Hedgehog Signaling Regulates Brain Tumor-Initiating Cell Proliferation and Portends Shorter Survival for Patients with PTEN-Coexpressing Glioblastomas

Hedgehog Signaling Regulates Brain Tumor-Initiating CellProliferation and Portends Shorter Survival for Patients withPTEN-Coexpressing Glioblastomas

QIJIN XU, XIANGPENG YUAN, GENTAO LIU, KEITH L. BLACK, JOHN S. YU

Maxine Dunitz Neurosurgical Institute, Department of Neurosurgery, Cedars-Sinai Medical Center, Los Angeles,California, USA

Key Words. Glioblastoma • Cancer stem cells • Hedgehog • PTEN

ABSTRACT

The identification of brain tumor stem-like cells (BTSCs)has implicated a role of biological self-renewal mechanismsin clinical brain tumor initiation and propagation. The mo-lecular mechanisms underlying the tumor-forming capacityof BTSCs, however, remain unknown. Here, we have gen-erated molecular signatures of glioblastoma multiforme(GBM) using gene expression profiles of BTSCs and haveidentified both Sonic Hedgehog (SHH) signaling-dependentand -independent BTSCs and their respective glioblastomasurgical specimens. BTSC proliferation could be abrogatedin a pathway-dependent fashion in vitro and in an intracra-nial tumor model in athymic mice. Both SHH-dependent

and -independent brain tumor growth required phospho-inositide 3-kinase-mammalian target of rapamycin signal-ing. In human GBMs, the levels of SHH and PTCH1expression were significantly higher in PTEN-expressingtumors than in PTEN-deficient tumors. In addition, weshow that hyperactive SHH-GLI signaling in PTEN-coex-pressing human GBM is associated with reduced survivaltime. Thus, distinct proliferation signaling dependencemay underpin glioblastoma propagation by BTSCs. Mod-eling these BTSC proliferation mechanisms may providea rationale for individualized glioblastoma treatment.STEM CELLS 2008;26:3018 –3026

Disclosure of potential conflicts of interest is found at the end of this article.

INTRODUCTION

The cancer stem cell (CSC) hypothesis not only has provided aframework for understanding cancer heterogeneity, tumorigen-esis, cancer progression, and cancer therapy but also has offeredan alternative approach to modeling human cancer. Recentidentification of cancer-initiating stem cells in brain tumor [1,2], prostate cancer [3], colon cancer [4, 5], and breast cancer [6,7] suggested that CSCs may play a central role in the propaga-tion of several cancer types. CSCs have also been shown to beresponsible for prevalent radioresistance and chemoresistance inglioma [8]. Compared with conventionally cultured human can-cer cell lines, CSCs have been shown to recapitulate humanbrain tumors in phenotype and in cancer genetics and thus maymore faithfully model mechanisms of tumorigenesis and tumorpropagation [9].

Glioblastoma multiforme (GBM) is the most malignantform of human primary brain tumor and can be initiated frombrain tumor stem-like cells (BTSCs) [8–10]. The capability ofBTSCs to sustain brain tumor growth apparently lies in theiractive self-renewal and/or suppressed cell differentiation [2].Several major signaling pathways that are critical in brain de-velopment have also been implicated in tumorigenesis, includ-

ing bone morphogenetic protein (BMP) [11], Notch [12], SonicHedgehog (SHH) [13, 14], epidermal growth factor receptor(EGFR) [15, 16], PTEN/ phosphoinositide 3-kinase (PI3K)/mammalian target of rapamycin (mTOR) [12, 15, 17], platelet-derived growth factor receptor (PDGFR) [18], and OLIG2 [19].Recently, a gene expression profiling of gliomas has shown thatSHH signaling is active in a subset of gliomas [20]. This studyfurther showed that SHH signaling is essential for glioma CSCself-renewal and CSC-initiated brain tumor growth [20]. It ispostulated that the relatively homogeneous population of CSCs,rather than the heterogeneous tumor cells, may reveal key mech-anisms of tumor initiation and propagation of primary tumorsand hence predict tumor prognosis, therapy, and drug response.

Here, we performed gene expression profiling of BTSCsand unveiled salient signaling pathway signatures. We iden-tified both SHH signaling-dependent and -independentBTSCs that can initiate brain tumors retaining their respec-tive characteristics of signaling dependence. BTSC prolifer-ation could be abrogated in a pathway-dependent fashion invitro and in an intracranial tumor model in SCID mice.Furthermore, hyperactive SHH-GLI signaling in PTEN-co-expressing tumors was associated with reduced survivaltimes in glioblastoma patients.

Author contributions: Q.X.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing,final approval of manuscript; X.Y. and G.L.: collection and/or assembly of data, data analysis and interpretation, final approval of manuscript;K.L.B.: financial support, provision of study material or patients, final approval of manuscript; J.S.Y.: conception and design, provision ofstudy material or patients, data analysis and interpretation, manuscript writing, final approval of manuscript.

