MET Alterations Are a Recurring and Actionable Resistance Mechanism in ALK-Positive Lung Cancer

This article is featured in Highlights of This Issue, p. 2437

ALK-rearranged (i.e., ALK-positive) non–small cell lung cancers (NSCLC) depend on constitutive ALK signaling for growth and survival (1). This oncogene dependency underlies the marked sensitivity of these tumors to tyrosine kinase inhibitors (TKI) targeting ALK (2, 3). Historically, the ALK/ROS1/MET TKI crizotinib was the standard first-line therapy for advanced ALK-positive NSCLC (4). However, crizotinib has recently been supplanted by more potent and selective second-generation ALK TKIs (e.g., ceritinib, alectinib, and brigatinib; refs. 2, 3, 5). In addition, the third-generation ALK TKI lorlatinib was recently approved for patients previously treated with a second-generation TKI (6).

Despite initial sensitivity to ALK TKIs, ALK-positive tumors invariably develop resistance. In approximately one-half of cases, resistance to second-generation ALK TKIs is due to secondary mutations in the ALK kinase domain (7). These tumors remain ALK-dependent and responsive to treatment with lorlatinib (8). The remaining cases have presumably developed ALK-independent resistance mechanisms, most often due to activation of alternative or bypass signaling pathways (7, 9). Recent studies of lorlatinib resistance suggest an even more complicated array of resistance mechanisms, including diverse compound ALK mutations in one-third of patients and ALK-independent resistance mechanisms in the remaining two-thirds of patients (9, 10).

To date, ALK-independent resistance mechanisms remain poorly characterized. In preclinical studies, multiple bypass mechanisms have been described, including activation of MET, EGFR, SRC, and IGF-1R (11–13). Of these, MET is a particularly attractive target given the availability of potent MET TKIs. In addition, MET activation has been previously studied as a bypass pathway in EGFR-mutant NSCLC where MET amplification has been reported in 5%–20% of resistant cases (14, 15). In phase I/II studies, cotargeting EGFR and MET can reinduce responses in resistant EGFR-mutant tumors with MET amplification (16). In contrast to EGFR-mutant NSCLC, relatively little is known about the contribution of aberrant MET activation to resistance in ALK-positive NSCLC. Several case reports suggest that MET amplification can mediate resistance to ALK TKIs and that the ALK/ROS1/MET TKI crizotinib may be able to overcome MET-driven resistance (17, 18). However, larger studies are needed to validate MET as a clinically relevant driver of resistance in ALK-positive NSCLC.

Here we analyze tumor and/or blood specimens from 136 patients with ALK-positive NSCLC relapsing on ALK TKIs. We identify genetic alterations of MET in approximately 15% of resistant cases. In cell lines and in patients, MET activation drives resistance to ALK TKIs but confers sensitivity to dual ALK/MET blockade. These studies support the development of novel combinations targeting ALK and MET in ALK-positive NSCLC.

Resistant specimens were obtained from 136 patients with metastatic ALK-positive NSCLC who underwent molecular profiling between 2014 and 2019 (Fig. 1A). Medical records from these patients were retrospectively reviewed to extract data on demographics, treatment histories, and molecular profiling results. Data were updated as of August 31, 2019. This study was approved by the Massachusetts General Hospital Institutional Review Board. Patient studies were conducted according to the Declaration of Helsinki, the Belmont Report, and the U.S. Common Rule.

MET copy number was quantified in tissue using FISH or the FoundationOne assay as described previously (19). For cases assessed by FISH, amplification was defined as a ratio of MET to centromere 7 (MET/CEP7) of ≥2.2 (20). Cases with MET/CEP7 ≥ 4 were classified as high-level amplification (20). For tumors genotyped with the FoundationOne assay, MET amplification was defined as ≥6 gene copies (19). With the Guardant360 assay, MET amplification was defined as absolute plasma copy number ≥2.1 (21). As MET copy number analysis based on Massachusetts General Hospital (MGH) SNaPshot next-generation sequencing (NGS) has not been formally validated, this assay was not used to detect MET amplification. MET rearrangements and mutations and ALK mutations were detected in tissue using either FoundationOne or the MGH SNaPshot/Solid Fusion Assays (19, 22). Guardant360 was used to identify MET and ALK mutations in plasma, but is not designed to detect MET rearrangements (21). All patients included in this study provided written informed consent for molecular analysis.

