Mobocertinib (TAK-788): A Targeted Inhibitor of EGFR Exon 20 Insertion Mutants in Non–Small Cell Lung Cancer

Adenocarcinoma, the most common histologic non–small cell lung cancer (NSCLC) subtype (https://seer.cancer.gov/csr/1975_2016), is divided into molecular subtypes with distinct and often actionable oncogenic driver mutations (1). EGFR mutations are present in 12% to 38% of lung adenocarcinomas (2, 3) and are more common in patients with NSCLC in East Asia (40%–64%) than in Europe and North America (7%–37%; refs. 4, 5). EGFR activating mutations are often found in the first four exons (18 to 21) of the EGFR tyrosine kinase domain (6). Common activating mutations, which account for approximately 78% to 85% of all EGFR activating mutations, include exon 19 in-frame deletions and L858R substitution in exon 21 (7–9). First-generation (erlotinib, gefitinib), second-generation (afatinib, dacomitinib), and third-generation (osimertinib) EGFR tyrosine kinase inhibitors (TKI) are approved by the FDA for treatment of patients with metastatic NSCLC with these common mutations and have been associated with high objective response rates (50%–77%; refs. 10–13).

The efficacy of first- and second-generation EGFR TKIs in patients with common activating mutations is impaired by the emergence of drug resistance, which occurs most frequently via acquisition of a secondary mutation, T790M, in the EGFR kinase domain “gatekeeper” residue, impairing drug binding (6, 14). Third-generation TKIs such as osimertinib were developed to address this issue and offer a therapeutic option to patients whose tumors bear combined common and gatekeeper mutations (15). Second- and third-generation TKIs bind irreversibly and covalently to cysteine 797 in the adenosine triphosphate (ATP) binding site of the EGFR kinase (16, 17). Uncommon mutations, which account for approximately 5% of all EGFR activating mutations, include point mutations at amino acids in exons 18 (e.g., G719X), 20 (e.g., S768X), and 21 (e.g., L861X; refs. 7, 8). Only afatinib is currently approved by the FDA for use in patients with a specific set of uncommon nonresistant mutations (S768I, L861Q, and G719X; ref. 18); the clinical activity of osimertinib against these mutations is under investigation (19).

The EGFR exon 20 insertion (EGFRex20ins) mutation class encodes a family of mutants with amino acid insertions clustered between positions 762 and 774 [e.g., V769_D770insASV (ASV)] that result in constitutive activation of EGFR (20). EGFRex20ins mutations represent approximately 6% to 12% of all EGFR-mutated NSCLC cases (7–9, 21) and are enriched in nonsmokers and Asian patients (8, 9). Almost all EGFRex20ins mutations confer in vitro and clinical resistance to first- and second-generation EGFR TKIs (7, 8, 20, 22–24), although osimertinib may have clinical efficacy against some resistant mutants (25–28). A small subset of atypical EGFRex20ins mutations [e.g., A763_Y764insFQEA (FQEA)] is associated with sensitivity to first-, second-, and third-generation EGFR TKIs (24, 29). No EGFR TKIs are currently approved for use in patients with EGFRex20ins mutations.

The main challenge of designing TKIs that target EGFRex20ins mutants is achieving selectivity over wild-type (WT) EGFR (20, 30). Common mutations (L858R, exon 19 deletion) result in EGFR kinases with destabilized inactive forms that display reduced ATP affinity relative to WT EGFR (20). First-generation EGFR TKIs take advantage of this difference in ATP affinity to obtain high affinity for mutant EGFR and selectivity over WT EGFR (31, 32). In contrast, structural modeling of EGFRex20ins variants suggests that the insertion of in-frame amino acids at the C-terminal of the C-helix leads to conformational changes closely resembling the active form of WT EGFR, resulting in less selectivity for first- and second-generation EGFR TKIs compared with WT EGFR (20, 24, 33). Inhibition of WT EGFR in normal tissues is associated with dose-limiting toxicities that typically manifest as cutaneous, mucosal, and gastrointestinal adverse events (34–36). EGFR TKIs that inhibit WT EGFR more potently than the targeted oncogenic EGFR mutant are unlikely to allow for dosing at clinically efficacious levels without the development of unacceptable levels of skin and gastrointestinal toxicities (33).

Mobocertinib (TAK-788; AP32788) is a novel oral EGFR TKI that was designed to address the unmet need in patients with EGFRex20ins-mutant NSCLC. Using an iterative, structure-guided strategy, similar to that employed in the development of other targeted TKIs (37), mobocertinib was designed to potently inhibit oncogenic variants containing activating mutations in exon 20, with selectivity over WT EGFR.

To define the relative incidence of EGFR TKI–resistant EGFRex20ins mutations globally, we extracted data from the American Association for Cancer Research Project Genomics, Evidence, Neoplasia, Information, Exchange (GENIE) database (38). Among 155 patients with lung adenocarcinoma with EGFRex20ins mutations, the most common mutations were A767_V769dup [V769_D770insASV (ASV), 21%], S768_D770dup [D770_N771insSVD (SVD), 18%], and N771_H773dup [H773_V774insNPH D770_N771insNPG (NPH), 9%; Fig. 1A; Supplementary Table S1]. All three EGFRex20ins mutations (ASV, SVD, and NPH) were captured in our screening cascade, as well as D770_N771insNPG (NPG) and FQEA.

