1521-0111/88/6/962–969$25.00 http://dx.doi.org/10.1124/mol.115.097774MOLECULAR PHARMACOLOGY Mol Pharmacol 88:962–969, December 2015Copyright ª 2015 by The American Society for Pharmacology and Experimental Therapeutics

Identification of an EGFRvIII-JNK2-HGF/c-Met–Signaling AxisRequired for Intercellular Crosstalk and Glioblastoma MultiformeCell Invasion s

Vanessa C. Saunders,1 Marie Lafitte,1 Isabel Adrados, Victor Quereda, Daniel Feurstein,YuanYuan Ling, Mohammad Fallahi, Laura H. Rosenberg, and Derek R. DuckettDepartment of Molecular Therapeutics (V.C.S., M.L., I.A., V.Q., D.F., Y.L., D.R.D.), Informatics Core, Scripps Research Institute,Jupiter, Florida (M.F.) and Cancer Research Technology Discovery Laboratories, Jonas Webb Building, Babraham ResearchCampus, Cambridge, United Kingdom (L.H.R.)

Received January 7, 2015; accepted September 30, 2015

ABSTRACTGlioblastoma multiforme (GBM) is the most aggressive andcommon form of adult brain cancer. Current therapeuticstrategies include surgical resection, followed by radiotherapyand chemotherapy. Despite such aggressive multimodal ther-apy, prognosis remains poor, with a median patient survival of14 months. A proper understanding of the molecular driversresponsible for GBM progression are therefore necessary toinstruct the development of novel targeted agents and enablethe design of effective treatment strategies. Activation of thec-Jun N-terminal kinase isoform 2 (JNK2) is reported in primarybrain cancers, where it associates with the histologic grade andamplification of the epidermal growth factor receptor (EGFR). Inthis manuscript, we demonstrate an important role for JNK2 inthe tumor promoting an invasive capacity of EGFR variant III,

a constitutively active mutant form of the receptor commonlyfound in GBM. Expression of EGFR variant III induces trans-activation of JNK2 in GBM cells, which is required for atumorigenic phenotype in vivo. Furthermore, JNK2 expressionand activity is required to promote increased cellular invasionthrough stimulation of a hepatocyte growth factor–c-Metsignaling circuit, whereby secretion of this extracellular ligandactivates the receptor tyrosine kinase in both a cell autono-mous and nonautonomous manner. Collectively, these findingsdemonstrate the cooperative and parallel activation of multipleRTKs in GBM and suggest that the development of selectiveJNK2 inhibitors could be therapeutically beneficial either assingle agents or in combination with inhibitors of EGFR and/orc-Met.

IntroductionGlioblastoma is the most malignant central nervous system

cancer and accounts for the majority of primary brain cancer–related deaths (Porter et al., 2010). Despite advances inmultimodality therapies, such as surgery, radiotherapy, andchemotherapy, the outcome for patients remains extremelypoor, with an average postdiagnostic survival of just over14 months (Behin et al., 2003; Stupp et al., 2005; Louis et al.,2007). The high mortality rate results from the universalresurgence of tumors post-treatment, which occurs due toinfiltrating tumor cells that escape initial surgery and exhibitprofound resistance to irradiation and current chemotherapytreatments (Claes et al., 2007). Thus, identification of novel

tractable targets for improved therapeutics is desperatelyneeded.Genomic and proteomic analyses have identified a number

of key oncogenic drivers of glioblastoma multiforme (GBM)tumorigenesis and therapeutic resistance, including receptortyrosine kinases (RTKs) (Beroukhim et al., 2007; Huang et al.,2007a,b). In particular, amplification of the epidermal growthfactor receptor (EGFR) is present in approximately half of allGBMs (Hurtt et al., 1992; Jaros et al., 1992), with a largeproportion also expressing activating mutations, such asdeletion of exons 2–7, which results in a ligand-independent,constitutively active mutant commonly referred to as EGFRvariant III (EGFRvIII) (Nishikawa et al., 1994; Batra et al.,1995; Huang et al., 1997). While it is unclear how EGFRvIIImutations are generated, it appears to occur late during tumorprogression and its coexpression with the wild-type receptorhave been reported to confer poor prognosis and a shorterpatient survival time (Shinojima et al., 2003; Heimbergeret al., 2005). This phenomenon is also observed in orthotopicxenograft mouse studies, whereby expression of EGFRvIII inhumanGBMcell lines leads to an overwhelming enhancement

This work is supported by The Rendina Family Foundation, 2011–2013Florida Center for Brain Tumor Research (FCBTR)/Accelerate Brain CancerCure (ABC2) Grant, and funds from the State of Florida to Scripps Florida. Theauthors declare no competing financial interests.

1V.C.S. and M.L. contributed equally to this work.dx.doi.org/10.1124/mol.115.097774.s This article has supplemental material available at molpharm.

aspetjournals.org.

ABBREVIATIONS: CM, conditioned media; DMEM, Dulbecco’s minimum essential medium; EGFR, epidermal growth factor receptor; EGFRvIII,epidermal growth factor receptor variant III; FBS, fetal bovine serum; GBM, glioblastoma multiforme; HGF, hepatocyte growth factor; IPA, ingenuitypathway analysis; JNK2, c-Jun N-terminal kinase isoform 2; p–c-Met, activation of c-Met; RTK, receptor tyrosine kinase; shRNA, small hairpin RNA.

