EGFR Exon 20 Insertion Mutations Display Sensitivity to Hsp90 Inhibition in Preclinical Models and Lung Adenocarcinomas

Precision medicine is an important palliative modality in the management of advanced lung adenocarcinomas (1), with oral kinase inhibitors approved for advanced tumors with epidermal growth factor receptor (EGFR) mutations (gefitinib, erlotinib, icotinib, afatinib, olmutinib, and osimertinib), anaplastic lymphoma kinase (ALK) rearrangements (alectinib, crizotinib, ceritinib, and brigatinib), ROS proto-oncogene 1 (ROS1) rearrangements (crizotinib), or B-Raf proto-oncogene, serine/threonine kinase (BRAF) V600E mutation (dabrafenib plus trametinib). EGFR mutations, first identified in 2004, mostly cluster around the kinase domain of the receptor (2). The most frequent EGFR mutations, more than 75% of mutations found in non–small cell lung cancer (NSCLC), are indels around exon 19 and the exon 21 L858R, which are oncogenic and hypersensitive to reversible (first generation), irreversible (second generation), and covalent mutation–specific (third generation) EGFR tyrosine kinase inhibitors (TKIs); hence the approval labels based on these biomarkers (2). Other less prevalent but clinically relevant EGFR mutations that predict for response to gefitinib, erlotinib, osimertinib, or afatinib include exon 18 indels/E709X, exon 18 G719X, exon 19 insertions, exon 20 A763_Y764insFQEA, exon 20 S768I, exon 21 L861Q, exon 18–25 kinase domain duplication, and EGFR rearrangements (2). EGFR TKI monotherapy is not curative in the advanced setting, and heterogeneous acquired resistance to gefitinib, erlotinib, icotinib, and afatinib occurs due to EGFR mutations, oncogene bypass tracks, or histologic transformation (3, 4). The most prevalent resistance is mediated by EGFR-T790M, a gatekeeper mutation that can be inhibited by covalent mutation–specific EGFR TKIs; with osimertinib worldwide and olmutinib in South Korea being the approved clinical drugs (5).

However, not all EGFR mutants share the same sensitivity pattern to approved EGFR inhibitor classes. The third most prevalent kinase domain EGFR mutation class encompasses in-frame insertions and indels within or following the regulatory C-helix amino-acids of exon 20 (6, 7), which account for approximately 10% of all EGFR mutations in NSCLC. These structural oncogene variants still engage phosphoinositide 3-kinase (PI3K)/protein kinase B (AKT) and extracellular-signal-regulated kinase (ERK)/mitogen-activated protein kinase (MAPK) downstream pathways with EGFR dependency (7). Our group has previously shown that the segment of the kinase domain pocket, in which first- and second-generation drug binding occurs, of exon 20 insertion mutants is structurally similar to wild-type (WT) EGFR, thereby negating a therapeutic window for use of gefitinib, erlotinib, and afatinib (7) with clinically achievable WT EGFR–sparing doses (7, 8). The lack of significant clinical responses to approved doses of first-, second-, and third-generation EGFR TKIs corroborates the preclinical data (5–11), and reiterates that EGFR exon 20 insertion–mutated tumors have an unmet clinical need within the realm of precision oncology. Although new EGFR exon 20 mutation–selective inhibitors are undergoing preclinical and clinical development (Supplementary Table S1), none is expected to receive regulatory approval within the immediate future. To circumvent this issue, we have started to identify non-EGFR–specific therapies that may be effective against EGFR exon 20 insertion–mutated NSCLCs. The heat shock protein 90 (Hsp90) chaperone complex protects cellular proteins from degradation by the ubiquitin–proteasome system and this system is coopted by driver oncogenes (such as EGFR mutants) plus cell-cycle proteins (12). In this report, we provide preclinical rationale and early clinical data to support the use of the non-geldanamycin resorcinol-containing triazolone Hsp90 inhibitor in clinical development luminespib (13–15) for this unique class of lung cancers.

The clinical and radiographic data used for this study was obtained from ongoing Institutional Review Board (IRB) approved protocols at our institutions. The case was identified as a participant from clinical trial NCT01124864 (15). Written informed consent was obtained from the patient when entering NCT01124864 and the review of data was conducted in accordance with recognized ethical guidelines.

