Discovery and Therapeutic Exploitation of Mechanisms of Resistance to MET Inhibitors in Glioblastoma

Glioblastoma (GBM) is the most common and most lethal primary malignant brain tumor (1). Current standard of care includes surgical resection, radiotherapy, and chemotherapy. Prognosis remains poor with a median survival of 15 months (2). New therapies targeting common aberrations in signal transduction pathways in GBM are currently being investigated (3). Dysregulation of receptor tyrosine kinases (RTK) have been found in approximately 90% of GBM (4, 5). Consequently, tyrosine kinase inhibitors have been developed for anticancer therapy.

MET is an RTK that is essential for embryonic development and tissue repair (2, 6). Hepatocyte growth factor (HGF), the only known ligand for MET, activates MET and downstream signaling pathways including RAS/MAPK, PI3K/AKT and STAT (7–9). MET is commonly dysregulated in GBM via various mechanisms including somatic mutations, rearrangement, amplification and overexpression of MET and HGF that leads to autocrine loop formation (10–12). Furthermore, MET expression inversely correlates with patient survival (12, 13) and is upregulated in GBM (5, 14, 15).

Several MET inhibitors are under investigation with mixed results in clinical trials (14). Crizotinib is an FDA-approved ATP-competitive small-molecule inhibitor of MET. Approved for the treatment of advanced and metastatic ALK-positive non–small cell lung cancer (NSCLC), crizotinib is in phase II clinical trials for CNS and solid brain tumors (5, 16). In NSCLC patients, crizotinib displays initial potent anticancer activity. However, acquired resistance frequently ensues, rendering this drug ineffective as a monotherapy (17). Onartuzumab is a monovalent monoclonal antibody that competes with HGF for binding to MET (18). Previous clinical trials involving onartuzumab as a monotherapy in NSCLC patients have been disappointing, highlighting the importance of understanding resistance mechanisms to the drug.

Although MET inhibitors have displayed initial efficacy, acquired resistance to single agent modalities invariably occurs (19). The development of acquired resistance to monotherapy encompasses multiple mechanisms such as acquisition of secondary mutations in therapeutic targets, activation of bypass signaling pathways, or immune evasion (20). Deciphering the exact mechanism of acquired resistance would be advantageous to halt disease progression in GBM patients by using combinatorial therapies (20–22). This study had two aims: (i) to elucidate the bypass signaling pathways that are activated when GBM cells acquire MET inhibitor resistance; (ii) to develop combinatorial drug therapies against MET inhibitor–resistant GBM. Our data uncovered a number of signaling molecules that are altered in MET inhibitor–resistant cell lines compared with sensitive cell lines. Upregulation of RTKs such as fibroblast growth factor receptor (FGFR) and EGFR, signaling molecules such as mTOR and signal transducer and activator of transcription 3 (STAT3), as well as elevation of cyclooxygenase 2 (COX-2) emphasized the extent of cross-talk between multiple signaling pathways in MET inhibitor–resistant GBM. Multitargeted combinational therapies against these molecules overcame single-agent MET inhibitor resistance, paving the way for new therapeutic approaches that could be tested in clinical trials.

Human GBM cell lines (U87 and U373) and stem cell line (GSC827) were used. U87 and U373 were from the ATCC and were authenticated through short tandem repeat (STR) profiling. GSC827 was isolated from GBM specimens and characterized for tumorigenesis, pluripotency, self-renewal, stem cell markers, and neurosphere formation (23). The specimens were obtained with written informed consent by the patients and the studies conducted in accordance with recognized ethical guidelines and approved by the institutional review board of the Cleveland Clinic. All cell lines were tested for mycoplasma.

U87, U373, and GSC827 cells were exposed to increasing concentrations of either crizotinib (from 1 to 100 nmol/L) or Onartuzumab (form 1 to 300 nmol/L) over a period of 6 months. Cell lines were then tested for MET inhibitor resistance through trypan blue death assay as previously described (24). For this, cells were treated with MET inhibitor [crizotinib 100 nmol/L or Onartuzumab 300 nmol/L] for 48 hours, collected and stained with trypan blue reagent. MET inhibitor–resistant cells were grown continually in the presence of either 100 nmol/L crizotinib or 300 nmol/L Onartuzumab.