962

http://molpharm.aspetjournals.org/content/suppl/2015/10/09/mol.115.097774.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from

in tumor growth, invasion, and resistance to radiation andchemotherapies (Nagane et al., 1996; Huang et al., 1997).The c-Jun NH2-terminal family of kinases (JNK) consists of

three isoforms, JNK1, JNK2, and JNK3,which are activated by avariety of stimuli, including UV light, cytokines, and growthfactor signaling (Davis, 2000). JNKs have many cellular sub-strates and play diverse cellular roles, from induction ofapoptosis to proliferation, depending upon the cell type andinitial stressor (Kennedy and Davis, 2003). In GBM, constitutiveJNKactivation is observed in up to 86% of human glioblastomas,where activation strongly correlates with both histologic gradeand expression of EGFR (Antonyak et al., 2002; Cui et al., 2006;Li et al., 2008). JNK1 and JNK2 are further known to regulateGBM stem cell–like characteristics and tumor initiating poten-tial. Most notably, genetic knockdown of JNK2 impairs in-tracranial tumor formation and extends survival in GBMmouse models, suggesting that this particular isoform is impor-tant for the pathology of the disease (Matsuda et al., 2012; Yoonet al., 2012).Here, we have investigated the involvement of the JNK2

isoform in EGFRvIII-driven GBM tumorigenesis. We demon-strate that JNK2 is necessary for the growth of aggressiveEGFRvIII-positive tumors in vivo and invasiveness ex vivo.Our mechanistic studies identify a critical role for theEGFRvIII-JNK2 signaling axis in the regulation of HGFproduction and c-Met activation. Importantly, JNK2-dependent secretion of HGF is sufficient to activate c-Met,both in a cell intrinsic and extrinsic manner, where condi-tioned media from EGFRvIII-expressing cells promotes c-Metactivation in cells that do not express the mutated receptor.Finally, we demonstrate functional importance of this signal-ing circuit in the promotion of cellular invasion. The additionof exogenous hepatocyte growth factor (HGF) is sufficient topartially restore invasion of JNK2-depleted cells. Together,these data define a critical role for JNK2 in the activation of aHGF/c-Met signaling downstream of EGFRvIII in GBM.

Materials and MethodsCell Lines and Reagents. U87 and T98G human glioblastoma

cells were purchased from the American Type Culture Collection(Manassas, VA). DK-MG cells were purchased from Leibniz-InstitutDSMZ (Braunschweig, Germany). Modified U87 cell lines were akind gift from Dr. Frank Furnari (Ludwig Institute for CancerResearch, University of California at San Diego, San Diego, CA).GBM6 cells were a kind gift from Dr. Jann N. Sarkaria (Departmentof Radiation Oncology, Mayo Clinic, MN). GBM6 cells were main-tained by serial passage in the flanks of nudemice and cultured shortterm as described previously (Carlson et al., 2011). Cell lines weremaintained in Dulbecco’s minimum essential medium (DMEM) withglutamax (GIBCO by Life Technologies, Grand Isle, NY) supple-mented with 10% fetal bovine serum (FBS) (Sigma, St. Louis, MO).Cells were incubated at 37°C, 5% CO2, and 95% humidity. U87vIIIand U87vIII-KD cells were maintained in media containing G418(Life Technologies). To study the effects of conditioned media (CM),cells were seeded for 24 hours and subsequently serum starved for anadditional 24 hours. CMwas obtained from cells in serum-freemediaand clarified by centrifugation.

Lentiviral Transduction. Sequence-specific small hairpin RNA(shRNA) pGIPZ vectors for the inhibition of JNK2 or nonsilencing controlwere purchased from Open Biosystems (Thermo Fischer Scientific,Pittsburg, PA). shRNA JNK2 sequences were 1) 59-CTAGCAA-CATTGTTGTGAA-39; 2) 59- CTTCTGAAGTTATCTCTTA-39; and 3) 59-GCATTAAAGCAGCGTATC-39. Lentiviral particles were generated using

the Trans-Lentiviral shRNA packaging kit (TLP5912; Thermo FischerScientific) per themanufacturer’s recommendations. U87vIII cells (1�106) were seeded in 10-cm2 dishes and transduced with shRNAlentiviruses for a period of 72 hours. Transduced cells were selectedby the addition of 1 mg/ml of puromycin (Sigma). Stable cell lines weremaintained at 37°C in a humidified 5% CO2 atmosphere in DMEMwith glutamax (GIBCO by Life Technologies) supplemented with 10%FBS.

Western Blot Analysis. GBM cell lines were seeded into six-wellplates at a concentration of 2.5 � 105 cells/well, cell media wasremoved, and cells were washed in ice-cold phosphate-bufferedsaline and pelleted. Cells were lysed in 100 ml of ice-cold RIPAbuffer (Boston BioProducts, Ashland, MA) containing protease andphosphatase inhibitors (Roche Applied Science, Indianapolis, IN)and centrifuged at 14,000g for 15 minutes at 4°C. Protein concen-trations were determined using the Pierce BCA protein assayreagent (Thermo Scientific, Rockford, IL). Samples were analyzedusing NuPAGE Novex Bis-Tris polyacrylamide gel electrophoresis(Life Technologies). Gels containing the separated protein weretransferred to nitrocellulose membranes using standard protocols.Membranes were probed with the following antibodies: anti-EGFR,anti–phospho-EGFR (Y1068), anti–c-Met, anti–phospho-c-Met (Y1234/1235), anti–phospho-c-Met (Y1349), anti–phospho-GRB2-associatedbinding protein 1 (Y307), anti-JNK2, anti–glyceraldehyde 3-phosphatedehydrogenase, or anti–a-tubulin antibodies. All primary antibodieswere purchased from Cell Signaling Technology (Danvers, MA) andused at a 1:1000 dilution. After repeated washes with Tris-bufferedsalinewith 0.1%Tween 20 (20mMTris, pH7.6, 140mMNaCl, and 0.1%Tween 20), membranes were incubated with the appropriate IRDye-conjugated secondary antibody (LI-COR Biosciences, Lincoln, NE) in a1:10,000 dilution. Membranes were imaged using the LI-COR Odysseyinfrared imaging system (LI-CORBiosciences). HumanHGF polyclonalantibody was purchased from R&D Systems (Minneapolis, MN).

