Selective MET Kinase Inhibition in MET-dependent ... | ResearchGate

Author Manuscript Published OnlineFirst on July 27, 2017; DOI: 10.1158/1541-7786.MCR-17-0177 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Selective MET Kinase Inhibition in MET-dependent Glioma Models Alters Gene Expression and Induces Tumor Plasticity Corina N.A.M. van den Heuvel#1, Anna C. Navis#1 , Tessa de Bitter1, Houshang Amiri2, Kiek Verrijp1, Arend Heerschap2, Karen Rex3, Isabelle Dussault3, Sean Caenepeel3, Angela Coxon3, Paul N. Span4, Pieter Wesseling1, Wiljan Hendriks5 and William P.J. Leenders1 1

Radboud University Medical Centre, Department of Pathology, PO Box 9101, 6500 HB, Nijmegen,

The Netherlands 2

Radboud University Medical Centre, Department of Radiology, PO Box 9101, 6500 HB, Nijmegen,

The Netherlands 3

Amgen Inc. Thousand Oaks, Department of Oncology Research, CA, USA

4

Radboud

University

Medical

Centre,

Department

of

Radiation

Oncology,

Radiotherapy&Oncoimmunology laboratory, PO Box 9101, 6500 HB, Nijmegen, The Netherlands 5

Radboud Institute for Molecular Life Sciences, Department of Cell Biology, PO Box 9101, 6500 HB,

Nijmegen, The Netherlands

Running title: Rescue Pathways by MET inhibition using Compound A Keywords: tyrosine kinase receptor, MET, Compound A, precision medicine, rescue kinase, plasticity

#

CNAMvdH and ACN contributed equally to this study

Current addresses ACN: Karolinska Institute, Division of Molecular Neurobiology, Stockholm, Sweden; HA: VU Medical Centre, Department of Radiology, Amsterdam, The Netherlands and Neuroscience Research Center, Institute of Neuropharmacology, Kerman University of Medical Sciences, Kerman, Iran; PW: Department of Pathology, VU University Medical Center, Amsterdam, the Netherlands and Department of Pathology, Princess Máxima Center for Pediatric Oncology and University Medical Center Utrecht, Utrecht, The Netherlands 1

Downloaded from mcr.aacrjournals.org on July 31, 2017. © 2017 American Association for Cancer Research.


Financial support: StichtingStopHersentumoren.nl (grant number 2015-001) to CNAMH; Eurostars grant (Deltamet 9616) to CNAMH, RadboudUMC grant to ACN. Corresponding author: William P.J. Leenders, PhD, Radboudumc, Dept of Pathology, 846, PO Box 9101, 6500 HB Nijmegen, The Netherlands, Phone: +31-243614289, email: [email protected]

Conflict of interest: Compound A used in this study was obtained from Amgen Inc. KR, SC and AC are shareholders in Amgen Inc. The other authors have no conflicts of interest to disclose.

Word count: 5286 Number of Figures: 6 Supplementary files: 5

2



Abstract The receptor tyrosine kinase (RTK) MET represents a promising tumor target in a subset of glioblastomas. Most RTK inhibitors available in the clinic today, including those inhibiting MET, affect multiple targets simultaneously. Previously it was demonstrated that treatment with cabozantinib (MET/VEGFR2/RET inhibitor) prolonged survival of mice carrying orthotopic patient derived xenografts (PDX) of the MET-addicted glioblastoma model E98, yet did not prevent development of recurrent and cabozantinib-resistant tumors. To exclude VEGFR2 inhibition-inflicted blood brain barrier normalization and diminished tumor distribution of the drug, we have now investigated the effects of the novel MET-selective inhibitor Compound A in the orthotopic E98 xenograft model. In vitro, Compound A proved a highly potent inhibitor of proliferation of MET-addicted cell lines. In line with its target selectivity, Compound A did not restore the leaky blood brain barrier, and was more effective than cabozantinib in inhibiting MET phosphorylation in vivo. Compound A treatment significantly prolonged survival of mice carrying E98 tumor xenografts, but did not prevent eventual progression. Contrasting in vitro results, the Compound A treated xenografts displayed high levels of AKT phosphorylation despite the absence of phosporylated MET. Profiling by RNA sequencing (RNA-seq) showed that in vivo transcriptomes differed significantly from those in control xenografts. Implications: Collectively, these findings demonstrate the plasticity of paracrine growth factor receptor signaling in vivo and urge for prudency with in vitro drug testing strategies to validate monotherapies. Keywords: glioblastoma; MET; Compound A; cabozantinib; resistance; personalized therapy

