Identification of MEK162 as a Radiosensitizer for the Treatment of Glioblastoma

Glioblastoma (GBM) is the most prevalent and deadly primary malignant brain tumor. Current treatment consists of surgery followed by radiotherapy and temozolomide, which results in a median survival rate of 15 months (1, 2). Among the three treatment modalities, radiotherapy has been proven to be the most important component in GBM treatment (3). A dose of 60 Gy fractionated radiation increased overall survival from 4.2 with surgery alone to 12.1 months (4, 5). Sensitizing cancer cells to DNA-damaging agents such as radiation can be achieved by directly targeting key proteins of the DNA damage response (DDR) such as double strand break (DSB) “sensor” proteins ATM and ATR or by targeting key repair proteins such as DNA-PKcs. Alternatively, cells can be sensitized indirectly by interfering with cell-cycle checkpoints that exist to check DNA integrity before cell division (G₂–M checkpoint) or DNA replication (G₁–S checkpoint). If these checkpoints are abrogated, cells undergo mitotic catastrophe under stress of DNA-damaging agents including radiotherapy. Recently, through molecular subtyping, GBM tumors have been subdivided into the proneural, neural, classical, and mesenchymal subtypes defined by distinct somatic mutation and gene expression profiles (6, 7). These genetic profiles show that >90% of all glioblastomas have activating mutations in the PI3K/MAPK pathways rendering components of these pathways as potential targets for therapeutic intervention. Inhibitors targeting key proteins in these pathways have been extensively studied preclinically and in clinical trials for many types of cancers including brain cancers (8, 9). The PI3K/MAPK pathways are canonically under control of various receptor tyrosine kinases (RTK) on the membrane that are activated upon ligand binding and act as main drivers of growth, proliferation, motility, and survival. Inhibition of these pathways can reduce cancer growth and induce apoptosis. Most importantly, targeting these pathways can sensitize cancer cells to radiation and other DNA-damaging agents in various cancers, including glioma (10–12). However, to date none of these strategies have shown significant therapeutic improvements over current standard of care for GBM patients in the clinic (13).

In this study, we investigated the single-agent efficacy and radiosensitizing potential of novel targeted agents in the PI3K/MAPK pathways in glioma. Out of this panel, MEK162, an allosteric inhibitor of MEK1/2, was found to enhance the effect of irradiation on glioma cells in vitro as well as in vivo. By linking in vitro measured responses to RNA expression data, we identified a potential patient subgroup that is expected to benefit from this combination therapy. Herewith, we demonstrate for the first time MEK inhibition combined with radiotherapy as a novel therapeutic avenue for GBM patients and provide a strong basis for further clinical evaluation.

The established cell lines U251, T87G, and U87 were acquired from the ATCC. Primary GBM8 cells were kindly provided by Dr. Bakhos Tannous (Harvard/MGH, Boston, MA). The primary glioma sphere cultures (GSC) were kindly provided Dr. Roel Verhaak (MD Anderson Cancer Center, Houston, TX). Cells were certified mycoplasma free by regular testing (www.microbiome.nl). Established cell lines were maintained as monolayer in DMEM supplemented with 10% FCS and 100 U/L penicillin/streptomycin. Primary cultures (GBM8 and GCSs) were cultured in Neurobasal-A-Medium (NBM) supplemented with N2, B27 without vitamin A, Glutamax, human EGF, human FGF basic, heparin, and penicillin/streptomycin. MK2206, BKM120, RAD001, BEZ235, MLN0128, MEK162, and Torin2 were purchased from Selleckchem (Selleckchem.com; for chemical structures, see Supplementary Fig. S1). Cells and spheroids were irradiated at room temperature using a Cobalt-60 source at a dose rate of 516 Gy/h (Gammacell 220; Atomic Energy of Canada). 3D spheroid cultures were formed by plating cells at 3,000 cells/well in ultra-low attaching repellent plates (Corning #3471). Four days were allowed for spheroid formation. Images were automatically captured on a Leica DMI3000 microscope (Leica) using Universal Grab 6.3 software (DCILabs). Spheroid sizes were determined using Scratch Assay 6.2 (DCILabs). Spheroid volumes were quantified as a function of time which allow analyses in (i) size reduction, (ii) growth rate, and (iii) long-term regrowth time [time to reach 5 times the initial volume (V5)].

