Response Heterogeneity of EGFR and HER2 Exon 20 Insertions to Covalent EGFR and HER2 Inhibitors

Genotype-directed therapy is now the standard of care for subsets of patients with advanced non–small cell lung cancer (NSCLC) with tumors harboring oncogenic alterations. This is best exemplified in NSCLC patients with EGFR-mutant or anaplastic lymphoma kinase (ALK)–rearranged lung cancers. In both instances, treatment with EGFR or ALK tyrosine kinase inhibitors (TKI), respectively, is more effective and associated with a higher response rate (RR) and progression-free survival (PFS) than platinum-based chemotherapy (1–5).

The vast majority of clinical trials evaluating EGFR TKI therapy as initial treatment for advanced or recurrent EGFR-mutant NSCLC have included only patients harboring the common drug-sensitive EGFR exon 19 deletion and L858R mutations (2, 5, 6). Collectively, these two mutations account for 85% of all EGFR mutations (7). The remaining 15% of EGFR mutations are comprised of rarer point mutations in exon 18 (G719X) or exon 21 (L861Q) and the exon 20 insertion mutations (7). The exon 20 insertions comprise approximately 4% to 10% of all EGFR mutations, and the majority occur after residue M766 of EGFR (8–11). Unlike other EGFR mutations, patients with EGFR exon 20 insertions rarely respond to gefitinib or erlotinib. A review of 84 patients with exon 20 insertions across different series treated with either gefitinib or erlotinib demonstrated an RR of only 11% with a PFS of 2.4 months (12). Similarly, treatment with afatinib in this patient population is also associated with a low RR and PFS (8.7% and 2.7 months, respectively; ref. 13). Overall survival of patients with EGFR exon 20 insertion mutations is similar to that of patients without EGFR-mutant NSCLC but inferior to that of patients with EGFR exon 19 deletion or L858R advanced NSCLC (9).

Notably, EGFR exon 20 insertion mutations occur in a structurally analogous position as exon 20 insertion mutations in HER2. Mutations in HER2 are oncogenic both in vitro and in vivo (14–16). Unlike EGFR exon 20 mutations, the spectrum of HER2 exon 20 mutations is more narrow, with the A775_G776insYVMA mutation accounting for most of the mutations seen in NSCLC (17–21). As with EGFR exon 20 mutations, there has been limited success in treating patients with HER2 exon 20 mutant NSCLC (22). Strategies to date have included the use of either single-agent HER2 kinase inhibitors or a combination of a HER2 kinase inhibitor with agents targeting downstream signaling. A recent randomized phase II trial compared neratinib with the combination of neratinib and temsirolimus in patients with HER2 mutation–positive NSCLC. Although none of the patients treated with neratinib alone responded (RR: 0%), 3 of 14 (RR: 21%) patients treated with the combination of neratinib/temsirolimus had a PR (23). Collectively, for both EGFR and HER2 exon 20 insertion NSCLC patients, there remains a critical need to develop more effective therapies.

Despite the general lack of efficacy of EGFR or HER2 kinase inhibitors in EGFR or HER2 exon 20 mutant cancers, it is notable that a small but distinct group of patients have had substantial clinical benefits following treatment with EGFR and/or HER2 inhibitors. For example, patients harboring the rare exon 20 A763_Y764insFQEA insertion mutation remain sensitive to erlotinib (8). Further inquiry into the relationship between a specific mutation(s) and corresponding drug sensitivity may both provide biological insights into drug efficacy and identify subsets of patients who could benefit from a treatment strategy using existing drugs.

Dacomitinib is a covalent inhibitor of both EGFR and HER2. In patients harboring EGFR exon 19 deletion or L858R mutations, dacomitinib led to an RR of 76% and PFS of 18.2 months (24). The activity in patients with either EGFR or HER2 exon 20 insertions has also been evaluated. In the phase I study of dacomitinib, 6 patients with EGFR exon 20 insertions were treated and 1 of 6 patients had a sustained PR (25). In a phase II study, 3 of 26 patients with HER2-mutant NSCLC (12%) had a partial response (26). None of the three responders harbored the common A775_G776insYVMA HER2 mutation. This heterogeneity in clinical responses among patients with different EGFR or HER2 exon 20 insertion mutations treated with dacomitinib prompted us to study the different mutations in vitro, identify common features among the dacomitinib-sensitive mutants, and determine whether the in vitro findings would be reflective of the differences observed clinically.

