Targeting the MET Receptor Tyrosine Kinase as a Strategy for Radiosensitization in Locoregionally Advanced Head and Neck Squamous Cell Carcinoma

Head and neck squamous cell carcinoma (HNSCC) is a leading cause of cancer with a yearly incidence of more than 650,000 new cases and over 350,000 cancer-related deaths per year (1). Patients with advanced disease are treated with radiotherapy (RT), alone or in combination with systemic agents, exclusively or following surgery (1, 2). Consequently, resistance to RT represents a major cause of treatment failure in patients with persistent and relapsing HNSCC. Radioresistance in turn is a complex and multifactorial phenomenon, in which activation of oncogenes, accelerated tumor cell repopulation, hypoxia, proficient DNA damage repair, and tumor–stroma interactions among others play crucial roles (3–5).

In this context, oncogenic signaling through MET, the RTK for scatter-factor/hepatocyte growth factor (SF/HGF), contributes to increased cell growth and proliferation, angiogenesis, invasiveness, migration, and metastatic potential (6). Moreover, accumulating evidence situates MET as a relevant actor in the cellular DNA damage response (DDR) in various preclinical cancer models (7–10). Currently, nearly 160 clinical trials study a battery of MET inhibitors alone or in combination with other therapeutic modalities in various cancers (11). In HNSCC, both MET and SF/HGF overexpression and paracrine loops enhance invasive features and potentially confer therapeutic resistance to cytotoxic drugs (12, 13). Moreover, Saintigny and colleagues suggested that MET activation represents an early driver and a target for chemoprevention of oral cancers (14). To date, however, the effects of MET targeting along with ionizing radiation (IR), the most common treatment modality for HNSCC, have not been comprehensively explored (13, 15, 16).

Here, we investigate the prevalence and impact of MET expression and mutations in primary tumors (PT) and lymph node metastases (LNM) in a cohort of advanced-stage HNSCC patients treated with RT. Moreover, we evaluate the effects of MET targeting in combination with IR in a panel of representative HNSCC cell lines both in vitro and in vivo, as well as in ex vivo organotypic tissue cultures (OTC) derived from patients with HNSCC. We show that MET expression in LNMs is associated with decreased disease-free survival and that MET targeting along with IR leads to heterogeneous responses, potentially resulting in radiosensitization in high MET-expressing models.

Patients' tissue and data were collected according to a predefined protocol approved by the regional ethical committee (Kantonale Ethikkomission Bern; Protocol Nr.: 050/14). Written informed consent has been obtained from the patients and the studies were conducted in accordance with recognized ethical guidelines (Declaration of Helsinki). Formalin-fixed paraffin-embedded (FFPE) archival blocks from HNSCC patients treated between 2004 and 2012 at Inselspital Bern were selected according to the following inclusion criteria: (i) histologically proven squamous cell carcinoma of the oral cavity, oropharynx, larynx and hypopharynx; (ii) presence of lymph node metastases involvement; (iii) primary neck dissection; (iv) primary or postoperative RT to the PT and the neck (± concomitant chemotherapy or cetuximab); (v) complete follow-up data for at least 2 years after the end of therapy for event-free patients; (vi) >18 years of age.

A tissue microarray (TMA) including PTs and/or LNMs, as well as normal tissue (from histologically normal salivary glands harvested during neck dissection), was constructed using a new-generation tissue arrayer as previously reported (17). At least two punches per sample were included to minimize sampling bias.

For IHC and H&E staining, 5-μm sections were cut from TMAs. Sections were deparaffinized and rehydrated. IHC was performed with the BOND RX automated system (Leica Biosystems). Endogenous peroxidase activity was blocked with H₂O₂ 3% for 4 minutes. Sections were incubated for 30 minutes with specific primary antibodies at room temperature: MET (Invitrogen, clone 3D4) and Ki67 (Cell Signaling Technology, clone D2H10). Antibody detection was performed with the Bond Polymer Refine Detection kit (Leica Biosystems, DS9800) according to the manufacturer's instructions. Expression of p-MET could not be detected with two different commercially available antibodies in FFPE tissues in the presence of positive controls (from fresh-frozen tissues). Semiquantitative assessment of IHC stainings was performed as previously described (18).

The presence of 34 previously reported MET gene mutations was determined using the iPLEX Assay (Agena GmbH), with genomic DNA isolated from 400 FFPE blocks as described in Supplementary Methods.