Enzyme-Linked Immunosorbent Assay. Cells (2.5 � 105) wereplated in six-well plates for 24 hours and subsequently serum starvedfor an additional 24 hours. Human hepatocyte growth factor levelswere determined using a humanHGF enzyme–linked immunosorbentassay kit per the manufacturer’s recommendation (Invitrogen by LifeTechnologies).

Invasion Assays. Invasion assays were performed using theCultureCoat 24-well basement membrane extract–coated cell in-vasion assay (Trevigen Inc., Gaithersburg, MD). The assay consist ofan 8-mm pore sized modified Boyden chamber coated with a thinlayer of basement membrane extract. Cells were cultured to 80%confluence and subsequently serum starved for 4–6 hours, resus-pended in serum-free media, and added to the top of each chamberper the manufacturer’s recommendations. Recombinant humanHGF (40 ng/ml) was added to the cells (R&D Systems). DMEM withglutamax supplemented with 10% FBS was added to the bottom ofeach chamber (500 ml). Invasive cells on the underside of themembranes were dissociated in a solution containing calcein-AM.The plates were analyzed using a SpectraMax (Molecular Devices,Sunnyvale, CA) plate reader, with fixed excitation and emissionwavelengths (485/520 nm).

Microarray Analysis. TotalmRNAwas extracted fromGBM celllines using the RNeasy Plus Mini Kit (Qiagen Sciences, German-town, MD). Using the Applied Biosystems High Capacity cDNAReverse Transcription Kit (Applied Biosystems, Foster City, CA),RNA (10 mg) was reverse transcribed into single-stranded cDNA perthe manufacturer’s recommendations. Samples were heated to 95°Cfor ∼1 minute to denature RNA/cDNA hybrids, followed by treat-ment with RNaseA for 30 minutes at 37°C, and DNA was purifiedusing a QIAquick PCR Purification Kit (Qiagen Sciences) per themanufacturer’s recommendation. DNA concentrations were calcu-lated using the NanoDrop 1000 spectrophotometer (Thermo Scien-tific). Hybridization of the cDNA to the Human Gene Expression12x135K array was performed at the Florida State UniversityNimbleGenMicroarray Facility. Differentially expressed genes were

JNK2 Mediates EGFRvIII-Driven HGF Production 963

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from

identified using the following criteria: fold change . 2 in bothdirections and unpaired t test P value , 0.05. Three hundred andsixty probes were identified as the “JNK2 signature” and were im-ported into the Ingenuity pathway analysis tool (IPA) for pathwayand upstream regulator analysis.

In Vivo Tumor Xenograft. MaleNu/Numice were obtained fromCharles River Laboratories (Wilmington, MA), received food andwater ad libitum, and were kept in a controlled environment at anambient temperature on a 12-hour light/dark cycle. Experiments wereperformed when animals reached 5–6 weeks of age, upon which micewere subcutaneously inoculated in the flankwith 7.5� 105 GBM cells.Tumors were measured using calipers, and tumor volumes weredetermined using methodologies where width, length, and heightare used [e.g., v5 (p/6)� (L�W�H)] (Tomayko and Reynolds, 1989).All animal studies were approved by the Scripps-Florida InstitutionalAnimal Care and Use Committee, and care and maintenance were inaccordancewith the principles described in theGuide for Care andUseof Laboratory Animals (National Institutes of Health Publication 85-23, 1985).

Statistical Analysis. All values in the figures are presented as themean6 standard deviation of at least three independent experiments,except for the in vivo study, where the values are presented as themean 6 standard error of the mean. All the experiments were

analyzed using a Student’s t test, where significant P values werepresented as *P # 0.05, **P # 0.01, and ***P # 0.001.

ResultsJNK2 Is Required for EGFRvIII-Driven GBM Tumor-

igenicity. To investigate the role of JNK2 in EGFRvIII-mediated GBM tumorigeneis, we first examined JNK2pathway activation in GBM cells expressing EGFRvIII. U87cells were engineered to express the mutated receptor andanalyzed for activated signaling by measuring phosphoryla-tion of the downstream transcription factor c-Jun. Using thisapproach, we identified that EGFRvIII expression in bothU87cells as well as the patient-derived line GBM6 resulted in theactivation of JNK2. In contrast, phosphorylation of thedownstream transcription factor c-Jun was not detected inU87 cells in the absence of the mutant EGFR, and silencing ofJNK2 was sufficient to impair activation of c-Jun (Fig. 1A).These findings suggest an important role for the JNK2 isoformin the activation of signaling downstream of EGFRvIII.Since previous studies have identified a role for JNK2in the promotion of a tumorigenic phenotype in GBM

Fig. 1. JNK2 is required for the tumorigenicity and invasiveness of EGFRvIII-expressing GBM cells. (A) Western blot analysis of EGFR and c-Junphosphorylation in lysates fromU87, U87vIII, andU87vIII-JNK2shRNA cells and a primary (GBM6) cell line. (B) Growth of U87vIII subcutaneous flanktumors expressing indicated shRNAs. (C) Heatmap generated frommicroarray data showing differentially expressed genes (P, 0.05;.2-fold change) inJNK2knockdown versus nontargeting knockdown (shNT)U87vIII cells. (D) Canonical pathways significantlymodulated by JNK2, as identified by IPA ofmicroarray data. (E) Invasive capacity of U87 cells versus U87vIII cells expressing either JNK2 or nontargeting shRNAs over 24 hours (***P# 0.001). (F)Invasive capacity of DK-MG EGFRvIII-expressing cells with either JNK2 or nontargeting shRNAs over 24 hours (***P # 0.001).

964 Saunders et al.

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from

(Antonyak et al., 2002; Cui et al., 2006), we examined theeffects of JNK2 shRNAs on the growth of U87vIII cells in vivo.JNK2 knockdown resulted in a significant reduction in tumorgrowth over time when compared with either nontransducedcells or those expressing nontargeting shRNAs (Fig. 1B).

