Identification of a novel MET mutation in high-grade glioma resulting in an auto-active intracellular protein

MET has gained interest as a therapeutic target for a number of malignancies because of its involvement in tumorigenesis, invasion and metastasis. At present, a number of inhibitors, both antibodies against MET or its ligand hepatocyte growth factor, and small molecule MET tyrosine kinase inhibitors are in clinical trials. We here describe a novel variant of MET that is expressed in 6 % of high-grade gliomas. Characterization of this mutation in a glioma cell line revealed that it consists of an intronic deletion, resulting in a splice event connecting an intact splice donor site in exon 6 with the next splice acceptor site being that of exon 9. The encoded protein lacks parts of the extracellular IPT domains 1 and 2, encoded by exons 7 and 8, resulting in a novel pseudo-IPT and is named METΔ7−8. METΔ7−8 is located predominantly in the cytosol and is constitutively active. The auto-activating nature of METΔ7−8, in combination with a lack of transmembrane localization, renders METΔ7−8 not targetable using antibodies, although the protein is efficiently deactivated by MET-specific tyrosine kinase inhibitors. Testing of MET-expressing tumors for the presence of this variant may be important for treatment decision making.


Introduction
The MET proto-oncogene (chromosome 7q31.2) encodes the tyrosine kinase membrane receptor MET (also called Scatter Factor Receptor), which is essential during development. Signaling from the receptor controls epithelialto-mesenchymal transition (EMT) of myogenic precursor cells during differentiation into skeletal muscle cells [5], a process that involves migration over long distances in the embryo. In adults, MET is involved in tissue regeneration upon injury [6].
MET is produced as a glycosylated single-chain precursor protein of ~190 kDa which, during transport to the membrane, undergoes furin-mediated cleavage in the trans-Golgi network [8,29]. The resulting mature receptor consists of an extracellular 50 kDa α-chain, covalently attached via a disulfide bond to a membrane-spanning 140 kDa β-chain [20,56]. The extracellular segment consists of an N-terminal Sema domain that is involved in ligand binding, a small cysteine-rich domain, and four IPT (Immunoglobulin-like fold shared by plexins and transcription Abstract MET has gained interest as a therapeutic target for a number of malignancies because of its involvement in tumorigenesis, invasion and metastasis. At present, a number of inhibitors, both antibodies against MET or its ligand hepatocyte growth factor, and small molecule MET tyrosine kinase inhibitors are in clinical trials. We here describe a novel variant of MET that is expressed in 6 % of high-grade gliomas. Characterization of this mutation in a glioma cell line revealed that it consists of an intronic deletion, resulting in a splice event connecting an intact splice donor site in exon 6 with the next splice acceptor site being that of exon 9. The encoded protein lacks parts of the extracellular IPT domains 1 and 2, encoded by exons 7 and 8, resulting in a novel pseudo-IPT and is named 1 3 factors) domains, which connect the Sema and cysteinerich domains with the C-terminal β-subunit [18].
Upon binding of the ligand hepatocyte growth factor (HGF, scatter factor), receptor dimerization occurs followed by trans-phosphorylation in the intracellular tyrosine kinase (TK) domain at tyrosine (Y) residues 1230, 1234 and 1235 [14,34]. The TK domain subsequently induces auto-phosphorylation of Y1349 and Y1356, which act as docking sites for signal transduction molecules including GAB1, GRB2, phospholipase-C and SRC [50]. Phosphorylated GAB1 interacts with molecules like PI3-K and SHP2, which together induce several downstream signaling pathways. MET signaling is mediated by, among others, the PI3-K/AKT and RAS/MAPK pathways, which induce cell cycle progression, survival, cytoskeletal changes and invasion (reviewed in [17]). In addition to its role downstream of HGF, MET can also be involved in signaling of other transmembrane receptors, including VEGFR2, CD44v6, EGFR and Plexin B1 [19,26,35,44]. Upon ligand-induced receptor dimerization, MET is internalized via endocytosis and may be recycled [27]. Phosphorylation of Y1003 in the juxtamembrane (JM) domain of the receptor leads to ubiquitination and subsequent proteasomal degradation [25]. Thus, levels of MET in the cell are tightly regulated.
Aberrant activation of MET signaling is a tumor-promoting event in a variety of malignancies and can be induced by several mechanisms, including alternative mRNA splicing, exon skipping and crosstalk with other receptors [13]. In high-grade gliomas, the frequently occurring oncogenic EGFR mutant EGFRvIII can induce overexpression of both MET and HGF, a process that is balanced by wild-type EGFR activation [33]. MET amplifications have been found in a number of tumor types including glioblastoma (GBM) [9,10] and missense mutations in the Sema, the TK and the JM domain have been reported to affect HGF binding, kinase activation and receptor degradation, respectively [1,30,32,36,38,43,48,49]. Recently, gene fusions between the protein tyrosine phosphatase PTPRZ1 and MET, resulting in constitutive activation of MET, were described in 16 % of secondary GBMs [2]. Activation of MET signaling has been proposed as a mechanism of resistance to EGFR inhibitors, likely a result of the similarities in downstream signaling events from both receptors [3].
The significant role that MET plays in tumor progression and metastasis has made it a prime therapeutic target in oncology. MET tyrosine kinase inhibitors and therapeutic antibodies against the extracellular domain of MET and against HGF, all preventing HGF-mediated MET activation, are currently in clinical trial (www.clinicaltrials. gov). In a previous study, we have shown that the combined VEGFR2/MET tyrosine kinase inhibitor cabozantinib (XL-184, CoMETRIQ) potently inhibits MET phosphorylation, cell proliferation and migration and consequently prolongs survival of mice carrying orthotopic E98 glioma xenografts [42]. Here, we identify a novel intragenic MET deletion in E98 cells, which results in a truncated protein that is constitutively active and lacks membranous expression, thereby having important implications for therapeutic strategies targeting MET. We show that this mutation occurs in 6 % of glioblastomas and, like the EGFR mutation EGFRvIII [4], is relatively specific for this tumor type.