To identify an EGFRex20ins inhibitor with the necessary degree of selectivity for mutant EGFR over WT EGFR to be clinically viable, we established an assay that could predict clinical efficacy of first-, second-, and third-generation EGFR TKIs against five classes of EGFR mutants. Activity against WT EGFR was assessed by determining concentrations required to inhibit EGFR phosphorylation in A431 cells, and activity against EGFR mutants was assessed by determining concentrations required to inhibit viability of 15 Ba/F3 cell lines engineered to have their survival dependent on mutant human EGFR activity. We then compared mean erlotinib, gefitinib, afatinib, and osimertinib concentrations required to inhibit WT and mutant EGFR activity by 50% (IC₅₀; Fig. 1B; Supplementary Table S2).

Consistent with their approved indications, all four EGFR TKIs inhibited EGFR with common activating mutations [E746-A750 del (D) and L858R (L)] more potently than WT EGFR, but only osimertinib inhibited T790M-containing mutants [E746-A750 del + T790M (DT) and L858R + T790M (LT)] more potently than WT EGFR. None of the four TKIs inhibited the activity of a triple mutant (L858R + T790M + C797S; LTC), consistent with known clinical resistance mechanisms. Also consistent with their approved indications, afatinib, but not erlotinib or gefitinib, inhibited all five uncommon EGFR mutants more potently than WT (the clinical benefit/risk of osimertinib, which inhibited some but not all uncommon mutants more potently than WT EGFR, is being explored; ref. 19). Importantly, consistent with the lack of meaningful clinical activity, erlotinib, gefitinib, and afatinib all exhibited greatly reduced activity against all five EGFRex20ins mutants (NPG, ASV, NPH, SVD, and FQEA) compared with WT EGFR, except for afatinib against the structurally distinct FQEA mutant. In addition to FQEA, osimertinib inhibited NPG and ASV mutants more potently than WT EGFR but inhibited SVD and NPH mutants less potently than WT. Thus, these models were used to guide discovery of a TKI for patients with EGFRex20ins mutations based on its ability to inhibit all EGFRex20ins mutants more potently than WT EGFR.

The design and discovery of mobocertinib involved structure-guided design and structure–activity relationship development, including validation of various in vitro and in vivo nonclinical assays. The conformation of EGFRex20ins mutants largely resembles that of WT EGFR in the ATP binding site (Fig. 2A). With no amino acid substitutions in the binding site, it was challenging to address selectivity directly. However, we adopted an approach to target potential structural nuances between the proteins in the vicinity of the α C-helix to gain selectivity by targeting portions of the binding site not exploited by osimertinib. One such nuance was the observation that, upon NPG insertion, both the P-loop and the C-helix are shifted, making the binding pocket smaller (39).

Like afatinib and osimertinib, mobocertinib was designed to form a covalent interaction with cysteine 797 in EGFR. Compared with a reversible binding mechanism, this irreversible binding mechanism leads to increased potency via higher affinity binding, more sustained EGFR kinase activity inhibition, and greater overall selectivity, as only a limited number of other kinases possess a cysteine in the equivalent position. Amino acid insertion in exon 20 results in a shift of the C-helix to the active conformation of EGFR; this alteration is quite distant from the ATP binding pocket (Fig. 2A–D). An osimertinib docking model in the EGFRex20ins NPG mutant (Fig. 2B) revealed an unoccupied pocket accessible by substitution on the pyrimidine ring. Mobocertinib's isopropyl ester was designed specifically to interact with the gatekeeper residue within this pocket and to probe subtle conformational differences between EGFRex20ins mutants and WT EGFR. This key structural feature results in increased affinity for the EGFRex20ins mutant compared with osimertinib.

An LC/MS study was conducted to determine whether mobocertinib forms an irreversible complex with WT EGFR and an EGFRex20ins mutant. After incubation with mobocertinib, LC/MS analysis revealed a shift in the observed mass of WT and mutant EGFR that corresponded to the exact molecular weight of mobocertinib (Supplementary Table S3), demonstrating formation of a covalent, irreversible interaction.

The inhibitory profile of mobocertinib was evaluated in the Ba/F3 screening assay. Mobocertinib inhibited the activity of 14 of 15 mutant EGFR variants tested (IC₅₀s 2.7–22.5 nmol/L) more potently than WT EGFR (IC₅₀ 34.5 nmol/L) (Fig. 2E and F). The only exception was the C797S-containing triple mutant (i.e., LTC; Fig. 2E; Supplementary Table S2), which conferred resistance to mobocertinib (IC₅₀ > 10,000 nmol/L). Taken together with the above MS results, this is consistent with its mechanism of covalent binding. Mutants inhibited more potently than WT EGFR included all four variants containing common activating mutations with or without the T790M resistance mutation (D, L, DT, and LT: IC₅₀s 2.7–21.3 nmol/L) and all five variants containing uncommon activating mutations (G719A, G719S, S768I, L861Q, and L861R; IC₅₀s 3.5–20.2 nmol/L). Most importantly, mobocertinib inhibited all five variants containing EGFRex20ins mutations (FQEA, NPG, ASV, NPH, and SVD; IC₅₀s 4.3–22.5 nmol/L) more potently than WT EGFR.