JNK2Mediates EGFRvIII-Induced Cellular Invasion.To elucidate the molecular mechanisms by which JNK2confers tumorigenicity to U87vIII, cells expressing eithernontargeting or JNK2 shRNAs were subjected to microarrayanalysis. Using this approach, we identified 360 differentially

Fig. 2. JNK2 is required for activated HGF/c-Met signaling in EGFRvIII-expressing GBM cells. (A) IPA reveals multiple direct and indirect interactionsbetween genes involved in cell movement. Interactions between signature genes (differentially expressed) based on the Ingenuity Knowledge Base and(B) direct downstream targets of HGF: downregulated in JNK2 knockdown cells (green) and upregulated in JNK2 knockdown cells (red). (C) Enzyme-linked immunosorbent assay analysis of HGF production in supernatants derived fromU87 cells or U87 cells expressing EGFRvIII versus a kinase deadversion of the receptor (KD) (**P # 0.01). Immunoblots showing the effects of EGFRvIII activity on activation of c-Met and GRB2-associated bindingprotein 1 (Gab1) in (D) U87 cells and (E) T98G. (F) Enzyme-linked immunosorbent assay analysis of HGF levels in media derived from U87 cells and inU87vIII cells expressing indicated shRNAs (***P# 0.001). (G) Immunoblots showing the effects of EGFRvIII expression and JNK2 knockdown on c-Metand Gab1 phosphorylation status in (G) U87 cells and (H) DK-MG EGFRvIII-expressing cells.

JNK2 Mediates EGFRvIII-Driven HGF Production 965

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from

expressed genes (P# 0.05;.2-fold change), with the majorityof transcripts showing decreased expression upon JNK2knockdown (Fig. 1C). IPA further revealed key canonicalpathways regulated by JNK2, which included categories, suchas cancer, tissue development, cellular growth and prolifera-tion, and inflammation (Fig. 1D). Notably, the most signifi-cantly enriched category was cellular movement, with 68genes related to this pathway showing differential expressionupon JNK2 knockdown (Supplemental Table 1). These obser-vations are consistent with a large body of evidence showingthat EGFRvIII mediates the invasiveness of GBM tumors(Zhu et al., 2009; Dunn et al., 2012) and implicates a role forJNK2 in this phenotype. We therefore examined the effects ofJNK2 knockdown on cellular invasion. Since JNK2 is impli-cated in GBM cell proliferation and to distinguish this effectfrom the invasive phenotype, all the experiments wereperformed in a short-term period after modulation of JNK2expression where the proliferation was not modulated. Asexpected, EGFRvIII-expressing U87 cells showed an almost2-fold increase in invasive capacity using a Boyden chamberassay. Further, silencing of JNK2 in U87 and DK-MGEGFRvIII-expressing cells resulted in a significant reductionin invasion through the basementmembrane coated transwell(Fig. 1, E and F), suggesting that JNK2 is indeed an importantmediator of GBM invasiveness in EGFRvIII-expressing cells.JNK2 Is Required for HGF Production in EGFRvIII-

Expressing GBM Cells. To identify key effectors of JNK2-mediated invasiveness, we performed further pathwayanalyses on our microarray data. Using the IPA software,we evaluated known gene interactions within the signaturelist (Fig. 2, A and B). A number of genes whose products are

known to be associated with cellular movement were identi-fied as either direct or indirect signaling mediators, includingHGF, insulin-like growth factor binding protein 5, matrixmetallopeptidase 7, platelet derived growth factor receptoralpha, and signal transducer and activator of transcription 5A(Fig. 2A; Supplemental Table 1). Importantly, HGF wasidentified as the most significantly predicted upstream regu-lator of the gene signature associated with JNK2 knockdown(Fig. 2B) (IPA overlap P value 5 5.72 � 105). HGF is the onlyknown activating ligand for the RTK c-Met (Lai et al., 2009)and is known to have tumor-promoting roles in GBM, in-cluding promotion of cell proliferation, cell migration, andmaintenance of self-renewal, leading to therapeutic resistanceof glioma stem cells (Joo et al., 2012).Like JNK2, increased HGF expression and c-Met activation

is associated with an advanced tumor grade and poor progno-sis in patients with GBM (Laterra et al., 1997; Lamszus et al.,1999). Accordingly, the treatment of GBM xenografts witheither neutralizing anti-HGF antibody or c-Met kinase in-hibitor (crizotinib) impairs glial tumor growth (Li et al., 2005;Kim et al., 2006; Martens et al., 2006; Rath et al., 2013).Moreover, studies indicate that constitutive EGFR/EGFRvIIIsignaling results in transactivation of additional RTKs, suchas platelet derived growth factor receptor, vascular endothe-lial growth factor receptor, as well as c-Met itself (Huang et al.,2007b; Stommel et al., 2007). Based on these findings, wehypothesized that JNK2 is required for the activation of HGF/c-MET signaling in EGFRvIII-expressing GBM cells. Toinvestigate this, we first examined HGF production in U87cells, U87 cells expressing EGFRvIII, and those with a kinasedead version of the mutant receptor. Notably, EGFRvIII

Fig. 3. Paracrine activation of c-Met by EGFRvIII requires HGF secretion. (A) Phosphorylation status of c-Met in lysates derived from U87 cells treatedwith conditioned media from either U87vIII or U87vIII-KD cells. (B) Quantification of p–c-Met levels shown in A (**P# 0.01; ***P# 0.001). (C) p–c-Metlevels in T98G lysates treatedwith conditionedmedia from parental U87 or EGFRvIII-expressing cultures pretreatedwith anti-HGF (2 mg/ml) polyclonalantibodies. (D) Activation of c-Met in T98G cells treated with conditionedmedia fromU87vIII cells transfected with either nontargeting small interferingRNAs (siRNAs) or those directed against HGF. (E) HGF production by enzyme-linked immunosorbent assay in U87vIII cells transfected with siRNAsagainst HGF versus a nontargeting control (*P , 0.05).