EGFR and HER2 exon 20 mutations were identified from patients with NSCLC at the Dana-Farber Cancer Institute as part of a routine genotyping effort. The methods of detection included Sanger sequencing or targeted next-generation sequencing and have been previously described (27–29). All patients provided written informed consent, were conducted in accordance with the Declaration of Helsinki, and were approved by the Institutional Review Board at Dana-Farber Cancer Institute (DFCI, Boston, MA).

The full-length wild-type cDNAs of EGFR and HER2 were cloned into pDNR-Dual (BD Biosciences). All EGFR and HER2 mutations were introduced using site-directed mutagenesis using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Tech) with mutant-specific primers according to the manufacturer's instructions. Generation of retroviruses, infection into Ba/F3 and NIH-3T3 cells, and selection of stably expressing cell populations was performed as previously described (30, 31). Ba/F3 cells were a generous gift from the laboratory of Dr. David Weinstock at DFCI (in 2014). Transformed Ba/F3 cells were maintained in RPMI1640 supplemented with 10% FBS, streptomycin, and penicillin. Ba/F3 cells were not authenticated, because their STR profile has not been made publicly available. All cell lines used in the study tested negative for mycoplasma as determined by the Mycoplasma Plus PCR Primer Set (Agilent) in August 2016. All experiments were performed within two to three passages following thawing of the cell lines.

Inhibition of growth was assessed by MTS assay according to previously established methods (31, 32). Ba/F3 cells or NSCLC cell lines were exposed to treatment for 72 hours. All experimental points were set up in six wells. The data were graphically displayed using GraphPad Prism version 5.0 for Windows (GraphPad Software Inc.).

Cells grown under the previously specified conditions were lysed in a Lysis Buffer (Cell Signaling Technology). Western blot analyses were conducted after separation by SDS-PAGE electrophoresis and transfer to polyvinylidene difluoride-P immobilon membranes (Millipore). Immunoblotting was conducted according to the antibody manufacturer's recommendations. Anti–phospho-Akt (Ser473), total AKT, phospho EGFR (Tyr1068), total EGFR, phospho Erk (Thr202/Tyr204), total Erk, phospho HER2 (Tyr1221/1222), total HER2, and PARP antibodies were obtained from Cell Signaling Technology. Anti–α-tubulin antibody was purchased from Sigma-Aldrich.

NIH-3T3 cells expressing various HER2 mutations were suspended in growth medium containing 0.35% Noble agar (Sigma-Aldrich) with various concentration of each drug and plated on a bottom layer of 0.5% agar in 6-well plates. Plates were incubated for 2 weeks. Then, the cells were stained with 0.005% crystal violet, and plates were returned to the incubator for 3 hours. The number of viable colonies was quantified using ImageJ software.

N-ethyl-N-nitrosourea (ENU) mutagenesis was carried out as previously described (30, 31). Ba/F3 cells expressing InsGY mutation in EGFR and InsYVMA mutation in HER2 were exposed to 50 μg/mL of ENU (Sigma Aldrich) for 24 hours. Cells were then washed 3 times and expanded in growth media. After expansion, cells were cultured in 96-well plates in the presence of respective drugs (100 nmol/L and 1 μmol/L dacomitinib for EGFR, and 200 nmol/L neratinib and 200 nmol/L dacomitinib for HER2). Culture medium containing each drug was changed every few days, and resistant wells were expanded. Individual drug-resistant clones were isolated and confirmed to be drug resistant. DNA was extracted from resistant cells, and sequencing of tyrosine kinase domain of each gene was performed.

DFCI58 was established from a murine xenograft model and grown in ACL-4 media (Invitrogen) with 10% FBS. Balb/c nude mice were used for subcutaneous transplantation of tumor cells from a pleural effusion. Harvested xenografts were finely minced with autoclaved scissors, filtered through a 100 μm Nylon Cell Strainer, and seeded in complete ACL-4 media. The pleural effusion–derived cell line DFCI127 was directly developed initially in complete ACL-4 media and subsequently transferred and maintained in complete RPMI1640 media.

Between 2004 and 2015, 2,789 patients were tested for EGFR mutations, and between 2009 and 2015, 1,901 patients were tested for HER2 mutations. EGFR mutations were detected in 678 patients (24%). Forty-four EGFR mutations were insertion mutations in exon 20 (7% of EGFR-mutant cases and 1.7% of all tested cases). Forty-seven HER2 mutations were detected in exon 20 (2.5% of all tested cases).

