Effects of epidermal growth factor receptor blockade on ependymoma stem cells in vitro and in orthotopic mouse models



Some lines of evidence suggest that tumors, including ependymoma, might arise from a subpopulation of cells, termed cancer stem cells (CSCs), with self-renewal and tumor-initiation properties. Given the strict dependence of CSCs on epidermal growth factor (EGF) through EGF receptor (EGFR), we investigated the effects of EGFR inhibitors in ependymoma-stem cells (SCs) in vitro and in orthotopic mouse models. We established two ependymoma-SC lines from two recurrent pediatric ependymoma. Both lines expressed markers of radial glia—the candidate SCs of ependymoma—and showed renewal ability, multipotency, and tumorigenicity after orthotopic implantation, despite markedly different expression of CD133 (94 vs. 6%). High phosphorylated-EGFR/EGFR ratio was detected, which decreased after differentiation. EGFR inhibitors (gefitinib and AEE788) reduced clonogenicity, proliferation and survival of ependymoma-SC lines dose-dependently, and blocked EGF-induced activation of EGFR, Akt and extracellular signal-regulated kinase 1/2. Overall, AEE788 was more effective than gefitinib. EGFR blockade as well as differentiation strongly reduced CD133 expression. However, ex vivo treatment with AEE788 did not impair orthotopic tumor engraftment, whereas ex vivo differentiation did, suggesting that CD133 does not absolutely segregate for tumorigenicity in ependymoma-SCs. Orally administered AEE788 prolonged survival of mice bearing ependymoma-SC-driven orthotopic xenografts from 56 to 63 days, close to statistical significance (log-rank p = 0.06). Our study describes for the first time EGFR signaling in ependymoma-SCs and the effects of EGFR blockade in complementary in vitro and in vivo systems. The experimental models we developed can be used to further investigate the activity of EGFR inhibitors or other antineoplastic agents in this tumor.

Ependymoma accounts for ∼10% of all intracranial tumors of childhood. Despite histological benignancy, the clinical outcome is dismal in ∼50% of patients.1, 2 So far, surgery is the mainstay of treatment, with some benefit coming from postoperative radiotherapy. Conventional anticancer drugs have limited efficacy, although they have recently been reconsidered as adjuvant therapy or to delay the need for radiotherapy.3 Attempts to develop more effective antineoplastic agents have been hindered by the paucity of in vitro and in vivo models.

Accumulating evidence in brain cancers strengthens the hypothesis that only a rare fraction of cells within a tumor, termed cancer stem cells (CSCs), possess the ability to self-renew and proliferate and uniquely maintain tumor growth and recurrence.4–6 CSCs generate phenotypically similar tumorigenic daughter cells, as well as differentiate into phenotypically diverse nontumorigenic cells that form the tumor bulk.7 Because of their distinct properties, CSCs are believed to be spared by current treatments.6, 8 Alternatively, the molecular targeting of abnormally active pathways on which CSCs rely might result in their selective elimination, hence a complete tumor regression.

The gold standard assay for functionally identifying stem cells (SCs) in brain tumors is the ability to propagate serially in an undifferentiated state (evidenced by the formation of floating cell clusters termed neurospheres) and generate tumors in animals.5, 9 Cells that fulfill “stemness” defining criteria have been isolated from ependymoma and identified in restricted populations of cells with features of radial glia (RG)10, 11—the progenitors that give rise to postmitotic neurons, astrocytes and ependymal cells.12 Indeed, ependymoma-SCs express the RG marker brain lipid binding protein (BLBP) in addition to the neural stem markers nestin and CD133.10 Until recently, CD133 has unanimously been considered a marker that segregates for tumorigenicity in SCs isolated from glioblastoma, medulloblastoma and ependymoma.10, 13 However, a substantial number of studies, mostly in glioma, have now provided experimental evidence for the existence of CD133 CSCs.14–16

It is well established that the spherogenic and proliferative properties of neural and brain tumor SCs are mainly dependent on epidermal growth factor (EGF) through EGF receptor (EGFR).5, 17 In addition to a mitogenic role in SCs, EGFR signaling has been implicated in the pathogenesis of different tumors, including ependymoma.3, 18–20 Overexpression of EGFR has been correlated with poor prognosis and associated with increased tumor proliferative activity in ependymoma.21 Therefore, inhibitors of EGFR signaling might represent a novel strategy against this tumor and clinical studies with such agents in ependymoma are ongoing (http://www.clinicaltrials.gov).

Among EGFR tyrosine kinase inhibitors are the monospecific EGFR inhibitor gefitinib (Iressa, ZD1839, AstraZeneca, Macclesfield, UK)22 and the bispecific EGFR/HER2 inhibitor AEE788 (Novartis Pharmaceuticals, Basel, Switzerland),23 the latter blocking other tyrosine kinases as well, although at higher doses. So far, only one study has been published on EGFR inhibitors in ependymoma preclinical models, which shows that gefitinib as a single agent exerts no antitumor effects in vivo.19 Tumor heterogeneity has been accounted for limited efficacy of molecularly designed therapeutics. Therefore, models highly dependent on a targeted signaling might be useful to address the biological relevance of that signaling and the clinical importance of its blockade.