Using methods described previously (13), we generated cell lines from pleural fluid collected from MGH915 prior to lorlatinib (MGH915-3) and from a xenograft of pleural fluid at relapse on lorlatinib (MGH915-4). The cell lines were sequenced to confirm the presence of ALK rearrangement, MET rearrangement, and MET amplification. The ST7-MET fusion transcript was confirmed by sequencing a 700 bp region spanning the fusion breakpoint in cDNA generated from RNA extracted from MGH915-4. The corresponding ST7-MET fusion sequence was constructed by fusing ST7 exon 1 (NM_021908.3) to exons 2-21 of MET (NM_000245.4) which includes the full-length MET coding sequence. This resulting fusion sequence incorporates the ST7 ATG start codon corresponding to position 41 exon 1 followed by 151 bp of ST7 exon 1 and 14 bp of MET exon 2 5′ to the wild-type MET ATG start codon. Wild-type MET sequence was obtained from an open reading frame cDNA clone (GenScript OHu28578, NM_001127500.2). The sequence was cloned into pEN TTmcs (Addgene #25755), recombined with pSLIK-Neo (Addgene #25735) by using Gateway LR Clonase Enzyme (Invitrogen), and lentiviral particles were generated in HEK293FT cells by transfection with ViraPower packaging Mix (Invitrogen) following the manufacturers' instructions. H3122 cells were transduced by lentiviral infection and selected with G418 (Gibco) for 5 days. The expression of ST7-MET was induced by doxycycline followed by addition of lorlatinib to select the ST7-MET–expressing resistant population. After selection, cells were maintained with continuous culture in 1 μg/mL of doxycycline and 300 nmol/L of lorlatinib. For drug sensitivity assays using DOX-controls, doxycycline was removed 3 days prior to the experiment to allow for ST7-MET expression levels to return to baseline.

Cells (2,000) were plated in triplicate into 96-well plates. Viability was determined by CellTiter-Glo (Promega) 3 days after drug treatment in drug sensitivity assays. Luminescence was measured with a SpectraMax M5 Multi-Mode Microplate Reader (Molecular Devices). GraphPad Prism (GraphPad Software) was used to analyze data. For Western blotting, a total of 0.25–2 × 10⁶ cells was treated in 6-well plates for 6 hours. Equal amounts of total cell lysate were processed for immunoblotting.

Whole-genome sequencing (WGS) was performed as described in the Supplementary Methods.

Fisher exact test was used to compare MET amplification frequency between treatment groups. All P values were based on a two-sided hypothesis and computed using Stata 12.1.

To survey the landscape of molecular alterations associated with resistance to ALK TKIs, we analyzed 207 tissue (n = 101) and plasma (n = 106) specimens from 136 patients with ALK-positive NSCLC relapsing on ALK TKIs (Fig. 1A). Four assays were used for the analysis (see Materials and Methods), including two tissue-based NGS assays (FoundationOne and MGH SNaPshot/Solid Fusion Assay; refs. 19, 22), a plasma NGS assay (Guardant360; ref. 21), and FISH for MET amplification (Supplementary Table S1).

The clinicopathologic characteristics of the study patients were consistent with an ALK-positive NSCLC population (Supplementary Table S2). Of the 207 specimens, 12 (6%) were from patients relapsing on crizotinib who had not been exposed to other ALK TKIs. The majority of specimens (n = 136/207, 66%) were obtained at progression on a second-generation ALK TKI. The remaining 59 (29%) specimens were collected from patients relapsing on lorlatinib. Overall, ALK kinase domain mutations were detected in 34 (47%) of 73 tissue biopsies genotyped by NGS, including 25/46 (54%) specimens obtained after second-generation TKIs and 9/25 (36%) post-lorlatinib biopsies (Supplementary Fig. S1). These results suggest that a significant fraction of resistant cases lack secondary ALK mutations and likely harbor ALK-independent mechanisms of resistance.

