Introduction

Around 90% of all head and neck cancers are squamous cell carcinoma [1] and almost one third originates in the oral cavity [2]. In general, the management of oral squamous cell carcinoma includes single modality surgery, radiotherapy, or combination of these modalities with or without systemic adjuvant therapy (chemotherapy and or targeted therapies) [2,3,4,5]. Radiotherapy with or without chemotherapy is given; as an adjuvant to primary surgery to improve locoregional control in cases with unfavorable histopathological features, as primary treatment for cases unsuitable for surgery or as part of organ preservation schemes, and as a salvage treatment in cases with persistent or recurrent disease [2,3,4,5]. Although increasingly aggressive chemotherapy and chemoradiotherapy result in high response rates and improved survival, the possibility exists that the point has been reached that further advances are outweighed by increased toxicities [6]. This explains the interest for the use of targeted therapies in the treatment of these tumors [6].

One target of interest is MET [6, 7], a transmembrane receptor tyrosine kinase [8] known to orchestrate invasive growth in head and neck squamous cell carcinoma [9]. Although high MET expression is associated with poor prognosis in various solid cancers [8]—including head and neck squamous cell carcinoma [10]—and numerous targeted therapies are under investigation [11, 12], major survival benefits have not yet been established [12, 13]. This, together with costs of approximately $100,000 per year of treatment per patient hampers the use of MET inhibitors in clinical practice [14, 15].

One explanation for the lack of evidence of clinical benefit of MET inhibitors may be found in the absence of suitable companion diagnostics [13], defined as “a medical device, often an in vitro device, which provides information that is essential for the safe and effective use of a corresponding drug or biological product” [16]. The development of companion diagnostics for targeted therapies directed against MET has been proven difficult for several reasons [13]. Some are of technical nature, such as the lack of reliable antibodies and optimal scoring methods [10, 11, 13], while other reasons may be due to biology, such as ectodomain shedding [13].

Similar to other receptors having intrinsic kinase activity, MET is subjective to proteolysis either dependent or independent of ligand stimulation [17]. While performing the validation of a series of C-terminal MET antibodies using a panel of cell lines [18], we observed specific degradation patterns under reducing conditions. On the one hand, these patterns were explained by caspase cleavage of MET under conditions of stress resulting in an intracellular proapoptotic 40 kDa fragment (p40β) [19, 20]. On the other hand, they were explained by MET ectodomain shedding [21] and presenilin-dependent regulated intramembrane proteolysis [22]. Ectodomain shedding is the release of substantial parts of the extracellular domain of proteins located at the membrane facilitated by proteinases such as ADAMs [23, 24]. The remaining C-terminal transmembrane-associated fragment for type 1 proteins like MET, may be internalized by endocytosis or might undergo further cleavage such as regulated intramembrane proteolysis. During presenilin-dependent regulated intramembrane proteolysis, the extracellular domain of MET is shed from the cell and the remaining 55 kDa membrane-anchored C-terminal fragment—referred to as p55β—is cleaved by the γ-secretase complex at the membrane [25]. The resulting 50 kDa protein fragment, being the intracellular domain of MET and referred to as p50β, is released into the cytosol and degraded by the proteasome [22]. It has been described that both internalized C-terminal fragments and intracellular domains alter cellular functions [26] and that MET ectodomain shedding unleashes the aggressive nature of the MET oncogene [27].

The goal of this study was to investigate whether MET ectodomain shedding occurs in oral squamous cell carcinoma cell lines and tissues (fresh frozen and formalin-fixed paraffin-embedded) and consider whether its presence is of predictive clinical value using 156 formalin-fixed paraffin-embedded oral squamous cell carcinoma sampled on a tissue microarray. To achieve this, two reliable antibodies directed against the C- or N-terminus of MET were used in parallel, e.g., D1C2 [18] and A2H2-3 [28].

Materials and methods

Cell lines

MET antibody validation cell line panel

The reliability of A2H2-3 and its performance in relation to D1C2 was investigated using a previously described cell line panel (HT-29, DU145, DU145#Sh167, PC3, HeLa, LNCaP, and SK-BR-3). Details regarding handling of the cells have been described [18].

Oral squamous cell carcinoma cell line panel

A series of six oral squamous cell carcinoma cell lines (SCC-4, SCC-9, SCC-25, UM-SCC-14C, BICR-16, and CAL-27) was used to examine whether MET ectodomain shedding and regulated intramembrane proteolysis occur in vitro. The cell lines were obtained from different departments within the Erasmus Medical Center and the University Medical Center Utrecht. Detailed information concerning the properties of these cell lines is provided in Supplementary Table 1, while information on culture conditions is provided in Supplementary Table 2.

Ethics statement

Human tissues and patient data were used according to “The Code of Conduct for Responsible Use” and “The Code of Conduct for the Used of Data in Health Research” as stated by the Federation of Dutch Medical Scientific Societies [29].

