Cool-1-Mediated Inhibition of c-Cbl Modulates Multiple Critical Properties of Glioblastomas, Including the Ability to Generate Tumors In Vivo

Authors

  • Brett M. Stevens,

    1. Department of Biomedical Genetics, University of Rochester Medical Center, Rochester, New York, USA
    Current affiliation:
    1. Current address: Brett M. Stevens, Ph.D., University of Colorado Anschutz Medical Campus Division of Hematology, Hematologic Malignancies and Stem Cell Transplantation 12700 E 19th Avenue Rm 9122 RC2, MS B170 Aurora CO 80045
    Search for more papers by this author
  • Christopher J. Folts,

    1. Department of Biomedical Genetics, University of Rochester Medical Center, Rochester, New York, USA
    Search for more papers by this author
  • Wanchang Cui,

    1. Department of Biomedical Genetics, University of Rochester Medical Center, Rochester, New York, USA
    2. Department of Radiation Oncology, University of Maryland School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Addie L. Bardin,

    1. Department of Biomedical Genetics, University of Rochester Medical Center, Rochester, New York, USA
    Search for more papers by this author
  • Kevin Walter,

    1. Department of Neurosurgery, University of Rochester Medical Center, Rochester, New York, USA
    Search for more papers by this author
  • Eleanor Carson-Walter,

    1. Department of Neurosurgery, University of Rochester Medical Center, Rochester, New York, USA
    Search for more papers by this author
  • Angelo Vescovi,

    1. Department of Biotechnology and Biosciences, University of Milano-Bicocca and StemGen SpA, Milan, Italy
    Search for more papers by this author
  • Mark Noble

    Corresponding author
    1. Department of Biomedical Genetics, University of Rochester Medical Center, Rochester, New York, USA
    • Correspondence: Mark Noble, Ph.D., University of Rochester Medical Center, 601 Elmwood Avenue, Box 633, Rochester, New York 14642, USA. Telephone: +1-585 273-1448; Fax: +1-585 273-1450; e-mail: Mark_noble@urmc.rochester.edu

    Search for more papers by this author

Abstract

We discovered that glioblastoma (GBM) cells use Cool-1/β-pix to inhibit normal activation of the c-Cbl ubiquitin ligase via the redox/Fyn/c-Cbl pathway and that c-Cbl inhibition is critical for GBM cell function. Restoring normal c-Cbl activity by Cool-1 knockdown in vitro reduced GBM cell division, almost eliminated generation of adhesion-independent spheroids, reduced the representation of cells expressing antigens thought to identify tumor initiating cells (TICs), reduced levels of several proteins of critical importance in TIC function (such as Notch-1 and Sox2), and increased sensitivity to BCNU (carmustine) and temozolomide (TMZ). In vivo, Cool-1 knockdown greatly suppressed the ability of GBM cells to generate tumors, an outcome that was c-Cbl dependent. In contrast, Cool-1 knockdown did not reduce division or increase BCNU or TMZ sensitivity in primary glial progenitor cells and Cool-1/c-Cbl complexes were not found in normal brain tissue. Our studies provide the first evidence that Cool-1 may be critical in the biology of human tumors, that suppression of c-Cbl by Cool-1 may be critical for generation of at least a subset of GBMs and offer a novel target that appears to be selectively necessary for TIC function and modulates chemoresistance in GBM cells. Targeting such proteins that inhibit c-Cbl offers potentially attractive opportunities for therapeutic development. Stem Cells 2014;32:1124–1135

Introduction

Glioblastomas (GBMs)—the most malignant of brain tumors—are among the most deadly cancers known and express such a plethora of mechanisms for evading therapy that it has proven extremely difficult to identify means of treating them. Resistance to chemotherapeutic agents, radiation, and other cell death inducers, the ability to colonize tissue well-removed from the tumor's primary origin, and continued cell division all represent challenging obstacles to treatment.

Identification of promising new strategies for GBM treatment faces the formidable obstacles of attacking targets that control vital aspects of GBM biology, including tumor initiation and sensitivity to therapeutic agents, without also damaging normal precursor cells of the central nervous system (CNS). This is particularly challenging due to the vulnerability of glial precursor cells of the CNS to many therapeutic agents of importance in cancer treatment [1, 2], including BCNU/carmustine (which is used in treatment of gliomas, multiple myeloma, and lymphoma), cisplatin, cytarabine, and 5-fluorouracil, all of which are more toxic for glial progenitor cells and oligodendrocytes than they are for cancer cells.

We recently reported, in work on basal-like breast cancer (BLBC) cells, that the redox/Fyn/c-Cbl (RFC) pathway may offer an attractive new target for discovering cancer-specific therapeutic approaches [3]. In this pathway, which was originally discovered in oligodendrocyte/type-2 astrocyte progenitor cells (also referred to as oligodendrocyte precursor cells, here abbreviated as O-2A/OPCs), cellular oxidation causes sequential activation of Fyn kinase and c-Cbl ubiquitin ligase [4] (Supporting Information Fig. S1). This leads to ubiquitylation and degradation of c-Cbl's target proteins, including such important cancer targets as the epidermal growth factor receptor (EGFR). In BLBC cells, in contrast, c-Cbl activation is prevented by Cdc42. Genetic or pharmacological Cdc42 inhibition rendered BLBC cells vulnerable to pro-oxidative stimuli, and in particular to estrogen-receptor independent effects of tamoxifen in vitro and in vivo. Moreover, genetic knockdown of Cdc42 reduced BLBC cells' tumor-initiating capacity. In contrast, Cdc42 inhibition did not increase tamoxifen sensitivity of nontumorigenic MCF10A mammary epithelial cells.