To evaluate the role of MET as a resistance mechanism, we first assessed MET copy number in 86 tumor biopsies using FISH or the FoundationOne assay (Fig. 1A; Supplementary Table S1). Eleven (13%) biopsies harbored MET amplification (Fig. 1B), including four with low-level MET amplification (MET/CEP7 2.4–3.9) and six with high-level MET amplification based on FISH (MET/CEP7 5.2 to >25) or NGS (16–19 MET copies). One sample had focal MET amplification by NGS, and FISH was too variable to estimate copy number. In 3 of 11 cases, there was available tissue to assess MET copy number in a biopsy from an earlier timepoint, none of which had MET amplification, supporting the notion that MET amplification was newly acquired. No coalterations in other genes potentially associated with resistance or bypass signaling were identified in the 11 biopsies. In addition, MET amplification was mutually exclusive with ALK resistance mutations, with the exception of one case, MGH9226 (Fig. 1B; Supplementary Fig. S1). This patient developed ALK I1171N (without MET amplification) in plasma after relapsing on first-line alectinib, and then responded to second-line brigatinib for 5.5 months. After relapsing on brigatinib, tumor biopsy demonstrated persistence of ALK I1171N and acquisition of MET amplification, suggesting that MET was driving resistance.

All 11 cases with MET amplification were patients relapsing on second- or third-generation ALK TKIs (Fig. 1B). MET amplification was identified in six (12%) of 52 biopsies taken after a second-generation ALK TKI and in five (22%) of 23 post-lorlatinib biopsies. MET amplification was not detected in patients relapsing on crizotinib. In addition, 6 (55%) of the 11 positive cases had never received crizotinib (Fig. 1B and C). To determine whether the development of MET amplification may be impacted by prior crizotinib exposure, we evaluated the frequency of MET amplification according to prior TKI therapy. Tumors from patients previously treated with crizotinib followed by next-generation TKI(s) were significantly less likely to harbor MET amplification than those from patients treated only with next-generation TKI(s) (9% vs. 33%, P = 0.019; Fig. 1C). Thus, prior exposure to crizotinib, an ALK inhibitor with documented MET activity, may have suppressed emergence of MET-amplified clones in these patients.

We then analyzed MET copy number in 106 plasma specimens from patients relapsing on next-generation ALK TKIs using the Guardant360 assay (Fig. 1A). MET copy number gain was detected in eight (7.5%) plasma specimens (Supplementary Fig. S2), one of which was due to polysomy of chromosome 7. The seven cases with focal amplification of MET included two (3%) of 77 plasma specimens from patients relapsing on a second-generation ALK TKI and five (17%) of 29 post-lorlatinib specimens (Fig. 1B). Among 23 patients with contemporaneous plasma and tissue specimens, MET amplification was detected in both plasma and tissue in four cases (Supplementary Table S3). Eighteen pairs demonstrated absence of MET amplification. One patient, MGH960, had low-level MET amplification (2.2 copies) in plasma which was not detected in a contemporaneous biopsy. Thus, when tissue was used as the reference, plasma genotyping demonstrated 100% sensitivity, 95% specificity, and 80% positive predictive value for detecting MET amplification. The remaining 2 patients with high-level MET amplification (4.9–6.1 copies) in plasma did not have a paired tumor biopsy.

Four of the seven specimens with focal MET amplification harbored both an ALK mutation and MET amplification in plasma (Supplementary Fig. S2), including MGH9226, as described above. Apart from MGH9226′s ALK I1171N mutation, the remaining three plasma specimens contained the gatekeeper ALK L1196M mutation. ALK L1196M is known to cause resistance to crizotinib but is sensitive to next-generation ALK TKIs (7). Indeed, all 3 patients had received crizotinib prior to the next-generation ALK TKI(s) and presumably developed this secondary mutation after crizotinib exposure. The known activity of next-generation ALK TKIs against ALK L1196M suggests that MET amplification rather than the ALK mutation was likely mediating drug resistance. In addition to MET amplification (6.1 copies), MGH9224′s plasma demonstrated KRAS amplification (2.6 copies) and a PIK3CA E545K mutation (allelic frequency 0.04%) which may have also contributed to resistance. The remaining plasma specimens did not contain other putative drivers of resistance.

We next analyzed resistant specimens for other genetic alterations that could lead to aberrant MET activation. To identify MET mutations, we reviewed molecular profiling results from 179 tissue and plasma specimens genotyped by MGH SNaPshot/Solid Fusion NGS (n = 43), FoundationOne (n = 30), or Guardant360 (n = 106; Supplementary Figs. S1 and S2). One plasma specimen from a patient relapsing after sequential alectinib and brigatinib harbored MET exon 14 skipping without MET amplification (Fig. 1B). Two additional plasma specimens contained MET mutations (D1101H and G500C; Supplementary Fig. S2) of unknown significance. MET mutations were not detected in any of the tissue specimens.