Among the 44 EGFR exon 20 mutations, there were 19 different mutations (Table 1). The most common one was a duplication of codon 768 to 770, resulting in D770_N771insSVD. The second most common one was a duplication of codon 767 to 769, resulting in V769_D770insASV. These two insertions are very similar in their locations and together accounted for 36.3% (16 in 44 cases) of all EGFR exon 20 insertions. Eleven of 19 variants were identified in only 1 patient each. Some of these patients were enrolled in clinical trials of dacomitinib treatment, and 1 patient with D770delinsGY showed partial response (25).

In contrast with EGFR, only eight different variants of HER2 exon 20 mutations were identified (Table 1). The most common mutation resulted in a duplication of codon 772 to 775 (A775_G776insYVMA) and was found in 33 of 48 cases of all HER2 exon 20 insertions (68.8%). Five of the 8 variants were detected only once. Some HER2 exon 20 patients were also enrolled into clinical trials of dacomitinib, and 2 patients with a P780_Y781insGSP and 1 patient with a M774delinsWLV achieved a partial response (26).

In order to study whether the various EGFR exon 20 insertion mutations impart differential drug sensitivities to quinazoline-based EGFR inhibitors, we engineered and expressed the mutations in Ba/F3 cells. We generated Ba/F3 cells harboring five different EGFR exon 20 mutations, including the three most common mutations [D770_N771insSVD (InsSVD), V769_D770insASV (InsASV), and H773_V774insH (InsH)], the dacomitinib-sensitive mutant [D770delinsGY (InsGY)], one rare mutant [Y764_V765insHH (InsHH)], and wild-type EGFR (EGFR WT). All exon 20 EGFR-mutant Ba/F3 cells were able to grow without IL3, with the exception of EGFR WT Ba/F3 cells, which required 10 ng/mL of EGF to sustain growth (data not shown). All of the EGFR exon 20 insertion–containing Ba/F3 cells were resistant to gefitinib, whereas the EGFR WT cells remained sensitive (data not shown). Dacomitinib was a more potent inhibitor of WT EGFR Ba/F3 cells (Fig. 1A; Supplementary Fig. S1). In addition, the InsGY Ba/F3 cells were significantly more sensitive to dacomitinib (IC₅₀, 17.5 nmol/L) than the other four exon 20 insertion Ba/F3 cell lines [P = 0.013; mean IC₅₀ in InsGY mutant (n = 4) vs. mean IC₅₀ in other exon 20 insertion mutants; t test]. Notably, this particular insertion was the one identified in an NSCLC patient who had clinically achieved a dramatic and sustained response to dacomitinib (25). The InsGY Ba/F3 cells were also slightly more sensitive to afatinib while all five exon 20 insertion mutations were relatively more resistant to neratinib. The differential sensitivity of dacomitinib in the InsGY cells was mirrored by its ability to inhibit pEGFR in these cells (Fig. 1B).

We also generated and studied WT and five different HER2 exon 20 insertion Ba/F3 cell lines. The mutations included the most common HER2 mutations [A775_G776insYVMA (InsYVMA) and G776delinsVC (InsVC)]; two mutations identified in patients who clinically responded to dacomitinib [P780_Y781insGSP (InsGSP) and M774delinsWLV (InsWLV)]; and one rare mutation [G778_S779InsCPG (InsCPG)]. All mutants, including WT HER2, were able to grow without IL3 (data not shown). We evaluated the efficacy of neratinib, dacomitinib, and afatinib in all six HER2 exon 20 insertion mutant Ba/F3 cell lines (Fig. 1C). In general, the HER2-mutant Ba/F3 cells were more sensitive to all three drugs compared with the EGFR exon 20 mutant Ba/F3 cells (Fig. 1A; Supplementary Fig. S1). The InsYVMA Ba/F3 cells were the most resistant with IC₅₀ values higher than that seen for WT HER2 (Fig. 1C). These cells were more sensitive to neratinib, although not as sensitive as those the WT HER2 cells (Fig. 1C). As previously noted, the dacomitinib-sensitive Ba/F3 cell lines (InsGSP and InsWLV) contained mutations previously associated with clinical sensitivity to dacomitinib (26). Ba/F3 cells harboring the InsCPG were similarly sensitive to all three covalent inhibitors (Fig. 1C). The IC₅₀ values of the Ba/F3 cells harboring the dacomitinib-sensitive mutations (InsGSP, InsWLV, and InsCPG) were significantly lower than those containing the more resistant HER2 mutants or WT HER2 (P = 0.031; mean IC₅₀ of sensitive vs. resistant mutants; t test). Immunoblot analyses demonstrating inhibition of HER2 phosphorylation corresponded to the relative cellular sensitivities for each of the tested drugs (Fig. 1D). We also performed colony formation assays in soft agar and assessed drug sensitivities of the different HER2-mutant 3T3 cells (Supplementary Fig. S2A and S2B). Similar to the Ba/F3 cells, neratinib was most effective at inhibiting the growth of the InsGSP and InsYVMA colonies, whereas dacomitinib was most effective against the InsWLV cells (Supplementary Fig. S2A and S2B).