Protein expression levels were assessed by Western blot analysis. Proteins were stained using Cell Signaling Technology antibodies (Supplementary Table S1). For combination experiments, cells were treated with 1 μmol/L MEK162 one hour before irradiation (4 Gy), and samples were collected at indicated time points for analysis.

Female athymic nude-Fox1nu mice kept in cages up to 6 mice (age 8–10 weeks; Harlan) were maintained in accordance with animal welfare guidelines and regulations of the VU University. GBM8 primary sphere cultures, stably expressing firefly luciferase and mCherry, were intracranially injected in a volume of 5 μL (0.5 × 10⁶ GBM8-FM cells). Cells were injected stereotactically into the striatum. Coordinates (see the mouse brain atlas; ref. 14): 2.0 mm lateral, 2.5 mm ventral of lambda, and at a 2.0 mm depth. Before start of treatment (11 days after injection of the tumor cells) tumor engraftment was determined by measuring Firefly luciferase (Fluc) activity. Mice were injected intraperitoneally with 100 mg/kg d-luciferin (Gold Biotechnology) in the morning and imaged with a CCD-camera (IVIS). Mice without tumor engraftment (Fluc activity < 10⁴ RLU) were excluded from the experiment. Mice were subsequently stratified into 4 treatment groups of six animals each by distributing the maximal dynamic range over all groups: vehicle (5% Kolliphor P188, 1% DMSO in PBS), MEK162 (50 mg/kg), vehicle/irradiation (3 daily fractions of 2 Gy each), and MEK162/irradiation. MEK162 was administered by oral lavage 1 hour before irradiation. Mice were anesthetized using ketamine and xylazine less than 5 minutes prior to treatment. Each group of 6 mice received cranial anterior–posterior 6 MV photon irradiation on their heads only. The bodies of the mice were shielded from the irradiation using lead. Dose was delivered with a Clinac D/E (Varian Medical Systems) at a dose rate of 360 Gy/hours. in the isocenter. Experiments were covered by the ethical review permission DEC1403 and are reported according to the ARRIVE animal research reporting guidelines.

For cell-cycle analysis, BrdUrd (Invitrogen, #000103) 1:100 was added to the cells 2 hours prior to fixation with ethanol. Cells were stained with propidium iodide (20 μg/mL) and Rabbit α-BrdU (Rockland Immunochemicals, #600-401-c29). For apoptosis analysis, cells were harvested and stained with Annexin V-FITC (Immunotools, #31490013SP; 1:25) antibody, and propidium iodide (2 μg/mL).

Synergy was determined by calculating the combination index (CI) using the formula of Chou and Talalay (refs. 15, 16; equation 1). In the formula V₁ = Viability (% of control) of cell line after treatment with therapy 1 (i.e., irradiation); V₂ = viability (% of control) of cell line after treatment with therapy 2 (i.e., drug treatment); V_1,2 Viability (% of control) of the combined effect after treatment with therapy 1 and 2. The radiation-induced growth delay in days was calculated by (Drug^{4 Gy} – Drug^{0 Gy}) and was assessed when the average spheroid size reached 5 times the initial volume (V5) in vitro or when median BLI signal reached 10⁸in vivo.

Bioinformatic analyses were performed using the bioinformatics tool R2 (for a complete description see http://r2.amc.nl). The dataset used in this study was the Tumor Glioblastoma - TCGA - 540 - MAS5.0 - u133a. The MEK162/radiotherapy signature was generated by selecting the top 20 positive and top 20 negative correlation genes to the synergistic response, using the <correlate with a track> option with a false discovery rate of less than 0.05. Stringency parameters: <Minimal presentcalls>: 1, <Minimal highest expression>: 100 relative units. Selected genes are given in Supplementary Table S2.

We previously found that cell culture dimensionality profoundly influences the drug/radiotherapy interaction in glioma cells when pharmacologically inhibiting AKT (17) in agreement with earlier studies (18–20). Therefore, we chose to study the effect of a series of inhibitors of the PI3K and MAPK pathways in U87 glioma cells when cultured as 3D multicellular spheroids. As reported previously, these inhibitors did not radiosensitize glioma cells when grown as a monolayer (17). We used similar culture techniques to maintain 3D structural integrity for a longer period of time (>40 days; see Materials and Methods). Cells were exposed to drugs for 4 days with or without 4 Gy given one hour after administration of drugs (a schematic overview is given in Fig. 1A).

