Targeting EGFR Exon 20 Insertions in Non–Small Cell Lung Cancer: Recent Advances and Clinical Updates

The discovery of activating mutations in EGFR in non–small cell lung cancer (NSCLC) dramatically altered the clinical landscape of this disease. Early trials of first-generation EGFR tyrosine kinase inhibitors (TKI) demonstrated impressive efficacy, far surpassing historic standards in NSCLC (1, 2). Subsequent development of next-generation EGFR-targeted therapies further extended median overall survival (OS) for patients with EGFR-mutant tumors to more than three years (3).

The majority of EGFR driver mutations occur as in-frame deletions in exon 19 (del19) and point mutations in exon 21 (L858R). These two classes of canonical activating EGFR mutations confer sensitivity to first-generation EGFR TKIs such as erlotinib and gefitinib and second- and third-generation TKIs such as afatinib, dacomitinib, and osimertinib (1–5). The availability of EGFR-directed therapies ushered in the practice of routine genotyping as a critical aspect of standard care for advanced NSCLC. Although the first assays were mutation-specific targeted PCR-based tests designed to identify only common canonical alterations, increasingly sophisticated sequencing platforms have now been developed for both clinical and investigative purposes. Expanded genotyping, together with widespread research efforts, led to increased appreciation of a broader array of activating EGFR mutations, including in-frame insertion mutations in EGFR exon 20 (EGFR ins20) and activating point mutations such as G719X, S768I, and L861Q, among others (ref. 6; Fig. 1).

EGFR ins20 are the third most common subtype of EGFR mutation, found in approximately 10% of EGFR-mutant NSCLCs, and are mutually exclusive with other known driver mutations (refs. 6–9; Fig. 1). Though some heterogeneity exists, as a whole, ins20 mutations are not sensitive to early-generation EGFR TKIs (8). However, recent years have demonstrated significant progress in developing effective targeted therapies for these patients. Herein, we describe both the historical challenges and recent advances in the field of EGFR ins20–positive NSCLC.

When considered separately from canonical EGFR mutations, EGFR ins20 comprise approximately 1% to 2% of all NSCLC cases, a similar frequency as RET and ROS1 rearrangements (Fig. 1; refs. 6, 10). Although the demographic features of patients harboring tumors with EGFR ins20 NSCLC can vary, like those with classic EGFR mutations, they tend to be never-smokers and are more commonly female, and of East Asian descent (8–10). Nevertheless, given the historical lack of effective targeted therapies, clinical outcomes in EGFR ins20 patients have been similar to EGFR–wild-type NSCLC (10).

Driver mutations in the EGFR family member ErbB2 (HER2) are sometimes confused with EGFR ins20 because they too are typically insertion mutations in exon 20 of HER2 (11). As described below, although the biology of these mutations is similar and some therapies have activity against both types of exon 20 insertion mutations, the molecular features and spectrum of mutations found in HER2 ins20-positive NSCLC follows a different pattern than EGFR ins20, with less heterogeneity.

Exon 20 of EGFR encompasses amino acids (AA) 762–823 and contains two important regions: the regulatory C-helix domain (AA762–766) and the adjacent loop that follows it (AA767–774). Exon 20 insertions in EGFR-mutant NSCLC include in-frame insertions and/or duplications of 3–21 base pairs typically occurring between AA761 and AA775 (refs. 6–8; Fig. 2) and serve to activate the receptor in a ligand-independent manner by pushing the C-helix into its inward position, inducing active receptor conformation. Among the more than 60 unique activating EGFR ins20 identified to date, the majority are composed of 1–4 AA insertions located in the loop following the C-helix (refs. 6, 7, 12; Fig. 2). The significant heterogeneity of insertions identified is in striking contrast to both EGFR del19 mutations, which demonstrate less variability with a small range of in-frame indels identified, and HER2 ins20 mutations, which most commonly occur as A775_G776insYVMA (11).

When studied in vitro, exon 20 insertions confer intermediate catalytic activity, greater than wild-type EGFR but not as high as EGFR L858R mutants (7). Early studies of crystal structures of representative insertions showed that the most common EGFR ins20, unlike EGFR del19 and L858R, do not directly affect the structure of the ATP-binding pocket of EGFR (7). Thus, although EGFR del19 and L858R mutations increase the relative affinity for EGFR TKIs over ATP compared with wild-type EGFR—a molecular feature that allows for a large therapeutic window for TKI inhibition of the mutant receptor—this effect is not seen with EGFR ins20 (7). More recently, 3-D modeling with the solved crystal structures of EGFR ins20 D770insNPG compared with EGFR T790M and wild-type EGFR suggested that this representative EGFR ins20 demonstrated similar structure to the EGFR T790M model in terms of positioning of the gatekeeper residue, confirming the reason for resistance to noncovalent, first-generation TKIs (13). These analyses also suggested that the shift of the C-helix as well as the phosphate-binding loop (P-loop) toward the drug-binding pocket results in some degree of steric hindrance that inhibits TKI binding.

These structural differences explain the central challenge in developing effective small-molecule targeted inhibitors of EGFR ins20 and further underscore the lack of clinical activity of first- and second-generation EGFR TKIs against tumors harboring EGFR ins20 compared with canonical EGFR activating mutations (7, 8, 10, 14). One prominent exception to this finding is the EGFR ins20 A763_Y764insFQEA, which has been shown to have 10-fold higher binding affinity for first-generation EGFR TKIs compared with other EGFR ins20, correlating to a higher ratio of TKI to ATP affinity and associated sensitivity to TKI-induced receptor inhibition in vitro and in vivo (7). Isolated case reports have also described sensitivity of H773dup, H773_V774insNPH, and N771delinsKG mutations to afatinib (15–17), although the structural aspects of these unique variants that could explain this sensitivity have not been well described.

