MET-Dependent Solid Tumors: Molecular Diagnosis and Targeted Therapy

Abstract

MET amplifications, mutations, and fusions are drivers of oncogenesis. More contemporary diagnostic assays can maximize the likelihood of detecting these alterations. de novo MET alterations can serve as primary drivers of cancer growth or as secondary drivers that mediate resistance to targeted therapy for another oncogene such as sensitizing EGFR mutations. Single-agent or combination MET-directed targeted therapy is active in MET-amplified cancers. While the likelihood of benefit increases with increasing levels of MET amplification, no consensus exists on diagnostic cutoffs for MET copy number gains by fluorescence in-situ hybridization or next-generation sequencing. MET mutations include those affecting the kinase or extracellular domains, and those that result in exon 14 skipping biology. The activity of type I or II MET tyrosine kinase inhibitors (TKIs) varies by MET mutation category. Type I MET TKIs bind the kinase ATP pocket in the active state, whereas type II inhibitors bind the inactive state. The biology and response to targeted therapy of MET fusions are underexplored. While MET expression in the absence of a genomic marker of MET dependence is a poor predictor of MET-targeted therapy benefit, MET expression in the context of pathogenic MET alterations may select for response.

INTRODUCTION

Dysregulation of the c-MET tyrosine kinase (hereafter simplified as MET) is an established driver of oncogenesis¹. Compared to many other proto-oncogenes, MET is unique in that three different genomic states can lead to clinically-relevant oncogenesis: amplification, mutation, and fusion. All three of these states present diagnostic challenges in the clinic. Furthermore, these can be identified in two major contexts - as primary or secondary drivers of cancer growth. Primary MET dependence is exemplified by tumors that rely solely on overactive MET signaling to fuel growth. Secondary MET dependence is characterized by reliance on another oncogenic driver (e.g. mutant EGFR) and concurrent dependence on MET. Secondary MET dependence can be de novo or acquired, following the selective pressures of inhibitors directed against the primary driver.

Identifying tumors that are oncogenically addicted to MET is crucial because multiple MET-directed therapeutics are available in the clinic. This has been hindered on a diagnostic level due to (1) the lack of standardized cutoffs and testing methodology for MET-dependent states such as MET amplification that are measured as a continuous variable, and (2) the inability of older assays to more reliably capture both MET copy number gains and the wide variety of MET mutations and MET fusions that lead to oncogenesis. While no MET-directed targeted therapy is currently approved for MET-dependent tumors, several agents have recently gained breakthrough designation from regulatory authorities. This has happened largely secondary to the adoption of more advanced diagnostic technologies that more effectively identify MET-dependent cancers, and the contemporary strategy of molecular enrichment for these tumors on prospective targeted therapy trials.

MET AMPLIFICATION

MET copy number gains can occur either through polysomy or amplification. Polysomy occurs when multiple copies of chromosome 7 that carries MET are present. This can occur through chromosomal or whole genome duplication^10,11. The presence of multiple chromosomes results in an increase in the number of MET copies. With amplification, MET undergoes regional or focal copy number gains without chromosome 7 duplication¹² (Figure 1). In contrast to polysomy, true amplification is more likely to lead to oncogene addiction¹². These findings parallel data in breast cancer where tumors with HER2 copy number gains secondary to polysomy behave similarly to HER2-negative tumors¹³. Focal MET amplification can lead to elevations in MET expression, receptor activation, and ligand-independent downstream signaling in preclinical models^14,15.

Figure 1. MET amplification diagnosis.

(A) The identification of MET gene copy number by FISH only requires a single colored probe (yellow) against MET that is counted to determine the number of copies of the gene. This strategy cannot differentiate polysomy from true focal amplification as the absolute number of chromosomes that contain MET cannot be determined. In contrast, the use of an additional probe targeting centromere 7 (CEP7, blue) allows this determination. The MET/CEP7 ratio thus helps differentiate whole genome duplication/polysomy (low MET/CEP7 ratio) from focal amplification (high MET/CEP7 ratio). (B) Using next-generation sequencing, focal MET amplification can be distinguished from broad chromosomal gains that include MET. In the latter, adjacent genes such as LINC01510 and CAPZA2 are concurrently amplified. Focal MET amplification is associated with a higher likelihood of MET-dependence for oncogenesis.

Diagnosis

Various assays can detect MET copy number changes. These include fluorescence in-situ hybridization (FISH), quantitative real-time polymerase chain reaction (qRT-PCR), and next-generation sequencing (NGS)¹⁶. The latter can be utilized for tumor or plasma circulating tumor DNA (ctDNA) testing. Unfortunately, cutoff points that define MET amplification vary within each assay.

Fluorescence in situ hybridization

FISH is a commonly used technique employing fluorophore-coupled DNA fragments to recognize and tag genomic regions of interest. One or more colored fluorophores may be used during testing. Following fluorophore treatment, the gene sequences of interest in the cell nucleus will fluoresce with one or more probe colors. The number of signals identified in a cell nucleus indicates the number of copies of the gene of interest (Figure 1). Signals from a predetermined number of cells are counted and the number of signals per cell are averaged. Because FISH is performed under a microscope, signals from malignant cells may be differentiated from those of normal cells. While challenging, this can be helpful in samples with low tumor content. However, some tumors may be prone to tissue sectioning artifacts and signals from one cell may overlap another. This reduces the number of evaluable cells and double signals may appear due to overlap.

MET amplification can be defined by FISH in two major ways. The first method relies on determining gene copy number (GCN). Using the Cappuzzo criteria, MET amplification is defined as the presence of 5 or more copies of MET per cell (MET GCN ≥5)^17–19. Alternate definitions include a MET GCN of ≥6²⁰ and a MET GCN of ≥15^21,22. Unfortunately, GCN is unable to distinguish true amplification from polysomy. The second method surpasses this liability by taking the ratio of MET to CEP7 (centromere 7 enumeration probe); a separate fluorophore is used for the latter. Determination of this ratio adjusts for the number of chromosomes present and more accurately identifies MET amplification in the absence of chromosomal duplication. A MET/CEP7 ratio ≥2.0 is commonly used to define amplification^{18,19,23–27}. Others have categorized the degree of amplification into three groups using MET/CEP7 ratios: low ≥1.8 to ≤2.2; intermediate >2.2 to <5, and high ≥5¹².

Next-generation sequencing

Methods for determining MET copy number and cutoffs for MET amplification vary by NGS platform^28–30. Similar to FISH, there is no consensus on a single definition. In general, the sequenced gene of interest is normalized against a comparator such as archived sequences from normal samples. Some platforms go one step further and incorporate normal samples from patients (i.e. peripheral blood) into the normalization in addition to using a comparator²⁹. The subsequent bioinformatic pipelines by which GCNs are called also vary by platform and are often proprietary. Advantages to NGS include the identification of other genomic alterations, in addition to differentiating broad chromosomal gains from focal amplification as different regions on a chromosome are tiled, sequenced, and averaged over the cells sampled.

Two types of assays are used in the clinic: amplicon-based NGS and hybrid capture-based NGS. These differ by the method of tumor DNA enrichment³¹ and are more extensively discussed in a subsequent section of this review. Briefly, hybrid capture-based NGS can more accurately assess of copy number variations of MET and other genes. By contrast, with the amplicon-based approach (in which the amplified regions are limited solely to stretches of DNA flanked by established primers) the genomic territory covered is limited, significant sequence bias can be introduced and sequence replicates cannot be removed, and thus the true sequence coverage depth is affected.

