Glioblastoma multiforme (GBM) is the most common and most malignant adult brain tumor. A characteristic of GBM is their highly invasive nature, making complete surgical resection impossible. The most common gain-of-function alteration in GBM is amplification, overexpression, and mutations of the epidermal growth factor receptor (EGFR). The constitutively activated mutant EGFR variant III (EGFRvIII), found in ∼20% of GBM, confers proliferative and invasive advantage. The signaling cascades downstream of aberrant EGFR activation contributing to the invasive phenotype are not completely understood. Here, we show myristoylated alanine-rich protein kinase C substrate (MARCKS), previously implicated in cell adhesion and motility, contributes to EGFR-mediated invasion of human GBM cells. EGFRvIII-expressing or EGF-stimulated human GBM cells increased expression, phosphorylation, and cytosolic translocation of MARCKS in a protein kinase C-α–dependent manner. Down-regulation of MARCKS expression with small interfering RNA in GBM cells expressing EGFRvIII led to decreased cell adhesion, spreading, and invasion. Elucidation of mechanisms that promote EGFRvIII-mediated tumorigenesis in GBM, such as MARCKS, provides additional understanding and potential biological targets against this currently terminal human cancer. [Cancer Res 2009;69(19):7548–56]
Glioblastoma multiforme (GBM) is the most common and malignant adult brain tumor ( 1). The highly invasive nature of these tumors precludes complete surgical resection and ultimately leads to the demise of patients. Amplification and overexpression of the epidermal growth factor receptor (EGFR), the most common gain-of-function mutation, is a major contributor to the invasive phenotype. EGFR amplification (found in ∼50% of GBM) is frequently associated with intragenic rearrangements and/or deletions ( 2) that lead to expression of several mutant EGFRs ( 3). The most common mutant receptor, found in ∼40% of GBM that have EGFR gene amplification or ∼20% of all GBM, is EGFR variant III (EGFRvIII), known to increase the invasiveness of glioma cells.
Unlike wild-type EGFR (wtEGFR), EGFRvIII does not bind ligand; however, it is constitutively activated ( 4– 6). The extent to which EGFRvIII differs from wtEGFR in terms of its phosphorylation status, ability to dimerize, and be down-regulated as well as the extent of activation of downstream signaling pathways are the current topics of controversy. What is well established, however, is that GBM cell lines that express EGFRvIII have both an in vitro and an in vivo growth advantage, as its expression has been shown to decrease cell death and increase proliferation, angiogenesis ( 7), and invasion ( 8). The alterations in the proteome of EGFRvIII GBM cells that lead to increased tumorigenic properties, including invasion, are not completely known. The purpose of this study was to gain a better understanding of the molecular mechanism of invasion in EGFRvIII-expressing cells.
Several studies have shown that wtEGFR regulates protein kinase C (PKC) activity ( 9, 10). Furthermore, it is well known that PKC plays a major role in glioma cell invasion ( 11– 13). One of the most common substrates of PKC is myristoylated alanine-rich PKC substrate (MARCKS). In addition, wtEGFR has been shown to lead to MARCKS phosphorylation through PKC ( 10). MARCKS is an ubiquitously expressed protein involved in cell adhesion, spreading, and motility ( 14). Several microarray studies have found MARCKS to be overexpressed in several cancers, such as lung, ovarian, bladder, and GBM, 7 although its detailed expression profile and functional role in gliomas remains unknown. Because of the link between EGFR-PKC and MARCKS and the potential role of MARCKS in tumorigenesis including the invasive process, we hypothesized that MARCKS may contribute to EGFRvIII-mediated glioma cell invasion.
MARCKS cycles between the plasma membrane, where it is unphosphorylated, and the cytosol. Following phosphorylation by PKC, MARCKS translocates to the cytosol, with dephosphorylation resulting in membrane relocalization ( 14). At the membrane, MARCKS interacts directly with phosphatidylinositol-4,5-bisphosphate and inhibits its association with several actin-binding proteins, thus influencing cell spreading and motility. Following PKC phosphorylation of MARCKS and its subsequent translocation to the cytosol, a pool of free phosphatidylinositol-4,5-bisphosphate is released, allowing phosphatidylinositol-4,5-bisphosphate hydrolysis and/or interactions to occur ( 15).