Correspondence: John S. Yu, M.D., Maxine Dunitz Neurosurgical Institute, Department of Neurosurgery, Cedars-Sinai Medical Center, Suite800 East, 8631 West 3rd Street, Los Angeles, California 90049, USA. Telephone: 310-423-0845; Fax: 310-423-1038; e-mail:[email protected] Received May 10, 2008; accepted for publication September 3, 2008; first published online in STEM CELLS EXPRESS

September 11, 2008. ©AlphaMed Press 1066-5099/2008/$30.00/0 doi: 10.1634/stemcells.2008-0459

CANCER STEM CELLS

STEM CELLS 2008;26:3018–3026 www.StemCells.com

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

MATERIALS AND METHODS

Primary Brain Tumor Cell CulturePrimary brain tumor spheres were cultured as previously described[1]. Briefly, brain tumor stem-like cells were grown in Dulbecco’smodified Eagle’s medium (DMEM)/Ham’s F-12 medium (F12)supplemented with B-27 (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), 20 ng/ml basic fibroblast growth factor (bFGF),and 20 ng/ml epidermal growth factor (EGF) (Peprotech, RockyHill, NJ, http://www.peprotech.com). Alternatively, dispersed braintumor stem-like cells were grown on laminin-coated surface in thesame medium as described above. Primary human fetal neural stemcells were derived from primary cells obtained from Cambrex (EastRutherford, NJ, http://www.cambrex.com). The GBM cell line andadherent primary glioma cells were cultured in DMEM/F12 con-taining 10% fetal bovine serum. Some frozen primary GBM tissueswere used to compare gene expression profiles of BTSCs and theirparental tumors.

Subsphere Formation AssaysSubsphere formation assays were described before [1]. Briefly, cellsin single-cell suspension were diluted and plated at a density ofthree to five cells per well. After plating, the cells were observed,and only wells containing a single cell were considered. Cells werefed by changing half of the medium every 2 days. The wells werescored for sphere formation after 14 days.

Reverse Transcription-Polymerase Chain Reactionand Real-Time Polymerase Chain Reaction AnalysisTotal RNAs were isolated using the RNeasy Mini Kit (Qiagen,Hilden, Germany, http://www1.qiagen.com). cDNAs were synthe-sized by oligo(dT)-priming methods. Real-time polymerase chainreaction (PCR) was performed using the SYBR Green Supermix(Qiagen) according to the manufacturer’s instructions (primers usedare listed in supplemental online Table 1). Expression levels of�-actin or glyceraldehyde-3-phosphate dehydrogenase were usedfor normalization and quantification of gene expression levels.

ImmunoblottingCells were lysed and homogenized in RIPA lysis buffer containingfresh protease inhibitors by standard procedures. Protein concentra-tions were quantified with the BCA protein assay kit (Pierce,Rockford, IL, http://www.piercenet.com), and 30 �g of proteinswere separated in 4%–12% SDS-polyacrylamide gel electrophoresisgels, transferred to polyvinylidene difluoride membranes, and hy-bridized with a SUFU antibody (Santa Cruz Biotechnology Inc.,Santa Cruz, CA, http://www.scbt.com) by standard procedures.Signals were detected by chemiluminescence using ECL detectionreagents (GE Health, Piscataway, NJ, http://www.gelifesciences.com).

Tissue Array and ImmunohistochemistryGlioblastoma (grade IV) tissue arrays were obtained from USBiomax (Rockville, MD, http://www.biomax.us). Paraffin-embed-ded tissue sections were dried for 2 hours at 60°C, dewaxed inxylene, and rehydrated with distilled water. Incubation with anti-human GLI1 (1:500; Millipore, Billerica, MA, http://www.millipore.com), anti-phospho-AKT (1:100; Cell Signaling Technol-ogy, Beverly, MA, http://www.cellsignal.com), and anti-PTEN (1:200; Santa Cruz Biotechnology) was performed overnight at 4°C.Immunodetection was performed using the Elite Vector Stain ABCSystem (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com). The expression levels of PTEN and nuclear GLI1were detected by immunohistochemistry and blindly scored on ascale of 1–10. Samples were separated into two halves on the basisof phospho-AKT scores, and average GLI scores of the two groupswere compared.

Cell Proliferation, 5-Bromo-2�-DeoxyuridineLabeling, and Apoptosis AssaysFor cell proliferation assays, brain tumor stem-like cells wereseeded in growth medium at a density of 5 � 103 cells per 96-wellplate with or without inhibitors. At the indicated time of culture, thenumber of cells was determined using WST-1 colorimetric assay(Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com). For the 5-bromo-2�-deoxyuridine (BrdU) labelingassays, cells were cultured for 48 hours after transfection with apulse of BrdU (10 �M) for the last 3 hours. The cells were thenfixed and stained with anti-BrdU antibody. The proportion ofBrdU(�) cells was determined by fluorescence-activated cell sort-ing analysis performed in triplicate, in at least two independentexperiments. Apoptosis assays were performed using the annexin Vapoptosis kit (Roche Diagnostics).