Several technical issues should be considered when utilizing NGS. First, identifying copy number gains/losses is dependent on tumor purity and sample selection as normal cells are often admixed with malignant cells during analysis^28–30. Second, poor DNA quality (e.g. from old tumor samples) can increase noise and make it difficult to accurately analyze copy number alterations. Third, large genomic alterations can present a computational challenge and decrease call accuracy³². Finally, studies of the concordance between MET amplification by NGS and FISH have not been done. This makes it difficult to interpret NGS results considering FISH has been better studied.

Other assays

NGS of plasma ctDNA has been used to detect MET amplification. Methods for calling are similar to that of NGS of tumor^33,34. Although ctDNA is less invasive and has the potential to pick up more tumor heterogeneity than tumor biopsies, plasma calls depend on the sensitivity and resolution of the platform and relies on rates of tumor shedding^33–36. As such, studies have shown that amplification can be missed using this method^33,37,38. qRT-PCR of tumor has also been used to detect MET amplification, although the performance of this assay is not well characterized compared to FISH/NGS^39–47. Again, cutoffs vary and no clear standard definition for MET amplification has been proposed.

Clinical Features

De novo MET amplification

De novo MET amplification is found across a wide variety of solid tumors. It is identified <1%−5% of NSCLC^25,48–51, <1%−10% of gastric cancers^{19,21,23,52–54}, 2–4% of colorectal cancers^55,56, 13% of type I papillary renal cell carcinomas (PRCCs), 3% of type II PRCCs⁵⁷, and at lower frequencies in esophageal carcinoma and hepatocellular carcinoma (HCC)^18,24,58. In NSCLC, studies have not shown a strong association between MET amplification and smoking^48,59. In the TCGA and cBioPortal databases, MET amplification is also found in less well-studied cancer types, including glioblastoma, melanoma, gynecologic cancers, and lymphoma^60–64. In many of these cancers, MET amplification confers a poor prognosis, although these studies were conducted before effective MET-targeted therapies were widely tested^{17,19,21,23,25}. The prognostic implications of MET amplification are unknown against the current backdrop of targeted therapeutics. It is important to recognize that the actual frequency MET amplification varies from study to study and the true frequency of MET amplification within a particular context is challenging to determine. This is unsurprising given that no consensus exists on which assay or cutoff to use, as was discussed extensively in the prior section on diagnosis.

The NSCLC literature has been instrumental in highlighting that, compared to lower-level MET amplification, high-level MET amplification is more likely to indicate oncogenic MET-dependence. In one series, NSCLCs that had high-level MET amplification by FISH (MET/CEP7 ratio ≥5) did not harbor other oncogenic drivers (such as EGFR mutations or ALK fusions), whereas low (MET/CEP7 ≥1.8 to ≤2.2) or intermediate (MET/CEP7 >2.2 to <5) levels of MET amplification were found to overlap more commonly with other oncogenes¹². A recent study of crizotinib (a MET TKI) in MET-amplified NSCLC showed that response was enriched in the intermediate to high MET-amplified tumors compared to those with low-level MET amplification or polysomy (MET GCN >6 and MET/CEP7 <1.8), lending support to the use of MET/CEP7 ratio over MET GCN in evaluating MET-dependency⁶⁵. The cutpoints of the intermediate and high groups were modified to intermediate (MET/CEP7 >2.2 to <4) and high (MET/CEP7 ≥4) to enrich the high MET group for potential responders⁶⁶. In a separate report, focal MET amplification by NGS was thought to more likely represent a true oncogenic driver state compared to broad gains on chromosome 7 that includes MET⁶⁷.

Acquired resistance

Whereas tumors with de novo high-level MET amplification are primarily dependent on the MET pathway for growth, tumors that are reliant on other oncogenes (such as mutant EGFR) can develop secondary dependence on the MET pathway as a mechanism of resistance to targeted therapy⁶⁸. In NSCLC, depending on the cutpoints and assays used, acquired MET amplification is identified in 5–20% of cancers with sensitizing EGFR mutations that develop resistance to 1^st, 2^nd, and 3^rd generation EGFR TKIs^69–72. EGFR-mutant cell lines with MET amplification rely on MET-mediated bypass signaling through the PI3K signaling pathway, and that MET-mediated activation of ERBB3 is needed to initiate PI3K signaling^43,68. MET amplification has also been described as a mechanism of resistance to ALK inhibitors in ALK-fusion positive NSCLC⁷³. Acquired MET amplification has also been identified in unselected colorectal cancers treated with anti-EGFR monoclonal antibodies⁷⁴ and in BRAF V600E-mutant colon cancers treated with EGFR inhibitors^41,75.

When MET amplification occurs as a secondary driver of resistance, this can occur in a subclonal population. Cutoffs for selecting MET-dependency in this context will need to consider that, when employing an assay that interrogates nucleic acids derived from a mixed population of tumor cells, the existence of MET amplification in a subpopulation of cells and its absence in other cells may result in a lower apparent level of MET copy number change or the lack of a detectable change in MET compared to if single cells were sequenced or if FISH were performed⁷⁶. The use of complementary assays such as FISH may thus need to be considered in situations where NGS testing fails to detect subclonal MET copy number changes.

Targeted Therapy

De novo amplification

One of the earliest studies to examine the activity of MET-directed targeted therapy by the degree of MET amplification was PROFILE 1001⁷⁷. This phase I trial with an expansion cohort for MET-amplified NSCLCs examined the activity of crizotinib by varying levels of amplification: low (MET/CEP7 ≥1.8 to <2.2), intermediate (MET/CEP7 ≥2.2 to <5), and high (MET/CEP7 ≥5)⁷⁸. The objective response rate (ORR) was highest (50%) in the high MET amplification group compared to an ORR of 33% and 20% in the low and intermediate groups, respectively. In an update of PROFILE 1001 for which the cutoff between intermediate- and high-level MET amplification was changed to a MET/CEP7 ratio of 4⁶⁶, the best outcomes were consistently achieved in high-level MET-amplified NSCLCs (ORR 40%, median PFS 6.7 months) (Figure 2). A separate clinical trial studied the activity of the selective MET TKI capmatinib in MET-amplified NSCLCs⁷⁹. Tumors were classified by MET GCN: GCN <4, GCN ≥4 to <6, and GCN ≥6. Similarly, the ORR was highest (47%) in tumors with a MET GCN ≥6.

Figure 2. Targeted therapy for MET amplification.

In six prospective phase I/II clinical trials, patients with cancers possessing higher levels of MET amplification or gene copy number derived increased benefit from MET-directed targeted therapy. Trials included patients with non-small cell lung cancer (NSCLC), papillary renal cell cancer (PRCC), and other solid tumors. While cutoffs for MET amplification/gene copy number varied, the cutoff of a MET/CEP7 ratio or MET gene copy number of 4 was chosen for consistency for trials that used FISH; patient-level data were reviewed to calculate response rates. For the only trial which employed next-generation sequencing (NGS), the trial cutoff of MET gene copy number ≥6 was used. The objective response rate is depicted by cancers that fall below these cutoffs (light red circles) and cancers that met or exceeded these cutoffs (red circles). The size of each circle represents the size of the subpopulation within each trial (with the smallest circle representing an n of 2), and each row represents a single trial. ^§MET amplification was determined by FISH. ^ǂMET amplification was determined by NGS.

Finally, the activity of the selective c-MET TKI savolitinib was studied in MET-amplified PRCCs⁸⁰. As opposed to the two prior studies that used FISH, NGS was utilized and MET amplification was defined as ≥6 copies of MET. MET-amplified PRCCs were more likely to respond (ORR 43%) compared to PRCCs with a GCN <6 (ORR 0%). A summary of targeted therapy outcomes by MET amplification status is featured in Table 1.

