The EBC-1 cell line, which is an NSCLC cell line that harbors a MET amplification, was purchased from the JCRB Cell Bank (Osaka, Japan). Capmatinib-resistant EBC-1 cell lines (EBC-CR1, -CR2, and -CR3) were established via stepwise exposure to capmatinib at final concentrations of 1.5, 2.2, and 2.4 μmol/L, respectively and were not selected from one cell without cloning. EBC-CR1, -CR2, and -CR3 cells were maintained in 1 μmol/L of capmatinib over 2 months and selected due to diverse resistance mechanisms; these are referred to as capmatinib-resistant cell lines. All cell lines were incubated in RPMI1640 medium (Gibco, Carlsbad, CA) with 10% FBS, 2 mmol/L of L-glutamine (Gibco), and 1% penicillin/streptomycin (Gibco) in a 37˚C incubator. Capmatinib (MET inhibitor), crizotinib (anaplastic lymphoma kinase/MET inhibitor), afatinib (irreversible EGFR inhibitor), BYL719 (PI3Kα inhibitor), PD1730-74 (fibroblast growth factor receptor [FGFR] inhibitor), and BGJ398 (FGFR inhibitor) were purchased from Selleck Chemicals (Boston, MA). Capmatinib was prepared at 5 mmol/L and the other drugs were prepared as 10 mmol/L stock solutions in 100% dimethyl sulfoxide. Recombinant heparin-binding EGF-like growth factor (HBEGF) was purchased from Prospec (East Brunswick, NJ). All cell lines used in this study were authenticated by Short Tandem Repeat STR analysis using PowerPlex kit (Promega, Madison, WI).

To determine the sensitivity of the cell lines to these inhibitors in vitro, cells were seeded into 96-well microplates at 5,000 cells/well and incubated in medium for 24 hours prior to exposure to the drugs. The cells were incubated with serial dilutions of the inhibitors in a final volume of 200 μL/well for 72 hours. After the 72-hour incubation, Ez-cytox (Dogen, Seoul, Korea) was added to each well and the plates were incubated for 3 hours in a 37°C incubator. The optical density was measured at a test wavelength of 450 nm using an Eon Microplate Spectrophotometer (Biotech, Winooski, VT). The IC₅₀ values were calculated by Sigma Plot 12 (Hulinks, Tokyo, Japan). Cells were seeded in 6 well plates at a density of 5×10⁵ per wells for cell cycle and apoptosis analyses. Twenty-four hours later, the cells were treated with inhibitors for 72 hours. Both adherent and floating cells were harvested and stained with propidium iodide (1.0 mg/mL) and RNase A (10 mg/mL) (Sigma-Aldrich, St. Louis, MO). Quantification for the cell-cycle distribution and the sub-G1 population was performed by flow cytometric analysis using BD FACSCalibur (BD, East Rutherford, NJ) and ModFit LT verity software (Verity Software House Inc., Topsham, ME).

EBC-1 parental and EBC-CR1, -CR2, and -CR3 cell lines, grown to adequate densities in 6-well plates, were treated for 2 hours with capmatinib alone, 24 hours with the combination of capmatinib and afatinib, or 24 hours with BYL719. Rabbit monoclonal antibodies to detect phospho-MET (Tyr-1234/1235), phospho-EGFR (Tyr1068), phospho-STAT3 (Tyr705), phospho-AKT (Ser473), total MET, total EGFR, total AKT, total p44/42 MAPK, and glyceraldehyde 3-phosphate dehydrogenase, and mouse monoclonal antibody to detect total STAT3 were purchased from Cell Signaling Technology (Danvers, MA). Mouse anti-phospho-p44/42 MAPK (Thr-202/Tyr204) was purchased from Santa Cruz Biotechnology (Dallas, TX). Western blotting chemiluminescence reagents were purchased from GE Healthcare (Chicago, IL). All antibodies and reagents were used according to the manufacturers’ instructions.

Phospho-RTK was detected using the Human Phospho-RTK Array kit (R&D Systems, Minneapolis, MN). An LAS-3000 imaging system (Fuji Photo Film Co., Stanford, CT) was used for image analysis.

Cells were washed once with modified Dulbecco’s phosphate-buffered saline and lysed in buffer containing 0.025 M Tris, 0.15 M NaCl, 0.0001 M EDTA, 1% NP-40, and 5% glycerol (pH 7.4). Protein G/A agarose resin (Thermo Fisher Scientific, Waltham, MA) was incubated for 2 hours with rabbit monoclonal anti-MET antibody (Cell Signaling Technologies), anti-EGFR antibody (Cell Signaling Technologies) or rabbit IgG control (Invitrogen, Carlsbad, CA) as a negative control and cell lysate at 1 mg of protein was mixed with antibody-coupled resin (Cell Signaling Technologies) overnight at 4°C. Co-immunoprecipitation was performed using the Pierce Co-IP Kit (Thermo Fisher Scientific). Immunoprecipitates were washed, eluted, and separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis. Western blot was performed as described above.