In our study, we characterized EGFR signaling in ependymoma-SCs and explored the antitumor effects of EGFR inhibitors in vitro and, for the first time, in ependymoma-SC-driven xenografts. We established two lines from two recurrent pediatric ependymoma in serum-free, EGF-containing media, which eliminates differentiated cells and forces undifferentiated stem-like cells into an active proliferative state.17 These lines were proven to be enriched with self-renewing, multipotent, tumor-initiating cells and showed high phosphorylated-EGFR/EGFR ratio. In vitro, gefitinib and AEE788 reduced clonogenicity, proliferation, survival, EGF-induced signaling and the CD133+ fraction. In vivo, EGFR blockade partially impaired the growth of orthotopic ependymoma-SC-driven xenografts, underlining the importance of parallel in vitro and in vivo investigations to correctly evaluate effective treatment strategies.

Material and Methods

Chemical compounds

Gefitinib and AEE788 were dissolved in dimethyl sulfoxide (Sigma-Aldrich, Dorset, UK) to a 10-mM stock solution for in vitro studies. For in vivo applications, AEE788 was dissolved immediately before use in N-methylpyrrolidone and polyethylene glycol 300 (1:9; Sigma-Aldrich).

Establishment of ependymoma lines

The ependymoma-SC lines, referred to as EPP and EPV, and the serum-cultured, nonstem line, referred to as EP1, were derived from three pediatric infratentorial ependymoma (Supporting Information Table S1). Tumor specimens were obtained in accordance with Institutional Review Board Approval. Tissues were disaggregated to single-cell suspension by mechanical and enzymatic techniques.17, 24 To establish SC lines, cells were grown in Neurocult medium (Stem Cell Technologies, Vancouver, BC, Canada) supplemented with EGF (20 ng/ml; Sigma-Aldrich) and basic fibroblast growth factor (bFGF; 10 ng/ml; Promega, Woods Hollow Rd, WI). To establish the EP1 line, cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 10% fetal bovine serum (FBS). All of the in vitro and in vivo experiments were performed with cells between passage #15 and passage #26.

Proliferation assay

Proliferation assay was performed as previously described.25 Cells were plated in 25-cm2 flasks at a density of 2.5 × 105 cells/flask and let grow for 7 days. Cell numbers were determined by an automated cell counter (NucleoCounter NC-100™, ChemoMetec, Allerød, Denmark) that allows for discrimination between live and dead cells. Successively, 2.5 × 105 cells were replated in new flasks and counted after 7 days. Plating and counting were repeated weekly for a total of 5 weeks. We calculated the weekly fold expansion of cells (weekly cell number/2.5 × 105 cells). These factors were used to calculate the number of cells if all cells would have been plated at each time point.

Limiting dilution assay

Limiting dilution assay was performed as previously described11: neurosphere-derived cells were seeded in 96-well plates at dilutions from 200 to 5 cells/well. After 7 days, the percentage of wells without neurospheres for each cell plating density was calculated and plotted against the number of cells per well. The x-intercept values, which represent the number of cells required to form at least one tumor sphere in every well, were determined by regression lines. We used a diameter of 100 μm to define a standard neurosphere, which corresponds to ∼150–200 cells.

Differentiation assay

Differentiation was induced as previously described.10 Dissociated neurospheres were seeded onto vessels coated with poly-L-ornithine and laminin (Invitrogen, Carlsbad, CA) and allowed to grow for 24 hr. Medium was changed to Neurobasal medium (Invitrogen) with 10% FBS without mitogens for up to 5 days.

Immunofluorescence staining

Cells were fixed using 4% paraformaldehyde solution in phosphate-buffered saline (PBS) at room temperature for 20 min. After a 1-hr blocking with 10% normal goat serum (NGS)-PBS, cells were incubated with primary antibodies diluted in 10% NGS-PBS with 0.1% Triton X-100 for 2 hr at room temperature. The primary antibodies used were: anti-CD133, anti-BLBP (Abcam, Cambridge, UK); anti-Tuj-1 (Covance, Emeryvelle, CA); anti-nestin, anti-2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase; Millipore, Temecula, CA); anti-glial fibrillary acidic protein (GFAP; DAKO, Glostrup, Denmark). After washing with PBS, coverslips were incubated for 1 hr with goat anti-mouse IgG and IgM–Alexa 546 or goat anti-rabbit IgG–Alexa 488 secondary antibodies (Molecular Probes, Invitrogen) diluted in 10% NGS-PBS with 0.1% Triton X-100. Cell nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI, Molecular Probes, Invitrogen), mounted and visualized with fluorescence microscope Axioplan (Carl Zeiss, Jena, Germany).

Flow cytometry

Cells were incubated with monoclonal CD133/1-phycoerythrin (CD133-PE)-conjugated antibody or a PE-conjugated mouse IgG1 isotype control antibody (Miltenyi Biotec, Bergisch Gladbach, Germany) for 30 min at 4°C and washed with 1 ml of cell culture medium. Cells were then centrifuged at 500g for 5 min, resuspended in 0.5-ml culture medium and analyzed by flow cytometry (CyAn Flow Cytometer, Beckman Coulter, Orange County, CA).

RNA preparation and reverse transcription-quantitative polymerase chain reaction

RNA extraction and reverse transcription-quantitative polymerase chain reaction (RT-qPCR) were performed as previously described.26 TaqMan gene expression assays for CD133, nestin, GFAP, βIII-tubulin and the reference normalization gene hypoxanthine guanine phosphoribosyltransferase (HPRT) were obtained from Applied Biosystems (Foster City, CA). Each qPCR reaction was performed in triplicate, and the average of the three threshold cycles (Ct) of each gene was normalized to that of HPRT for each sample to get ΔCt. The ΔCt was then converted to the relative amount of tissue target mRNA by the formula 2math image.