A total of 100 ng of genomic DNA were used to perform polymerase chain reactions (PCR) with PGMY primers (19). HPV genotype was determined using the Linear Array HPV Genotyping Test (Roche Molecular Diagnostics). This test allows detection of 16 high-risk (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68, 73, 82) and 21 low-risk HPV types (6, 11, 26, 40, 42, 53, 54, 55, 61, 62, 64, 67, 69, 70, 71, 72, 81, 83, 84, IS39, and CP6108). For HPV-negative cases, beta-globin of 400 bp was amplified by PCR to test the integrity of extracted DNA.

FaDu cells were purchased from the ATCC and Detroit-562 cells from Cell Line Services GmbH and authenticated by the suppliers; both lines were grown in MEM (Sigma) supplemented with fetal calf serum (FCS 10% vol/vol; Sigma), antibiotic–antimycotic (penicillin 100 U/mL, streptomycin sulfate 100 U/mL, amphotericin B as Fungizone 0.25 mg/mL; Gibco, Invitrogen Corp.), and nonessential aminoacids (1% vol/vol; Sigma). SCC-61, SQ-20B, and HN5 cells were provided by Prof. M. Pruschy (University of Zurich, Switzerland) and no further authentication was performed. University of Michigan Squamous Cell Carcinoma (UM-SCC) cell lines were obtained from Prof. T. Carey's laboratory (University of Michigan), cultured and authenticated as previously reported (20). More detailed information about these cell lines is provided in Supplementary Methods. All cells are MET wild-type. For all experiments, the cells were maintained in culture upon thawing for up to 8 passages (4 weeks).

The MET tyrosine kinase inhibitor tepotinib (or MSC2156119J; ref. 21) was kindly provided by Merck HealthCare KGaA (22). PHA665752 (23) was purchased from Pfizer. Both drugs were dissolved in DMSO, and DMSO concentrations were normalized for all treatment conditions. Recombinant human SF/HGF was purchased from R&D Systems and prepared following the manufacturer's instructions. Treatments were carried out as described in Results.

All in vitro experiments were performed at least 3 times unless otherwise indicated.

In vitro and ex vivo IR was delivered using a ¹³⁷Cs research irradiator (MDS Nordion) with a dose rate of 0.75 Gy/min. Irradiation was carried out at the indicated time points following tepotinib treatment of cells or OTCs.

Immunoblotting was performed as previously described (18). Overnight incubation with the following primary antibodies took place: p-Y1234/Y1235 MET (p-MET), MET, p-Ser473 AKT, and p-Thr202/Tyr204 ERK (all from Cell Signaling Technology). Anti-β-actin was obtained from Millipore Corporation and pSer139 H2AX from Upstate Biotechnology Inc. Membranes were incubated with appropriate secondary antibodies, and signals were detected with the ECL kit (Amersham Pharmacia Biotech) or with infrared fluorescence with an Odyssey imager (Li-Cor Biosciences). Densitometry values were determined using ImageJ and β-actin was used for value normalization (imagej.nih.gov/ij/).

Colony-forming assays were performed using standard methodology (24). The surviving fraction was normalized to the plating efficiency (PE = colonies formed/cells plated). Radiosensitization was evaluated using the radiation enhancement ratio (RER). Briefly, the RER is the ratio between area under the curve (AUC) of two treatment conditions. A RER significantly superior to 1 according to one-sample t test is deemed to indicate radiosensitization (25).

Maximal effective concentration (EC₅₀) values were determined as previously described (24).

Caspase-3 enzymatic activity was determined by using a fluorogenic assay based on the caspase-3–specific substrate Ac-DEVD-AMC (Calbiochem). Cell death/viability was assessed with the Live/Dead Assay Kit (Molecular Probes) following the manufacturer's instructions (18). Staining for senescence-associated β-galactosidase was done after 7 days of treatment as previously described (18). Images were acquired with a Leica DC 300F inverted microscope at 40× magnification.

For cell-cycle analysis, cells were plated in 10-cm dishes and treated as indicated for 2 days. At the endpoint, cells were fixed and stained with propidium iodide and acquired on LSRII flow cytometer (Beckton Dickinson). Cell cycle was evaluated by Dean-Jett-Fox using the FlowJo software (FlowJo LLC; ref. 18).