Our findings on BLBCs raise the question as to whether suppression of c-Cbl activity by Cdc42 contributes to TIC function in other types of cancer, and our studies in GBM cells now indicate that suppression of c-Cbl function via expression of inhibitory proteins may be such a critical component of TIC biology that cancer cells have multiple means of achieving this outcome. In contrast with BLBC cells, GBM cells inhibit the RFC pathway through a distinct mechanism in which c-Cbl is directly bound by Cool-1/β-Pix. This inhibition controls multiple critical aspects of GBM cell biology, appears to be required for tumor generation and offers a new and selective candidate target for treatment of GBMs.

Materials and Methods

Detailed methods are provided in Supporting Information.

Tissue Culture

GBM cells were derived from GBM specimens in conditions shown to enhance the isolation of TICs and are from the same cluster of GBM TIC lines described in other publications [5, 6]. O-2A/OPCs were isolated and analyzed as previously described [4].

Western Blotting and Immunoprecipitation

Anti-Fyn polyclonal (Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com), anti-EGFR polyclonal (Cell Signaling Technology, Beverly, MA, http://www.cellsignal.com), anti-c-Cbl, anti-P-cbl, anti-P-tyr (BD Biosciences, San Diego, CA, http://www.bdbiosciences.com), anti-Cool-1(Millipore, Billerica, MA, http://www.millipore.com) antibodies were analyzed as previously [4]. For coimmunoprecipitation assays, anti-c-Cbl monoclonal antibody (BD Pharmingen, San Diego, CA, http://www.bdbiosciences.com/index[lowen]us.shtml) was added into precleared cell lysates, and precipitates were resolved on SDS-PAGE gels and probed with anti-Cool-1 antibody.

Fyn Kinase Assay

Fyn kinase activity was quantified using the Universal Tyrosine Kinase Assay Kit (Takara, Otsu, Japan, http://www.takara.co.jp), as previously [4].

Lentiviral-Mediated shRNA Gene Knockdown

Knockdown constructs (Open Biosystems) were based on constructs designed and used by the Broad Institute, expressed in viruses as described in Supporting Information. Twenty-four hours postinfection, cells were selected in growth medium + 20 ng/mL puromycin.

Cytotoxicity Assays

Cytotoxicity was assessed using Alamar Blue Assays.

In-Cell Western Blotting

To measure c-Cbl phosphorylation across multiple time points and multiple doses simultaneously, a modified FACE-maker ELISA assay was used (Invitrogen, Carlsbad, CA, http://www.invitrogen.com).

Flow Cytometry

For cell cycle analysis, GBM cells expressing Cool-1 shRNA or Scrambled shRNA were stained with 4',6-diamidino-2-phenylindole (DAPI) and analyzed on a BD-FACS CANTO (BD Biosciences, San Diego, CA, http://www.bdbiosciences.com). Live cells expressing Cool-1 shRNA or scrambled shRNA were stained with CD133-phycoerythrin (PE) (Miltenyi Biotec, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com) and with CD15-APC (Miltenyi Biotec) followed by of DAPI.

Tumorigenicity Analysis

Tumorigenicity was evaluated by intracranial transplantation of GBM cells expressing luciferase and confirmed knockdown of Cool-1 (or Scrambled shRNA expression). Tumor size was monitored using IVIS 100 Imaging station after i.p. luciferin injection.

Results

BCNU Induces C-Cbl Activation and Reductions in EGFR Levels in Normal Glial Progenitor Cells but Not in GBM Cells

To examine function of the RFC pathway in GBMs, we initially focused on the effects of BCNU, which has been used in treatment of gliomas and other cancers. While BCNU can create DNA crosslinks, it is also an inhibitor of glutathione reductase [7], which is required for maintaining normal levels of reduced glutathione. BCNU exposure thus also causes increases in intracellular oxidative status. As a positive control, we examined effects of BCNU exposure on O-2A/OPCs, in which the RFC pathway was originally discovered[8].

Direct comparison between O-2A/OPCs and GBM cells revealed that these cells differ in the ability of BCNU to cause c-Cbl activation and reductions in levels of EGFR (a c-Cbl target frequently overexpressed in GBMs and that is also a therapeutic target in GBMs and other cancers (e.g., [9, 10])). Exposure of O-2A/OPCs to 5–15 μM BCNU (Fig. 1) caused twofold to threefold increases in Fyn kinase activity (Supporting Information Fig. S3), dose-dependent twofold to fivefold increases in c-Cbl phosphorylation (Fig. 1A, 1B), and decreases in EGFR levels in cells exposed to EGF (Fig. 1D, 1E). In contrast, no such outcomes were observed in GBM27, GBM13, and GBM10 cells (which generate isotypic tumors when transplanted into immunocompromised mice [5, 6]) (Fig. 1; Supporting Information Fig. S2). Even when exposed for 24 hours to several-fold higher concentrations (25–100 μM) of BCNU, GBM27, GBM13, and GBM10 cells showed no increases in c-Cbl phosphorylation or decreases in EGFR levels (Fig. 1A, 1B, 1D, 1E; Supporting Information Fig. S2).

Figure 1.

The redox/Fyn/c-Cbl pathway is activated by BCNU in O-2A/OPCs but not in GBM cells.(A): BCNU treatment in O-2A/OPCs leads to increases in phosphorylated c-Cbl across time as measured by Western blotting in (A) and quantified in (B). In GBM cells there was a failure to phosphorylate c-Cbl in response to BCNU across multiple time points and doses as measured by western blotting in (A) and quantified in (C). While O-2A/OPCs showed a clear decrease in EGFR levels in response to BCNU (D, E), in GBM cells BCNU failed to cause a decrease in levels of EGFR (D, F). Results are expressed as mean ± SD (*, p < .05) and represent three to five independent experimental replicates. Abbreviations: BCNU, carmustine; EGFR, epidermal growth factor receptor; GBM, glioblastoma; OPC, oligodendrocyte precursor cell.