Patient tissues

To examine whether proteolytic processing of MET can be detected in tissues that are processed in a diagnostic setting, the following approach was taken. Tissues from 34 primary oral squamous cell carcinoma that were surgically removed between 2004 and 2011, were collected from the tissue bank of the department of Pathology of the Erasmus Medical Center. Using a microtome, 10 × 10-μm-thick slices were cut from the fresh frozen cancer tissues for protein isolation in view of western blotting and 3-μm-thick whole tissue sections were cut from the formalin-fixed paraffin-embedded cancer tissues in view of immunohistochemical analyses.

To examine the association between MET ectodomain shedding and survival, a tissue microarray representing 212 formalin-fixed paraffin-embedded primary oral squamous cell carcinomas—surgically removed between 1996 and 2005 in the University Medical Center Utrecht—was included in the study. In total, six tissue cores (0.6-mm diameter) were sampled from each cancer; three from the center and three from the periphery of the tumor. Further details concerning tissue microarray construction and histopathological scoring of the sampled cancers are described in [20]. Using a microtome, 4-μm-thick slices were cut from the formalin-fixed paraffin-embedded cancer tissues for immunohistochemical analyses.

The primary form of treatment for all included cancers (Erasmus Medical Center and University Medical Center Utrecht) was surgery with or without post-operative radiotherapy.

RNA isolation, cDNA generation, and quantitative real-time PCR

Culturing of the cell lines, RNA isolation, cDNA generation, quantitative real-time PCR were performed as previously described [18]. Included positive controls for MET were HeLa, PC3, and HT-29. LNCaP served as negative control. MET (Hs01565584_m1; Life TechnologiesTM/Applied Biosystems®, Grand Island, NY, USA) and HPRT1 (Hs02800695_m1; Life TechnologiesTM) mRNA expression levels were determined at a threshold of 0.2. Next, the expression levels of MET was standardized using the expression levels of HPRT1 according to relative standard curve method (4371095 Rev A; Life TechnologiesTM). Average raw Ct values for MET and HPRT1 can be found in Supplementary Table 3.

MET antibodies

Western blotting, immunohistochemistry and ELISA were performed using antibodies directed against the C- and/or N-terminus of MET. More specifically, D1C2 (Cell Signaling Technology®; Danvers, MA, USA) and A2H2-3 (Eli Lilly and Company; Indianapolis, USA). Further details concerning the antibodies are described in [18] for D1C2 and [28] for A2H2-3.

Protein isolation and western blot analysis using D1C2

Culturing of the cell lines, protein isolation, and western blot analysis using D1C2 under reducing conditions were performed as previously described [18]. Intensities of the relevant bands were evaluated as follows: absent, weak, moderate, and strong. Representative images can be found in the main text.

Assessment of MET ectodomain shedding and regulated intramembrane proteolysis using D1C2 under reducing conditions

For protein isolation, representative HE sections of the included fresh frozen tissues were used to verify if the tumor/stroma ratio of the analyzed samples was above 50%. Protein isolation and western blot analysis were performed using the same protocol as used for the cell lines.

Together, HeLa, PC3, and HT-29 served as a possible control for full length MET, MET C, and N-terminal fragments resulting after ectodomain shedding and regulated intramembrane proteolysis (Supplementary Tables 4, 5). Since all C- and N-terminal fragments are derived for the mature MET protein product, it is assumed that the intensities of C-terminal fragments under reducing conditions should be similar or weaker to the intensity observed for p145β within samples.

Protein isolation and western blot analyses using A2H2-3

Cells were cultured until 75–90% confluence, scraped in HBSS (Life TechnologiesTM), and frozen at −80 °C. Tumor cells were then lysed and immunoblotted with the A2H2-3 antibody essentially as described in [28]. Intensities of the relevant bands were evaluated as follows: absent, weak, moderate, and strong.

ELISA

The occurrence of MET ectodomain shedding was also examined by performing an ELISA on the culture media used to grow HeLa, PC3, LNCaP, HT-29, SCC-4, SCC-9, SCC-25, CAL-27, and UM-SCC-14C cell lines. The ELISA plate was coated with 2 μg/mL anti-MET antibody at 4 °C overnight. Following wash and block steps, either 100 μL of (a) MET ectodomain standard in diluent buffer or (b) tumor cell culture supernatants (dilutes 1:50 with diluent buffer) were added per well. After incubation, the plate was washed and 100 μL of 0.5 μg/mL biotinylated anti-MET antibody—that recognizes a different epitope than the anti-MET antibody used for coating—was added per well. Detection was carried out as described [28]. Signal was read at 450 nm with 570 nm correction on a Spectramax plate reader (Molecular Devices, LLC, San Jose, CA, USA) and evaluated as follows: absent, weak, moderate, and strong.

Immunohistochemistry

Staining of cell lines and whole tissue sections using D1C2 and A2H2-3 was essentially performed as described before [18, 28]. Two observers (SW and MDH) independently scored D1C2 and A2H2-3 immunoreactivity across the cell lines, whole tissue sections and tissue microarrays. Well-differentiated cancer cells that show no nuclei were omitted during scoring. Since the staining intensities for both antibodies were comparable for internal positive controls—e.g., endothelial cells lining veins (Supplementary Fig. 1)—present in the whole tissue sections and tissue microarrays, it was assumed that D1C2 and A2H2-3 membranous immunoreactivities can be compared when examining formalin-fixed paraffin-embedded tissues processed in a routine diagnostic setting.