Cell migration was evaluated with the Oris Cell Migration Assembly Kit (AMS Biotechnology) after 3 days of treatment. Pictures were captured at baseline and at experimental endpoint (i.e., confluent controls) using a Leica DC 300F microscope. The invaded area in each treatment condition was determined with the ImageJ software (imagej.nih.gov/ij/).

The antitumor efficacy of tepotinib (MSC2156119J) in combination with irradiation was investigated in mouse xenograft models using immunodeficient mice. All mice were housed in a pathogen-free barrier room in the animal care facility at the Merck HealthCare KGaA and handled using aseptic procedures. All procedures were approved by the local authority. CD-1 female nude mice (Charles River Laboratories) or HSD-athymic female nude mice (Envigo) were injected subcutaneously at the lower back with the human head and neck cancer cell lines FaDu and Detroit-562 (5 × 10⁶ cells in 100 μL PBS, 1:1 matrigel). As soon as the tumors reached the linear growth phase (FaDu tumors: 90–225 mm³, Detroit 562 tumors: 100–200 mm³), tumor-bearing mice (10 mice/group) were irradiated with an X-RAD 320 biological irradiator (2 Gy, 5 days on/2 days off) followed by daily administration of tepotinib (125 mg/kg, per os) or vehicle (20% kolliphor/80% 100 mmol/L sodium-acetate buffer, pH 5.5). Body weight and tumor size (length [L] and width [W]) were measured twice weekly. The tumor volume was calculated using the formula L × W²/2. At the end of the treatment period, animals were euthanized, and samples of the tumor tissue were collected and formalin-fixed. Histone H3 phosphorylation was analyzed in FFPE tumor tissue as described in Supplementary Methods.

We generated 300-μm OTCs with a Vibratome VT1200 (Leica Microsystems) from freshly resected fragments of HNSCCs as previously described (Supplementary Methods; ref. 26). A total of 34 fresh patient samples were initially collected but 8 had to be excluded due to lack of tumor tissue, significant morphologic alterations upon culture, or loss of Ki67-positive nuclei (>10%) in vehicle-treated OTCs. Culture contamination occurred in 2 consecutive cases within the first 48 hours of treatment. Altogether, OTCs from 24 patients were included in the final TMAs (Supplementary Fig. S1A), comprising 2 to 6 punches (median 4) per uncultured tumor and 2 to 5 punches (median 2) for OTCs (Supplementary Fig. S1B). Methodological validation was performed using 10 tissues to evaluate morphologic changes and variations of MET and Ki67 expression during culture (Supplementary Fig. S1C–S1F). TMAs were manually constructed as described in Supplementary Methods.

Multitarget analysis of relevant proteins and phosphoproteins involved in intracellular signaling pathways was performed as previously described by means of single-cell mass cytometry (CyTOF; Supplementary Methods; ref. 27).

Summary statistics were calculated and presented as relative averages ± standard deviation (SD). For intergroup comparisons, if not otherwise stated, Mann–Whitney t test or one-way ANOVA followed by Tukey–Kramer multiple comparison tests were performed using GraphPad Prism 7 or SPSS. To compare in vivo growth curves, two-way ANOVA followed by Bonferroni multiple comparisons was used. Univariate survival analysis was plotted according to the Kaplan–Meier method and compared with the log-rank test. Multivariable survival analysis was performed using Cox proportional hazards survival regression. All P values were two-sided. An alpha level <0.05 denoted statistical significance.

High MET expression in HNSCC has been reported at rates ranging between 55% and 85% of cases, with inconsistent correlation between MET expression and both several clinicopathologic characteristics and patients' outcome (15). Patterns of differential MET expression in PTs and LNMs have not been previously explored in HNSCC. Here we assessed MET expression in 798 tissues (250 PTs, 317 LNMs, and 231 normal tissues). MET was highly expressed in 62.4% of PTs, 53.63% of LNMs, and only 3.39% normal tissue samples (Supplementary Fig. S2A and S2B). Expression of MET was significantly higher in PTs than LNMs (P = 0.040), and in both PTs and LNMs significantly higher than in normal tissues (P < 0.001). We next evaluated MET expression with clinicopathologic features and outcome in a cohort of 89 patients with HNSCC for whom matched sets of PTs–LNMs–normal tissues and complete clinical data were available (Supplementary Table S1). Median follow-up after therapy was 31.03 months (range, 3.23–165.30), and patients' median age was 60.02 years (range, 20.07–88.08). High MET expression was detected only in PTs or only in LNMs in 25.84% of the cases, low in both in 23.60% and high in both in 50.56% (Fig. 1A and B). MET expression levels were not significantly different across the various HNSCC anatomic subsites (Supplementary Fig. S2C) and did not correlate with any specific clinical or pathologic characteristics (Supplementary Table S1). Regarding outcome, high MET expression in LNMs and both primaries and LNMs was significantly associated with decreased regional recurrence-free survival (P = 0.026 and P = 0.024, respectively; Fig. 1C and D). No independent risk factors were identified on the various multivariable analyses performed.