Although significant progress has been made recently, the wide spectrum of specific insertion mutations found in EGFR exon 20 has limited the systematic characterization of each unique alteration's structure and sensitivity to EGFR inhibitors. Although some insertions (in particular, V769-D770 insX, D770-N771 insX, and H773-V774 insX) are more common than others, the number of individual insertions/duplications observed clinically means that even large data sets have few cases of the less common mutations, making it difficult to identify universally shared targetable structural features (10).

From a clinical perspective, an equally significant challenge has been achieving reliable identification of this spectrum of mutations in patients. The wide variability in clinical genotyping assays used around the United States and internationally leads to missed diagnoses in a subset of patients. For example, older, mutation-specific PCR-based assays that were historically used for EGFR genotyping often did not capture alterations across exon 20 (10). However, the emergence of next-generation sequencing (NGS) platforms now in common use for clinical genotyping has enabled the more effective and comprehensive identification of EGFR ins20 mutations in patients with NSCLC. Liquid biopsies assaying circulating tumor DNA (ctDNA) can also detect EGFR ins20 alterations, although the sensitivity of these assays is limited by degree of tumor DNA shed into the bloodstream. Overall, these diagnostic improvements have facilitated greater clinical trial enrollment and created momentum in the development of new therapies.

Despite initial challenges in targeting EGFR exon 20 insertions, over the past 5 to 10 years a number of emergent therapies and clinical trials have been specifically developed for this unique molecular subgroup (Fig. 3). With a variety of mechanisms of action and varying levels of clinical activity, these novel therapies, summarized below and in Table 1, represent important steps forward in improving treatment options for patients with EGFR ins20 mutations.

Although early-generation EGFR TKIs were ineffective against most EGFR ins20 mutants, several TKIs with broader ErbB family inhibitory activity have shown increased promise against this subtype of EGFR-mutant NSCLC. One such TKI is poziotinib (formerly known as HM781-36B), which was shown by Cha and colleagues to covalently bind both EGFR and HER2, inhibiting growth of NSCLC cell lines harboring canonical sensitizing EGFR mutations and the first-generation EGFR TKI-resistance mutation T790M, as well as HER2-overexpressing NSCLC cell lines (18). Structurally, poziotinib is similar to the second-generation EGFR TKI afatinib, small and flexible with a quinazoline core (13). Unique from afatinib, poziotinib has more freedom of rotation around the amine and ether groups and increased halogenation of the terminal benzene ring (13).

On the basis of these and other structural characteristics, Robichaux and colleagues demonstrated in vitro activity of poziotinib against the most common EGFR and HER2 exon 20 insertion mutants in Ba/F3 models, including average IC₅₀ values approximately 100 times more potent than osimertinib and 40 times more potent than afatinib in vitro in EGFR ins20-expressing cell lines. In the same preclinical studies, EGFR ins20 mutants were found to be 65 times more sensitive to poziotinib than EGFR T790M. Notably, poziotinib demonstrated comparable potency against wild-type EGFR in vitro, suggesting a limited therapeutic window. Using 3-D modeling, Robichaux and colleagues showed that EGFR ins20 mutations create changes in the drug-binding pocket that result in steric hindrance preventing binding of most third-generation EGFR inhibitors, but that poziotinib could more effectively bind EGFR ins20, potentially due to its smaller size and flexibility (13). Specific comparisons between poziotinib and afatinib predicted lower free energy for binding of poziotinib to EGFR ins20 D770insNPG, suggesting a tighter binding affinity.

On the basis of these in vitro studies, follow-up experiments in genetically engineered mouse models and patient-derived xenografts (PDX), and phase I studies determining recommended phase II doses (RP2D), a single-center phase II clinical trial enrolled previously treated patients with NSCLC harboring exon 20 insertions in EGFR or HER2 (Table 1; ref. 19). Among 40 evaluable patients with EGFR ins20, the 8-week unconfirmed objective response rate (ORR) was a promising 58% (23/40). The ORR among patients with prior TKI treatment was 62% (8/13). The most common treatment-related adverse events were rash and diarrhea, notably with 60% of patients experiencing grade 3 toxicities and 45% requiring dose reduction. Among the HER2 cohort (n = 13), there was a 50% ORR and a similar toxicity profile, with the addition of one case of grade 5 pneumonitis.

Based upon these initial clinical data, a follow-up multicenter phase II study (ZENITH20; NCT03318939; Table 1) was launched for patients with either EGFR or HER2 exon 20 insertions with at least one prior line of therapy (20). Unfortunately, the results of the ZENITH20 trial showed more limited clinical activity against EGFR ins20. Among 115 patients with EGFR ins20 and a median of two prior lines of therapy, the confirmed ORR was 15% with a median progression-free survival (PFS) of 4.2 months. There were also high rates of skin and gastrointestinal toxicities, with 65% of patients requiring dose reduction from the starting dose of 16 mg daily. This type of EGFR–wild-type side effect profile illustrates the difficulty in obtaining an adequate therapeutic index for TKIs in EGFR ins20 NSCLC, a challenge that has also limited the clinical utility of afatinib and dacomitinib.

