Abstract
Targeted therapies that inhibit receptor tyrosine kinases (RTKs) and the downstream phosphatidylinositol 3-kinase (PI3K) signaling pathway have shown promising anticancer activity, but their efficacy in the brain tumor glioblastoma multiforme (GBM) and other solid tumors has been modest. We hypothesized that multiple RTKs are coactivated in these tumors and that redundant inputs drive and maintain downstream signaling, thereby limiting the efficacy of therapies targeting single RTKs. Tumor cell lines, xenotransplants, and primary tumors indeed show multiple concomitantly activated RTKs. Combinations of RTK inhibitors and/or RNA interference, but not single agents, decreased signaling, cell survival, and anchorage-independent growth even in glioma cells deficient in PTEN, a frequently inactivated inhibitor of PI3K. Thus, effective GBM therapy may require combined regimens targeting multiple RTKs.
GBM, the most prevalent tumor in the central nervous system of human adults, is among the most lethal cancers, with a median survival of ∼12 months (1). Aberrant activation of PI3K pathway components appears to be universal in human cancer, including GBM. PI3K is activated upon binding phosphorylated RTKs and/or adaptor proteins at the plasma membrane and signals to multiple downstream effectors, such as Akt (2). Over 80% of GBMs show robust Akt activation, and 40 to 50% have lost or mutated PTEN (3), underscoring the importance of the PI3K pathway in gliomagenesis (4). Activation of the RTK epidermal growth factor receptor (EGFR) is a critical pathogenetic event, with amplification, mutation, and rearrangement observed in more than 40% of cases, making it a compelling target for therapeutic inhibition (5, 6). Other RTKs, such as the platelet-derived growth factor receptors α and β (PDGFRα and PDGFRβ) and MET (1, 7), have been reported to bealtered in GBM, albeit at lower frequencies. Notably, anti-PDGFR therapy has failed in GBM patients (8), and only 10 to 20% of patients benefit from EGFR inhibition (9), pointing to confounding factors that attenuate the response to RTK inhibition.
Coexpression of wild-type (WT) PTEN and a constitutively active EGFR (vIII mutant) in GBM correlates with clinical response to EGFR inhibitors, indicating that PTEN is a response biomarker for anti-EGFR therapy and that its loss renders these agents ineffective by dissociating EGFR inhibition from the abrogation of PI3K pathway activity (10). Alternatively, the activation state of critical survival pathways, such as PI3K and mitogen-activated protein kinase (MAPK), may be determined by the sum of multiple inputs, and multiple RTKs besides EGFR may be simultaneously or sequentially used by GBM cells to maintain signal flux through such pathways. In such a multiple-input system, single-agent anti-EGFR inhibition might be incapable of sufficiently suppressing PI3K signaling in the context of unopposed activation by PTEN loss, resulting in a lack of clinical efficacy. That is, the total signal flux through the PI3K pathway may dictate the response to upstream RTK inhibition, and multiple inputs to PI3K signaling would thus confer insensitivity to the inhibition of any single agent.
To evaluate this possibility, we sought evidence that multiple PI3K activators coexist in glioma cells (11). Because PI3K is activated by binding phosphorylated proteins to its regulatory subunit, p85α, we performed anti-p85α immunoprecipitations to identify PI3K-activating proteins. Multiple tyrosine-phosphorylated proteins were found to be in the PI3K complex (fig. S1A). Guided by Scansite [http://scansite.mit.edu (12)], which identifies potential p85α-binding proteins, we confirmed the endogenous PI3K interaction with specific RTKs by coimmunoprecipitation assays. For example, 5 out of 14 cell lines showed activated ERBB3 (fig. S1B), which mediates the binding of EGFR and ERBB2 to PI3K (13), and 9 out of 20 cell lines had activated growth factor receptor–bound protein 2 (Grb2)–associated binder 1 (GAB1), a docking protein that binds activated RTKs directly or through Grb2 (14). In seven of these 20 cell lines, highly phosphorylated GAB1 coimmunoprecipitated activated MET (fig. S2B) (15).