Previous studies in HNSCC and other tumor entities showed that MET mutations tend to cluster in the juxtamembrane and extracellular domains (13, 28). In our cohort of 223 patients with locoregionally advanced HNSCC, we identified 5 different mutations in 10 patients (prevalence 4.48%), located in the extracellular domain of MET (E168D, n = 1 patient; N375S, n = 4), the juxtamembrane domain (R998C, n = 1; T1010I, n = 3), and the tyrosine kinase domain (M1178I; n = 1; Supplementary Fig. S2D). When present, mutations were detected in both primaries and LNMs, without significant differences in terms of mutated allele's abundance or clinicopathologic features (Supplementary Fig. S2E–S2G).

The biological effects of MET targeting were evaluated with two distinct specific MET inhibitors (PHA665752 and tepotinib) in an initial panel of 5 HNSCC cell lines. Total MET as well as p-MET could be detected in all cell lines (Fig. 2A). High levels of the activated MET receptor as inferred from its phosphorylation status were observed in 3 of 5 cell lines, whereas the remaining 2 cell lines had detectable but low p-MET levels (Fig. 2A). MET phosphorylation could be effectively abrogated by using MET inhibitors (Fig. 2A) and, even though effects on both ERK and AKT phosphorylation were usually modest, tepotinib (50 nmol/L) tended to be more effective than PHA665752 (300 nmol/L; Fig. 2A). Consequently, for subsequent assays we used the more novel MET inhibitor tepotinib, currently under clinical development in phase I and II trials (22). Although EC₅₀ values for tepotinib were generally on the low micromolar range (Supplementary Fig. S3A–S3E), tepotinib at a concentration of 100 nmol/L was sufficient to significantly impair plating efficiency in all 3 HNSCC cell lines with higher p-MET levels (Supplementary Fig. S3F).

In addition, expression of MET and p-MET as well as effects of tepotinib on MET downstream signaling was assessed in two pairs of patient-matched UM-SCC cell lines: UM-SCC-10A (primary laryngeal tumor) and UM-SCC-10B (lymph node metastasis), and UM-SCC-81A (primary laryngeal) and UM-SCC-81B (second oropharyngeal PTs after radiotherapy; Fig. 2B). All 4 cell lines expressed considerable p-MET levels that were decreased upon MET inhibition by tepotinib, however with moderate impact on p-AKT and p-ERK activation (Fig. 2B).

MET's ligand SF/HGF is a well-established motogenic molecule that has been shown to protect neoplastic cells from cytotoxic agents while promoting invasive features such as cell migration (7, 29). As mentioned, SF/HGF-MET paracrine loops in a background of receptor/ligand overexpression are a common pathogenic feature in HNSCC (12). We investigated the effect of MET inhibition on migration in cell lines with high basal p-MET levels (FaDu and Detroit-562) and in cells with low basal p-MET levels stimulated with SF/HGF (SCC-61 and HN5). Tepotinib impaired migration in a dose-dependent manner in FaDu and Detroit-562 (Supplementary Fig. S3G and S3H). Moreover, as expected, SF/HGF significantly stimulated cell migration both without and with IR, an effect that could be fully abrogated by MET inhibition (Supplementary Fig. S3I).

MET signaling potentially confers resistance to IR as well as other cytotoxic therapeutic modalities in several cancer models (7–10). Here, we sought to evaluate the ability of MET inhibition to radiosensitize HNSCC cells by conducting clonogenic assays. As a first step, we tested whether treatment with tepotinib prior to or following IR had an impact on colony formation in UM-SCC-81A and UM-SCC-81B cells. The two treatment schedules had no significant impact on results (Supplementary Fig. S4A and S4B), and further assays were conducted by pretreating the samples.