Taken together, poziotinib is a structurally unique, irreversible EGFR TKI with moderate activity against EGFR ins20 NSCLC, but whose clinical activity is tempered by significant side effects related to inhibition of wild-type EGFR. Thus, the generalizable activity of poziotinib (when dosed once daily) appears to be more limited against EGFR ins20 than initially hoped, particularly given other competitor compounds in development. However, the ongoing cohorts of the ZENITH20 trial are exploring alternative dosing strategies, such as 8 mg twice daily, against both EGFR and HER2 ins20.

Mobocertinib (formerly known as TAK-788 and AP32788) is another EGFR and HER2 TKI that binds covalently to C797 and was shown in preclinical Ba/F3 models to have improved selectivity for inhibition of EGFR ins20 over wild-type compared with other TKIs (21, 22). The initial clinical data from the phase I/II multicenter study of mobocertinib (NCT02716116) included 28 patients with advanced, previously treated EGFR ins20 NSCLC treated at the RP2D (160 mg daily) and showed a confirmed ORR (by investigator review) of 43% and median PFS 7.3 months (ref. 23; Table 1). More recently, among 96 previously treated patients (median 1 prior line of therapy) enrolled in the phase II EXCLAIM cohort of the study, the confirmed ORR was slightly lower at 23% by blinded independent review and 32% by investigator review, with a median PFS of 7.3 months (24). Similar to poziotinib, the most common side effects were gastrointestinal, with 92% of patients treated at 160 mg having any grade diarrhea (16% grade ≥ 3) and 28% having nausea. Rash was seen in 45%. Treatment-related adverse events led to dose reduction in 21% of patients and treatment discontinuation in 10% of patients.

Based upon these preliminary results, the FDA granted mobocertinib orphan drug designation (2019) and subsequently breakthrough therapy designation (April 2020). This was one of two breakthrough therapy designations granted in 2020 for previously treated advanced EGFR ins20 NSCLC, a significant clinical advancement for a tumor type without any approved targeted therapies at the time. The ongoing phase III EXCLAIM-2 study is further evaluating the efficacy of mobocertinib versus platinum-doublet chemotherapy among treatment-naïve patients (NCT04129502; Table 1), and will be the first study to prospectively evaluate the role of an EGFR ins20–targeted therapy compared with the current first-line standard of care.

Osimertinib (AZD9291) is a third-generation mutant-selective EGFR TKI with a mono-anilino-pyrimidine compound structure that has activity against a wide variety of EGFR mutants, including del19, L858R, T790M, G719, and other activating point mutations, and is currently approved for first-line and adjuvant use for del19 and L858R as well as second-line use for T790M (25–29). A key feature of osimertinib is its wide therapeutic window against mutant forms of EGFR relative to inhibition of wild-type EGFR, suggesting a potential therapeutic advantage for targeting EGFR ins20 mutations. Although the initial publication describing preclinical activity of osimertinib did not show efficacy against EGFR ins20 in one model (ref. 30; H773-V774HVdup in HEK293 cells), further work suggested the potential for a reasonably wide therapeutic window (31). Specifically, Hirano and colleagues found in cell line models that all tested EGFR ins20 mutants except for EGFR A763_Y764insFQEA were resistant to first-generation EGFR TKIs, but observed that afatinib and osimertinib had similar IC₅₀ values against three representative first-generation TKI-resistant EGFR ins20 variants (Y764_V765insHH, A767_V769dupASV, and D770_N771insNPG; ref. 31). The authors subsequently compared IC₅₀ values of EGFR TKIs against wild-type versus ins20 mutants, as a surrogate to approximate therapeutic index. In these studies, they found that only osimertinib had a favorable IC₅₀ ratio (approximately 10-fold higher for wild-type EGFR as compared with the EGFR ins20 variants). However, although osimertinib had a higher ins20 mutation specificity compared with other EGFR TKIs, the IC₅₀ values of osimertinib for EGFR ins20 were still 10 to 100 times higher than those for 19del, L858R, and T790M, suggesting higher dosing may be required to achieve clinical efficacy in the treatment of EGFR ins20 NSCLC.

In similar types of experiments, Floc'h and colleagues later introduced EGFR ins20 D770_N771insSVD and V769_D770insASV into NSCLC cell lines and collected available PDX models of EGFR ins20 to test osimertinib efficacy in an environment more closely mimicking a clinical primary tumor (as compared with Ba/F3 models; ref. 32). In these experiments, they were able to confirm in both in vitro and in vivo models that osimertinib has appreciable antitumor activity in these two common variants of EGFR ins20 NSCLC.

Based upon these data, osimertinib is now being tested clinically in EGFR ins20 NSCLC. Our group published the first report of a clinical response to osimertinib in EGFR ins20 NSCLC involving an 80-year-old female never-smoker who presented with metastatic lung cancer harboring EGFR ins20 S768_D770dup. After whole-brain radiotherapy for symptomatic metastases, the patient elected for off-label high-dose (160 mg daily) osimertinib and had an extracranial partial response that was sustained for 11 months (33). Similarly, a Chinese group reported retrospective results on six patients harboring EGFR ins20 treated with osimertinib 80 mg daily, with four partial responses (12). Finally, ECOG-ACRIN EA5162 is a prospective, phase II clinical trial evaluating osimertinib at 160 mg daily for patients with EGFR ins20 NSCLC who have had at least one prior line of therapy (NCT03191149; Table 1; ref. 34).

