MET Copy Number as a Secondary Driver of Epidermal Growth Factor Receptor Tyrosine Kinase Inhibitor Resistance in EGFR-Mutant Non–Small-Cell Lung Cancer

In a subset of patients with non–small-cell lung cancer (NSCLC), increases in MET copy number have been hypothesized to lead to excessive amounts of MET protein, autoaggregation, ligand-independent MET signaling, and oncogenic addiction to the MET pathway.¹ The same process has also been described as one mechanism of acquired resistance to epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs) in EGFR-mutant NSCLC.^2-4 Changes in MET copy number represent a continuous variable that can be assessed in different ways: either the ratio of MET relative to another region of chromosome 7, such as CEP7 (to show true amplification), or the absolute number of MET copies, which can also be increased by polysomy.¹ Consequently, the persistent problem with MET copy number studies has been to define a threshold for any given methodology above which a MET-directed therapy will likely be active. Less stringent criteria may include more patients but could dilute the clinical benefit in the treated population. More stringent criteria will identify fewer patients: those who may derive the greatest clinical benefit, but potentially excluding others who will still derive some benefit.

In the article that accompanies this editorial, Lai et al⁵ conducted MET fluorescence in situ hybridization (FISH) analysis on 200 metastatic EGFR-mutant NSCLC samples. Using a mean MET/cell value of more than five to define MET high and a mean MET:CEP7 ratio of 2.0 or greater to define MET amplified, 26% of treatment-naïve patients were able to be defined as MET high, but only 3% exhibited amplification. Among 154 patients subsequently treated with EGFR TKI, the baseline MET-high group at either greater than 5 or greater than 6.8 copies showed no differences in median time to treatment failure (TTF), the proportion of patients with a TTF of 6 months or fewer, or objective response rate (ORR) compared with the MET-low group. In contrast to the MET-high group, in whom median TTF was 12.5 months, median TTF among the five EGFR-treated MET-amplified patients was only 5 months. The two patient cases with the highest MET:CEP7 ratio—3.7 and 4.5—both had progression as their best response and TTFs of 1.0 and 0.5 months, respectively.

MET copy number gain has been described as the mechanism of acquired resistance in 5% of EGFR-mutant cases post–first- and –second-generation EGFR TKIs and approximately 20% of cases post-osimertinib.^2-4 As treated patients in the work by Lai et al⁵ had only received first- and second-generation TKIs, a 3% frequency of clinically relevant MET amplification seems plausible. Data from their use of FISH in pretreatment specimens also offers the promise of predefining a population that might benefit most from upfront combined MET and EGFR inhibition. Without preselection, demonstrating efficacy from a first-line MET-EGFR combination when only a minority of patients may be sensitive to this approach has been challenging. For example, the combination of emibetuzumab, a monoclonal antibody directed against surface MET expression, and erlotinib showed no difference compared with erlotinib alone in treatment-naïve EGFR mutant NSCLC (median PFS, 9.3 months and 9.5 months, respectively; HR, 0.89; 90% CI, 0.64 to 1.23; P = .534).⁶

Results presented by Lai et al⁵ are also consistent with data that examined MET copy number as a primary oncogenic driver in which increases in mean MET/cell, at least up to seven copies, were unlikely to reflect a MET-dependent state and estimates of true amplification were a potentially better determinant of MET dependency.⁷ Specifically, among previously untreated adenocarcinoma cases screened using MET FISH, a coincident, alternative driver oncogene was commonly identified across all categories of mean MET/cell assessed.⁷ Even in the highest mean MET/cell category of seven or more copies, oncogene overlap occurred in 40% of cases. Similarly, oncogene overlap was identifiable in approximately 50% of low (ratio, 1.8 or greater to 2.2 or less) and intermediate (ratio, greater than 2.2 to less than 5.0) MET:CEP7 categories. In contrast, in the highest MET:CEP7 category (ratio, 5 or greater), no oncogene overlap was observed, which is more consistent with this level of amplified MET acting as a true oncogenic driver. Although only four such cases were available for analysis, this also represents the same MET:CEP7 category with the highest reported ORR (67%; four of six) to crizotinib to date.⁸

In EGFR-mutant NSCLC, MET amplification as a mechanism of acquired resistance to EGFR TKIs is believed to represent the selection of a subclone, rather than a truncal genetic event. Indeed, overt clinical waxing and waning of MET copy number in response to EGFR TKI selection pressure can be observed.⁹ For this reason, exact copy number levels that are suggestive of MET dependence may seem to be lower in the established EGFR acquired resistance setting compared with in the primary driver setting as a result of the potential dilutional impact of non–MET-driven cells in any analyzed sample. In the treatment-naïve setting, this dilution should be even more pronounced. Prior studies have shown that EGFR-mutant NSCLC cell lines and clinical samples that later manifest MET amplification in response to EGFR TKIs do have detectable preexisting MET amplification, but only in less than 1% of cells.¹⁰ Consequently, the observation by Lai et al⁵ that baseline MET:CEP7 ratios of 2.0 or more detected with routine FISH testing, which averages signals more than 50 to 100 cells, can still be linked to different TKI outcomes is unexpected and deserves additional study.