3



Introduction Therapy for high-grade diffuse gliomas (glioblastomas) consists of surgery followed by radiochemotherapy, but unfortunately these treatments only prolong median survival of patients to 15 months after first diagnosis (1). A high percentage of glioblastomas carry genetic alterations in receptor tyrosine kinases (RTK) such as EGFR, PDGFRα and MET (2-4). Constitutive activation of these RTKs contributes to oncogenic signaling, identifying these proteins as therapeutic targets. Of special interest as a tumor target in diffuse glioma is MET and its ligand hepatocyte growth factor (HGF). Whereas normal MET function is essential for wound healing, embryogenesis and organogenesis (5-9), MET dysregulation is associated with tumorigenesis and in glioblastomas is linked to cellular growth and migration, shorter patient survival time and poor treatment response (3,5,10-14). Besides playing an important role in tumor development and progression, MET is involved in the stem-cell like phenotype in glioma and may provide a rescue pathway for tumor progression under treatment with drugs targeting EGFR, ALK (15-20), FGFR (21) and mutated BRAF (22-25) in different cancer types. In addition, MET has been described as a mediator of resistance to radiotherapy (16,26,27). Therefore, inhibition of MET may provide direct anti-tumor effects and prevent development of resistance to other therapies. Several small-molecule tyrosine kinase inhibitors (TKIs) have been developed that effectively inhibit RTK activity. However, due to the conserved tyrosine kinase catalytic domain, many of these drugs have activity against multiple RTKs. An example is cabozantinib, a TKI with activity against MET, VEGFR2, KIT, RET, TEK and FLT3 (28), which was recently tested for glioblastoma in a phase I/II clinical trial (29). We previously reported that cabozantinib inhibits phosphorylation of MET and its downstream signaling intermediates AKT and ERK1/2 (p44/p42) in E98 glioma cells, a patientderived astrocytoma cell line which harbors amplification of a mutated MET gene that is representative for some 8% of glioblastoma cases (30). Inhibition of MET signaling using cabozantinib in vitro resulted in absence of phosphorylated (P-)MET, diminished migratory capacity

4



and in cytotoxicity. Mice carrying orthotopic xenografts generated from E98 glioma cells experienced a survival benefit when treated with cabozantinib. Notwithstanding these positive effects, treatment with cabozantinib in the end did not prevent progression of diffuse infiltrative tumors and, opposing the in vitro data, treated tumors presented with prominent P-MET levels (11). This brought us to postulate that efficacy of cabozantinib on tumor cells in vivo may have been compromised by concomitant inhibition of VEGFR2 on the tumor-associated endothelial cells, leading to vascular normalization, restoration of the blood-brain barrier and subsequent inefficacy of cabozantinib as a tumor drug (31,32). In the current study we therefore tested two novel small-molecule MET tyrosine kinase inhibitors, Compound A (MET-selective) (33) and AMG 458 (multi-specific, inhibiting among others MET and RON (34)), in a number of MET-expressing cell lines and performed in vivo analysis of the effects of Compound A on mice carrying subcutaneous U87-MG and intracerebral E98 xenografts.

Materials and Methods Cell culture The PDX astrocytoma model E98 and the generation of the lentivirally transduced E98 cell line expressing firefly luciferase and mCHERRY (E98FM) have been described before (35,36). MKN45, HT-29, parental E98, E98FM, E98FMA (generated from Compound A-resistant E98FM xenografts), LN18, Hs746T and TOV-112D cells were grown in DMEM medium containing 4.5 g/l glucose and Lglutamin (Lonza, Basel, Switserland) supplemented with 10% fetal calf serum (FCS, Gibco, Waltham, MA, USA) and 40 µg/ml gentamicin (Centrafarm, Etten-Leur, The Netherlands) at 37°C in the presence of 5% CO2. Origin and relevant molecular characteristics of all cell lines are outlined in Supplementary Table S1. Human umbilical vein endothelial cells (HUVEC, isolated in-house from umbilical cords) were grown on 1% gelatin coated culture dishes in EndoPrime medium (PAA Laboratories/GE Healthcare, Little Chalfont, UK) supplemented with additives according to

5



manufacturer’s instructions. All cell lines, except HUVEC, expressed MET as validated by western blot analysis. Prior to intracerebral xenografting, E98FM cells were grown as spheroids for at least 2 passages in serum-free Neurobasal/B27 medium (Gibco/Invitrogen, Waltham, MA, USA), supplemented with EGF and basic FGF (both PromoCell, Heidelberg, Germany) as described previously (37,38). For passaging, spheres were dissociated with Accutase (Sigma-Aldrich, St. Louis, MI, USA). For in vitro analyses stock solutions of 10 mM of Compound A, AMG 458 (both Amgen, Thousand Oaks, CA, USA (33)) or cabozantinib (Exelixis, San Francisco, CA, USA (39)) were prepared in DMSO. Characteristics of AMG 458 and cabozantinib have been described before (28,34).