Gene copy numbers were determined by quantitative polymerase chain reaction (PCR). Genomic DNA (gDNA) was extracted from cell lines using the Exgene Cell SV Kit (Geneall Biotechnology, Seoul, Korea). Quantitative PCR was used to determine the gene copy number as follows: gDNA was amplified with specific primers (S1 Table) and SYBR Green PCR Master Mix (Applied Biosystems, Waltham, MA) using the GeneAmp PCR System (Applied Biosystems); images were captured with the Gel Logic 200 Imaging System (Kodak, Rochester, NY). Gene copy numbers were extrapolated from an external standard curve generated by serial dilutions of human gDNA (Merck Millipore, Darmstadt, Germany). The PCR products were purified using the PCR Purification Kit (Invitrogen), and sequenced by bidirectional Sanger sequencing.

Total RNA was extracted from EBC-1 parental and resistant cell lines using the RNA Mini Kit (Invitrogen) and reverse transcribed using Superscript III Reverse Transcriptase (Invitrogen). The StepOnePlus Real-Time PCR System (Applied Biosystems) was used for amplification of cDNA and quantitative real-time reverse transcriptase polymerase chain reaction was performed using the SYBR Green PCR Master Mix (Applied Biosystems) with specific primers (S2 Table). Relative gene expression was calculated using the ΔΔCt (cycle threshold) method with β-actin as the reference gene. Sequencing of the 3′-mRNA was performed using Quantseq technology (Lexogen, Vienna, Austria). RNA was extracted from EBC-1 and resistant cell lines as described above. Hierarchical clustering was analyzed on log2-fold changes in gene expression using the MultiExperiment Viewer (Dana Farber Cancer Institute, Boston, MA). Normalized read counts were calculated as (log2) > 6, and genes with an adjusted p < 0.05 were considered significantly differentially expressed.

Data are presented as mean±standard deviation of independently performed experiments. Statistical analyses were evaluated using Sigma Plot 12 (Systat Software Inc., San Jose, CA). Student's t tests were used to indicate statistically significant differences (p < 0.05).

EBC-1 cells were derived from NSCLC cell lines harboring MET amplification and showing high sensitivity to MET-TKIs. To explore the molecular mechanisms of resistance to MET-TKIs, we used capmatinib, an MET inhibitor that is currently being studied in clinical trials. Using the EBC-1 cell line, we established total three cell lines resistant to capmatinib (EBC-CR1, -CR2, and -CR3) by stepwise exposure to capmatinib from 10 nM to final concentrations of 1.5, 2.2, and 2.4 μmol/L, respectively to observe various resistant mechanisms. The EBC-CR3 cell line was derived from the EBC-CR1 cell line by continuous exposure to a higher concentration of capmatinib during additional 3 months (S3A Fig.). First, we evaluated the sensitivity of the all three resistant cell lines to capmatinib by cell viability assay. The MET-TKIs were not toxic to the resistant cell lines (IC₅₀, > 10 μmol/L on EBC-CR1–3 cells and 3.70±0.10 nmol/L on EBC-1 cells; and sub-G1, 18.15±3.31% on EBC-CR1, 8.37±4.61% on EBC-CR2, 12.17±3.6% on EBC-CR3, and 61.22±4.78% on EBC-1 cells) (Fig. 1A, S3B Fig.). Next, we confirmed the decrease of MET gene copy number by qPCR in resistant cell lines, which was accompanied by downregulation of MET phosphorylation (Fig. 1B). These results suggested that capmatinib strongly affected not only kinase activity but also MET copy number, which indicated its potent efficacy against MET-dependent tumors. Also, there were no additional MET-activating mutations identified by Sanger sequencing (data not shown).

In resistant cell lines, complete inhibition of MET phosphorylation by capmatinib demonstrated that there were no MET mutations that prevented the binding of capmatinib. Although complete inhibition of MET kinase was observed, phosphorylation of AKT and ERK1/2 were not completely inhibited in the resistant cell lines by a single treatment with capmatinib (Fig. 1C). These results suggest that ‘main’ survival pathways in resistant cells change from MET signal to other alternative kinase pathway.