966 Saunders et al.

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from

kinase activity resulted in an induction of HGF in these cells,as measured by enzyme-linked immunosorbent assay (Fig.2C). Further, activation of c-Met (p–c-Met) and its down-stream adaptor molecule GRB2-associated binding protein 1was observed upon expression of active EGFRvIII in U87 andT98G cells (Fig. 2, D and E) (Li et al., 2013). In the case of thekinase dead mutant, we observed a slight increase in p–c-Met(Y1234/1235) levels, but no increase in p–c-Met (Y1345) oractivation of downstream adaptor molecule GRB2-associatedbinding protein levels andHGF levels were significantly lowercompared with active EGFRvIII. Kinase independent roles forEGFR have been observed and may explain the slight increasein p–c-Met (Y1234/1235) levels (Zhu et al., 2010; Tan et al.,2015). We next investigated the requirement for JNK2 in thisprocess by studying the effects of JNK2 shRNAs on both HGFproduction and activation of c-Met signaling. In confirmationof our gene expression data, knockdown of JNK2 reducedHGFexpression in U87 cells expressing EGFRvIII (Fig. 2F). More-over, JNK2 loss resulted in impaired activation of both c-MetandGRB2-associated binding protein 1 in these cells aswell asin DK-MG EGFRvIII–expressing cells, demonstrating a crit-ical role for JNK2 in EGFRvIII-mediated activation of c-Met(Fig. 2, G and H).EGFRvIII/JNK2–Induced HGF Expression Mediates

Cellular Crosstalk. Expression of EGFRvIII in glioblas-toma is heterogeneous and usually observed in a subpopulationof tumor cells (Nishikawa et al., 2004). Indeed, EGFRvIII-expressing tumors rarely arise independently of amplifiedEGFR, and experimental models imitating the human condi-tion have revealed that tumors containing a small fractionof EGFRvIII-expressing cells significantly enhance tumorgrowth, reducing overall survival (Nagane et al., 1996; Indaet al., 2010). Interleukin 6 and leukemia inhibitory factor, forexample, were identified as EGFRvIII-induced paracrinefactors that stimulated tumorigenicity (Inda et al., 2010). Tothis end, we hypothesized that JNK2 mediated production ofHGF by EGFRvIII-expressing GBM cells may serve to trans-activate c-Met signaling in non–EGFRvIII-expressing tumorcells. To test this, CM derived fromU87 cells expressing eitherEGFRvIII or the kinase dead mutant was first added to U87cells and activation of c-Met was examined. CM derived fromEGFRvIII-expressing cells significantly induced phospho–c-Met in contrast to U87 cells grown in normal media. Further,the phospho–c-Met level was markedly reduced in cellsincubated with CM derived from EGFRvIII-kinase deadexpressing cells (Fig. 3, A and B). To confirm that HGFis the active component of the EGFRvIII CM, we useda neutralizing antibody that efficiently blocks activation ofc-Met in T98G cells treated with recombinant human HGF(Supplemental Fig. 1). Pretreatment of U87vIII CM with theHGF antibody impaired activation of c-Met in T98G targetcells (Fig. 3C). Consistent with this, transfection of U87vIIIcells with small interfering RNAs directed against HGF priorto collection of CM also abrogated activation of c-Met in T98Gcells, demonstrating a critical role for HGF in the activation ofc-Met by EGFRvIII (Fig. 3D) (Huang et al., 2007b).We next assessed whether inhibition of JNK2 can block the

HGF/c-Met paracrine signaling in GBM cells. Conditionedmedia was isolated from cultures of U87, U87vIII, U87vIII-shNT, or U87vIII-shJNK2 cells and incubated with T98Gcells. Cells treated with CM derived from JNK2 knockdowncells resulted in a significant reduction in phospho–c-Met levels

compared with those treated with the CM of cells expressing anontargeting shRNA (Fig. 4A). Moreover, exposure of thesecells to CM derived fromU87vIII cells pretreated with the pan-JNK inhibitor SP600125 (due to a lack of isoform-specificinhibitor) significantly reduced the activation of c-Met (Fig.4B). Finally, we addressed whether the role of JNK2 in cellularinvasion is dependent on the EGFRvIII-HGF signaling circuit.The addition of recombinant human HGF to invading cultureswas able to partially restore the invasive capacity of U87EGFRvIII-JNK2 knockdown cells (Fig. 4C). Moreover, theaddition of recombinant humanHGF did not rescue the growth

Fig. 4. Inhibition of JNK2 is sufficient to impair EGFRvIII-induced c-Metsignaling. (A) c-Met phosphorylation in lysates from T98G cells exposed toconditioned media from U87vIII-shJNK2 versus U87vIII-shNT cells. (B)T98G cells pretreated with either SP6000125 or vehicle control (-) andincubated with conditioned media derived from either U87vIII-shJNK2versus U87vIII-shNT cells were analyzed for p–c-Met levels. (C) Effect ofJNK2 knockdown on invasion capacity of U87vIII cells (**P , 0.01).Addition of recombinant human HGF (rhuHGF) partially restores theinvasive capacity of U87 EGFRvIII-JNK2 knockdown cells (*P , 0.05).

JNK2 Mediates EGFRvIII-Driven HGF Production 967

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from

inhibition of the U87 EGFRvIII shJNK2 cells (SupplementalFig. 2). This suggests thatwhile the EGFRvIII-JNK2 axis playsa role inGBMcell proliferation, this is driven through a pathwayindependent of HGF/c-Met. Collectively, our findings elucidatecritical signaling interactions between EGFRvIII-JNK2 and theHGF–c-Met pathway in GBM cell invasion (Fig. 5).