Immunoblotting Cell lines were seeded in 6-well plates and allowed to adhere. E98 cells at ~80% confluence were treated with Compound A (0 to 300 nM) or DMSO vehicle (0.03%) for 20 min. MKN45 and Hs746T cells were stimulated with HGF for 10 min, following 20 min treatment with Compound A or AMG 458 (0-3000 nM). After treatment, cells were washed twice with ice-cold PBS and lysed in 1x RIPA buffer (Cell Signaling Technology, CST, Danvers, MA) with 1 mM phenylmethylsulfonyl fluoride (PMSF), according to manufacturer’s instructions. Protein lysates were subjected to electrophoresis on 10% SDS-PAGE gels and electroblotted onto nitrocellulose membranes (Whatman Optitran BA-S85, GE Healthcare, Little Chalfont, UK). After blocking aspecific binding sites in blocking buffer (1:1 PBS/Odyssey blocking buffer [LI-COR Biosciences, Lincoln, NE, USA]) blots were incubated o/n at 4°C with primary antibodies (see Supplementary File 2 for detailed information). Primary antibodies were detected with appropriate IRDye680- or IRDye800-conjugated secondary antibodies (1:20,000, Invitrogen Molecular Probes, Waltham, MA, USA) and signals were visualized using the Odyssey imaging system (LI-COR Biosciences, Lincoln, NE, USA).

6



Proliferation assays Cells were seeded in 96-well plates (5,000, 10,000 and 20,000 cells/well for U87-MG/TOV112D/LN18, HUVEC/Hs746T/HT-29/MKN45 and E98 cells, respectively). The next day increasing concentrations of Compound A or vehicle only were added to the medium. Metabolic activity of cells was measured four days after start of Compound A treatment by incubation with 0.5 mg/ml 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) in PBS (Sigma-Aldrich, St. Louis, MO, USA). After 3.5 hr incubation at 37°C formazan crystals were dissolved in MTT solvent (0.1% NP40 and 3.4 mM HCl in isopropanol) and optical densities were measured at 560 nm. IC50 concentrations were determined using sigmoidal dose-response (variable slope) statistics in GraphPad Prism. Data represent the mean ± SEM (N=4).

Animal experiments All animal experimentation at AMGEN Inc. was conducted in an American Association for Laboratory Animal Care (AALAC)-accredited facility with Institutional Animal Care and Use Committee (IACUC) approved protocols. Animal experiments performed at Radboudumc were approved by the Radboudumc local ethical committee for animal experimentation. Based on the results of our previous experiments with cabozantinib, animal numbers were calculated by power analysis and kept as low as possible while retaining sufficient power. Pharmacodymamic assays are described in Supplementary File 3. U87-MG (5x106) cells were injected subcutaneously into the right flank of female CD1 nu/nu mice (Charles River Laboratories, MA, USA). Treatment was initiated when tumor volume reached approximately 400 mm3. Compound A (suspension in soybean oil) was administered either once or twice daily by oral gavage at the indicated doses. Vehicle group received 100% soybean oil by oral gavage once daily. Tumor volume was measured twice weekly as length (mm) x width (mm) x height (mm) and expressed as cubic mm using a PRO-MAX electronic digital caliper (Japan Micrometer Mfg. Co. Ltd.). Statistical analysis of observed differences between growth curves was performed by 7