In this study, we screened GBM xenograft explants, cell lines, and operative specimens and found expression, phosphorylation, and cytosolic translocation of MARCKS to be dependent on activation of wtEGFR or expression of the constitutively activated EGFRvIII. Knockdown of MARCKS using small interfering RNA (siRNA) in U373vIII GBM cells led to decreased cell adhesion, spreading, and invasion. These results show altered expression and phosphorylation of MARCKS to play a role in the invasive phenotype of GBM cells with aberrant EGFR signaling, especially those expressing EGFRvIII.
Materials and Methods
Cell lines and xenograft specimens. EGFRvIII was cloned into the tetracycline-inducible Tet-Off vector (pTRE; Clontech) and cotransfected with the regulatory plasmid (pTA) into human U87 and U373 astrocytoma cells. Clones expressing the pTRE-EGFRvIII (U87vIII and U373vIII) or the empty pTRE vector (U87P and U373P) were grown in DMEM supplemented with 10% tetracycline-approved fetal bovine serum (Clontech).
Fresh GBM operative specimens were obtained from patients undergoing a craniotomy at the University Health Network with ethical approval from the University Health Network Research Ethics Board and grown subcutaneously in NOD-SCID mice. GBM pathology was confirmed by a neuropathologist (S.C.). Western blot analysis was done to characterize these tumors for wtEGFR and EGFRvIII expression.
Western blot analysis. Cell lines, xenograft tumors, and GBM operative specimens were analyzed by SDS-PAGE as described previously ( 16). Membranes were probed for 1 h at room temperature, unless otherwise stated, with MARCKS (1:1,000), wtEGFR (1:1,500), 4G10 (overnight at 4°C; 1:1,000; all from Upstate), β-actin (1:20,000; Sigma-Aldrich Canada), phosphorylated MARCKS (overnight at 4°C; 1:1,000; Cell Signaling), PKC-α (1:2,000; BD Transduction Laboratories), transferrin (1:1,000; Zymed), and glyceraldehyde-3-phosphate dehydrogenase (1:2,000; Chemicon).
Laser capture microscopy and quantitative real-time PCR. Frozen serial transverse sections (5 μm) were prepared from flash-frozen GBM samples (n = 5). Laser capture microscopy (LCM) sections were stained with HistoGene LCM Frozen Section staining kit (Arcturus Biosciences) according to the manufacturer's protocol. These sections were evaluated by a neuropathologist (S.C.) to confirm the diagnosis of a GBM and to outline the central and peripheral regions of the tumor. Tumor cells from the peripheral and central regions were captured with PixCell II system using Capsure LCM HS caps (Arcturus). Total RNA was isolated from the captured cells by Picopure RNA isolation kit (Arcturus).
The fidelity and specificity of the RNA extracted from the LCM isolated GBM cells were checked by real-time PCR (RT-PCR) with tumor cell-specific primers. The primers used were glial fibrillary acid protein (5′-CTTGCGGTCCCTTCTTACTCAC-3′ and 5′-CTCAGTCAAAGCAGAGTGGGTG-3′) and S100β (5′-CACTGCTGTTCTTTAAATGC-3′ and 5′-GTGCTGGAGGCACGTTGGAG-3′). Quantitative RT-PCR was carried out on these samples using SYBR Green fluorescence (Invitrogen) and primers against EGFR (5′-GGAGAACTGCCAGAAACTGACC-3′ and 5′-GCCTGCAGCACACTGGTTG-3′), MARCKS (5′-TTGTTGAAGAAGCCAGCATGGGTG-3′ and 5′-ATTCTCCTGTCCGTTCGCTTTGGA-3′), and HPRT1 (5′-ATGCTGAGGATTTGGAAAGG-3′ and 5′-CTCCCATCTCCTTCATCACA-3′), which was used as a control, on the ABI PRISM 7000 Sequence Detection System (Applied Biosystems). Reactions were done in triplicate in three independent experiments and the average fold difference between samples was calculated using the Applied Biosystems software.