Scoring of D1C2 and A2H2-3 immunoreactivity across cell lines and whole tissue sections as well as MET ectodomain shedding across whole tissue sections

For the cell lines, staining intensity was evaluated for the nucleus, cytoplasm, and membrane separately on a scale from 0 to 3 (Table 1). Each location was scored separately. For the patient derived whole tissue sections, intensities and scores were only evaluated for the membrane. Representative images of membranous A2H2-3 immunoreactivities observed on whole tissue sections are depicted in Fig. 1. Representative images of D1C2 intensities have been reported [18].

Table 1 Definition of staining intensities (SIs) and scores used to evaluate D1C2 and A2H2-3 immunoreactivity across cell lines and tissues and tissue microarray.
Full size table
Fig. 1
figure 1

Images of membranous A2H2-3 immunoreactivities observed for tissues (×20 objective).

Full size image

The percentage of cells subjective to ectodomain shedding across whole tissue sections was assessed using the differences between C- and N-terminal MET immunoreactivity observed using parallel sections (Table 1).

Scoring of D1C2 and A2H2-3 immunoreactivity, MET protein status, and ectodomain shedding across the tissue microarray

Scoring of membranous D1C2 and A2H2-3 across the tissue microarray was performed as described in Table 1. Seemingly suboptimal fixed cores were omitted during scoring.

Disagreement between the two observers was assessed using a Bland and Altman diagram and reevaluation was performed simultaneously until agreement was reached. To evaluate the consistency of A2H2-3 immunoreactivity within the cancer center and periphery, the intraclass correlation coefficient was calculated for both regions. To evaluate the difference between D1C2 as well as A2H2-3 immunoreactivity scored in the center and periphery, a paired t-test was performed. Calculations were performed with SPSS Statistics (version 24; IBM; Armonk, NY, USA). Unless otherwise mentioned, statistical significance was set at P value < 0.05.

Bubble plots used to illustrate N-terminal MET immunoreactivity across cancer surfaces were generated with Microsoft® Excel (version 2010; Microsoft®; Redmond, WA, USA).

To examine whether MET degradation can be observed using the tissue microarray, it was necessary to examine overall N-terminal MET immunoreactivity alongside overall C-terminal MET immunoreactivity. Therefore, central and peripheral scores for both C- and N-terminal MET were averaged. This was valid, since the results of paired t-tests show that there are no significant differences between the center and periphery for both A2H2-3 and D1C2 (results not shown).

To examine whether differences between C- and N-terminal MET immunoreactivities are associated with survival, all cancers—scored for D1C2 and A2H2-3—were assigned to one of three categories: MET negative, MET decoy receptor, or transmembranous MET with or without the ectodomain (Table 1).

Negative for MET was assigned when immunoreactivity was considered to be negative for both C- and N-terminal MET. Negative immunoreactivity—for both termini—was defined using the same boundaries as previously described [18] for the C-terminus being <10% immunoreactivity in both the center and periphery.

MET decoy receptor was assigned if—on average—more immunoreactivity for N-terminal than C-terminal MET was observed.

Transmembranous C-terminal MET having or not having the ectodomain was assigned if—on average—more or equal amounts of C-terminal MET compared with N-terminal MET was observed. Within this category the average % of cells subjective to ectodomain shedding was calculated by subtracting the average % of A2H2-3 positive cells from the average % of D1C2 positive cells (Table 1).

Association between MET ectodomain shedding and survival

The optimal threshold for ectodomain shedding in view of survival analysis was determined using the receiver operating characteristic curve analysis for disease-free survival and overall survival. Only cancers positive for transmembranous C-terminal MET were included and the area under the curve was used as a performance measure.

Survival analyses

Overall survival was defined as the time in months from the date of primary surgery to: the date of death due to any cause or the cutoff time (set at 60 months). Disease-free survival was defined as the time in months from the date of primary surgery to: the date of first evidence of any disease (local, regional, distant, or secondary primary) progression, the date of death due to any cause or the cutoff time (set at 60 months). Individuals who were: lost to follow-up or survived beyond the cutoff time, were considered as censored observations. For both survival measures, the cutoff time was set at 60 months. MET protein status and ectodomain shedding disease-free survival as well as overall survival curves were calculated by means of the Kaplan–Meier method and significance of differences in survival times was assessed with the log-rank test. Univariable as well as multivariable Cox proportional hazards regression models were used to evaluate the prognostic value of MET ectodomain shedding, demographical, clinical, and histopathological characteristics. Calculations were performed with SPSS Statistics (version 24; IBM). Unless otherwise mentioned, statistical significance was set at P value < 0.05.