repeated-measures analysis of variance followed by Dunnett’s test for multiple comparisons. A paired t-test was performed to determine regression using JMP software v7.0 interfaced with SAS v9.1 (SAS institute, Cary, NC, USA). For generation of intracerebral glioma xenografts, E98FM cells grown as spheroids for at least 2 passages, were injected orthotopically via standard procedures (35). In short, five to six week-old female BALB/c nu/nu mice (Janvier Labs, Le Genest-Saint-Isle, France) were anesthetized by inhalation of 3% isoflurane/O2. Per animal 200,000 cells in 10 μl PBS were injected. Mice were checked weekly for tumor development using bioluminescent imaging (BLI) on an IVIS imager (procedure is described below). When tumor growth was observed (typically 3-5 weeks after tumor cell injection) treatment was initiated by daily oral gavage of 60 mg/kg Compound A in 100 µl soy bean oil (Spectrum, New Brunswick, NJ, USA) (N=12) or soy bean oil only (N=10). Two Compound A-treated and two control animals were sacrificed after 10 days of treatment to analyze the short-term effects of therapy. Fifteen minutes before sacrifice these animals were injected intraperitoneally with BrdU (Sigma-Aldrich, St. Louis, MO, USA; 1.25 mg/animal). The other animals were sacrificed when signs of tumor-related symptoms started to develop (usually >15% weight loss within two days). Prior to sacrifice, vehicle (N=3) and Compound A-treated animals (N=4) were subjected to Gd-DTPA-MRI (see Supplementary File 4). Brain, liver and kidneys from each animal were partially snap frozen in liquid nitrogen and partially processed to formalin-fixed paraffinembedded (FFPE) tissue blocks. Statistical significance in survival was determined in Graphpad Prism v5.03 (GraphPad Software, LaJolla, CA) using the Log-rank Mantel-Cox test. To generate a cell line from Compound A-resistant xenografts, tumors were macroscopically dissected from brains of Compound A-treated animals, cut in small pieces in PBS containing 4.5 g/l glucose and enzymatically dissociated in PDD buffer (HBSS containing 12.4 mM MgSO4, 0.01% papain (SigmaAldrich, St. Louis, MO, USA), 0.1% Dispase 2 and 0.01% DNase [both Roche Diagnostics, Basel, Switzerland]). After 1 hr at 37°C tissue was processed to a single-cell suspension by pipetting and filtering over a 70 μm mesh nylon filter. After washing with PBS, cells were suspended in serum-

8



containing DMEM and cultured under standard conditions. MTT assays were performed after 2 and 3 passages as described above.

In vivo bioluminescence Tumor burden in mice carrying orthotopic E98FM xenografts was measured via luciferase bioluminescence imaging at least once a week and prior to sacrifice. Mice were imaged under 3% isoflurane/O2 anesthesia in an IVIS Lumina system (Caliper Life Sciences, Hopkington, MA, USA) 10 min after intraperitoneal injection of 200 μl PBS containing 15 mg/ml D-luciferin (Caliper Life Sciences, Hopkington, MA, USA). Images were analyzed using the Living Image 3.0 software (Caliper Life Sciences, Hopkington, MA, USA). Two-dimensional regions of interest (ROIs) were drawn around the skull. Luciferase activity was expressed as photons emitted per second per square centimeter.

Immunohistochemical analysis Immunohistochemistry (IHC) on 4 μm FFPE tissue sections was performed as previously described 38. In short, after epitope retrieval (10 min boiling in citrate buffer pH 6.0 or EDTA) and permeabilization in 1% Triton-X-100 (10 min, for P-histone H3 stainings only) tissue sections were incubated o/n at 4°C with primary antibodies against MET, P-MET Tyr1234/Tyr1235, P-AKT S473, Ki-67, BrdU, P-histone H3 Ser10, and caspase-3A. Appropriate biotinylated secondary antibodies were used for detection using the AvidinBiotin Complex (ABC) method (Vector Laboratories, Burlingame, CA, USA). Mouse IgG in tissue sections was visualized via direct incubation with biotinylated horse-anti-mouse-IgG, followed by ABC-visualization. See Supplementary File 2 for detailed primary antibody information. Antibody complexes were visualized by a 7 min incubation with 3,3'-

9



diaminobenzidine solution (Power-DAB, ImmunoLogic, Duiven, The Netherlands). All sections were counterstained with haematoxylin and mounted in Quick-D mounting medium (Klinipath BV, Duiven, The Netherlands). Additionally, for some stainings IHC was performed on 4 µm frozen sections that were airdried and fixed with 4% paraformaldehyde for 10 min. After washing with distilled water and PBS, sections were incubated for 1h at RT with antibodies against MET and P-MET Tyr1234/Tyr1235 in PBS/1%BSA (see Supplementary File 2 for detailed information). The ABC method and Power-DAB were used for detection and visualization as described above. All sections were counterstained with haematoxylin and mounted in Quick-D mounting medium. Immunohistochemical stainings were interpreted by two observers, blinded to the origin of the material. Diffuse and more compact areas of tumor were scored separately as negative (-), heterogeneous/weakly positive (+/-) or positive (+). Statistical significance was determined in Graphpad Prism v5.03 (GraphPad Software, LaJolla, CA) using a Kruskal Wallis post-hoc Dunn's multiple comparison test (placebo N=6, Compound A and cabozantinib N=5).