Size reduction analysis showed that all drugs combined with 4 Gy display synergy (CI ≤ 0.8) up till day 15, with MEK162 showing the strongest synergy (CI = 0.43, Supplementary Fig. S2A–S2C; Fig. 1B). Growth rate analysis showed a significant reduction for the MEK162 + radiotherapy combination (Fig. 1C), showing MEK162 to significantly enhance the effect of irradiation (CI = 0.82; Fig. 1D). Growth delay analysis showed that the irradiation effect was only surpassed when combined with by MEK162, leading to a 10-day lag phase (Fig. 1E).

Given this long-term effect onto sphere regrowth, we analyzed DNA damage parameters using Western blot analysis. Irradiation-induced damage, as indicated by the DNA double-strand break (DSB) marker γH2AX, returned to baseline levels after 24 hours. Addition of MEK162 sustained high levels of γH2AX, reflecting attenuated DNA damage repair (Fig. 1F). Interestingly, treatment of MEK162 alone also increased γH2AX levels (although not to the same extent as MEK162 + radiotherapy) suggesting that MEK1/2 by itself controls DNA integrity in glioma spheroids.

We then evaluated the dose dependency of both modalities as assessed by growth delay. U87 spheroids were treated with increasing concentrations of MEK162 without/with 5 × 2 Gy (Fig. 2A and B, respectively) showing that MEK162 enhanced irradiation induced growth delay in a dose-dependent way (Fig. 2C). The radiation dose response was also evaluated by increasing the number of fractions which also showed a dose-dependent effect in spheroid regrowth in the presence of MEK162, culminating in complete ablation of regrowth after 3 weeks of treatment (Fig. 2D and E). Importantly, neither increasing the number of radiation (2 Gy each) fractions in the absence of MEK162 (Fig. 2C) nor prolonged treatment with MEK162 alone (Fig. 2D) significantly impacted spheroid regrowth. No such effect was observed when an mTOR inhibitor (MLN0128) was used (Supplementary Fig. S3, experimental setup shown in Supplementary Fig. S4).

We used primary glioma sphere cultures (GSC) to confirm these data. GSCs are considered more representative of human tumors because they retain a level of intratumoral heterogeneity and tumor-specific genetic and epigenetic signatures in vitro (21). GBM8 is a primary model extensively described in vitro and in vivo as a clinically relevant glioma model (22, 23). In vitro, MEK162 showed statistically significant radiosensitizing effects in GBM8 sphere size (CI = 0.22 at day 14), growth delay (from 6 to 15 days) and sustained DNA damage (Supplementary Fig. S5). We assessed whether radiosensitization by MEK162 was due to indirect (cell cycle) or direct effects (DDR sensors) by cell cycle and Western blot analysis. MEK162 treatment showed a decrease of the S-phase and G₂–M phase over time and a significant increase in sub-G₁ (apoptotic) after 72 hours (Fig. 3A). After a 2-hour bromodeoxyuridine (BrdUrd) pulse before MEK162 treatment, we noticed a 6-fold increase of BrdUrd-positive cells in sub-G₁ while the overall fraction of BrdUrd⁺ cells only decreased minimally (Supplementary Fig. S6). These data indicate that MEK inhibition alone reduces proliferation while possibly abrogating the G₂–M checkpoint and thereby pushing cells into apoptosis after 72-hour exposure. We therefore checked expression and phosphorylation of key proteins by Western blot analysis, which showed that both CDK1 and WEE1 were reduced after 24 hours of MEK162 treatment (Fig. 3B). Furthermore, we found that the phosphorylation of ATM and its downstream effector CHK2, both proteins of the DDR, were reduced. This was accompanied by an increase in γH2AX levels after 24 hours of MEK162 treatment. Combining MEK162 with 4 Gy resulted in decreased ATM phosphorylation and retention of DNA damage by γH2AX (Fig. 3C, see red boxes). Furthermore, the combination resulted in a synergistic increase of the sub-G₁ fraction after 24 hours (Fig. 3D) as well as a synergistic increase in late apoptotic cells after 72 hours (Fig. 3E).

