Specific Recurrent CNAs Identified in Tumor Samples from Patients with Acquired Resistance vs. Those from Untreated Resected EGFR Mutant Tumors.

We next compared results from tumors with acquired resistance to those obtained from a separate aCGH analysis of 38 mutant EGFR lung adenocarcinomas resected from patients who had never received treatment with kinase inhibitors. DNA from the untreated tumors was analyzed by using 44K Agilent chips (16). The recurrent genomic gains and losses in these samples appeared grossly similar to the acquired resistance set (Fig. 1 Lower).

After mode-centering, comparison of the two sets (at the location of each of the 44K probes; see Materials and Methods) revealed three major loci of recurrent CNAs unique to samples from patients with acquired resistance (Fig. 1 Upper and Table 1). One locus, at 7p11-12, includes EGFR and was amplified compared with the untreated set in 3 of the 12 tumor samples (nos. 5, 6, and 10a). These results are consistent with the notion that EGFR mutation and amplification occur frequently in tumors from patients with acquired resistance (11). A second locus, at an interval encompassing 7q31.2, was found in two samples (nos. 6 and 10a) [supporting information (SI) Fig. 4]. The gene encoding MET lies in this interval and encodes a receptor tyrosine kinase implicated in the development, maintenance, and progression of cancers in both animals and humans (reviewed in ref. 17). The third CNA occurred on 5p15.2–15.3 and was found in two samples (nos. 5 and 10a); candidate genes in this region remain to be identified (Table 1). We did not observe any genomic deletions that were significantly overrepresented in either treated or untreated groups.

Genomic Gains on Chromosome 7 in Tumor Cells from Patients with Acquired Resistance.

We next examined the individual aCGH profiles of the region of interest on chromosome 7 at higher resolution in all of the samples. Eleven of the samples showed broad gains of chromosome 7, including the region of EGFR (data not shown). Samples 5, 6, and 10a displayed further focal amplification at the locus encompassing EGFR, and samples 6 and 10a had additional focal amplifications at the locus, including MET (SI Fig. 4). None of these samples had focal amplification of the gene encoding hepatocyte growth factor (HGF), the ligand for MET, located at 7q21.1. An additional tumor sample from patient number 10 (10b, a metastatic lymph node) also displayed focal amplifications at both EGFR and MET.

To determine the proportion of drug-resistant tumor cells with amplified MET, we assessed MET gene copy number per cell by dual-color fluorescent in situ hybridization (FISH) in the one tumor sample (no. 6) for which we had sufficient viable tumor cells for analysis. Cells were labeled with probes that hybridized to the centromere of chromosome 7 (CEP7; green) or to MET (red). We found that the tumor sample comprised a mixed population of cells. All were polysomic for chromosome 7. However, although some cells displayed equal numbers of copies of CEP7 and MET (averaging 4–6 copies), others harbored greater numbers of copies of MET than CEP7 (SI Fig. 5). Taken together with the aCGH results on the same tumor sample, these data suggest that tumor cells in patient no. 6 have ≈4–6 copies of chromosome 7 with additional focal amplifications of the region containing MET in approximately half of the cells.

Confirmation of MET Status by Quantitative PCR and Analysis of Additional Patient Tumor Samples.

To confirm further the results from the aCGH studies, we next performed quantitative “real-time” PCR (qPCR) to determine the status of MET in DNA samples from four independent cohorts of tumor samples. As a control gene, we selected one [(MTHFR (5,10-methylenetetrahydrofolate reductase; located at 1p36.3)] from a genomic region that showed no CNA in any sample by aCGH and is not subject to germ-line copy number polymorphism.

In the first cohort—the tumor samples already analyzed by aCGH—quantitative PCR results confirmed the results (Table 2, nos. 1–12). We then tested four additional tumors from our own patients with acquired resistance (Table 2, nos. 13–16) and three additional drug-resistant tumors from Taiwan (Table 2, nos. 17–19). In total, in these tumor samples from 19 patients, four displayed MET amplification (e.g., a fold change relative to MTHFR > 1.5). Here, we chose a ratio of MET:MTHFR > 1.5 to define MET amplification based on corresponding data from FISH and aCGH analysis of tumor cells from patient no. 6 (SI Fig. 5) and of cell lines (see below and Table 2).