RNA next generation sequencing Total RNA was isolated from 5x 20 µm sections of snap-frozen control, Compound A- or cabozantinib-treated E98FM xenograft tissues using TRIzol reagent (Life Technologies, ThermoFisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions. After validation of RNA quality by a RIN score >8 in Bioanalyzer assays (Agilent Technologies, Amstelveen, The Netherlands) and depletion of ribosomal RNA, whole RNA next generation sequencing was performed by ServiceXS (Leiden, The Netherlands). Each sample yielded 30-50 million reads (pairedend sequencing protocol) and resulting dataset was analyzed using the ‘Tuxedo’ protocol; reads were

10



mapped against the RefSeq human genome (hg19) with TopHat and final transcript assembly was done with the Cufflinks package (40). Normalization was done with Cuffquant. Relative values for each transcript were expressed as transcript per million mapped reads (TPM). Human reads were separated from mouse reads by using the Xenome algorithm (41). Statistical analysis (N=1) was performed using RStudio 1.0.136.

Quantitative real-time PCR RNA was isolated from 5x 20 µm sections of snap-frozen control, Compound A- or cabozantinibtreated E98FM xenograft tissues using TRIzol reagent (Life Technologies, ThermoFisher Scientific, Waltham, MA, USA) and reverse-transcribed into cDNA using Superscript II (Invitrogen, Waltham, MA, USA) according to the manufacturer’s instructions. Primer sequences for qPCR were: β2M-Fw 5’-ATGAGTATGCCTGCCGTGTG-3’, β2M-Rv 5’-CCAAATGCGGCATCTTCAAAC-3’, IFI6-Fw 5’-CTTGTGCTACCTGCTGCT-3’, IFI6-Rv 5’-TTCTTACCTGCCTCCACC-3’ (Biolegio, Nijmegen, The Netherlands). Power SYBR Green PCR Master Mix (Applied Biosystems) was used for qPCR on a 7500 Fast real-time PCR system (Applied Biosystems) and data was normalized to the human housekeeping gene β2M. Experiments involved Compound A treated xenografts (N=4) and cabozantinib treated xenografts (N=1), and for each sample 4 technical replicates were performed.

Results Compound A and AMG 458 inhibit MET signaling in MET amplification-positive cell lines in vitro In order to test whether MET inhibitors that do not act upon VEGFR2, and thus leave the blood brain barrier leaky to support drug delivery to brain tumor regions, provide added value for glioblastoma treatment, we tested the MET inhibitors Compound A and AMG 458. Compound A (Supplementary Figure S1A) is a tyrosine kinase inhibitor with high selectivity for MET (IC50 in MET phosphorylation assays of 5-7nM) (33), while AMG 458 (Supplementary Figure S1B) 11



inhibits multiple RTKs including MET and RON (IC50 in MET phosphorylation assays of 60 nM) (34). We tested the effects of both compounds in MKN45 gastric cancer cells, carrying a MET amplification, and Hs746T gastric cancer cells, carrying both a MET amplification and activating METΔ14 mutation (42). Compound A dose-dependently inhibited MET phosphorylation and activation of the downstream effector molecules AKT and ERK1/2 (p44/p42) in both cell lines (Figure 1A), while not affecting total MET levels (43). Parallel tests with AMG 458 showed that Compound A was 10-fold more potent in inhibiting phosphorylation of MET and its downstream targets (Figure 1A), and this translated in increased cell killing activity of Compound A as compared to AMG 458 in MKN45 cells (IC50 values of 20.0 nM and 445.0 nM for Compound A and AMG 458, respectively, Figure 1B) identifying Compound A as the superior compound with respect to MET inhibition. We therefore continued our study using Compound A in the E98 glioma model, which carries an amplification of a mutated MET gene encoding the auto-active METΔ7-8 oncoprotein (30). Compound A dose-dependently inhibited MET phosphorylation and activation of downstream effector molecules AKT and ERK1/2 (p44/p42) (Figure 1A) and potently inhibited E98 cell proliferation with an IC50 value of 9.4 nM (Figure 1B). IC50 values of Compound A on cell lines with low to moderate levels of MET ranged from 947.5 nM to 6570.0 nM (Figure 1C), providing a therapeutic window for METspecific tumoricidal effects.

Compound A inhibits MET signaling and improves survival of MET-dependent xenografts in vivo In preparation for in vivo studies, we first investigated the pharmacodynamic (PD) properties of Compound A as described in Supplementary File 3. Compound A dose-dependently inhibited HGFmediated MET phosphorylation in liver with an approximate ED90 of 10 mg/kg and an associated plasma exposure of approximately 800 ng/ml (Supplementary File 3, Figure S2A, p