The activity of mobocertinib against EGFRex20ins and other classes of mutants was also assessed in cancer cell lines. To characterize mobocertinib's activity against EGFRex20ins mutations, we used two patient-derived cell lines, CUTO14 and LU0387. CUTO14 cells were derived from the pleural effusion of a patient with lung adenocarcinoma harboring an EGFRex20ins ASV mutation who had no prior TKI treatment (39). LU0387 cells were derived from a patient with lung adenocarcinoma harboring an EGFRex20ins NPH mutation. These cells lack activating mutations in a variety of other potential driver oncogenes (KRAS, ALK, BRAF, MET, PIK3CA, AKT, and ERK), consistent with the EGFRex20ins mutation being the primary oncogenic driver. In CUTO14 cells, mobocertinib robustly inhibited EGFR signaling, reaching 80% and 100% inhibition of phosphorylated EGFR (pEGFR) at concentrations of 100 nmol/L and 1,000 nmol/L, respectively, whereas osimertinib induced 38% and 63% inhibition at the same concentrations (Fig. 3A). Mobocertinib inhibited viability of CUTO14 cells (IC₅₀ 33 nmol/L) with greater potency than erlotinib, gefitinib, afatinib, and osimertinib (IC₅₀s 2,679, 1,021, 66, and 575 nmol/L, respectively). Mobocertinib inhibited viability of LU0387 cells (IC₅₀ 21 nmol/L) with greater potency than WT EGFR phosphorylation (IC₅₀ 35 nmol/L) and with potency substantially greater than that of erlotinib, gefitinib, and osimertinib (IC₅₀s 2,793, 364, and 195 nmol/L, respectively; Fig. 3B). Although the potency of afatinib (IC₅₀ 20 nmol/L) in LU0387 cells was similar to that of mobocertinib, it had fivefold greater potency against WT EGFR (IC₅₀ 3.9 nmol/L) and maximally inhibited LU0387 cell survival by only ≈60%, compared with 85% for mobocertinib.

Inhibitory activity against the most common EGFR activating mutants, with or without a T790M resistance mutation, was assessed by determining concentrations of mobocertinib required to inhibit pEGFR in three NSCLC cell lines expressing EGFR mutants: HCC827 [exon 19 deletion (D)], H4011 [L858R (L)], and H1975 [L858R plus T790M (LT)]. Mobocertinib potently inhibited EGFR with common activating mutations (IC₅₀s 1.3–4.0 nmol/L) or with a T790M mutation (IC₅₀ 9.8 nmol/L, LT) more potently than WT EGFR (Fig. 3C). Potent inhibition of EGFR and downstream signaling by mobocertinib were confirmed in all three cell lines by Western blot analysis (Fig. 3D–F).

Examination of the in vitro and in vivo metabolism of mobocertinib in preclinical studies resulted in the identification of two N-demethylated metabolites (AP32914 and AP32960; Fig. 2D), likely the result of CYP450-mediated oxidative dealkylation. The cellular activities of AP32914 and AP32960 were similar (IC₅₀s typically within twofold) to those of mobocertinib for both WT and mutant EGFR, suggesting the potential for these metabolites to contribute to the pharmacologic activity of mobocertinib (Supplementary Table S2).

We used a mutant C797S version of EGFRex20ins NPG to evaluate the reversible binding affinity of mobocertinib in a low-temperature Biacore surface plasmon resistance (SPR) assay. Mobocertinib demonstrated substantially better binding affinity to the NPG EGFRex20ins mutant by SPR and better selectivity over WT EGFR compared with osimertinib and afatinib (Fig. 3G). The better selectivity observed upon initial protein-ligand formation is an important factor contributing to the selectivity of mobocertinib toward EGFRex20ins mutants.

The kinase selectivity profile of mobocertinib was evaluated by in vitro kinase assays of 490 recombinant human protein kinases (372 unique kinases and 118 mutant variants; Fig. 3H). At 1 μmol/L, mobocertinib inhibited 28 of 490 kinases (6%) by >50%, including all 14 members of the EGFR family tested (EGFR, HER2, HER4, and 11 EGFR variants with activating or resistance mutations; Supplementary Table S4). Dose–response curves were established in a subset of kinases, including the 28 kinases mobocertinib inhibited most potently in the single-point screen and 10 other kinases reported to be involved in critical cellular processes. Of these 38 kinases, 15 were inhibited with IC₅₀s <2 nmol/L, including all 14 members of the EGFR family tested and BLK (Supplementary Table S5). Six additional kinases (JAK3, TXK, BMX, ACK1, BTK, and BTK^E41K) were inhibited with IC₅₀s <20 nmol/L. Similar results were obtained with the mobocertinib metabolites AP32914 and AP32960 (Supplementary Table S5). Overall, mobocertinib displayed potent kinase inhibitory activity mainly restricted to the EGFR family of proteins, with limited off-target activity against the rest of the kinome.

We evaluated mobocertinib in two mouse tumor models containing an EGFRex20ins ASV mutation. In a tumor model using engineered Ba/F3 cells, once-daily oral dosing of mobocertinib showed dose-dependent inhibition of tumor growth, or tumor regression, at well-tolerated doses (Supplementary Fig. S1A and S1B). In a xenograft model using CTG-2842 cells from a patient with EGFRex20ins ASV-mutant NSCLC who had been treated with erlotinib (without response), mobocertinib dosed orally at 15 mg/kg daily induced substantial tumor regression (92%; Fig. 4A) and was well tolerated (Supplementary Fig. S2).