Cell viability assay

Cells were seeded onto six-well plates. After 24 hr, vehicle or serial concentrations of EGFR inhibitors were added to the medium and cells were cultured for 72 hr. Cell number and viability were assessed by NucleoCounter and expressed as a percentage of the control.

Clonogenic assay

To determine sphere-forming capacity in the presence of EGFR inhibitors, a modified limiting dilution assay was used.27 One hundred cells were plated in medium containing AEE788 or gefitinib (0.1–10.0 μM) or vehicle. After 7 days, the number of wells containing spheres was calculated and expressed as percentage of control wells.

EGF stimulation and protein analysis

Dissociated neurospheres were cultured overnight in medium without mitogens. After incubation with EGFR inhibitors (0.1–10.0 μM) for 2 hr, cells were stimulated for 10 min with 25 ng/ml EGF and immediately harvested in lysis buffer, as previously reported.26 Lysates were electrophoresed on polyacrylamide gels and transferred to a Hybond nitrocellulose membrane (Amersham Pharmacia, Piscataway, NJ). Membranes were probed with the following primary antibodies: anti-EGFR, anti-phospho-EGFR (Tyr1173), anti-poly(ADP-ribose) polymerase (PARP), anti-actin (Santa Cruz Biotechnology, Santa Cruz, CA); anti-Akt, anti-phospho-Akt (Ser473), anti-extracellular signal-regulated kinase (ERK)1/2, anti-phospho-ERK1/2 (Thr202/Tyr204; Cell Signaling Technology, Beverly, MA). After incubation with horseradish peroxidase (HPR)-conjugated secondary antibodies (Vector Laboratories, Burlingame, CA), the immunoblots were visualized using the Enhanced Chemiluminescence (ECL) detection system (Amersham Pharmacia).

Stereotactic injection of ependymoma-SCs and animal treatment

All of the animals were experimented in strict accordance with the guidelines of the Istituto Superiore di Sanità (Rome, Italy) and the Ethical Committee of Catholic University. To generate orthotopic xenografts, 3 × 105 cells in 10 μl PBS were implanted into the fourth ventricle of nude mice using a stereotaxic injection frame (Kopf Instrument, Better Hospital Equipment Corp, Miami Lakes, FL), after administration of general anesthesia (80 mg/kg ketamine + 10 mg/kg xylazine, Sigma-Aldrich). The injection coordinates were: anteroposterior, −6.0 mm from Bregma; lateral to right, 0.2 mm; dorsoventral, −4 mm (http://www.mbl.org/atlas/atlas.php). Mice were then randomized into drug-treated and control group and treatment was started 3 days from implantation. AEE788 was administered orally (50 mg/kg daily for 3 days/week) for the duration of the survival study. Control group received only vehicle. On the appearance of brain tumor symptoms, mice were killed and brains were collected. The results were examined using the Kaplan–Meier method, and significance testing (p = 0.05) performed based on the log-rank test.

Histology and immunohistochemistry

Xenograft specimens were fixed with 4% paraformaldehyde, paraffin embedded and cut into 3-μm sections. Sections were deparaffinized, and endogenous peroxidase was blocked with 3% hydrogen peroxide in PBS. After microwaving sections for antigen retrieval in 0.01 M citrate solution (pH 6.0) for 5 min × 4 times, nonspecific protein binding was blocked with 20% normal goat/rabbit serum. Sections were then incubated overnight at 4°C with anti-Ki-67 antibody (Novocastra Laboratories, Newcastle, UK) according to the manufacturer's specifications. Sections were then incubated with the secondary anti-rabbit antibody HRP-conjugated (DAKO Cytomation, Carpinteria, CA). The signal was detected using 3,3′-diaminobenzidine (DAB) substrate (DAKO). Slides were counterstained with Mayer hematoxylin and were finally mounted.


bFGF: basic fibroblast growth factor; BLBP: brain lipid binding protein; CNPase: 2′,3′-cyclic nucleotide 3′-phosphodiesterase; CSCs: cancer stem cells; EGF: epidermal growth factor; EGFR: epidermal growth factor receptor; ERK: extracellular signal-regulated kinase; GAPDH: glyceraldehyde-3-phosphate dehydrogenase; GFAP: glial fibrillary acidic protein; HPRT: hypoxanthine phosphoribosyltransferase; RG: radial glia; RT-qPCR: reverse transcription quantitative polymerase chain reaction; SCs: stem cells


Ependymoma contains tumorigenic SCs

We established ependymoma-SC lines from two recurrent pediatric ependymoma by using neural SC permissive conditions. Within ∼7 days of primary culture, single-cell suspensions formed neuronal sphere-like clusters (Fig. 1a). These spheres were serially propagated giving rise to SC lines—referred to as EPP and EPV—which showed extensive proliferation with a doubling time of 38.6 ± 4.4 hr and 46.0 ± 4.4 hr, respectively (Fig. 1b). Over passaging, EPP line continued to proliferate as floating clusters, whereas EPV line grew as both neurosphere-like structures and substrate-adherent cells. Self-renewal capacity of ependymoma neurosphere-forming cells was assessed by limiting dilution assay11: the percentage of tumor cells able to generate clonally derived neurospheres was 1.1 and 0.8% for EPP and EPV lines, respectively (Fig. 1c). Ependymoma-SC cultures exhibited abnormal karyotypes, which confirmed that these cells were transformed and were not normal neural SCs contaminating the tumor cell population (Supporting Information Fig. S1).