Genetic analysis of E98
Genomic DNA from E98 cells was analyzed by semi-conductor sequencing (IonPGM, Life Technologies) using the comprehensive cancer panel (Life Technologies) that targets 409 cancer-related genes. The IonPGM E98 library generation was performed according to the manufacturer's protocol. In short, 10 ng of DNA per pool was amplified in 21 cycles by PCR using the Ion AmpliSeqTM mastermix, followed by barcode and adapter ligation. Amplified products were purified with Agencourt AMPure XP beads (Beckman Coulter Genomics, High Wycombe, UK). The library was diluted to 20 pM. Emulsion PCR was performed using the Ion OneTouchTM 200 Template kit following the protocol of the Ion OneTouchTM System. Next, Ion Sphere Particles (ISPs) were recovered and enriched for template-positive ISPs using Dynabeads MyOne Streptavidin C1 beads (Life Technologies) in the Ion OneTouchTM ES instrument (Life Technologies). ISP enrichment was quantified using the Qubit 2.0 fluorometer (Life Technologies). Sequencing primer and polymerase were added to the final enriched spheres before loading onto an Ion 318 chip according to the Ion PGMTM 200 sequencing kit protocol. The gene copy number analysis was performed as follows. The relative number of sequence reads aligned to a specific gene was determined by dividing by the total number of aligned reads of E98, and was divided over the relative number of sequence reads of the same gene in non-neoplastic blood cells. The relative ratios are plotted in a graph based on the genomic position of the gene.

FISH
Formalin-fixed, paraffin-embedded sections (4 μm) on SuperFrost glass (dried >45 min at 56 °C) were deparaffinized and rehydrated in ddH 2 O. After boiling in a microwave in sodium citrate buffer (pH 6), slides were allowed to cool to RT, washed in ddH 2 O and incubated for 5 min in 10 mM HCl. Proteins were digested with pepsin (200 U/ ml, Sigma) for 15 min at 37 °C. Subsequently, slides were rinsed in 10 mM HCl and PBS and postfixed for 5 min in 1 % paraformaldehyde (PFA, Merck)/PBS. Sections were washed in ddH 2 0, dried and hybridized with 10 μl probe mix (1 μl cep7 Spectrum Green (06J37007) + 1 μl LSI MET Spectrum Red (06N05-001) + 7 μl hybridization buffer in MQ, all Vysis) under a cover slip. Sections were denatured at 80 °C for 10 min, followed by hybridization o/n at 37 °C in a Hybridizer (Dako). After removing the coverslip by soaking for 5 min in 2× SSC buffer (Maxim Biotech) at 42 °C, slides were washed 3 times in 2x SSC buffer at 73 °C, once in 2xSSC (5 min) and once in ddH 2 O. After dehydration in EtOH, slides were air-dried in the dark and mounted in Vectashield/DAPI (3 parts Vectashield/ DAPI + 1 part Vectashield, all vector). Slides were analyzed on a Leica Fluorescence microscope.

