A Urokinase Receptor–Bim Signaling Axis Emerges during EGFR Inhibitor Resistance in Mutant EGFR Glioblastoma

The epidermal growth factor receptor (EGFR) is a frequently amplified and mutated gene in glioblastoma (GBM; ref. 1). The constitutively active deletion mutant, ΔEGFR (also known as EGFRvIII), is a therapeutic target as it confers enhanced tumorigenicity (2) and is required for maintenance of glioma growth in vivo (3). EGFR tyrosine kinase inhibitors (TKI) are approved for treatment of certain malignancies in which kinase domain-mutated EGFR is a driver oncogene, such as non–small cell lung cancer (NSCLC), and patients with tumors that harbor these mutations are sensitive to the drugs (4, 5). In contrast, the efficacy of EGFR TKIs in GBM has been limited by inherent and acquired TKI resistance (6).

Although resistance to EGFR-targeted therapy occurs in other solid tumors (7, 8), the same mechanisms do not appear to be operational in most GBMs (9–11). For example, the T790M gatekeeper mutation is a predominant mechanism of acquired resistance to EGFR inhibitors in NSCLC (12), but is not found in TKI-treated GBM tumors. Thus, there is precedence for the existence of distinct resistance mechanisms in GBM. Some GBMs demonstrate PTEN deletion, which is associated with EGFR TKI resistance due to activation of Akt signaling (13). Activation of TKs other than EGFR also may contribute to resistance (14), as in the cases of Src- and FGFR-induced phosphorylation of PTEN at tyrosine 240 (15).

uPA receptor (uPAR) is a glycosylphosphatidylinositol-anchored membrane protein, which associates with integrins and RTKs to form a potent signaling complex (16, 17). uPAR-initiated cell signaling inhibits apoptosis, promotes release of tumor cells from dormancy (18, 19), induces stem cell–like properties in cancer cells, and promotes invasion and metastasis (20–22). We have previously shown that in GBM cells, uPAR expression is required for escape from ΔEGFR oncogene dependence in vivo (23), which suggests that uPAR-mediated signaling may serve as an EGFR TKI resistance mechanism. However, gene silencing in vivo may not model changes that occur in GBM treated with targeted drugs. Herein, we describe novel model systems of GBM acquired resistance in which EGFR TKIs were administered chronically in vitro or in vivo. Using these model systems, we demonstrate that uPA–uPAR signaling mediates repression of Bim, a proapoptotic protein, to promote TKI resistance in GBM. Regulating Bim levels by targeting uPAR or by modulating signaling pathways downstream of uPAR may represent a strategy to overcome resistance to EGFR-targeted therapies in GBM.

Ink4a/arf^−/− ΔEGFR murine astrocytes were a gift from Ronald DePinho (University of Texas MD Anderson Cancer Center, Dallas, TX) and Robert Bachoo (University of Texas Southwestern Medical Center, Dallas, TX) (24). U373 ΔEGFR-dependent and escaper cells were cultured as described previously (3). U373 and murine astrocytes were recently authenticated using short tandem repeat (STR) DNA fingerprinting by ATCC and DDC, respectively. Gefitinib, erlotinib, and U0126 were obtained from LC Labs. GBM39 cells were a gift from David James, University of California San Francisco, CA (25). ABT-737 was obtained from Selleck Chemicals and provided by the Small Molecule Development group at the Ludwig Institute for Cancer Research. GSK1120212 was kindly provided by Tona Gilmer of GlaxoSmithKline. Cisplatin was obtained from Calbiochem.

Ink4a/arf^−/− ΔEGFR astrocytes (24) were seeded in 0.3% agar on top of a base layer of 0.6% agar at 2 × 10³ cells per well in a 6-well dish in 1 μmol/L gefitinib, erlotinib, or vehicle (DMSO). A feeder layer containing drug or vehicle was replenished twice per week. After 2 weeks, colonies greater than or equal to approximately 50 cells were counted in five random fields per well, and the drug dose was escalated to 2 μmol/L. Three to 5 weeks later, drug-resistant colonies visible to the naked eye were isolated using a blunt-ended pipette tip, dissociated, plated, and maintained in 2 μmol/L gefitinib or erlotinib.