To identify alternative survival pathways of capmatinib resistance, EBC-1, EBC-CR1, -CR2, and -CR3 cells were screened for gene expression by 3′-mRNA sequencing. First, we analyzed expression of RTK-related genes by hierarchical clustering. RTK-related genes were stratified into several clusters and almost 20% of RTK-related genes were significantly upregulated in the resistant cell lines (Table 1, Fig. 1D). According to the results of 3' mRNA sequencing, MET mRNA expression was slightly decreased and EGFR mRNA expression was increased in EBC-CR1, -CR2, and -CR3 compared to the parental cell line (Table 1). We then confirmed increased EGFR protein phosphorylation compared to multiple RTKs in the resistant cell lines but observed no change in EGFR copy number (S1C and S3D Fig.). The mostly upregulated genes were KIT and EFEMP1 (Table 1, Fig. 1D). Although KIT mRNA was highly upregulated, KIT activation was not observed in phospho-RTK array (S3C Fig.). In contrast, EGFR signal was activated at both mRNA and protein levels. In addition, EFEMP1 that binds to EGFR can activate the downstream signals and eventually increase EGFR signal [16]. These results suggested that the EGFR signaling pathway might be an alternative pathway for MET. Although EGFR activation was increased in all resistant cell lines, cell viability assay to afatinib showed that only EBC-CR1 remained sensitive to afatinib alone (Fig. 2A). Interestingly, capmatinib inhibited EGFR kinase phosphorylation in EBC-1 cells. To determine the mechanism of EGFR pathway activation in the EBC-CR1 cell line, we tested the expression of multiple EGFR ligands (including epidermal growth factor [EGF], amphiregulin, epiregulin, transforming growth factor-α, and HBEGF) and observed that HBEGF expression was elevated in EBC-CR1 cells (Fig. 2B, S4A Fig.). To further explore the role of HBEGF in EGFR activation, we added exogenous HBEGF to the EBC-1 and EBC-CR1 cell lines and found that HBEGF was sufficient to rescue to MET inhibition in the MET-dependent NSCLC cell line, EBC-1 (S3E Fig.). In EBC-CR1, cultures with HBEGF were more resistant to EGFR inhibition than those without ligand (p < 0.01) (S3F Fig.). Unlike the parental cell line, a single treatment with afatinib effectively inhibited phosphorylation of EGFR, AKT, and ERK1/2 despite persistent MET kinase activation (Fig. 2C). As a result, EGFR activation was highly dependent on HBEGF as the main survival signal in EBC-CR1 cells.

In contrast to EBC-CR1 cells, in EBC-CR2 cells, the combined blockade of MET and EGFR kinase significantly abolished phosphorylation of the downstream signaling molecules, AKT and ERK1/2 (Fig. 3A). While activation of both MET and EGFR kinase was diminished in the parental cell line, EGFR phosphorylation was increased in EBC-CR2 cells by capmatinib treatment and MET phosphorylation was increased by afatinib. Intriguingly, EBC-CR2 cells displayed more dominant MET-EGFR heterodimers than the parental cell line (Fig. 3B). To determine the efficacy of combined treatment with capmatinib and afatinib, we performed the cell viability assay to compare the effect of capmatinib, afatinib, and their combination in EBC-CR2 cells. We observed that the IC₅₀ was lower in the cells treated with the combination compared to either compound alone, which suggested that capmatinib and afatinib synergistically inhibited cell proliferation (Fig. 3C). These results indicated that dual blockade of MET and EGFR obstructed the MET-mediated EGFR survival signal caused by MET and EGFR heterodimerization. Accordingly, the MET-EGFR heterodimer is a much more dominant form of activated EGFR kinase, which is the main survival signal in EBC-CR2 cells.

Despite continuous exposure of the EBC-CR3 cell line (derived from EBC-CR1 cells) to capmatinib but not EGFR-TKI resulted in the loss of sensitivity to afatinib in the cell viability assay (Fig. 4A). In Western blot, EBC-CR3 showed a complete block of EGFR or MET activation, but persistent AKT phosphorylation after afatinib treatment (S5B Fig.). We performed qRT PCR and Sanger sequencing to investigate the mechanism underlying EGFR-dependent resistance to afatinib. We found that PIK3CA mRNA expression was increased and gene copy number was amplified in EBC-CR3 cells, which was unlike what we observed in the other cell lines (Fig. 4B, S5C Fig.); however, we did not find any hot spot mutations in PIK3CA (data not shown). PIK3CA amplification was observed frequently and might play a pathogenic role in lung cancer, especially squamous cell carcinoma [17]; thus we hypothesized that PIK3CA amplification is acquired resistant mechanism to capmatinib. To confirm this hypothesis, we tested the efficacy of BYL719 on PIK3CA inhibition in EBC-CR3 cells, because this molecular target of BYL719 had been reported previously. [18] As a result, we showed resistance after a single treatment with BYL719; however, the cells with concurrent EGFR overexpression were sensitive to BYL719 in the presence of afatinib in the cell viability assay (IC₅₀ for BYL719, 140.84±19.65 nmol/L) (Fig. 4A). Although EBC-1 and EBC-CR2 cells were resistant to combined treatment of BYL719 and afatinib, it showed additive effect on EBC-CR1 cells (S6A Fig.). In addition, the combination of BYL719 with afatinib effectively inhibited phosphorylation of AKT, despite the persistence of the MET signal (Fig. 4C), and inactivation of RAS/MAPK pathway (data not shown). The combined treatment incompletely inhibited AKT phosphorylation on EBC-1 and -CR2 (S6B Fig.). Although one or more colony might be formed in EBCCR3 cells, there was no other genetic modification in other colonies except PIK3CA amplification. These results demonstrated that wild-type PIK3CA amplification could be an additional mechanism of resistance to EGFR-TKI in capmatinib-resistant cancers; however, the combined therapy effectively overcame the resistance to TKIs in a MET-independent manner in the presence of additional genetic alterations.