Transfection and protein purification
pIRESneo-MET or pIRESneo-MET Δ7-8 were transfected into HEK-293T or TOV-112D cells in 6-well culture dishes (Greiner Bio-One, Krëmsmunster, Austria) using Fugene HD transfection reagent (Promega, Fitchburg, WI, USA) according to the manufacturers' instructions. After 48 h, cell monolayers were washed with PBS and cell extracts prepared in RIPA buffer containing protease and phosphatase inhibitors (Cell Signaling Technology, CST, Danvers, MA, USA).

Protein domain modeling
A homology model for the new hybrid IPT-domain using the WHAT IF & YASARA Twinset was generated [31,60]. We used the experimentally solved 3D structure 2uzx, which contains the human tyrosine kinase MET. The first 60 residues of the hybrid domain are identical to this structure, whereas the following 38 residues were modeled based on homology between the 2 domains.

Biotinylation assay
E98 and A549 cells were allowed to adhere and grown to 80 % confluence in a 6-well plate (Cellstar, Greiner Bio-One, Kremsmünster, Austria). Cells were washed 3 times with ice-cold PBS and incubated for 30 min at 4 °C with 0.5 mg/ml EZ-Link Sulfo-NHS-LC-Biotin (ThermoFisher Scientific). This compound does not penetrate cells and biotinylation is therefore confined to exposed membrane proteins. In parallel, Tris-free RIPA extracts, prepared from equivalent numbers of E98 or A549 cells by repeated washing over Amicon Ultra K10 centrifugal filters (Merck Millipore, Billerica, MA, USA), were treated similarly to biotinylate all cellular proteins. The reaction was quenched by washing cells or cytosolic proteins three times with 100 mM glycine/PBS and two times with PBS (using K10 columns in the case of protein lysates). Intact biotinylated cells were then subjected to lysis with RIPA buffer. Protein concentrations in all lysates were determined using the BCA protein concentration assay (ThermoFisher Scientific). MET was immunoprecipitated from 200 µg total protein in a volume of 200 μl using anti-MET (1:50, clone D1C2, CST) for 1 h at 4 °C. Immune complexes were captured by incubation for 30 min at 4 °C with 10 μl prot A agarose slurry (Roche Diagnostics, Basel, Switzerland), followed by centrifugation (14,000 rpm, 4 min) and three PBS washes. Immune complexes were solubilized by heating (5 min 95 °C) in 30 μl 2× SDS-PAGE sample buffer (0.2 % SDS, 62.5 mM Tris-HCl pH 6.8, 10 % glycerol and 0.2 μM DTT). Samples were subjected to 10 % SDS-PAGE and western blotting as described above.

E98 MET protein is auto-activated in an HGF-independent fashion
Our previous experiments have shown that MET in orthotopic E98 xenografts is phosphorylated in tumor cells in a homogeneous fashion (Fig. 1a, see also [42]). In vitro, E98 cells also show high levels of phosphorylated MET when grown under serum-free conditions (Fig. 1b). HGF treatment did not further increase phosphorylation levels of MET, in contrast to A549 control cells in which HGF was required for MET activation. MET phosphorylation in E98 cells was not the result of endogenous HGF expression as revealed by RT-PCR analysis (Fig. 1c), while analysis of E98 xenograft RNA revealed the presence of mouse HGF only, as determined by Sanger sequencing (not shown). Since mouse HGF is not an activating ligand for human MET [62], we conclude that constitutive activation of MET in E98 cells and xenografts is not the result of an autocrine HGF-activation loop.