Cell death was assessed by trypan blue assay as previously described (24). Cell proliferation was assessed by cell counting as previously described (25). All experiments were performed in triplicate. Apoptotic cell death was assessed by pre-treating cells with ZVAD (20 nmol/L) followed by combinational drug treatment for 48 hours. Cell death was determined as described above.

Proteomic screening was performed by reverse phase protein array (RPPA) as previously described (26–28). The cells were treated as described, in triplicates. Protein signaling analytes were chosen based on their previously described involvement in key aspects of tumor biology. Detection was performed using a fluorescence-based tyramide signal amplification strategy using Streptavidin-conjugated IRDye680 (LI-COR Biosciences, Lincoln NE) detection reagent. All antibodies were validated for single band specificity and for ligand-induction (for phospho-specific antibodies) by immunoblotting before use on the arrays as previously described (26–28). Each array was scanned using a TECAN LS (Vidar Systems Corporation). Spot intensity was analyzed, data were normalized to total protein and a standardized, single data value was generated for each sample on the array by MicroVigene software V2.999 (VigeneTech).

Immunoblotting was performed as previously described (29). Antibodies used were p.FGFR1, FGFR1, p.ERK, ERK, p.AKT, AKT, p.MET, p.mTOR, mTOR, COX-2 and p.STAT3 (Cell Signaling Technology), MET and GAPDH (Santa Cruz Biotechnology).

Functional rescue experiments were performed to determine whether molecules identified through RPPA screenings mediate resistance to MET inhibitors. Cells were pretreated with celecoxib (inhibits COX-2), debio-1347 (inhibits FGFR1), erlotinib (inhibits EGFR), rapamycin (inhibits mTOR) or STAT3 inhibitor as described for 2 hours then treated with either crizotinib or Onartuzumab for 48 hours. Cell death and proliferation was assessed as described above.

U87 cells were transfected with either scrambled siRNA (control), si-FGFR1, si-COX2, si-RON, si-RET, si-Vimentin or si-ERBB3 (Thermo Fisher Scientific) then treated with crizotinib (100 nmol/L) for 48 hours. Knockdown was confirmed by immunoblotting. Cells death was determined as described above.

Combination therapies to overcome resistance were assessed using an orthotopic xenograft mouse model. U87 cells (3 × 10⁵) were stereotactically implanted into the right corpus striatum of immunodeficient mice (n = 10 per treatment group). Six days after implantation, the animals were treated with either control vehicle (DMSO), agent either alone [debio-1347 (25 mg/kg) or celecoxib (10 mg/kg), crizotinib (25 mg/kg)] or in combination [debio-1347+crizotinib or celecoxib+crizotinib] by oral gavage daily from day 6 post-tumor implantation for 7 days. Tumor volumes were visualized and quantified by MRI. These studies were approved by the University of Virginia Animal Care and Use Committee.

The continuous variable RPPA data generated were subjected to both unsupervised and supervised statistical analyses, as previously described (24). Statistical analyses were performed on final microarray intensity values obtained using R version 2.9.2 software (The R Foundation for Statistical Computing). If the distribution of variables for the analyzed groups were normal, a two-sample t test was performed. If the variances of two groups were equal, two-sample t test with a pooled variance procedure was used to compare the means of intensity between two groups. Otherwise, two-sample t test without a pooled variance procedure was adopted. For non-normally distributed variables, the Wilcoxon rank sum test was used. Significance levels were set at P < 0.05. To evaluate the statistical significance of the difference between wildtype and resistant GBM cell proliferation and death, treated and control, we used a two sample t test and significance levels were set at P < 0.05. To evaluate the statistical significance of the difference between each treated and control animal groups in vivo, we used both two-sample t test and non-parametric Wilcoxen rank-sum test.