Although the study did not meet its primary endpoint, initial results from ECOG-ACRIN EA5162 demonstrated three confirmed partial responses, one unconfirmed response, and one complete response among the first 17 evaluable patients in the trial, for a confirmed ORR of 4/17 (24%). The median PFS was 9.6 months. Osimertinib 160 mg daily led to higher rates of diarrhea (76% any grade), fatigue (67% any grade), and acneiform rash (38% any grade) than is typical at 80 mg daily, but no cases of grade 3 gastrointestinal or dermatologic toxicities were observed. On the basis of these preliminary results, expansion of EA5162 is planned to further investigate the role of osimertinib 160 mg in a larger cohort of patients with EGFR ins20.

Of note, there are no clinical data in EGFR ins20 patients directly comparing standard-dose osimertinib 80 mg with the higher 160 mg daily dose. However, the higher dose is being empirically tested in EGFR ins20 disease for two reasons. First, despite osimertinib's known sparing of wild-type EGFR inhibition, the preclinical data discussed above suggest higher IC₅₀ values are required to inhibit EGFR ins20 compared with canonical mutations. Second, osimertinib has previously been demonstrated to be safe and effective at 160 mg daily in patients with central nervous system (CNS) disease in which a higher drug concentration is needed to achieve adequate response (35–37).

CLN-081 (formerly TAS6417) is a novel EGFR TKI with a unique core structure (6-methyl-8,9-dihydropyrimido[5,4-b]indolizine) that fits into the ATP-binding site in the hinge region of EGFR and was specifically designed to inhibit EGFR ins20 (38). Similar to other irreversible EGFR TKIs, mass spectrometry analysis of CLN-081 showed that this novel small molecule also covalently modifies the Cys797 residue of EGFR proteins harboring exon 20 insertions. Hasako and colleagues further demonstrated the potency and selectivity of CLN-081 in in vitro models of NIH/3T3 and Ba/F3 cell line models harboring various EGFR ins20 mutations and found CLN-081 showed a wild-type to mutant IC₅₀ ratio of 134-fold for A763_Y764insFQEA, 17.4-fold for D770_N771insSVD, 17.2-fold for D770_N771insG, 6.37-fold for V769_D770insASV, 4.55-fold for H773_V774insPH, and 4.51-fold for H773_V774insNPH (38). By comparison, the first- and second-generation EGFR TKIs erlotinib and afatinib showed minimal selectivity for mutant receptor in these studies, with less than 1-fold wild-type to mutant IC₅₀ ratios except for the known TKI-sensitive insertion mutation A763_Y764insFQEA. Similarly, Udagawa and colleagues showed that CLN-081 has a roughly 10- to 30-fold lower IC₅₀ value in two EGFR ins20 cell lines as well as a significantly higher selectivity index (39). Consistent with these studies, CLN-081 also inhibited growth and induced apoptosis in human NSCLC cell lines with EGFR ins20 and resulted in significant tumor regression when administered in vivo against xenograft models harboring V769_D770insASV, D770_N771insSVD, and H773_V774insNPH (38). Finally, CLN-081 was shown to prolong survival of genetically engineered mouse models bearing EGFR ins20 NSCLC in a dose-dependent fashion (38).

On the basis of these preclinical data, CLN-081 is now being tested in an open-label, multicenter phase I/IIa trial (NCT04036682; Table 1). Early data from the first 17 EGFR ins20 patients enrolled demonstrate an unconfirmed ORR of 35% with disease control rate of 100% (17/17; ref. 40). The RP2D has not yet been established and enrollment to the phase I/II study is ongoing.

Another newly developed TKI specifically designed to irreversibly and selectively inhibit noncanonical, allosteric ErbB receptor mutations is BDTX-189. Initially shown in preclinical studies to have an inhibitory effect on both EGFR and HER2 ins20 mutations as well as mutations in the extracellular domain, BDTX-189 is now being tested in a phase I/II clinical trial (NCT04209465; Table 1). Patients enrolled on this trial have metastatic disease without alternative therapy options and tumors that harbor EGFR or HER2 ins20 mutations or one of a number of other oncogenic alterations in ErbB family members.

Amivantamab (formerly known as JNJ-61186372) is a bispecific antibody targeting both EGFR and cMET (41). First-in-human studies of amivantamab in patients with advanced EGFR-mutant NSCLC harboring a variety of different EGFR mutations showed an encouraging preliminary efficacy and safety profile (CHRYSALIS; NCT02609776; Table 1; ref. 42). Of note, several early responses were seen in patients with EGFR ins20 (41), prompting preclinical analyses to further describe the potential of amivantamab as a therapy specifically for EGFR ins20.

Utilizing Ba/F3 cells engineered to express EGFR V769_D770insASV, D770delinsGY, H773_V774insH, and D770_N771insSVD, Yun and colleagues showed that amivantamab inhibits cell viability in a dose-dependent manner and to a significantly greater extent than osimertinib (41). In two of three patient-derived models of EGFR ins20 NSCLC (P772ins_H773insPNP and H773_V774insNPH), amivantamab showed significant growth inhibition of cell lines as well as associated organoid and PDX models. Of note, amivantamab demonstrated greater activity in PDX models than either poziotinib or cetuximab, and it also caused less skin toxicity (a side effect of inhibiting wild-type EGFR) than poziotinib (41). Mechanistically, these authors showed that treatment with amivantamab resulted in downregulation of EGFR and cMET in cell lines harboring various EGFR ins20 mutations, similar to what was previously shown in EGFR-mutant NSCLC cell lines harboring canonical driver mutations (43). This process is mediated by internalization and lysosomal degradation of the receptors, causing inhibition of downstream signaling and induction of apoptosis (41). This work also showed that antibody-dependent cell-mediated cytotoxicity plays a central role in the inhibition of EGFR and MET signaling following exposure to amivantamab (41).