Pretreatment with tepotinib (50 nmol/L) significantly sensitized 8 of the 9 cell lines evaluated in this study to IR when assessing their survival, however with very variable degrees of radiosensitization (significant RERs 1.08–1.23). Indeed, Detroit-562, UM-SCC-81A, and UM-SCC-81B cell lines displayed RERs that were significant but very close to 1, denoting poor radiosensitization with 50 nmol/L of tepotinib. In contrast, combination of IR with tepotinib 100 nmol/L resulted in higher RERs, except in FaDu and UM-SCC-10B (the latter consistently did not reach a significant RER when comparing tepotinib 50 vs. 100 nmol/L). An increase in tepotinib concentration from 50 to 100 nmol/L further enhanced the radiosensitizing effect of MET inhibition only in 4 of 9 cell lines in a dose-dependent manner, this effect not being marked in UM-SCC-10A with an RER of 1.037 (Fig. 3A; Supplementary Fig. S4C).

Based on the potential radiosensitization in vitro, we evaluated tepotinib along with a clinically relevant fractionated IR dose of 2 Gy in subcutaneous xenografts. FaDu and Detroit-562 were used for in vivo experiments. Animals were irradiated daily 5 days a week with a 2-day hiatus. Tepotinib was administered daily. Tepotinib alone did not display a significant effect on FaDu cell lines but it did in Detroit-562 (P < 0.05, Fig. 3B and C). Concerning IR, both xenograft models showed significant responses to this modality alone (P < 0.001). Nevertheless, although IR resulted in tumor regression in Detroit-562 xenografts, it led to baseline tumor sizes in the FaDu model. Given the strong response of Detroit-562 to IR alone, combination with tepotinib did not have an additional effect in these xenografts (Fig. 3B and C). In contrast, combination of IR and tepotinib in FaDu xenografts led to significantly higher levels of tumor control (P < 0.001, Fig. 3B and C). Phospho-histone H3 was used as surrogate for proliferation, equally confirming a more marked resistance to IR in FaDu models, an effect that could be countered by combination with tepotinib (Fig. 3D).

Next, we evaluated the radiosensitization potential of MET targeting in translationally relevant patient-derived OTCs, a model that allows preserving tumoral histoarchitecture ex vivo (26). For this purpose, after an incubation time of 6 to 12 hours in culture medium, OTCs were treated with tepotinib 1 μmol/L (drug concentration was increased as compared with in vitro experiments in order to ensure appropriate tissue penetration) or DMSO for 16 hours and subsequently irradiated (6 Gy). OTCs were cultured for a total time of 72 hours since the beginning of treatment. Evaluation of proliferation patterns upon different therapeutic modalities revealed substantial heterogeneity in responses (Supplementary Fig. S5A). Ki67 expression was significantly reduced upon IR and combination of MET inhibition and IR when compared with vehicle (DMSO) or tepotinib alone, but combination of the two perturbations was not globally superior to IR as single therapy. Importantly, however, when considering individual OTCs and clustering those with similar responses to IR/combination, four basic patterns emerged (Fig. 4A and B): (i) OTCs sensitive to IR not further sensitized by concomitant MET inhibition (6/24); (ii) OTCs sensitive to IR further sensitized by MET inhibition (4/24); (iii) OTCs not significantly responsive to either IR or combination (8/24); and (iv) OTCs initially resistant to IR sensitized by concomitant MET inhibition (6/24). High MET expression was observed in tissues both responsive and resistant to MET inhibition, but pattern 4 contained only MET-high tissues (Fig. 4C). These results suggest that increased MET expression, while necessary in order to achieve radiosensitization, is not sufficient for this effect. It is important to note that p-MET expression could not be detected (discussed below). Finally, tepotinib alone at a concentration of 1 μmol/L did not result in significant decrease of Ki67 expression in any of the patterns (Supplementary Fig. S5B).

