Top-level MET gene copy number gain defines a subtype of poorly differentiated pulmonary adenocarcinomas with poor prognosis.

BACKGROUND: MET amplifications occur in human tumors, including non-small cell lung cancer (NSCLC). MET inhibitors have demonstrated some clinical activity in MET amplified NSCLC, presumably with a gene dose effect. However, the definition of MET positivity or MET amplification as a potential oncogenic driver is still under debate. In this study, we aimed to establish the molecular subgroup of NSCLC with the highest unequivocal MET amplification level and to describe the prevalence, and histologic and clinical phenotype of this subgroup.
METHODS: A total of 373 unselected patients with NSCLC were consecutively tested for MET gene copy number (GCN) by FISH. Mean GCN, MET/CEN7 ratio and other FISH parameters were identified and correlated with morphological and molecular pathological characteristics of the tumors as well as with clinical data.
RESULTS: Based on the variability of obtained data a top-level category of MET amplification was newly defined (>90th percentile of average GCN; ≥10 MET gene copies per tumor cell). This criterion was fulfilled in 2% of analyzed tumors. These tumors were exclusively poorly differentiated adenocarcinomas with a predominant solid subtype and pleomorphic features. Rarely, co-alterations were detected (KRAS mutation or MET exon 14 skipping mutation). In this top-level group, there were no EGFR mutations or ALK or ROS1 alterations. The most important clinical feature was a significantly shortened overall survival (HR 3.61; median OS 8.2 vs. 23.6 months). Worse prognosis did not depend on initial stage or treatment.
CONCLUSIONS: The newly defined top-level category of MET amplification in NSCLC defines a specific subgroup of pulmonary adenocarcinoma with adverse prognosis and characteristic morphological features. Lower levels of MET gene copy number seem to have probably no specific value as a prognostic or predictive biomarker. 2020 Translational Lung Cancer Research. All rights reserved.

Introduction

Lung cancer is still the main cause for cancer related deaths worldwide. Understanding the mechanisms of molecular carcinogenesis of non-small cell lung cancer (NSCLC) is crucial to discover specific therapeutic targets and has led to improved outcome (1). However, although an increasing number of targeted therapies and immuno-oncology related treatments is available nowadays, NSCLC still remains a deadly disease since only a minority of patients can be cured (2). One of the biologically and therapeutically relevant targets in NSCLC and many other human cancers is the mesenchymal-epithelial transition receptor (MET) and its ligand, hepatocyte growth factor (HGF) (3,4). The MET proto-oncogene was initially described by Cooper et al. in an osteosarcoma derived cell line in 1984 (5). The MET gene is located on chromosome 7q and its product, a heterodimeric transmembrane receptor tyrosine kinase, consists of an extracellular α- and a transmembrane β-chain (1,3).

MET as a receptor tyrosine kinase can be activated by a multitude of biologic mechanisms, such as gene fusions, activating mutations, gene amplification and also simply by overexpression of the receptor protein or by ligand dependent activation. MET activation itself leads to dimerization and transphosphorylation followed by activation of downstream signaling via PI3K/AKT, RAS-RAC/RHO, MAPK and phospholipase C pathways (6). The effects are manifold: the MET/HGF pathway has an impact on multiple cellular functions, such as differentiation, cell cycle progression, proliferation and angiogenesis (7). Its dysregulation occurs in many different types of cancer (4) and leads to several effects in tumorigenesis, such as cancer cell proliferation, invasion, survival, motility and the development of metastases (8).

Activating missense mutations in the tyrosine kinase domain have been described in papillary renal cancer (9). Another type of activating MET mutations affects the splice site donor and acceptor regions around exon 14. Alternative splicing with consecutive skipping of exon 14 causes a stabilization and accumulation of catalytically active MET protein on the cell surface due to reduced ubiquitinylation and proteasomal degradation. Originally discovered in small cell lung cancer, MET exon 14 skipping mutations have also been described in 3–6% of adenocarcinoma of the lung and about 1–2% of tumors with other NSCLC histologies (10-14). Moreover, MET exon 14 skipping mutations were identified as an independent prognostic factor that predict poor survival (15,16). MET amplification has been described in about 3-5% of newly diagnosed NSCLC (15,17,18) and increased MET gene copy number seems to be a negative prognostic factor (17,19-21).