The superior efficacy of amivantamab over other EGFR mAbs is not fully understood; however, experts have speculated that its bispecific nature uniquely contributes to its antitumor efficacy. In other words, it may be that the process of amivantamab binding both EGFR and MET at the same time creates a synergistic inhibitory effect, as the interconnected downstream signaling pathways are simultaneously affected. This hypothesis is supported by preclinical work showing greater inhibitory effect of amivantamab than the combination of two separate anti-EGFR and anti-MET monovalent antibodies in multiple models of both EGFR- and MET-driven NSCLC (43). Some studies also suggest that bispecific antibodies have improved selectivity for receptor-overexpressing cells because they need to bind both independent epitopes at one time (44). It should be noted that the biological context is especially important when considering mechanism of antibody-based therapies, as even subtle differences in degree of ligand dependence, receptor density on the cell surface, and other factors can make important differences.

Updated results from the dose-escalation and EGFR ins20 dose-expansion cohorts of the CHRYSALIS phase I study were recently presented and demonstrate clinical efficacy (Table 1). To date, 81 patients with EGFR ins20 NSCLC have been treated with amivantamab at the RP2D of 1,050 mg (1,400 mg for patients ≥80 kg), with an ORR of 40% with a median PFS of 8.3 months (45, 46). Skin rash was reported as a common side effect in 86% of patients, with only 4% grade 3 rash. Other common treatment-related adverse events were infusion-related reaction (66% with only 3% grade ≥ 3, mostly with the first infusion) and paronychia (42%), but overall treatment-related grade 3 events were seen in only 16%. Amivantamab became the first EGFR ins20–targeted therapy to be granted FDA breakthrough therapy designation in March of 2020 and the first to be granted FDA approval in May of 2021. A randomized first-line study of amivantamab plus chemotherapy versus chemotherapy alone in EGFR ins20 is ongoing (PAPILLON, NCT04538664.)

Although both TKIs and EGFR-targeted antibodies are being studied for clinical efficacy as monotherapy in EGFR ins20 NSCLC, the potential for combination of these two therapeutic modalities remains an area of interest as well. This is especially true for EGFR mAbs, as the single-agent efficacy has historically been less promising in models of canonical TKI-sensitive EGFR-mutant NSCLC, and they also appear to have less efficacy in EGFR ins20 disease as described above. Antibody–TKI combinations currently being investigated include cetuximab, necitumumab, and amivantamab in combination with multiple second- and third-generation EGFR TKIs.

The combination of afatinib plus cetuximab has previously shown clinical efficacy in canonical EGFR-mutant NSCLC, although this combination led to high rates of dermatologic toxicities and a phase III study did not improve PFS compared with afatinib (47, 48). More recently, Hasegawa and colleagues tested the growth-inhibitory effects of afatinib/cetuxmab in preclinical models of EGFR ins20 NSCLC compared with either therapy alone, and found that the combination induced more effective growth inhibition of Ba/F3 cells engineered to express EGFR A763_Y764insFQEA, Y764_V765insHH, A767_V769dupASV, and D770_N771insNPG both in vitro and in vivo as subcutaneous xenografts (49). In a case series of four patients with EGFR ins20 NSCLC previously treated with platinum-based chemotherapy ± EGFR TKI, afatinib/cetuximab achieved a partial response in three of the patients (Table 1; ref. 50). The range of response duration was 2.7 to 17.6 months, and dose reduction was required in two of the four patients. A phase II single-arm clinical trial evaluating afatinib/cetuximab in EGFR ins20 NSCLC is ongoing (NCT03727724; Table 1).

Using a similar approach, a recent phase I dose-escalation study of necitumumab and osimertinib in patients with advanced EGFR TKI–resistant NSCLC showed 2 of 4 responses in EGFR ins20 patients, with median PFS of 5.3 months in this small cohort (NCT02496663; Table 1; ref. 51). Investigators in China are also studying the EGFR mAb JMT101 plus EGFR TKI (either osimertinib or afatinib) in an ongoing phase Ib trial for EGFR ins20 NSCLC (NCT04448379; Table 1), and amivantamab is also currently being tested in combination with the EGFR TKI lazertinib as part of the combination and dose-expansion cohort of part 2 of the CHRYSALIS phase I trial described above (45).

About ten years ago, hsp90 inhibitors were being studied in a wide range of patients with NSCLC. EGFR is a Hsp90 client, requiring the chaperone for stabilization of the protein to conformational maturity (52). Initial preclinical studies demonstrated that the HSP90 inhibitor luminespib (AUY922) caused inhibition of downstream signaling and induction of apoptosis across a broad panel of different EGFR ex20ins mutants (53). Included in this initial study was also the first report of efficacy in a patient, as demonstrated by a confirmed radiographic partial response to luminespib lasting 18 weeks in a patient with EGFR ins20 NSCLC (53).

Our group tested the clinical activity of luminespib in EGFR ins20 in a phase II single-arm, investigator-initiated clinical trial (Table 1; ref. 54). Twenty-nine patients with at least one prior therapy were enrolled, yielding an ORR of 17%, median PFS of 2.8 months, and median OS of 9.9 months. Six patients (21%) required dose reduction due to treatment toxicity, with the most significant treatment-related adverse effects being diarrhea, ocular toxicity (requiring two cases of treatment discontinuation), fatigue, and hypertension. Unfortunately, this compound is no longer being clinically developed.