Figure 1.

Ependymoma-derived neurospheres display long-term proliferation, self-renewal capacity and tumorigenicity. (a) Phase-bright photomicrographs showing EPP and EPV neurospheres derived from single-cell suspensions of two pediatric ependymoma. Cells were grown in a serum-free medium with EGF and bFGF. Original magnification, ×20. (b) In vitro proliferation of ependymoma-SC EPP and EPV lines was evaluated over 5 weeks. For each time point, 2.5 × 105 cells were plated and counted after 7 days. (c) Limiting dilution analysis of EPP and EPV SC lines. Neurosphere-derived cells were seeded in 96-well plates at dilutions ranging from 200 to 5 cells/well. After 7 days, the percentage of wells without neurospheres for each cell plating density was calculated and plotted against the number of cells per well. The x-intercept value indicates the number of cells required to form at least one neurosphere per well. (d) Hematoxylin–eosin staining of ependymoma xenografts generated after orthotopic injection of EPP cells in the mouse brain. (e) Representative immunohistochemical analysis using Ki-67 antibody shows a high proliferation index (70–80%). Original magnifications both ×40.

To unequivocally identify our presumptive ependymoma-SCs as such, we assessed their in vivo tumorigenicity. After orthotopic injection of EPP and EPV cells, brain tumor symptoms occurred within 67 ± 7 days and 95 ± 12 days, respectively, with 100% penetrance. Immunohistochemical analysis confirmed the growth of xenograft tumors in all mice (Fig. 1d, and Supporting Information Fig. S2). Tumors were well demarcated, with occasional invasion into surrounding normal brain, and displayed high cellularity, high mitotic index, poor differentiation (hematoxylin–eosin staining; Fig. 1d) and a high degree of proliferation as detected by Ki-67 immunoreactivity (70–80%; Fig. 1e).

Ependymoma-SC lines display differential levels of RG markers, which are modulated by differentiation

Once confirmed tumor-initiation properties in our ependymoma lines, we determined whether they displayed an RG phenotype, as previously reported for the candidate SCs of ependymoma.10 Therefore, we examined the expression of the RG markers nestin, CD133, BLBP and GFAP.10, 12, 28 Western blot analysis (Fig. 2a) and parallel immunofluorescence staining (Supporting Information Fig. S3) demonstrated similar levels of nestin, GFAP and BLBP in EPP and EPV lines. Of note, abundant expression of CD133 protein was observed only in the former, although both lines were tumorigenic. Differential expression of CD133 was confirmed by RT-qPCR (Fig. 2b), which demonstrated a ninefold higher level of CD133 mRNA in EPP cells. By cytofluorimetric analysis, CD133 was expressed in ∼94% of EPP cells, and in only 6% of EPV cells (Fig. 2c). Interestingly, the serum-derived primary culture EP1—which was established in growth conditions not selective for SCs and was not tumorigenic—showed no CD133 at both mRNA and protein levels, and a pattern of nestin, GFAP and BLBP expression opposite to that of the SC lines (Figs. 2a and 2b).

Figure 2.

Ependymoma-SC lines show an RG phenotype and multilineage differentiation properties. (a) Western blot analysis of nestin, CD133, GFAP, BLBP in EPP and EPV SC lines. The serum-cultured, nonstem EP1 line also was probed. GAPDH was used as a loading control. (b) RT-qPCR analysis of expression of CD133 and nestin in the aforementioned lines. Levels were normalized to the level of the reference gene HPRT. Average values from two independent amplifications each in triplicate are shown. (c) Flow cytometric histograms demonstrate the percentage of CD133+ cells in ependymoma-SC lines. (d) Cells were grown in proliferation conditions (P, serum-free medium with mitogens) or differentiation conditions (D, 5 days in 10% serum-containing medium without growth factors). Cell lysates were subjected to immunoblot analysis with antibodies to nestin, CD133, GFAP, Tuj-1, CNPase and BLBP. GAPDH was used as a loading control. (e) RT-qPCR analysis of CD133 and GFAP in EPP and EPV cells exposed to differentiation conditions for 5 days. Levels were normalized to the level of HPRT in each sample. Fold changes were calculated relative to undifferentiated controls and plotted as log2. Means of two independent experiments are shown. (f) Representative images of immunofluorescence staining of RG (nestin, CD133, BLBP), glial (GFAP), neuronal (Tuj-1) and oligodendroglial (CNPase) markers. Nuclei were counterstained with DAPI. Original magnifications ×20.

To test whether the ependymoma-SC lines displayed multilineage differentiation properties, we cultured single-cell suspensions in the absence of mitogens. In these conditions, both lines showed dramatic morphological changes toward neuronal- and astrocytic-like cells (Fig. 2f). Differentiated cells expressed surrogate markers of the glial (GFAP), neuronal (Tuj-1) and oligodendroglial (CNPase) lineages, but retained expression of nestin and BLBP also (Figs. 2d and 2f). However, the majority of differentiated EPP cells lost expression of CD133 as compared to undifferentiated neurospheres (Fig. 2f, and Supporting Information Fig. S3). Western blot and RT-qPCR demonstrated that the morphological changes were associated with an increase in GFAP expression in both lines, accompanied by a decrease in CD133 expression in EPP (Figs. 2d and 2f).