DiscussionMajor clinical challenges in the treatment of GBM include

the ability of infiltrating tumor cells to disperse into distantbrain tissue and the refractory nature of these cells to currenttherapies (Lefranc et al., 2005; Claes et al., 2007). Accordingly,identification of oncogenic signaling pathways driving GBMinvasiveness and disease progression is required to developtargeted therapies, leading to prolonged overall survival.GBMs are, however, known to be highly heterogeneous interms of their molecular profiling, with complex signalinginteractions governing differential responses to therapeuticintervention. For example, several protumorigenic mecha-nisms have been linked to signaling via EGFRvIII, theconstitutively active EGFR deletion mutant common toGBM (Huang et al., 2007b; Stommel et al., 2007; Bonaviaet al., 2012). EGFR signaling is a known driver of GBMpathogenesis, where it has been shown to promote tumorinitiation and development as well as infiltration of surround-ing brain tissue and resistance to therapy. Importantly,EGFRvIII is able to transactivate multiple RTKs within thesame tumor, resulting in overlapping and redundant signal-ing pathways (Huang et al., 2007b; Stommel et al., 2007). Forexample, c-Met is activated by EGFRvIII in GBM, where it isassociated with advanced tumor grade and poor prognosis(Laterra et al., 1997; Lamszus et al., 1999). Importantly, suchcomplexity is thought to account for the lack of significantefficacy of EGFR-targeted agents in the clinic (Taylor et al.,2012), with recent studies demonstrating that combined

inhibition of EGFRvIII and c-Met synergistically impairstumor cell growth in mouse models of GBM and lung cancer(Huang et al., 2007b; Lai et al., 2009; Lal et al., 2009; Pillayet al., 2009). These findings suggest that an understanding ofthe complex signaling interactions in human cancers arerequired for the rational design of combined and effectivetargeted therapeutic strategies.Amplification of c-Met and HGF has been associated with

highly invasive and metastatic tumors and poor prognosis inmultiple tumors types. In GBM, EGFRvIII overexpression isknown to drive activation of c-Met via a previously uncharac-terized mechanism (Huang et al., 2007b). In this study, wedefine a role for JNK2 as a central mediator of the EGFRvIII-HGF/c-Met signaling circuit in GBM. Specifically, EGFRvIII-dependent activation of JNK2 is required for increasedcellular invasion through transcriptional upregulation of acast of genes involved in tumor cell movement, includingHGF.We show that JNK2-dependent secretion of HGF leads toactivation of c-Met in EGFRvIII-expressing cells, both in aparacrine- and autocrine-dependent manner. Phospho–c-Metis observed not only in GBM cells expressing the mutatedreceptor, but also in cells that lack EGFRvIII expressionfollowing incubation with conditioned media isolated fromEGFRvIII-expressing cells. Confirmation of a role for secretedHGF ligand EGFRvIII cellular crosstalk was achieved usingneutralizing antibodies and small interfering RNAs targetedagainst HGF. Finally, we established that JNK2-HGF signal-ing is important for GBM cellular invasion and recombinantHGF rescues the inhibitory effects of JNK2 knockdown.Collectively, our findings elucidate critical signaling interac-tions in GBM cell invasiveness and cellular crosstalk, both ofwhich are important pathologic features of GBM. Further, ourfindings suggest that selective inhibitors of JNK2 couldrepresent potential therapeutics for use in GBM, particularlywhen combined with additional RTK inhibitors, includingthose targeted against EGFR and c-Met.

Fig. 5. Proposed model for the signaling interactions between EGFRvIII-JNK2 and the HGF–c-Met pathway in glioblastoma. EGFRvIII-mediatedactivation of JNK2 induces secretion of HGF, leading to the activation of c-Met, both in a paracrine- and an autocrine-dependent manner. HGF-mediatedintercellular crosstalk will enhance neighboring non-EGFRvIII cell tumorigenicity.

968 Saunders et al.

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from

Acknowledgments

We sincerely thank Ms. Pamela Clark-Spruill for secretarialassistance and members of the Duckett Laboratory for input andediting of the manuscript. This work was supported in part by a grantfrom the Florida Center for Brain Tumor Research, Accelerate BrainCancer Cure, the Rendina Family Foundation, and funds from theState of Florida to Scripps Florida.

Authorshipship Contributions

Participated in research design: Saunders, Lafitte, Adrados,Quereda, Feurstein, Duckett.

Conducted experiments: Saunders, Lafitte, Adrados, Quereda,Ling.

Contributed new reagents or analytic tools: Fallahi.Performed data analysis: Saunders, Lafitte, Fallahi, Adrados,

Quereda, Ling, Rosenberg, Duckett.Wrote or contributed to the writing of the manuscript: Saunders,

Lafitte, Rosenberg, Adrados, Quereda, Fallahi, Duckett.

References

Antonyak MA, Kenyon LC, Godwin AK, James DC, Emlet DR, Okamoto I, Tnani M,Holgado-Madruga M, Moscatello DK, and Wong AJ (2002) Elevated JNK activationcontributes to the pathogenesis of human brain tumors. Oncogene 21:5038–5046.

Batra SK, Castelino-Prabhu S, Wikstrand CJ, Zhu X, Humphrey PA, Friedman HS,and Bigner DD (1995) Epidermal growth factor ligand-independent, unregulated,cell-transforming potential of a naturally occurring human mutant EGFRvIII gene.Cell Growth Differ 6:1251–1259.

Behin A, Hoang-Xuan K, Carpentier AF, and Delattre JY (2003) Primary brain tu-mours in adults. Lancet 361:323–331.

Beroukhim R, Getz G, Nghiemphu L, Barretina J, Hsueh T, Linhart D, Vivanco I, LeeJC, Huang JH, and Alexander S et al. (2007) Assessing the significance of chro-mosomal aberrations in cancer: methodology and application to glioma. Proc NatlAcad Sci USA 104:20007–20012.