We next investigated antitumor activity of oral mobocertinib in mice engrafted with LU0387 tumors expressing an EGFRex20ins NPH mutant. At 10 mg/kg daily, mobocertinib induced 56% tumor growth inhibition, and at 30 mg/kg daily induced 87% tumor regression relative to pretreatment tumor size (P < 0.0001; Fig. 4B). In contrast, erlotinib induced 38% tumor growth inhibition at 50 mg/kg. Significant inhibition of tumor growth was also achieved with osimertinib 30 mg/kg daily, although to a lesser degree (mean 13% tumor regression; 3/10 mice with >50% regression) than that achieved with an equivalent dose of mobocertinib (mean 87% tumor regression; 10/10 mice with >50% regression; Fig. 4B). The 10- and 30-mg/kg daily doses of mobocertinib were well tolerated (Supplementary Fig. S3A) with no clinical signs or drug-related mortality. Increased efficacy was associated with increased plasma levels of mobocertinib and its active metabolites (Supplementary Fig. S3B). Pharmacodynamic activity of mobocertinib was assessed by measuring the levels of pEGFR (Tyr 1068) in LU0387 tumors. Representative Western blot experiments are shown in Fig. 4C. There was no substantial impact on pEGFR levels in mice treated with 10 mg/kg mobocertinib or 50 mg/kg erlotinib, doses associated with relatively modest tumor growth inhibition. At 30 mg/kg, the dose associated with 87% tumor regression, mobocertinib strongly inhibited tumor pEGFR levels. These data demonstrating dose-dependent pEGFR inhibition in tumors indicate that mobocertinib suppressed growth of EGFRex20ins mutant tumors through EGFR inhibition. Combination treatment with mobocertinib and cetuximab improved efficacy in a head and neck squamous cell carcinoma model (CTG-2130) with EGFRex20ins D770_N771insGL (Fig. 4D).

Mobocertinib showed robust antitumor efficacy and was well tolerated in tumor models driven by other EGFR mutations, including the PC9 xenograft model harboring the common exon 19 deletion (D; Supplementary Fig. S4A and S4B) and the H1975 xenograft model harboring the LT mutation (Fig. 4E; Supplementary Fig. S5A and S5B). Mobocertinib demonstrated minimal efficacy in a brain metastasis model in mice intracranially implanted with H1975-luc tumors (Supplementary Fig. S6).

To further evaluate mobocertinib efficacy against EGFRex20ins expressing tumors in vivo, we generated a genetically engineered mouse model (GEMM) with inducible expression of human EGFR D770_N771insSVD (SVD). We electroporated mouse embryonic stem (ES) cells with a DNA construct containing LoxP–Stop–LoxP–hEGFR SVD and injected these positive ES cells into blastocysts to produce chimeras. The chimera mice were further crossed with WT mice to generate germlines. We intranasally delivered adeno-Cre virus into adult EGFRex20ins SVD mice to initiate EGFRex20ins SVD expression specifically in murine lung epithelial cells (Fig. 5A). MRI monitoring showed that lung tumors appeared 14 weeks after viral delivery (Fig. 5B). Immunohistochemistry staining demonstrated that SVD-expressing tumors were positive for adenocarcinoma markers thyroid transcription factor 1 (TTF1) and surfactant protein C (SPC), as well as negative for the squamous cell carcinoma marker p63 (Fig. 5C), indicating that this model mimicked the clinical setting, as most EGFR-mutant lung tumors are adenocarcinomas. Long-term monitoring showed the growth kinetics of these tumors (Fig. 5B) and indicated that mice had a median survival time of 21.7 weeks after adeno-Cre induction (Fig. 5D).

To investigate the short-term efficacy of mobocertinib, we performed a pharmacodynamic study in tumor-bearing EGFRex20ins SVD mice. After 1 week of treatment, mobocertinib-treated SVD mice had significantly reduced tumor volumes compared with those treated with vehicle control (Fig. 5E). We harvested lung tumors and stained tissues for pEGFR and pERK1/2. Immunohistochemistry results demonstrated that mobocertinib effectively abolished pEGFR and inhibited its major downstream signaling target pERK (Fig. 5F), supporting the on-target efficacy of mobocertinib against EGFRex20ins SVD in vivo.

Next, we treated EGFRex20ins SVD mice with continuous mobocertinib or vehicle and evaluated long-term antitumor efficacy by MRI every 2 weeks. Vehicle-treated mice developed progressive disease at 2 and 4 weeks. In contrast, all mobocertinib-treated mice (n = 8) showed significant tumor growth inhibition, with up to a 90% decrease from baseline in tumor volume (Fig. 5G and H), with responses sustained for 10 weeks (Fig. 5I). These studies demonstrated a sustained benefit of mobocertinib for EGFRex20ins SVD tumors, which was consistent with the early observed efficacy and pharmacodynamic results. Taken together, these data indicated that mobocertinib has on-target, efficacious, and durable antitumor effects in EGFRex20ins SVD lung cancer.

EGFRex20ins-mutant NSCLC is a life-threatening disease involving rare mutations for which no currently approved oral EGFR-targeted therapy is available. Mobocertinib is an irreversible TKI specifically designed to address this need. We demonstrated that mobocertinib has better potency and selectivity over WT EGFR for all EGFRex20ins mutants tested (IC₅₀s 4.3–22.5 nmol/L) than erlotinib, gefitinib, afatinib, and osimertinib. Mobocertinib also demonstrated potent activity against EGFR with common activating mutations with or without the T790M resistance mutation (D, L, DT, and LT: IC₅₀s 2.7–21.3 nmol/L) and uncommon activating mutations (G719A, G719S, S768I, L861Q, and L861R; IC₅₀s 3.5–20.2 nmol/L). Mobocertinib inhibited all three classes of mutants with potency greater than WT EGFR, although it is unclear whether the degree of selectivity is greater than that observed for EGFR TKIs approved for patients with these mutations. In addition, mobocertinib showed robust efficacy in tumor models harboring common exon 19 deletion (D) and T790M mutations. These results are consistent with those of previous studies reporting that mobocertinib inhibits T790M and L858R mutants, as well as EGFRex20ins ASV, SVD, FQEA, and H773_V774insH mutants in Ba/F3 cells, more potently than WT EGFR (40). Mobocertinib also demonstrated dose-dependent inhibition of cell viability in patient-derived cancer cell lines bearing FQEA or N771_H772insH mutations in previous studies (40). The consistent selectivity for a broad range of mutant EGFR variants over WT EGFR suggests that mobocertinib may have lower potential for dose-limiting toxicities associated with EGFR inhibition in normal tissues (e.g., skin and gastrointestinal adverse events) than drugs with less selectivity for these mutants, although gastrointestinal and skin adverse events were reported in the phase I/II trial (41).