The failure to cause c-Cbl activation and decreases in EGFR levels in GBM cells was not due to a failure of BCNU to enter cells, to render them more oxidized, or to activate Fyn kinase. BCNU caused a twofold increase in Fyn kinase activity in GBM27 cells, which was blocked by the Src-family kinase inhibitor PP1 (Supporting Information Fig. S3). Exposure to 100 μM BCNU decreased levels of total glutathione (GSH) by ∼50% and increased the proportion of oxidized glutathione (GSSG) in GBM27 cells from levels almost too low to detect to up 50% of total GSH, a change in the GSH/GSSG ratio largely prevented by pretreatment with 5 mM N-acetyl cysteine (NAC) (Supporting Information Fig. S4).

BCNU-Induced Activation of the Redox/Fyn/c-Cbl Pathway in GBM Cells is Prevented due to Sequestration of c-Cbl by Cool-1

As BCNU causes oxidation and Fyn activation in GBM cells, the lack of BCNU-induced activation of c-Cbl and decreases in levels of EGFR must be due to changes at other points in the RFC pathway. In BLBC cells, Cdc42 inhibits c-Cbl function [3, 11, 12]. Studies on experimental transformation of NIH3T3 cells by v-src revealed that Cool-1 also can inhibit function of Cbl-b [13], a member of the Cbl family frequently overexpressed in breast cancers, but other studies reported weak-to-no binding of Cool-1 to c-Cbl [14]. Unlike cdc42, moreover, there is no published information on functional contributions of Cool-1 to naturally occurring tumors, either in humans or experimental animals.

Comparison of lysates of human glioma specimens with normal brain suggested that Cool-1, but not Cdc42, was of potential interest for closer examination. In lysates of multiple GBM biopsies, we saw no changes in cdc42 protein content as compared with normal brain (Supporting Information Fig. S5). In contrast, examination of Cool-1 expression revealed a striking shift between normal brain and tumor in isoform expression and phosphorylation (Fig. 2). In normal cortex, levels of isoforms 1/6 were greatest, with isoforms four being lower, and the lowest expression being for isoforms 2/3/5. In contrast, in all GBM lysates examined, isoforms 2/3/5 showed the highest levels of expression with relatively little expression of isoforms 1/6 or 4. Similar reductions in isoforms 1/6 and 4 were seen in four-of-four anaplastic astrocytoma lysates, and two-of-four anaplastic astrocytomas showed increases in isoforms 2/3/5 (Fig. 2A, 2B).

Figure 2.

Increased Cool-1 phosphorylation and physical interaction with c-Cbl occurs specifically in GBM cells. (A): Western blot analysis revealed increased expression of specific isoforms (2/3) in GBM patient tumor lysates in comparison to lysates of normal human cortex. The increased expression is significant for these isoforms over cortex samples (B): Results are expressed as Mean ± S.D. *, p < .05. (C): Complexes between Cool-1 and c-Cbl were found in lysates of GBM biopsy specimens by immunoprecipitation with anti-c-Cbl antibodies and probing of gels with anti-Cool-1 antibodies. Such complexes were not found in lysates of normal cortical tissue. (D): Cultured GBM cells contained c-Cbl/Cool-1 complexes, as revealed by IP with anti-c-Cbl antibodies and analysis immunoblotting (IB) with anti-Cool-1 antibodies. (E): Lysates from GBM samples, but not from normal cortex, also showed increased phosphorylation of Cool-1, as determined by IP with anti-Cool-1 antibodies and probing of lysates with anti-phospho-tyrosine antibodies. (F): Cultured O-2A/OPCs do not contain c-Cbl/Cool-1 complexes as revealed by IP with anti-c-Cbl antibodies and IB with anti-Cool-1 antibodies. Abbreviations: GBM, glioblastoma; IP, immunoprecipitation; OPCs, oligodendrocyte precursor cells; WB, Western blot; WCL, whole cell lysate.

Further analysis of Cool-1 in GBM lysates showed complex formation with c-Cbl and increased Cool-1 phosphorylation. Coimmunoprecipitation of c-Cbl and Cool-1 was seen in all GBM samples but not in normal human cortex (Fig. 2C) or white matter (Supporting Information Fig. S6) lysates. In addition, analysis of lysates from five GBM cell lines demonstrated Cool-1 was contained in a complex with c-Cbl, as shown by immunoprecipitation of c-Cbl and Western blot analysis of Cool-1 (Fig. 2D; Supporting Information Fig. S6). Coimmunoprecipitation of Cool-1 with c-Cbl was not dependent on the presence of EGF (Supporting Information Fig. S6). In contrast, c-Cbl immunoprecipitation from lysates of cultured O-2A/OPCs revealed no evidence of Cool-1/c-Cbl complexes (Fig. 2F), in agreement with the lack of such complexes in normal gray matter and white matter lysates (Fig. 2C; Supporting Information Fig. S6). Marked increases in levels of phosphorylated Cool-1 also were seen in five separate GBM biopsies but not in normal cortex (Fig. 2E).

We next found that short hairpin (shRNA)-mediated reductions in Cool-1 levels (Supporting Information Fig. S7), in multiple GBM cells, decreased basal levels of EGFR and enabled BCNU exposure to induce c-Cbl phosphorylation and further decreases in EGFR (Fig. 3A–3C; Supporting Information Fig. S8). Exposure to 100 μM BCNU now caused a time-dependent increase in c-Cbl phosphorylation, with more than twofold increases in phosphorylation occurring over 180 minutes of exposure, as determined by the ratio of phosphorylated to total c-Cbl (Fig. 3A, 3B; Supporting Information Fig. S8). Moreover, exposure to 100 μM BCNU caused an approximately 50% decrease in EGFR levels over 6 hours (Fig. 3C, 3D for GBM27 cells, Supporting Information Fig. S8 for GBM10, GBM13, and GBM21 cells). In contrast, expression of scrambled shRNA had no effects on BCNU-induced c-Cbl phosphorylation or reductions in EGFR levels. Reductions in EGFR levels were c-Cbl-dependent, as coexpression of shRNA for c-Cbl in Cool-1 knockdown cells eliminated effects of BCNU on EGFR levels (Fig. 3C, 3D). Cool-1 knockdown had no apparent effect on cell viability.