We tested the effect of combined MEK162 treatment and irradiation in an orthotopic in vivo model using GBM8 cells (Fig. 4A). Assessing the radiation-induced growth delay by MEK162 showed an increase from 2 to 9 days (Fig. 4C). Differences in median tumor growth rate (Fig. 4D) resulted in a synergistic growth reduction for the combination treatment (CI = 0.52, Fig. 4F). A significantly longer median survival was observed in the presence of MEK162 compared with vehicle control (P = 0.016), whereas median survival for each monotherapy was not different from vehicle-treated animals (Fig. 4G). Overall survival was increased for single agent and combination arms but these differences were not statistically significant (Fig. 4E). Treatments were well tolerated for all treatment groups, as determined via scoring of the animals' conditions and weight during the 6-week follow-up period. Hence, the endpoint of the study, that is, severe deterioration as a result of tumor size, which was verified via a bioluminescence signal, was not limited by toxicity.

Raw data for each mouse are provided in Supplementary Fig. S7. The presence of MEK162 in combination with irradiation showed marked synergy onto tumor size as measured by bioluminescence signal (Supplementary Fig. S7B).

Identifying patients that respond to the combination therapy would be of great value. We used a panel of 9 primary GSCs (24) to generate an RNA response signature. For this, GSCs were treated with a fixed MEK162 (1 μmol/L) and irradiation (3 × 2 Gy) dose and assessed for spheroid size reduction at day 14 (Fig. 5A and B). We subsequently correlated the CI at day 14 with the top 20 positive- and negative-correlating genes which resulted in a therapy response RNA signature (Fig. 5C). This gene signature was used to stratify 540 patients from the TCGA glioma dataset based on their z-score and define them as expected responders and nonreponders to MEK162 + radiotherapy. This resulted in a clear distinction and defined a group that might have a benefit from the combination therapy (n = 43) and a group with no expected benefit (n = 461). Patients of the expected benefit group were found to have a low overall survival with current standard of care (n = 46, P = 0.013; Fig. 5D). In line with this finding, these patients are categorized as mesenchymal [i.e., they matched the Verhaak mesenchymal gene signature (7), R = 0.575, P < 0.0001] that inversely correlated with the proneural gene signature (R = −0.545, P < 0.0001; Fig. 5E).

We report that MEK1/2 inhibition synergizes with radiation in glioblastoma sphere cultures in vitro and orthotopically in vivo. This interaction appears specific for the MAPK axis as it was not observed for any of the tested inhibitors in the PI3K/mTOR pathway. Previous research in glioma and other cancers have identified similar radiosensitizing interactions in monolayer cell lines addicted to the MAPK pathway by activating mutations in either KRAS or BRAF (25–27). Here we report that even in the absence of these activating mutations; MEK162 is able to enhance the irradiation response of glioma cells, but only when cultured as spheroids, underlining the importance of MEK signaling in these more physiologically relevant culture conditions. Multicellular spheroids better mimic the 3D cell-to-cell contacts and the pathophysiologic gradients of nutrients and oxygen of tumors in situ (28). This can lead to altered intracellular signaling and thereby different responses to pharmaceutical agents (29), irradiation (30), and resultant synergistic interactions compared with monolayer cultures. Recently, a phospho-proteomic study of multiple established glioma cell lines cultured as spheroids reported increased signaling through MEK in spheroids compared with cells grown as monolayers (31). Moreover, this importance of MEK signaling was also found through phospho-proteomic network modeling across eight primary GBM cultures (including GBM8; ref. 32). Both studies demonstrate that under more physiologic culture conditions MEK is highly active and cells rely on MEK for their DNA damage response.

We evaluated MEK162 mechanistically in GBM8 spheroids and found it to decrease levels of key proteins of both G₂–M and G₁–S cell cycle checkpoints as well as DDR proteins. Cell-cycle analyses showed the cells exiting the cell cycle during or after S-phase and enter sub-G₁. This combinatorial disruption of cell-cycle checkpoints and DDR likely underlies increases in γH2AX after MEK inhibition alone. The interaction between MEK1/2 and the G₂ checkpoint has been described in other studies and indicates that synergy in glioblastoma occurs through mitotic catastrophe (27, 33, 34).