All animal procedures were approved by the IACUC at the University of California, San Diego. A total of 1 × 10⁶ ink4a/arf^−/− ΔEGFR astrocytes were injected subcutaneously into the flanks of 6- to 8-week old female athymic mice. When average tumor volume reached approximately 500 mm³, mice received vehicle, 200 mg/kg gefitinib, or 150 mg/kg erlotinib daily five times per week (Monday to Friday) by oral gavage. To generate TKI resistance, mice continued to receive TKIs until tumors grew to 1,500 mm³ or became ulcerated. Erlotinib was decreased to 75 mg/kg after 10 doses due to drug-induced skin rash. To establish cell lines, tumors were removed from mice, and mechanically dissociated in sterile PBS and cultured in media containing 1 μg/mL puromycin to ensure elimination of any contaminating mouse cells.

Cells (1 × 10³) were seeded in a 96-well plate in media containing 1% FBS. Twelve to 15 hours later, drugs or vehicle were added in quadruplicate wells and incubated for 72 hours. Viability was assessed with the WST-1 cell proliferation reagent (Clontech). Drug treatments were normalized to vehicle-treated control to calculate percent cell viability.

Fifteen to 40 μg of cell lysate was resolved on a 7.5% to 15% polyacrylamide gel. Proteins were transferred, blocked with 5% milk or 3% BSA followed by primary antibody incubation. Primary antibodies included EGFR pY1068, AKT pS473, Bim, phospho-S6 ribosomal protein, S6 ribosomal protein, phospho-p42/44 MAPK, p42/44 MAPK, PARP (Cell Signaling Technology), EGFR clone 13 (BD Transduction Labs), PTEN, Tubulin (Santa Cruz Biotechnology), and β-actin (Sigma). Membranes were washed three times with TBS + 0.1% Tween-20 and incubated with anti-mouse or anti-rabbit-HRP antibody (Sigma). Chemiluminescence was detected using Amersham ECL Western Blotting Detection Reagent (GE Healthcare) or SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific). Images were obtained using ChemiDoc MP imaging system and Image Lab 4.1 software (Bio-Rad).

Cells were transduced with the pLV.FLuc.GFP lentivirus as previously described (26). Forty-eight hours following infection, in vitro luminescence was determined using the luciferase assay system (Promega), or 1 × 10⁵ cells were stereotactically injected into the brains of 6- to 8-week old female athymic mice, 2 mm lateral and 1 mm anterior to the bregma. Tumor growth was monitored by bioluminescent imaging (BLI), whereby d-luciferin was injected intraperitoneally at a dose of 150 mg/kg, and after 5 minutes luminescence was determined as measured by total flux using IVIS spectrum (PerkinElmer). Once tumors were detectable, mice were treated five times per week (M-F) by oral gavage with 200 mg/kg gefitinib or vehicle (0.5% Tween-80, 0.5% methylcellulose). IVIS imaging software was used to quantify total flux of luminescence in regions of interest.

RNA was isolated with the RNeasy Kit (Qiagen). cDNA was synthesized with the iScript cDNA Synthesis Kit (Bio-Rad). qPCR was performed with a System 7300 (Applied Biosystems) and a one-step program: 95°C for 10 minutes, 95°C for 30 seconds, and 60°C for 1 minute for 40 cycles. uPA and HPRT-1 gene expression were measured using TaqMan assays (Invitrogen) in triplicate with internal duplicate determinations.

uPA-specific siRNA (25 nmol/L; 5′-CAUGUUACUGACCAGCAAC-3′) or nontargeting control (NTC) siRNA (Dharmacon) siRNAs were introduced into cells using Lipofectamine 2000 (Invitrogen) in serum-free medium (SFM) for 4 hours. Cultures were allowed to recover in serum-containing medium for 12 hours and serum starved for 36 hours. The extent of gene silencing was determined by qPCR and immunoblot.

GBM 39 cells (1 × 10⁶) in PBS were injected subcutaneously into the flanks of nude mice in a 1:1 mixture with Matrigel (BD Biosciences) in a final volume of 100 μL. Following tumor establishment as measured using calipers, mice were randomized to receive vehicle or erlotinib at a dose of 150 mg/kg delivered by oral gavage daily for 4 days.