E98 cells express a truncated version of an amplified MET gene
To examine the underlying mechanism of MET auto-activation in this model, we PCR-cloned MET cDNA from E98 cells, using primers flanking the open reading frame (ORF, NM_000245.2). The MET PCR product from E98 was 240 bp smaller than that from U87 and CaCo2 cDNA (Fig. 2a), and Sanger sequencing analysis revealed an inframe deletion of nt 2050-2289 in the coding sequence, corresponding to exons 7 and 8. The same transcript was also found in E98 xenografts (not shown). Genomic analysis of E98 cells using semi-conductor sequencing revealed high copy MET amplification (Fig. 2b). Average amplification of MET (except for exons 7/8) was about 14-fold. FISH analysis using probes specific for MET and chromosome 7 centromere confirmed the amplification (Fig. 2c). Consistent with Sanger sequencing, amplicons in exon 7 and the first part of exon 8 were absent (see insert in Fig. 2b), suggesting that the lack of exons 7 and 8 results from a genomic rearrangement (similar to the EGFR variant III . A matching H&E staining is shown as reference. Note that tumor shows diffuse infiltration in the brain parenchyma and that the tumor cells are highly positive for this activated MET. b Western blot analysis of MET expression in serum-starved E98 and A549 cells in absence of presence of HGF. Note that the processed form of MET (arrowhead) in E98 is somewhat smaller than that of A549 while the preform is predominantly present in E98 (arrow). α-Tubulin was used as a loading control. c RT-PCR for HGF on E98, TOV-112D and HEK-293T cell line cDNA. HMBS was used as a control housekeeping gene [15]), instead of alternative splicing. PCR on genomic DNA using exon 6-and 9-specific primers resulted in amplification of a 910-bp fragment (predicted size from the wild-type allele, present in U87, is 3014 bp, see Fig. 2d). Sequencing of this product revealed an intronic deletion of 2114 bp (between position g.116 395 653 located in intron 6, and 116 397 766, located in exon 8). This deletion results in an intact splice donor site from exon 6, juxtaposed to the exon 9 splice acceptor site, and explains the lack of exons 7 and 8 in the resulting mRNA (Fig. 2e). Of note, a wtMET allele could not be detected in E98 cells (Fig. 2d).
In the MET protein, this rearrangement leads to loss of the C-terminus of IPT1 and the N-terminus of IPT2 (Fig. 3a). Yasara modeling, using X-ray crystallographic data of the MET ectodomain, predicts the formation of a novel IPT, composed of the remaining parts of IPT1 and  (Fig. 3b, arrow).