Many tyrosine kinase inhibitors with anti-MET activity are currently being explored in cancers with MET activation, among them MET amplified and mutated NSCLC. Early data from clinical trials is available mainly for crizotinib, capmatinib and tepotinib (22). Recently, Camidge et al. presented an update of the PROFILE 1001 study reporting on MET targeting therapy with crizotinib in 40 NSCLC patients (23). Those with high MET amplification [defined by MET/centromere 7 (CEP7) ratio ≥4] showed clinically meaningful antitumor activity with rapid and durable responses. Objective response rates were lower in tumors with lower MET amplification levels. Thus, based on available data, MET amplification is probably both, a negative prognostic and a potential predictive biomarker for MET tyrosine kinase inhibitors. However, generally accepted criteria for MET positivity in NSCLC do not yet exist. Moreover, even methods to detect clinically meaningful MET alterations are still under discussion. MET mutations, i.e. those mutations which cause exon 14 skipping, and gene fusions can be detected by DNA-based next generation sequencing of the intron-exon borders around exon 14 of the MET gene. Additionally, RNA-based approaches are employed. Also, gene copy number gains can be detected by some sequencing assays. However, fluorescence in situ hybridization (FISH) has been used to select patients with MET amplification in clinical trials on MET inhibitors so far (23,24). Detections of MET protein expression by immunohistochemistry (IHC) was shown to be associated with amplification to a certain extent (18). However, a clinical trial with the therapeutic monoclonal MET antibody onartuzumab failed to demonstrate a clinically meaningful predictive value of MET IHC (25,26). Based on currently available treatment approaches in NSCLC with MET inhibitors including clinical trials, two types of predictive biomarkers seem to be the most promising: (I) DNA or RNA sequencing for exon 14 skipping mutations, and (II) FISH for amplification.

However, various and different, sometimes even contradictory criteria for MET amplification or MET copy number gains have been proposed. Some authors have used a high MET/centromere 7 copy number ratio as a measurement for amplification (17,19,23). Since high level MET copy number gains can also occur against the background of simultaneously increased copies of centromeric regions (resulting in a “negative” ratio <2.0), we have previously suggested a more general approach to describe copy number changes of MET (18). This approach specifically emphasizes average gene copy number and has been adopted in current clinical trials (24).

Data on the frequency of different amplification levels in NSCLC patients are still sparse. Moreover, MET gene copy number gains have not been comprehensively correlated with clinical data so far. In this study, we aimed at elucidating the frequencies of MET amplification levels in an unbiased series of consecutive clinical samples of NSCLC patients and correlating these levels with different histologies, molecular subtypes and outcome of patients. Since effects of MET inhibitors seem to be related to a gene-dose effect at least to a certain degree, we furthermore aimed to establish the molecular subgroup of NSCLC with the highest unequivocal MET amplification level and to describe the histologic and clinical phenotype of this newly defined subgroup.

Methods

Patients

A total of 390 unselected consecutive NSCLC have been included in this study. Patients were tested for MET amplification between January 1st, 2015 and June 30th, 2017 as part of the routine molecular diagnostics at the Institute of Pathology of the University Medical Center Göttingen, Germany. Seventeen patients were excluded due to missing clinical data, resulting in a series of 373 consecutive NSCLC patients. Tumor stage was determined based on the 8th edition of UICC TNM Staging System of lung cancer (27). Cases which were initially staged on the basis of the 7th edition were double-checked and re-staged if appropriate. For subsequent data analysis, stages IVA and IVB were aggregated to stage IV and compared with lower stages. All patients were treated at a dedicated lung tumor center (Lungentumorzentrum Universität Göttingen). Clinical and follow-up data were obtained from their medical records.