Cell adhesion and spreading assays. Two siRNAs against MARCKS were selected from the Stealth Select RNA interference database (Invitrogen) with the following RNA sequences si1: 5′-UUCGCUGCGGUCUUGGAGAACUGGG-3′ and CCCAGUUCUCCAAGACCGCAGCGAA-3′, si2: 5′-AGCGUCGCCGUUUACCUUCACGUGG-3′ and 5′-CCACGUGAAGGUAAACGGCGACGCU-3′, and Stealth interference negative control (S). U373 pTRE-EGFRvIII (U373vIII) and U373P cells were transfected with 10 nmol/L MARCKS siRNA or negative control siRNA or mock transfected.
For the cell attachment assays, 96-well plates were coated with 5 μg/cm2 collagen, fibronectin, or laminin (Sigma; ref. 17). Cells were transfected with MARCKS siRNA, control siRNA, or mock transfected for 72 h and then replated onto the coated 96-well plate. After 30-min incubation at 37°C, unattached cells were washed away thrice in PBS. Attached cells were stained with sulforhodamine B (Sigma) according to the manufacturer's protocol, with absorbance readings taken at 570 nm.
Cell spreading assays were done on 5 μg/cm2 collagen-coated coverslips. Seventy-two hours post-transfection, cells were reseeded on the coated coverslips, fixed, and stained with rhodamine phalloidin (Molecular Probes) at a dilution of 1:100 for 1 h at room temperature and 300 nmol/L 4′,6-diamidino-2-phenylindole (Molecular Probes) for 5 min at room temperature. Images were taken on a Zeiss Axiovert200 equipped with a Hamamatsu Orca AG CCD camera and spinning disk confocal scan head using Volocity acquisition software.
Cell invasion assays. Matrigel invasion chambers (BD Biosciences) were prepared according to the manufacturer's protocol. Cells were transfected with MARCKS, negative control siRNA, or mock transfected, as described above, for 48 h or treated with the PKC-α inhibitor RO-32-0432 (Calbiochem) for 24 h. Cells were then reseeded in the 24-well Matrigel invasion chamber in 0.5% fetal bovine serum–containing medium. The outer well contained DMEM with 10% fetal bovine serum as the chemoattractant. After 16 h of incubation, cells were processed according to the manufacturer's protocol. The number of invading cells was quantified by counting them in at least six random fields (×200).
PKC and EGFRvIII inhibition studies. Cells were grown in the absence or presence of the PKC-α inhibitor RO-32-0432 or the EGFRvIII tyrosine kinase inhibitor AG1478 (Calbiochem) as per the indicated concentrations and times. Following inhibition treatment, cells were stimulated with 100 ng/mL EGF (Chemicon) for 30 min, lysed, and analyzed by SDS-PAGE as described above. Immunoprecipitation was done on 1 mg/mL cell lysate with 4 μg wtEGFR antibody incubated at 4°C overnight.
Fractionation studies. Cell fractionation was carried out using the Qproteome Cell Isolation Kit (Qiagen) as per the manufacturer's protocol using 5 × 105 cells stimulated with EGF and/or treated with the PKC or tyrosine kinase inhibitor as described above.
Statistical analysis. Statistical analysis was done using the two-tailed Student's t test. P values < 0.05 were considered significant.
MARCKS is overexpressed and phosphorylated in wtEGFR- and EGFRvIII-expressing xenograft GBM explants tumors, cell lines, and operative specimens. MARCKS, a protein that plays a role in cell motility, has previously been linked to wtEGFR-PKC signaling, both of which are involved in glioma invasion. We therefore screened a panel of wtEGFR/EGFRvIII-expressing xenograft GBM explant tumors, cell lines, and operative specimens to determine the expression pattern of MARCKS.
Western blot analysis showed that the expression of MARCKS is higher in EGFRvIII-expressing xenografts compared with the non-EGFRvIII-expressing xenografts ( Fig. 1A, xeno 4 ). MARCKS, as well as phosphorylated MARCKS, was also overexpressed in additional EGFRvIII- and wtEGFR-expressing xenografts but was not expressed in EGFR negative xenografts ( Fig. 1A). Transfection of U373 and U87 cell lines with EGFRvIII increased the expression levels of MARCKS and phosphorylated MARCKS ( Fig. 1B). Finally, Western blot analysis of GBM operative specimens showed an association between wtEGFR and MARCKS as well as phosphorylated MARCKS expression levels ( Fig. 1C), where 80% of GBM with wtEGFR also expressed MARCKS/phosphorylated MARCKS. These data indicate that MARCKS expression and phosphorylation are associated with EGFRvIII and wtEGFR expression.