Erlotinib, afatinib, and rociletinib were purchased from LC Laboratories. Luminespib was purchased from Selleckchem. All reagents were dissolved in dimethyl sulfoxide and stored at −80°C.

Ba/F3 cell lines, BID007, NCI-H1975 (H1975), NCI-H3255 (H3255), and PC9 cells were obtained, authenticated, and maintained in RPMI-1640 medium (Mediatech) supplemented with 10% FBS, as described previously (7). The Ba/F3 EGFR-WT cell system (Supplementary Fig. S1) was derived in an identical fashion as other EGFR mutants (7) and, as an overexpression system for WT EGFR, it is used to compare the differential effects of inhibitors with EGFR-mutant proteins (8). All cells were grown at 37°C in a humidified atmosphere with 5% CO₂. Cells were thawed and tested for lack of Mycoplasma contamination (MycoAlert Mycoplasma Detection Kit, Lonza) prior to experiments that were initiated within the initial one to five passages.

Cells were plated in each well of 96-well plates and then were treated in the appropriate medium with or without EGFR or Hsp90 inhibitors for up to 3 days. Cell viability was determined by CellTiter 96 Aqueous One solution proliferation kit (Promega), as described previously (7). Inhibitory proliferation curves and the 50% inhibitory concentration (IC₅₀) were generated using GraphPad Prism 6 (GraphPad Software).

Western blot lysates and preparation were performed as described previously (7). Immunoprecipitation was achieved using protein G agarose to sepharose beads for preclearing lysates and capturing the immunocomplex after incubation with the immunoprecipitating antibody. In cell line experiments, treatment with inhibitors for the specified number of hours (figure legends) was performed to determine the response of Hsp90, EGFR, and its downstream signals to inhibitors. Immunoglobulin G (IgG) and anti-human influenza hemagglutinin (HA) tag antibodies were obtained from Abcam. Total EGFR and β-actin antibodies were obtained from Santa Cruz Biotechnology. Total ERK 1/2 antibody was obtained from BD Transduction Laboratories. Phospho-EGFR (pT1068) antibody was purchased from Invitrogen. Hsp90, heat shock protein 70 (Hsp70), phospho-AKT (pS473), total AKT, phospho-ERK (pT202/pY204), Poly ADP-ribose polymerase (PARP), and cleaved PARP antibodies were obtained from Cell Signaling Technology. Primary antibodies were diluted 1:1,000, while secondary antibodies were diluted 1:10,000.

Proliferation experiments were performed using three replicates (when not otherwise detailed). Median IC₅₀s and SDs were generated using GraphPad Prism 6 (GraphPad Software).

A panel of representative EGFR exon 20 mutations (A763_Y764insFQEA, Y764_V765insHH, M766_A767insAI, A767_V769dupASV, D770_N771insSVD, D770_N771insNPG, H773_V774insH), EGFR TKI–hypersensitive mutations (exon 19 deletion delL747_P753insS and L858R), and first-/second-generation EGFR TKI–resistant mutations (delL747_P753insS+T790M and L858R+T790M) driving an isogenic Ba/F3 cell line system was utilized to probe inhibitory concentrations in comparison with WT EGFR (Fig. 1). When we analyzed the representative first-generation EGFR TKI erlotinib, only the EGFR-exon 19 deletion, L858R, and A763_Y764insFQEA–dependent Ba/F3 cells had inhibitory concentrations that were exceedingly lower than those of the WT EGFR–driven cell (Fig. 1A). The identical pattern was observed with the second-generation irreversible EGFR TKI afatinib (Fig. 1B). In all other cells, the driver EGFR mutant led to IC₅₀s in 72-hour proliferation assays that exceed WT EGFR for erlotinib (Fig. 1A) and afatinib (Fig. 1B). In the case of the third-generation mutation-selective covalent EGFR TKI rociletinib (an inhibitor with superb in vitro profile but limited pharmacokinetic translation; Supplementary Table S1), the cells with the lowest IC₅₀s harbored delL747_P753insS+T790M and L858R+T790M followed by delL747_P753insS and L858R (Fig. 1C). All EGFR exon 20 insertions had IC₅₀s that exceed that of the WT EGFR-driven system (Fig. 1C). The phosphorylation of EGFR was only repressed in cells that were inhibited by the respective EGFR TKI (Fig. 1). These results showed that EGFR exon 20 insertion mutants do not have a therapeutic window in relation WT EGFR when treated with common first-, second-, and third-generation EGFR TKIs.