To extend these findings to another independent cohort, we performed qPCR analysis on DNA from EGFR mutant tumors obtained from 24 Taiwanese patients with acquired resistance to gefitinib (SI Table 3, nos. 20–43). Matched pretreatment tumor DNA samples were available for comparison. We detected MET amplification in tumors from five patients. In four of these samples, MET amplification was found in only posttreatment samples, suggesting that selection for cells with MET amplification occurred while these patients were on gefitinib. Note that here, we used a more stringent criterion for MET amplification (MET:MTHFR ratio > 5) because the samples were tested with an independent protocol (see Materials and Methods and SI Table 3 legend) for which concurrent aCGH or FISH data were unavailable.

In two patients (SI Table 3, nos. 30 and 32), MET amplification was detected in tumor samples obtained before treatment. In one of these patients (no. 32), the untreated sample was a surgically resected lung tumor. Disease recurred more than three years later, and the posttreatment specimen was derived from omentum 15 months after an initial response to gefitinib. The other patient (no. 30) was diagnosed with a CT-guided lung biopsy and had a confirmed partial radiographic response on gefitinib. The “acquired resistance” specimen was obtained when pleural fluid developed 8 months after starting gefitinib. These results could be due to genetic heterogeneity within individual tumors and/or tumor heterogeneity within individual patients. The observed amplification of MET in some tumors after treatment with TKIs could be attributed to selection of subpopulations of cells with MET amplification.

When data (using qPCR, aCGH, and/or FISH) from the multiple independent cohorts were combined, we found MET amplification in 9 of 43 patients with acquired resistance, compared with 2 of 62 untreated patients (P = 0.007, Fisher's Exact test). The common EGFR^T790M resistance mutation was found in 20 of 43 (46.5%) patients with acquired resistance. Interestingly, 4 of the 10 tumor samples (from nine patients) with MET amplification harbored the EGFR^T790M mutation as well. Thus, tumors with acquired resistance to gefitinib or erlotinib may exhibit amplification of MET in the absence or presence of a second-site drug-resistant mutation in EGFR. There was no correlation between increased copies of MET and type of primary drug-sensitive EGFR mutation (exon 19 deletion vs. exon 21 point mutation) or duration of drug treatment (data not shown).

An Established Lung Adenocarcinoma Cell Line Contains an Exon 19 Deletion Associated with Drug-Sensitivity, an Exon 20 Point Mutation Associated with Drug-Resistance, and Increased Copies of MET.

We performed qPCR analysis of MET in lung adenocarcinoma cell lines with EGFR mutations. Surprisingly, we found that one cell line—H820—contained not only drug-sensitive (del E746-E749) and drug-resistant (T790M) EGFR kinase domain mutations (data not shown) but also MET amplification (Table 2). Consistent with these results, we found by using FISH that the majority of H820 cells harbored 4–6 copies of chromosome 7 (CEP7) and 7–9 copies of MET (Fig. 2A). As judged by aCGH, large regions of chromosome 7 were amplified to varying degrees but no specific focal amplifications were observed at loci containing either EGFR or MET (data not shown).

Open in a separate window

Fig. 2.

MET status and sensitivity of H820 cells to the MET inhibitor, XL880. (A) H820 cells display, as assessed by dual-color FISH analysis (see Materials and Methods), ≈4–6 copies of chromosome 7 (CEP7), with additional focal amplifications at loci containing MET. The cells shown are representative of the whole population. Scale bar: 10-μm. CEP7 probe, green; MET probe, red; nuclei, blue (DAPI). (B) Lysates from H820 and PC-9 cells were immunoblotted by using anti-phospho-(p)-MET (Y^1234/5), anti-total MET, and anti-actin antibodies. H820 cells harbor an exon 19 deletion (E746-E749), a T790M mutation, and amplified no MET, whereas PC-9 cells harbor an exon 19 deletion (E746-A750), no T790M mutation, and no MET amplification. (C) Growth inhibition curves of H820 and PC-9 cells treated with MET and EGFR inhibitors at various concentrations. Error bars indicate one standard deviation from six replicates.