Table 1.

Targeted therapy outcomes by MET copy number status

Drug	Trial (n)	Amplification Criteria	Assay Used		All Patients	MET Amplification Status
Solid tumors						MET-Amplified
SAR125844193	Phase I (n=72)*	MET GCN >4 and MET/CEP7 ≤2.0a	FISH	ORR	-	17% (5/29)
capmatinib198	Phase I (n=38)€	MET/CEP7 ≥ 2.0 or MET GCN ≥5b	FISH			MET GCN <4	MET 4< GCN <6	MET GCN ≥6
				ORR	0% (0/36)	0% (0/22)	0% (0/6)	0% (0/3)
				95%CI	(0–9.3%)	-	-	-
AMG 337199	Phase I* (n=111)	MET/CEP7 ≥2.0	FISH			MET/CEP7 <4	MET/CEP7 ≥ 4
				mDOR	6.6 mo	-	-
				ORR	10% (11/111)	0% (0/2)	60% (6/10)
				95%CI	(5–17%)	-	-
Gastroesophageal cancers						MET-Amplified
AMG 337200	Phase II (n=60)¥	MET/CEP7 ≥2.0	FISH	mDOR	-	6.0 mo
				95%CI	-	(3.7–16.7)
				ORR	-	18% (8/45)
				95%CI	-	(8%–32%)
foretinib201	Phase II (n=74)	MET/CEP7 ≥2.0	FISH	mPFS	1.7 mo	-
				95%CI	(1.6 –1.8)	-
				ORR	0% (0/71)	0% (0/3)
Hepatocellular carcinoma
capmatinib202	Phase II (n=30)	MET/CEP7 ≥2.0 or GCN ≥5a	FISH	ORR	-	10% (3/30)
				95%CI	-	(2.1–27%)
Non-small cell lung cancers						MET/CEP7 ≥1.8 and ≤2.2	MET/CEP7 >2.2 – <4.0	MET/CEP7 ≥4.0
crizotinib	Phase I66 (n=37)	MET/CEP7 ≥1.8	FISH	mPFS	-	1.8 mo	1.9 mo	6.7 mo
				95%CI	-	(0.8–14.0)	(1.3–5.5)	(3.4–7.4)
				ORR	-	33% (1/3)	14% (2/14)	40% (8/20)
				95%CI	-	(0.8–91%)	(2%–43%)	(19%–64%)
	Phase II203 (n=25)*	MET GCN ≥6 and IHC 2+/3+	FISH			MET/CEP7 ≥1.8 and ≤2.2	MET/CEP7 >2.2 –<5.0	MET/CEP7 ≥5.0
				mPFS	5.0 mo	-	4.4 mo	NE
				95%CI	(2.7–7.3)	-	(2.9–5.9)	-
				ORR	31% (5/16)	-	36%	0% (0/2)
				95%CI	(5.2–71.4%)	-	(6.0–77%)	-
	Phase II65 (n=25)*	MET GCN ≥6 and IHC 2+/3+	FISH	mPFS	3.2 mo	-	-	-
				95%CI	(1.9–3.7)	-	-	-
				ORR	32% (8/25)	-	-	-
capmatinib79	Phase I (n=44)	MET/CEP7 ≥2.0 or MET GCN ≥5c,d	FISH			MET GCN <4	MET 4< GCN <6	MET GCN ≥6
				ORR	20% (11/55)	0% (0/17)	17% (2/12)	47% (7/15)
				95%CI	-	-	(2–48%)	(21–73%)
Non-small cell lung cancers (EGFR-mutant)						MET-amplified
osimertinib/ savolitinib 300 mg daily204	Phase I (n=36)	Previously treated with 1st/2nd generation EGFR-TKI and T790M negative; MET/CEP7 ratio ≥2 or MET GCN ≥5 by FISH or MET GCN ≥5 by NGSd	FISH or NGS	mPFS	-	9.1 mo
				95%CI	-	(5.4–12.9)
				ORR	-	64%
				95%CI	-	(46%–79%)
osimertinib/ savolitinib 300 or 600 mg daily204	Phase I (n=51)	Previously treated with 1st/2nd generation EGFR-TKI and T790M negative; MET/CEP7 ratio ≥2 or MET GCN ≥5 by FISH or MET GCN ≥5 by NGSd	FISH or NGS	mPFS	-	9.0
				95%CI	-	(5.5–11.9)
				ORR	-	65% (33/51)
				95%CI	-	(50%–78%)
osimertinib/ savolitinib 300 or 600 mg daily204	Phase I (n=18)	Previously treated with 1st/2nd generation EGFR-TKI and T790M positive MET/CEP7 ratio ≥2 or MET GCN ≥5 by FISH or MET GCN ≥5 by NGSd	FISH or NGS	mPFS	-	11.0 mo
				95%CI	-	(4.0-NE)
				ORR	-	67% (12/18)
				95%CI	-	(41–87%)
osimertinib/ savolitinib 300 or 600 mg daily204	Phase I (n=69)	Previously treated with 3rd generation EGFR-TKI; MET/CEP7 ratio ≥2 or MET GCN ≥5 by FISH or MET GCN ≥5 by NGSd	FISH or NGS	mPFS	-	5.4 mo
				95%CI	-	(4.1–8.0)
				ORR	-	30% (21/69)
				95%CI	-	(20–43%)
gefitinib/ tepotinib205	Phase II (n=12)	MET/CEP7 ratio ≥2 or MET GCN ≥5; after prior EGFR-TKI and T790M negative	FISH	mPFS	-	16.6 mo
				90%CI	-	(8.3-NE)
				ORR	-	67% (8/12)
				90%CI	-	(39%–88%)
gefitinib/ savolitinib206	Phase I (n=44)	MET/CEP7 ratio ≥2 or MET GCN ≥5; after prior EGFR-TKI	FISH	ORR	-	25% (11/44)
gefitinib/capmatinib82	Phase II (n=100)	GCN ≥5 or IHC 2+/3+ in >50%; then, GCN ≥5 plus IHC 2+/3+; then, GCN ≥4 or IHC 3+; MET/CEP7 ≥1.8	FISH			MET GCN <4	MET 4 ≤ GCN <6	MET GCN ≥6
				mPFS	5.5 mo	3.9 mo	5.4 mo	5.5 mo
				95%CI	(3.8–5.6)	(3.7–5.6)	(3.7–7.5)	(4.2–7.3)
				ORR	29% (29/100)	12% (12/41)	22% (4/18)	47% (17/36)
Papillary renal cell carcinoma						MET GCN <6 and not MET-driven§	MET GCN ≥6
savolitinib80	Phase II (n=79)*	Focal MET GCN ≥6	NGS	ORR	10% (8/79)	0% (0/32)	43% (3/7)

MET positivity also defined as

^aimmunohistochemistry (IHC) ≥50% with 2+/3+

^bH-score ≥150 or IHC ≥50% with 2+/3+ or H-score ≥50 for hepatocellular carcinoma and glioblastoma

^cH-score ≥150 or IHC ≥50% with 2+/3+

^dIHC ≥50 with % 3+.

^*Available patient level data with MET gene copy numbers or MET/CEP7 ratio and response results were used to produce these results.

^ǂResponse and outcome data derived from dosing regimen #3 cohort of the study, but response rates by MET amplification status in this cohort was not available.

^€Only data from patients receiving a stable dose in the dose expansion cohort were used.

^$One patient with partial response had both a MET exon 14 alteration and MET copy number 2.3.