MET Δ7-8 is aberrantly processed
MET is synthesized as a 190-kDa precursor protein which is proteolytically cleaved by furin between residues 307 and 308, to yield an extracellular α-chain of approximately 45 kDa, covalently linked via a disulfide bridge to the transmembrane β-chain. Reducing SDS-PAGE, followed by Western blot analysis of E98 cell extracts showed that the majority of MET protein was in a 180-kDa phosphorylated form, corresponding to the uncleaved truncated preform (Fig. 1b, arrow). In contrast, in A549 cells the matured cleaved MET protein was the predominant form (arrowhead in Fig. 1b). To investigate whether this was a specific feature of E98 cells, we analyzed the protein structure in cells after transfection with the full-length cDNAs encoding MET wt or MET Δ7-8 . In both HEK-293T and TOV-112D cells, wtMET was properly processed to an αand β-chain, indicating that these cells are not defective in furin-mediated processing. In contrast, MET Δ7-8 was predominantly present in the uncleaved preform in both cell types (Fig. 4a, arrow). Thus, improper MET cleavage is an intrinsic property of MET Δ7-8 . Overexpression studies in HEK-293T and TOV-112D cells resulted in phosphorylated Y1234/1235 residues in both MET wt and MET Δ7-8 proteins (Fig. 4a, P-MET). Because both cell lines produce HGF (Fig. 1c), this may be a result of HGF-dependent autocrine activation. MET is cleaved by the endoprotease furin, which is localized predominantly in the trans-Golgi network, but also in vesicles and near the plasma membrane [8,29,52]. To test whether the inefficient cleavage of MET Δ7-8 in E98 cells is related to intracellular transport defects, we analyzed the subcellular localization of MET Δ7-8 in detail via confocal microscopy. MET Δ7-8 did not co-localize  with the cell surface marker CD44 and was confined to the cytosol (Fig. 4b). In contrast, in U87 cells MET wt did co-localize with CD44. Additional intracellular staining in U87 cells reflects de novo synthesized material that is being processed for constitutive secretion. Immunostainings with the early endosome marker EEA-1 and the rough endoplasmic reticulum (RER) marker CLIMP-63 suggested that MET Δ7-8 in E98 cells is predominantly retained in the RER (Fig. 4c).
To confirm the absence of cell surface expression of MET Δ7-8 on E98 cells, we labeled intact E98 cells or cell lysates with NHS-biotin and immunoprecipitated MET using specific antibodies, followed by SDS-PAGE/Western blot and staining for biotin and MET. Whereas METs' N-terminal α-chain was readily biotinylated in E98 cell lysates, no detectable MET biotinylation occurred when intact cells were labeled (Fig. 4d, lane 1). In contrast, biotinylated MET was readily detected in A549 cells upon labeling of intact cells, as shown by the biotin-labeled α-chain (Fig. 4d, lane 3). Thus, these data confirm that MET Δ7-8 is predominantly localized intracellular and is poorly exposed on the cell surface of E98 cells.
To further confirm a defect in intracellular trafficking of MET Δ7-8 , we analyzed secretion patterns of extracellular domains of wtMET or MET Δ7-8 (ending with residue D 929 , numbering according to MET variant 2 (NP_000236.2), containing a C-terminal biotin tag. Whereas the extracellular domain of wtMET was properly processed and secreted into the culture medium (as illustrated by the presence of the 309-929 biotinylated extracellular β-chain and the MET25-308 α-chain, Fig. 4e, lane 4), no secreted MET products were found in medium of cells, transfected with the MET Δ7-8 ectodomain (Fig. 4e, lane 3). Instead, all biotinylated MET products were located intracellularly (Fig. 4e, lane 1).

185-kDa MET Δ7-8 is not affected by antagonistic anti-MET antibodies but is inhibited by cabozantinib
It was previously reported that incubation of A549 cells with VHH G2, a recombinant single-domain llama antibody against MET, results in low MET activation levels, while inhibiting the strong activation which is induced by HGF [22]. To test whether and how G2 affects phosphorylation of MET Δ7-8 , we treated serum-starved E98 cells with G2, either followed or not followed by HGF, using A549 cells as control. As shown in Fig. 5, HGF did not increase overall MET phosphorylation levels, although, interestingly, a slight activation was seen in the minority of processed β-fragment of MET. For G2, a similar effect was observed. Consistent with high overall MET phosphorylation levels in all samples, neither G2 nor HGF treatment resulted in altered levels of P-AKT and P-ERK1/2, both targets of MET. In A549 cells G2 and HGF induced MET phosphorylation, which increased levels of P-AKT and P-ERK1/2. Thus, in contrast to A549 cells that express MET wt , E98 cells are not responsive to antibodies against or ligands of MET. However, both E98 and A549 cells responded well to the MET tyrosine kinase inhibitor cabozantinib (Fig. 5). Δ7-8 Data mining of the COSMIC and TCGA databases did not uncover the intronic deletion in gliomas and other tumor types (not shown). Intragene deletions spanning multiple exons are, however, in general more difficult to recognize in whole-exome sequencing (WES). Furthermore, the coverage of the exact location of the deletion is poor in the mostly used WES protocols, and as can be seen in Fig. 2b, even targeted sequencing reveals the mutation only after detailed bioinformatic analysis. We, therefore, assume that detection of MET Δ7-8 requires dedicated PCR protocols.