RNA Interference and Cell TransfectionShort interfering RNA (siRNA) (5�-CUCCACAGGCAUACAG-GAUUU-3�) against human GLI1 [21], with or without fluores-cence-labeled Alexa Fluor 488 at the 3� end, was obtained byQiagen custom synthesis. Pretested control siRNA was obtainedfrom Qiagen. Brain tumor stem-like cells were transfected usingHiPerfect (Qiagen) following the manufacturer’s instructions.Briefly, brain tumor spheres were triturated into small aggregatesand cultured for 2 hours in the growth medium without heparin.For every 105 cells, 0.5 �g of control siRNA or GLI1 siRNA wasdiluted and mixed with 24 �l of HiPerfect reagent. After mixingand incubation for 10 minutes, the transfection mixture wasadded to the cells. Cells were examined for transfection effi-ciency at 8 –16 hours and were used for functional assays 24hours after transfection.

Intracranial Cell Transplantation into Nude MiceAthymic nude mice (nu/nu; 6 – 8 weeks old; Charles River Lab-oratories, Wilmington, MA, http://www.criver.com) were anes-thetized with i.p. ketamine and medetomidine and were ster-eotactically implanted with glioblastoma sphere cells (50,000 permouse) in the right striatum. The experiment was repeated onceunder identical conditions. The implanted mice were euthanizedat 6 –12 weeks, followed by intracardiac perfusion-fixation with4% paraformaldehyde. Brain tissues were retrieved for frozensection and analysis. For BrdU incorporation assays, mice wereinjected i.p. with 10 mg/ml BrdU (Sigma-Aldrich, St. Louis,http://www.sigmaaldrich.com) at 50 mg/kg once per day for 2days before sacrifice. BrdU-positive cells were detected by im-munofluorescence using anti-BrdU antibody (1:500; BD Bio-sciences, San Diego, http://www.bdbiosciences.com). All ani-mals used were experimented on in strict accordance with theInstitutional Animal Care and Use Committee guidelines en-forced at the Cedars-Sinai Medical Center.

Treatment of Brain Tumors Using siRNA andRapamycinBrain tumor stem-like cells were transfected with control siRNA orGLI1 siRNA (1.0 �g per 2 � 105 cells) at 2 hours prior totransplantation. For rapamycin treatment, cells were incubated withrapamycin (100 nM) for 2 hours prior to transplantation. At days 7and 14 postoperation, animals were given siRNA (1.0 �g), rapa-mycin (200 ng), or the control, intratumorally.

GBM Patient Sample AnalysisFrozen tissues were obtained from 55 GBM patients (age range,29 –75) who were treated at Cedars-Sinai Medical Center, LosAngeles, CA. Tumor specimens were obtained according to aprotocol approved by the Institutional Review Board of Cedars-Sinai Medical Center. The mRNA expression of PTEN, SHH,PTCH1, and GLI1 in 55 GBM tissues was determined by real-time PCR assays as described above. The expression levels ofSHH, PTCH1, and GLI1 in PTEN-expressing (1%–100% of themaximal value) and PTEN-deficient (less than 1% of the maxi-mal value) GBM tissues were analyzed. There was no significant

3019Xu, Yuan, Liu et al.

www.StemCells.com

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

difference in age between groups (PTCH1: 50 � 11 vs. 51 � 11,p � .78; SHH: 49 � 10 vs. 52 � 1, p � .60; GLI1: 54 � 9 vs.47 � 11, p � .19). For the survival function analysis, thePTEN-expressing samples were grouped into two groups, greaterthan or less than the median values of SHH, PTCH1, and GLI1expression, respectively, and the survival function was analyzedbetween the two groups for SHH, PTCH1, and GLI1 usingKaplan-Meier analysis. Survival was measured from the date ofsurgery of first diagnosis. All patients were diagnosed with denovo glioblastoma and underwent treatment with external beamradiation therapy to 60 Gy and chemotherapy.

Statistical MethodsTo test whether variables differed across two groups, we usedStudent’s t test (unpaired, two-tailed). The Kaplan-Meier (KM)analysis of survival distributions was performed using SPSS soft-ware (SPSS, Chicago, http://www.spss.com).

RESULTS

Pathway Gene Expression Profiles for Brain TumorStem-Like CellsPrevious studies in murine models have shown that the SHHsignaling pathway plays a key role in the neural developmentand in neural stem cell self-renewal. We show that primaryhuman fetal neurospheres grown in the EGF- and bFGF-supple-mented medium express neural stem cell (NSC) markers CD133and nestin (supplemental online Fig. 1). When the human NSCswere incubated with the SHH signaling inhibitor cyclopamine,cyclopamine dose-dependently inhibited cell proliferation. Fur-thermore, overexpression of GLI1 in human NSCs, using anadenoviral vector, enhanced human NSC proliferation, suggest-ing that SHH signaling is essential for human NSC proliferation(supplemental online Fig. 1). Although both NSCs and BTSCsexpress CD133 and form spheres in culture, it is unclear whetherthe requirement for SHH signaling in NSC proliferation isshared by BTSCs.