Tarloxotinib is a novel agent developed as a prodrug that requires pathophysiologic hypoxic conditions for activation of the irreversible dual EGFR/HER2 TKI called tarloxotinib-E (55–57). Initially tested in multiple settings of EGFR- or HER2-mutant NSCLC, including T790M-negative acquired resistance to first-generation EGFR inhibitors (58), the driving hypothesis for its unique clinical application was the hope that selective activation in hypoxic tumor environments would allow for a greater therapeutic window for inhibition of EGFR signaling preferentially in cancer cells over normal cells. Given that this is a central challenge in inhibiting EGFR ins20, multiple groups performed initial preclinical experiments testing tarloxotinib in models of EGFR ins20 NSCLC (55, 57).

Estrada-Bernal and colleagues found that in patient-derived NSCLC cells harboring A767_V769dupASV, N771_H773dupNPH, and S768_770dupSVD, the IC₅₀ for tarloxotinib-E was either similar or significantly lower than afatinib, and in xenograft models of the same cell lines harboring A767_V769dupASV and N771_H773dupNPH variants, tarloxotinib-E induced significantly greater tumor regression compared with afatinib (55, 59). From a toxicity perspective, it is notable that the prodrug tarloxotinib is 100-fold less potent against wild-type EGFR than the active metabolite tarloxotinib-E, and xenograft studies demonstrate higher exposure of active metabolite in tumor tissue compared with nonmalignant skin tissue (59). This suggests that a wide therapeutic index is achievable for antitumor activity relative to on-target wild-type EGFR toxicity in adjacent normal tissues.

Similarly, Nishino and colleagues utilized Ba/F3 cells with the introduced EGFR ins20 mutations A763insFQEA, V769insASV, D770insSVD, H773insH, and H773insNPH to test efficacy of tarloxotinib-E as compared with other EGFR TKIs (57). They found tarloxotinib-E had lower IC₅₀ values for inhibition of each of these ins20 mutants than afatinib and osimertinib, but poziotinib was a more potent inhibitor of all mutant receptors except A763insFQEA. These authors also demonstrated that the IC₅₀ for inhibition of ins20 was on average greater than 72.1 times higher for tarloxotinib-E compared with prodrug, again suggesting potential for a wide therapeutic index.

Despite these compelling preclinical data, an ongoing phase II study of tarloxotinib showed no responses among 11 patients with EGFR ins20 (Table 1; ref. 60). However, responses to tarloxotinib were seen among NSCLCs harboring HER2 exon 20 insertions and HER2, NRG1, and other fusions, with enrollment ongoing in these groups.

With multiple drugs in clinical development and showing signs of preliminary activity for patients with EGFR ins20–mutant NSCLC, further work is required to identify optimal treatment strategies based on specific ins20 mutation, treatment history, and sites of disease. In addition, it is clear that overcoming acquired resistance and determining the best way to sequence these different therapies will be critical to improving long-term patient outcomes.

Given the spectrum of EGFR ins20 mutations identified in patients, preclinical studies testing new targeted therapies must compare efficacy across multiple different mutation subtypes. Although we know that EGFR A763_Y764insFQEA mutation behaves more like the canonical EGFR oncogenic mutations del19 and L858R in terms of TKI sensitivity, this appears to be the only specific EGFR ins20 mutation to do so. Comparing the remainder of the ins20 mutations identified across the rest of the C-helix domain and the loop following the C-helix domain, the response of each of them to specific therapies does not consistently or significantly differ in available preclinical studies.

Whenever possible, clinical trial data must be analyzed for mutation-specific therapeutic efficacy in addition to general response rates in all patients with EGFR ins20 mutations; however, available clinical data for variant-specific therapeutic response to targeted therapies thus far suggest limited differentiation between any of the non-A763_Y764insFQEA variants in terms of treatment efficacy. In the case of poziotinib, the analysis of response by specific mutation reveals the predominance of responses in tumors harboring more proximal insertions between M766-D770 (18%; 8/44 responders). For mobocertinib and amivantamab, there did not appear to be a clear association of likelihood or depth of response with specific EGFR ins20 variants in early trial data (46). We do not yet have data for variant-specific response rates for osimertinib or other emerging promising therapies. Similarly, analysis of response data for luminespib by subtype of EGFR ins20 mutation revealed no correlation with response data, and responses in early trials were seen among patients with D770_N771insSVD, D770_P772dup, N771_H773dupPH, and P773dup.

Although the data discussed in this review demonstrate potential clinical activity for multiple new targeted agents against EGFR ins20 NSCLC, we know from experience with other targeted therapies that the duration of clinical response is likely to be limited by acquired resistance. Of the leading targeted therapies for EGFR ins20 mutations, poziotinib is currently the compound for which we have the most information about potential resistance mechanisms. It is known that an acquired point mutation in the 797 residue of EGFR that serves as a binding site for covalent inhibitors (C797S) can cause resistance to irreversible inhibitors of EGFR in other clinical contexts (30, 61, 62). Introduction of C797S into Ba/F3 cell lines also confers resistance to poziotinib, suggesting that this site is also important for poziotinib binding (13). Another study using ENU mutagenesis screens in Ba/F3 cells with multiple different EGFR ins20 variants (A763insFQEA, V769insASV, D770insSVD, and H773insNPH) confirmed EGFR C797S as a likely mediator of poziotinib resistance and further demonstrated that the EGFR T790M mutation also confers in vitro poziotinib resistance (57). Interestingly, this study showed that both C797S and T790M emerged as resistance mechanisms to tarloxotinib as well, and in the case of both tarloxotinib and poziotinib, the specific exon 20 insertion affected the likelihood of cells developing T790M versus C797S in the presence of drug (57).