Mobocertinib was well tolerated and demonstrated sustained dose-dependent inhibition of tumor growth in both patient-derived xenografts and murine orthotopic models of NPH, ASV, and SVD EGFRex20ins-mutant NSCLC. Both mobocertinib and osimertinib induced tumor regression in the LU0387 NPH tumor model, but the extent of regression was greater with mobocertinib 30 mg/kg daily (87%) compared with osimertinib 30 mg/kg daily (13%). Combination treatment with mobocertinib plus cetuximab improved efficacy in a head and neck cancer model with an EGFRex20ins mutation (D770_N771insGL) that was less sensitive to mobocertinib, consistent with results for afatinib or osimertinib in combination with cetuximab in D770_N771insNPGEGFRex20ins mutant cell lines (42).

Similar to our data, prior studies have demonstrated that osimertinib potently inhibits some EGFRex20ins mutants in Ba/F3 cells (25). However, osimertinib demonstrated limited clinical activity (objective response rate, 5%) in patients with EGFRex20ins NSCLC at the approved dose (80 mg/d; ref. 43); a higher dose (i.e., 160 mg/d) may be needed for substantial efficacy in patients with EGFRex20ins mutations (26–28).

Other novel EGFR TKIs (e.g., poziotinib, TAS6417/CLN-081, DZD9008, tarloxotinib, and DS-2087b) have demonstrated preclinical activity and are under investigation in EGFRex20ins-mutated NSCLC (39, 44–49). Poziotinib demonstrated limited efficacy in patients with EGFRex20ins mutations, and doses of 16 mg/d or less were associated with high rates of WT EGFR-mediated toxicity and dose reduction (48, 49). Tarloxotinib did not demonstrate any objective responses among 11 patients with EGFRex20ins mutations in a phase II trial (46). Preclinical and clinical data suggest that amivantamab, a bispecific antibody that targets EGFR and MET, may have efficacy for patients with NSCLC with EGFRex20ins mutations (50–52).

In conclusion, our preclinical data demonstrated that mobocertinib is a pan-mutation-selective irreversible EGFR TKI with a broad spectrum of in vitro and in vivo activity against clinically relevant EGFR mutations, including several of the most common EGFRex20ins mutations. Based on its favorable preclinical pharmacokinetic (PK) and toxicity profile, mobocertinib was selected for clinical development as a treatment for patients with EGFRex20ins-mutated NSCLC. Mobocertinib demonstrated antitumor activity in patients with NSCLC with diverse EGFRex20ins variants with a safety profile consistent with other EGFR inhibitors in the first two parts of a three-part phase I/II study (ClinicalTrials.gov identifier: NCT02716116). The phase I/II results are reported in a companion article in this issue by Riely and colleagues (53).

Data from GENIE v8.0 (38) were extracted for patients with EGFR mutation in all cancer types. Only tumors with in-frame insertions and insertion–deletions in EGFR were considered; excluding pathologies other than lung cancer resulted in identification of 226 data points. After removal of duplicate data, 157 patients with EGFR insertions or insertion–deletions were identified; 155 patients had an insertion/insertion–deletion in the exon 20 coding region (amino acids 763–775).

All cell lines were grown at 37°C, with 5% CO₂, in medium supplemented with 10% FBS. Sources of cell lines, growth medium, and additional supplements used are as follows: A431 cells (ATCC; RPMI 1640 growth medium), HEK293 cells (Thermo Fisher Scientific; DMEM growth medium), and parental Ba/F3 cells (German Collection of Microorganisms and Cell Cultures GmbH; RPMI 1640 growth medium supplemented with 10 ng/mL IL3). Per standard protocol, all Ba/F3 EGFR mutants and A431 cells were tested and negative for Mycoplasma. Mobocertinib, AP32914, AP32960, erlotinib, gefitinib, afatinib, and osimertinib were synthesized at ARIAD Pharmaceuticals, Inc., or purchased from commercial sources.

Cloning of human EGFR-mutant coding sequence into the pLVX.IRES.Puro lentiviral vector (Takara Bio USA) was performed by GenScript. All cloned genes were confirmed by sequencing. Viral particles were produced by transfecting pLVX.IRES.Puro vectors into HEK293 cells using the Trans-Lentiviral ORF Packaging Kit (GE Healthcare). Forty-eight hours posttransfection, virus-containing supernatants were harvested and incubated for another 48 to 72 hours with parental Ba/F3 cells in the presence of IL3. Transduced Ba/F3 cells were then selected with puromycin (0.5 μg/mL). Upon reaching confluence, cells were grown in IL3-deprived medium to render them solely dependent on transduced EGFR-mutant gene activity for survival.

Cells were plated and incubated with compound, in duplicate, using a Tecan HP D300 Instrument (Tecan), for 3 days at 37°C. Cell viability was measured using Cell Titer-Glo assay (Promega). Dose–response curves were generated and used to calculate IC₅₀ values.

The potency with which compounds inhibited WT EGFR activity was assessed by measuring levels of EGFR phosphorylated at Tyr1068 (pEGFR) in EGFR-amplified A431 (epidermoid carcinoma) cells. A431 cells were incubated with compound for 2 hours and stimulated for 20 minutes with EGF (25 ng/mL), and lysates were prepared. Total EGFR and pEGFR levels in lysates were simultaneously measured by immunoassays using the Meso Scale Discovery Multi-Array platform [Meso Scale Discovery (MSD)] per the manufacturer's protocol. IC₅₀s for inhibition of WT EGFR activity were calculated by examining levels of pEGFR normalized to total EGFR levels in dose–response curves.