Radionsensitization following EGFR targeting in HNSCC has previously been shown to be the consequence of enhanced apoptosis and DNA damage (30). To assess whether MET inhibition enhanced IR-induced cell death in HNSCC models, cells were exposed to tepotinib (100 nmol/L) or DMSO for 6 hours prior to IR (6 Gy), and caspase-3 enzymatic activity was determined 3 days after treatment. As expected, IR significantly induced caspase-3 activation in all 9 evaluated cell lines (Fig. 5A). Combination of tepotinib and IR was significantly superior to IR alone in inducing caspase-3 activity in 6 of 9 cell lines (all of which have high basal levels of p-MET). The findings from caspase-3 assays were further confirmed with live/dead assays (Fig. 5B; Supplementary Fig. S6A). To elucidate potential molecules and pathways involved in radiosensitization, we selected UM-SCC-10B as a model of MET inhibition–radiosensitized cell line and HN5 (nonsensitized by MET inhibition). Cells were treated with either vehicle or tepotinib (100 nmol/L) 16 hours prior to IR. To capture early (though not immediate) events, cells were lysed 6 hours following IR and subjected to CyTOF as described in Materials and Methods and Supplementary Methods. The basal expression levels of all tested proteins differed in both cell lines, especially EGFR, EphA2, p-S6, and p-p53. Importantly, treatments had basically no effect on HN5, but were able to modulate several proteins in UM-SCC-10B. Most relevant perhaps was the downregulation of EphA2 and p-S6 upon MET inhibition (whether alone or in combination with IR; Supplementary Fig. S6B and S6C). Next, we combined the results of the live/dead assays presented in Fig. 5B (combination vs. IR alone) with the CyTOF data and used logistic regression to model the cell death probability. We finally computed the death odds ratio (OR) between the 10th and 90th percentiles of the transformed protein expression values. Figure 5C shows the proteins from the CyTOF panel with the largest OR values in the tepotinib-radiosensitized cell line UM-SCC-10B (a full list of ORs for each protein is provided in Supplementary Table S2). In this model, an OR above 1 indicates that increased expression of the protein is associated (although not necessarily in a causal way) with cell death, whereas an OR below 1 implies the opposite. Our model points out that expression of the RTK EphA2 and the PI3K downstream effector p-S6 seem to be associated with cell survival upon IR. As expected, expression of activated p-p53 and Ki67 was associated with cell death. The analysis results for the 19 proteins included in the CyTOF panel are shown in Supplementary Fig. S6D. The relevance of EGFR expression being associated with cell death is difficult to interpret from a biological perspective. However, because EGFR expression can be a mechanism of compensation upon MET inhibition (31), it could be hypothesized that upon MET inhibition and/or IR, EGFR expression occurs in a cell about to engage in cell death. In such a scenario, our model would interpret EGFR expression not as a prosurvival mechanism, but as an event associated with cell death.

To further explore the effects of tepotinib in HNSCC cell lines, cell-cycle analysis was carried out in cells treated with either tepotinib (100 nmol/L), IR (6 Gy), or their combination for 48 hours. IR led to G₂–M accumulation in both cell lines, whereas combination of tepotinib and IR further increased the G₂–M fraction only in UM-SCC-10B cells (Fig. 6A). As IR-induced prolonged G₂-arrest has been shown to potentially induce G₂–M checkpoint–regulated senescence (32) and we previously demonstrated that MET targeting enhances IR-induced senescence in gastric models featuring MET oncogene addiction (18), we retrospectively reviewed the morphology of irradiated but living UM-SCC-10B cells in live–dead assays, noticing a flattened and swollen shape suggestive of senescence. Based on these observations, we performed senescence-associated β-galactosidase (SA-β-gal) stainings in UM-SCC-10B and could show that in addition to induction of cell death, IR significantly increased the percentage of SA-β-gal–positive cells, which was further increased upon addition of MET inhibition (Fig. 6B).

Finally, we assessed whether tepotinib enhanced IR-induced DNA damage, thus leading to the observed radiosensitization and increase in cell death and cellular senescence. Phosphorylated levels of the histone variant H2AX at Ser139 (γH2AX) were used as surrogate marker of DNA double-strand breaks (33). UM-SCC-10B and HN5 cells were lysed at 2 and 8 hours after IR. Pretreatment with tepotinib resulted in increased levels of γH2AX 2 hours after IR in both cell lines, but whereas γH2AX levels in HN5 cells considerably decreased 6 hours later, tepotinib (100 nmol/L) pretreatment in UM-SCC-10B cells led to slightly more sustained H2AX phosphorylation after 8 hours, indicating slower resolution of DNA damage (Fig. 6C). Similar findings were observed in UM-SCC-81A and UM-SCC-81B (Supplementary Fig. S6E).