All patients were treated according to local standards which are based on national and international guidelines, and if necessary, received systemic therapy. Treatment information was analyzed retrospectively in full detail for all patients with MET gene copy number gains (defined by MET level ≥1; see below; n=141). 70.2% (99/141) of these patients received systemic therapy at any time. First line therapy consisted of platinum-based combination treatments (n=90 patients, with pemetrexed, paclitaxel, docetaxel, gemcitabine, or etoposide). In 22 of these patients, a triple combination with bevacizumab was given. In patients who received definitive radio-chemotherapy, cisplatin was combined with vinorelbine or given as a monotherapy. In addition, single agent systemic therapies in stage IV disease included pemetrexed, and erlotinib or afatinib in EGFR mutant cancers. 48.5% (48/99) of patients were treated in a 2nd line setting who received docetaxel with or without nintedanib, platinum-based combinations with paclitaxel, vinorelbine, or etoposide, triple combinations with bevacizumab, or a monotherapy with pemetrexed, afatinib, erlotinib, or gefitinib. 11 patients were treated with PD-L1 or PD-1 inhibitors (atezolizumab, nivolumab or pembrolizumab in 1, 7 and 3 patients, respectively), and 7 patients received an anti-MET tyrosine kinase inhibitor (TKI) as 2nd line treatment. 3rd line therapy was given in 21/99 patients (21.2%; nivolumab: n=11; MET TKI: n=1, further therapies included erlotinib, pemetrexed, carboplatin plus paclitaxel, and docetaxel with or without ramucirumab).

Only a small number of patients (9/99, 9.1%) was also treated in a 4th line setting. These therapies were based on atezolizumab in one patient and nivolumab in another two patients, furthermore docetaxel with or without ramucirumab, gemcitabine, vinorelbine, and erlotinib. Two patients were treated with docetaxel in a 5th line setting, one with further vinorelbine at 6th line and with nivolumab at 7th line.

This study was conducted after approval of the local Ethics committee (5/1/17).

Histology, subtyping and molecular profiling

The majority of the specimens (60%) consisted of primary tumor tissue, predominantly transbronchial biopsies followed by core needle biopsies and resection specimens, and one third were surgical resections and biopsies of distant metastases. Cytology specimens (i.e., smears or cytospin preparations) were excluded. Tumor diagnoses were established based on the current WHO and IASLC classification (28,29). In brief, tumors with strong and predominant expression of p40 were classified as squamous cell carcinomas; lesions with convincing expression of neuroendocrine markers, e.g., CD56, chromogranin A, synaptophysin, were regarded as neuroendocrine tumors; p40-negative non-neuroendocrine carcinomas, either gland forming or solid, with or without TTF-1 expression were recognized as adenocarcinomas. Additional diagnostic biomarkers were applied if appropriate. For further statistical analyses tumors were grouped into four subgroups: (I) adenocarcinomas, (II) squamous cell carcinomas, (III) sarcomatoid carcinomas (comprising pure spindled or pleomorphic carcinomas), (IV) others (include large cell neuroendocrine carcinomas, typical/atypical carcinoids). Frequencies of tumor subtypes are summarized in .