Avi-tagged constructs EGFR WT-Avi and EGFRex20ins NPG-Avi were expressed in baculovirus, enzymatically biotinylated, and incubated with ≈80 μmol/L mobocertinib under conditions maintaining native fold. After a 30-minute incubation, the mixture was subjected to LC/MS analysis under denaturing conditions. Liquid chromatography was performed using a POROS R2/20 2.1-mm × 30-mm column (Applied Biosystems, Thermo Fisher Scientific; mobile phase A: 0.1% formic acid in water, mobile phase B: 0.1% formic acid in acetonitrile; flow rate, 0.5 mL/min). LC/MS experiments were conducted using a Xevo G2-S QTof mass spectrometer equipped with an Acquity UPLC (Waters Corporation) under positive ion mode. Full-scan (200–2,000 m/z) electrospray ionization and MS data were collected at a resolution of 32,500 at full width at half maximum (FWHM) using MassLynx software version 4.1 (Waters Corporation).

HCC4011 cells (UT Southwestern) and HCC827 and H1975 cells (ATCC) were maintained in RPMI 1640 medium. LU0387 primary cells (Crown Bioscience) were maintained in DMEM Nutrient Mixture F-12 growth medium. CUTO14 cells (University of Colorado, Aurora) were maintained in RPMI 1640 growth medium. All cell lines were grown at 37°C, with 5% CO₂, in medium supplemented with 10% FBS. CUTO14 cells tested negative for Mycoplasma and were authenticated by short-tandem repeat profiling at the source laboratory prior to freezing and shipment to Takeda. For in vitro and in vivo experiments, cells were used between passages 3 and 20. Per standard protocol, all HCC827 and HCC4011 cells were tested and negative for Mycoplasma.

Viability assays using LU0387 cells were performed at Crown BioScience. Cells were plated into 96-well plates and dosed with a nine-point, threefold dilution series of compounds (10 μmol/L to 1.5 nmol/L) and incubated for 7 days. Viability was measured using the CellTiter-Glo assay (Promega). Dose–response curves were generated and used to calculate IC₅₀ values.

The potency with which compounds inhibited pEGFR signaling was assessed in CUTO14, HCC827, H4011, and H1975 cells by incubating them with compound (CUTO14: mobocertinib or osimertinib at concentrations of 0.1–1,000 nmol/L; HCC827, HCC4011, and H1975 cells: erlotinib 10–1,000 or 100–10,000 nmol/L, or mobocertinib 0.3–1,000 nmol/L) for 6 hours and evaluating levels of pEGFR in cellular lysates using Western blotting for pEGFR, EGFR, pERK1/2, ERK1/2, pAKT, and AKT with antibodies from Cell Signaling Technology [pAKT (Ser473) (D9E) XP rabbit monoclonal antibody (mAb) #4060; AKT antibody rabbit #9272; EGFR (D38B1) XP rabbit mAb #4267; pEGFR (Tyr1068) (D7A5) XP rabbit mAb #3777; p-p44/42 MAPK (ERK1/2) (Thr202/Tyr204) (D13.14.4E) XP rabbit mAb #4370; p44/42 MAPK (ERK1/2) (137F5) rabbit mAb #4695; and Vinculin (E1E9V) XP rabbit mAb #13901].

NSCLC cells were treated with compound for 2 hours, then lysed in a Multi-Array platform (MSD) lysis buffer. Total EGFR and pEGFR (Tyr1068) levels were measured using the MSD Multi-Array immunoassay system. EGFR phosphorylation ratios (pEGFR/total EGFR) were calculated and used to determine dose–response curves and IC₅₀s using GraphPad Prism (GraphPad Software).

Measurement of WT and EGFRex20ins NPG EGFR binding affinities was performed using SPR [Biacore T200 using a CM5 chip (Cytiva); chip compartment temperature 10°C; sample compartment temperature, 20°C) in the presence of different concentrations of mobocertinib, afatinib, and osimertinib to assess dissociation constants (K_d). The EGFR proteins used were truncated versions of EGFR containing a C797S mutation and had an AviTag (Avidity LLC) appended to the C-terminal end. An amine coupling was performed by injecting a mixture of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) and N-hydroxysuccinimide (NHS) followed by injection of NeutrAvidin (Thermo Fisher Scientific) and capping the chip with ethanolamine. The two EGFR proteins were immobilized by injecting a solution of the protein typically diluted to 1 mg/mL of the running buffer (25 mmol/L HEPES, pH 7.4; 250 mmol/L NaCl; 10% glycerol; 2 mmol/L Tris (2-carboxy-ethyl)-phosphin-HCl; 0.1% CHAPS; 1% trehalose; 2% DMSO), and then the chip was capped with 1 mmol/L solution of biotin in running buffer. K_d values were determined using the T200 Software Biacore package (Cytiva).

In vitro profiling of the 490-member kinase panel was performed at Reaction Biology Corporation using the Kinase HotSpot platform. First, a broad panel of 490 protein kinases, including 372 unique human kinases and an additional 118 mutant variants, was assayed using a single mobocertinib concentration of 1 μmol/L. A second screen was performed on a panel of 38 kinases, including 28 kinases inhibited most strongly in the first assay and 10 additional kinases of interest using a broad range of mobocertinib, AP32914, and AP32960 concentrations to allow determination of IC₅₀ values. Percentage inhibition data for each kinase was mapped to the kinome tree (54) using Kinome Render (55).