Bonavia R, Inda MM, Vandenberg S, Cheng SY, Nagane M, Hadwiger P, Tan P, SahDW, Cavenee WK, and Furnari FB (2012) EGFRvIII promotes glioma angiogenesisand growth through the NF-kB, interleukin-8 pathway. Oncogene 31:4054–4066.

Carlson BL, Pokorny JL, Schroeder MA, and Sarkaria JN (2011). Establishment,maintenance and in vitro and in vivo applications of primary human glioblastomamultiforme (GBM) xenograft models for translational biology studies and drugdiscovery. Curr Protoc Pharmacol 52:1–14.

Claes A, Idema AJ, and Wesseling P (2007) Diffuse glioma growth: a guerilla war.Acta Neuropathol 114:443–458.

Cui J, Han SY, Wang C, Su W, Harshyne L, Holgado-Madruga M, and Wong AJ(2006) c-Jun NH(2)-terminal kinase 2alpha2 promotes the tumorigenicity of hu-man glioblastoma cells. Cancer Res 66:10024–10031.

Davis RJ (2000) Signal transduction by the JNK group of MAP kinases. Cell 103:239–252.

Dunn GP, Rinne ML, Wykosky J, Genovese G, Quayle SN, Dunn IF, Agarwalla PK,Chheda MG, Campos B, and Wang A et al. (2012) Emerging insights into themolecular and cellular basis of glioblastoma. Genes Dev 26:756–784.

Heimberger AB, Hlatky R, Suki D, Yang D, Weinberg J, Gilbert M, Sawaya R,and Aldape K (2005) Prognostic effect of epidermal growth factor receptor andEGFRvIII in glioblastoma multiforme patients. Clin Cancer Res 11:1462–1466.

Huang HS, Nagane M, Klingbeil CK, Lin H, Nishikawa R, Ji XD, Huang CM, GillGN, Wiley HS, and Cavenee WK (1997) The enhanced tumorigenic activity of amutant epidermal growth factor receptor common in human cancers is mediated bythreshold levels of constitutive tyrosine phosphorylation and unattenuated sig-naling. J Biol Chem 272:2927–2935.

Huang PH, Cavenee WK, Furnari FB, and White FM (2007a) Uncovering therapeutictargets for glioblastoma: a systems biology approach. Cell Cycle 6:2750–2754.

Huang PH, Mukasa A, Bonavia R, Flynn RA, Brewer ZE, Cavenee WK, Furnari FB,and White FM (2007b) Quantitative analysis of EGFRvIII cellular signaling net-works reveals a combinatorial therapeutic strategy for glioblastoma. Proc NatlAcad Sci USA 104:12867–12872.

Hurtt MR, Moossy J, Donovan-Peluso M, and Locker J (1992) Amplification of epi-dermal growth factor receptor gene in gliomas: histopathology and prognosis.J Neuropathol Exp Neurol 51:84–90.

Inda MM, Bonavia R, Mukasa A, Narita Y, Sah DW, Vandenberg S, Brennan C,Johns TG, Bachoo R, and Hadwiger P et al. (2010) Tumor heterogeneity is an activeprocess maintained by a mutant EGFR-induced cytokine circuit in glioblastoma.Genes Dev 24:1731–1745.

Jaros E, Perry RH, Adam L, Kelly PJ, Crawford PJ, Kalbag RM, Mendelow AD,Sengupta RP, and Pearson AD (1992) Prognostic implications of p53 protein,epidermal growth factor receptor, and Ki-67 labelling in brain tumours. Br JCancer 66:373–385.

Joo KM, Jin J, Kim E, Ho Kim K, Kim Y, Gu Kang B, Kang YJ, Lathia JD, CheongKH, and Song PH et al. (2012) MET signaling regulates glioblastoma stem cells.Cancer Res 72:3828–3838.

Kennedy NJ and Davis RJ (2003) Role of JNK in tumor development. Cell Cycle 2:199–201.

Kim KJ, Wang L, Su YC, Gillespie GY, Salhotra A, Lal B, and Laterra J (2006)Systemic anti-hepatocyte growth factor monoclonal antibody therapy induces theregression of intracranial glioma xenografts. Clin Cancer Res 12:1292–1298.

Lai AZ, Abella JV, and Park M (2009) Crosstalk in Met receptor oncogenesis. TrendsCell Biol 19:542–551.

Lal B, Goodwin CR, Sang Y, Foss CA, Cornet K, Muzamil S, Pomper MG, Kim J,and Laterra J (2009) EGFRvIII and c-Met pathway inhibitors synergize againstPTEN-null/EGFRvIII1 glioblastoma xenografts. Mol Cancer Ther 8:1751–1760.

Lamszus K, Laterra J, Westphal M, and Rosen EM (1999) Scatter factor/hepatocytegrowth factor (SF/HGF) content and function in human gliomas. Int J Dev Neurosci17:517–530.

Laterra J, Nam M, Rosen E, Rao JS, Lamszus K, Goldberg ID, and Johnston P (1997)Scatter factor/hepatocyte growth factor gene transfer enhances glioma growth andangiogenesis in vivo. Lab Invest 76:565–577.

Lefranc F, Brotchi J, and Kiss R (2005) Possible future issues in the treatment ofglioblastomas: special emphasis on cell migration and the resistance of migratingglioblastoma cells to apoptosis. J Clin Oncol 23:2411–2422.

Li JY, Wang H, May S, Song X, Fueyo J, Fuller GN, and Wang H (2008) Consti-tutive activation of c-Jun N-terminal kinase correlates with histologic grade andEGFR expression in diffuse gliomas. J Neurooncol 88:11–17.