Crizotinib-resistant and onartuzumab-resistant GBM cell lines from U87, U373, and GSC827 were generated through dose escalation of each drug over 6 months until the cells were no longer sensitive to crizotinib at a concentration of 100 nmol/L or onartuzumab at a concentration of 300 nmol/L (Fig. 1A–D). When compared with the wild-type cells, the resistant cell lines exhibited no change in cell survival and proliferation when treated with either crizotinib or onartuzumab for 48 hours. Treatment of wildtype GBM cells with crizotinib or onartuzumab resulted in an antiproliferative effect and induction of apoptosis. Conversely, in crizotinib-resistant and onartuzumab-resistant GBM cells, cell survival and proliferation remained unaffected.

To elucidate the mechanism(s) of resistance of GBM cells to crizotinib and onartuzumab, we used proteomic screening with RPPA to compare protein expression and activation changes between wildtype and resistant GBM cells when treated with the MET inhibitors for 48 hours. The experiment was performed in triplicate. The data revealed several upregulated signaling pathways in the resistant cells. The most changed molecules were: p.EGFR, cABL, ALDH, Cyclin A and PDL1 increased 26.7, 4.3, 3.8, 3.3, and 3 fold respectively in crizotinib-resistant U373 cells; p.EGFR, p.ATPCL, and PDL1 increased 7.7, 2.6, 2.3-fold, respectively, in onartuzumab-resistant U373 cells; p.NFkB p65 and ERBB3 increased 2.7, 2-fold, respectively, in crizotinib-resistant U87 cells and COX-2, p.SRC, and p.ERK increased 2.7, 2.4, and 2.3-fold, respectively, in onartuzumab-resistant U87 cells compared with the wildtype after MET inhibitor treatment (Fig. 2). Additional significantly changed molecules are shown in Supplementary Figs. S2 and S3. In particular, MET inhibitor resistance resulted in the activation of the EGFR pathway, the FGFR pathway, the mTOR pathway, STAT3, and COX-2. Increased activation of these molecules in resistant cells was further induced by additional treatment with MET inhibitors, forcing the resistant cells to upregulate existing bypass signaling pathways. These expression changes were validated by immunoblotting (Fig. 2C). Several other changed molecules were also assessed but could either not be confirmed via immunoblotting, had no effect on the reversal of MET resistance, or had no available specific pharmacological inhibitors. These molecules were therefore not further investigated (Supplementary Table S1). Overall, the data demonstrate that MET inhibitor resistance is mediated through multiple key regulatory pathways that compensate for inhibition of the MET pathway (Fig. 2A and B, Supplementary Figs. S2 and S3). Celecoxib and debio-1347 were selected for evaluation in vivo, attributed to this, inhibition of COX-2 by celecoxib and FGFR by debio-1347, along with the effect on downstream pathways, was confirmed though immunoblotting (Supplementary Fig. S1).

To determine whether inhibition of select bypass-signaling pathways can re-sensitize the resistant cells to MET inhibitors, we treated the cells with MET inhibitors in combination with an inhibitor of one bypass signaling pathway and assessed the cell proliferation and death. A total of nine bypass signaling targets and their inhibitors were selected on the basis of the availability of drugs that inhibit them. These targets included FGFR1, EGFR, mTOR, STAT3, COX-2, RON, RET, Vimentin and ERBB3. Wildtype and resistant U87, U373, and GSC827 cells were pretreated for 2 hours with (i). COX-2 inhibitor, celecoxib (100 nmol/L, 25 nmol/L or 5 nmol/L, respectively), (ii). FGFR1 inhibitor, debio-1347 (10 μmol/L for U87 or 5 μmol/L for U373 and GSC827), (iii). mTOR inhibitor, rapamycin (100 nmol/L for U87 or 25 nmol/L for U373 and GSC827), (iv). STAT3 inhibitor (STAT3i; 50 μmol/L for U87 or 25 μmol/L for U373 and GSC827) or (v). EGFR inhibitor, erlotinib (100 nmol/L for U87 or 25 nmol/L for U373 and GSC827) then subjected to either crizotinib (100 nmol/L) or onartuzumab (300 nmol/L) for 48 hours. When used alone, all five inhibitors decreased cell proliferation and increased cell death in wildtype cells but not in the resistant cells. However, when the resistant cells were treated with a MET inhibitor and one of the five inhibitors, we observed increased cell death (Fig. 3A and B, 4A and B and 5A and B, Supplementary Figs. S4A–S4C, S5A–S5B, and S6A–S6E) and decreased proliferation (Fig. 3C–F, 4C–F and 5C and D, Supplementary Figs. S4D–S4I, S5C–S5F, and S6F–S6O), indicating restored sensitivity to MET inhibitors. Inhibitions of RON, RET, Vimentin and ERBB3 either did not successfully restore sensitivity to MET inhibitors or clinically applicable inhibitors were not available for use (Supplementary Fig. S1H–S1K). In addition, to determine the predominant mode of cell death mediated by the combination of celecoxib and crizotinib, we pretreated U87 WT and CR cells with ZVAD for 30 minutes before treating with celecoxib followed by crizotinib for 48 hours then assessed the effect on cell death. We show that ZVAD reduced cell death caused by the combinational treatment, indicating that apoptosis as the preliminary mode of cell death (Supplementary Fig. S4J and S4K). The above data demonstrate that when combined with either crizotinib or onartuzumab, celecoxib, debio-1347, rapamycin, STAT3i and erlotinib are all partially but significantly effective at overcoming MET inhibitor resistance in GBM cells. P values are stated in the figure legends.