Bubble plots used to illustrate the association between N-terminal MET immunoreactivity and survival were generated with Microsoft® Excel (version 2010; Microsoft®; Redmond, WA, USA).

Results

Reliability and performance of A2H2-3 in relation to D1C2

Under nonreducing conditions, the cell lines HT-29, DU145, PC3, HeLa, and SK-BR-3 showed immunoreactivity with bands migrating as p170 and p145β using this antibody (Fig. 2a). All observed intensities—ranging from weak to strong—were in accordance with measured MET mRNA expression levels. No immunoreactivity was observed for bands migrating otherwise than for MET protein products and NTFs (Supplementary Tables 4, 5). Although A2H2-3 immunoreactivity observed under formalin-fixed paraffin-embedded conditions was heterogeneous, all but one cell line (SK-BR-3) showed scores that are in line with MET mRNA expression levels (Fig. 2b).

Fig. 2: Performance of A2H2-3 across the antibody validation cell line panel.
figure 2

a Immunoreactivities observed for p145N-term. MET (A2H2-3) under nonreducing conditions, immunoreactivities observed for p145C-term. MET (D1C2) under reducing conditions, quantitative real-time PCR (qRT-PCR) results showing average MET fluorescence standardized to average HPRT1 fluorescence. b Negative controls (NCs) and membranous (M), cytoplasmic (C), and nuclear (N) immunohistochemical reactivities observed for A2H2-3 and D1C2. c Observed immunoreactivities using western blotting, mRNA expression levels, and immunoreactivities using immunohistochemistry. Reference results retrieved from [18].

Full size image

Figure 2a also illustrates that D1C2 and A2H2-3 perform comparably in terms of specificity and sensitivity in the detection of p170 and p145β under reducing and nonreducing conditions.

Figure 2b shows that when using immunohistochemistry both antibodies specifically detect cytoplasmic MET with comparable results across all cell lines. Moreover, it shows that although both antibodies specifically detect membranous MET, A2H2-3 has a higher sensitivity compared with D1C2 (PC3 and SK-BR-3).

Occurrence of MET ectodomain shedding and regulated intramembrane proteolysis in oral squamous cell carcinoma cell lines and tissues

Measured MET mRNA expression levels for the included controls (HeLa, PC3, HT-29, and LNCaP) were as expected (Fig. 3a). Moreover, all oral squamous cell carcinoma cell lines reliably express MET with varying levels (Fig. 3a). With the exception of SCC-25, immunoreactivities observed for D1C2 and A2H2-3 are in line with MET mRNA expression levels (Fig. 3a, b). Observed C-terminal p145β immunoreactivity under reducing conditions is similar to N-terminal MET p145β immunoreactivity under nonreducing conditions.

Fig. 3: Occurrence of MET ectodomain shedding and regulated intramembrane proteolysis in oral squamous cell carcinoma cell lines.
figure 3

a Quantitative real-time PCR (qRT-PCR) results showing average MET fluorescence standardized to average HPRT1 fluorescence. b Immunoreactivities observed for p145N-term. MET (A2H2-3) under nonreducing conditions, immunoreactivities observed for p145C-term. MET (D1C2) under reducing conditions. c Specific immunoreactivities observed using D1C2 showing C-terminal MET protein products and fragments under reducing conditions that are the result of shedding and/or regulated intramembrane proteolysis ([18], Supplementary Table 5). d Immunoreactivities for MET ectodomain shedding observed with A2H2-3. e Legend for observed mRNA expression levels, immunoreactivities using western blotting and immunoreactivities using ELISA. Reference results retrieved from [18] for HeLa, PC3, LNCaP, and HT-29.

Full size image

SCC-4, SCC-25, CAL-27, and UM-SCC-14C showed weak immunoreactivity for p70β, p90β, and p95β indicating that they are subjective to ectodomain shedding. In contrast, SCC-9 shows no immunoreactivity indicating ectodomain shedding (Fig. 3c). An independent nonreduced blot, generated using A2H2-3, confirms that all examined oral squamous cell carcinoma cell lines express p145β and p170 (Supplementary Fig. 2). An independent ELISA—using the same antibody—on the culture media used to grow the oral squamous cell carcinoma cell lines, shows that all examined cell lines shed the N-terminus of MET with varying levels (Fig. 3d).