MARCKS and EGFR are highly expressed in the periphery versus central regions of GBM. GBM are characterized by regions of central necrosis, consisting of hypoxic pseudopalisading tumor cells and peripheral regions consisting of tumor cells invading into normal surrounding brain tissue ( 18, 19). We hypothesized that the variations in MARCKS and EGFR expression by the GBM cells within these regions of GBM may account for their phenotypic differences, specifically their invasive capacity. LCM was used to dissect out tumor cells and also associated endothelial cells as an internal control from frozen sections of GBM operative specimens. RT-PCR was used with glial fibrillary acid protein and S100-β primers (Supplementary Fig. S1) to confirm the astrocytic identity of the isolated tumor cells. Quantitative RT-PCR analysis showed that the expression levels of MARCKS and EGFR were significantly (P < 0.05) increased by 9- and 3-fold respectively, in the GBM periphery compared with GBM center ( Fig. 1D).
Knockdown of MARCKS by siRNA decreases cell attachment and spreading. Previous studies have implicated MARCKS in cell attachment and spreading ( 20– 22); thus, we sought to determine if MARCKS plays a similar role in gliomas by knocking down MARCKS using siRNA in U373P and U373vIII cells ( Fig. 2A ). Adhesion assays were carried out on 96-well plates coated with collagen, fibronectin, or laminin, which are known extracellular matrix components overexpressed in gliomas. Compared with mock-transfected cells, knockdown of MARCKS in both U373P and U373vIII cells showed the greatest decrease in attachment (between 48-62% and 46-55%, respectively; P < 0.01) to collagen ( Fig. 2B). Attachment to fibronectin was decreased between 22% and 37% in U373P cells and 25% and 32% in U373vIII cells (P < 0.01; Fig. 2C) compared with mock-transfected cells. Finally, there was no significant decrease in attachment to laminin in either cell line following MARCKS knockdown compared with mock-transfected cells ( Fig. 2D). Therefore, the greatest effect of MARCKS knockdown on attachment was seen on collagen, a moderate effect was seen on fibronectin, and no effect was seen on laminin in both U373P and U373vIII cells. Furthermore, the effect of knockdown on adhesion between U373P and U373vIII cells did not differ significantly.
Because we observed the largest difference in adhesion on collagen, we determined if seeding on this extracellular matrix component affected cell spreading. Figure 3 shows that MARCKS siRNA-transfected cells failed to spread and remained rounded compared with negative control siRNA- or mock-transfected U373P and EGFRvIII cells.
EGFRvIII-induced invasion is mediated through MARCKS. Prior reports have shown that EGFRvIII expression in gliomas increases their invasive potential ( 8). To determine if MARCKS contributes to the invasive phenotype of U373vIII cells, a Matrigel invasion assay was done. Figure 4 shows that knockdown of MARCKS in U373vIII cells decreased invasion to approximately the same levels of U373P with MARCKS knockdown; however, the invasion was reduced by ∼70% in U373vIII with MARCKS knockdown (P < 0.001) compared with ∼40% decrease in invasion in U373P cells (P < 0.01). Thus, we observed a greater effect on invasion in MARCKS knockdown in the U373vIII cells compared with U373P cells.
MARCKS is phosphorylated by EGF stimulation and constitutively phosphorylated by EGFRvIII through PKC. EGF stimulation of EGFR can activate PKC ( 23), with PKC activation leading to MARCKS phosphorylation ( 14). We found that EGF stimulation of U373P cells increases MARCKS phosphorylation ( Fig. 5A ), which could be inhibited by administration of AG1478, a EGFR tyrosine kinase inhibitor. Inhibition of MARCKS phosphorylation was even more effectively inhibited by addition of RO-32-0432, a PKC-α inhibitor, followed by EGF stimulation. Collectively, these results suggest that MARCKS is phosphorylated following EGFR activation via activation of the PKC-α pathway.