In two (3%) of 73 tissue specimens, we detected a rearrangement fusing exons 2–21 of MET with exon 1 of ST7 (suppressor of tumorigenicity 7), both of which reside in close proximity on chromosome 7q (Fig. 2A). Both cases were detected using an RNA-based NGS assay (MGH Solid Fusion Assay; ref. 22). In the first case, the ST7-MET fusion was seen in MGH9284′s pericardial fluid at progression on first-line alectinib without concurrent MET amplification. The patient had primary progression on second-line lorlatinib at which time molecular profiling of pleural fluid did not demonstrate the rearrangement but revealed low-level MET amplification (MET/CEP7 2.4, Fig. 1B). In the second case, concomitant ST7-MET and high-level MET amplification were detected in MGH915′s pleural fluid after failure of lorlatinib (Fig. 1B), neither of which were present prior to lorlatinib. Interestingly, contemporaneous biopsy of an axillary node showed only high-level MET amplification and no detectable ST7-MET.

To evaluate the functional role of MET in driving resistance to ALK TKIs, we established cell lines from MGH915′s pre-lorlatinib and post-lorlatinib malignant pleural effusion (MGH915-3 and MGH915-4, respectively). Consistent with the pleural fluid analysis, the MGH915-4 cell line harbored both MET amplification (MET/chromosome 7 ratio 4.2) and the ST7-MET rearrangement, but MGH915-3 did not (Supplementary Fig. S3A and S3B). High MET expression was confirmed in MGH915-4 cell line at the mRNA and protein level (Supplementary Fig. S3B and S3C). In cell viability studies, the MGH915-4 cell line was resistant to second- and third-generation ALK TKIs, none of which have MET activity, but was sensitive to crizotinib (Fig. 2B). Treatment with MET-specific TKIs capmatinib or savolitinib, none of which have ALK activity, partially suppressed proliferation (Supplementary Fig. S3D). However, the combination of the ALK-selective TKI lorlatinib with capmatinib, savolitinib, or crizotinib potently suppressed cell proliferation (Fig. 2C; Supplementary Fig. S3D). Consistent with the cell growth assays, lorlatinib monotherapy potently inhibited ALK phosphorylation in immunoblotting studies, but failed to inhibit downstream ERK or PI3K/AKT signaling (Fig. 2D). Only dual inhibition of ALK and MET by crizotinib or by the combination of lorlatinib plus a MET TKI effectively suppressed both ALK and downstream signaling pathways (Fig. 2D; Supplementary Fig. S3E).

As MGH915-4 harbors the ST7-MET rearrangement, we next evaluated whether ST7-MET rearrangement could drive resistance to ALK inhibition. The sensitive ALK-positive cell line H3122 was transduced with lentivirus expressing a TET-inducible ST7-MET construct corresponding to the ST7-MET fusion identified in MGH915-4 (Fig. 2A; Supplementary Fig. S3C). Whereas the H3122 cells without expression of ST7-MET were highly sensitive to lorlatinib and other ALK TKIs, doxycycline-induced expression of ST7-MET in H3122 cells caused resistance to all ALK TKIs except crizotinib (Fig. 3A). Treatment with MET TKIs restored sensitivity to lorlatinib (Fig. 3B). Similarly, treatment with a selective MET TKI alone did not suppress proliferation of H3122 cells expressing ST7-MET, but the addition of lorlatinib to the MET TKIs resensitized the cells to treatment (Fig. 3B). The combination of lorlatinib plus crizotinib had greater antiproliferative effect than crizotinib alone, likely due to more potent inhibition of ALK (Fig. 3B).