CTG-2842 cells (Champions Oncology) are derived from a lung adenocarcinoma, established from the primary site of a patient previously treated with carboplatin/pemetrexed with response and subsequent progression at 6 months, with erlotinib with no response, and with pembrolizumab/coxsackievirus with no response. CTG-2130 cells (Champions Oncology) are derived from a head and neck squamous cell carcinoma, established from a skin metastasis of a patient who was previously treated with and had no response to cisplatin and cetuximab; subsequently, the patient was treated with and had no response to an anti–PD-L1 inhibitor/anti-LAG3 inhibitor and paclitaxel. Tumor fragments generated from CTG-2842 cells or CTG-2130 cells were inoculated subcutaneously into the left flank of female athymic Nude-Foxn1nu mice (Envigo). When average tumor volume reached 150 to 300 mm³, animals were treated with vehicle or mobocertinib 15 mg/kg orally daily ± cetuximab 5 mg/kg i.p. once every 3 days for up to 60 days. Tumor size was measured twice weekly.

Ba/F3 cells expressing EGFRex20ins ASV were used to generate subcutaneous tumors in severe combined immunodeficiency disease (SCID) mice (8 weeks old; Charles River Laboratories). When the average tumor volume reached about 150 mm³, animals were treated with vehicle or mobocertinib 30 or 50 mg/kg by oral gavage once daily for 7 days. Tumor size was measured at least twice a week. Mice were euthanized on day 7 at 2, 6, or 24 hours after the last dose, when blood samples were collected for PK analysis.

Female Nu/Nu mice (Beijing Vital River Laboratory Animal Technology Co., Ltd) were inoculated subcutaneously in the right flank with one LU0387 tumor fragment (2–3 mm in diameter; Crown Bioscience). When average tumor volume reached 150 mm³, animals were treated with vehicle, mobocertinib 10 or 30 mg/kg, erlotinib 50 mg/kg, or osimertinib 30 mg/kg once daily for 21 days. Tumor size was measured every 3 days. Mice were euthanized on day 21, 2 or 6 hours after the last dose, and blood samples were collected for PK analysis. LU0387 tumor samples were homogenized in RIPA buffer and analyzed by Western blotting with antibodies against total EGFR and pEGFR that recognize EGFR tyrosine 1068 (Cell Signaling Technology) and β-actin (Sigma).

Female nude BALB/c mice (Beijing Vital River Laboratory Animal Technology Co., Ltd) were injected subcutaneously with a suspension of 2.0 × 10⁶ PC9 human lung adenocarcinoma cells. When the mean tumor volume reached approximately 240 mm³, mice were orally dosed with vehicle or mobocertinib 10 or 30 mg/kg once daily for 21 days. Tumor growth and body weight were measured twice per week.

H1975 cells (10⁷ cells) were injected subcutaneously into the right flank of SCID mice (8 weeks old; Charles River Laboratories). When the average tumor volume reached about 210 mm³, animals were treated with vehicle or mobocertinib 3, 10, or 30 mg/kg once daily for 14 days. Tumor size was measured at least twice a week. Mice were euthanized on day 14 at 2, 6, or 24 hours after the last dose, when blood samples were collected for PK analysis.

Luciferase-expressing H1975 cells were generated from the parental H1975 line. Female Athymic Nude-Foxn1^nu mice were implanted intracranially with 10⁵ cells in 10 μL on day 0. Treatment began on day 6 at an overall mean tumor burden of 1.5 × 10⁶ p/s. In vivo bioluminescence images were acquired after injection of 150 mg/kg D-luciferin on days 6, 13, 20, 27, and 34.

In all mouse studies of tumor models, the vehicle solution for oral drug delivery was 10% N-methyl-2-pyrrolidone/90% polyethylene glycol. Stock solutions of mobocertinib were prepared in vehicle, and a constant dose volume was used to deliver mobocertinib doses ranging from 15 to 50 mg/kg.

In each of the studies, plasma levels of mobocertinib, AP32914, AP32960, and other analytes (afatinib or erlotinib) were measured using nonvalidated LC/MS methods (Supplementary Table S6).

All animal studies were approved by a local Institutional Animal Care and Use Committee (IACUC) and conducted in accordance with the regulations of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) and local IACUC guidelines.

The coding region of human EGFRex20ins SVD with the Kozak sequence (GCCGCCACC) was synthesized at GENEWIZ, Inc. and introduced into pGV at the EcoRI cloning site. Sequence-confirmed human EGFRex20ins SVD plasmid was coelectroporated with plasmid-expressing flippase (FLP) recombinase into mouse ES cells. Electroporated ES cells were selected with hygromycin, and positive clones were identified by PCR. Positive ES clones were injected into mouse blastocysts for chimera generation. Chimera mice were crossed with WT mice to generate mice with germline mutations. Mouse genomic DNA was used as the PCR template, and the human EGFRex20ins SVD sequence was confirmed with Sanger sequencing. The genotyping primers used were DSVD-forward: TCCGTGGACAATCCCCACGT and DSVD-reverse: CAGCGTCAATCATCCAACAC. Detailed methods for this strategy were previously described (56). All animal experiments, including breeding and treatment studies, were approved by the NYU Langone Medical Center IACUC.

EGFRex20ins SVD mice were monitored by MRI for tumor development after intranasal induction with adeno-Cre (5 × 10⁷ plaque-forming units). Tumor-bearing mice were dosed with mobocertinib (30 mg/kg, orally, daily) and monitored by MRI every 2 weeks. Mice were euthanized and lung tissues were collected and fixed with 10% formalin.