To define the compendium of coactivated RTKs in GBM, we used an antibody array that allows simultaneous assessment of the phosphorylation status of 45 RTKs. Consistent with figs. S1 and S2, three or more activated RTKs, including EGFR, ERBB3, PDGFRα, and MET, were detected in each of 19 out of 20 cell lines (Fig. 1A and table S1). Most activated RTKs remained phosphorylated under serum deprivation and in tumor cell xenografts (Fig. 1, B and C), indicating that the RTK activation in the transformed cells is not due to ligands in serum-containing culture media. Finally, RTK coactivation is not a distinctive feature of glioma cells, because similar patterns were detected in other solid tumor types such as lung and pancreatic adenocarcinoma cell lines (fig. S3).
Multiple RTKs are activated simultaneously in glioma cell lines. (A) Whole-cell extracts (WCEs) from the glioma cell lines LN382, SF763, LN18, and HS683 were incubated on RTK antibody arrays, and phosphorylation status was determined by subsequent incubation with anti-phosphotyrosine horse-radish peroxidase. Each RTK is spotted in duplicate: The pairs of dots in each corner are positive controls. Each pair of positive RTKdots isdenoted by a red numeral, with the identity of the corresponding RTKs listed below the arrays. These arrays are representative of various RTK coexpression patterns in the 20 glioma cell lines examined in table S1. (B) RTK antibody arrays were used as in (A) with WCEs from an immortalized human astrocyte cell line [E6/E7/hTERT NHA (19)] or the glioma line LNZ308 grown in 10% serum (log) or grown for 48 hours in 0.05% serum (serum starved). (C) RTK antibody arrays were used as in (A) to compare RTK activation in WCEs from xenograft tumors derived from the glioma cell lines SF767 or LN340 or from the corresponding in vitro cultured cells.
To explore the therapeutic implications of RTK coactivation, we used U87MG glioma cells that constitutively express WT EGFR, EGFRvIII (EGFR*), or a kinase-dead EGFRvIII (EGFR*-KD) at levels comparable to those observed in primary GBMs (16). Although MET is phosphorylated and bound to GAB1 in U87MG cells (Fig. 2A and fig. S2B), activated MET was substantially displaced by EGFR in the GAB1/PI3K complex when WT EGFR and EGFR* are expressed. This outcome required the catalytic activity of EGFR, because EGFR*-KD only modestly displaced MET (lane 4 in Fig. 2A). Because EGFR*-KD was expressed at levels similar to those of WT EGFR and EGFR*, it is unlikely that the displacement was simply a consequence of ectopic overexpression. This apparent “swapping” of RTKs within the PI3K complex did not alter downstream signaling (right panel in Fig. 2A), indicating that MET and EGFR act as redundant but independent inputs to their signaling networks. Consequently, coactivated MET (or other RTKs) would be expected to render anti-EGFR inhibition ineffective in extinguishing downstream signaling by replacing activated EGFR in the PI3K complex.
Inhibition of multiple RTKs is necessary to abrogate PI3K and reduce cell survival. (A) The PI3K/GAB1 adaptor complex can readily switch between MET and EGFR binding with little discernable effect on downstream signaling. U87MG parental cells or cells constitutively expressing WT EGFR, the activating vIII deletion mutant (EGFR*), or the vIII mutant with an inactivating mutation in its kinase domain (EGFR*-KD) were immunoprecipitated (IP) with an antibody to the RTK/PI3K adaptor protein, GAB1 (left), and then immunoblots were probed with the indicated antibodies. Heavy chain (hc) is shown to demonstrate equal immunoprecipitation efficiency. WCEs from the same cells were immunoblotted with the indicated antibodies. (B) (Top) U87MG-EGFR* cells were treated with each of the RTK inhibitors [10 μM erlotinib (E), 10 μM SU11274 (S), and/or 10 μM imatinib (I)], and then WCEs were immunoprecipitated with an antibody to GAB1, eluted, and immunoblotted with antibodies to p85α or GAB1. Note the faster migration of GAB1 in lysates from RTK inhibitor–treated cells, which is consistent with a decrease in phosphorylation. (Bottom) WCEs were immunoblotted with the indicated antibodies. (C) Treatment with multiple RTK inhibitors decreases U87MG-EGFR* cell viability. Cells were treated for 72 hours with combinations of 10 μME, 10 μMS, and 10 μM I, or with 100 nM actinomycin D (actD) in 0.05% serum-containing medium, and then cell viability was assayed by adenosine triphosphate quantitation. Error bars indicate SEM; n = four experiments. unt, untreated. (D) Impact of single and combination RTK-inhibitor treatments on soft-agar colony formation of U87MG-EGFR* cells. Cells were plated in 10% serum and 0.4% agarose-containing growth medium with 10 μM of each of the indicated RTK inhibitors. Colonies were counted after 18 days.