Prevalence of MET
PCR with deletion-spanning primers revealed the presence of the deletion-specific 105-bp fragment in E98 cDNA, while a number of other cell lines presented with the wt 345-bp amplicon only (Fig. 6a). Since frozen material from the patient tumor that was used to generate E98 is unavailable, we could unfortunately not obtain genomic DNA and cDNA of sufficient quality to confirm the presence of the mutation in the originating tumor. We did however perform immunohistochemistry on formalin-fixed, paraffin-embedded tumor material from this patient and observed a highly heterogeneous staining for MET, with only a small percentage of strongly positive tumor cells, apparently with intracellular staining (Fig. 6b). Of note, such immunostainings cannot discriminate between MET wt and MET Δ7-8 . FISH analysis confirmed the presence of the MET amplification in the original tumor (Fig. 6c).
To investigate the prevalence of the MET Δ7-8 mutation further, we performed the exon 6-9 PCR on cDNA, generated from a series of gliomas (n = 102) and a number of other tumor types in which MET has been suggested to play an important role (castration-resistant prostate carcinomas, n = 43; Ewing sarcoma, n = 21; rhabdomyosarcoma, n = 22) [24,28]. MET Δ7-8 was found in 6 out of 102 gliomas, both grade III and IV (5.8 %; 2 out of 5 anaplastic oligodendrogliomas, 2 out of 16 anaplastic astrocytomas and 2 out of 61 glioblastomas), and was not detected in the other tumor types tested (Table 1). All grade III tumors with MET Δ7-8 were IDH1-R132H mutated. Most tumors that contained the mutation were heterozygous, containing also the wild-type transcript (example in Fig. 6d). Further research with higher numbers of patients is needed to investigate any correlations between MET Δ7-8 and other molecular aberrations.
Sanger sequencing of the smaller PCR products of one patient with glioma confirmed the deletion of exons 7 and 8 (not shown). Of note, due to the lack of high-quality genomic DNA from these clinical samples we could not discriminate in these samples whether MET Δ7-8 resulted from a mutation or from exon skipping or alternative splicing, such as suggested for MET variants lacking exon 10 or 14 [37].