Table 1

Patients’ characteristics

Target	MET amplification negative	MET gene copy number gain					Total tested
		Level 1	Level 2	Level 3	Level 4	MET gene copy number gain (total)
		≥40% of tumor cells with ≥4 MET signals and criteria for level 2, 3 or 4 amplification are not fulfilled	≥50% of cells containing ≥5 MET signals and criteria for level 3 or 4 amplification are not fulfilled	MET/CEN7 ratio ≥2.0 or average MET GCN per cell of ≥6 but <10 or ≥10% of tumor cells containing ≥15 MET signals	Average MET gene copy number per cell of ≥10
Total	232 (62.2%)	84 (22.5%)	22 (5.9%)	27 (7.2%)	8 (2.1%)	141 (37.8%)	373 (100%)
Gender
Male	144 (62.1%)	46 (54.8%)	11 (50%)	17 (63.0%)	7 (87.5%)	81 (57.4%)	225 (60.3%)
Female	88 (37.9%)	38 (45.2%)	11 (50%)	10 (37.0%)	1 (12.5%)	60 (42.6%)	148 (39.7%)
Age
Median (range)	64.1 (36.4–84.4)	69.1 (39.5–87.3)	64.0 (40.8–79.1)	63.3 (47.2–80.1)	65.2 (48.7–90.5)	66.1 (39.5–90.5)	64.8 (36.4–90.5)a
Mean	63.6	66.9	63.1	63.7	68.7	65.8	64.4
Stage
I	35 (15.1%)	11	4	4	1	20 (14.2%)	55 (14.7%)
II	24 (10.3%)	8	0	1	0	9 (6.4%)	33 (8.8%)
III	45 (19.4%)	16	5	8	2	31 (22.0%)	76 (20.4%)
IV	117 (50.4%)	40	12	13	5	70 (49.6%)	187 (50.1%)
Ukn.	11 (4.7%)	9	1	1	0	11 (7.8%)	22 (5,9%)
Histologic type
SCC	55 (23.7%)	20	1	5	0	26 (18.4%)	81 (21.7%)
AC	159 (68.5%)	61	21	20	8	110 (78.0%)	269 (72.1%)
Sarcomatoid	1 (0.4%)	3	0	2	0	5 (3.5%)	6 (1.6%)
Other	6 (2.6%)	0	0	0	0	0 (0%)	6 (1.6%)
Unknown	11 (4.7%)	0	0	0	0	0 (0%)	11 (2.9%)
Tumor subtypes
KRAS mutation
Positive	57 (24.6%)	22	3	5	1	31 (22.0%)	88 (23.6%)
Negative	114 (49.1%)	40	16	15	5	76 (53.9%)	190 (50.9%)
Unknown	61 (26.3%)	22	3	7	2	34 (24.1%)	95 (25.5%)
EGFR mutation
Positive	18 (7.8%)	8	7	3	0	18 (12.8%)	36 (9.7%)
Negative	160 (69.0%)	62	15	17	8	102 (72.3%)	262 (70.2%)
Unknown	54 (23.3%)	14	0	7	0	21 (14.9%)	75 (20.1%)
ALK/ROS fusion
Positive	2 (0.9%)	0	0	0	0	0 (0.0%)	2 (0.5%)
Negative	147 (63.4%)	66	21	19	8	114 (80.9%)	261 (70.0%)
Unknown	83 (35.8%)	18	1	8	0	27 (19.1%)	110 (29.5%)
PD-L1 expression
PD-L1 <1	52 (22.4%)	26	5	6	1	38 (27.0%)	90 (24.1%)
PD-L1 1–49	31 (13.4%)	24	3	7	3	37 (26.2%)	68 (18.2%)
PD-L1 ≥50	20 (8.6%)	17	6	7	0	30 (21.3%)	50 (13.4%)
Ukn.	129 (55.6%)	17	8	7	4	36 (25.5%)	165 (44.2%)
MET mutation
Positive	0 (0.0%)	1	0	1	1	3 (2.1%)	3 (0.8%)
Negative	0 (0.0%)	21	5	9	4	39 (27.7%)	39 (10.5%)
Unknown	232 (100.0%)	62	17	17	3	99 (70.2%)	331 (88.7%)
Therapy
MET inhibitor	0 (0.0%)	2	2	1	3	8 (5.7%)	8 (2.1%)
PD-L1 inhibitor	Unknown	10	6	6	3	25 (17.7%)	25 (6.7%)

Clinical baseline characteristics of the entire cohort are summarized in the last column. MET copy number changes, distribution across clinical, morphologic and molecular subtypes are shown. Clinical stage refers to initial diagnosis, later progression to higher stages occurred. SCC, squamous cell carcinoma; AC, adenocarcinoma; CEN7, centromere 7; GCN, gene copy number; a, missing data: n=6.

All non-squamous NSCLC cases underwent further molecular characterization. ALK and ROS1 testing was done by FISH as previously described (30-32). For EGFR and KRAS sequencing, the EGFR and KRAS therascreen assays, (Qiagen, Hilden, Germany) were applied according to the manufacturer’s recommendations. Next generation sequencing was done as previously described (33). All cases were further tested for PD-L1 expression by using the clone 28-8 on a DAKO Omnis platform (staining protocol described in detail in Koppel et al. 2018 (34); evaluation and scoring is described in Schildhaus et al. 2015 (18).

MET FISH

FISH has been carried out as previously described (18,35). In brief, 4 µm thick sections have been hybridized by using the ZytoLight SPEC MET/CEN7 Dual Color Probe (ZytoVision, Bremerhaven, Germany). The numbers of MET and centromere 7 signals were counted in 60 nuclei obtained from three different areas with the highest gene count. Average gene copy number and ratio (MET/CEN7) as well as the percentages of tumor cells with ≥4, ≥5 and ≥15 gene copies were calculated for each tumor. All FISH assays were evaluated by two pathologists with specific experience in this field (KS, HUS). Tumors were categorized into amplification levels, based on previously published criteria after modification (see below, ).