Collectively, these results indicate that differentiation induces a predominant astroglial phenotype in ependymoma-SCs, similar to the astroglial commitment previously reported in differentiated glioma-SCs.29, 30 However, the persistent expression of stemness markers suggests that ependymoma-SCs do not fully execute the differentiation program, in keeping with the aberrant differentiation properties of CSCs.5, 31

Ependymoma-SCs show a higher p-EGFR/EGFR ratio than differentiated cells

To determine whether ependymoma-SCs display differential activation of EGFR pathway compared to non-SCs, we examined the expression levels and the activation status of mediators of this signaling in our cultures. On binding to EGF, activated EGFR homodimerizes and heterodimerizes and triggers signaling through ERK1/2 and Akt, which promote cell proliferation and survival.32 Western blot analysis showed that total and phosphorylated (activated) EGFR were detectable in all lines, although EGFR was highly expressed in EP1 and EPP cells, whereas p-EGFR in only EPP cells (Fig. 3a). However, both SC lines displayed a much higher p-EGFR/EGFR ratio than EP1 line. Levels of total and activated Akt were higher in SC lines, whereas comparable amounts of ERK1/2 and p-ERK1/2 were found. Interestingly, differentiation of EPP and EPV cells led to a decrease in the phosphotyrosine content of EGFR, despite an increase in EGFR expression at both mRNA and protein levels (Fig. 3b and data not shown). Thus, both differentiated ependymoma-SC lines and an unrelated serum-derived, nonstem ependymoma line show a reduced p-EGFR/EGFR ratio.

Figure 3.

EGFR signaling is activated in ependymoma-SC lines and blocked by EGFR inhibitors. (a) Immunoblot of lysates of EPP and EPV cultures with antibodies to total and phosphorylated (p) EGFR, Akt and ERK1/2. The nonstem EP1 line was also probed. Equal loading is shown by GAPDH. (b) Expression levels of EGFR and p-EGFR in ependymoma-SC lines after differentiation. Cells were grown in either serum-free medium with mitogens (proliferation conditions, P) or in the presence of 10% serum without growth factors for 5 days (differentiation conditions, D). Cells were growth factor- or serum-starved for 24 hr prior to Western blot analysis. (c) Dose-dependent inhibition of EGF-induced activation of EGFR, Akt and ERK1/2 by gefitinib and AEE788 in ependymoma-SC lines. Cells were grown overnight in a mitogen-free medium. Cells were then treated with increasing concentrations of EGFR inhibitors for 2 hr prior to a 10-min exposure to EGF (25 ng/ml). Cell lysates were subjected to immunoblot analysis with antibodies to phosphorylated (p) and total EGFR, Akt and ERK1/2. Actin was used as a loading control. (d) Dose-dependent inhibition of activation of EGFR, Akt and ERK1/2 by gefitinib and AEE788 in ependymoma-SC lines grown in EGF and bFGF-containing media. Cells were treated with increasing concentrations of EGFR inhibitors for 3 hr. Western blot analysis was performed on the phosphorylated levels of EGFR, Akt and ERK1/2. Equal loading was verified by actin immunoblotting.

The preferred heterodimerization partner of EGFR is the ligandless HER2, which potentiates the signaling elicited by EGF.32 HER2 is expressed in RG and its down-regulation coincides with the RG to astrocyte transformation during neurogenesis, whereas activation of HER2 maintains RG proliferation.33, 34 Expression of HER2, but not of its phosphorylated form, was found only in EP1 cells (Supporting Information Fig. S4), indicating that EGF signaling is mainly—if not exclusively—mediated by EGFR in our ependymoma-SC lines.

The EGFR inhibitors gefitinib and AEE788 interfere with EGF-induced signaling in ependymoma-SC lines

We examined the inhibitory effects of gefitinib and AEE788 on EGFR signaling in ependymoma-SC lines. Cells were grown overnight in a mitogen-free medium, treated for 2 hr with increasing concentrations of inhibitors, and then stimulated with EGF. In both SC lines, EGF strongly activated its cognate receptor (Fig. 3c). Gefitinib and AEE788 were able to prevent the phosphorylation of EGFR in a dose-dependent manner. In parallel with EGFR, Akt and ERK1/2 were activated by EGF and reduced to basal levels by the two agents. Levels of total EGFR, Akt and ERK1/2 did not change over the experiment. Overall, AEE788′s inhibitory effects were greater than gefitinib's, and EPP cells were more sensitive to both EGF-induced activation of EGFR signaling and its blockade than EPV cells.

We then addressed the inhibitory properties of AEE788 and gefitinib in regular proliferation medium, which contains bFGF in addition to EGF. Both the agents were able to down-regulate chronically activated EGFR, Akt and ERK1/2 (Fig. 3d), indicating these inhibitors as potentially effective on EGF signaling even in the presence of heterologous growth factors that might be present in vivo.

EGFR inhibitors decrease clonogenic potential and proliferation of ependymoma-SC lines

To determine whether targeting EGFR activity would preferentially reduce the SC compartment in ependymoma-SC lines, we investigated neurosphere formation—which is a measure for the presence of CSCs35—by using modified limiting dilution assays.27 Dissociated EPP and EPV neurospheres were plated in the presence of AEE788, gefitinib or vehicle. In drug-treated wells, the number of equally sized tumor spheres decreased dose-dependently as compared to vehicle-treated controls (Fig. 4a), suggesting a depletion of the SC component.