In this study, we evaluate the clinical impact of MET expression in PTs and LNMs as well as the potential radiosensitizing effect of MET inhibition in preclinical HNSCC models.

The MET RTK and its ligand SF/HGF are commonly overexpressed in HNSCC, and their high expression is in turn correlated with poor prognosis (13, 34–37). In terms of potential use of MET expression as a biomarker in HNSCC, our findings demonstrate that MET expression in PTs and LNMs is associated with decreased rates of regional disease control in univariable survival analyses. Classic prognostic factors were not significant independent risk factors in our multivariable disease control analysis. These results are most likely due to the specific composition of our study cohort, in which only patients with locoregionally advanced HNSCC treated with neck dissection and RT were included. Given that presence of LNMs is per se one of the most significant prognostic factors in HNSCC and that all patients in our cohort had LNMs, it is likely that the multivariable survival analysis did not reflect classic findings in cohorts of patients encompassing both early- and late-stage HNSCC (38). Consequently, our results are to be interpreted within the frame of aggressive locoregionally advanced disease, the presentation most likely to require multimodal therapeutic approaches.

We also report the findings of a MET mutational screening approach in a large cohort of primary HNSCCs and LNMs (patient-matched in 79.37% of the cases). In line with previous studies, MET mutations were found in almost 5% of patients, clustering predominantly in the extracellular and juxtamembrane receptor domains. The presence of mutations did not correlate with specific clinicopathologic features or survival (13, 39–41). Moreover, abundance of mutated alleles in PTs did not significantly differ from that in LNMs.

In this study, we provide the first evaluation of the radiosensitization potential of tepotinib, a novel and highly selective MET inhibitor, in HNSCC (22). So far, the role of MET targeting as a strategy for radiosensitization in HNSCC has not been comprehensively investigated. To the best of our knowledge, only a study by Baschnagel and colleagues (16) using the multikinase inhibitor crizotinib, which reported negative findings in vivo, has addressed this issue. Discordance between such negative results and our present findings is most likely due to differences in experimental design, models used (and specifically evaluation of OTCs in our study), treatments and potential use of crizotinib instead of a selective MET inhibitor. To the best of our knowledge, we provide the first evidence of MET inhibition-based radiosensitization in an in vivo HNSCC.

We found that tepotinib effectively blocked MET phosphorylation and reduced colony-forming ability in HNSCC cell lines. In combination with IR, tepotinib led to differential degrees of radiosensitization by enhancing IR-induced cell death, G₂–M cell-cycle arrest, senescence, and impairing repair of DNA double-strand in sensitive cell lines. An interesting observation was that, although tepotinib alone did not have any significant impact on cell proliferation at submicromolar concentrations, concentrations of 50 to 100 nmol/L were sufficient to result in radiosensitization. It can be argued that experimental settings and evaluated endpoints partly account for such observation, as metabolism-based cell proliferation assays evaluate short-term effects, whereas clonogenic assays focus on long-term outcome at the cellular level. Nevertheless, our results underline that MET signaling plays an important role in the stress-and-response process following IR (42). Indeed, in baseline conditions, HNSCCs do not seem to strongly rely on MET signaling. This is probably the result of the well-known RTK-mediated cooperation and redundancy phenomenon in HNSCC (15, 31, 43). In the presence of IR-induced genotoxic stress, however, prosurvival cell signaling circuitry undergoes significant rewiring. In such conditions, impairment of MET signaling may represent a major disadvantage at least in a subset of HNSCCs, ultimately resulting in radiosensitization (31, 42–44). Significantly, our CyTOF analysis pointed out the EphA2 RTK as a new potential compensatory mechanism upon MET inhibition.

The heterogeneity in responses to targeted therapy is most relevant in clinical terms as it is estimated that the vast majority of new anticancer drugs fail in clinical trials, an observation partly attributed to the preclinical models used (45). Indeed, cell lines and their derived murine xenografts, whereas of immense discovery and mechanistic value, fail to recapitulate the complexity of human neoplasms and their heterogeneity among other features (26). We opted to further explore the findings in cell lines and xenografts using OTCs from patients with HNSCC. OTCs allow preservation of in situ tumor complexity in ex vivo conditions. Even though the ability of OTCs to predict actual clinical responses remains to be demonstrated, and in spite of the fact that OTCs allow only short-term readouts, our current findings suggest the potential benefit of combined MET inhibition with IR seen in cell lines in vitro and in vivo and in particular subsets of HNSCC patients. Moreover, the ensemble of our preclinical observations strongly emphasizes the importance of pretherapeutic stratification.