Unlike the common EGFR-activating mutations (L858R and Exon 19 deletions), there are no patient-derived EGFR exon 20 mutant cell lines that contain the common EGFR exon 20 insertion mutations. Accordingly, we established two patient-derived cell lines (DFCI58 and DFCI127) harboring EGFR exon 20 insertion mutations (Supplementary Fig. S3A). DFCI58 was established from the adenocarcinoma of a 60-year-old male. He was diagnosed with stage IV disease and, an EGFR H773_V774insNPH exon 20 mutation was identified. Some of this effusion was collected prior to any systemic therapy, propagated in a murine xenograft, and used for establishment of the cell line. DFCI127 was established from the adenocarcinoma of a 42-year-old female. She was diagnosed with stage IV disease and had an EGFR P772_H773insPNP exon 20 mutation. She was treated with carboplatin/pemetrexed/bevacizumab, pemetrexed/bevacizumab, and docetaxel. Following progression on docetaxel, a thoracentesis was performed, which was used to establish the cell line.

We evaluated the efficacy of different EGFR inhibitors in both the DFCI 58 and DFCI 127 cell lines (Supplementary Fig. S3B and S3C). Although higher concentrations of gefitinib had some efficacy in both cell lines, afatinib and dacomitinib were substantially more effective at inhibiting cell growth as well as phosphorylation of EGFR, AKT, and ERK1/2. Immunoblot analysis revealed higher levels of cleaved PARP in response to dacomitinib treatment than to gefitinib treatment (Supplementary Fig. S3C). These data suggested that the apparent inhibition of cell growth induced by the dacomitinib treatment was mediated by enhanced apoptosis in both DFCI cell lines.

The structural basis for the inhibitor resistance of most exon 20 insertion mutants is poorly understood. Only one crystal structure is available for an exon 20 variant (the insNPG mutant), and it reveals an inhibitor-binding site that is essentially indistinguishable from that of the inhibitor-sensitive L858R mutant (8). This insertion and most other insertions encoded within exon 20 are not directly in contact with the ATP-site of the kinase (Fig. 2A), suggesting that inhibitor resistance may arise, at least in part, from differences in conformational dynamics of the kinase.

We next examined our dacomitinib-sensitive versus dacomitinib-resistant exon 20 mutants, in light of available structural information, to gain insight into possible determinants of inhibitor sensitivity. In doing so, we noticed that all of the inhibitor-sensitive mutants contained a glycine in a position two residues beyond the end of the C-helix (Table 2 and Fig. 2A). In wild-type EGFR, this residue is Asp770. This residue lies in a region that is rearranged in the transition between the active and inactive conformations of the kinase. EGFR activity is regulated by repositioning of the C-helix, which rotates into the ATP-site in the active state while rotating out in the inactive state. These conformational transitions are also likely to be important for binding of either substrate or ATP-site inhibitors. In EGFR, Asp770 and the exon 20 insertions lie at the pivot-point of the C-helix. As part of the switch to the inactive conformation, the side chain of Arg776 rearranges to hydrogen bond with the carbonyl group of Ala767 at the end of the C-helix (Fig. 2B). Although this position is accessible in wild-type EGFR, it is blocked by Asp770 in the insertion mutant (Fig. 2C). The insertion residues reposition Asp770, thus sterically blocking Arg776 from accessing the end of the C-helix. Accordingly, we hypothesized that replacement of Asp770 with a glycine, as occurs in the inhibitor-sensitive mutants (Table 2), might restore access of Arg776 and thereby facilitate restoration of wild-type C-helix conformational changes and inhibitor binding.