However, we and others have described that in-development doses of approved (osimertinib) or novel exon 20 mutation-specific (poziotinib and TAK-788) EGFR TKIs have shown some degree of preclinical activity (8, 16–18) and are ongoing further clinical development (Supplementary Table S1).

To evaluate whether the Hsp90 complex is responsible for degradation of EGFR exon 20 mutants, as it is for other EGFR-mutant proteins (12), we performed a coimmunoprecipitation reaction using the HA tag—that is part of our EGFR-mutant proteins (7)—with Hsp90 (Fig. 2A). Both the typical EGFR exon 20 insertion mutant D770_N771insSVD and the atypical mutant A763_Y764insFQEA disclosed physical association with the Hsp90 complex (Fig. 2A). The Hsp90 inhibitor luminespib, in the same experiments, led to dissociation and degradation of EGFR-mutant proteins (Fig. 2A). In a time-dependent manner using preclinical models driven by EGFR-D770_N771insSVD and A763_Y764insFQEA–mutated proteins, luminespib led to augmentation of Hsp70 (a protein signature of Hsp90 inhibition in view of the difficulty in measuring Hsp90 functionally), degradation of EGFR, downregulation of the activity of AKT and ERK, and induction of apoptosis as measured by cleavage of PARP at the cellular level (Fig. 2B).

Turning back to our extensive panel of representative Ba/F3 cells driven by EGFR exon 20 mutations, we showed that the Hsp90 inhibitor luminespib disclosed a similar pattern of IC₅₀s in a dose-dependent manner when comparing these mutant cells with EGFR-exon 19 deletions, L858R-, or T790M-dependent cells (Fig. 2C); and even with cells driven by overexpression of WT EGFR (Fig. 2C; Supplementary Fig. S1). The range of median IC₅₀s in all EGFR mutants tested did not veer from the 1.8 to 8.2 nmol/L range, in significant contrast to the many fold difference in EGFR TKI experiments (Fig. 1). The pattern of short-term EGFR degradation induced by increasing concentrations of luminespib was also similar between EGFR-D770_N771insSVD, A763_Y764insFQEA, L858R, or L858R+T790M (Fig. 2C).

As the aforementioned data was obtained using the Ba/F3 system, we confirmed the role of Hsp90 inhibition using NSCLC cell lines. The EGFR exon 20 insertion–mutated cell line BID007 (harboring as the driver oncogene EGFR-A763_Y764insFQEA) had inhibitory curves to increasing concentrations of luminespib that were near identical to other EGFR-mutated cell lines (Fig. 3A). The levels of EGFR degradation induced by luminespib were also similar between BID007, PC9 (EGFR-delE746_A750), and H1975 (EGFR-L858R+T790M) in a dose-dependent manner (Fig. 3B). Luminespib also induced Hsp70, degraded EGFR, downregulated the activity of AKT and ERK, and induced apoptosis over the latter time points during the course of a day in BID007 (Fig. 3D). These data confirmed in distinct preclinical models that Hsp90 inhibition should undergo further clinical testing in EGFR exon 20 insertion–mutated NSCLCs.

Our investigators cared for a patient with advanced NSCLC that was enrolled in the previously published phase II trial of luminespib in molecularly defined patients with advanced lung cancer (ClinicalTrials.gov identifier NCT01124864; ref. 15). The patient was a never smoker woman who presented at age 79 years old with recurrent lung adenocarcinoma with intra- and extrathoracic disease burden. Initial therapy with oral erlotinib at 150 mg/day was ineffective and resulted in symptomatic tumor progression within 1 month of use.

The patient then received, on the aforementioned clinical trial, intravenous luminespib at 70 mg/m² weekly and achieved a partial response (Fig. 4A) that was confirmed in subsequent imaging studies (Fig. 4B). The patient continued on luminespib for 18 weeks prior to radiographic progressive disease (Fig. 4B), and died 6 months later without intervening systemic therapy. This case spawned the design of a dedicated phase II trial of luminespib for EGFR exon 20 insertion–mutated NSCLC, NCT01854034, as discussed below.