Cells were serum starved in phenol-red free media overnight and conditioned media concentrated 20-fold using centricon (Millipore). Concentrated conditioned media were diluted 1:15 with 0.65 mmol/L Val-Leu-Lys-p-nitroanilide (VLK-pNA; Bachem) and 0.94 mmol/L plasminogen purified from human plasma as previously described (27) in PBS. After incubating samples at 37°C for 1 hour, VLK-pNA hydrolysis was detected by measuring absorbance at 406 nm.

Tissue sections were cut from formalin-fixed paraffin-embedded xenograft tissue. Four-micrometer sections were stained with anti-phospho-p44/42 MAPK (Thr202/Try 204; Cell Signaling Technology) on a Ventana Discovery Ultra (Ventana Medical Systems). Antigen retrieval was performed using CC1 for 24 minutes at 95°C. Primary antibody was incubated at 1:1,800 for 1 hour at 37°C followed by UltraMap (Ventana Medical Systems) and DAB detection. Slides were dehydrated through alcohol and xylene and coverslipped. Phospho-epitope labeling was validated by eliminating staining after treating sections with lambda protein phosphatase (New England Biolabs; 2,400 U) for 2 hours at 37°C before application of primary antibody. Before phosphatase treatment, tissue sections were deparaffinized and antigen retrieval was performed as described above.

The Cancer Genome Atlas (TCGA) glioblastoma transcriptomal profiles were downloaded from TCGA Data Portal (https://tcga-data.nci.nih.gov/tcga/). Level 3 normalized, Agilent 244K gene expression mRNA array data was downloaded and genes were median centered. For each specimen, MAPK pathway activity score was calculated using the t score method developed by Creighton and colleagues (28). In brief, each gene in the MAPK signature that was overexpressed was assigned the value of +1; genes that were underexpressed were assigned −1. The normalized expression values of the signature genes in the clinical specimen were then plotted against these assigned values, and Pearson correlation coefficient was determined. The correlation ranged from −1 (denoting low pathway activity) to +1 (denoting high pathway activity).

All statistical analyses were performed using the GraphPad Prism 6 software. The specific test used for each experiment to determine significance (P < 0.05) is indicated in the text of the results and/or figure legends. Data are representative of results obtained in at least three independent experiments.

To define markers of EGFR TKI response in GBM, we examined the ΔEGFR-expressing patient-derived xenograft GBM39, which is sensitive to EGFR blockade (29). Erlotinib caused dramatic tumor regression in mice (Fig. 1A), which coincided with upregulation of Bim (Fig. 1B), a proapoptotic Bcl2-family member (30). Moreover, Bim levels were, on average, higher in tumors from patients treated with the EGFR/HER2 inhibitor lapatinib (29) who were therapeutic responders than nonresponders (Supplementary Fig. S1A).

We previously developed a genetic model in which doxycycline-mediated suppression of ΔEGFR inhibits U373 MG xenograft tumor growth until the eventual emergence of escaper tumors that have circumvented ΔEGFR oncogene dependence. This model is analogous to sensitivity and resistance to EGFR TKIs, respectively. The Ras–MAPK pathway is thought to contribute to the escape process (3), and cell lines established from escaper tumors, which harbor sustained ΔEGFR inhibition, demonstrate increased uPA expression and dependency on uPAR signaling to maintain ERK1/2 activation (23, 31). Because Bim was associated with response to pharmacologic EGFR inhibition (Fig. 1A and B), we asked if there was a relationship between uPAR and Bim in this genetic model. Cells were transfected with uPA-specific or nontargeting control siRNA, and uPA gene-silencing was greater than 95% as determined by qPCR (23, 31). Interestingly, when uPA was silenced in cells lacking ΔEGFR, a very robust increase in Bim and a decrease in phospho ERK1/2 were observed (Fig. 1C, lanes 4, 6, and 8). These results suggest that uPAR signaling plays a key role in suppressing Bim expression when ΔEGFR is neutralized in GBM cells, possibly through the Ras–MAPK pathway.