We next examined the effect of ST7-MET expression on ALK and MET signaling pathways. In the absence of doxycycline induction of ST7-MET, we observed very little MET phosphorylation, and suppression of ALK phosphorylation with ALK TKIs was sufficient to inhibit downstream MAPK and PI3K signaling (Fig. 3C). Upon doxycycline-induced expression of ST7-MET, we observed increased phospho-MET, which could be suppressed with MET TKIs. Consistent with the effects on cell viability, the combination of MET plus ALK inhibitor, but not either agent alone, was sufficient to inhibit downstream signaling (Fig. 3C). Of note, we were unable to distinguish between ST7-MET and endogenous MET by Western blot analysis due to their similar sizes and lack of suitable antibodies capable of detecting ST7 exon 1. However, upon doxycycline induction of ST7-MET, we detected an increase in two bands corresponding to immature pro-MET (175 kDa) and the mature MET beta-chain (145 kDa), with an increase in phosphorylation of mature MET only after doxycycline induction (Fig. 3C). To compare the functional effect of ST7-MET relative to amplification of wild-type MET, we also generated H3122 cells overexpressing wild-type MET. Similar to ST7-MET, doxycycline-induced expression of MET caused resistance to all ALK inhibitors except crizotinib, and the combination of ALK and MET inhibitors suppressed downstream signaling and cell proliferation (Supplementary Fig. S4). In our overexpression system, the effect of ST7-MET and wild-type MET were comparable. While the nature of our experimental system precludes the ability to discriminate between increased MET activity resulting from a hyperfunctional ST7-MET protein or increased MET expression level, these studies mimic the MGH915-4 clinical scenario and suggest that resistance resulting from increased expression of ST7-MET is potentially targetable.

Based on our preclinical findings, we treated two resistant ALK-positive patients with combination ALK/MET TKIs. The first patient, MGH939, received first-line alectinib and relapsed after 9 months (Fig. 4A). Biopsy of a progressing lung lesion demonstrated no ALK mutations. However, high-level MET amplification with MET/CEP7 > 25 was identified (Fig. 1B). Similarly, plasma testing showed no detectable ALK mutations, but did show evidence of MET amplification (2.5 copies). The patient was treated with carboplatin/pemetrexed for 5 cycles and then transitioned to crizotinib after disease progression. Scans performed 5 weeks after initiating crizotinib demonstrated significant response to therapy (Fig. 4B). However, the patient relapsed 10 weeks later. Biopsy of a chest wall mass confirmed persistent MET amplification. There was insufficient tissue for NGS, but plasma testing revealed multiple ALK resistance mutations that were not detected pre-crizotinib/post-alectinib, in addition to the known MET amplification. These findings suggest that while crizotinib is active against both ALK and MET, its activity may ultimately be limited by secondary ALK mutations.

The second patient, MGH915, was treated with first-line ceritinib with response lasting 31 months. Subsequent therapies included alectinib-based combinations, chemotherapy plus immunotherapy, chemotherapy with alectinib, and lorlatinib (Fig. 5A). As noted previously, at the time of relapse on lorlatinib, analysis of pleural fluid demonstrated ST7-MET and high-level MET amplification (MET/CEP7 ratio 5.2, Fig. 5B). A concurrent biopsy of a growing axillary node also revealed MET amplification (MET/CEP7 ratio 5.7) but did not demonstrate the MET rearrangement. On the basis of these findings, the patient received the combination of lorlatinib and crizotinib. Due to potential overlapping toxicities, the lorlatinib dose was reduced to 50 mg when crizotinib 250 mg twice a day was added. After 2 weeks on the combination, lorlatinib was escalated to 75 mg daily. The patient tolerated combined therapy well with side effects of grade 1 nausea and grade 1 lipid elevation. He experienced rapid symptomatic improvement, and first restaging scans confirmed improvement in disease (Fig. 5C). However, after 3 months, the patient developed progression of lung and axillary metastases (Fig. 5C). Biopsy of an axillary node demonstrated persistent high-level MET amplification with MET/CEP7 ratio >25 (Fig. 5B).