The clinical management of advanced lung cancers with EGFR exon 20 insertion mutations remains elusive, without an approved or commercially active oral inhibitor obtainable. As detailed here and in prior reports (5–11), the currently available EGFR TKIs gefitinib, erlotinib, icotinib, afatinib, olmutinib, and osimertinib at their approved doses are unable to induce significant radiographic regressions. The development of novel EGFR TKIs that are selective for EGFR exon 20 insertions is of utmost importance for thoracic oncologists and a few novel inhibitors have shown initial preclinical and clinical activity (Supplementary Table S1). Poziotinib (formerly HM781-36B) is a pan-ErbB TKI that had limited activity in the setting of tumors with EGFR-L858R and EGFR-exon 19 deletions with enhancement for EGFR-T790M (19, 20), and therefore had initial clinical development halted. More recently, a moderate-size EGFR inhibitor screen against EGFR exon 20 insertion mutations in preclinical models disclosed selective activity of poziotinib and a single institution clinical trial has reported unconfirmed partial responses in the majority of cases (17). A confirmatory phase II trial of poziotinib (NCT03066206) for this population commenced on 2017 and is expected to end by 2021. TAK-788 (formerly AP32788) is a novel EGFR inhibitor designed to match the unique crystal structure model of the most common EGFR exon 20 insertion–mutant proteins (7) and it has shown consistently in initial preclinical models a favorable therapeutic window (lower inhibitory concentrations than EGFR WT) in EGFR exon 20–mutant models (18). The first-in-human clinical trial of TAK-788 (NCT02716116) was initialed in 2016 and due to promising initial results has been expanded from dose escalation to expansion cohorts (18), with an estimated completion date by 2020. Because of the close structural similarity of EGFR (also known as ErbB1) and Erb-B2 Receptor Tyrosine Kinase 2 (ErbB2, also known as HER2) and the promiscuity of TKIs (21), the EGFR exon 20 mutation–selective inhibitors poziotinib and TAK-788 are also being developed for ERBB2 exon 20 mutated lung adenocarcinomas (17, 18). We hope one of these drugs has a favorable toxicity profile and is found to be effective in the majority of EGFR and ERBB2 exon 20–insertion mutated NSCLCs.

Notwithstanding, it seems that the clinical path for approval of poziotinib and/or TAK-788 will take time even in the absence of major toxicities and, as for all EGFR/ErbB2 TKIs, biologic resistance will hinder long-term outcomes of monotherapy with any approved TKI. To circumvent these fallacies, the preclinical and clinical development of therapies with distinct mechanisms of action is valid. We attempted to highlight the Hsp90 system as a potential target. The non—mutant EGFR is a potential client of the Hsp90 chaperone complex but kinase mutant EGFRs (including exon 19 deletions, L858R, and T790M) are intimately dependent on the chaperone properties of Hsp90 (22, 23). The data presented here establish that EGFR exon 20 insertion mutants are bona fide clients of Hsp90, in a manner similar to more prevalent EGFR-mutant proteins, and that exploiting this dependency with Hsp90 inhibitors has merit.

Various compounds have been developed to target the Hsp90 chaperone system with different levels of clinical development. Some are geldanamycin Hsp90 inhibitors (tanespimycin/17-AAG, alvespimycin/17-DMAG, and retaspimycin/IPI-504), purine-scaffold–based inhibitors (BIIB021/CNF2024), and nongeldanamycin resorcinol-containing triazolone inhibitors (luminespib and ganetespib/STA-9090) among others (23). Although none has reached regulatory approval, there have been encouraging reports of activity in subgroups of NSCLC, most notably in oncogene-driven cohorts (15). Other groups have reported the preclinical activity of the geldanamycin inhibitor tanespimycin in EGFR- and ERBB2-mutated cell lines (24), but this compound has significant clinical toxicities. In preclinical models of EGFR- and ERBB2-mutated cells (22, 24), Hsp90 inhibition is active to a degree that exceeds the clinical experience with this class of agents and novel models are necessary to better reflect the full spectrum of response/resistance in patients. The limitations of Hsp90 inhibitors that have undergone clinical development, including luminespib, need to be addressed. These include lack of optimized biomarkers for adequate pharmacodynamic/pharmacokinetic dosing schemes in patients, the inherited toxicities associated with Hsp90 inhibition that can limit maximum dosing, the induction/upregulation of prosurvival proteins (such as can be seen with Hsp70 as noted in Figs. 2 and 3) upon Hsp90 monotherapy that may limit the effects of client oncogene inhibition, the lack of selectivity of chaperone dependency for a particular driver oncogene (as seen in Fig. 2 in relation to different EGFR proteins), and the lack of complete understanding of downstream signaling cascades plus other chaperones that may not be inhibited after Hsp90 inhibition in humans (22–26).