In this respect, the genomic background has been shown to determine to a great extent responses to RTK targeting in several tumor entities. Indeed, gene amplification, activating mutations, and specific lesions such as MET exon 14 skipping, as well as RNA overexpression, have all been shown to render malignant cells dependent on MET signaling to sustain growth, proliferation, and other oncogenic features (46). Some of these alterations confer extreme sensitivity to MET inhibitors. In models featuring such exquisite sensitivity to MET inhibitors (referred to as oncogene addiction), targeting MET results in effective radiosensitization (8–10, 18). Consequently, the presence of such lesions could be used as a pretherapeutic stratification marker. In HNSCC, however, preclinical findings by us and others do not support the notion of MET oncogene addiction (13, 24). In order to identify pathways potentially contributing to MET inhibition–mediated radiosensitization, we explored the activation status of several signaling pathways in cell lines by means of CyTOF. This approach demonstrated that MET inhibition could modulate PI3K activity in the tepotinib-radiosensitized cell line UM-SCC-10B, but not in the resistant HN5 cells. The seeming relevance of PI3K signaling in mediating responses to MET targeting in HNSCC stands in contrast with previous findings in gastric cancer models, suggesting the MAPK pathway as the link between MET inhibition and radiosensitization (18). This observation indicates that different tumor entities and even different tumor types within an anatomic location may selectively use and rely upon different signaling nodes to regulate the cellular stress response following IR. Consequently, determining activation status or expression abundance of selected markers in each tumor entity could be beneficial for pretherapeutic stratification.

Finally, our results both in vivo and in OTCs indicate that MET expression seems to be necessary but not sufficient to predict radiosensitization. A limitation that needs to be acknowledged was the fact that p-MET expression could not at all be detected in archival FFPEs or OTCs, most likely due to a suboptimal tissue fixation protocol in spite of using a robust and thoroughly standardized staining used for routine histopathologic diagnosis (47). Consequently, we are not able to elaborate on any associations between OTCs' responses and MET activation status.

O. Elicin is a consultant at the advisory board meeting of AstraZeneca and Serono. M. Friese-Hamim is an Associate Director In Vivo Pharmacology Oncology for Merck Healthcare KGaA. C. Wilm is principal scientist at Merck Healthcare KGaA. C. Stroh is Director Translational and Biomarker Research at Merck KGaA. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Nisa, D.M. Aebersold, Y. Zimmer, M. Medová

Development of methodology: L. Nisa, R. Giger, O. Elicin, C. Wilm, M. Buchwalder, D.M. Aebersold, Y. Zimmer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Nisa, P. Francica, R. Giger, O. Elicin, M. Friese-Hamim, C. Wilm, C. Stroh, B. Bojaxhiu, M.D. Caversaccio, M.S. Dettmer, M. Buchwalder, T.M. Brodie

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Nisa, P. Francica, M. Medo, M. Friese-Hamim, C. Stroh, Y. Zimmer, M. Medová

Writing, review, and/or revision of the manuscript: L. Nisa, R. Giger, O. Elicin, M. Friese-Hamim, C. Stroh, M.D. Caversaccio, M.S. Dettmer, M. Buchwalder, D.M. Aebersold, Y. Zimmer, T.E. Carey, M. Medová

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Nisa, R. Giger, B. Bojaxhiu, A. Quintin, M.D. Caversaccio, M.S. Dettmer, M. Medová

Study supervision: L. Nisa, Y. Zimmer, M. Medová

Other (provided cell line models and read and approved manuscript): T.E. Carey

We cordially thank Dr. Vinko Tosevski for assistance with establishment of mass cytometry measurements and Prof. Manfred Claassen for helpful advises on single-cell data analysis. We are gratefully indebted to Mr. Bruno Streit and Dr. Wieslawa Blank-Liss for their outstanding technical help. The assistance of Mrs. Irina Ciorba-Nisa, MSc, in production of this manuscript, is most gratefully acknowledged. This work was supported by the Research Grant of the Inselspital for Young Clinicians (to L. Nisa), by a Bernese Cancer League grant and by the Stiftung zur Krebsbekämpfung (both to M. Medová). These funding sources had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.