^¥Only patients from cohort 1 was used.

^§Not MET-driven defined as no MET focal amplification, no MET kinase domain mutation, no HGF amplification, and no chromosome 7 gain. CI, confidence interval; ORR, objective response rate; mPFS, median progression-free survival; mDOR, median duration of response; GCN, gene copy number; NE, not estimable; NGS, next-generation sequencing; FISH, fluorescence in-situ hybridization; IHC, immunohistochemistry; mo, month(s); HCC, hepatocellular carcinoma; GBM, glioblastoma; PRCC, papillary renal cell carcinoma

Acquired resistance

In tumors with primary dependence on another oncogene and secondary dependence on MET, combination therapy targeting both the primary driver and MET is active. For example, the combination of an EGFR TKI and a MET TKI is active in EGFR-mutant NSCLCs that acquire MET amplification after progression on a prior EGFR TKI. In a trial of osimertinib and capmatinib, the ORR was 25%⁸¹ in EGFR-mutant NSCLCs that acquired MET amplification (GCN ≥5 or MET/CEP7 ratio ≥2 by FISH) after progression on osimertinib. In a trial of gefitinib and tepotinib, the ORR was 67% in EGFR-mutant NSCLCs that acquired MET amplification (GCN ≥5 or MET/CEP7 ratio ≥2 by FISH) after progression on a 1^st generation EGFR TKI. Interestingly, the degree of activity of combination therapy also increases with increased MET copy number (Figure 1). One study of EGFR-mutant NSCLCs with MET amplification⁸² evaluated the activity of gefitinib and capmatinib by MET GCN: GCN <4, GCN ≤4 to <6, and GCN ≥6. The ORR was highest (47%) in patients with GCNs ≥6. Other agents such the EGFR-MET bispecific antibody JNJ-372 have shown promising activity in EGFR-mutant cancers that have become resistant to EGFR TKI therapy and the activity of these drugs should be characterized further in tumors with acquired MET amplification.

Taken together, higher levels of MET amplification can predict for increased benefit from MET-directed targeted therapy. This applies to (1) single-agent MET inhibitor therapy for MET-amplified cancers, and (2) combination therapy that includes a MET inhibitor for cancers that develop MET amplification as a mechanism of resistance to targeted therapy directed against the primary oncogenic driver. The need for standardized definitions of MET amplification thus has substantial implications not only for the diagnosis of MET-dependent cancers, but for the identification of cancers that are more oncogenically addicted to MET with a higher likelihood of benefit from MET-directed targeted therapy.

MET MUTATION

Genomic diversity

Activating mutations can occur at a diverse range of positions along MET. These include alterations involving the kinase domain, intronic splice sites that flank exon 14, and the extracellular domain.

Kinase domain mutations

MET mutations were first described in hereditary HPRCC⁸³. These germline MET mutations include V1092I, H1094R/Y M1131T, V1188L, V1220I, M1250T, and D1228H/N/V. MET mutation increases MET kinase activity and leads to phenotypic transformation or tumor focus formation in vitro⁸⁴. It also induces tumor formation in mouse models in vivo^84,85. Later studies discovered that MET mutations occur in sporadic PRCCs. These are found in up to 15% of cases^57,86,87, predominantly in type 1 and less commonly in type 2 PRCC⁵⁷. The spectrum of MET mutations in sporadic PRCCs includes V1092I, H1094l/R/Y, N1100Y, H1106D, M1131T, V1188L, L1195V, V1220I, D1228H/N/V, Y1230H, Y1230A/C/D/H, Y1235D, and M1250T/I^83,88,89 (Figure 3). Interestingly, some of these mutations are less active than others and concomitant MET amplification may be necessary to increase oncogenesis^85,86,90. Activating MET mutations have been found in other cancers, including HCC and head and neck cancers^91,92. In the latter, MET Y1235D mutation is found in up to 14% of cases⁹³. In addition to being found de novo, MET kinase domain mutations can emerge as a mechanism of acquired resistance to MET TKI therapy. In MET exon 14 NSCLCs (discussed in detail below), MET Y1230C, Y1230H, D1228H, and D1228N have been found to mediate resistance to prior MET TKI therapy with crizotinib^94–98 by disrupting drug binding⁹⁴. MET kinase domain mutations have also emerged as resistance mechanisms in EGFR-mutant lung cancers who were treated with a combination of an EGFR TKI and a MET TKI for MET-amplification-mediated acquired resistance⁹⁹.

Figure 3. MET mutations.

The MET protein consists of the extracellular semaphorin (SEMA) domain in red, a plexin-semaphorin-integrin (PSI) domain in purple, four immunoglobulin-plexin-transcription (IPT) repeats in yellow, a juxtamembrane domain (encoded by exon 14) in blue, and a kinase domain in green. A) MET exon 14 splice site alterations result in exon 14 exclusion. This results in the absence of the ubiquitin binding site of the juxtamembrane domain, impaired MET degradation, and increased MET signaling. B) Missense mutations in the juxtamembrane domain prevent spliceosome binding or modify the Y1003 ubiquitination site. These ultimately recapitulate the biology of MET exon 14 splice site alterations. C) Mutations in the kinase domain lead to the increased activation of the MET kinase and can be associated with conformational changes that favor the xDFG-out state. D) Other mutations can occur in the semaphorin domain where hepatocyte growth factor, the ligand for MET, binds. The impact of these mutations on MET function are unclear.

MET exon 14-alterations

When MET is activated, the Y1003 residue of the juxtamembrane domain that is encoded by exon 14 is phosphorylated. This mediates c-Cbl E3-ligase binding and subsequent ubiquitination and degradation of the MET receptor¹⁰⁰. MET exon 14 alterations are a heterogeneous group of MET mutations that convergently interfere with this process. While MET exon 14 alterations were first identified in small cell lung cancer (SCLC)¹⁰¹, later studies revealed that these mutations are more commonly found in NSCLC, with a prevalence of 3–4%^102–105. These are further enriched in sarcomatoid carcinomas (occurring in 9–22% of cases), an aggressive NSCLC variant that can be highly resistant to chemotherapy¹⁰⁶. MET exon 14 alterations occur at lower frequencies in other cancers, including gastric cancers and neuroblastomas^103,107,108. Beyond being found de novo in various malignancies, MET exon 14 alterations have been shown to mediate resistance to EGFR TKIs in EGFR-mutant NSCLC^109,110. For NSCLCs with MET exon 14 skipping, a higher proportion of patients have a smoking history compared to those whose cancers harbour other drivers such as ALK/ROS1/RET fusions, although several independent reports have noted that a proportion of patients with these alterations are never smokers, consistent with the GEOMETRY trial results^111–113.

Most MET exon 14 alterations are mutations that interfere with RNA splicing. In the wild-type state, the intronic regions of MET pre-mRNA are spliced out before the transcript is translated into a protein. Mutations that occur in regions that flank MET exon 14 (i.e. the polypyrimidine tract, splice donor, or splice acceptor regions) effectively disrupt normal splicing and result in exon 14 being skipped¹⁶. The loss of the encoded juxtamembrane domain leads to the loss of the Y1003 ubiquitin binding site on MET. Consequently, MET degradation is decreased and MET expression increases, thus driving oncogenesis^114–116. Most of these splice site mutations take the form of insertions or deletions with a wide range of sizes. In addition, missense mutations that result in D1010 substitution (i.e. D1010H/N/Y) disrupt splicing¹¹⁷.