Because EGFRvIII is constitutively activated, we determined if EGFRvIII expression results in constitutive MARCKS phosphorylation. Figure 5A shows that the levels of phosphorylated MARCKS are higher in U373vIII cells compared with U373P cells. When U373vIII cells were grown in the presence of the EGFRvIII inhibitor AG1478 for 24 h, the levels of phosphorylated MARCKS significantly decreased. In addition, incubation of U373vIII cells for 3 days with AG1478 led to decreased expression levels of MARCKS ( Fig. 5B).
To determine if EGFRvIII phosphorylation of MARCKS occurs through PKC-α, U373vIII cells were grown in the presence of the PKC-α inhibitor RO-32-0432 for 24 h. Phosphorylated MARCKS was undetectable after this incubation ( Fig. 5A), suggesting that EGFRvIII expression leads to increased levels of constitutively phosphorylated MARCKS through PKC-α.
EGFRvIII and MARCKS regulate the levels of PKC-α. Transfection of U373 with EGFRvIII resulted in an increase in PKC-α levels that were reduced following tyrosine kinase inhibition ( Fig. 5C). In addition, knockdown of MARCKS in U373vIII cells resulted in a decrease in PKC-α levels ( Fig. 5D). Elevated levels of PKC have been shown to increase invasion of glioma cells ( 13). To assess whether the increased invasion of EGFRvIII-expressing cells was due to increased levels of activated PKC, we performed a Matrigel invasion assay on cells treated with the PKC-α inhibitor RO-32-0432 ( Fig. 5E). PKC inhibition of U373 EGFRvIII cells decreased invasion to approximately the same levels of U373 parental cells treated with the PKC inhibitor; however, U373EGFRvIII invasion was decreased ∼17 times (P < 0.01) compared with ∼11 times decrease in U373 parental invasion (P < 0.001). Taken together, our results suggest that the invasive phenotype conferred by EGFRvIII may act in part through MARCKS and PKC.
MARCKS is predominantly localized to the cytosol in U373vIII cells and to the membrane in U373P cells. To determine the subcellular localization of MARCKS in U373P and EGFRvIII cells, cell fractionation was carried out under various conditions ( Fig. 6A ). Transferrin and glyceraldehyde-3-phosphate dehydrogenase were used to confirm purification of the membrane and cytosol fractions, respectively. MARCKS was found almost exclusively in the membrane fraction in U373P cells. After EGF treatment, ∼90% of total MARCKS was localized to the cytosol ( Fig. 6B). Conversely, pretreatment of U373P cells with AG1478 or RO-32-0432 before EGF stimulation prevented MARCKS from shifting to the cytosol. In contrast, in U373vIII cells, MARCKS was predominantly located in the cytosol, with only ∼34% of total MARCKS membrane bound ( Fig. 6B). Following 24 h of AG1478 or RO-32-0432 treatment, MARCKS was translocated from the cytosol to the membrane, where ∼49% or ∼62% of total MARCKS was found, respectively. In summary, the majority of MARCKS in U373P cells was localized to the membrane in EGF-unstimulated conditions or to the cytosol in EGF-stimulated conditions, whereas, in U373vIII cells, the majority of MARCKS was constitutively localized to the cytosol.
We show for the first time that MARCKS as well as its phosphorylated form is up-regulated in both wtEGFR- and EGFRvIII-expressing glioma cell lines, xenografts, and operative specimens. In addition, LCM on paraffin-embedded GBM operative specimens identified that both MARCKS and EGFR expression are increased in the infiltrative tumor periphery relative to the tumor center. These observations suggest that MARCKS plays a role in both wtEGFR- and EGFRvIII-mediated invasion. These findings have significant therapeutic implications, as amplification and overexpression of wtEGFR and EGFRvIII account for ∼50% of all GBM. Thus, the MARCKS pathway may be a common target in this subset of GBM. The purpose of this study, however, was to gain a better understanding of how EGFRvIII-MARCKS signaling contributes to glioma invasion. Additional studies are being carried out to investigate whether MARCKS plays an important role in wtEGFR-mediated glioma invasion.