Previously we had isolated BTSCs from human GBM tis-sues and demonstrated that BTSCs, as opposed to the differen-tiated brain tumor cell counterparts, can generate tumors inimmune-deficient mice [1]. It had been shown that BTSCs

derived from glioblastoma manifest gene expression patternsdifferent from those of corresponding serum-cultured cell linesand more closely recapitulate the phenotypes and biology ofhuman glioblastoma [9]. Using real-time PCR analysis, weexamined the expression of selected genes involved in majorsignaling pathways in BTSCs and in comparison with matchedtumor cells and NSCs. As an example shown in Figure 1A,BTSC1 is different from both tumor cells (GBM1) and NSC intheir gene expression patterns. For instance, the SHH pathwaygenes had the highest expression levels in BTSC1, whereasNSCs expressed the highest level of BMI1. Although the geneexpression patterns of GBM1 and BTSC1 both showed highlevels of expression of GLI1, suggesting active SHH signalingin these cells, only the BTSC1 profile revealed stronger andmore consistent gene expression levels of the ligand (SHH), thereceptor (PTCH1), and the downstream effectors (N-Myc andCyclin D2). Therefore, BTSCs manifest gene expression pro-files different from that of total tumor cells.

Next, the gene expression profiles of five BTSCs fromdifferent GBM patients were compared and analyzed. All fiveBTSCs had a distinct gene expression profile (Fig. 1B). Fur-thermore, these BTSCs can be separated into two groups on thebasis of their gene expression patterns. BTSCs 1, 2, and 3 formone group characterized by high expression levels and activitiesin SHH, Notch, and platelet-derived growth factor signalingpathways. The other group, including BTSCs 4 and 5, showedlow expression of genes in these signaling pathways. Instead,these BTSCs seemed to have lost expression of one or more keytumor suppressor genes, such as PTEN, p21CIP, or BMP recep-tor. To confirm that SHH signaling is active only in BTSCs 1–3,the expression of GLI1 protein was examined in each BTSC. Asshown in Figure 1C, strong nuclear immunostaining of GLI1was present only in BTSCs 1–3 and not in BTSCs 4 and 5(supplemental online Fig. 2). Thus, the tumor-initiating poten-tial of BTSCs is associated with distinct molecular programs.

Identification of SHH Pathway-Dependent and-Independent BTSCsThe robust expression of the SHH-GLI1 pathway components ina group of BTSCs implied that hyperactive SHH-GLI1 signalingmay play a role in cell proliferation of these BTSCs. We sought

Figure 1. Gene expression profiling of different BTSCs revealed distinct genetic programs and signaling pathway activities in brain tumor cells. (A):A molecular signature of BTSCs compared with that of human NSCs and their parent brain tumor cells (without culture). Both BTSCs and NSCs weregrown in identical defined media. Frozen primary GBM tissues were used to compare gene expression profiles of BTSCs and their parental tumors.The mRNA expression levels for selected genes were quantified by real-time polymerase chain reaction (PCR). (B): Heat map image showing geneexpression profiling of five different BTSCs. mRNA expression levels were quantified using real-time PCR and were normalized to glyceraldehyde-3-phosphate dehydrogenase expression. For each gene, the expression values were assigned a color using the minimum/maximum method. Red andgreen colors represent high and low mRNA expression levels, respectively. BTSC1–BTSC5 indicate five different BTSCs. (C): Detection of GLI1protein expression and subcellular localization using immunocytochemistry assays. Panels 1–5 indicate BTSC1–BTSC5, respectively. Panel 6 showsan LN18 cell as a positive control. Abbreviations: BTSC, brain tumor stem-like cell; GBM, glioblastoma multiforme; hNSC, human fetal neural stemcell; NSC, fetal neural stem cell; SHH, Sonic Hedgehog.

3020 Hedgehog Signaling in Glioblastomas

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

to inhibit this signaling by knocking down the downstreamtranscription factor GLI1 using RNA interference to determinethe effect of signal inhibition on BTSC proliferation and self-renewal. The expression of GLI1 in tumor cells was suppressedby 90% using GLI1-targeting siRNA (supplemental online Fig.3). siRNA of the same sequence had previously been used toeffectively silence GLI1 expression [21]. Knockdown of GLI1significantly inhibited BTSC1 cell proliferation but had noeffect on BTSC4 (Fig. 2A). When both siRNA against GLI1 andsiRNA against GLI2 were combined, the inhibiting effects oncell proliferation were similar (supplemental online Fig. 4). Wealso showed that GLI1 siRNA could inhibit cell proliferation inBTSC3, which has a SHH signaling signature, but not in SHHnonexpressing BTSC5 (supplemental online Fig. 5). To deter-mine whether SHH signaling may be important for BTSC self-renewal, we performed subsphere formation assays. As shownin Figure 2A, the percentage of subsphere-forming BTSC1 cells,but not BTSC4 cells, was significantly reduced in GLI1 knock-down cells relative to control cells, suggesting that SHH signal-ing may promote BTSC1 cell proliferation through enhancedcell self-renewal.