Figure 4.

EGFR inhibitors decrease clonogenicity, proliferation and survival of ependymoma-SC lines, without inducing apoptosis. (a) A modified limiting dilution assay was performed to determine the effects of EGFR inhibitors on the clonogenic potential of EPP and EPV lines. One hundred cells/well—approximately the number of cells that form one neurosphere per well—were plated in the absence or presence of various concentrations of AEE788 or gefitinib. After 7 days, the number of drug-treated wells containing spheres was calculated and expressed as percentage of vehicle-treated wells. (b) Ependymoma-SC lines are more sensitive to inhibition of proliferation and survival by EGFR inhibitors than the serum-derived EP1 line. Cells were treated with AEE788 or gefitinib for 72 hr. Viable and nonviable cells were counted by NucleoCounter. (c) Western blot analysis of EPP and EPV SC lines exposed to EGFR inhibitors for up to 72 hr. Membranes were probed with antibodies to PARP (72-hr drug exposure), p27 and the loading control actin (24-hr drug exposure). Loading control for the 72-hr treatment is shown in Fig. 5a.

We next compared proliferation and survival of EPP and EPV cultures to those of EP1 cultures after treatment with EGFR inhibitors for 72 hr. AEE788 and gefitinib exerted a dose-dependent reduction in the number of viable cells with a concomitant increase in the number of nonviable cells (Fig. 4b). Overall, SC lines were more sensitive to EGFR inhibitors than the serum-cultured line, and EPP line was more sensitive than EPV, in keeping with molecular and clonogenicity data. Indeed, 10-μM AEE788 reduced cell viability by ∼90 and 60% in SC and non-SC lines, respectively. An equimolar concentration of gefitinib determined a 50% growth inhibition in EPP and EPV cells, being almost ineffective on EP1 cells, where, unexpectedly, low doses of gefitinib caused a small, although reproducible, stimulation of cell proliferation. Differentiated EPP and EPV SC lines showed reduced cell death by the two inhibitors (Supporting Information Fig. S5).

Decrease in cell number can be ascribed to impaired proliferation and/or different types of cell death, including apoptosis.36 Therefore, we checked cells for the expression of a mitotic inhibitor (p27), and an apoptotic marker (cleavage of PARP). Cells were exposed to EGFR inhibitors over a 72-hr period. Levels of p27 increased dose-dependently already after 24 hr of treatment (Fig. 4c, and data not shown). No cleavage of PARP was observed up to 72 hr.

EGFR inhibition decreases CD133 expression, but not tumor-initiation properties of ependymoma-SCs

We next examined whether inhibition of EGFR signaling is accompanied by a loss of stemness features. AEE788 and gefitinib reduced CD133 expression dose-dependently in EPP line, whereas levels of nestin remained stable in both lines (Fig. 5a). Cytofluorimetric analysis demonstrated a reduction of the percentage of CD133+ cells by the two EGFR inhibitors, more remarkable after AEE788 treatment (Fig. 5b, and Supporting Information Fig. S6). Indeed, an exposure to 5-μM AEE788 decreased the CD133+ subpopulation of EPP line by ∼56%. No sign of morphological differentiation was observed even after prolonged exposure to low doses of the agents, and, consistently, treatment did not induce upregulation of either of the differentiation markers GFAP or βIII tubulin (Fig. 5c). To address whether loss of CD133 expression translates to a reduction in tumor engraftment, we stereotactically injected an equal number of viable EPP cells after ex vivo treatment with vehicle or AEE788 (1 and 5 μM). Tumors originated with similar growth kinetics in the three groups (five animals per group), resulting in no statistically significant difference in survival of mice (Fig. 5d). Of interest, ex vivo differentiation—that yet decreased CD133 expression to a similar extent (compare Figs. 2d and 5a)—reduced tumorigenicity and significantly prolonged the survival of xenografted mice (log-rank, p < 0.02).

Figure 5.

EGFR blockade reduces CD133+ subpopulation of EPP line, but not its tumor-initiation properties. (a) Ependymoma-SC lines were exposed to the indicated concentrations of AEE788 or gefitinib for 72 hr. Cell lysates were subjected to immunoblot analysis to determine expression of the SC markers nestin and CD133. Actin was used as a loading control. (b) Changes of the CD133+ fraction of EPP line treated with AEE788 or gefitinib by flow cytometric analysis. Representative histograms are shown. (c) GFAP and βIII-tubulin transcript levels in ependymoma-SC lines after treatment with AEE788 or gefitinib (1 μM) for 7 days. mRNA of the target gene was quantified by RT-qPCR and normalized to the level of the reference gene HPRT in each sample. Means ± SD relative to untreated controls, which were used as calibrators (1 = no change). (d) Survival of mice (five animals per group) after orthotopic injection of an equal number of viable EPP cells treated with vehicle or AEE788 (1 and 5 μM) or after differentiation (diff). Mice were sacrificed when brain tumor symptoms developed. Survival was examined using the Kaplan–Meier method.

Collectively, these data demonstrate that CD133 expression on its own is not strictly correlated with the tumorigenic potential of ependymoma-SCs and that EGFR inhibition does not select a subpopulation of more differentiated, less tumorigenic cells.