To test this hypothesis, we mutated Asp770 to glycine in the context of the dacomitinib-resistant Ba/F3 cell lines harboring the H773_V774 insH or D770_N771 insSVD mutants (Supplementary Table S1). Both mutants led to transformation and IL3-independent growth (data not shown). In addition, the proliferation rates were similar with or without the glycine substitution (data not shown). However, the glycine mutation markedly sensitized both variants to inhibition by dacomitinib (Fig. 2D). In the D770_N771 insSVD mutation, the Asp770Gly mutation shifted the IC₅₀ 10-fold [P = 0.046; mean IC₅₀ (n = 3) insSVD vs. GinsSVD; t test]. The effect in the H773_V774 insH variant was more dramatic, with an IC₅₀ of 5 nmol/L [Fig. 2D; P < 0.001; mean IC₅₀ (n = 3) insH vs. GinsH; t test]. We also carried out the converse experiment using the D770delinsGY-mutant cell line by mutating the inserted glycine to alanine, arginine, or aspartic acid (Supplementary Table S1). Notably, contrary to our hypothesis, these substitutions did not uniformly confer resistance; the alanine mutant was slightly more resistant, whereas the arginine and aspartic acid mutants exhibited increased sensitivity (Fig. 2D).

This patient was a 45-year-old female never-smoker who initially presented with advanced lung adenocarcinoma metastatic to the bone. Routine targeted next-generation sequencing of her tumor identified a HER2 V777_G778insGSP mutation (Table 2). Over the course of 2 years, her disease was controlled on first-line cisplatin/pemetrexed chemotherapy (7 months) and then second-line docetaxel (15 months). Because of her HER2-mutant NSCLC, she then initiated third-line vinorelbine/trastuzumab x2 cycles without any evidence of response in bilateral pulmonary disease, complicated by subsequent development of symptomatic brain metastases requiring whole brain radiation. Following radiation, she was increasingly symptomatic with fatigue and ambulatory hypoxia requiring 2-4L of continuous O2 supplementation. The HER2 V777_G778insGSP mutation leads to the same changes in the HER2 protein as does the HER2 P780_Y781insGSP mutation (Table 2). Because of the potential sensitivity of this mutant to covalent EGFR/HER2 inhibitors, the patient was treated with 30 mg of afatinib given once daily. She experienced significant tumor shrinkage (Fig. 3), accompanied by a decrease in the requirement for supplemental oxygen. The response was sustained for 7 months with eventual clinical progression due to leptomeningeal carcinomatosis.

In order to identify how drug-sensitive cancers with EGFR or HER2 exon 20 insertion mutations may develop acquired drug resistance, we performed an ENU mutagenesis assay. This approach has previously led to identification of clinically relevant drug resistance mutations to EGFR TKIs (30, 31). We used the EGFR exon 20 InsGY cells and selected resistant clones in the presence of either 100 nmol/L or 1 μmol/L dacomitinib following ENU exposure. Seventeen drug-resistant clones were expanded and identified to be drug resistant: 12 from the 100 nmol/L and 5 from the 1 μmol/L treated cells. Sequencing of EGFR revealed a T790M mutation in 10 of 12 100 nmol/L and 5 of 5 1 μmol/L treated cells. The remaining 2 drug-resistant clones from the 100 nmol/L treated group did not contain any additional EGFR mutations (data not shown). We next generated Ba/F3 cells using EGFR InsGY that contained either T790M or C797S in cis. EGFR C797 is the site of covalent binding of neratinib, dacomitinib, and afatinib. Mutations in C797 (to C797S) have recently been demonstrated to lead to resistance to mutant selective EGFR inhibitors both in vitro (WZ4002, rociletinib, osimertinib) and in lung cancer patients (31, 33). Both T790M and C797S resulted in drug resistance to dacomitinib and afatinib (Fig. 4A). Immunoblot analysis revealed that EGFR phosphorylation was not inhibited by dacomitinib in Ba/F3 cells harboring either the T790M or C797S mutations when compared with the parental InsGY cells (Fig. 4B).