Subsequently, it was examined whether ectodomain shedding and regulated intramembrane proteolysis can be detected in patient derived fresh frozen tissues (n = 44) under reducing conditions using D1C2. Six of the 44 examined tissues were excluded from further analysis because of the absence of immunoreactivity of both C-terminal MET and the endogenous control (b-actin). The remaining 38 tissues represent 30 surgically removed primary cancers (Fig. 4). The immunoreactivity of (b-actin) was low (scores 0 and 1) in 10 of the 37 analyzed tissues (27%). The purple heatmap (Fig. 5a), depicting D1C2 immunoreactivity under reducing conditions, shows that C-terminal MET protein fragments can be detected that are consistent with ectodomain shedding (p95β, p90β, p70β, p60β, and p55β) and regulated intramembrane proteolysis (p50β). The observed immunoreactivity patterns can be classified into five categories: (1) negative for p145β and C-terminal fragments (n = 4), (2) positive for p145β and negative for C-terminal fragments (n = 18), (3) positive for p145β and C-terminal fragments consistent with ectodomain shedding (n = 2), (4) positive for p145β and C-terminal fragments consistent with regulated intramembrane proteolysis (n = 2), and (5) positive for p145β and C-terminal fragments consistent with ectodomain shedding and regulated intramembrane proteolysis (n = 11, Fig. 5a, b). The results for multiple samples derived from 1, 2, and 25 are consistent. They show no signs of ectodomain shedding and/or regulated intramembrane proteolysis. In contrast, the results for the biological duplicates of cancers 26, 27, and 30, show inconsistent results for the presence of ectodomain shedding and/or regulated intramembrane proteolysis. The latter suggests that if ectodomain shedding and regulated intramembrane proteolysis occur, it might be heterogeneous. Taking everything into consideration, it is concluded that using fresh frozen tissues there are cancers without C-terminal MET immunoreactivity (n = 4, 13%) and cancers positive for C-terminal MET immunoreactivity (n = 26, 87%). Within the latter group (positive for C-terminal p145β), about half are not subjective to ectodomain shedding and/or regulated intramembrane proteolysis (n = 12, 46%) while others are subjective to ectodomain shedding (n = 2, 8%), regulated intramembrane proteolysis (n = 1, 4%), or both (n = 11, 42%).

Fig. 4: D1C2 immunoreactivities obtained under reducing conditions for 44 fresh frozen tissues.
figure 4

NB 1 specimen (T12*) was excluded from further analysis, since it was the only included regional lymph node metastasis. If multiple samples were analyzed from the same cancer, the T numbers were indexed with.X (e.g., T1.1 and T1.2).

Full size image
Fig. 5: Occurrence of MET ectodomain shedding and regulated intramembrane proteolysis in corresponding oral squamous cell carcinoma fresh frozen and formalin-fixed paraffin-embedded tissues.
figure 5

The abbreviations RIPping, WB, and IHC used in this figure stand for regulated intramembrane proteolysis, western blot, and immunohistochemistry. a Heatmap depicting observed immunoreactivities for C-terminal MET protein products and fragments (Supplementary Tables 4, 5) under reducing conditions observed with D1C2 and a ribbon depicting shedding and/or regulated intramembrane proteolysis status per tissue specimen based on D1C2 immunoreactivity under reducing conditions. (1) Negative for p145β and C-terminal fragments. (2) Positive for p145β and negative for C-terminal fragments. (3) Positive for p145β and C-terminal fragments that are the result of ectodomain shedding. (4) Positive for p145β and C-terminal fragments that are the result of regulated intramembrane proteolysis. (5) Positive for p145β and C-terminal fragments that are the result of ectodomain shedding and regulated intramembrane proteolysis. b Shedding status under reducing conditions per cancer specimen based on the results depicted in a and shedding status based on difference in immunoreactivity positivity between D1C2 and A2H2-3 (formalin-fixed paraffin-embedded). Cancers were scored positive for ectodomain shedding if at least 10% of the cancer cells are subjective to ectodomain shedding. c Legend for observed D1C2 immunoreactivities under reducing conditions and occurrence of MET ectodomain shedding under reducing and formalin-fixed paraffin-embedded conditions.

Full size image

Finally, it was examined whether ectodomain shedding can be detected in corresponding formalin-fixed paraffin-embedded tissues. Comparing D1C2 and A2H2-3 immunoreactivities observed across parallel slides, it can be concluded that ectodomain shedding can be detected using immunohistochemistry and that it occurs in 50% (n = 15) of the examined cancers (Fig. 5b). Figure 5b also shows that 13 cancers are subjective to ectodomain shedding according to western blotting and 15 cancers are subjective to ectodomain shedding according to immunohistochemistry, showing an overlap of eight cases. A representative ectodomain shedding area observed using immunohistochemistry is depicted in Supplementary Fig. 3. All cancers subjective to shedding were positive for D1C2 (Supplementary Fig. 4).

General results tissue microarray

The intraclass correlation coefficient for the center cores was 0.88 (95% CI, 0.84–0.91) and for the periphery cores was 0.87 (95% CI, 0.82–0.91). This indicates optimal agreement for both regions.

Evaluation of average membranous D1C2 and A2H2-3 immunoreactivity in both cancer regions (center and periphery) was possible for 156 oral squamous cell carcinoma. The baseline characteristics of these cancers are indicated in Table 2.

Table 2 Baseline characteristics for the entire sample population (n = 156).
Full size table

Evaluation of N-terminal MET immunoreactivity across cancers using a scatter plot based analysis and its association with prognosis

N-terminal MET immunoreactivity is generally low across cancers and it shows no relationship with disease-free survival or overall survival of the corresponding patients (Supplementary Figs. 5A, B). This was confirmed for overall survival by performing a receiver operating characteristic curve analyses for average A2H2-3 immunoreactivity. In contrast, receiver operating characteristic curve analysis for disease-free survival showed an association (P = 0.018) between an average A2H2-3 immunoreactivity ≤2% and disease-free survival using the area under the curve as a performance measure (results not shown).