AEE788 at a clinically achievable dose is more active than gefitinib in ependymoma-SC lines

To explore the effects of an exposure mimicking the in vivo administration of the inhibitors, we treated the cells with a clinically achievable concentration (1 μM) of either agent for up to 7 days. After 3 days, AEE788 and gefitinib showed similar and limited growth-inhibitory activity; however, the cytotoxic profile of the two agents was very different by day 7. Indeed, AEE788 reduced the proliferation of EPP and EPV cells by 91 and 66%, respectively, whereas gefitinib by only 34 and 26% (Fig. 6a). At this time point, AEE788-treated cultures were mostly made up of nonviable cells (72 and 40% in EPP and EPV lines, respectively), with rare, small neurospheres (Fig. 6b). Gefitinib did not considerably increase the fraction of nonviable cells even after 7 days (Fig. 6a), although it reduced the size of neurospheres (Fig. 6b). Unlike short-term treatment, long-term exposure to AEE788 (but not to gefitinib) induced apoptosis to some extent in both lines, as detected by the cleavage of PARP (Fig. 6c). Collectively, these data indicate that gefitinib acts through a prevalent cytostatic mechanism, whereas AEE788 exerts cytotoxic effects also.

Figure 6.

AEE788, the more effective agent in vitro at a clinically achievable dose (a)–(c), causes a close-to-significance prolongation of the survival of mice bearing ependymoma-SC-driven xenografts (d). (a) Ependymoma-SC lines were treated with 1 μM of either AEE788 or gefitinib for up to 7 days, with medium changed every other day. The percentage of live and dead cells was determined on day 3 and day 7. Each point shows the average value of three independent experiments. Bars, SD. (b) Representative images of vehicle- or drug-treated neurospheres photographed at day 7 are shown. Original magnification, ×20. (c) Time-course of the activation of PARP and of EGFR signaling molecules in ependymoma-SC lines after treatment with EGFR inhibitors. Cells were exposed to AEE788 or gefitinib at 1 μM for 3 or 7 days. Cell lysates were subjected to immunoblot analysis with antibodies to PARP and to phosphorylated and total EGFR and Akt. (d) EPP cells (3 × 105) were injected orthotopically into nude mice. Mice were then randomly divided into two groups of eight animals per group, and either vehicle or 50 mg/kg AEE788 was administered orally thrice a week for about 8 weeks. On the appearance of brain tumor symptoms, animals were sacrificed and brains removed for further analyses. Survival was examined using the Kaplan–Meier method (p = 0.06).

As for EGFR signaling, prolonged treatment with either agent determined a time-dependent increase in the amount of EGFR protein and mRNA in both lines (Fig. 6c, and data not shown). AEE788 was yet able to effectively block EGFR activation, whereas gefitinib did so only in EPV line. Akt inactivation occurred in parallel and was accompanied by a decrease in the amount of unphosphorylated protein, more remarkable after AEE788 treatment.

AEE788 prolongs survival of mice bearing ependymoma-SC-driven orthotopic xenografts, approaching statistical significance

We next evaluated whether the critical effects of EGFR blockade in vitro translate to in vivo survival difference by targeting EGFR in intracranial tumor propagation. Preliminary experiments indicated 50 mg/kg AEE788 thrice a week via oral gavage as the maximum tolerated dose for long-term in vivo treatment. Mice were administered with AEE788 for ∼2 months and sacrificed when brain tumor symptoms developed. Immunohistochemical analysis confirmed the growth of xenograft tumors in all mice (data not shown). AEE788 treatment caused a close-to-significance prolongation of survival from 56 ± 3 to 63 ± 3 days (log-rank, p = 0.06; Fig. 6d). Treatment-related toxicity was minor, with a less than 20% body weight loss at worst.


Despite advancements of the clinical management of several cancers, there has been no substantial improvement in the outcome of ependymoma patients during the past 10 years.1–3 CSCs have only recently been discovered in ependymoma and their role in tumor formation and progression is yet to be determined. An increased understanding of the biological characteristics of these cells is necessary to discover potential therapeutic targets.

In our study, we established and fully characterized two ependymoma lines that were proved to be enriched with cells that display the defining features of stemness, that is, self-renewal and tumor propagation, and are usually lost in serum-cultured cell lines.9, 37 So far, very few permanent ependymoma cell lines are available, which have been established in serum-containing media,31, 38, 39 except one line, recently established in neural SC permissive conditions.40 Similar to the latter, our ependymoma lines have been cultured in conditions that favor propagation of SCs since the surgical resection of patients' tumors. Both our ependymoma-SC lines can be induced to differentiate along multiple lineages, which, on the whole, mimic the phenotypic heterogeneity of the tumor bulk. In addition, their ability to develop orthotopic tumors in vivo permits to investigate the cellular outcome to antineoplastic agents in an environment that mirrors that of human ependymoma more closely. Because therapy responsiveness is linked to the biological characteristics of the tumor cells, including their differentiation status and related proliferation capacities, tightly controlled stem/nonstem cultures and complementary in vitro/in vivo systems might be more predictive models in the preclinical screening of novel agents.