To confirm that the cell death induced by celecoxib and debio-1347 was attributed to inhibition of COX2 and FGFR1 respectively, we silenced both COX2 and FGFR1 with siRNA and analyzed the effect on cell death. Cells were transfected with either si-control, si-COX2 or si-FGFR1 then treated with crizotinib for 48 hours. siRNA-mediated knockdown of COX2 and FGFR1 was verified by immunobloting (Supplementary Fig. S1G). To assess whether COX2 or FGFR1 silencing sensitized-resistant cells to MET inhibitors as celecoxib (inhibits COX2) and debio-1347 (inhibits FGFR1), cells were transfected as above and a trypan blue assay was performed. The data showed that silencing of COX2 and FGFR1 restored MET inhibitor sensitivity in crizotinib-resistant cells (Supplementary Fig. S1E and SF). The above data show that silencing of either COX2 or FGFR1 expression leads to comparable anticancer effects on crizotinib-resistant U87 cells as celecoxib or debio-1347 treatment.

GSCs play an important role in mediating resistance to cytotoxic therapies (5). MET inhibition reduces GSC population sensitizing the tumor to therapies (5). With this in mind, we tested these combinational therapies on resistant and wildtype GSC827 and assessed the effect on cell death and proliferation. Although the concentration of all inhibitors (except MET inhibitors) had to be decreased due to toxicity, we observed a similar pattern as seen with U87 and U373 cells. All single-agent treatments, although effective in wildtype GSC827 cells, were inadequate in MET inhibitor–resistant GSC827 cells whilst simultaneous inhibition of a bypass signaling pathway (EGFR, FGFR, mTOR, STAT3 or COX-2) and MET inhibitor demonstrated significant induction of cell death and a dramatic suppression of cell proliferation. Importantly, combinational treatment with celecoxib and crizotinib decreased tumor cell proliferation and increased cell death significantly in crizotinib-resistant GSC827 cells. These data suggest that similar pathways are implicated in MET inhibitor resistance in GSCs and GBM cells, indicating that these therapies may prove to be an extremely effective therapy for resistant GBM (Supplementary Fig. S6).

To examine whether these bypass signaling pathways can overcome MET inhibitor resistance in vivo, we assessed the effect of crizotinib alone and in combination with celecoxib and debio-1347 on tumor growth in immunodeficient mice bearing GBM xenografts. Celecoxib is an FDA-approved drug that demonstrates potent anti-cancer properties through the modulation of both the pro-survival BCL-2 (30) family and COX-2, and is currently in clinical trials for the treatment of numerous neoplasms. Debio-1347 is in clinical trials for the treatment of advanced solid tumors and metastatic breast cancer with FGFR alterations. We injected either wildtype or resistant U87 cells into the striata of immunodeficient mice, six days after implantation the animals were treated with either control (DMSO), either agent alone [Debio-1347 (25 mg/kg), celecoxib (10 mg/kg) or crizotinib (25 mg/kg)] or in combination [Debio-1347+crizotinib or celecoxib+crizotinib] by oral gavage daily for 7 days. Tumors were visualized by MRI and volumes were quantified. The data show that both therapeutic combinations inhibited tumor growth in crizotinib-resistant mice significantly more than either agent alone (Fig. 6A and B). The resistant cells displayed significant tumor volume reduction when subjected to combinational treatment indicating restored sensitivity to MET inhibitors.