Discussion
A major problem in glioma treatment is diffuse growth in the neuropil, and it has been suggested that the MET oncogene is causally involved in this phenotype in subsets of tumors [16]. Several signal transduction pathways that are induced by MET activation are shared with other tyrosine kinase receptors and it is increasingly recognized that activation of MET bypasses the need for EGFR activation, generating resistance to EGFR-targeted therapies [13]. Also, there is an interesting crosstalk between MET and EGFR: the EGFR variant III (EGFRvIII) stimulates expression of both MET and its ligand HGF, possibly contributing to oncogenicity of EGFRvIII [16], but this phenomenon is counteracted by EGF-mediated activation of wild-type EGFR in the same complex [33].
We previously reported that MET is crucially implicated in proliferation, survival and migration of (EGFR-negative) E98 glioma cells in vitro. To further investigate the involvement of MET in the biological behavior of E98 cells, we   [42]. Based on the observation that the pattern of MET phosphorylation in E98 xenografts was remarkably homogeneous in the absence of its ligand HGF, we analyzed the MET product in E98 cells in more detail and found that the protein is expressed as a truncated product, which lacks effective furin cleavage and is predominantly retained intracellular in its active, phosphorylated form. The truncation generates a novel IPT domain consisting of a fusion between the carboxyterminus of IPT1 and the amino terminus of IPT2. Modeling of the novel fusion IPT domain suggests that it adopts a similar structure as the IPT domains in wild-type MET, although a small stretch of extra amino acid residues is accommodated in a loop, extending towards the Sema domain. Since this loop approaches the furin cleavage site in the Sema domain, it may directly affect furin cleavage. Interestingly, the colorectal adenocarcinoma cell line Lovo lacks furin protease and consequently cannot properly process MET [40]. In this cell line, the unprocessed preform is still expressed on the cell surface, binds HGF and is capable of signaling. Thus, defective processing alone does not explain intracellular retention of MET Δ7-8 and we therefore hypothesize that the novel IPT1-2 fusion domain in MET Δ7-8 is involved in defective furin cleavage, intracellular retention of the uncleaved perform and auto-activation of the uncleaved preform.
Recently, effects of various activating mutations in the MET TK domain have been analyzed in detail. These mutants are expressed on the cell surface but are subject to increased rates of turnover, resulting in accumulation in early endosomes where they can still signal [27]. MET Δ7-8 did not accumulate in early endosomes, suggesting that the underlying mechanism of intracellular retention is different [41]. Our confocal microscopy and biotinylation experiments strongly suggest that only the very small fraction of MET Δ7-8 that is cleaved reaches the cell membrane, but this fraction is insignificant with respect to the auto-active component in the cytosol.
From a genetic perspective, the MET Δ7-8 mutant resembles the auto-active EGFRvIII variant which occurs concomitant with EGFR amplification in 25-64 % of all GBMs [51,59]. This variant results from a genetic deletion of exons 2-7 and its activation is also independent of the ligands EGF or TNFα. Like the EGFRvIII mutation [15], the MET Δ7-8 alteration appears to be restricted to only few cancer types, as we detected it in glioma but not in a number of other tumor types. The mutation was detected in grade III gliomas of both oligodendroglial and astrocytic origin and GBMs. Investigations of MET Δ7-8 expression in other tumor types is however warranted and is ongoing in our lab.
Another example of a tyrosine kinase receptor which often carries deletions in GBM is PDGFRA. Forty percent of the GBMs with amplified PDGFRA contain an inframe deletion in the ectodomain, leading to constitutive activation of the TK domain [46]. Recently, also in pediatric high-grade gliomas, ligand-independent and tumorigenic in-frame PDGFRA deletion variants have been reported [47,58]. Data on the subcellular localization of such mutated oncoreceptors are frequently lacking, and our data call for in-depth analysis of the cellular localization of receptors that are considered targetable. Interestingly, activating mutations in the RON tyrosine kinase receptor have been identified which resemble the MET Δ7-8 mutation in that it also involves the first IPT domain [39]. Of note, these alterations do not lead to a loss of expression at the cell surface.
Importantly, we did not detect a MET wt transcript in E98 cells, in contrast to the clinical gliomas which were analyzed in this study and which all showed abundant wtMET, also in MET Δ7-8 tumors. We formally cannot exclude that the wild-type MET amplicons in our PCR derive from 'contaminating' non-neoplastic stromal cells in the tumor biopsies that were tested. There may, however, be another explanation for the relatively low levels of MET Δ7-8 in clinical tumors: in the patient tumor from which the E98 model was generated, a low percentage of MET-expressing tumor cells was detected, although it is impossible to determine whether these cells carry the MET Δ7-8 mutation because the antibodies used do not discriminate between MET and MET Δ7-8 . It is tempting to speculate that during the generation of the E98 model, a small subset of MET Δ7-8 tumor cells in the primary tumor experienced a growth advantage, ultimately resulting in clonal outgrowth during xenograft formation. This scenario fits with the notion that clinical tumors not only show inter-but also intratumoral heterogeneity [54], and derived preclinical tumor models may only be representative for the most malignant population of tumor cells.
Expression of MET Δ7-8 may have important consequences for choice of therapy. MET is increasingly recognized as an important target in multiple tumor types, including glioma, and therapeutic antibodies against HGF or the HGF binding site on MET have been developed. Since HGF is not involved in MET Δ7-8 activation and MET Δ7-8 is retained intracellular, MET-mutated cells will not be responsive to these therapies. Indeed, we were able to show that E98 cells do not respond to the anti-MET VHH G2. Selection of MET-mutated cells in tumors that initially respond to antibody-based MET-directed treatment is expected to result in recurrence of treatment-resistant clones. In this respect, it will be important in future studies to assess the occurrence of the Δ7-8 mutation in paired samples of primary and recurrent tumors after MET antibody-based therapies, but also anti-EGFR therapies since cells may use MET Δ7-8 to bypass EGFR signaling [3,53]. With this in mind, the use of specific tyrosine kinase inhibitors of MET may have preference over antibody-based therapy for resistant tumors, at least in the ones that are KRAS and RAF wild type, [3,21,57,61] since MET Δ7-8 is sensitive to these inhibitors [42]. Such inhibitors have already shown to improve overall survival of patients with non-small cell lung carcinoma with MET amplification and renal papillary carcinoma with MET mutations [11,45]. A recent clinical study with anti-MET antibody MetMab for lung cancer failed to meet the primary endpoint of prolonged survival [55]. A study of MET mutations and intracellular localization patterns in this patient group may be highly informative for future therapeutic directions [23].
In conclusion, we describe a highly active, non-liganddependent mutant of MET in 6 % of gliomas, which is not exposed on the cell surface and is predicted to be non-targetable with therapeutic antibodies against MET and/or HGF.