Additional preclinical studies in EGFR-mutant NSCLC cell lines with acquired resistance to first-generation EGFR TKIs mediated by phenotypic changes such as epithelial-to-mesenchymal transition also conferred cross-resistance to poziotinib in vitro (13). Analysis of resistance mechanisms was also performed using tissue biopsies and cell-free DNA from patients in the original phase II trial of poziotinib; comparing pretreatment and postprogression samples from 20 patients with acquired resistance to poziotinib, four patients developed acquired EGFR kinase domain mutations (T790M, V774A, and D770A; ref. 63). Other potential acquired resistance mechanisms identified to date in clinical samples include PIK3CA E545K, MAP2K2 S94L, MET amplification, EGFR amplification, and CDK6 amplification (63).

Although specific mechanisms of resistance to osimertinib in EGFR ins20 NSCLC have not yet been elucidated, we do have emerging data about osimertinib resistance in other subtypes of EGFR-mutant NSCLC, including second-site kinase domain mutation at the osimertinib binding residue (C797S), bypass pathway activation (MET amplification, acquired fusions in RET, ALK, ROS1), and histologic transformation to small-cell lung cancer or squamous cell carcinoma (64–66). Whether these resistance mechanisms will emerge in osimertinib-treated EGFR ins20 NSCLC, and at what relative frequencies, remains unknown.

Potential resistance mechanisms for other targeted therapies in clinical development, most notably CLN-081, mobocertinib, and amivantamab, remain unknown at this time. One could speculate that similar mechanisms may be seen for CLN-081 and mobocertinib as for other covalent small-molecule inhibitors, such as EGFR kinase domain mutations (e.g., C797S) that affect drug-binding affinity, or activation of bypass receptor signaling pathways circumventing EGFR signaling. In contrast, we might expect mechanisms of resistance to amivantamab to be more likely to include activating mutations farther downstream in signaling pathways (such as RAS mutations seen in cetuximab resistance in colorectal cancer), or alterations causing changes in cell-surface expression and therefore epitope availability of either EGFR or MET. Ultimately, characterization of these resistance mechanisms and development of strategies to overcome them will rely on clinical data from patients treated with these targeted agents.

A final important consideration when evaluating the efficacy of emerging targeted therapies for EGFR ins20 NSCLC is the CNS activity of each of these compounds. At this time, we do not know the specific CNS activity for most of the compounds discussed above, but limited data from the mobocertinib clinical trial show lower ORR among patients with CNS metastases than those without, suggesting a lack of CNS activity for this compound (21, 24). Similarly, initial data from the ZENITH trial suggest that patients with brain metastases experienced lower response rates to poziotinib than those without CNS disease (67). In contrast, although high-dose osimertinib has been shown to have less mutant selectivity than other compounds in development for EGFR ins20 disease, its strong CNS activity has been demonstrated in previous clinical studies with canonical EGFR oncogenic mutations (35).

Preliminary data in a limited number of patients treated with amivantamab suggest comparable outcomes in patients with brain metastases versus those without; however, we would expect that its large molecular size would preclude strong CNS activity. This is the rationale for clinical trials testing it in combination with a CNS-penetrant third-generation EGFR TKI such as lazertinib. At this time, we do not have any clinical data for CNS activity of CLN-081 or tarloxotinib.

Although clinical outcomes for patients with classic, canonical, sensitizing EGFR mutations in NSCLC have improved drastically since the development of EGFR TKIs, patients with tumors harboring EGFR ins20 have not yet seen the same degree of clinical advances. Historic challenges include limitations in methods for diagnosis (as some early genotyping assays lacked coverage for detection of EGFR ins20), the sheer number of different ins20 variants that exist, and the relative paucity of preclinical models for laboratory study. These barriers initially limited progress by delaying our ability to robustly quantify the frequency of, and characterize the biology of, NSCLC tumors harboring EGFR ins20. With rapid improvements in sequencing technology and careful retrospective review of clinical data, the past decade has brought clearer understanding of EGFR ins20 as the third most common subtype of EGFR-mutant NSCLC, one with unique biology requiring its own dedicated approach to development of targeted therapy.

The progress made in overcoming these systemic challenges has greatly accelerated the pace of recent advances in EGFR ins20–targeted therapies, as evidenced by the recent FDA approval of amivantamab (Fig. 3). It should also be noted that the pace of recent discovery has been influenced by the dedicated involvement of patient advocacy groups. These groups, such as the Exon 20 Group (Exon20group.org), have enhanced the organization and dissemination of information to patients about new therapies/clinical trials via multiple platforms, including social media. Moving forward, patient advocacy groups will continue to be a critical partner in our ability to improve clinical outcomes for patients with rarer disease subsets.

To sustain the momentum of discovery and improve outcomes in patients with EGFR ins20 NSCLC, several key scientific challenges lie ahead. From a drug development perspective, the structural similarities between wild-type EGFR and EGFR ins20 make achieving a reasonable therapeutic window for targeting these mutations exceedingly difficult. Without the advantage of the increased affinity for TKI versus ATP binding that occurs with classic sensitizing mutations, in EGFR ins20 disease, TKIs that are found to have activity against the mutant receptor often also cause significant inhibition of wild-type EGFR in normal tissues. This can result in sometimes-unmanageable side effects and limit clinically achievable doses. Development of more mutant-selective TKIs or other creative alternative strategies for tumor-specific inhibition of EGFR signaling remains a high priority for future studies. Also, as we have learned from other subsets of oncogene-driven NSCLCs, improving the CNS penetration of targeted therapies is another major opportunity for advancing clinical outcomes. Without both CNS activity and sufficient mutant selectivity to achieve clinically effective doses in the brain, CNS progression will be a risk for patients on these therapies.