Although the MET pathway is often dysregulated in GBM, MET inhibitors have not been particularly effective in treating patients with cancer due to acquired resistance. One mechanism of acquiring resistance against MET inhibitors is via activation of bypass pathways that compensate for the loss of survival signaling when MET is inhibited. Elucidating these bypass pathways offers the potential to develop combinatorial drug therapy to re-sensitize GBM cells to MET inhibitors. In this study, we developed GBM cell and animal models of resistance to MET inhibitors. Resistant cells revealed increased levels of active p.MET that could not be suppressed by the MET inhibitors, thus proving resistance and suggesting that MET receptor activation could contribute to this resistance. This finding differs from published mechanisms of resistance to EGFR inhibitors which involve loss of oncogenic mutant EGFRvIII (31). Using this model, we investigated the proteomic changes that occur when GBM cells become resistant to two clinically applicable MET inhibitors (crizotinib and onartuzumab), uncovered several important bypass pathways that include mTOR, FGFR1, EGFR, STAT3, and COX-2 and showed that targeting these pathways in combination with MET inhibitors, reverses resistance to the MET inhibitors.

The mTOR pathway is highly activated in GBM (32) and, although mTOR has been implicated in acquired resistance in small cell lung cancer (33), less is known about its role in GBM therapy resistance. MET inhibition leads to downregulation of PI3K signaling, which, in turn, leads to decreased activation of mTOR resulting in the induction of apoptosis and decreased cell proliferation (34). However, aberrant activation of PI3K/AKT signaling as a bypass mechanism results in increased mTOR activation that promotes cancer progression, metastasis, and invasion (35). We demonstrate that mTOR phosphorylation is significantly increased in MET inhibitor–resistant GBM cells, suggesting a role for the mTOR pathway in MET inhibitor resistance. Rapamycin, an FDA-approved inhibitor of mTOR, alone did not significantly enhance cell death but did have antiproliferative effects in MET inhibitor–resistant GBM cells. Combinational treatment of resistant GBM cells with rapamycin and either MET inhibitor induced apoptosis and further suppressed cell proliferation indicating restored MET inhibitor sensitivity.

Aberrant activation of RTKs such as FGFR1 and EGFR is a recognized mechanism by which malignant cells acquire resistance to other RTK monotherapies (36). FGFR1 and EGFR compensate for the loss of MET-mediated survival signaling through reactivation of downstream PI3K and STAT signaling (37). We demonstrate that FGFR1 is implicated in MET inhibitor resistance in GBM. FGFR1 is upregulated in MET inhibitor–resistant GBM cells and shows a trend towards correlation with GBM patient survival based on TCGA data analysis. Therefore, we assessed the effect of debio-1347, an FGFR1 inhibitor, in combination with crizotinib or onartuzumab on MET inhibitor–resistant GBM cells and we showed that FGFR1 inhibition can circumvent MET inhibitor resistance and simultaneous inhibition of both FGFR1 and MET is advantageous for reversal of MET inhibitor resistance (38).

EGFR and MET are frequently co-expressed in cancer and HGF can transactivate EGFR, which in turn, activates MET resulting in synergistic tumor growth (36, 37). Furthermore, MET has been reported to play a role in acquired resistance to EGFR-targeted therapies in many cancers (17, 39). Attributed to this and the fact that cross-talk exists between MET and EGFR, MET inhibitor–resistant GBM cells were subjected to combined treatment with the EGFR inhibitor, erlotinib, and either crizotinib or onartuzumab. Erlotinib alone did not significantly induce apoptotic cell death in MET inhibitor–resistant GBM cells; however, concomitant treatment with erlotinib and either crizotinib or onartuzumab restored MET inhibitor sensitivity leading to enhanced cell death and decreased cell proliferation.