Table 2

FISH criteria for levels of MET gene copy number gains and their frequencies among NSCLC patients

Level	Description	Criteria	Percentage of patients in this series (n=373)	Percentage of patients from Schildhaus et al., 2015 (n=693)
4	Top level	Average MET gene copy number per cell: ≥10	2.1%	0.9%
3	High level	Criterion of level 4 not fulfilled AND	7.2%	2.3%
		• MET/CEN7 ratio ≥2.0 OR• Average MET GCN per cell ≥6 but <10 OR• ≥10% of tumor cells containing ≥15 MET signals
2	Intermediate level	Criteria for levels 3 and 4 not fulfilled; ≥50% of cells containing ≥5 MET signals	5.9%	6.2%
1	Low level	Criteria for levels 2 to 4 not fulfilled; ≥40% of tumor cells with ≥4 MET signals	22.5%	23.4%
0	Negative	Criteria for levels 1 to 4 not fulfilled	62.2%	67.0%

Definition of the top-level MET amplification in the context of pre-existing FISH criteria [levels 0 to 3, according to Schildhaus et al. 2015 (15)]. Data from the previous report have been retrospectively re-calculated. The top-level category has not been described before. Therefore, cases from earlier reports have not been specifically tested for this feature.

Statistics

SPSS software (IBM Corp., IBM SPSS Statistics for Windows, Version 25.0. Armonk, NY USA) was used for statistical analysis. Categorial variables were tested by chi-square or Fisher’s exact test. Survival times were analyzed with the Kaplan-Meier method and for comparing the survival times across different groups, the log rank test was applied. Survival data were available for 371 patients. Overall survival (OS) was calculated based on the date of first diagnosis to the date of last follow up or death of patient. Patients who were still alive or were lost to follow-up were censored at the time of the last contact. Cox regression was used to assess univariate tests and multivariate models. Parameters with P values <0.1 at Cox regression in univariate analysis were tested in multivariate models. Wald test was performed testing different subtypes of categorical variables. Forest plot graph was performed using GraphPad Prism software (GraphPad Software, GraphPad Prism for Windows, version 8.0.1, La Jolla, California USA). All tests were two-sided and statistical significance was defined as P<0.05.

Results

Frequencies of MET gene copy number alterations and definition of a top-level amplification category

Two hundred thirty-two samples (62.2%) were MET amplification negative and 141 showed MET gene copy number gains at various levels. Distribution of MET copy number changes across clinical, morphologic and molecular subgroups together with baseline characteristics of the entire cohort are shown in . Five out of six pure sarcomatoid carcinomas showed MET gene copy number (GCN) gains ().

Average MET gene copy numbers per nucleus ranged from 2.4 to a maximum of 25.2 (median 4.4). MET/CEN 7 ratios were found between 0.8 and 10.1 (median: 1.4). To determine the patients’ subgroup with the highest unequivocal MET amplification level, we determined that parameter with the broadest numerical range (i.e., average gene copy number) and calculated the 90th percentile which was found at 10.8 MET gene copies per tumor cell. Therefore, a top-level MET amplification category (level 4) was defined by an average gene copy number of ≥10 MET signals per nucleus (). MET positivity was found in 84, 22, 27 and 8 patients at levels 1, 2, 3 and 4, respectively (). Among level 4 samples, MET/CEN 7 ratio ranged from 1.4 to 10.1. Four out of eight MET top-level patients had a MET/CEN7 ratio ≥4.0; the remaining cases showed co-amplification of centromeric sequences (CEN7). However, a high ratio ≥4.0 was also found in two patients with less MET gene copies at lower amplification levels. MET GCN gain—at any level—and ALK or ROS1 gene fusions did not co-occur. Eighteen patients (50%) with an activating EGFR mutation showed a simultaneous MET amplification which was detectable prior to EGFR TKI treatment. Three (2.1%) patients were tested positive for MET exon 14 skipping mutation and were classified with a level 1, 3 and 4 MET GCN gain. Two of those patients were initially diagnosed with stage IV and one with stage III.