To address whether coactivated RTKs confer resistance to single anti-RTK inhibition, we examined the consequences of single and combined inhibition of EGFR and MET in U87MG-EGFR* cells with these two coactivated RTKs, using P-Akt and P-S6ribosomal protein as molecular surrogates of downstream RTK signaling. We first confirmed that treatment with either an EGFR inhibitor (erlotinib) or a MET inhibitor (SU11274) effectively blocked phosphorylation of their intended targets in U87MG-EGFR* cells (fig. S4A). Although treatment with either inhibitor alone had no discernible effect on PI3K association with GAB1, combined inhibition with both erlotinib and SU11274 resulted in the release of p85α from the RTK/GAB1 complex and faster migration of GAB1, which is consistent with a reduction in GAB1 phosphorylation. Accordingly, downstream signaling as measured by P-Akt and P-S6 was inhibited only when two inhibitors were combined (Fig. 2B). Moreover, combined RTK inhibition significantly decreased viability of cultured U87-EGFR* cells (Fig. 2C) and reduced the number and size of soft-agar colonies formed in a stringent anchorage-independent growth assay (Fig. 2D and fig. S4B), whereas single inhibitors were less effective. In line with the lack of detectable phospho-PDGFR or c-KIT in U87MG-EGFR* cells (fig. S4A), treatment with the PDGFR/c-KIT/abl kinase inhibitor imatinib only slighty affected PI3K activation (Fig. 2B); however, when combined with erlotinib and SU11274, triple inhibitor treatment eliminated residual P-Akt and P-S6 activity (compare lanes 4 and 8 in Fig. 2B) and conferred the most dramatic inhibition on viability and anchorage-independent growth (Fig. 2, C and D), possibly reflecting the inhibitory activity of imatinib on other kinases in these cells (17).
Although it has been reported that PTEN loss abrogates the response of GBM to EGFR inhibitors (10), our finding that combination treatments significantly reduced P-Akt and P-S6 in PTEN-mutant U87MG-EGFR* cells (Fig. 2B) suggests that, even with PTEN deficiency, combined signaling from multiple RTKs is required to maintain downstream pathway activation, such as that of PI3K. To examine whether inhibition of coactivated RTKs via combination therapy mitigates PI3K activity in other PTEN-deficient glioma cells, we subjected multiple cell lines, either WT or mutant for PTEN, to single and combination treatment with erlotinib, SU11274, and imatinib. Consistently, PI3K signaling was reduced or completely abrogated with combined inhibition of coactivated RTKs irrespective of PTEN status (Fig. 3A and fig. S5A). Additionally, inhibition of PI3K activity via RTK inhibition abrogated anchorage-independent growth and cell viability (Fig. 3, B and C, and fig. S5, B and C). The decrease in cell viability was mediated in part by inhibition of PI3K signaling, because transient transfection of either myristoylated Akt or p110α-CAAX (C, cysteine; A, aliphatic amino acid; X, any amino acid) increased cell viability in drug-treated cells (fig. S6, P < 0.001). At the same time, the inability to completely rescue viability indicates that PI3K is not the sole mediator of functional RTK signaling. In these treatment studies, we are aware that (i) many kinase inhibitors exhibit activities against multiple kinases in addition to their primary targets (17) [e.g., SU11274 diminished P-PDGFRβ in LN382 cells (fig. S5D)] and (ii) a particular RTK may not be represented or be detected as activated on an antibody array. Nevertheless, in three different glioma cell lines, colony formation and cell viability were partially inhibited by single and dual treatment with RTK inhibitors but were most affected by combined treatment with all three inhibitors (Fig. 3, B and C, and fig. S5, B and C). Given the potential nonspecific actions of these agents, particularly SU11274 [the U.S. Food and Drug Administration (FDA)–unapproved MET inhibitor], we used RNA interference (RNAi) against MET to verify that genetic inhibition of MET can similarly confer enhanced anti-oncogenic activity of erlotinib and imatinib in LN382. Indeed, transfection of LN382 cells with MET small interfering RNAs (siRNAs), in combination with erlotinib and imatinib, resulted in a nearly similar level of inhibition in soft-agar colony formation in an anchorage-independent growth assay (Fig. 3D). Similar trends were observed with RNAi against PDGFR and EGFR (fig. S7). These results support the view that, even in PTEN mutant cells, more robust anti-oncogenic effects can be achieved through combined inhibition of relevant upstream signaling inputs.