Figure 3.

Genetic knockdown of Cool-1 with shRNA rescues BCNU-induced decreases in levels of EGFR. (A, B): BCNU treatment in Cool-1 knockdown cells caused time-dependent increases in c-Cbl phosphorylation and decreases in EGFR levels. (C): The decreases in EGFR were Src family dependent and were prevented by PP1 co-pretreatment (C; Supporting Information Fig. S7A). Decreases in EGFR levels were c-Cbl dependent (C, D), as further knockdown of c-Cbl in the Cool-1 knockdown cells caused a failure of EGFR to be decreased in response to BCNU. Reductions in EGFR levels were blocked by exposure to the PP1 or to NAC, as predicted if c-Cbl activation is regulated by the redox/Fyn/c-Cbl pathway [4]. Results in (B, D) are expressed as mean ± SD *, p < .05, a is p < .05 compared to scrambled shRNA, b is p < .05 compared to Cool-1 shRNA and represent quantification of at least three independent experimental replicates. Abbreviations: BCNU, carmustine; EGFR, epidermal growth factor receptor; GBM, glioblastoma; NAC, N-acetyl cysteine; shRNA, short hairpin RNA.

As predicted by our original analyses on the RFC pathway [4], BCNU-induced decreases in EGFR levels were prevented by exposure of Cool-1 knockdown GBM cells to NAC or to PP1, or if BCNU levels were reduced (Fig. 3D; Supporting Information Fig. S9).

Cool-1 Knockdown Alters Multiple Properties of GBM Cells but Does Not Alter Division or Sensitivity to BCNU and Temozolomide of O-2A/OPCs

Along with increasing the sensitivity of GBM cells to BCNU as an RFC pathway activator, we also increased sensitivity to BCNU's cytotoxic effects (Fig. 4A; Supporting Information Fig. S10). For example, in GBM10 cells, exposure to 50 μM BCNU for 5 days killed approximately 56% of cells expressing Cool-1 shRNA, as compared with approximately 20% of cells infected with control (scr) vectors. In a shorter 24-hour exposure in GBM27 cells (Supporting Information Fig. S10), significant increases in sensitivity were only seen with exposure levels above 100 μM, concentrations seen with surgical implantation of Gliadel® wafers [15]).

Figure 4.

Cool-1 knockdown increases vulnerability to BCNU and temozolomide (TMZ), and alters cell cycle parameters, in GBM cells but not primary oligodendrocyte precursor cells (O-2A/OPCs). (A): Knockdown of Cool-1 in GBM cells leads to increased sensitivity to BCNU compared to scrambled shRNA cells at 5 days. (B): Cool-1 knockdown cells from multiple GBM lines exposed to TMZ showed decreased survival compared to scrambled shRNA control cells at 5 days. (C): Cool-1 knockdown caused an increase in the population of cells in G0/G1 phase and decreases in the percentage of the population in G2M phase for multiple GBM lines with no significant change seen in O-2A/OPCs. (D): Knockdown of Cool-1 decreased the percentage of cells positive for Ki-67 in comparison to scrambled shRNA cells only in GBM cells with no change in Ki-67 for O-2A/OPC cells. (E): Knockdown of Cool-1 leads to a decrease in proliferation in comparison to scrambled shRNA in multiple GBM lines with no significant change in proliferation of O-2A/OPC cells. (F): Cool-1 knockdown cells exhibited significantly decreased migration in a Boyden chamber assay and wound healing assay in comparison to control cells over 24 hours. (G, H): Cool-1 knockdown in O-2A/OPCs does not change sensitivity to BCNU (G) or TMZ (H). Results in (C–F) are expressed as mean ± SD (*, p < .05.) and represent three to five experimental replicates. Results in (A), (B), (G), and (H) are expressed as mean ± SEM (*, p < .05) for three experimental replicates with three samples at each dose per replicate. Abbreviations: BCNU, carmustine; DAPI, 4',6-diamidino-2-phenylindole; GBM, glioblastoma; shRNA, short hairpin RNA.

More importantly, Cool-1 knockdown also increased sensitivity to temozolomide (TMZ, a more critical agent than BCNU in GBM treatment) at multiple exposure levels in GBM10, GBM21, and GBM27 cells. Unlike BCNU, TMZ does not have known pro-oxidant activities, as also shown by the failure of TMZ to cause decreases in glutathione levels (Supporting Information Fig. S4). In contrast, shRNA-mediated Cdc42 knockdown had no effects on BCNU or TMZ sensitivity (Supporting Information Fig. S11).

Cool-1 knockdown also modulated multiple other GBM cell functions, independent of sensitivity to BCNU or TMZ. Proliferation of Cool-1 knockdown GBM10, GBM11, GBM13, and GBM27 cells was significantly decreased as measured by cell growth over 10 days (Fig. 4E) and division was decreased as indicated by Ki-67 staining (Fig. 4D). A significant shift in cell cycle parameters also was seen by fluorescence-activated cell sorter (FACS) analysis (Fig. 4C; Supporting Information Fig. S12), which showed in GBM27 cells, for example, an approximately 40% increase in the G0/G1 population versus cells expressing scrambled shRNA constructs and an approximately 55% decrease in the G2 population in Cool-1 knockdown cells versus Scr shRNA constructs (p < .05). Cool-1 knockdown, but not expression of Scr constructs, also suppressed GBM cell migration, an aspect of these tumors of importance due to their ability to spread throughout the CNS. Analysis in Boyden chambers (Fig. 4D) showed decreased migration in GBM10 and GBM27 cells of 90% and 50%, respectively, with GBM11 cells exhibiting a trend toward decreased migration.