SHH signaling seemed to enhance BTSC proliferationthrough increased cell cycle progression, rather than promotingcell survival, as GLI1 knockdown in BTSC1 decreased BrdUincorporation rate from 32% to 19% (Fig. 2B). GLI1 knock-down in BTSC1 did not increase cell apoptosis, but ratherdecreased it (Fig. 2C). To further test whether SHH signaling isindeed critical for BTSC1 proliferation, cells were transiently

transfected with a vector expressing SUFU, a negative regulatorof SHH signaling (Fig. 2D). Overexpression of SUFU in BTSC1significantly decreased cell proliferation at both 72 and 96 hoursbut had little effect on BTSC4 cells (Fig. 2E). Thus, we haveidentified both SHH signaling-dependent BTSCs and SHH sig-naling-independent BTSCs.

Inhibition of Tumor Growth from SHHPathway-Dependent BTSCs by Targeting GLI1These BTSCs had previously been shown to initiate tumorgrowth in immune-compromised mice [1]. To test whethertumor growth from BTSCs can be inhibited by targeting GLI1,BTSC1 (SHH-dependent) and BTSC4 (SHH-independent) wereeach implanted into nude mice after transfection with GLI1siRNA or control siRNA, followed by intratumor injection ofGLI1 siRNA or control. Animals implanted with BTSC1 orBTSC4 carried brain tumors in less than 12 weeks (Fig. 3A).Although BTSC4-initiated tumors were more malignant andinvasive, the BTSC1 tumors were more confined (Fig. 3A).GLI1 siRNA treatment significantly reduced BTSC1 tumor vol-ume (Fig. 3B) but had no effect on the BTSC4 tumor size.Characteristically, BTSC1-initiated brain tumors, but notBTSC4 tumors, contained a broad layer of GLI1-positive cellsalong the outer layer of the tumors, which were greatly reducedin GLI1 siRNA-treated tumors (Fig. 3C). Consistent with thetumor volume reduction, GLI1 siRNA treatment significantlyreduced GLI1-positive cells and proliferating cells in BTSC1

Figure 2. Sonic hedgehog-GLI1 signaling is required for the cell proliferation of some BTSCs. (A): BTSC1 (upper panel) and BTSC4 (middle panel)were transiently transfected with short interfering RNA (siRNA) (1.67 �M) targeting human GLI1 gene expression (f) or control siRNA (�),followed by cell proliferation assays at the indicated time points. �, p � .04 by Student’s t test. Bottom panel: subsphere formation assay resultsshowing percentages of dissociated BTSCs treated with GLI1 siRNA (open bars) or control siRNA (filled bars) forming secondary spheres. Resultsare from three independent experiments. ��, p � .0002 by Student’s t test. (B): GLI1 knockdown in BTSCs did not increase apoptotic cell death.Percentage of apoptotic cell death with (filled bars) or without (open bars) GLI1 knockdown were measured using an annexin V/PI staining kit. (C):BrdU incorporation assays in BTSCs without (upper panel) or with (lower panel) GLI1 knockdown. Results are representative of two experiments.(D): Misexpression of human SUFU in BTSCs as shown in Western blots. BTSC1 (lanes 1 and 2) and BTSC4 (lanes 3 and 4) were transientlytransfected with hSUFU-expressing plasmid (lanes 2 and 4) or control plasmid (lanes 1 and 3). Cell lysates were analyzed at 48 h. (E): Effect of SUFUexpression on BTSC proliferation. Cells were transfected with SUFU vector (filled bars) or empty vector (open bars) for 24 h, followed by WST-1assays at the indicated times. Misexpression of SUFU in BTSC1 (upper panel), but not in BTSC4 (lower panel), inhibited cell proliferation. Resultsare from two independent experiments. ��, p � .001 (72 h) and p � .003 (96 h) by Student’s t test. Abbreviations: BrdU, 5-bromo-2�-deoxyuridine;BTSC, brain tumor stem-like cell; h, hours; PI, propidium iodide.

3021Xu, Yuan, Liu et al.

www.StemCells.com

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

tumors, as indicated by BrdU incorporation detection (Fig. 3D).Therefore, targeting GLI1 significantly inhibited SHH-GLI1signaling-dependent brain tumor growth.

PI3K-mTOR Signaling Is Critical for the Growth ofBrain Tumors Derived from BTSCsThe identification of SHH signaling-independent BTSCs andbrain tumors prompted us to examine alternative signalingpathways. The PTEN loss-of-function and active PI3K-AKTpathway has been shown to be prevalent and critical inmalignant brain tumors [12, 15, 17]. PI3K-AKT signaling isrequired in both SHH-dependent and -independent BTSCproliferation, as several PI3K-AKT pathway inhibitors, in-cluding LY294002 and AKT inhibitor VIII, inhibited cellproliferation in both BTSC1 and BTSC4 (supplemental on-line Fig. 6). A key downstream effector of PI3K-AKT sig-naling is mTOR [22]. Previous studies had implicated mTORsignaling in the progression of a wide range of cancers,including kidney [23], pancreatic [24], breast [25–27], pros-tate [28], lung [29], and glioma [17]. To test whether mTORsignaling is important in BTSC proliferation, we incubatedBTSC1 and BTSC4 with or without rapamycin, an inhibitorof mTOR. As shown in Figure 4A, rapamycin dose-depen-dently inhibited the cell proliferation of both BTSC1 andBTSC4. Therefore, PI3K-mTOR signaling is critical for thecell proliferation of both SHH-dependent and -independentBTSCs.