To investigate the evolutionary origin of MET alterations, we performed WGS on MGH915′s serial tumor specimens and patient-derived models (Fig. 6A; Supplementary Fig. S5A and S5B). We first focused on the MET locus and observed that the post-lorlatinib specimens (MGH915-4) harbored a focal amplification (copy number 14-16) extending from the intergenic region preceding the MET gene to intron 1 of ST7 (Fig. 6B). This region was contained within a larger approximately 3 Mb amplification (Supplementary Fig. S5C), consistent with an initial tandem duplication of the MET-ST7 locus, followed by amplification of the resulting ST7-MET and wild-type MET loci (∼7 copies each; Fig. 6C). The duplication placed exon 1 of ST7 22kb upstream of the MET transcriptional start site, likely generating a chimeric read-through fusion whereby mRNA processing causes the first exon of ST7 to be spliced to MET exon 2. Neither the tandem duplication nor MET amplification was present in the pre-lorlatinib (post-alectinib) pleural effusion (MGH915-3). In the post-lorlatinib/crizotinib combination sample (MGH915-9), we observed an independent MET amplification event that did not involve the amplified 3 Mb region housing the ST7-MET tandem duplication (Fig. 6B; Supplementary Fig. S5C). When all non-MET mutation clusters and MET alterations were assessed, there was significant overlap between mutations in the post-alectinib pleural effusion (MGH915-3) and the post-lorlatinib/crizotinib axillary node (MGH915-9), but MET amplification only occurred in MGH915-9 (Fig. 6D). The post-lorlatinib pleural effusion (MGH915-4), however, represented a distinct branch of the phylogenetic tree which included more than 3,400 private mutations in addition to MET amplification and ST7-MET. Thus, reconstruction of phylogenetic relationships supports convergent evolution of independent genomic events leading to MET-driven acquired resistance to ALK TKIs.

We analyzed more than 200 tissue and plasma specimens to identify genetic alterations leading to MET bypass signaling in resistant ALK-positive NSCLC. We detected MET amplification and/or a MET fusion in 15% of tumors after failure of a next-generation ALK TKI. Importantly, resistant cell line models and patients harboring MET amplification or rearrangement could be resensitized to treatment with dual ALK/MET inhibition.

MET gene amplification was initially discovered as a bypass mechanism in EGFR-mutant NSCLC in 2007 and has subsequently been identified in up to 20% of resistant specimens (14, 23). In this study, we demonstrate that the prevalence of MET alterations in patients relapsing on ALK TKIs may be comparable to that seen in EGFR-mutant NSCLC. Interestingly, the incidence of MET amplification in EGFR-mutant NSCLC is higher after exposure to the third-generation EGFR TKI osimertinib compared with less potent EGFR TKIs (14, 15). Similarly, in our series, we observed that the frequency of MET amplification was highest in tumors resistant to the third generation, broad-spectrum ALK TKI lorlatinib (Fig. 1C). These findings suggest that an association may exist between TKI potency and likelihood of developing ALK-independent resistance mechanisms such as MET amplification. In addition to TKI potency, TKI selectivity may also impact the evolution of resistance. In support of this hypothesis, we observed MET amplification in one-third of cases that received front-line second-generation ALK TKIs (which have no activity against MET) compared with less than 10% of patients initially treated with crizotinib.

MET signaling can be activated through a variety of mechanisms, including copy number gain, mutation, fusions, and ligand upregulation (24). In this study, we found that copy number gain was most frequently used to activate MET bypass signaling. The range of MET copies in tissue was broad, spanning MET/CEP7 ratios of 2.4 to greater than 25 (Fig. 1B). In cases where MET amplification was detected in paired tissue and plasma specimens, the number of MET copies in plasma was consistently lower than in tissue. Tumor DNA represents a small fraction of total cell-free DNA and some disease sites are more likely to shed tumor DNA than others. Thus, calculation of MET copy number in plasma is challenging. The relationship between MET copy number and dependency on MET bypass signaling in resistant ALK-positive NSCLC remains to be determined. While NSCLCs with de novo high-level MET amplification appear to be more sensitive to crizotinib than NSCLCs with low-level MET amplification, MET/CEP7 ratio of 4–5 is an imperfect threshold for predicting sensitivity to crizotinib and some tumors with low-level amplification respond to crizotinib (20). In the context of acquired resistance, the relationship between MET copies and sensitivity to MET TKIs may be more complex. Specifically, in resistant ALK-positive NSCLCs incompletely inhibited by ALK TKIs, it is conceivable that even low-level MET amplification may be adequate to restore proliferative signals. Furthermore, intratumoral molecular heterogeneity resulting from exposure to multiple lines of therapy may impact response to MET TKIs even among ALK-positive tumors with high-level MET amplification.