To address the hypothesis that uPAR signaling suppresses Bim and promotes cell survival during EGFR inhibition, we developed novel model systems of EGFR TKI resistance. We used ΔEGFR-expressing, PTEN wild-type primary ink4a/arf^−/− murine astrocytes that form highly invasive intracranial tumors in mice similar to GBMs (24). Importantly, ΔEGFR and wild-type PTEN are associated with response to EGFR TKIs in patients with GBM (13, 24). These cells and the tumors generated from them are hereafter referred to as “parental.” To generate resistance, gefitinib and erlotinib were administered chronically to parental cells seeded in soft agar and to mice harboring established parental xenograft tumors. As anticipated, the TKIs inhibited colony formation in soft agar (Supplementary Fig. S1B) and caused dramatic regression of established tumors in vivo (Supplementary Fig. S1C).

Resistance developed in vitro within 17 to 32 days following initial response. TKI-resistant colonies were isolated and expanded to generate clonal cell lines isolated as resistant to gefitinib or erlotinib, denoted “G-” and “E-,” respectively (Supplementary Fig. S1D). Chronic TKI treatment of mice harboring xenograft tumors elicited a sustained period of dormancy for approximately 20 days, after which tumors became resistant and rapidly increased in size (Fig. 2A). Cell lines generated from vehicle-treated sensitive parental tumors as well as from the gefitinib or erlotinib-resistant tumors were denoted “GR-” and “ER-,” respectively, to distinguish them from the in vitro resistant cell lines.

TKI resistance was confirmed in vitro in cell lines generated from soft agar (Fig. 2B, Supplementary Fig. S1D) and from tumors (Fig. 2C). Parental cell viability was significantly inhibited in a dose-dependent manner by gefitinib or erlotinib, whereas this was not the case for the resistant cell lines (Fig. 2B and C, Supplementary Fig. S2A–S2C). To test whether EGFR TKI resistance was maintained orthotopically in vivo, parental and G12 cells transduced with a luciferase-expressing lentivirus (Supplementary Fig. S3A) were grown intracranially in mice and, once established as tumors, were treated with gefitinib or vehicle. Parental tumors decreased in size in response to gefitinib (Fig. 2D, P < 0.05), but G12 tumors were unaffected (Fig. 2D). Thus, TKI resistance occurs as a result of chronic dosing of ink4a/arf^−/− ΔEGFR–expressing astrocytes and xenograft tumors.

We first assessed ΔEGFR expression, phosphorylation, and PI3K pathway activity as potential mechanisms contributing to EGFR TKI resistance. In resistant cells and tumors, ΔEGFR was expressed and phosphorylated upon drug removal, indicating that the cells were not resistant due to loss of the drug target (Fig. 3A and D and Supplementary Fig. S3B and S3C). Acute treatment of parental cells with gefitinib caused a slight decrease in total EGFR, and many of the resistant cell lines exhibited similar total EGFR levels. This observation was consistent with previous reports of cells that have decreased levels of the EGFR family member ErbB2 upon treatment with afatinib, a dual EGFR/ErbB2 inhibitor (Fig. 3D; ref. 32). Gefitinib or erlotinib inhibited EGFR phosphorylation in resistant samples (Fig. 3A and D and Supplementary Fig. S3B), which indicated that the drug maintained access to and could effectively inhibit the target. These observations ruled out two candidate mechanisms of TKI resistance: efflux pump-mediated decreased intracellular drug concentration and a point mutation analogous to T790M, which decreases the ability of drug to bind to and inhibit EGFR kinase activity.

The PI3K–AKT pathway is a canonical EGFR signaling pathway that has been implicated in resistance to EGFR-targeted drugs (13). Phosphorylation of S6, a downstream effector of PI3K, was suppressed by gefitinib or erlotinib treatment to the same extent in parental and resistant cells from the in vitro (Fig. 3A) and in vivo (Fig. 3B) model systems. Expression of PTEN, a negative regulator of the PI3K signaling pathway, was unaltered in most resistant clones with the exception of clone G5 from the in vitro system, which exhibited decreased PTEN levels compared with parent cells (Fig. 3C). As anticipated for a cell line with decreased expression of a PI3K signaling negative regulator, clone G5 had attenuated inhibition of AKT when treated with gefitinib (Supplementary Fig. S3B). Decreased PTEN also occurred infrequently in the in vivo resistance system, in two of the 20 cell lines from resistant tumors (Fig. 3D: GR-2, GR-8). Overall, our results suggest that the predominant mechanism of acquired resistance in these model systems is not associated with modulation or function of the target itself and involves pathways other than PI3K–Akt–S6.