Critically, Cool-1 knockdown had little effect on primary glial progenitor cells. Expression of Cool-1 shRNA in O-2A/OPCs did not alter cell cycle parameters (Fig. 4C) or division (as determined by Ki67 staining, Fig. 4D). Increases in cell number over 10 days were also not significantly decreased (Fig. 4E). The one parameter showing a marked decrease was in O-2A/OPC migration (Supporting Information Fig. S12), as predicted by the role of Cool-1 in modulating cytoskeletal function [13, 16]. Perhaps most importantly, Cool-1 knockdown did not alter the sensitivity of O-2A/OPCs to BCNU or TMZ (Fig. 4G, 4H).

Cool-1 Knockdown Modifies Multiple Aspects of TIC Function In Vitro

One of the useful in vitro measures of TIC function is the ability to generate multicellular spheroids when grown on nonadhesive surfaces, a function that was highly Cool-1 dependent. Cultures of GBM10, 11, 13, and 27 cells expressing scrambled shRNA constructs for Cool-1 generated spheroids that increased in size continuously for more than 8 days. In contrast, cells expressing shRNA for Cool-1 showed severely reduced spheroid generation (Fig. 5A; Supporting Information Fig. S13). GBM27 cells expressing Cool-1 shRNA generated approximately 90% fewer spheres per well than cells expressing Scr shRNA, with similar outcomes for GBM10, GBM11, and GBM13 cells (Fig. 5A). Moreover, as illustrated for GBM27 cells (Supporting Information Fig. S13) spheroids generated from Cool-1 knockdown cells were much smaller than those generated from Scr-expressing cells.

Figure 5.

Cool-1 knockdown modulates aspects of GBM biology thought to be relevant to the capacity to initiate tumors. (A): Cool-1 knockdown decreases sphere growth in suspension cultures of multiple GBM lines, quantified in (A), and in soft agar assays (representative images in Supporting Information Fig. S12). (B): Cool-1 knockdown decreases the protein levels of Notch-1 and Sox2 in multiple GBM lines as measured via immunoblot (B) and quantified in (C) and (D). β-Catenin levels also were reduced in two third cell lines examined. Cool-1 knockdown led to significant decreases in the percentage of CD15 of multiple GBM lines (E, F) and of CD133 expression (F) and of CD15/CD133 double positive cells as measured by flow cyometry. Decreases in the CD15 positive population in GBM27 cells were partially blocked by c-Cbl knockdown in cells in which Cool-1 also had been knocked down (E). Results are expressed as mean ± S.D. *, p < .05 and represent four to six independent experiments. Abbreviations: GBM, glioblastoma; shRNA, short hairpin RNA.

We next found that Cool-1 knockdown caused reductions in levels of Notch-1 and β-catenin, two c-Cbl targets [17, 18] known to be a critical regulators of TIC function in GBMs [19-21], and also caused decreases in levels of Sox2, another protein required for GBM tumor initiation [22-25]. Cool-1 knockdown caused decreases in Notch 1 and Sox2 protein levels in 3/3 GBM cell lines (Fig. 5B–5D). Levels of β-catenin also were markedly reduced in two out of three of these cell lines (Fig. 5B).

Cool-1 knockdown also reduced expression of CD133 and CD15, two antigens reported to identify glioma cells with tumor initiation capacity [5, 26]. In GBM27 cells, Cool-1 knockdown caused a >80% reduction in prevalence of CD133+ cells, an approximately 50% reduction in CD15+ cells and a >85% reduction in CD133/CD15 double-positive cells (Fig. 5E, 5F). As CD15 has been reported to be more effective in defining cells with tumor initiation capacity than CD133 [26], we also examined its expression in other cell lines and found similar reductions in CD15 expression in Cool-1 knockdown GBM10 and GBM21 cells (Fig. 5F; Supporting Information Fig. S14).

Expression of c-Cbl shRNA constructs in Cool-1 knockdown GBM27 cells restored CD15 expression (Fig. 5E) and significantly restored neurosphere-forming ability (Supporting Information Fig. S13), supporting the importance of c-Cbl as a contributor to effects of Cool-1 knockdown.

Reduction of Cool-1 Expression Prevents Tumor Generation

The definitive test of TIC function is to examine the capacity of cells to generate tumors, and such analyses demonstrated Cool-1 knockdown was very effective at suppressing this aspect of GBM biology. GBM21 and GBM27 cells containing an established Cool-1 knockdown or expressing Scr shRNA were transplanted intracranially into the right striatum of NOD/SCID mice. Tumor size and progression was monitored using luciferase-based whole animal imaging of cells expressing the firefly luciferase gene.

Cool-1 knockdown reduced the capacity of GBM cells to establish tumors to such an extent as to raise the possibility of total inhibition of TIC function (Table 1, Fig. 6). Readily detectable tumors developed in approximately 60% of mice within 8 weeks after transplantation of as few as 10,000 GBM27 Scr construct-expressing cells. In contrast, in Cool-1 knockdown cells, no tumors were generated following transplantation of 10,000 or even 50,000 cells (Fig. 6A–6E). Tumors only arose in rare animals transplanted with 200,000 (5/30 mice) or 500,000 Cool-1 knockdown cells (1/5 mice), with the other mice in these groups remaining tumor-free for at least 150 days. In GBM21-transplanted mice, 200,000 Cool-1 knockdown cells did not cause any tumors (Table 1), versus 2/5 animals developing tumors when injected with Scr shRNA-expressing cells. Similar outcomes were seen with two different shRNA constructs. The lack of tumor initiation was not due to a lack of viable cells, as transplanted cells were readily detected 2 weeks post-transplantation (Supporting Information Fig. S15A).