Our In vivo data confirmed that MEK inhibition enhanced the effect of fractionated radiotherapy on the endpoints tumor growth rate, tumor regrowth time, and tumor size. This led to a significant increased median survival for the combination therapy compared with untreated mice. This is, to our knowledge, the first evidence of radiosensitization of an orthotopic xenograft model for glioblastoma by inhibition of MEK1/2.

Currently, MEK inhibitors are widely investigated in the clinic in particular for RAS- or BRAF-mutated tumors. There are two trials recruiting patients for a phase I/II study for children with low-grade gliomas with activating mutations BRAF or loss of NF1 (NCT02285439 with MEK162/NCT01089101 with selumetinib). So far, only trametinib has gained FDA approval (35), whereas MEK162 (binimetinib) is in phase III (36, 37). Common adverse effects of MEK inhibitors include rash, fatigue, and diarrhea which are manageable. More serious but rarer side effects include central serous retinopathy and, when present in the CNS, MEK inhibitors can cause retinal vein occlusion and neuropathy (9). Trametinib was however shown to not enter intact brain in suggesting such toxicities can be avoided but thereby possibly making it unsuitable for brain tumors (38). Ascierto and colleagues (39) showed that MEK162 was effective against NRAS melanoma brain metastases without inducing CNS-related side-effects. It has been shown that fractionated radiation can effectively disrupt the blood–brain barrier and increase drug accumulation in an orthotopic in vivo brain tumor model (40). Exploiting this approach might improve drug accumulation at the tumor site and should therefore be further evaluated for radiosensitizing compounds (41).

Currently most trials for MEK inhibitors as monotherapy include patients with activating mutations in the MAPK pathway. In this study, we have stratified patients for possible therapy response based on a set of nine primary GSCs. This limited panel of cell lines yielded a gene signature that positively correlated with the mesenchymal subtype (which has a low overall survival), indicating these patients most likely benefit from this therapy. The mesenchymal subtype is frequently associated with a loss of NF1 leading to high MAPK activity that could guide patient selection as well (42). Further expansion of the GSC dataset by more clinical samples might allow for accurate identification of patients and thereby enable an individualized approach. On the basis of these data first steps, a multicenter phase I/II study for MEK inhibition in combination with radiotherapy and temozolomide has been initiated.

No potential conflicts of interest were disclosed.

Conception and design: B.J. Slotman, D.A. Haas-Kogan, L.J.A. Stalpers, B.A. Westerman, J. Theys, P. Sminia

Development of methodology: R.S. Narayan, T. Lagerweij, L.J.A. Stalpers, J. Theys, P. Sminia

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.S. Narayan, A. Gasol, P.L.G. Slangen, F.M.G. Cornelissen, T. Lagerweij, T. Wurdinger, J. Theys, P. Sminia

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.S. Narayan, A. Gasol, P.L.G. Slangen, F.M.G. Cornelissen, T. Lagerweij, B.J. Slotman, D.A. Haas-Kogan, B.A. Westerman, J. Theys, P. Sminia

Writing, review, and/or revision of the manuscript: R.S. Narayan, B.J. Slotman, D.A. Haas-Kogan, L.J.A. Stalpers, B.G Baumert, B.A. Westerman, J. Theys, P. Sminia

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.S. Narayan, P.L.G. Slangen, F.M.G. Cornelissen, J. van den Berg, B.J. Slotman, P. Sminia

Study supervision: B.J. Slotman, L.J.A. Stalpers, B.G Baumert, B.A. Westerman, J. Theys, P. Sminia

Other (biotechnician): H.Y.Y.E. Veldman

This project was funded by the Dutch Cancer Foundation (KWF), grant number VU2010-4874 (to P. Sminia, L.J.A. Stalpers, and B.G. Baumert) and grant number NINDS 5R01NS091620-02 (to D.A. Haas-Kogan). A. Gasol was supported in part by the foundation STOPHersentumoren.nl, grant number 2002657 (to P. Sminia and A.G. Denkova).

We would like to thank Dr. Roel Verhaak (MD Anderson) for providing us the Glioma Sphere Cultures and Dr. Bakkhos Tannous (Harvard/MGH) for providing us with GBM8. We thank Naomi Petersen (VUmc) and Kim Paesmans (MAASTRO) for assistance with the in vivo work. We thank Dr. Phil Koken en Dr. Stan Heukelom for their help with the irradiation of the mice.

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.