Evaluation of MET protein status across the tissue microarray and its association with prognosis

Using the definitions for MET protein status, it can be stated that 36 cancers are negative for MET (23%), 14 cancers are positive for the MET decoy receptor (9%), and 106 cancers are positive for transmembranous C-terminal MET (68%, Supplementary Fig. 6).

Univariable survival analyses reveals that patients with cancers negative for MET immunoreactivity perform significantly worse than those with cancers positive for transmembranous C-terminal MET in terms of disease-free survival (HR = 1.84; 95% CI, 1.16–2.93 and P = 0.010; Fig. 6a; Supplementary Table 6) and overall survival (HR = 2.00; 95% CI, 1.24–3.23 and P = 0.005; Fig. 6b; Supplementary Table 7). The decoy receptor did not significantly contribute to the model for either disease-free survival or overall survival.

Fig. 6: Kaplan–Meier curves.
figure 6

a Disease-free survival for all patients, stratified by MET receptor status. b Overall survival for all patients, stratified by MET receptor status.

Full size image

Evaluation of MET ectodomain shedding in C-terminal MET positive cancers and its association with prognosis

Based on the receiver operating characteristic analyses for disease-free survival and overall survival, absence of MET ectodomain shedding was defined as ectodomain shedding in <35% of cancer cells and presence of ectodomain shedding was defined as ≥35% of cancer cells (Supplementary Tables 8–11; Supplementary Figs. 7, 8). Univariable survival analyses performed for ectodomain shedding and the set of clinico-histopathological characteristics listed in Table 3 show that patients with cancers subjective to ectodomain shedding (n = 38, 36%) perform significantly worse in terms of disease-free survival (HR = 2.05; 95% CI, 1.20–3.51 and P = 0.009; Fig. 7a) and overall survival (HR = 1.83; 95%, CI 1.04–3.23 and P = 0.038; Fig. 7b). To test the independent value of MET ectodomain shedding for disease-free survival and overall survival, multivariable analyses were performed correcting for age at diagnosis, pT, pN, extranodal growth, degree of differentiated, and MET ectodomain shedding (Table 4). The results show that MET ectodomain shedding only remains significantly associated with disease-free survival (HR = 2.41; 95% CI, 1.35–4.30 and P = 0.003) opposed to overall survival (HR = 1.64; 95% CI, 0.920–2.94 and P = 0.095).

Table 3 The—univariable—effect of clinico-histopathological characteristics on disease-free survival and overall survival of patients diagnosed with cancers positive for transmembranous C-terminal MET with or without the ectodomain.
Full size table
Fig. 7: Kaplan–Meier curves.
figure 7

a Disease-free survival for all patients with cancers positive for transmembranous C-terminal MET, stratified by MET ectodomain shedding. b Overall survival for all patients with cancers positive for transmembranous C-terminal MET, stratified by MET ectodomain shedding.

Full size image
Table 4 Multivariable analysis of disease-free survival and overall survival for patients having cancers showing transmembranous C-terminal MET with or without the ectodomain.
Full size table

Discussion

The results presented here illustrate that A2H2-3 is specific in the detection of N-terminal MET under nonreducing and formalin-fixed paraffin-embedded conditions and is particularly sensitive using immunohistochemistry. ELISA is the most sensitive method to detect ectodomain shedding in vitro and shows that all investigated oral squamous cell carcinoma cell lines shed their N-terminus. Staining of C- and N-terminal MET using formalin-fixed paraffin-embedded parallel whole tissue sections shows that MET ectodomain shedding occurs in oral squamous cell carcinoma, which is associated with poor survival.

In spite of the finding that ≤2% of average N-terminal MET immunoreactivity is associated with poor disease-free survival, the prognostic relevance of this result was not further pursued because such a low threshold is impractical in a diagnostic setting.

It was decided to use western blotting to demonstrate the occurrence of ectodomain shedding in fresh frozen tissues, because the results show which—potentially oncogenic [21, 27]—C-terminal transmembrane-associated fragments are present. Although immunoreactivity of b-actin was low in about a quarter of the samples (Fig. 4), they were not excluded because b-actin can vary depending on grade and composition of the tumor [30, 31]. Since other traditional reference genes, e.g., GAPDH, used as normalizers for mRNA expression data have also been described to show variable gene expression levels across cancers, we decided not to switch to another internal positive control protein [32]. Verifying the absence of D1C2 immunoreactivity using ELISA on harvested culture media of the examined oral squamous cell carcinoma cell lines, could have been an additional confirmation of the occurrence of ectodomain shedding. As it is strongly believed that the totality of the presented data are supportive for the occurrence of ectodomain shedding in vitro, such an experiment was not performed.