We chose to focus on luminespib in view of its reported clinical activity in traditional EGFR-mutated NSCLCs and the fact that a dedicated clinical trial of luminespib 70 mg/m² weekly was underway for patients with EGFR exon 20 mutated NSCLC. In a phase II study, luminespib was associated with a confirmed radiographic response rate that exceeded 15% in patients with all types of EGFR-mutated NSCLC with an acceptable adverse event profile notable for mild vision-related disorders in the majority of cases (15). The only responder, in this initial trial, with EGFR exon 20–mutated NSCLC is reported in detail here (Fig. 4) and spawned our group to complete a dedicated clinical trial of luminespib in this patient population (NCT01854034). Initial results of this Simon two-stage design clinical trial were reported in 2015, with sufficient responses within the initial strata of cases to expand the trial to its conclusion (25). The final results of the NCT01854034 trial, which completed accrual in 2016, will be reported in full elsewhere detailing the response rate, progression-free survival and toxicities of the complete cohort of 29 cases with EGFR exon 20–mutated NSCLC. If the final response rate exceeds 15%, as seen for other oncogene-driven NSCLCs (15), this will provide the proof-of-concept clinical information to stimulate further study of Hsp90 inhibition for these NSCLCs—this level of response activity is generally superior to that seen with evidence-based second-line strategies (either the cytotoxic agent docetaxel or immune checkpoint inhibitors) in unselected cases of advanced lung cancer.

It is possible to speculate that some type of Hsp90 inhibitor alone or in combination with other therapies could be clinically developed for the management of EGFR- and/or ERBB2-mutated NSCLC in the TKI-naïve or TKI-resistant setting, including tumors harboring EGFR exon 20 insertion mutations. Additional preclinical and clinical studies are warranted.

Z. Piotrowska is a consultant/advisory board member for Takeda/Ariad, AstraZeneca, Guardant Health, Novartis, and Abbvie. G. R. Oxnard is a consultant/advisory board member for Takeda, AstraZeneca, and Inivata. L. V. Sequist is a consultant/advisory board member for AstraZeneca, Pfizer, Merrimack Pharmaceuticals, Genentech, Blueprint Medicines, Novartis, and Boehringer Ingelheim. S. S. Kobayashi reports receiving speakers bureau honoraria from Ono Pharmaceutical and Boehringer Ingelheim. D. B. Costa reports receiving speakers bureau honoraria from Pfizer, and is a consultant/advisory board member for AstraZeneca, Pfizer, and Takeda. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D.B. Costa, Z. Piotrowska, S.S. Kobayashi

Development of methodology: D.B. Costa, S.E. Jorge, A.R. Lucena-Araujo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.B. Costa, A.R. Lucena-Araujo, H. Yasuda, Z. Piotrowska, G.R. Oxnard, M.S. Huberman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.B. Costa, S.E. Jorge, S.S. Kobayashi

Writing, review, and/or revision of the manuscript: D.B. Costa, S.E. Jorge, A.R. Lucena-Araujo, Z. Piotrowska, G.R. Oxnard, D. Rangachari, L.V. Sequist, S.S. Kobayashi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.B. Costa, S.E. Jorge, S.S. Kobayashi

Study supervision: D.B. Costa, S.S. Kobayashi

This work was funded in part through a Lung Cancer Foundation of America-International Association for the Study of Lung Cancer grant (to D. B. Costa), an American Cancer Society grant RSG 11-186 (to D.B. Costa), and NIH/National Cancer Institute (NCI) grants R37 CA218707 (to D.B. Costa), R01 CA169259 (to S.S. Kobayashi), and R21 CA178301 (to S.S. Kobayashi).

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