uPAR-mediated signaling was essential for escape of ΔEGFR-silenced U373 MG xenograft tumors from oncogene dependence (23). Moreover, GBM cells typically express high levels of uPAR (33), which mediates antiapoptotic signaling (19) and induces release of cells from dormancy (18, 19). uPAR signaling is dependent upon activation by the uPAR ligand, uPA, which is often secreted by cancer cells or other cells in the tumor microenvironment (16). uPA mRNA expression was significantly increased compared with parental cells in five of seven of the TKI-resistant cell lines from the in vitro system (Fig. 4A). uPA activity in conditioned medium was significantly increased in all the seven resistant cell lines (Fig. 4B). The slight difference in between mRNA and activity may be explained by the presence of endogenous uPA inhibitors, which, if downregulated, would contribute to elevated uPA activity.

We also examined uPA mRNA in the cell lines that acquired resistance to erlotinib or gefitinib in vivo. There was variability in the level of uPA mRNA in the different cell lines; however, in some cells, the increase in mRNA exceeded 50-fold. Twelve of 18 cell lines demonstrated a statistically significant increase in uPA mRNA expression compared with parental cells (Fig. 4C).

Because the uPA–uPAR signaling axis is involved in survival signaling and resistance to TKIs could be caused by differences in cell death, we investigated the ability of gefitinib or erlotinib to induce apoptosis. PARP cleavage, an indicator of apoptosis, was substantial in parental cells treated with gefitinib or erlotinib but was far less pronounced in TKI-resistant cells (Fig. 5A). Furthermore, TKI treatment caused significant elevation of caspase-3/7 activity (Supplementary Fig. S4A) and Annexin/PI staining (Supplementary Fig. S4B) in parental cells and not resistant cells (Supplementary Fig. S4A). Interestingly, TKI-resistant cells exhibited decreased apoptosis in response to the cytotoxic agent cisplatin (Supplementary Fig. S4C). These results indicate that an antiapoptotic TKI-resistance mechanism is induced in these model systems, and suggest that it is general in nature.

To determine if the induction of Bim observed in erlotinib-treated GBM39 tumors (Fig. 1B) and after genetic suppression of ΔEGFR in U373 cells (Fig. 1C) is altered in our models of TKI resistance, we treated parental and resistant cells with gefitinib or elotinib. Bim was greatly induced in the TKI-sensitive cells upon treatment with either drug Fig. 5B and C). However, Bim induction was diminished in TKI-resistant cell lines generated in vitro (Fig. 5B) and in vivo (Fig. 5C). Together, these results support our findings from other GBM systems that expression of Bim is associated with TKI sensitivity and suggest that altered apoptotic signaling and decreased expression and/or function of Bim are central features of TKI resistance.

The Mek–ERK pathway plays an important role in mediating uPAR-dependent antiapoptotic signaling (19). Thus, we tested whether resistance to EGFR TKIs in the astrocyte model systems reflects ERK1/2 activation. TKI-resistant cell lines and tumors possessed elevated levels of ERK1/2 phosphorylation compared with their sensitive counterparts when treated with TKIs (Fig. 6A–C).

The functional importance of elevated ERK1/2 activity in promoting TKI resistance was determined by assessing whether ERK inhibition affected the resistant phenotype and TKI-induced Bim expression. In parental cells, addition of the MEK inhibitor U0126 to gefitinib produced a similar effect to gefitinib alone in suppressing cell viability and inducing Bim expression (Fig. 6D and Supplementary Fig. S5A). In resistant cells, Bim induction was attenuated and cell viability was unchanged in response to treatment with gefitinib, as anticipated (Fig. 6D and Supplementary Fig. S5A). Addition of a Mek inhibitor to gefitinib significantly decreased cell viability compared with either drug alone and caused increased Bim expression (Fig. 6D and Supplementary Fig. S5A). Similar results were obtained with a different, allosteric Mek inhibitor, GSK1120212 (Supplementary Fig. S5B). Thus, inhibition of Mek–ERK signaling conferred sensitivity to TKI-resistant GBM cells.