Table 1. Cool-1 knockdown decreases tumor initiation in a c-Cbl dependent manner in vivo
Cell numberGBM 27GBM 27GBM 27GBM 27GBM 21GBM 21GBM 21
Scrambled shRNACool-1 shRNA # 1Cool-1 shRNA #2Cool-1 & c-Cbl shRNAScrambled shRNACool-1 shRNA # 1Cool-1 shRNA#2
  1. Cool-1 knockdown cells derived from two different parental GBM cell lines and expressing two different shRNA constructs each were xenografted into immune compromised mice and showed markedly decreased ability to generate tumors. Tumor initiation was rescued by co-knockdown of c-Cbl and Cool-1, as shown for GBM27 cells.

  2. Abbreviations: GBM, glioblastoma; shRNA, short hairpin RNA.

10,0003/50/50/53/5  0/5
50,0003/50/50/54/5  0/5
200,00010/185/303/57/83/50/50/5
500,0005/51/5 5/54/42/5 
Figure 6.

Cool-1 knockdown decreases tumor initiation and tumor growth in vivo. (A–D): Cool-1 knockdown cells xenografted into immune compromised mice showed a markedly decreased ability to generate tumors (Table 1). (A): H&E stained brain sections from mice transplanted with Cool-1 knockdown cells and Scr-shRNA-expressing cells, sacrificed at the same time point. (B): Increased magnification views of tumor-bearing section from (A). (C): Analysis of brains of mice transplanted similarly as for (A) but sacrificed 6 weeks after transplantation and labeled with antibody against human mitochondria (green) and DAPI. The higher magnification images (D) of sections from tumor-bearing animals show extensive tumor cell migration and infiltration. (E, F): Even in animals in which Cool-1 knockdown cells generated tumors (5/30 Cool-1 shRNA transplanted mice vs. 10/18 mice transplanted with scrambled shRNA-expressing cells), we observed increased survival, presumably due to a slower rate of initial tumor growth and/or initiation (as also indicated by smaller tumor sizes at 4 weeks post-transplantation (F). The red line in (E) indicates majority of mice transplanted with Cool-1 knockdown cells that showed no evidence of tumor generation even 150 days after transplantation. Abbreviations: DAPI, 4',6-diamidino-2-phenylindole; hMT, human mitochondria; shRNA, short hairpin RNA.

As further confirmation of the importance of Cool-1 knockdown in preventing tumor generation, those rare tumors that emerged all showed re-expression of Cool-1 and high levels of EGFR. Cool-1 expression was readily detectable by immunofluorescence (Supporting Information Fig. S15B) and Western blot analysis (Supporting Information Fig. S15D) despite these tumors being <20% the size of tumors arising in mice injected with Scr shRNA-expressing cells (Fig. 6F; Supporting Information Fig. S15C). EGFR levels in these rare tumors were also similar to tumors arising from Scr shRNA-expressing cells (Supporting Information Fig. S15E). Thus, generation of tumors in mice transplanted with Cool-1 knockdown cells seems most likely dependent on a subset of cells in which Cool-1 was expressed. Consistent with the presence of a smaller number of cells capable of tumor generation, in mice in which Cool-1 shRNA cells generated tumors the average time to death was increased by almost 60%, from 72 days in animals injected with scrambled shRNA-containing cells to 114 days (Fig. 6E).

That the effects of Cool-1 are mediated by c-Cbl was tested by decreasing c-Cbl levels in GBM cells first treated to establish a Cool-1 knockdown. As predicted by our other observations, mice engrafted with cells coexpressing Cool-1 and c-Cbl shRNAs readily generated tumors. We found tumors in 60% of mice transplanted with 200,000 cells and in 100% of mice transplanted with 500,000 cells (2/3 and 5/5 animals, Table 1).

C-Cbl Targets and Cool-1 mRNA are Frequently Overexpressed in Human Tumors

Examination of existing databases suggests our findings may be applicable to a significant fraction of gliomas. Investigation of protein levels of c-Cbl targets in human tumor biopsies, using the Genome Cancer Atlas (TGCA), reveals increased expression of multiple such targets. Five receptors ubiquitylated by c-Cbl were examined (Notch1, VEGFR2, MET, EGFR, IGF1R) [27] and increased expression of 2/5 targets was found in 73% of samples (133/181), and 3/5 targets in 43% of samples (77/181).

Although the absence of information on c-Cbl or Cool-1 phosphorylation limits utility of these databases in interrogating c-Cbl function, the potential relevance of changes in Cool-1 expression was also supported when we plotted patient data for Cool-1 mRNA expression against EGFR expression in the TCGA protein array. 65% of samples (51/78) in which EGFR protein levels were increased above the mean also showed increased expression of Cool-1(ARHGEF7) mRNA. Additional analysis of Cool-1 mRNA expression in the TCGA AFFYU133A array dataset revealed increased transcript levels in 52% of samples (288/558). These changes were found almost entirely in tumors classified as proneural and classical GBM subtypes, but not in neural or mesenchymal subtypes.

Discussion

We found that GBM cells, but not normal CNS glial progenitors, inhibit RFC pathway activity via Cool-1 expression and that reducing Cool-1 levels and restoring normal c-Cbl activity has multiple beneficial outcomes. In vitro, these include reductions in GBM cell division, suppression of TIC function, reductions in levels of multiple proteins required for GBM initiation, and increases in BCNU and TMZ sensitivity. In vivo, Cool-1 knockdown appeared to eliminate the ability of GBM cells to generate tumors. Cool-1 knockdown did not, however, reduce division or increase BCNU or TMZ sensitivity in primary glial progenitor cells. In addition, analysis of normal brain (which contains large numbers of astrocytes) also showed an absence of Cool-1/c-Cbl complexes in normal CNS tissue, further indicating the tumor specificity of our findings. Thus, these studies identify a potential molecular intervention that suppresses TIC functions and increases sensitivity to relevant therapeutic agents while appearing to selectively affect cells from at least a subset of GBMs but not primary glial progenitors.