Phenotypic profiling revealed expression of the canonical markers of RG—that is, nestin, BLBP and GFAP10—in both lines, in agreement with previous publications that propose that ependymoma may arise from a small subpopulation of tumor-initiating cells with RG features.10, 31, 38, 41 These cells have been prospectively identified in the CD133+ fraction, because unsorted or CD133 ependymoma cells do not develop tumors in host animals.10 Even though we xenotransplanted only unsorted cells, some lines of evidence indicate that CD133 might not unambiguously identify tumorigenic cells in ependymoma. Indeed, EPP and EPV lines showed a markedly different expression of CD133 (94 vs. 6%, respectively), yet both lines generated tumors on orthotopic xenotransplantation. The tumorigenic potential of the two lines was identical in terms of tumor take rates (100% penetrance), although EPV-driven xenografts grew slower than EPP-driven xenografts, which was, however, consistent with the different doubling times in vitro. In addition, ex vivo treatment with AEE788—that yet reduced the proportion of CD133+ EPP cells by ∼60%—neither inhibited nor delayed the tumor development in brains. By contrast, ex vivo differentiation, which was accompanied by a comparable decrease in CD133 expression, did reduce tumorigenicity of ependymoma-SCs. Until recently, CD133 has been considered a defining marker of stemness for cancer cells—hence of tumorigenicity. However, this assumption is now being questioned in a growing number of studies.42 CD133 seems to be required not for initiation but for progression of glioblastoma.15, 16, 43 In a recent publication, EGFR expression in glioblastoma-SCs appears to determine tumor growth more than CD133 expression on itself, because EGFR+ cells, regardless of their CD133 status, are more tumorigenic than EGFR cells.44 Therefore, the role for CD133 as a marker of tumorigenicity of SCs of brain tumors, including ependymoma, should be further explored.

Consistent with the strict dependence on EGF signaling of SCs, both ependymoma-SC lines displayed a high p-EGFR/EGFR ratio, which was reduced by differentiation, indicating EGFR pathway as a possible target to selectively eliminate the SC compartment from the tumor bulk. The EGFR inhibitors gefitinib and AEE788 decreased sphere formation, proliferation and survival of ependymoma-SC lines in a dose-dependent manner, and concurrently blocked EGFR-mediated signaling. On the whole, undifferentiated SC-lines were more sensitive to growth inhibition and cell death by the two agents than their respective differentiated lines or an unrelated nonstem, serum-derived ependymoma line. Notably, EGFR inhibition decreased the proportion of CD133+ cells. However, xenotransplantation of equal numbers of vehicle- or AEE788-treated viable cells originated tumors with similar growth characteristics, indicating that EGFR blockade did not select a less tumorigenic subpopulation. Overall, EGFR inhibitors in vitro did not cause a loss of stemness features, as they did not induce either morphological or biochemical differentiation (no up-regulation of GFAP or βIII tubulin, no decrease in nestin), apart from down-regulation of CD133. Little is known about CD133 function and regulation.14 Although loss of CD133 expression after treatment with chemotherapeutics has consistently been interpreted as elimination of CSCs,27, 30 population changes could be due to direct effects on the CD133 levels. Differential expression of CD133 may reflect cycling cells rather than a differentially expressed stable SC lineage marker.45 By the same token, the down-regulation of CD133 that we found after AEE788 treatment might be due to a decrease in the cycling cells (as shown by the corresponding increase in p27 expression) rather than to a loss in SCs.

Orally administered AEE788 over a prolonged period of time increased the survival of mice bearing orthotopic ependymoma SC-driven xenografts from 56 to 63 days, with values close to significance. To the best of our knowledge, no study has so far addressed the response of ependymoma (or other brain tumor) SCs to EGFR inhibition in vivo. In our study, the dramatic response observed in ependymoma-SC cultures only partially translated in xenografts, underlining the importance of parallel investigations in in vitro and in vivo settings to correctly evaluate the therapeutic potential of antineoplastic agents. Inadequate biodistribution to the tumor site might account for the limited effects observed in our in vivo models, although AEE788 has been reported to be effective against intracranial glioma xenografts46—hence, to cross blood brain barrier. Other explanations can be ventured: (1) orthotopic xenografts reflect the tumor microenvironment more accurately, thus uncovering mechanisms responsible for CSCs proliferation that can be missed in vitro. For instance, other growth factors present in the tumor niche might substitute for EGF. Inhibitors of Hedgehog and Notch pathways have been shown to decrease tumor engraftment of glioblastoma-SCs, indicating these signaling as mitogenic for CSCs.47, 48 (2) Blockade of EGF signaling might lead to the acquisition of compensatory genetic and/or epigenetic events.44 In vitro, we found that EGFR inhibitors up-regulated the expression of EGFR at mRNA and protein levels in ependymoma-SCs. Whether a similar mechanism occurs in vivo, thus, possibly mediating some kind of resistance to EGFR inhibition could be investigated. (3) Because of the crucial importance of EGFR for CSCs, crosstalk with heterologous signaling systems might activate EGFR signaling through alternate pathways in vivo, thus escaping EGFR-specific blockade. (4) Tumor growth might be sustained by an ampler population of tumorigenic cells, whose features do not completely fit in those of canonical SCs.49 In this light, the dichotomic division of tumor cells in stem and nonstem subpopulations might not always reflect what occurs in vivo, where cancer cells need a higher degree of plasticity to adjust to different niches and optimize their chance of survival,50 hence their responsiveness to chemotherapeutics.

Despite solid rational basis, anti-EGFR therapy has not always lived up to expectations in the treatment of different tumors, including ependymoma.51 Insights into the molecular mechanisms underlying the responsiveness of tumor stem and non-SCs to such agents will help define novel and more effective therapeutic strategies.


We thank Dr. Jolanda Bajer for karyotypic analysis and Dr. Angelo Vescovi for supplies of serum-free, mitogen-containing media. Gefitinib and AEE788 were kindly provided by AstraZeneca and Novartis Pharmaceuticals, respectively.