When comparing C-terminal MET immunoreactivities obtained using western blotting and immunohistochemistry, discrepancies are observed (positive versus negative, Supplementary Fig. 4). This probably has to do with only scoring membranous immunoreactivity under formalin-fixed paraffin-embedded conditions. Absence of membranous immunoreactivity does not imply absence of cytoplasmic immunoreactivity (Fig. 8) representing either the precursor, the mature receptor and/or C-terminal fragments [17]. Discrepancies between reducing and formalin-fixed paraffin-embedded conditions can also be explained by tumor heterogeneity. The cells sampled in the fresh frozen tissues may represent a different clone of cancer cells than the cells sampled in the formalin-fixed paraffin-embedded tissues of corresponding cancers. Tumor heterogeneity can also explain the observation that two cancers were scored negative under reducing conditions and positive using immunohistochemistry (Fig. 8). Another possible explanation might be differences in histopathology of the examined oral squamous cell carcinoma. However, we find the number of included samples too low to perform such analyses.

Fig. 8: 4 × 4 matrix illustrating agreement and discrepancies for D1C2 −/+ immunoreactivity under reducing and formalin-fixed paraffin-embedded conditions (×5 objective).
figure 8

a Combinations of D1C2 −/+ immunoreactivity observed under reducing versus formalin-fixed paraffin-embedded conditions. Illustrated with photographs of observed D1C2 immunoreactivities in four examined formalin-fixed paraffin-embedded cancers. b Photographs of parallel sections stained with A2H2-3 of the same cancer regions depicted under a showing that both stainings can behave similarly.

Full size image

Tumor heterogeneity is also a possible explanation for the inconsistencies observed between reducing and formalin-fixed paraffin-embedded conditions with respect to ectodomain shedding (Fig. 9). Therefore, we assume that both methods—[1] western blotting using D1C2 and fresh frozen tissues and [2] immunohistochemistry using D1C2 and A2H2-3 on parallel formalin-fixed paraffin-embedded sections—reliably depict the occurrence of MET ectodomain shedding. Since immunohistochemistry is widely used in routine diagnostics, we opted to use this technique to investigate whether there is an association between MET ectodomain shedding and survival.

Fig. 9: 4 × 4 matrix illustrating agreement and discrepancies for the presence of MET ectodomain shedding under reducing and formalin-fixed paraffin-embedded conditions (×5 objective).
figure 9

a Combinations of MET ectodomain shedding −/+ observed under reducing versus formalin-fixed paraffin-embedded conditions. Illustrated with photographs of observed D1C2 immunoreactivities in four examined formalin-fixed paraffin-embedded cancers. b Photographs of parallel sections stained with A2H2-3 of the same cancer regions depicted under a illustrating that D1C2 and A2H2-3 immunoreactivity can be identical—first row—or that there can be MET ectodomain shedding—second row.

Full size image

We have established previously—using the same tissue microarray—that membranous D1C2 immunoreactivity is either constant (uniform negative or positive staining) across or differs between the center and periphery (variable staining) of oral squamous cell carcinoma (n = 157) and HPV negative oropharyngeal squamous cell carcinoma (n = 22) [18]. Univariable survival analyses revealed that both uniform negative (n = 41) and positive (n = 16) staining patterns perform significantly worse than the variable staining pattern (n = 122) in terms of disease-free survival (HR = 2.00; 95% CI, 1.29–3.08 and P = 0.002 for uniform negativity and HR = 1.92; 95% CI, 1.01–3.65 and P = 0.048 for uniform positivity) and overall survival (HR = 2.17; 95% CI, 1.39–3.41 and P = 0.001 for uniform negativity and HR = 2.23; 95% CI, 1.16–4.27 and P = 0.016 for uniform positivity). Although counterintuitive, the finding that both low and high expression of a receptor tyrosine kinase is associated with poor prognosis is for example also reported for ERBB2 in primary breast cancers [33]. The association of the uniform positive staining pattern with poor prognosis concurs with the knowledge that increased MET immunoreactivity has been described to be associated with poor prognosis of head and neck squamous cell carcinoma [10, 34]. Subsequent multivariable survival analyses—combining the uniform negative and positive staining patterns—revealed that the uniform staining pattern showed an interaction with vasoinvasive growth [18]. Assuming that the effect of the MET staining pattern was subservient to that of vasoinvasive growth, final multivariable survival analyses for disease-free survival (HR = 2.92; 95% CI, 1.80–4.76 and P < 0.001) and overall survival (HR = 3.48; 95% CI, 2.08–5.80 and P = 0.001) was restricted to patients showing no histopathological signs of vasoinvasive growth (n = 136). Given the low number of HPV negative oropharyngeal SCC, it was decided to exclude these cancers from the survival analyses performed in this paper.

The disease-free survival (HR = 1.72; 95% CI, 1.07–2.77 and P = 0.025) and overall survival (HR = 1.74; 95% CI, 1.06–2.87 and P = 0.030) of the 156 oral squamous cell carcinoma patients included for MET protein status analysis are significantly lower than those of the excluded oral squamous cell carcinoma patients (n = 54). A possible explanation might be the high number of stage I cancers in the excluded sample population (Supplementary Table 12).