To directly test the role of Bim as a downstream mediator of ERK1/2 in controlling GBM cell viability, we used ABT-737, a BH-3 mimetic compound that functions by inhibiting Bcl-2, Bcl-w, and Bcl-xL (34). TKI-sensitive and TKI-resistant cells were treated with ABT-737 and gefitinib, alone or in combination. In TKI-sensitive parental cells, treatment with ABT-737, gefitinib, or the combination significantly decreased cell viability (P < 0.05; Fig. 6E). In TKI-resistant cells, cell viability was decreased only upon treatment with the combination of gefitinib and ABT-737 (P < 0.05; Fig. 6E).

Finally, to determine if levels of uPA expression are directly correlative with MAPK signaling in clinical samples, GBM transcriptional profiles from the TCGA Data Portal were analyzed. There was a highly statistical correlation (P < 0.0001) between a gene signature indicative of MAPK activity (28) and uPA expression levels (Fig. 6F), indicating that uPA-mediated Bim repression through MAPK activity, which we detected in two models of EGFR TKI acquired resistance and in a genetic system of ΔEGFR-independent glioma growth, is an accurate reflection of activity of this signaling pathway in patients. Together, these data implicate an ERK–Bim signaling axis in GBM EGFR TKI resistance and support a model in which restoration of Bim function and TKI sensitivity can be achieved either by blocking MEK activity or with a Bim mimetic.

We have identified a novel function of the uPA–uPAR–ERK1/2 signaling axis in mediating Bim repression during resistance to EGFR inhibition in GBM. Bim has not been previously implicated in resistance or response of GBM to EGFR TKIs, but is known to be required for EGFR TKI-induced apoptosis in other solid tumors (35, 36). In our model systems of resistance, elevated uPA caused an increase in signaling through uPAR to Mek–ERK1/2. To our knowledge, this is the first study that directly implicates the uPA–uPAR signaling axis as playing a critical function in a pharmacologic model of resistance to EGFR targeted therapy. uPA levels are increased in primary glioblastoma cell lines resistant to cisplatin as a result of chronic drug dosing (37). Interestingly, the TKI-resistant cell lines we generated were also resistant to cisplatin-induced apoptosis. We hypothesized that uPA–uPAR is functionally important in EGFR TKI resistance based on our previous findings detailing the role of this signaling system in escape from ΔEGFR oncogene dependence (23). The in vitro and in vivo pharmacologic models that we generated here independently support this genetic model and confirm a role for uPA–uPAR signaling in EGFR TKI resistance.

The inherent molecular heterogeneity of GBM makes it likely that several resistance mechanisms are acting simultaneously within one tumor. Like other oncogenes, EGFR amplification and/or mutation can be subclonal (38). Expression of EGFR itself often varies in response to TKI treatment, and tumors could overcome EGFR oncogene dependence in multiple ways. We observed a minority of resistance tumors exhibiting decreased EGFR expression, and some others that appeared to have slightly decreased PTEN. The current study demonstrates the emergence of a uPA–uPAR signaling axis, but does not preclude that other, nonredundant mechanisms may exist. In fact, it is possible that Bim represents a central, downstream, resistance node and targeting this node would have the benefit of bypassing the heterogeneity of upstream resistance mechanisms.

Bim expression is induced transcriptionally by factors such as FoxO3a (39), and its function is suppressed posttranslationally by Mek and ERK1/2-mediated phosphorylation at serine 55 and 69, respectively (40, 41). Bim interacts with dynein light chains (DLC) in nonapoptotic cells, which sequester the protein in the cytosol (42). JNK-mediated phosphorylation of Bim within the DLC binding motif prevents this interaction, allowing Bim to bind to Bcl-2 family members and activate Bax and Bak at the mitochondria. In our model systems of resistance, Bim was induced by TKI treatment in sensitive cells and tumors, but expression was suppressed during TKI resistance. We were able to achieve Bim induction in resistant cells by using a combination of gefitinib and Mek inhibitor. Although we cannot rule out a transcriptional component of Bim control, we believe that the primary avenue by which Bim is suppressed in our models of resistant GBM cells and tumors is likely via a Mek-mediated mechanism of phosphorylation and subsequent degradation (41). In this study, we used the BH3 mimetic compound ABT-737 to restore Bim function, which conferred TKI sensitivity upon resistant cells. ABT-737 is selective for binding to Bcl-2, Bcl-w, and Bcl-xl and has no activity against Mcl-1 (43). Interestingly some tumors become resistant to ABT-737 by upregulating expression of Mcl-1(43), which is a possible scenario in the setting of combined chronic treatment with ABT-737 and EGFR TKI. Thus, it is likely that a more suitable BH3 mimetic would be one that also targets Mcl-1, such as TW-37(43).