To determine whether the high number of copies of MET caused increased MET enzymatic activity, we used a surrogate kinase assay in which immunoblots of H820 cell lysates were probed with polyclonal antibodies that recognize MET phosphorylated at tyrosine residues (Y^1234/5). These phosphorylation sites are located in the activation loop and are indicative of kinase activation (Fig. 2B). Total MET protein was measured by immunoblotting with an anti-MET monoclonal antibody. Both antibodies recognize the 170-kDa precursor form of the enzyme and the mature 140-kDa beta subunit. The smaller protein is the product of proteolytic cleavage of the larger precursor and also contains the tyrosine kinase catalytic domain (18). We compared results with H820 cells to those with PC-9 cells, a lung cancer line that harbors an exon 19 deletion (E746-A750) of EGFR but no T790M mutation and no MET amplification (Table 2). Extracts from PC-9 cells contained relatively small amounts of total and phosphorylated MET protein. By contrast, H820 cell extracts displayed much greater amounts of MET protein, much of which was phosphorylated at Y^1234/5 (Fig. 2B).

Sensitivity of H820 Cells to a MET Inhibitor but Not to EGFR Inhibitors.

To explore further the functional consequences of MET amplification in EGFR mutant tumor cells, we used fluorescence-based growth inhibition assays to measure the sensitivity of H820 cells to erlotinib, CL-387,785 [an irreversible EGFR inhibitor known to overcome T790M-mediated resistance (10, 19, 20)], and XL880. The third compound abolishes MET and VEGFR2 tyrosine kinase activity at subnanomolar levels and also inhibits the PDGFRβ, KIT, FLT3, TIE-2, and RON kinases in vitro with low nM potency (21, 22).

H820 cells were relatively insensitive to erlotinib, because the concentration of drug required to inhibit the growth of 50% of the cells (IC₅₀) was ≈10 micromolar (Fig. 2C Top). The IC₅₀ of CL-387,785 was also in the micromolar range (Fig. 2C Top and Middle). By contrast, H820 cells displayed sensitivity to XL880 in the nanomolar range (Fig. 2C Middle). At these concentrations of XL880, the kinase activity of MET was inhibited, as measured by examination of H820 cell lysates from XL880-treated cells using the surrogate kinase assays described above (Fig. 3Left). By comparison, PC-9 cells were insensitive to XL880 but were much more sensitive to CL-387,785 (Fig. 2C Bottom). These results suggest that MET inhibitors like XL880 may be able to overcome MET-mediated resistance to EGFR kinase inhibitors, even in cells that harbor the T790M mutation.

Open in a separate window

Fig. 3.

ERBB3 signaling in H820 cells depends on MET status. Lysates from H820 and PC-9 cells, either untreated (DMSO-only control; 0 μM) or treated with 0.3 μM XL880 or 0.3 μM CL-387,785, were immunoblotted by using anti-phospho-(p)-MET (Y^1234/5), anti-total MET, anti-phospho-(p)-ERBB3 (Y¹²⁸⁹), anti-total ERBB3, anti-phospho-(p)-EGFR (Y¹⁰⁹²), anti-total EGFR, and anti-actin antibodies.

Dependence of ERBB3 Signaling on MET in H820 Cells.

During the course of the work described here, others reported that MET amplification leads to gefitinib resistance in lung cancer by activating ERBB3 signaling (23). ERBB3 is a heterodimeric partner for EGFR that mediates phosphoinositide 3-kinase activity in gefitinib-sensitive lung cancer cell lines (24). We found that XL880 treatment strongly inhibited the phosphorylation of ERBB3 in H820 cells but not in PC-9 cells. Conversely, treatment with CL-387,785 strongly inhibited phosphorylation of ERBB3 in PC-9 cells but not in H820 cells (Fig. 3).

To determine whether the inhibition of ERBB3 phosphorylation in H820 cells treated with XL880 was due specifically to inhibition of the MET kinase, we transfected MET-specific siRNAs into H820 and PC-9 cells to knockdown expression of the protein. After transfection with MET siRNAs, both phospho-MET and total-MET were virtually undetectable by immunoblotting of H820 cell extracts (SI Fig. 6). In concordance with the results obtained with XL880, phospho-ERBB3 levels were also reduced in H820 cells after treatment with the MET siRNAs. By contrast, we observed no effect on phospho-ERBB3 status in similarly treated PC-9 cells and no effect on MET or ERBB3 status in cells treated with siRNAs against a control gene (GAPD) (SI Fig. 6). Thus, ERBB3 signaling appears to depend on MET protein in H820 cells, even though these cells harbor EGFR mutations.