Our studies emphasize the potential importance of c-Cbl inhibitors in tumor biology, by offering (together with our studies on BLBC cells [3]) the first demonstrations that such inhibition is critical in TIC function in human tumors and by raising the possibility that such inhibition is so critical to TIC function that two different tumor types use distinct proteins to inhibit c-Cbl. Previous studies in lymphomas and other tumors demonstrated the importance of disruption of c-Cbl function as neoplastic stimuli, by direct mutation or by mutation of receptors to make them resistant to c-Cbl modulation (see, e.g., [28, 29] for review). In comparison, disruption of c-Cbl by inhibitory proteins has received much less attention. Previous studies on human tumors showed that Cdc42 knockdown in breast cancer cells reduces proliferation and migration and enables reductions in EGFR levels (apparently in a c-Cbl dependent manner [11, 12, 30]), but these studies did not examine tumor initiation or consequences of Cdc42 knockdown on sensitivity to therapeutic agents. In addition, recent studies on GBM cells isolated and grown in conditions that do not select for TICs raise the possibility that c-Cbl also may be inhibited by transglutaminase-2 [31]. These studies, which were focused on interactions between these two proteins provided evidence for coimmunoprecipitation of transglutaminase-2 and c-Cbl in overexpression studies but did not, however, examine effects of such inhibition on tumor generation.

Our studies appear to provide the first evidence that Cool-1 is both functionally important and inhibits c-Cbl in human tumors. Previous studies on breast cancer cells demonstrated binding of Cool-1 to Cbl-b (a member of the Cbl family frequently overexpressed in breast cancer cells [13]) but weak-to-no binding to c-Cbl [14]. Effects of Cool-1 expression on cancer cell function were not reported. The sole prior indications that disruption of activity of a Cbl family member by Cool-1 might be of relevance in transformation come from studies on v-Src-transformed NIH 3T3 cells. These studies showed that overexpressed Myc-tagged-Cool-1 coimmunoprecipitated with overexpressed Cbl-b and that overexpression of a phosphorylation-defective Cool-1 mutant in v-Src transformed cells reduced (by ∼50%) growth of these cells in saturation density assays and reduced (by ∼85%) the size of tumors generated in 9 days in flanks of nude mice, while expression of Cool-1 siRNA reduced colony generation of transformed cells by approximately 60% in soft agar assays. Further studies on Src-transformed cells revealed that Cool-1 phosphorylation status modulates their migration and invasive activity [16]. In addition, previous studies (using overexpression analyses) reported that a complex of Cdc42 and Cool-1 is required to inhibit c-Cbl [11, 30]. In contrast, our studies demonstrate the importance of Cool-1 and its inhibition of c-Cbl in at least a subset of naturally occurring GBMs, one of the most deadly human cancers. Our studies also demonstrate Cdc42 and Cool-1 regulate c-Cbl function independently, such that Cdc42 knockdown had no effect in GBM cells and, conversely, Cool-1 knockdown had no effect on BLBC cells [3].

One question raised by our studies concerns the extent to which outcomes of Cool-1 or Cdc42 knockdown are due to restoring c-Cbl function. Secondary knockdown of c-Cbl in GBM cells with a Cool-1 knockdown restored tumor initiation, increased neurosphere formation, partially restored levels of CD15 expression, and prevented BCNU-induced decreases in EGFR levels. In BLBC cells, c-Cbl knockdown in cells with a Cdc42 knockdown (or cells exposed to pharmacological Cdc42 inhibition) also restored tumor initiation, prevented tamoxifen-induced decreases in EGFR, and prevented response to tamoxifen in vitro and in vivo [3]. Thus, inhibition of c-Cbl activity appears to be central to understanding effects of Cool-1 and Cdc42 in GBM and BLBC cells, respectively.

Our studies also provide possible insights into why restoration of c-Cbl function may be so effective at suppressing tumor initiation in showing reductions in levels of several proteins critical in tumor generation. While Notch1 and activated β-catenin are known c-Cbl targets [17] [18], thus suggesting that restoring c-Cbl function would lead to reductions in levels of these regulators of glioma TIC function (e.g., [21, 32]), our findings that Sox2 is also reduced following Cool-1 knockdown is completely unexpected. In addition, CD133 appears to be critical for GBM TIC function [33] and also was reduced in Cool-1 knockdown cells. Such reductions in regulators of TIC function—along with the importance of such other c-Cbl targets as EGFR, c-Met and Ephrin A2 in TIC biology in GBMs [6, 34-37]—may help to explain why Cool-1 knockdown was so effective at preventing GBM formation. Indeed, as all tumors generated after transplantation of Cool-1 knockdown cells showed Cool-1 expression levels like those seen in nonmanipulated cells, our data is consistent with the hypothesis that Cool-1-mediated suppression of c-Cbl may even be essential for initiation of some forms of GBM.

The enhanced TMZ response in Cool-1 knockdown cells was also of interest both for biological reasons and due to the widespread use of this agent in GBM treatment. Enhanced responses to BCNU for GBM cells and to noncanonical activities of tamoxifen for BLBC cells [3] might be predicted based on known pro-oxidant activities of these agents and the role of the RFC pathway as a modulator of c-Cbl activity. TMZ, in contrast, is an alkylating/methylating agent not known for pro-oxidant activity, and exposure does not decrease glutathione levels in GBM cells (Supporting Information Fig. S4). While previous attempts to enhance response to TMZ have focused on O−6-methylguanine-DNA methyltransferase inhibition [38, 39] or cotreatment with trans-sodium crocetinate (an oxygen diffusion-enhancing compound [40]), our results raise the question of whether restoration of c-Cbl activity might more broadly increase sensitivity to TMZ (and potentially other therapeutic agents).