MET mutations that do not directly affect splicing can recapitulate a similar biology. For instance, mutations that lead to Y1003 substitution (i.e. Y1003F/N/S) interfere with c-Cbl E3-ligase binding¹¹⁵. Like MET exon 14 RNA splice site mutations, Y1003 substitution transforms normal cells, leads to increased cell proliferation, and promotes tumor growth^114,115. Large deletions that encompass exon 14 result in the loss of the juxtamembrane domain that carries Y1003.

Other mutations

The semaphorin (SEMA) domain of MET protein interacts with HGF (the ligand of MET) and is involved in dimerization that leads to receptor activation¹¹⁸. Mutations in this domain include E34K, E150D, H150Y, L269V, L299F, S323G, M362T, N375S, and C385Y (Figure 3)^{102,119–122}. N375S is the most common of these and occurs in 3–14% of NSCLCs¹²³. Whether SEMA domain mutations (particularly N375S) are activating is a point of contention. While N375S mutation has been shown to participate in carcinogenesis by activating downstream SRC and ERK1/2 signaling¹²⁴, data have shown that some SEMA domain mutations decrease the binding affinity of HGF and are found in individuals without cancer^123,125.

Diagnosis

DNA-based sequencing

As MET exon 14 alterations are highly heterogenous, it is important to select an NGS assay that is poised to capture this wide variety of mutations¹⁰³. As mentioned previously, amplicon-based NGS and hybrid-based NGS are the two main types of assays used in the clinic. In amplicon-based NGS, genes of interest are sequenced using an established set of DNA primers that flank defined genomic regions¹²⁶. Although this approach can offer a shorter turn-around time and improved capture of difficult regions, it is less tolerant of sequence variability and more prone to sequencing bias and allelic dropout, especially for repetitive sequences. Many MET exon 14 alterations, particularly indels that result in splicing defects, can lie outside these amplified regions and be missed^127,128. Furthermore, mutations such as large insertion/deletions may involve a primer binding site and interfere with primer binding; this likewise prevents amplicon-based assays from detecting these alterations¹²⁶. Consistently, studies have demonstrated that a significant fraction of MET exon 14 alterations can be missed by amplicon-based NGS^127,128.

In contrast, hybrid capture-based NGS platforms avoid some of the issues associated with amplicon-based NGS¹²⁸. Tumor DNA is sheared, captured by long oligonucleotide baits, and then amplified. There is overlap between these unique DNA fragments so duplicate sequences can be identified and removed and the challenges of sequencing repetitive sequences can be adjusted for with careful bait design. By tiling over appropriate intronic regions, hybrid capture-based NGS forestalls primer binding issues and can detect mutations that are further into the intron. The use of a capture based approach also corrects for some of the sequencing bias and allele drop-out issues that occur with amplicon-based NGS¹²⁶. With respect to calling missense mutations, hybrid capture-based NGS also outperforms amplicon-based NGS. Although both platforms usually have high coverage depth of genes of interest to ensure accuracy, amplicon-based platforms are more commonly associated with false-positives or false-negatives consequently affecting the threshold of sensitivity that the assay can attain.

RNA-based sequencing

It is important to point out that while DNA-based sequencing detects mutations that lie within regions that are predicted to result in MET exon 14 skipping, it cannot confirm the absence of the exon that occurs at the RNA level. As such, RNA-based sequencing platforms can complement DNA-based NGS¹²⁸. Given the diversity of MET exon 14 splice site alterations, it can be difficult to interpret whether some mutations identified by DNA-based NGS truly result in exon 14 skipping. Interpretations can be made based on proximity to splice donor/acceptor regions, but this becomes challenging with alterations that occur deeper into the intron. In contrast, RNA sequencing directly identifies the loss of exon 14 transcription. In addition, the challenges associated with the need to sequence large intronic regions in DNA-based NGS are avoided with RNA-based sequencing methods, given the absence of intronic regions in RNA. Furthermore, whereas the effect of a new mutation on METex14 can be difficult to interpret with DNA-based platforms, RNA sequencing directly demonstrates whether exon 14 has been transcribed.

Assays using anchored multiplex PCR (AMP, Archer Technologies) have been utilized to identify MET exon 14 alterations. Using this method, RNA is used as template to generate complementary DNA (cDNA) from tumor¹²⁹. As the cDNA is sequenced and amplified, molecular barcodes are incorporated to quantify sequences and correct sequencing errors. The loss of exon 14 is then identified in cDNA. One study examined the utility of AMP RNA sequencing to identify MET exon 14 alterations in 232 NSCLCs that were “driver-negative” by DNA-based hybrid capture NGS¹³⁰. Six MET exon 14 skipping alterations were identified by RNA sequencing that DNA-based NGS did not detect. An analysis of the genomic sequences identified by DNA-based NGS was then retroactively performed. In five of six cases, novel MET exon 14 splice site alterations were detected after further manual review¹²⁷. In the remaining case, MET mutation was not detected, suggesting that MET exon 14 skipping occurred by another mechanism.

Molecular counting using the nCounter system (Nanostring Technologies) has also been used to identify MET exon 14 loss at the RNA level. The system uses probes to detect MET transcripts of interest, utilizing a fluorescently tagged 5’ reporter probe and biotinylated 3’ capture probes¹³¹. In contrast to AMP, it does not require the reverse transcription of RNA to cDNA and does not require amplification. Instead, target-specific color-coded probes detect the sequences of interest and the color codes are counted and tabulated using a digital analyzer for image acquisition¹³¹. Further studies of this method are needed, particularly as it has not been compared to AMP for the detection of MET exon 14 skipping^132,133.

Targeted therapy

Kinase domain mutations

MET-directed targeted therapy is active against select MET kinase domain mutations. It is critical to recognize that the activity of various MET TKIs against specific mutations is variable. Specifically, while type II MET inhibitors such as cabozantinib and foretinib have preclinical activity against several kinase domain mutations (e.g. D1228N, M1250T, H1094Y/L¹³⁴), type I MET inhibitors such as crizotinib do not have substantial activity against these alterations. As such, MET kinase domain mutations have emerged as mechanisms of resistance to crizotinib in MET-amplified and MET exon 14-altered cancers. The ability of these mutations to switch the conformation of the MET kinase from an active (xDFG-in) to an inactive (xDFG-out) conformation may be contributory as type I MET inhibitors preferentially bind to the former while type II MET inhibitors bind to the latter¹³⁵.

Data in tumors with de novo MET kinase domain mutations is largely limited to PRCC. In a study of foretinib in patients with HPRCC with germline MET mutations, the ORR was 50%¹³⁶. In cancers where MET kinase domain mutations were acquired as mechanisms of resistance to prior MET TKIs, data is limited to NSCLCs^94–96,98. Patients were identified with cancers that developed MET kinase domain mutations as resistance mechanisms to the combination of an EGFR TKI and crizotinib (administered to EGFR-mutant cancers with MET amplification-mediated resistance to a prior EGFR TKI)¹³⁷. TKI type switching to the type II MET inhibitor cabozantinib resulted in the re-establishment of disease control.

MET exon 14 alterations

As opposed to kinase domain mutations that can alter the conformation of the MET kinase, MET exon 14 alterations harbor a MET kinase domain that is presumably similar to the wild-type MET receptor kinase domain¹⁰⁰. Thus, both type I and type II MET TKIs are poised to inhibit MET exon 14-altered cancers and have shown preclinical activity in these models. The clinical activity of MET inhibition in MET exon 14-altered NSCLCs was first prospectively established with crizotinib, a type Ia multikinase MET inhibitor⁷⁷. The phase I PROFILE 1001 study accrued patients with advanced MET exon 14-altered NSCLCs to an expansion cohort of the trial. The ORR was 32% and the median PFS was 7.3 months (Table 2). These results supported the inclusion of crizotinib in the NCCN guidelines for this indication. It also supported the US FDA’s decision to grant crizotinib Breakthrough Therapy Designation for MET exon 14-altered NSCLCs in 2018¹³⁸.