In addition to MET amplification, we detected identical ST7-MET rearrangements in 2 patients relapsing on next-generation ALK TKIs. Rare ST7-MET rearrangements with different breakpoints have been described in gliomas and ovarian cancer, but until now the pathogenicity of these fusions was unknown (25, 26). MGH915′s patient-derived cell line harbored both ST7-MET and MET amplification, preventing us from assessing the independent contribution of each MET alteration to ALK TKI resistance. We demonstrate that introducing ST7-MET into a sensitive ALK cell line is sufficient to confer resistance to ALK-targeted therapy. However, as engineering the cells to express the ST7-MET rearrangement also increases the total expression ST7-MET, we are unable to conclusively determine whether the fusion alone (in the absence of overexpression) is sufficient to drive resistance. Interestingly, ST7-MET was only identified in one of MGH915′s two contemporaneous biopsies, while high-level MET amplification was present in both of the resistant disease sites. In MGH9284′s case, ST7-MET was identified in pericardial fluid while low-level MET amplification without the rearrangement was detected in pleural fluid 2 months later. These findings suggest that a single tumor can acquire distinct resistance alterations that converge on MET activation and also highlight the potential for intratumoral heterogeneity of MET genetic alterations.

In 2 patients with MET-driven resistance, inhibiting both ALK and MET led to clinical responses. However, despite initial responses, both patients relapsed after approximately 3 months. At relapse on crizotinib monotherapy, MGH939′s plasma analysis demonstrated multiple crizotinib-resistant (but next-generation ALK TKI-sensitive) ALK mutations, providing rationale for exploring combinations pairing more potent next-generation ALK TKIs with a MET TKI. Indeed, MGH915 did receive combination therapy with lorlatinib and crizotinib, but still had a short-lived response. Repeat biopsy after relapse on the combination revealed persistent MET amplification without additional genetic alterations associated with resistance to ALK and/or MET TKIs (Figs. 3B and 4A). Interestingly, the MET/CEP7 ratio increased from 5.7 to greater than 25 at relapse on the lorlatinib/crizotinib combination. As there can be cell-to-cell variability in MET copy number which may be difficult to capture by FISH, it is possible that there was expansion of a more highly MET amplified subpopulation under the selective pressure of crizotinib. In this scenario, it is conceivable that the MET copy number increase was not the primary driver of relapse in MGH915′s case. Resistant cells may have developed other alterations (including nongenetic aberrations) that decreased sensitivity to dual ALK/MET combination treatment. To inform the clinical development of ALK/MET TKI combinations, additional studies will be required to characterize molecular determinants of resistance to ALK/MET combination therapies, to determine whether sensitivity to combination therapy is impacted by line of therapy, and to assess antitumor activity of other more potent MET TKIs like capmatinib and savolitinib (Supplementary Fig. S3D). Indeed, there is clinical rationale for exploring ALK/MET combinations besides lorlatinib/crizotinib. For example, while cross-trial comparisons suggest that most adverse events, including nausea/vomiting and edema, occur at comparable rates with the three MET TKIs evaluated in cell lines (crizotinib, capmatinib, savolitinib; refs. 4, 27, 28), the blood–brain barrier penetration of capmatinib is superior to crizotinib. Capmatinib may therefore be a more optimal partner for lorlatinib in patients with brain metastases.

This study has several important limitations. First, due to limited tissue availability, we could not perform analyses for all MET alterations in all specimens. In addition, we used several assays to detect MET alterations, each with its own limitations. Second, patients in our study did not consistently undergo repeat biopsy each time they relapsed on an ALK TKI, precluding definitive timing of the acquisition of the MET alteration in the majority of cases. Third, we limited our analysis to genetic alterations of MET in resistant tumors and did not examine nongenetic mechanisms of MET activation, including upregulation of MET's ligand hepatocyte growth factor. Fourth, in the WGS analysis, clonal relationships were inferred from patient-derived cell lines and xenografts and biopsies obtained from a variety of sites. It is possible that patient-derived models may not fully recapitulate the spectrum of subclones in the tumor and that spatial heterogeneity contributed to differences in sequential biopsies. Fifth, most of the patients in this study had previously received crizotinib which could have prevented expansion of resistant subclones harboring MET alterations, thereby decreasing the chance of MET activation emerging as a bypass pathway with subsequent next-generation ALK TKIs. As more selective second-generation ALK TKIs have recently replaced crizotinib in the front-line setting, the frequency of MET bypass signaling driving resistance may be higher than estimated in this study. Finally, although we identified rare ST7-MET rearrangements in a subset of cases, we could not fully validate the functional role of the rearrangement or confirm that the rearrangement was an independent mediator of resistance due to limitations of our experimental system. Additional studies—ideally using patient-derived models that do not harbor concurrent MET amplification—are needed to robustly characterize these rearrangements.