To test whether suppressing mTOR signaling with rapamy-cin can inhibit BTSC-initiated brain tumor growth, BTSC4 wastransplanted intracranially into nude mice, with or without rapa-mycin treatment, immediately before surgery and again at day 7and day 14 postoperation. Whereas mock-treated animals de-veloped highly invasive brain tumors, the rapamycin-treatedtumors had a more defined boundary (Fig. 4B). Furthermore,rapamycin treatment increased animal survival (p � .004) (Fig.4C). Consistent with previous data showing that BTSC4, but notBTSC1, had lost PTEN expression, there was little PTEN ex-pression in the BTSC4-initiated brain tumors compared with theBTSC1 brain tumors (Fig. 4B). These data suggested thatmTOR signaling is critical for SHH-signaling-independent braintumor growth.

Differential SHH-GLI Signaling Activities inPTEN-Expressing and PTEN-Deficient GBMsTo survey the general status of SHH signaling activity and itsrelation to PTEN expression in human GBM tissues, weexamined the expression of GLI1 and PTEN using tissuearrays with a panel of 40 GBM tissues. There were wideranges of expression levels of GLI1 and PTEN, but highGLI1 expression was present in most GBM tissues (Fig. 5A).Interestingly, the GLI1 expression level in PTEN-expressingGBM tissues was significantly higher than that in PTEN-deficient GBM tissues (Fig. 5B). To further test the associ-ation of SHH signaling activity with PTEN expression status,

Figure 3. Suppression of GLI1 expressionas a potential therapy for sonic hedgehogsignaling-active brain tumors. (A): BTSC1and BTSC4 were implanted into the brainsof nude mice with or without GLI1 siRNAtreatment. Brain tissues were prepared andanalyzed with H&E staining at 12 weeks.(B): GLI1 knockdown inhibited BTSC1 tu-mor growth as indicated by the tumor vol-ume reduction (n � 5). ��, p � .01 byStudent’s t test. (C): Brain tumors derivedfrom BTSC1, but not those derived fromBTSC4, manifested strong GLI1 expression,particularly in the tumor border regions.Brain tumors from BTSC1 with GLI1siRNA treatment showed significantly re-duced GLI1 expression. Note the nuclearstaining of GLI1 in the GLI1-positive cells(insets). (D): GLI1 siRNA inhibited BTSC1proliferation in vivo. At 12 weeks postim-plantation, tumor cell proliferation withoutor with GLI1 siRNA treatment was assessedby BrdU incorporation assays. Abbreviations:BrdU, 5-bromo-2�-deoxyuridine; BTSC, braintumor stem-like cell; siRNA, short interfer-ing RNA.

3022 Hedgehog Signaling in Glioblastomas

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

we determined the gene expression levels of SHH, PTCH1,and GLI1 in tumor tissues from another 55 GBM patients.These samples were separated into a PTEN-expressing groupand a PTEN-deficient group on the basis of their PTENexpression levels. As shown in Figure 5B, PTEN-deficient

tumors manifested a significantly higher level of PTCH1gene expression than PTEN-expressing tumors. However, theexpression levels of SHH and GLI1 were significantly higherin PTEN-expressing cells than in PTEN-deficient cells (Fig.5B). Therefore, there is a correlation between SHH-GLI

Figure 4. Mammalian target of rapamycin(mTOR) activity was required for brain tu-mor growth. (A): Rapamycin, an mTOR in-hibitor, dose-dependently inhibited BTSC1(upper panel) and BTSC4 (lower panel) cellproliferation. BTSCs were incubated withRap10, Rap100, or Rap2k for specified pe-riods of time. Cell proliferation was deter-mined using WST-1 assays. (B): BTSC4formed invasive brain tumors that had littlePTEN expression compared with BTSC1-derived brain tumors. Rapamycin-treated tu-mor showed a less invasive morphology.(C): Rapamycin treatment on animals bear-ing BTSC4 brain tumors had a survival ben-efit over the control group. p � .004 byKaplan-Meier analysis (n � 10). Rapamycintreatment was administered immediately be-fore surgery and again intratumor at days 7and 14 postoperation. Abbreviations: BTSC,brain tumor stem-like cell; DMSO, dimethylsulfoxide; h, hours; Rap10, 10 nM rapamy-cin; Rap100, 100 nM rapamycin; Rap2k, 2�M rapamycin.