Clinical and morphologic phenotype of patients with top-level MET amplification (level 4)

We identified eight MET top-level patients, seven of them were male (87.5%). Predominance of male gender was the highest among these patients if compared with all other MET status. Median age was 65.2 years (range, 48.7–90.5 years). Most of the patients were diagnosed with multiple metastases; five patients were initially with stage IV, two with stage III and one with stage I at the time of initial diagnosis. Many level 4 patients had metastatic lesions at uncommon locations, such as skin, muscle or pararectal soft tissue. We obtained information about the smoking status for 6 out of 8 patients. All of them were heavy smokers with a history of at least 40 pack years.

Patients with MET top-level amplifications survived significantly shorter than all other patients in our series [median overall survival (mOS) 8.2 vs. 23.6 months; P=0.002, Log Rank test; ]. Uni- and multivariate analysis demonstrated stage, histotype and MET amplification level as the only independent parameters for outcome. It is noteworthy that adenocarcinomas per se are significantly associated with better outcome than all other NSCLC histologies. This effect, however, is overcompensated by top-level MET amplification which occurred exclusively in adenocarcinomas. Level 4 is associated with a more than threefold increase of the likelihood to die of the cancer (hazard ratio: 3.61; ) which is independent of the clinical stage at initial presentation. Survival of patients with top-level MET amplification (level 4) was shorter even if compared with high level amplified tumors (level 3) and even if systemic treatment was administered in a multitude of therapy lines. demonstrates the treatments applied to MET level 3 and 4 patients, and visualizes the individual progress of disease at different levels of MET amplification. Off label treatment with a MET tyrosine kinase inhibitor was beneficial in one level 4 patient and led to a partial response.

Figure 1

Overall survival analysis. (A) Comparing NSCLC patients of all stages at the time of initial diagnosis, top-level MET copy number gain (level 4, n=8) was associated with a significantly shorter overall survival (HR 3.61; median OS 8.2 vs. 23.6 months for MET Level 0–3, n=363). (B) Notably, this difference was also significant if MET levels are split up. There was no significant survival difference between level 3 (n=27) and lower MET amplification levels (level 1, n=84; level 2, n=22) or MET negative patients (level 0, n=230). This underlines the specific prognostic value of MET level 4 (top-level gain; defined by ≥10 gene copies per cell) over former definitions of high-level MET amplification which are included in level 3. (C) Comparison of OS for the subgroup of stage IV patients who did not receive anti-MET tyrosine kinase inhibitors (TKI) (n=182); median OS of MET level 4 patients (n=4) was significantly shorter than for level 0–3 patients (n=178; 3.5 vs. 14.1 months). (D) This statistical significance disappears if patients are included who were treated with anti-MET-TKI (see for details). Median OS of level 4 patients (n=5) was 5.1 months vs. 14.1 for level 0–3 (n=182).

Table 3

Univariate and multivariate analysis of overall survival

Parameter	Cox regression							HR 95% CI
	Univariate				Multivariate
	HR	95% CI	P value		HR	95% CI	P value
Gender (female vs. male)	0.844	0.641–1.113	0.229		NT	NT	NT
Age (≥65 vs. <65 years)	1.291	0.985–1.690	0.064		1.242	0.931–1.659	0.141
Histology (AC vs. non-AC)	0.765	0.558–1.050	0.097a		0.564	0.397–0.800	0.001
Stage (IV vs. I–III)	2.443	1.815–3.289	<0.001		2.714	1.988–3.705	<0.001
MET GCN gain (level 4 vs. level 0–3)	3.152	1.479–6.719	0.003		3.610	1.674–7.786	0.001
PD-L1 (TPS) (≥50% vs. <50%)	0.844	0.551–1.294	0.437		NT	NT	NT

a, Wald-Test, testing for histology (4 strata: adenocarcinoma, squamous, sarcomatoid, other), revealed a significant survival difference (P=0.001). Therefore histology (adenocarcinoma vs. non-adenocarcinoma) was included in multivariate Cox regression as well. HR, hazard ratio; CI, confidence interval; NT, not tested; AC, adenocarcinoma; non-AC, other NSCLC histologies excluding adenocarcinomas (include 26 patients with squamous cell carcinoma and 6 patients with sarcomatoid carcinomas); GCN, gene copy number; TPS, tumor proportion score.