In summary, by performing the most comprehensive analysis of MET alterations in ALK-positive NSCLC to date, we demonstrate that MET is an important bypass mechanism in tumors exposed to next-generation ALK TKIs. The detection of MET amplification in one-third of patients who received next-generation ALK TKIs in the front-line setting provides justification for exploring ALK/MET combinations as an initial treatment strategy for advanced ALK-positive NSCLC.

I. Dagogo-Jack is a paid consultant for Boehringer Ingelheim and Foundation Medicine, reports receiving other commercial research support from Genentech and Pfizer, and reports receiving other remuneration (travel support) from Pfizer and Array. J.J. Lin is a paid consultant for Nuvalent, C4 Therapeutics, and Pfizer, reports receiving commercial research grants from Novartis, Turning Point Therapeutics, Neon Therapeutics, and Hengrui, and reports receiving other remuneration from OncLive. S.R. Digumarthy reports receiving speakers bureau honoraria from SiEMENS. A. Saxena is a paid consultant for Genentech, AstraZeneca, and Medtronic, and reports receiving speakers bureau honoraria from AstraZeneca and Takeda. R.J. Nagy is an employee of and holds ownership interest (including patents) in Guardant Health, Inc. C.H. Benes reports receiving commercial research grants from Novartis and Eli-Lilly. J.F. Gainor is a paid consultant for Bristol-Myers Squibb, Merck, Genentech, Blueprint, LOXO, Lilly, Regeneron, Pfizer, Ironwood Pharmaceuticals, Moderna, and Jounce, and reports receiving other commercial research support from Novartis. A.T. Shaw is an employee of Novartis and a paid consultant for Pfizer, Genentech/Roche, Ariad/Takeda, Syros, Blueprint Medicine, KSQ Therapeutics, TP Therapeutics, Chugai, Daiichi-Sankyo, LOXO/Bayer, Achilles, Archer, Foundation Medicine, and Guardant. A.N. Hata reports receiving commercial research grants from Pfizer, Novartis, Amgen, Roche/Genentech, Eli Lilly, and Relay Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: I. Dagogo-Jack, J.F. Gainor, A.T. Shaw, A.N. Hata

Development of methodology: I. Dagogo-Jack, S. Yoda, J.K. Lennerz, K. Prutisto-Chang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Dagogo-Jack, S. Yoda, J.K. Lennerz, J.J. Lin, M.M. Rooney, K. Prutisto-Chang, A. Oh, N.A. Adams, A. Do, S.E. Stevens, A. Saxena, R.J. Nagy, C.H. Benes, C.G. Azzoli, J.F. Gainor, A.T. Shaw

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Dagogo-Jack, S. Yoda, J.K. Lennerz, A. Langenbucher, K. Prutisto-Chang, B.Y. Yeap, H.D. Marble, R.J. Nagy, C.H. Benes, M.S. Lawrence, J.F. Gainor, A.T. Shaw

Writing, review, and/or revision of the manuscript: I. Dagogo-Jack, S. Yoda, J.K. Lennerz, J.J. Lin, M.M. Rooney, B.Y. Yeap, A. Do, S.R. Digumarthy, A. Saxena, R.J. Nagy, C.G. Azzoli, J.F. Gainor, A.T. Shaw, A.N. Hata

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Dagogo-Jack, M.M. Rooney, N.A. Adams, E. Chin, A. Do, H.D. Marble, A.T. Shaw

Study supervision: A.T. Shaw, A.N. Hata

Other (image interpretation): S.R. Digumarthy

This work was supported by a Conquer Cancer/Amgen Young Investigator Award (to I. Dagogo-Jack), an institutional research grant from American Cancer Society (to I. Dagogo-Jack), a NCI Career Development Award (K12CA087723-16, to I. Dagogo-Jack), a Lung Cancer Research Foundation Annual Grant Program award (to S. Yoda), a Conquer Cancer/AstraZeneca Young Investigator Award (to J.J. Lin), grants from the NCI (R01CA164273, to A.T. Shaw; R01CA225655, to J.K. Lennerz), by Be a Piece of the Solution, and by Targeting a Cure for Lung Cancer Research Fund at MGH (Boston, MA).

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.