Figure 5. SHH signaling activities in PTEN-positive brain tumors correlated with patient survival. (A): Variable GLI1 expression in human GBMtissues. Forty GBM tissues and normal brain tissues in a GBM tissue array were stained with GLI1 and PTEN antibodies and were scored.Representative pictures show GLI1 staining results for GBM tissues with different scores. Right: the average GLI1 expression level in thePTEN-expressing GBM tissues (open bar) was significantly higher than that in the PTEN-deficient GBM tissues (filled bar). �, p � .048. (B): Theexpression levels of SHH, PTCH1, and GLI1 in PTEN-expressing and PTEN-deficient GBM tissues. The mRNA expressions of PTEN, SHH, PTCH1,and GLI1 in 55 GBM tissues (ages 29–75) were determined by real-time polymerase chain reaction and were grouped (described in Materials andMethods). The gene expression levels in PTEN-expressing (open bars) and PTEN-deficient (closed bars) samples were compared. �, p � .006(PTCH1), p � .008 (SHH), and p � .004 (GLI1) by Student’s t test. (C): Effect of SHH-GLI signaling on the survival of PTEN-expressing GBMpatients. GBM patient samples (age-matched) were grouped into lower expression (solid line) and higher expression (dotted line) on the basis of theexpression levels of SHH (left panel), PTCH1 (middle panel), and GLI1 (right panel). The survival curves were generated using Kaplan-Meier analysismethod. Abbreviations: GBM, glioblastoma multiforme; SHH, Sonic Hedgehog.

3023Xu, Yuan, Liu et al.

www.StemCells.com

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

signaling activity and PTEN activity in GBM tissues; SHH-GLI signaling activity appears to be higher in PTEN-express-ing tumors than in PTEN-deficient tumors.

Hyperactive SHH-GLI Signaling in Brain Tumorswith PTEN Coexpression Is Associated withReduced Survival TimeSince SHH-GLI signaling activity is high in PTEN-express-ing brain tumors and this signaling pathway is also essentialfor BTSC self-renewal and proliferation, we next studied thepotential connection between SHH-GLI signaling activityand GBM patient survival time. For this purpose, the KMsurvival function analysis was performed within the PTEN-expressing group of GBM samples, using the gene expressionlevels of SHH, PTCH1, or GLI1 as the variables. As shownin Figure 5C, higher expression levels of SHH, PTCH1, andGLI1 were associated with reduced survival time. For thegroups with high expression levels of SHH, PTCH1, andGLI1, the median survival times were 48, 52, and 53 weeks,respectively, compared with 70 weeks for the groups withlower SHH, PTCH1, and GLI1 expression. In particular, thesurvival benefits for the groups with lower expression ofSHH and PTCH1 are statistically significant (Fig. 5C). To-gether, these data suggest that active SHH signaling is asso-ciated with poor prognosis for patients bearing PTEN-coex-pressing glioblastoma.

DISCUSSION

This study has generated the first molecular signatures ofGBM cancer stem cell proliferation to identify both SHHsignaling-dependent and -independent BTSCs and brain tu-mors. Previously, gene expression profiling using GBM celllines of primary tumor cells has identified numerous, yetdiverse, signature genes or gene sets [30 –32]. Using genet-ically homogeneous BTSCs, tumor classifications and theirmolecular signatures can be revealed with relatively smallsample sizes because of reduced noise levels. We also showthat the growth of brain tumors initiated from SHH-depen-dent BTSCs can be inhibited by targeting GLI1, whereas bothSHH-dependent and -independent brain tumor growth re-quires PI3K-mTOR signaling. Furthermore, we found thatthe expression levels of SHH and GLI1 are significantlyhigher in PTEN-expressing cells than in PTEN-deficientcells. Finally, we present evidence indicating that hyperactiveSHH-GLI signaling in human brain tumors with PTEN-co-expression is associated with reduced survival time.

SHH signaling pathway is a key regulatory mechanism inneural development, and abnormal SHH signaling has beenimplicated in the tumorigenesis of pediatric brain tumors,such as medulloblastoma [33–36]. Recently, SHH-GLI sig-naling was also found to be essential for glioma stem cellself-renewal [20]. Our study of signaling in GBM, however,demonstrated that there are both SHH signaling-dependentand -independent brain tumors and BTSCs. SHH signaling-dependent BTSCs, such as BTSC1, express high levels ofSHH, PTCH1, and GLI1 and initiate brain tumors character-ized by a cancer stem cell zone of GLI1-positive cells. Bothoverexpression of SUFU and knockdown of GLI1 expressioninhibited BTSC1, but not BTSC4, cell proliferation. Unlikethe study by Clement et al. [20], our study found that cyclo-pamine could not inhibit BTSC growth at normal dose levels(no more than 10 �M). This discrepancy could be due toexperimental conditions, or it could reflect a difference inligand dependence, cancer genetics, or drug resistance of the

cancer cells. However, we could significantly inhibit SHHsignaling by targeting GLI1 with siRNA both in vitro and invivo. Because the duration of siRNA action is transient, theinhibition of BTSC1 by siRNA in culture also showed tran-sient effects. Future work using stable inducible systemswould provide further insight into this mechanism. Consis-tent with our data demonstrating a critical role for SHHsignaling in some BTSCs, a recent study showed that inhi-bition of SHH signaling could deplete glioblastoma stemcells [37].