Another potentially intriguing aspect of restoration of c-Cbl function is the ability to also impact on non-TICs, a particularly important concept as it seems increasingly likely that cells in the putative non-TIC compartment also can maintain and initiate tumors. In BLBC cells, which—unlike the GBM cells we examined—were not grown in ways to enrich for TICs, we observed suppression of cell division, increased cell death, and suppression of tumor growth that affected far more cells than the likely 0.1%–1% representation of TICs in the populations studied [3].

The degree to which benefits obtained by Cool-1 knockdown are dependent upon restoring RFC pathway activity will be the subject of future research. It is intriguing to note, however, that RFC pathway inhibition appears to provide a novel means by which tumor cells can escape normal functional consequences of being oxidized. We found in BLBC cells that oxidized glutathione (GSSG) represented approximately 10% of the total glutathione pool, as compared with generally representing ≤2% of total glutathione in normal cells [41]. In BCNU-treated GBM27 cells, GSSG represents ∼40% of the total glutathione (Supporting Information Fig. S4). Such oxidative challenges are severely toxic for primary cells, yet both BLBC cells and GBM cells are not noticeably impaired by such GSSG abundance. As GSSG has been reported to activate Fyn [42], inhibition of RFC pathway-mediated activation of c-Cbl provides a means of escaping such an oxidative challenge.

Our studies raise the question of how broadly relevant protein-based inhibition of c-Cbl function might be in tumor biology. We previously found identical results on seven independent BLBC cell lines representing Basal B triple-negative cells, Basal A cells, and Basal A cells with Her2 amplification [3]. In our present studies, five independent GBM cell lines all yielded identical results, as did examination of biopsies from five different patients. In addition, TCGA analysis of five c-Cbl targets showed increased levels of at least two c-Cbl target proteins in 73% of cases and of at least three c-Cbl targets in 43% of cases. A more focused examination of tumors with increased EGFR protein expression showed increased expression of Cool-1 mRNA in 65% of samples, and in 52% of all GBM samples in the database. With the caveat that these databases provide no information on isoform changes and alterations in Cool-1 phosphorylation, such information nonetheless suggests a more extended examination of Cool-1 alterations in gliomas would be of interest. It was also potentially interesting that changes in Cool-1 expression were found almost entirely in proneural and classical GBM subtypes, while ∼60% of instances of increased levels of transglutaminase-2 were found in the mesenchymal GBM subtype [31], raising the possibility that different GBM types use different strategies for inhibiting c-Cbl but nonetheless converge on inhibiting this enzyme. Similarly, levels of Cdc42 in breast tumor lysates exceed by as much as 50-fold those seen in normal tissue from the same patients and elevated Cdc42 levels also are observed in lung cancer and in colorectal cancer [43, 44], raising the possibility that suppression of c-Cbl function by inhibitory proteins may be of broad relevance.

Our work may be of particular interest in respect to therapeutic development for a large subgroup of GBMs as inhibition of proteins that themselves inhibit c-Cbl provides more tractable approaches to treatment than restoration of activity in mutated c-Cbl proteins. The extensive interest in Rho GTPase family members led to discovery of the Cdc42 inhibitor used in our studies on BLBC cells [3], and our present studies suggest there would be similar value in developing Cool-1 inhibitors, perhaps targeted to isoforms that are more highly expressed in GBMs than in normal brain. As the importance of disruption of c-Cbl function in human cancers becomes increasingly clear [45-49], it also will be of interest to determine whether Cool-1 is responsible for suppressing c-Cbl activity in other cancers and whether targeting this interaction may provide a rational therapeutic strategy of broad potential relevance.

Conclusions

We found that GBM cells, but not normal CNS glial progenitors, inhibit RFC pathway activity via Cool-1 expression and that restoring normal c-Cbl activity by reducing Cool-1 levels has multiple beneficial outcomes. Cool-1 knockdown reduced GBM cell division, almost eliminated generation of adhesion-independent spheroids in vitro, reduced the representation of cells with antigenic profiles thought to identify TICs, and decreased levels of multiple proteins critical for GBM generation. In vivo, Cool-1 knockdown may even eliminate the ability of GBM cells to generate tumors as all tumors that emerged showed re-expression of this protein. Moreover, Cool-1 knockdown increased sensitivity of GBM cells to BCNU and TMZ. In primary O-2A/OPCs, in contrast, Cool-1 knockdown did not reduce cell division or increase sensitivity to BCNU or TMZ and we found no evidence of Cool-1/c-Cbl complexes in normal brain tissue. Thus, these studies appear to provide a single molecular intervention that has the unusual properties of suppressing TIC functions and increasing sensitivity to relevant therapeutic agents while selectively affecting GBM cells but not primary glial progenitors.

Acknowledgments

We are grateful to multiple colleagues for their insightful comments, and in particular to Craig Jordan, Chris Pröschel, Margot Mayer-Pröschel, Hartmut Land, and Helene McMurray. This research was supported with funding from the National Institutes of Health (ES012708 and CA131385), the Empire State Stem Cell Board (N09G195), and the Carlson Stem Cell Trust.

Author Contributions

B.M.S.: collection and/or assembly of data, data analysis and interpretation, and manuscript writing; C.J.F., W.C., and A.B.: collection and/or assembly of data and data analysis and interpretation; K.W., E.C.W., and A.V.: provision of study material or patient specimens; M.N.: conception and design, financial support, provision of study, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.