Table 2.

Targeted therapy in MET exon 14-altered lung cancers

Inhibitor Type	Type Ia	Type Ib
Drug	Crizotinib77	Capmatinib113	Tepotinib140	Savolitinib207
Objective response rate (ORR)	32% (95% CI 21–45) n=21/65			52% (95% CI NR) n=17/31
ORR by treatment
Naïve (no systemic therapy)  Pre-treated (no prior c-MET therapy)		68% (95% CI 48–84) n=19/28 41% (95% CI 29–53) n=28/69	44% (95% CI 22–69) n=8/18 45% (95% CI 28–64) n=15/33
ORR by mutation type
Insertion/deletion  Point Mutation	0% (95% CI NR) n=0/4 36% (95% CI NR) n=12/33	NR	NR	43% (95% CI NR) n=6/14 59% (95% CI NR) n=10/17
ORR by splice site
Splice acceptor  Splice donor	31% (95% NR) n=5/16 32% (95% CI NR150) n=12/37	NR	NR	42% (95% CI NR) n=5/12 58% (95% CI NR) n=11/19
ORR by MET amplification
Concurrently amplified  Wild-type	0% (95% CI NR) n=0/2 NR (95% CI NR) n=NR	NR	NR	100% (95% CI NR) n=5/5 35% (95% CI NR) n=8/23
Median progression-free survival	7.3 months (95% CI 5.4–9.1)		12.3 months (95% CI 6.3-NR)	NR
Median progression-free survival
Naïve (no systemic therapy)  Pre-treated (no prior MET TKI)		9.7 months (95% CI 5.5–13.9) 5.4 months (95% CI 4.1–7.0)

NR, not reported

Newer agents such as the selective type Ib MET inhibitors such as capmatinib, tepotinib, and savolitinib have since been explored in MET exon 14-altered NSCLCs^113,139,140 (Table 2). Notably, selective MET inhibitors such as capmatinib are more potent compared to crizotinib¹¹⁷. In the GEOMETRY trial, the ORRs with capmatinib were 68% and 41% in treatment-naïve and chemotherapy pre-treated patients, respectively¹¹³. In the VISION trial, the ORR with tepotinib was 44%¹⁴⁰. In a phase 2 trial of savolitinib, the ORR was 52%¹³⁹. The durability of disease control and relative toxicity of these agents compared to crizotinib has not yet been well characterized.

Given that the response rates to MET TKIs vary widely, several investigators have attempted to identify subgroups within MET exon 14-altered NSCLCs that are less likely to respond to therapy^77,139. The PROFILE 1001 study of crizotinib first noted that response did not significantly vary by MET exon 14 alteration region (splice acceptor vs donor site involvement), mutation type (insertion/deletion vs base substitution), or concurrent MET amplification⁷⁷ (Table 2). This was subsequently confirmed by studies of selective MET inhibitors such as savolitinib¹³⁹.

In terms of acquired resistance in patients with initial benefit, on-target mechanisms such as MET amplification and MET kinase domain mutations have been identified. While the role that HGF plays in resistance remains to be determined, HGF amplification has been found in the setting of acquired resistance^97,141. While preclinical data supporting the role of the latter in MET exon14-altered NSCLCs is not available, in models of MET-amplified HCC cancer cell lines, HGF can reduce the sensitivity to MET TKIs in MET-amplified cancer cell lines and patient-derived xenografts¹⁴².

MET FUSION

Clinicopathologic features

The MET gene was originally identified as an oncogene after chemically transformed osteosarcoma cell lines were found to have the TPR-MET fusion¹⁴⁷. MET fusions were thereafter identified in gastric cancers, thyroid carcinoma, PRCC, lung adenocarcinoma, HCC, glioma, and sarcoma. The exact frequency within these cancers is poorly defined, although these are enriched in gliomas and found in 12% of cases^148,149. Beyond TPR-MET, multiple other fusions have been identified^{60,61,105,148,150–161} (Figure 4). MET fusion can occur through intrachromosomal (e.g. PTPRZ1-MET, CLIP2-MET, CAPZA2-MET, ST7-MET) or interchromosomal fusion¹⁵⁵, each in about half of cases. In pediatric glioblastomas, intrachromosomal fusion appears to be more common and frequently involves PTPRZ1¹⁴⁸. MET fusion can also result from paracentric or pericentric inversion - the latter appears to be preferred.

Figure 4. MET Fusion.

A wide variety of MET fusions have been identified. These fusions can result in constitutive MET activation in different ways. The 5’ upstream partners CLIP2, TFG, KIF5B, BAIAP2L1, C8orf34, TPR have coiled-coil domains that promote chimeric oncoprotein dimerization. Other domains (e.g. the MENTAL domain of STARD3NL) can mediate alternate methods of homodimerization. The 3’ MET gene typically includes the kinase domain, however, fusions that include the juxtamembrane domain, or larger regions of the gene have been identified. Interestingly, certain fusions such as TPR-MET result in the exclusion of exon 14. The biology of this fusion is thus thought to be similar to that of MET exon 14-altered cancers. The exons at which the breakpoints for MET can occur are noted in the diagram.

These fusions often include the kinase domain on exon 15^{148,150,151,153,154,160,162,163} and many upstream partners provide dimerization domains, resulting in ligand-independent constitutive MET activation. In addition, select fusion events (e.g. TPR-MET) have been found to exclude exon 14, enabling behavior similar to MET exon 14 alteration biology¹⁶². Interestingly, fusions that include exon 14 appear (e.g. KIF5B-MET, PTPRZ-MET) to be less oncogenic than fusions that exclude exon 14¹⁴⁸. In PTPRZ-MET, the PTPRZ promoter is fused to the full-length MET gene that includes the MET dimerization domain on exon 2; this results in both MET overexpression and increased downstream signaling activation^148,164.

Diagnosis and targeted therapy

A number of assays can detect MET fusions. These include FISH, RT-PCR, and NGS^{148,150,165,166}. How well each of these performs has not been well-studied and complex/novel rearrangements can make FISH inadequate for detecting many MET fusions¹⁶⁷. This section will thus focus primarily on NGS, the increasingly preferred assay in the clinic. DNA-based NGS can reliably detect a wide variety of MET fusions, although several general features of fusions make it difficult for even DNA-based hybrid capture NGS to capture all events¹⁶⁸. First, repetitive intronic DNA sequences can occur in fusions breakpoints. Because these can also occur at other sites of the genome and hybrid capture produces short reads, using baits can result in reads that are unmappable to the reference genome³⁰. As a result, these baits are excluded in contemporary assays, increasing the chance of missing genomic fusion breakpoints. Second, the intronic regions of select fusion partners can be prohibitively long, making it challenging and impractical to tile these sequences¹³⁰. Third, DNA-based NGS is limited in its ability to detect novel gene fusion partners^167,169. As discussed earlier, these issues can be avoided with RNA-based AMP NGS^130,169,170. In support of this, the same study of DNA-based NGS “driver-negative” NSCLCs identified actionable gene fusions in 12% (27 of 232) of cases using RNA-based NGS¹³⁰. This suggests that AMP can complement DNA-based NGS in detecting MET fusions, especially in settings where no driver is found.