Immunohistochemistry staining was performed at iHisto, Inc. using the following antibodies: anti-TTF1 5883–1 (Epitomics, Inc.), anti-p63 ab53039 (Abcam), anti-pEGFR (Tyr1068; #3777; Cell Signaling Technology), and anti-pERK1/2 (Thr202/Tyr204; #4370; Cell Signaling Technology).

The data sets supporting the results reported in this article will be made available within 3 months from initial request to researchers who provide a methodologically sound proposal. The data will be provided in compliance with applicable laws, data protection, and requirements for consent and anonymization.

F. Gonzalvez is an employee and shareholder of Aligos Therapeutics and a former employee of ARIAD. S. Vincent is an employee and shareholder of Takeda. T.E. Baker is an employee and shareholder of MOMA Therapeutics and a former employee of ARIAD. A.E. Gould is an employee of Takeda. S.D. Wardwell is an employee and shareholder of Blueprint Medicines and a former employee of ARIAD. S. Nadworny is a former employee of ARIAD. Y. Ning is a former employee of ARIAD. S. Zhang is a former employee of ARIAD and an employee and stockholder of EMD Serono, a subsidiary of Merck KGaA, Darmstadt, Germany. W-S. Huang is an employee and shareholder of Theseus Pharmaceuticals and a former employee of ARIAD. Y. Hu is an employee of Takeda. F. Li is a former employee of ARIAD. M.T. Greenfield is a former employee of ARIAD. S.G. Zech is an employee and shareholder of Amgen Inc. and a former employee of ARIAD. B. Das is a former employee of ARIAD. N.I. Narasimhan is a former employee of ARIAD. T. Clackson is an employee and shareholder of Xilio Therapeutics and is a former employee of ARIAD. D. Dalgarno is an employee and shareholder of Theseus Pharmaceuticals and a former employee of ARIAD. W.C. Shakespeare is an employee and shareholder of Theseus Pharmaceuticals and a former employee of ARIAD. M. Fitzgerald is a former employee of Takeda. J. Chouitar is an employee of Takeda. R.J. Griffin is an employee of Takeda and a shareholder of GSK. S.K. Wong reports sponsored research agreements with Takeda, BMS, Mirati, Alkermes, Merus, Amgen, Ansun Biopharma, Enliven Therapeutics, Tvardi Therapeutics, Delfi Diagnostics, and Dracen Pharmaceuticals; a consulting agreement with Merck; and consulting and sponsored research agreements with AstraZeneca, Janssen, Pfizer/Array, Novartis, and Zentalis. X. Zhu is an employee and shareholder of Regor Pharmaceuticals, Inc. and a former employee of ARIAD. V.M. Rivera is an employee and shareholder of Theseus Pharmaceuticals and a former employee of Ariad. No disclosures were reported by the other authors.

F. Gonzalvez: Conceptualization, data curation, formal analysis, investigation, methodology, writing-original draft, writing-review and editing. S. Vincent: Conceptualization, data curation, formal analysis, investigation, methodology, writing-original draft, writing-review and editing. T.E. Baker: Data curation, formal analysis, investigation, methodology, writing-review and editing. A.E. Gould: Data curation, investigation, writing-original draft, writing-review and editing. S. Li: Investigation, methodology, writing-review and editing.S.D. Wardwell: Investigation, writing-review and editing. S. Nadworny: Investigation, writing-review and editing. Y. Ning: Investigation, writing-review and editing. S. Zhang: Data curation, investigation, writing-review and editing. W.-S. Huang: Conceptualization, formal analysis, investigation, writing-review and editing. Y. Hu: Investigation, writing-review and editing. F. Li: Investigation, writing-review and editing. M.T. Greenfield: Investigation, writing-review and editing. S.G. Zech: Investigation, writing-review and editing. B. Das: Investigation, writing-review and editing. N.I. Narasimhan: Investigation, writing-review and editing. T. Clackson: Conceptualization, supervision, investigation, writing-review and editing. D. Dalgarno: Writing-review and editing. W.C. Shakespeare: Conceptualization, supervision, writing-review and editing. M. Fitzgerald: Investigation, writing-review and editing. J. Chouitar: Resources, investigation, writing-review and editing.R.J. Griffin: Writing-review and editing. S. Liu: Resources, investigation, methodology, writing-review and editing. K. Wong: Writing-review and editing. X. Zhu: Conceptualization, investigation, writing-review and editing. V.M. Rivera: Conceptualization, data curation, formal analysis, investigation, methodology, writing-original draft, writing-review and editing. NOTE: The authors define “Investigation” as performing clinical experiments and collecting data.

We thank the American Association for Cancer Research (AACR) for its financial and material support in the development of the AACR Project GENIE registry, as well as members of the consortium for their commitment to data sharing. Interpretations are the responsibility of study authors. We also thank Dr. Robert Doebele and Dr. Anh Le, University of Colorado, for the use of CUTO14 cells, and Anna Kohlmann, Paul M. Taslimi, Tianjun Zhou, Lingling Zhang, Hyun G. Jang, Emily Y. Ye, Meera Tugnait, Hua Zou, Weston Lane, Robert Skene, R. Scott Rowland, and Andy Garner for their contributions to this study. Professional medical writing assistance was provided by Lauren Gallagher, RPh, PhD, and Lela Creutz, PhD, of Peloton Advantage, LLC, an OPEN Health company, Parsippany, New Jersey, and funded by Millennium Pharmaceuticals, Inc. Teodor G. Paunescu, PhD (Millennium Pharmaceuticals, Inc., Cambridge, Massachusetts, a wholly owned subsidiary of Takeda Pharmaceutical Company Limited) is acknowledged for editorial assistance. All studies were funded by Millennium Pharmaceuticals, Inc., a wholly owned subsidiary of Takeda Pharmaceutical Company Limited.