Li L, Puliyappadamba VT, Chakraborty S, Rehman A, Vemireddy V, Saha D, SouzaRF, Hatanpaa KJ, Koduru P, and Burma S et al. (2013) EGFR wild type antago-nizes EGFRvIII-mediated activation of Met in glioblastoma. Oncogene 34:129–134.

Li Y, Lal B, Kwon S, Fan X, Saldanha U, Reznik TE, Kuchner EB, Eberhart C,Laterra J, and Abounader R (2005) The scatter factor/hepatocyte growth factor:c-met pathway in human embryonal central nervous system tumor malignancy.Cancer Res 65:9355–9362.

Louis DN, Ohgaki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A, ScheithauerBW, and Kleihues P (2007) The 2007 WHO classification of tumours of the centralnervous system. Acta Neuropathol 114:97–109.

Martens T, Schmidt NO, Eckerich C, Fillbrandt R, Merchant M, Schwall R, WestphalM, and Lamszus K (2006) A novel one-armed anti-c-Met antibody inhibits glio-blastoma growth in vivo. Clin Cancer Res 12:6144–6152.

Matsuda K, Sato A, Okada M, Shibuya K, Seino S, Suzuki K, Watanabe E, Narita Y,Shibui S, and Kayama T et al. (2012) Targeting JNK for therapeutic depletion ofstem-like glioblastoma cells. Sci Rep 2:516.

Nagane M, Coufal F, Lin H, Bögler O, Cavenee WK, and Huang HJ (1996) A commonmutant epidermal growth factor receptor confers enhanced tumorigenicity on hu-man glioblastoma cells by increasing proliferation and reducing apoptosis. CancerRes 56:5079–5086.

National Institutes of Health Publication 85-23, 1985 (Guide for Care and Use ofLaboratory Animals).

Nishikawa R, Ji XD, Harmon RC, Lazar CS, Gill GN, Cavenee WK, and Huang HJ(1994) A mutant epidermal growth factor receptor common in human gliomaconfers enhanced tumorigenicity. Proc Natl Acad Sci USA 91:7727–7731.

Nishikawa R, Sugiyama T, Narita Y, Furnari F, Cavenee WK, and Matsutani M(2004) Immunohistochemical analysis of the mutant epidermal growth factor,deltaEGFR, in glioblastoma. Brain Tumor Pathol 21:53–56.

Pillay V., Allaf L., Wilding A. L., Donoghue J. F., Court N. W., Greenall S. A., Scott A.M., and Johns T. G. (2009). The plasticity of oncogene addiction: implications fortargeted therapies directed to receptor tyrosine kinases. Neoplasia 11:448–458.

Porter KR, McCarthy BJ, Freels S, Kim Y, and Davis FG (2010) Prevalence estimatesfor primary brain tumors in the United States by age, gender, behavior, and his-tology. Neuro-oncol 12:520–527.

Rath P, Lal B, Ajala O, Li Y, Xia S, Kim J, and Laterra J (2013) In vivo c-Metpathway inhibition depletes human glioma xenografts of tumor-propagating stem-like cells. Transl Oncol 6:104–111.

Shinojima N, Tada K, Shiraishi S, Kamiryo T, Kochi M, Nakamura H, Makino K,Saya H, Hirano H, and Kuratsu J et al. (2003) Prognostic value of epidermalgrowth factor receptor in patients with glioblastoma multiforme. Cancer Res 63:6962–6970.

Stommel JM, Kimmelman AC, Ying H, Nabioullin R, Ponugoti AH, Wiedemeyer R,Stegh AH, Bradner JE, Ligon KL, and Brennan C et al. (2007) Coactivation ofreceptor tyrosine kinases affects the response of tumor cells to targeted therapies.Science 318:287–290.

Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ, BelangerK, Brandes AA, Marosi C, and Bogdahn U et al.; European Organisation for Re-search and Treatment of Cancer Brain Tumor and Radiotherapy Groups; NationalCancer Institute of Canada Clinical Trials Group (2005) Radiotherapy plus con-comitant and adjuvant temozolomide for glioblastoma. N Engl J Med 352:987–996.

Tan X, Thapa N, Sun Y, and Anderson RA (2015) A Kinase-Independent Role for EGFReceptor in Autophagy Initiation. Cell 160:145–160.

Taylor TE, Furnari FB, and Cavenee WK (2012) Targeting EGFR for treatment ofglioblastoma: molecular basis to overcome resistance. Curr Cancer Drug Targets12:197–209.

Tomayko MM and Reynolds CP (1989) Determination of subcutaneous tumor size inathymic (nude) mice. Cancer Chemother Pharmacol 24:148–154.

Yoon CH, Kim MJ, Kim RK, Lim EJ, Choi KS, An S, Hwang SG, Kang SG, Suh Y,and Park MJ et al. (2012) c-Jun N-terminal kinase has a pivotal role in themaintenance of self-renewal and tumorigenicity in glioma stem-like cells. Oncogene31:4655–4666.

Zhu H, Acquaviva J, Ramachandran P, Boskovitz A, Woolfenden S, Pfannl R,Bronson RT, Chen JW, Weissleder R, and Housman DE et al. (2009) OncogenicEGFR signaling cooperates with loss of tumor suppressor gene functions in glio-magenesis. Proc Natl Acad Sci USA 106:2712–2716.

Zhu H, Cao X, Ali-Osman F, Keir S, and Lo HW (2010) EGFR and EGFRvIIIinteract with PUMA to inhibit mitochondrial translocalization of PUMA andPUMA-mediated apoptosis independent of EGFR kinase activity. Cancer Lett294:101–10.

Address correspondence to: Dr. Derek R. Duckett, The Scripps ResearchInstitute, 130 Scripps Way, Jupiter, FL 33458. E-mail: ducketdr@scripps.edu

JNK2 Mediates EGFRvIII-Driven HGF Production 969

at ASPE

T Journals on D

ecember 22, 2020

molpharm

.aspetjournals.orgD

ownloaded from