We analogously performed an ENU mutagenesis study using the HER2 InsYVMA Ba/F3 cells. Twelve drug-resistant clones were identified following selection in either 200 nmol/L of neratinib (n = 5) or 200 nmol/L of dacomitinib (n = 7). Sequencing of HER2 from the resistant cells revealed a secondary C805S mutation in all 12 clones that was not present in the drug-sensitive cells. C805 is the EGFR C797 analogous cysteine residue (Supplementary Fig. S4). We further generated Ba/F3 cells coexpressing an HER2-activating mutation (InsYVMA or InsWLV) and C805S. Cell proliferation assay revealed that these two cell lines were resistant to all three drugs compared with parental cells (Fig. 4C). Immunoblot analysis revealed that HER2 phosphorylation was not completely inhibited in the resistant cells (Fig. 4D).

Lung cancer patients harboring EGFR or HER2 exon 20 mutant tumors represent a unique subset of patients for whom there are currently no effective or approved targeted therapies. However, there is significant clinical and biochemical heterogeneity among EGFR or HER2 exon 20 mutations. Although the vast majority are associated with resistance to EGFR inhibitors, previous studies have identified a unique EGFR exon 20 mutation (A763_Y764insFQEA) that remained sensitive to erlotinib (8). In the present study, we identify four additional rare exon 20 mutations in EGFR (InsGY) and HER2 (InsGSP, InsWLV, and InsCPG) that are uniquely more sensitive to the covalent EGFR/HER2 inhibitor dacomitinib than the more common EGFR or HER2 exon 20 mutations (Fig. 1A and C). Three of these mutations were initially identified from the tumors of lung cancer patients who had clinically responded to dacomitinib, whereas the fourth was identified as an in vitro mutation conferring drug sensitivity (Fig. 1C). With the increased systematic use and availability of comprehensive tumor sequencing, even patients with rare EGFR mutations will continue to be identified, thus it will remain important to determine the correlation between specific mutations and their sensitivity to existing and available treatments.

In the phase II trial of dacomitinib, the mean trough concentration following dosing with 45 mg of dacomitinib once daily was 56.7 to 74.7 ng/mL (approximately 120–160 nmol/L; ref. 24). Of the EGFR exon 20 mutations tested, only the IC₅₀ value of the InsGY mutation was below this concentration (17.5 nmol/L; Fig. 1A; Supplementary Fig. S1), suggesting that this glycine-bearing mutation may have been the reason for the unique clinical sensitivity to dacomitinib in this patient. The findings for HER2 exon 20 mutations are similar albeit more subtle (Supplementary Fig. S1). The HER2 mutations identified from dacomitinib-responding patients were also the most sensitive to dacomitinib in vitro (Fig. 1C), but the differences in the IC₅₀ values between these and mutations from patients with no response (such as HER2 InsYVMA) were not as pronounced as in those with the EGFR exon 20 mutations (Fig. 1C). Collectively, these findings raise the possibility that more potent and specific inhibitors of EGFR or HER2 exon 20 insertion mutations—with IC₅₀ values closer to the Ba/F3 cell lines harboring dacomitinib-sensitive mutations—may be clinically effective against a larger fraction of these cancers.

Remarkably, by studying the drug-sensitive and -resistant EGFR and HER2 exon 20 insertion mutations, we uncovered a common feature among the drug-sensitive mutations: the presence of a glycine at position 770 (Table 2). Furthermore, by introducing a glycine into position 770 of two dacomitinib-resistant EGFR exon 20 mutations, we were able to substantially enhance the efficacy of dacomitinib (Fig. 2D). To date, no features among the different EGFR or HER2 exon 20 mutations have been identified that could predict clinical sensitivity to existing covalent EGFR or HER2 inhibitors. Although our model linking rearrangements of Arg776 and confirmation of the C-helix to inhibitor sensitivity remains speculative, taken together, our findings show that inhibitor resistance is not an inherent feature of activating exon 20 insertions in this region. Indeed, glycine mutations can restore inhibitor sensitivity without abrogating transforming activity. These mutations could act by removing the steric hindrance imposed by Asp770, or by simply increasing the flexibility of the inserted loop. Detailed structural and biophysical studies will be required to directly ascertain the effects of sensitive versus resistant exon 20 insertion mutants on the dynamics of the C-helix, the kinetics of inhibitor binding, and the coupling between the two. Based on our findings, we treated a patient with an HER2-mutant tumor harboring the glycine in position 770 (Table 2), who had previously failed a prior HER2-targeted therapy (trastuzumab). She ultimately had a clinically significant and durable response on afatinib treatment (Fig. 3). Additional prospective clinical validation, specifically among EGFR and HER2 exon 20 mutant patients whose tumors share this glycine residue at position 770, will be necessary to further support our preclinical findings.