Figure 2

Individual swimmer plots for NSCLC patients with MET amplification levels 3 and 4 (high and top-level copy number gain). Maximum survival for level 4 patients was 13.3 months from initial diagnosis, compared to 73.8 months in the group of level 3 patients. MET TKI treatment contributed to a prolonged survival in one patient with top-level gain (31.5 months). Each bar represents one patient with individual progress of disease since initial diagnosis; the color of the bars indicates the therapy applied; for description of colors and symbols used in this figure see legend; Level 4 patients listed in red; 1, patients with MET mutation. OS, overall survival.

Also, in terms of morphology, we observed a peculiar phenotype: All MET level 4 patients were basically classified as adenocarcinoma. No squamous cell carcinoma was found among these patients (excluded by lack of p40 expression in all samples). All of these cancers were poorly differentiated, predominantly solid adenocarcinomas with pleomorphic features in terms of interspersed pleomorphic giant cells. A pure or predominant pleomorphic or sarcomatoid differentiation, however, was not seen. Interestingly, five out of eight lesions showed TTF1-positivity which was, however, heterogeneous and moderate (). Regarding molecular subgroups, neither EGFR mutation nor ALK/ROS1 fusions were found among MET level 4 patients, but one KRAS mutation and one MET exon 14 skipping mutation were found. High PD-L1 expression (TPS ≥50%) could not be demonstrated in these tumors.

Figure 3

Morphologic phenotype of MET top-level amplified lung cancers. (A,B,C) All tumors were adenocarcinomas with minor features of pleomorphic carcinomas (descriptively referred to as adenocarcinomas with pleomorphic features) (A: HE, ×400; B: HE, ×400; C: HE, ×200). Histologically, solid growth pattern predominates. However, also gland forming structures were seen (arrows). Cases showed interspersed enlarged “pleomorphic” tumor cells (B, C, arrowheads) which were aggregated in focal tumor areas in some cases (C, bottom). (D) TTF1 was absent or weakly and unevenly expressed. All samples expressed strongly cytokeratin 7 but were negative for p40 and neuroendocrine markers (not shown) (D, TTF1 immunohistochemistry, ×400). (E) Fluorescence in situ hybridization reveals more than 10 MET gene copies on average per tumor cell (orange: centromere 7) (MET FISH, ×630).

Discussion

In this study we describe a novel subgroup of NSCLC patients which is defined by the highest unequivocal level of MET amplification. We examined a prospective series of unselected consecutive NSCLC samples by FISH. Based on a comprehensive descriptive approach, we defined the cut-off numerically at the 90th percentile of the average MET gene copy number per tumor cell. Among all tested parameters, i.e., MET/CEN 7 ratio, average gene copy number and percentages of tumor cells with ≥4, ≥5 and ≥15 MET gene copies per tumor cell, average MET gene copy number showed the broadest numerical range. Therefore, the MET top-level amplification category was defined by ≥10 MET gene copies per tumor cell on average.

Furthermore, we correlated cases fulfilling our newly described criteria with clinical and morphologic data and demonstrate a peculiar phenotype of these patients. A major finding is related to clinical outcome: Among all characteristics tested—including stage, histotype and molecular subtypes—patients with MET top-level amplification suffer from the shortest survival and the highest likelihood to die from their cancer. Thus, MET top-level amplification is an independent prognostic factor which is even unrelated to clinical stage at initial diagnosis. In our series the prevalence of MET top-level amplification which was 2.1%. This is basically in line with a previous report (18). In a retrospective re-calculation of data from our former publication the prevalence of this subgroup was 0.9%. Combined data (based on 1,066 prospectively and comparably tested patients from that publication and from this study), therefore, indicate that the prevalence of MET top-level amplification is in the range of 1% to 2% among Caucasian patients from Western countries. Furthermore, we could demonstrate that MET top-level amplification is mutually exclusive with EGFR mutations as well as with ALK and ROS1 gene fusions. However, activating KRAS mutations and MET exon 14 skipping mutations can co-occur whereas high PD-L1 expression has not been found in this subgroup.

