of treating disorders (the application describes the synthesis of WZ4002) that is owned by the Dana-Farber Cancer Institute and licensed to Gatekeeper Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Ercan, H.G. Choi, N.S. Gray, P.A. Jänne

Development of methodology: D. Ercan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Ercan, M. Capelletti, T. Xie, H.G. Choi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Ercan, C.-H. Yun, M. Capelletti, T. Xie, M.J. Eck, N.S. Gray, P.A. Jänne, H.G. Choi

Writing, review, and/or revision of the manuscript: D. Ercan, T. Xie, M.J. Eck, N.S. Gray, P.A. Jänne

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.S. Gray

Study supervision: P.A. Jänne

This study is supported by grants from the NIH RO1CA114465 (to P.A. Jänne), R01CA135257 (to P.A. Jänne), PO1CA154303 (to P.A. Jänne, M.J. Eck, and N.S. Gray), the Cammarata Family Foundation Research Fund (to M. Capelletti and P.A. Jänne), and the Nirenberg Fellowship at Dana Farber Cancer Institute (to M. Capelletti and P.A. Jänne).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.