Our studies also reveal how the rare but drug-sensitive EGFR or HER2 exon 20 mutant cancers may develop acquired drug resistance. In both cases described here, mutation in the covalent binding site of either EGFR (C797S) or HER2 (C805S) is sufficient to lead to drug resistance (Fig. 4). For EGFR, the T790M secondary mutation also results in drug resistance, to a similar degree as C797S. However, the impact of the C805S mutation may be different than that recently described for the EGFR C797S mutation. Cells expressing an EGFR-activating mutation and C797S in the absence of T790M are resistant to mutant selective EGFR inhibitors (WZ4002, osimertinib, and rociletinib) but remain sensitive to afatinib (31). In the case of HER2, unlike EGFR, the efficacy of afatinib is more likely dependent on covalent binding than on noncovalent inhibition. As HER2 patients treated with afatinib, neratinib, or dacomitinib develop acquired resistance, it will be interesting and clinically significant to determine whether C805S will also occur clinically as an acquired drug resistance mutation. In addition, these observations should inspire the development of alternative strategies to inhibit EGFR and HER2 even in the presence of these resistance mutations as such approaches may be clinically effective for a growing group of patients.

Furthermore, there continues to be a need to develop effective therapies for patients with the more common EGFR exon 20 mutation (InsSVD and InsASV, accounting for ∼1/3 of all exon 20 mutations) and HER2 exon 20 mutations (InsYVMA, accounting for ∼2/3 of all exon 20 mutations). Current approaches include using new EGFR/HER2 kinase inhibitors (AP32788; NCT02716116) or the use of Ado-Trastuzumab Emtansine (NCT02675829). Our studies highlight the heterogeneity among EGFR and HER2 exon 20 insertion mutations, which may affect the efficacy of specific therapies. As additional therapies are developed for this subset of patients, it will be important to continue to correlate efficacy (or lack thereof) with specific genomic variants of EGFR and HER2.

A.J. Redig is a consultant/advisory board member for Boehringer Ingelheim, Medtronic, and Roche/Genentech. G.R. Oxnard is a consultant/advisory board member for Ariad, AstraZeneca, Boehringer-Ingelheim, Genentech, and Novartis. M.J. Eck reports receiving commercial research grant from, and is a consultant/advisory board member for, Novartis. P.A. Jänne reports receiving commercial research grant from AstraZeneca, has ownership interest (including patents) in Gatekeeper Pharmaceuticals, is a consultant/advisory board member for Ariad Pharmaceuticals, AstraZeneca, Boehringer Ingelheim, Chugai Pharmaceuticals, Ignyta, LOXO Oncology, Merrimack Pharmaceuticals, Pfizer, and Roche/Genentech, and receives post marketing royalties from DFCI owned IP on EGFR mutations licensed to Lab Corp. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Kosaka, J. Tanizaki, M. Capelletti, M.J. Eck, P.A. Jänne

Development of methodology: T. Kosaka, M. Capelletti, C.E. Repellin, M. Bahcall

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Kosaka, J. Tanizaki, R.M. Paranal, H. Endoh, M. Capelletti, C.E. Repellin, A. Ogino, A. Calles, A.J. Redig, G.R. Oxnard

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Kosaka, J. Tanizaki, R.M. Paranal, H. Endoh, M. Capelletti, C.E. Repellin, A. Calles, D. Ercan, A.J. Redig, G.R. Oxnard, M.J. Eck, P.A. Jänne

Writing, review, and/or revision of the manuscript: T. Kosaka, M. Capelletti, C.E. Repellin, A. Calles, D. Ercan, A.J. Redig, M. Bahcall, G.R. Oxnard, M.J. Eck, P.A. Jänne

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Lydon, C.E. Repellin, A. Calles, A.J. Redig

Study supervision: P.A. Jänne

Other (Cell line contribution): J. Choi

This study was supported by the NCI of the NIH [R01CA135257 (P.A. Jänne and G.R. Oxnard) and R01CA114465 (P.A. Jänne and G.R. Oxnard), the Cammarata Family Foundation Fund (P.A. Jänne), the Denise and Kevin Hanlon Family Fund for Lung Cancer Research (P.A. Jänne), and HER2 Research LLC (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.