MET top-level amplified lung cancers seem to show also a specific morphologic phenotype which we describe as poorly differentiated adenocarcinomas with pleomorphic features. We acknowledge that MET top-level samples show some similarities with pleomorphic and/or sarcomatoid carcinomas where MET alterations, including MET mutations and lower levels of amplification, have been already described. However, adenocarcinoma-typic features such as gland formation TTF1 expression and/or KRAS mutation could be demonstrated in our cases which fulfilled the criteria for top-level amplification. Therefore, we feel encouraged to describe our cases as examples of a specific subtype of pulmonary adenocarcinomas rather than sarcomatoid carcinomas. In addition and supporting our recent finding, all patients with top-level amplification from that previous report turned out retrospectively as adenocarcinomas (18).

We are aware that the small size of our MET top-level cohort is a major limitation of our study. However, we provide here first evidence that a peculiar subgroup of NSCLC exists which is characterized by the highest unequivocal MET amplification and a specific morphologic and clinical phenotype. This observation may contribute to a more specific description of “MET amplification” which has been subject to rather vague definitions in the past.

One major finding of our study was the extremely poor prognosis of these MET top-level amplified cancers which constituted the worst prognostic subgroup of all NSCLC histologies in our cohort. This observation was not only unrelated to initial clinical stage but also nearly independent of any form of systemic treatment, including conventional chemotherapy and immune-checkpoint inhibitor treatment. Therefore, we conclude that MET blockade might become a reasonable systemic treatment option for these patients. Although we cannot provide systematic response data of MET inhibitors from our prevalence study, we believe that the newly defined MET top-level amplification may provide a reasonable inclusion criterion for ongoing clinical trials with MET inhibitors in NSCLC. Based on the extremely short survival of these patients, we suggest to give patients with MET top-level amplification early access to anti-MET treatment already in a first line approach. Otherwise, the aggressive biology of these tumors may overcompensate a potential benefit of MET blockade if patients are treated too late. Our limited clinical experience with MET inhibitors obtained in this study may support the hypothesis that early treatment of patients with top-level MET amplification may benefit from treatment with a MET TKI. This, however, needs to be proven by prospective clinical trials. Based on our data, we suggest (I) to establish prospective randomized trials enrolling patients with MET top-level amplified cancers in a first line setting, and (II) to re-analyze subgroups of top-level amplified NSCLC patients from ongoing or terminated trials.

The definition of “MET positivity” is still under debate. Whereas MET mutations are commonly accepted as an actionable target, it is still unclear whether MET amplification may also be actionable. Very recently, Camidge et al. presented an update of the PROFILE 1001 study reporting on MET targeting therapy with crizotinib in 40 NSCLC patients. Patients with high MET amplification [in this study defined by MET/centromere 7 (CEP7) ratio ≥4] showed clinically meaningful antitumor activity with rapid and durable responses. Objective response rates for low level MET/CEP7 ratio (1.8–2.1, n=1), medium level (2.2–3.9, n=14) and high level (≥4, n=20) were 33.3%, 14.3%, and 40.0%, respectively. Best median progression free survival was detected in the high level group (6.7 months), 1.8 and 1.9 months for low and medium level, respectively (23). This observation may point towards a gene dose effect which may be meaningful for a significant effect of these drugs. Moreover, we conclude that only patients with the highest unequivocal MET amplification level may be good candidates for anti-MET treatment. In this context, we need to emphasize that low level MET copy number gains have probably no specific value as a prognostic or predictive biomarker. Many cancers do show slight or moderate increases of MET copy numbers which do not necessarily reflect a specific biologic mechanism in terms of an oncogenic driver.

In summary, we describe a subtype of NSCLC which can be determined by very high MET gene copy number gains (i.e., top-level amplification as defined by ≥10 MET gene copies per tumor cell on average). We provide first evidence that tumors with this particular feature account for 1% to 2% of NSCLC cases and share a common clinical, genetic and morphologic phenotype. Patients with MET top-level cancers suffer from a deadly and aggressive tumor with extremely short overall survival which does not adequately respond to conventional chemotherapy or immune-therapy. MET top-level amplification is mutually exclusive with actionable EGFR, ALK or ROS1 alterations, whereas KRAS and MET mutations may co-occur. These particular tumors show a characteristic morphologic phenotype describable as adenocarcinoma with pleomorphic features.

Preliminary data from clinical trials with MET inhibitors point toward a gene dose effect. Therefore, we suggest including patients with MET top-level amplification specifically in clinical trials, also in first line settings.

The article’s supplementary files as