We also measured the effect of siRNA knockdown of MET on the growth of H820 cells. In multiple experiments, the number of viable cells remaining 72 h after treatment with MET siRNAs was reduced, whereas treatment with GAPD siRNA had no effect. However, the effect of MET siRNA was modest (between 80 and 90% of controls in both total cell count and growth inhibition assays; data not shown), less than that seen with the kinase inhibitor, XL880. Thus, in H820 cells, XL880 may inhibit other kinases in addition to MET that affect cell viability. Alternatively, the effect of kinase inhibition by XL880 is different from the effect of siRNA-mediated knockdown of MET, because the latter probably diminishes MET activity more slowly.

Mutational Analyses.

Genomic DNA was extracted from tumor specimens, and primers for EGFR (exons 18–24) analyses were as published in refs. 4 and 35. PCR-RFLP assays for exon 19 deletions and L858R and T790M missense mutations were performed as published in refs. 9 and 36. All mutations were confirmed at least twice from independent PCR isolates, and sequence tracings were reviewed in the forward and reverse directions by visual inspection. Primers for MET sequencing are listed in SI Table 4.

aCGH Profiling.

Genomic DNA was extracted from tumor samples, using standard techniques. Normal genomic DNA (Promega) was used as a reference for all samples. DNA was digested and labeled by random priming using Bioprime reagents (Invitrogen) and Cy3- or Cy5-dUTP. Labeled DNA was hybridized to Agilent 244K CGH arrays for acquired resistance samples, and 44K CGH arrays for untreated samples. For additional details, see SI Methods.

Quantitative Real-Time PCR.

See SI Methods for details.

Cell Lines and Viability Assays.

NCI-H820 cells were developed by A. Gazdar (University of Texas Southwestern Medical Center). PC-9 cells were a gift from M. Ono (Kyushu University, Fukuoka, Japan). Cells were grown in RPMI supplemented with FBS. Growth inhibition assays were performed with the CellTiter-Blue cell viability kit (Promega), per the manufacturer's instructions. All assays with H820 cells were performed at least three independent times; for PC-9 cells, all assays were performed at least two independent times (see SI Methods). The complete aCGH dataset is available at http://cbio.mskcc.org/Public/Bean_etal_PNAS_2007.

Nomenclature.

Two numbering systems are used for EGFR. The first denotes the initiating methionine in the signal sequence as amino acid −24. The second, used here, denotes the methionine as amino acid +1. There are also two mRNA transcripts for MET that differ by an in-frame deletion of 54 nt (28, 37); because the isoform lacking these 54 nt appears to be the most abundant isoform (38), we consider full-length MET protein to consist of 1,390 (not 1,408) aa.

Immunoblotting.

See methods and supporting text in ref. 4 for details on cell lysis and immunoblotting reagents. At least three independent experiments were performed for all analyses. See SI Methods for a list of the antibodies used.

FISH.

FISH was performed according to established protocols (see SI Methods).

We thank K. Politi, C. Sawyers, and H.E.V. for helpful discussions; T. Mitsudomi (Aichi Cancer Center Hospital, Nagoya, Japan) for providing DNA samples; Z. Zeng and M. Weiser (both of Memorial Sloan–Kettering Cancer Center) for providing the MET BAC probe for FISH; Exelixis for providing XL880; and all of the patients and/or their family members who consented for tissue acquisition. This work was supported by University of Texas Specialized Programs of Research Excellence in Lung Cancer National Cancer Institute Grants P50CA75907 (to A.G.), U01-CA84999 (to W.G.), and R21-CA115051 (to V.M.); Taiwan National Science Council Grants NSC95-2314-B-002-113-MY3 (to J.-Y.S.) and NSC95-2314-B-002-227-MY3 (to C.H.Y.); Chang-Gung Medical Research Fund Grant CMRPG 350031 (to W.-C.C.); National Health Research Institutes Grant 96-A1-MG-PP-04-014 (to S.F.H.); the Doris Duke Charitable Foundation (W.P.), the Thomas G. Labrecque Foundation (W.G. and W.P.), Joan's Legacy: The Joan Scarangello Foundation to Conquer Lung Cancer (W.P.); National Insitutes of Health Grants K08-CA097980 and R01-CA121210 (to W.P.); the Jodi Spiegel Fisher Cancer Foundation (W.P.); the Carmel Hill Fund (W.P.); and funds from the Miner Family.