Targeting multiple activated RTKs abrogates cell signaling, anchorage-independent growth, and viability. (A) The glioma cell line LN382 was treated with the indicated RTK inhibitors in 0.05% serum-containing medium and immunoblotted as in Fig. 2B. The activated RTKs detected in these cells on antibody arrays as in Fig. 1A are indicated beneath the blots. (B) Combined RTK-inhibitor treatments inhibit softagar colony formation in LN382 cells when plated and treated as in Fig. 2D. (C) The inhibition of multiple activated RTKs decreases cell viability. LN382 cells were treated with 10 μME, 10 μMS, and/or 10 μM I, or with 100 nM actD, and then cell death was assayed as in Fig. 2C. Error bars indicate SEM; n = four experiments. (D) Soft-agar colony inhibition by MET siRNAs combined with RTK inhibitors. LN382 cells were transfected with MET siRNA (siM) or a scrambled negative control (–) and plated and treated with the indicated RTK inhibitors as in Fig. 2D.
Finally, we assayed untreated primary human GBM tumors from newly diagnosed patients for evidence of RTK coactivation. In contrast with a normal brain specimen that had no detectable RTK activation, each of the 14 GBM samples examined by antibody array profiling harbored multiple phosphorylated RTKs (Fig. 4A and table S2). These included known GBM RTKs, such as EGFR, PDGFRα, and MET, as well as RTKs not previously linked to GBM, such as RET, MST1R, and CSF1R. Given the well-known intratumoral heterogeneity in GBM, we performed immunofluorescence staining with phosphospecific antibodies against multiple RTKs and observed coexpression of activated RTKs in individual dissociated cells from a primary GBM (Fig. 4B). Together with the in vitro data above, this evidence of in vivo RTK coactivation supports our hypothesis that concomitant activation of multiple RTKs serves to reduce dependence on any single RTK for the maintenance of critical downstream signaling in a complex tumor microenvironment, thus rendering such tumors refractory to single-agent RTK inhibition.
Multiple RTKs are concomitantly activated in primary GBMs. (A) Antibody arrays were performed as in Fig. 1 on protein lysates extracted from snap-frozen primary human gliomas or normal brain autopsy material. (B) Coexpression of phospho-RTKs in cells dissociated from primary GBM MSK199. Individual tumor-derived cells were immunofluorescently stained with the phospho-RTK antibodies P-EGFR, P-PDGFRα, P-InsR, or P-CSF1R. Each row depicts one field of cells from a slide simultaneously stained with the indicated antibodies. DNA is labeled with Hoechst 33342. Nestin is expressed in neural progenitor cells, tumor endothelial cells, and diffuse gliomas, including astrocytomas and GBMs (20), and olig2 is expressed in neural progenitors, normal oligodendroglia, and diffuse gliomas (21). The bottom row depicts cells stained only with secondary antibodies.
The findings of this study provide a rational explanation for the feeble clinical responses to RTK-inhibitor monotherapy for many solid tumor types and anticipate more favorable outcomes by combinations of drugs against different activated RTKs or single drugs with inhibitory activities against multiple activated RTKs. Moreover, by demonstrating the capability to rapidly profile the activation status of most members of the RTKs in resected GBM specimens and the use of such profiles to tailor rational combination therapies, this study provides proof-of-concept for the eventual implementation of a “personalized” therapeutic paradigm in human cancer (18). Because FDA-approved RTK inhibitors are available and additional drugs are under development, this treatment paradigm could be readily implemented for cancers that are currently highly refractory to existing therapies.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1142946/DC1
Materials and Methods
Figs. S1 to S7
Tables S1 and S2