Another important goal for ongoing research efforts is the development of a more nuanced understanding of the biology of specific EGFR ins20 variants, as well as the common mechanisms of resistance to each of the new therapies being developed. Except for EGFR ins20 A763_Y764insFQEA, all other EGFR ins20 variants identified to date have been found to be resistant to early-generation EGFR TKIs. However, the potential differential sensitivity of each of more than 60 known EGFR ins20 variants to these newly developed next-generation targeted therapies remains largely unknown. Although small case numbers often limit robust analyses such as these, ongoing careful clinical trial designs as well as expanded preclinical studies can continue to add to our collective expertise. In addition, as described in previous sections, early studies evaluating resistance mechanisms to EGFR ins20 inhibitors showed both on-target and off-target mechanisms of resistance, and there are preliminary preclinical data suggesting that specific resistance mutations may depend on the specific founder EGFR ins20 variant (63).

Putting these emerging data into context and taking a step back, the EGFR ins20–targeted therapy landscape is beginning to more closely resemble that of canonical EGFR mutations and other known targetable drivers in the field. Although the response rates to available therapies are currently not as impressive as NSCLC cohorts driven by ALK and ROS1 rearrangements, it is important to note that EGFR ins20 patients alone make up approximately the same total percentage of patients with lung cancer as each of these distinct subsets. Now that multiple new targeted therapies are in development for this patient group, the complexity of clinical decision-making around how to choose the right therapy at the right time for these patients will continue to grow. First and foremost, the choice of which targeted therapies to start with will depend on evolving clinical trial data comparing response rates, OS, and CNS activity across the panel of emerging therapies and compared with standard of care. Choices about subsequent lines of targeted therapy will require careful attention to available molecular data, continued collective discussion about correlation to patient outcomes, and further laboratory-based mechanistic studies of efficacy of sequential application of these therapies.

From a practical perspective, for clinicians treating patients with EGFR ins20–positive NSCLC today, we encourage exploration of clinical trial options whenever possible, including newly diagnosed patients. Although the landscape of available clinical trials is evolving rapidly, a current summary of ongoing clinical trials for EGFR ins20 is provided in Table 1. Given both the activity and safety profiles of mobocertinib and other EGFR ins20 TKIs, randomized studies such as EXCLAIM-2 and PAPILLON will be critical to determining the optimal first-line treatment strategies. For those unable or ineligible to participate in clinical trials, chemotherapy remains an effective treatment strategy. Although data are limited, the off-label use of osimertinib 160 mg daily can be considered in select patients who cannot access clinical trials and do not have other standard-of-care options available, particularly if CNS metastases are present.

The role of chemoimmunotherapy in this population remains largely unknown, though the combination regimen of carboplatin, paclitaxel, bevacizumab, and atezolizumab (Impower150 regimen) is approved in some countries for treatment of EGFR-mutant lung adenocarcinoma based on a randomized phase III trial that demonstrated improved OS in nonsquamous NSCLC regardless of PD-L1 expression level or the presence of EGFR or ALK alterations (68). Approval in the United States is currently limited to treatment of patients with tumors expressing high PD-L1 activity. Another potential chemoimmunotherapy option for these patients is pending the results of the ongoing KEYNOTE-789 (NCT03515837) study investigating pemetrexed plus platinum chemotherapy with or without pembrolizumab in patients with EGFR-mutant nonsquamous NSCLC with acquired resistance to TKIs. If this study demonstrates positive results, it may be a compelling nontargeted therapy option for patients with EGFR ins20 disease, and head-to-head trial comparisons of chemoimmunotherapy versus leading targeted therapy options will be important. Of note, however, a recent retrospective study of a series of 6,290 patients with NSCLC found that those harboring tumors with EGFR ins20 had longer OS and time to treatment discontinuation with platinum chemotherapy and no improvement in treatment duration with the addition of immune checkpoint inhibitors compared with patients without targetable mutations (69). Finally, the activity of immune checkpoint inhibition as monotherapy is likely to be limited on the basis of experience with canonical EGFR mutations (70, 71).

Although the diagnosis of EGFR ins20 NSCLC historically meant a dismissal of targeted therapy options, the type of precision medicine usually reserved for other subsets of EGFR-mutant NSCLC is currently becoming a reality for these patients. As such, a hopeful vision for the future is the ability for clinicians to more effectively select the most appropriate type and sequence of therapy for each individual patient, even those with rare driver mutations.

L.V. Sequist reports grants and personal fees from AstraZeneca, Genentech, and Merrimack; grants from Novartis, Boehringer-Ingelheim, LOXO, and Blueprint Medicines; and personal fees from Janssen outside the submitted work; in addition, L.V. Sequist has a patent for BLU-667 + osimertinib pending. Z. Piotrowska reports personal fees from AbbVie, Guardant, Eli Lilly, InCyte, Genentech, C4 Therapeutics, Blueprint Medicine, Janssen, and Jazz Pharmaceuticals; grants and personal fees from Novartis, Takeda, Spectrum, AstraZeneca; and grants from Tesaro, Cullinan Oncology, and Daichi-Sankyo outside the submitted work. No disclosures were reported by the other author.

The authors acknowledge support from Lungstrong, Targeting a Cure for Lung Cancer, The Susanne E. Coyne fund, and Be a Piece of The Solution.