To avoid apoptosis, tumor cells may upregulate any number of pathways known to suppress Bim. Signaling through PI3K–AKT or JNK (44) causes attenuation of Bim expression or proapoptotic function, and TKI-resistant tumors in which PTEN has been lost or mutated typically harbor a higher sustained level of PI3K–Akt activity. Mechanistically, Akt-induced phosphorylation and inactivation of FoxO3a prevents transcriptional upregulation of Bim (45). In support of this notion, it has been shown that in the context of HER2-amplified breast cancer with mutated PIK3CA, combination of lapatinib with a PI3K–mTOR inhibitor increases apoptosis significantly more than either agent alone (46). Our results suggest that the combination of EGFR TKIs with Mek inhibitors, explored for EGFR-mutant NSCLC resistant cell lines (47), might be an effective therapeutic strategy for TKI-resistant GBM. An alternative strategy is to interfere with the upstream mechanism responsible for driving Mek signaling and Bim suppression, which in our model system is the uPA–uPAR signaling axis. Although uPAR signaling has been associated with Bim downregulation in macrophages (48), a direct link between uPA–uPAR and Bim has not previously been shown in tumor cells.

One potential approach to targeting uPA–uPAR would be the use of uPA-neutralizing antibodies (49), or a peptide that interferes with uPA function (50). An alternative therapeutic approach for combating resistance that is supported by our findings ignores the upstream mechanisms of Bim suppression and involves the use of BH-3 mimetics to restore Bim function in combination with EGFR TKIs. In fact, the killing effect of EGFR TKIs against NSCLC cell lines is enhanced by BH-3 mimetics (35), and these drugs are being tested clinically as chemosensitizers for several different tumor types (43).

In conclusion, we have found that EGFR TKI-resistant GBM cells and tumors have upregulated uPA to achieve sustained signaling through Mek–ERK1/2 such that Bim expression is maintained at low levels, rendering them incapable of efficiently inducing apoptosis. Through the generation and characterization of multiple model systems, we have identified a novel role for the uPA–uPAR signaling axis in resistance of GBM to EGFR-targeted therapy and have provided the rationale for therapeutic strategies targeting uPA–uPAR or downstream signaling mediators to prevent or reverse resistance.

No potential conflicts of interest were disclosed.

Conception and design: J. Wykosky, S.L. Gonias, W.K. Cavenee, F.B. Furnari

Development of methodology: J. Wykosky, G.G. Gomez, S.R. VandenBerg, S.L. Gonias, F.B. Furnari

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Wykosky, J. Hu, G.G. Gomez, T. Taylor, G.R. Villa, D. Pizzo, S.R. VandenBerg, A.H. Thorne, S.L. Gonias

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Wykosky, J. Hu, G.G. Gomez, T. Taylor, G.R. Villa, C.C. Chen, P.S. Mischel, S.L. Gonias, W.K. Cavenee

Writing, review, and/or revision of the manuscript: J. Wykosky, J. Hu, T. Taylor, G.R. Villa, S.R. VandenBerg, C.C. Chen, P.S. Mischel, S.L. Gonias, W.K. Cavenee, F.B. Furnari

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Hu, T. Taylor, S.L. Gonias, W.K. Cavenee, F.B. Furnari

Study supervision: S.L. Gonias, W.K. Cavenee, F.B. Furnari

This work was supported by grants from the NIH (F32-NS066519) and the American Brain Tumor Association (J. Wykosky), support from the Defeat GBM Research Collaborative, a subsidiary of the National Brain Tumor Society (W.K. Cavenee, P.S. Mischel, and F.B. Furnari), R01-CA169096 (S.L. Gonias), R01-NS080939 and the James S. McDonnell Foundation (F.B. Furnari). W.K. Cavenee is a fellow of the National Foundation for Cancer Research.

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.