When investigating the association of MET protein status with disease-free survival and overall survival using univariable survival analysis, the results showed that patients with MET negative cancers perform significantly worse for both survival measures than those with cancers showing the MET decoy receptor or transmembranous MET with or without the ectodomain. This finding is in line with the result described above and in [18] that C-terminal MET negativity is associated with poor disease-free survival and overall survival. This is as expected, since the boundaries used here for MET negativity were the same as those used in [18] and all cancers negative for C-terminal MET were also negative for N-terminal MET.

Univariable survival analysis was also performed for MET ectodomain shedding and for clinico-histopathological characteristics of patients diagnosed with cancers positive for transmembranous C-terminal MET. Within this category, 14 (93.3%) cancers show the uniform positive staining pattern for C-terminal MET ([18], Supplementary Table 13). Therefore, it was excluded that MET ectodomain shedding showed an interaction with vasoinvasive growth. Consequently, all cancers defined as positive for transmembranous C-terminal MET were included for multivariable survival analyses. In view of power, multivariable correction was restricted to six variables known to be associated with poor survival [24, 35,36,37]. More specifically, ectodomain shedding, age at diagnosis, pT, pN, extranodal growth, and degree of differentiation. The fact that ectodomain shedding has a significant contribution for disease-free survival, makes it a promising predictive marker for oral squamous cell carcinoma patients. Moreover, this result can be explained on biological grounds, namely by the finding that deletion of the ectodomain unleashes the aggressive potential of cancer cells [21, 27].

There is no association between MET protein status and stage, nor between shedding and stage for C-terminal MET positive cancers (Supplementary Tables 14, 15). This implies that patients with both early and late stage oral squamous cell carcinoma might be eligible for treatment with biologicals directed against MET. Although shedding has no predictive value (only associated with disease-free survival), abundant C-terminal MET immunoreactivity has a predictive value (associated with disease-free survival) and overall survival [18]. Since 57.1% (n = 8) of patients showing abundant C-terminal MET immunoreactivity are subjective to ectodomain shedding, it might have an impact on the choice of therapy. In the future, it might be of interest to validate the occurrence of ectodomain shedding in oral squamous cell carcinoma using e.g., the expression of ADAMs as an alternative read-out. However, we find this beyond the scope of this paper.

In summary, we conclude that monitoring MET protein status by comparing C- and N-terminal immunohistochemistry might be of added value in the stratification of cancers eligible for treatment with targeted therapies directed against this receptor tyrosine kinase. It is hypothesized that patients having cancers classified as MET negative or positive for the decoy receptor—lacking the protein’s kinase domain—are not eligible for treatment with targeted therapies directed against MET. It is hypothesized that patients showing abundant immunoreactivity for the entire receptor might be eligible for treatment with either MET monoclonal antibodies or tyrosine kinase inhibitors. Finally, it is proposed that patients having cancers showing immunoreactivity for transmembranous C-terminal MET lacking the ectodomain—in more than 35% of the cancer cells—might benefit from MET tyrosine kinase inhibitors, treatment strategies directed against proteins that orchestrate ectodomain shedding—such as ADAMs [23, 38]—or MET monoclonal antibodies that induce ectodomain shedding having regulated intramembrane proteolysis as a consequence [39, 40]. The latter strategy assumes that these monoclonal antibodies induce physiological ectodomain shedding and possibly prevents the occurrence of the oncogenic C-terminal fragment.

To our knowledge we are the first to apply a combination of reliable C- and N-terminal MET antibodies to evaluate MET status in oral squamous cell carcinoma using immunohistochemistry. Taking this approach, we have not only shown the occurrence of MET ectodomain shedding in oral squamous cell carcinoma, but have also established its independent association with poor disease-free survival. Adding MET protein status to the diagnostic setting of patients diagnosed with oral squamous cell carcinoma, would add up to approximately $180 per patient [41]. This seems a small price to pay realizing that the use of biologicals adds up to approximately $100,000 per year of treatment per patient [14, 15]. These findings might aid in designing companion diagnostics for targeted therapies directed against MET (Fig. 10), ultimately improving overall survival of patients with cancers subjective to ectodomain shedding.

Fig. 10: Proposed companion diagnostics for the use of targeted therapies directed against MET in oral squamous cell carcinoma.
figure 10

a MET negative cancers not eligible for treatment with targeted therapies directed against MET. b Cancers positive for the decoy receptors not eligible for treatment with targeted therapies directed against MET. c Cancers abundantly positive for the entire—both C- and N-terminal—MET receptor eligible for treatment with monoclonal antibodies and tyrosine kinase inhibitors. d Cancers subjective to shedding—≥35% of cancer cells—eligible for treatment with tyrosine kinase inhibitors, treatment strategies directed against proteins that orchestrate ectodomain shedding (e.g., ADAMs), monoclonal antibodies that induce ectodomain shedding having regulated intramembrane proteolysis as a consequence.

Full size image