Although several studies have shown that EGFRvIII plays an important role in tumor invasion ( 8), the role of MARCKS in invasion has not been established. Several groups have found that MARCKS is involved in myoblast cell adhesion and spreading ( 20– 22, 24, 25), which are part of the invasive process. We show for the first time that MARCKS is involved in glioma cell adhesion and spreading. Knockdown of MARCKS using siRNA in both U373P and U373vIII cells resulted in a significant decrease in adhesion to collagen and fibronectin but not to laminin. This suggests that cell adhesion on collagen and fibronectin is mediated through MARCKS-dependent pathways, whereas adhesion to laminin is mediated through MARCKS-independent pathways. Furthermore, our observation that the effect of MARCKS knockdown on cell attachment and spreading was similar in both U373P and U373vIII cells suggests that MARCKS-mediated cell attachment and spreading is independent of EGFRvIII signaling. Our results are consistent with the findings of others that MARCKS knockdown decreases cell spreading ( 22). Similar to our observation that the role of MARCKS on adhesion is extracellular matrix–dependent, Disatnik and colleagues found that myoblast cells expressing mutant MARCKS failed to spread on fibronectin but were able to spread on laminin in an α5β3 integrin–dependent manner ( 21). Because aberrations in integrin expression are associated with gliomas, we are investigating the interactions between MARCKS and specific integrin isoforms that are responsible for glioma cell adhesion and spreading.
We found that MARCKS knockdown decreases cell invasion in U373P and U373vIII cells. This effect, however, was more pronounced in U373vIII cells compared with U373P cells, suggesting that MARCKS contributes to EGFRvIII-mediated cell invasion. Although one group has shown that hyperphosphorylation of MARCKS following phorbol 12-myristate 13-acetate stimulation led to a reduction in invasion of bladder carcinoma cells ( 26), this effect may be tumor type specific because it has been shown that phorbol 12-myristate 13-acetate stimulation (and presumably MARCKS phosphorylation) of glioma cells leads to increased cell invasion ( 27). The role of MARCKS in invasion has not been studied before; thus, we set out to decipher the mechanism of MARCKS-mediated invasion.
Several isoforms of PKC have been found to phosphorylate MARCKS, including PKC-α and PKC-ε ( 28), both overexpressed and implicated in glioma invasion ( 29, 30). To determine if the activation of PKC downstream of EGFRvIII/MARCKS is important to cell invasion in U373vIII cells, we blocked PKC-α activity and found ∼17 times decrease in U373vIII cell invasion through Matrigel compared with ∼11 times decrease in U373P cell invasion. These results suggest that PKC is playing an important role in EGFRvIII-mediated invasion. In addition, given that invasion was decreased more following PKC inhibition compared with MARCKS knockdown in both U373vIII and U373P cell lines suggests that PKC is upstream/downstream of other signaling pathways contributing to glioma invasion, not just the MARCKS signaling pathway.
Current investigations include further deciphering how MARCKS and PKC-α affect EGFRvIII-mediated invasion. Toward this, our results show that EGFRvIII and MARCKS increase expression of PKC-α, suggestive of a positive feedback loop between MARCKS and PKC-α. We are currently determining the effects of EGFRvIII and MARCKS on PKC-α kinase activity. It has been reported that overexpression and phosphorylation of MARCKS increases expression of PKC-α in neuroblastoma cells through phospholipase D ( 31, 32). Further studies are being carried out to determine if a similar mechanism is occurring in glioma cells.
In conclusion, we show for the first time that MARCKS is overexpressed and phosphorylated in EGFR/EGFRvIII-expressing human GBM. Mechanistically, our data suggest that EGFRvIII regulates MARCKS expression and phosphorylation through PKC-α and both EGFRvIII and MARCKS regulate PKC, which in turn regulates invasion.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: Sick Kids Hospital and OSOTF scholarships (J. Micallef), American Brain Tumor Association (J. Mukherjee), and National Cancer Institute of Canada (A. Guha).
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.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received December 30, 2008.
- Revision received June 28, 2009.
- Accepted July 27, 2009.
- ©2009 American Association for Cancer Research.