The utility of MET-directed targeted therapy has been incredibly underexplored in MET fusion-positive cancers. MET TKIs induce apoptosis in TPR-MET-transformed cell lines¹⁷¹ and isolated clinical responses to crizotinib have been described in MET fusion-positive lung adenocarcinomas and gliomas^148,155,172. In the PROFILE 1001 study, one patient was included in the MET exon 14 cohort given that his cancer harbored a MET fusion that resulted in exon 14 skipping⁷⁷. A confirmed response was achieved. Multiple clinical trials are underway to evaluate the efficacy of MET TKIs in cancers including those with MET fusions (www.clinicaltrials.gov: NCT02978261, NCT03993873, NCT01639508).

MET OVEREXPRESSION

The role of MET expression in oncogenesis should be considered in a number of contexts. First, MET can be transcriptionally induced in cancer cells in the setting of hypoxia/inflammation to activate proliferation, decrease apoptosis, and promote migration. Tumors can thus rely on MET signaling even in the absence of a genomic driver such as MET amplification, mutation, or fusion. Such states could theoretically be addressed by MET-directed targeted therapy although the clinical data in this setting with monoclonal antibodies has been disappointing. Thankfully, newer strategies such as biparatopic antibodies, antibody combinations, and antibody-drug conjugates are being explored. Second, MET can be overexpressed in cancers that harbor an activating genomic signature, including those with primary/secondary MET amplification or MET exon 14 alterations.

Diagnosis

Immunohistochemistry

A number of MET IHC antibodies have been utilized. These include monoclonal antibodies [e.g. SP44 (Ventana), cMET (R&D systems), MET4 (Dako)], polyclonal antibodies (e.g. polyclonal MET AF276 (R&D systems) and antibodies to phosphorylated MET [pMET Y1349 (Epitomics)]^173–177. Of these, SP44, a rabbit monoclonal anti-total MET antibody clone, is commonly used. The comparative performance of these antibodies is not known. The extent and intensity of IHC staining as assessed by a pathologist and provides a semi-quantitative measure of MET protein expression. Various scoring systems have been used to define both MET expression and overexpression by IHC^48,175,178. Most commonly, staining is based on a 0–3+ scale corresponding to negative (0), weak (1+), moderate (2+), or strong (3+)¹⁷⁵. Staining that is 1+ or greater provides evidence for MET expression. A common cutoff for MET overexpression is 2+ in at least 50% of the cells (MetMab criterion)¹⁷⁵. Another scoring system, the H-score, multiplies the percentage of cells with 1+, 2+, or 3+ staining by the staining intensity score¹⁷⁹. H-scores range from 0–300; ≥200 usually denotes overexpression but cutpoints vary^48,180. Investigators have also used a median H-score (of the range of H-scores obtained from samples exclusively within a given study) as a cut point for overexpression; this approach makes standardization across studies difficult.

Mass spectrometry

Selected reaction monitoring mass spectrometry (SRM-MS) utilizes serial steps of ionization and fragmentation of tumor proteins to quantify the mass of the MET protein per sample (in amol/μg). SRM-MS can measure MET from formalin-fixed paraffin-embedded tissue and maintains reproducibility in samples a year old and beyond¹⁸¹. Compared to IHC, SRM-MS can be less dependent on interobserver bias and may detect lower levels of expression. However, SRM-MS cannot differentiate between tumor and normal tissue and MET quantification may be influenced by admixed stroma or inflammatory infiltrates. Furthermore, SRM-MS is more complicated than IHC and requires more expensive technology, making it less widely adopted. IHC is more routinely used in diagnostic pathology laboratories while SRM-MS use has largely been investigational.

Screening

The presence of MET overexpression has been investigated as a screening tool for activating MET alterations. Unfortunately, MET overexpression has not reliably detected MET amplification or MET exon 14 alterations and there is little data on MET fusion detection⁵⁹. This contrasts with ALK-rearranged NSCLC where ALK overexpression by IHC strongly correlates with the presence of an ALK rearrangement by FISH¹⁸².

MET amplification

MET IHC overexpression has not been strongly correlated with MET amplification^183–185. This may be secondary to the inclusion of lower levels of MET amplification that do not result in substantial protein expression, or expression may be variably modulated by post-transcriptional/translational factors. In NSCLC, MET amplification (MET/CEP7 ratio >2.2) was only detected in one of 74 patients (1%) with MET overexpression (H-score ≥ 200) in a series from the Lung Cancer Mutation Consortium (LCMC)¹⁸⁶. In gastric cancer, a MET IHC H-score of 150 had a 75% sensitivity and 78% specificity to detect MET amplification (MET/CEP7 ratio >2.0 and GCN >4.0)¹⁸¹. All the above studies utilized FISH and there is little data on NGS.

MET mutation

Interestingly, while MET exon 14-altered NSCLCs are expected to overexpress MET, not all tumors are MET-positive by IHC or SRM-MS^97,112,187. In 25 patients with MET exon 14-altered lung cancer, only 16 patients (64%) were MET IHC positive (2+ and 3+) and about a third of cases were not found to express MET by SRM-MS¹⁸⁷. Similar to MET-amplified cancers, the expression of MET may be modulated by post-transcriptional/translational factors.

It is thus not surprising that MET overexpression does not highly correlate with the presence of MET exon 14 alteration. In the LCMC series discussed above, MET exon 14 alteration was only detected in two of 74 patients (3%) with MET overexpression (H-score ≥ 200)¹⁸⁶. The sensitivity and specificity of MET IHC overexpression for MET exon 14 alterations is variable. IHC had a 90% sensitivity and 47% specificity for predicting MET exon 14 mutation in NSCLC¹⁸⁸, and 20% sensitivity and 83% specificity in sarcomatoid lung carcinoma¹⁸⁰. IHC has not routinely been explored as a screening tool for other MET mutations (e.g. kinase domain mutations) that do not result in exon 14 alteration.

Targeted therapy

In isolation, MET overexpression is not consistently predictive of benefit from MET-directed therapy. Reasons include the challenge of defining expression versus overexpression for a continuous variable, and overexpression may not be not equivalent to a MET-dependent state. While many cancers overexpress MET by IHC, the frequency of overexpression is variable and dependent on cutoff point used. For example, this ranges from 24–66% in NSCLCs^184,188 and 28%−63% in gastric cancers^23,52,189. Multiple therapeutic anti-MET antibodies (onartuzumab, emibetuzumab)^190,191, anti-hepatocyte growth factor (HGF) antibodies (ficlatuzumab, rilotumumab)^178,192 and TKIs (crizotinib, cabozatinib, tivantinib, SAR125844, tepotinib)^193–196 have been tested in clinical trials. The overall activity of these drugs as monotherapies for MET-overexpressing cancers was low (Table 3). For example, in a phase III trial for MET-overexpressing HCCs, there was no difference (p=0.81) in the median OS between patients who received tivantinib (8.4 months) versus placebo (9.1 months). In the same population of MET-overexpressing HCCs, tepotinib resulted in a short median PFS of 2.8 months and the degree of MET IHC positivity did not select for improved activity. Given the limited therapeutic success with antibody-directed therapies, many drug development programs combined these therapies with chemotherapy, or EGFR-directed targeted therapy. The latter strategy was chosen given the preclinical synergy of this combination and the identification of secondary MET dependence following EGFR TKI resistance. Unfortunately, while select phase 2 trials seemed promising (with improvements in PFS and OS), subsequent confirmatory phase III trials did not show a benefit in MET-overexpressing cancers (Table 3) and the level of MET expression did not correlate with increased benefit. Additionally, patients with MET-overexpressing cancers by IHC were equally responsive to MET-directed therapy or placebo (Supplemental Table 1).