Many growth factor receptors, including MET activate STAT3 (40, 32). STAT3 is elevated in GBM, driving tumor growth, angiogenesis and invasion (41) through the regulation of downstream targets including c-myc, Bcl-2, and Bcl-XL and is emerging as a drug resistance mechanism in GBM (42). On the basis of our screening data, we assessed the effect of STAT3 inhibition on MET inhibitor resistance and found that combinational treatment of MET inhibitor–resistant GBM cells with a STAT3 inhibitor and MET inhibitor induced cell death and significantly inhibited cell proliferation indicating restored MET inhibitor sensitivity.

We also found that MET inhibitor resistance resulted in upregulation of COX-2, an inducible cyclooxygenase that is overexpressed GBM (43). Elevation of COX-2 stimulates increased angiogenesis and invasion of tumor cells and correlates with poor prognosis (44) although little is known about COX-2 and MET inhibitor resistance in GBM. Celecoxib, a COX-2 inhibitor, was used in conjunction with either crizotinib or onartuzumab to assess the effect of COX-2 inhibition on MET inhibitor–resistant GBM cells. MET inhibitor–resistant GBM cells demonstrated significant toxicity to combination therapy. Although COX-2 inhibition is thought to be the primary mode of action, celecoxib has also been reported to act in a COX-2–independent manner (45). Further evaluation may be needed to completely elucidate the mechanism by which celecoxib reverts MET inhibitor resistance.

GSCs, a small subpopulation responsible for self-renewal, have been implicated in GBM relapse (46, 47). Inhibition of MET, which is expressed in GSCs, halts GBM progression and decreases the expression of stem markers such as CD133 and Sox2. GSCs display enhanced sensitivity to MET inhibitors, indicating a vital role for MET in GSC maintenance (48). Attributed to this, the development of anti-cancer therapies that target GSCs appears imperative for optimal GBM treatment (49). We show that treatment with MET inhibitor and simultaneous inhibition of one of the bypass proteins reverses MET resistance in GSCs. Interestingly, RPPA exposed considerable overlap in the bypass signaling involved in resistance to both MET inhibitors. These commonly upregulated pathways are vital for cell survival and include important molecules such as EGFR, BCL 2, COX-2, and FGFR1. The ability to target a commonly altered pathway, that would be effective at reversing the effects of drug resistance to multiple inhibitors, is the ultimate goal for personalized therapy (50).

Our data identify the mechanisms of resistance to MET inhibitors in GBM and suggest new combination therapies that overcome resistance. Device-based and PDX model-based screens are currently being assessed for optimal, patient-specific oncogenic driver identification. Informed by our findings, individual patients that display drug resistance could be assessed for their unique oncogenic driver signature and receive the most effective combination therapy.

J. Wulfkuhle holds ownership interest (including patents) in Theranostics Health, LLC, and is a consultant/advisory board member for Baylor College of Medicine. S. Phan holds ownership interest (including patents) in Genentech/Roche. E. Petricoin is an employee of and holds ownership interest (including patents) in Perthera, Inc., and is a consultant/advisory board member for Perthera, Inc. and ADVX Investors Group, LLC. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N. Cruickshanks, Y. Zhang, R. Abounader

Development of methodology: N. Cruickshanks, Y. Zhang, S. Hine, F. Guessous

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Cruickshanks, Y. Zhang, S. Hine, F. Yuan, C. Grello, M. Pahuski, C. Dube, B. Wang, C. Deveau, K. Saoud, J. Wulfkuhle, D. Schiff, S. Phan, E. Petricoin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Cruickshanks, S. Hine, M. Gibert, F. Yuan, M. Oxford, C. Dube, F. Guessous, C. Deveau, I. Gallagher, J. Wulfkuhle, E. Petricoin

Writing, review, and/or revision of the manuscript: N. Cruickshanks, M. Gibert, F. Yuan, M. Oxford, C. Dube, J. Wulfkuhle, D. Schiff, S. Phan, E. Petricoin, R. Abounader

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Zhang, S. Hine, M. Gibert, B. Wang

Study supervision: R. Abounader

Other (access to onartuzumab and scientific support related to onartuzumab): S. Phan

This work was supported by NIH grants RO1 NS045209 and UO1 CA220841 (to R. Abounader).

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.