Lai et al⁵ performed detailed single-cell FISH analysis in 10 separate tumor areas that revealed significant spatial heterogeneity in both MET/cell and MET:CEP7 positivity. Nevertheless, although the exact degree of heterogeneity was not quantified, from the Data Supplement in Lai et al,⁵ spatial consistency did seem to be greater for MET amplified than MET polysomy-defined positive cells. Whereas these data continue to support the MET:CEP7 ratio as the more robust measure of the cancer’s biologic predisposition, concern still exists about the reliability of standard techniques for identifying MET amplification in treatment-naïve EGFR-mutant cases.

When considering MET copy number alterations that may represent only a subpopulation of tumor cells in EGFR-mutant NSCLC, the reliability of noncellular level assessments of MET is of even greater concern. Next-generation sequencing (NGS) techniques analyze DNA pooled across multiple areas of a tumor together with nontumor DNA. Between two MET-amplified and 16 MET-polysomy cases, only eight were deemed to have MET copy number gain (range, six to 22) by NGS (Oncomine; Thermo Fisher Scientific, Waltham, MA). Of the two patients with MET-amplified disease, the case with the lower ratio (MET:CEP7 = 2.0) was not identified by NGS, whereas a MET copy number of 13.32 was reported for the higher ratio case (MET:CEP7 = 3.4). These concerns should be even greater when NGS is applied to circulating DNA.

Compared with the treatment-naïve setting, MET–EGFR inhibitor combinations are more commonly being explored in variably preselected populations of EGFR-mutant NSCLC after acquired resistance to EGFR TKIs. ORR with savolitinib and osimertinib was 28% (seven of 25 patients) among those with EGFR-mutant NSCLC who experienced progression on a prior third-generation EGFR TKI with centrally FISH confirmed MET amplification (defined as five or more copies of MET/cell or a MET/CEP7 ratio of 2 or greater).¹¹ Unfortunately, no details on the exact cytogenetic values recorded in responders versus nonresponders have been shown to date. ORR with capmatinib and erlotinib was 47% (17 of 36 patients) among those with EGFR-mutant NSCLC who experienced progression on an EGFR TKI with a FISH-defined mean MET/cell of 6 or greater.¹² To some extent, this seems to be at odds with the prior data that favored MET:CEP7 over comparable mean MET/cell levels to identify a MET-driven state.^5,7,8 However, again, the exact cytogenetic values—mean MET/cell and MET:CEP7—observed in responders and nonresponders have not been shown to date.

Perhaps, at some high levels or in some settings in which the pretest probability is increased—as in established EGFR TKI acquired resistance—absolute copy number may be as valid a predictor of MET dependency/codependency as the MET:CEP7 ratio. Alternatively, some other relevant heterogeneity within the MET copy number–altered NSCLC population could be confounding the efficacy signals in these small data sets. Even in MET exon 14–mutated NSCLC, which, in theory, leads to primary oncogenic MET signaling in a manner that is comparable to MET copy number gain—increasing MET protein levels, but by altered ubiquitination, rather than by increased gene dosage—ORR to MET TKIs is consistently only approximately 30% to 40%.^13,14 Similarly, in the expanded crizotinib data set looking at NSCLC with primary MET:CEP7 ratios of 4 or greater, ORR was also 40% (eight of 20 cases).¹⁵ These ORRs seem to be significantly lower than with other actionable driver oncogenes. The next steps may therefore involve validating the assumptions that these genetic events actually lead to increased MET protein expression and pathway activation. Whereas MET protein expression alone is not a reliable indicator of a MET-driven state post–EGFR TKI, MET protein levels in the presence of increased MET copy number or other estimates of downstream MET activation—for example, using a MET:GRB2 proximity ligation assay—have the potential to increase the predictive value of any MET genetic alteration.^16,17 It is only by delving deeply into the exact techniques and values used to call MET copy number–driven states (or even MET exon 14–driven states) within individual patients and trials that we will be able to better define the optimal patient population for MET-directed therapy in NSCLC.

AUTHOR CONTRIBUTIONS

Conception and design: All authors

Data analysis and interpretation: D. Ross Camidge

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

MET Copy Number as a Secondary Driver of Epidermal Growth Factor Receptor Tyrosine Kinase Inhibitor Resistance in EGFR-Mutant Non–Small-Cell Lung Cancer

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/site/ifc.

D. Ross Camidge

Honoraria: Roche, G1 Therapeutics, Mersana, Takeda, AstraZeneca, Genoptix, Ignyta, Daiichi Sankyo, Hansoh, Lycera, Biothera, Revolution Medicines

Research Funding: Takeda

No other potential conflicts of interest were reported.