Table 3.

Targeted therapy outcomes by MET expression status

MET IHC	Drug	Disease	Trial (n)		IHC (−)	IHC (+)
Polyclonal antibody
AF276	rilotumumab + chemotherapy173	prostate cancer	phase II (n=144)			+ low	+ high
				mPFS	-	3.6 mo	2.9 mo
				80% CI	-	(2.8–4.0)	(2.7–3.5)
Monoclonal antibodies
MET4	rilotumumab + chemotherapy174	gastric/GEJ cancer	phase Ib/II (n=9 for Ib, n=121 for II)	mPFS	6.8 mo	4.4 mo
				mOS	11.1 mo	10.6 mo
				ORR	6/19 (32%)	20/40 (50%)
SP44	tivantinib194	hepatocellular carcinoma	phase II (n=71)	mOS	3.8 mo	9.0 mo
	tivantinib + chemotherapy208	colorectal cancer	phase II (n=117)	mPFS	11 mo	7.9 mo
				mOS	NE	22.3 mo
				ORR	11/36 (31%)	24/54 (44%)
	onartuzumab + chemotherapy209	triple negative breast cancer	phase II (n=185)			+ low	+ high
				mPFS	7.3 mo	5.7 mo	10.3 mo
				mPFS	-	8.5 mo	4.4 mo
	onartuzumab + erlotinib210	EGFR-mutant non-small cell lung cancer	phase II (n=61)	95% CI	-	(7.1–12.4)	(0.7-NE)
				ORR		35/53 (66%)	7/8 (87%)
	onartuzumab + erlotinib175	non-small cell lung cancer	phase II (n=137)			1+	2+	3+
				mPFS	1.4 mo	1.4 mo	4.1 mo	2.7 mo
				ORR	1/32 (3%)	3/35 (9%)
Unknown antibody	capmatinib	Non-small cell lung cancer	phase I (n=52)			1+	2+	3+
				ORR	-	-	2/12 (17%)	9/37 (24%)

New MET antibody strategies have thus been explored. These include mixtures of antibodies directed against different epitopes on the MET protein (e,g, Sym015) that have shown preclinical activity against MET overexpressing and MET exon 14-altered cell lines. In addition, antibody drug conjugates have been explored. These drugs have the advantage of binding MET-expressing cancer cells regardless of MET signal dependence and lower levels of expression may be sufficient for payload delivery. As an example, a study of the antibody-drug conjugate telisotuzumab vedotin did not find a correlation between MET expression levels and benefit in a variety of solid tumors. It is important to keep in mind that, in the same vein, as MET can be expressed on normal lung tissues¹⁹⁷, pulmonary toxicity can be an issue.

While MET expression is a poor predictor of benefit from MET-directed therapy in the absence of evidence of a genomic correlate for MET dependence, the predictive nature of MET expression is being recontextualized in cancers that are oncogenically addicted to MET. In one study of MET exon 14-altered NSCLCs, MET expression was surprisingly heterogeneous, where 31% (n=5/16) did not have MET protein detectable by SRM-MS⁹⁷. In addition, the ORR with crizotinib was higher in cases with detectable MET expression by SRM-MS (55%, n=6/11) compared to those that did not express MET (0%, n=0/5). This suggests that, in the right context, prospectively identifying MET expression may maximize benefit with MET-directed targeted therapy. Furthermore, the role that MET expression plays in MET-amplification has not been well defined; this should be explored prospectively on ongoing trials of MET-directed therapies.

CONCLUSIONS

The process of identifying MET-dependent cancers is complex. From a diagnostic perspective, clinically-meaningful cutoffs need to be standardized for continuous variables such as the level of MET amplification or MET expression. Diagnostic migration towards more comprehensive and technically sophisticated assays will likely be required to maximize the likelihood of detection MET amplification, mutation, and fusion. Assays such as next-generation sequencing should be considered for the detection of these alterations in both tumor and plasma, and ideally in both DNA and RNA. The effective detection of MET-dependent cancers is critical given that MET-directed targeted therapy is active in many of these cancers. Importantly, the level of activity of these therapies can be modulated by the type of alteration identified and the degree of oncogenic addiction to MET pathway signaling.

Box 1 |. Pathophysiology of MET in cancer and immunotherapy.

Under physiologic conditions, hepatocyte growth factor (HGF), the ligand for MET, regulates the epithelial-to-mesenchymal transition for tissue repair and embryogenesis². In cancer, increased MET activity promotes tumor growth by providing anti-apoptotic and pro-migratory signals³. Furthermore, the MET gene is regulated by transcriptional factors including hypoxia-inducible factor 1α (HIF-1α)⁴. In preclinical models, the inhibition of angiogenesis results in hypoxic stress that leads to MET-mediated local invasion and distant metastasis^5,6.

MET also plays a role in regulating the immune microenvironment. During tissue repair, upregulation of HGF and increased HGF-MET autocrine signaling promotes an immunosuppressive environment by converting immunologically active macrophages (M1 phenotype) to a growth-stimulating (M2) phenotype and by inducing tolerogenic dendritic cells^7,8. MET inhibition can abrogate these effects while concurrently increasing PD-L1 expression⁹. Thus, the combination of immunotherapy (e.g. with an immune checkpoint inhibitor) and MET-targeted therapy is currently being explored in multiple ongoing clinical trials (NCT03914300; NCT04151563; NCT02406208; NCT03866382; NCT03793166; NCT02819596; NCT03742349; NCT03484923).

Box 2 |. Type 1 and type 2 MET inhibitors.

Type I MET tyrosine kinase inhibitors (TKIs) target the ATP-binding pocket of the active form of MET¹⁴³. Type Ia agents, such as the multikinase ALK, ROS1, and MET inhibitor crizotinib, interact with MET moieties such as the Y1230 residue, the hinge region, and the solvent front G1163 residue. Type Ib inhibitors tend to be more MET-selective agents that, in contrast to type Ia agents, do not interact with G1163¹¹⁷. Type Ib inhibitors include capmatinib, tepotinib, and savolitinib. Unsurprisingly, these have been shown in vitro to overcome solvent front substitutions (e.g. G1163E/R) that render resistance to crizotinib.

Type II MET TKIs (cabozantinib, merestinib, and glesatinib) are likewise ATP-competitive, but bind the ATP pocket in the inactive state by extending to a hydrophobic back pocket^117,144,145. Binding to this configuration allows these agents to act against MET kinase domain mutations that result in resistance to type Ia and type Ib agents^117,146. These include the kinase domain mutations MET D1228E/G/H/N, and MET Y1230C/D/S/H/N.

Conversely, resistance to type II MET inhibitors can occur by MET L1195 and MET F1200 mutations^117,144,145. Thus, switching between type I and type II MET inhibitors might be effective in MET-dependent cancers depending on the specific resistance mutation that emerges.h

KEY POINTS.

• The degree of MET amplification is a continuous variable that can be measured by FISH or NGS. No consensus exists on diagnostic cutoffs for MET amplification.

• Solid tumors that harbor high-level MET amplification have a higher likelihood of benefit with single-agent or combination MET-directed therapy compared to those with lower MET copy number gains.

• MET mutations are highly heterogeneous and can range from those that involve the MET kinase domain to those that result in MET exon 14 alteration biology.

• The activity of type I or type II MET tyrosine kinase inhibitors can substantially differ by MET mutation type.

• A wide variety of MET fusions have been identified, but the biology of these alterations and their response to targeted therapy are not well characterized.

• MET overexpression in the absence of a known driver of MET dependence is a poor predictor of benefit from MET-directed targeted therapy.