In our study, gene expression profiling indicated that activeSHH signaling in the BTSCs is associated with high activities inthe signaling pathways of Notch, PDGFR, and OLIG2, whichare also known to be important in neural development andneural stem cell functions. Deregulated signaling of BMP [11],Notch [12], PDGFR [18], and OLIG2 [19] has been implicatedin the development and progression of GBM. For instance, it hasbeen shown that the BMP signaling pathway is suppressed inGBMs and that exogenous BMPs can arrest BTSC-initiatedtumor growth by blocking cell differentiation [11]. These dis-parate signaling pathways may contribute to GBM developmentindependently. Alternatively, SHH signaling may interact witheach of these signaling pathways. For instance, it has beenshown that activation of the SHH pathway could inducePDGFR� and OLIG2 expression during neural development[38, 39]. It remains to be seen whether PDGFR signaling andOLIG2 expression are regulated by SHH signaling in theseBTSCs.

The identification of both SHH signaling-dependent and-independent brain tumors in this study suggested that there aremolecularly distinct subclasses of GBMs that have an effect onprogression and prognosis. These findings are reminiscent of arecent study in which high-grade gliomas were classified into aproneural subclass and a mesenchymal subclass, resembling twostages in neurogenesis [12]. In that study, Phillips et al. discov-ered a prognostic model using PTEN-AKT and Notch signaturesto predict poor versus better glioma prognosis, respectively [12].In our study, the SHH signaling-dependent brain tumors showedhigh activities in Notch and PDGFR pathways, whereas theSHH signaling-independent brain tumors showed PTEN defi-ciency and high activity in the PI3K-AKT pathway. Our studyemphasized that BTSCs with distinct signaling patterns deter-mined the different GBM phenotypes and progression. Thepresence of SHH-independent GBM may also partly explain theresult in a recent study showing that SHH signaling is active ingrade II–III gliomas but not GBMs [40].

The pivotal role of PTEN expression status in GBMdevelopment and progression is further supported in ourstudy using BTSCs. PTEN expression is widely lost in GBMand other malignant tumors [7, 41– 44]. Coexpression ofPTEN with active EGFR in GBM determines clinical cancerresponses to an EGFR inhibitor [15]. PTEN loss and AKTactivation are associated with more invasive and malignantcancer and poor prognosis for GBM patients [12]. Further-more, a drug that potently inhibits GBM cell proliferationwas found to be an inhibitor of both PI3K� and mTOR [17].We found that there are both PTEN-expressing and PTEN-deficient BTSCs and that all BTSCs tested are inhibited byPI3K and AKT inhibitors.

Furthermore, we found that proliferation of the BTSCswas most strongly inhibited by rapamycin, an mTOR inhib-itor. Rapamycin also inhibited BTSC-initiated brain tumorand extended animal survival. mTOR is a critical effectordownstream of growth factor (through PI3K-AKT) and nu-trient signaling pathways [22]. mTOR has also emerged as akey target that is commonly deregulated in human cancers[22]. Recently, a combination therapy with rapamycin and an

3024 Hedgehog Signaling in Glioblastomas

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

EGFR inhibitor inhibited the growth of both peripheral lungcancer and chemoresistant bronchial lung cancer [29]. Clin-ical trials with mTOR inhibitors in the treatment of GBMhave yielded contradictory results [45– 47]. The inadequatepenetration of the drug across the blood-brain barrier into thetumor sites was cited as the reason for the failure [46]. Wedemonstrated that rapamycin effectively inhibited BTSCgrowth and that intratumor administration of rapamycin pro-longed animal survival. In addition, combination therapy ofan mTOR inhibitor and an EGFR/vascular endothelial growthfactor receptor inhibitor had been reported to offer increasedbenefit in glioma treatment [48].

Our study presents evidence indicating genetic interactionbetween the SHH signaling pathway and PTEN in human gli-oblastoma. After analysis of a tissue array of GBM tumors andadditional frozen GBM tissues, we found that the average SHHsignaling activity is significantly higher in PTEN-expressingtumors than in PTEN-deficient tumors. Although the expressionlevels of SHH and GLI1 are significantly higher in PTEN-expressing tumors, the expression level of PTCH1, a SHHreceptor, is significantly lower. PTCH1 is tumor suppressor anda negative regulator of SHH signaling in the absence of theligands. These results suggest that hyperactive SHH-GLI sig-naling is critical for brain tumor cell proliferation only when thetumor suppression mechanism of PTEN pathway is intact. WhenPTEN expression is lost through genetic or epigenetic alteration,alternative signaling pathways, including the PI3K-AKT-mTORpathway, are usually activated, leading to bypassing of therequirement for an active SHH-GLI signaling pathway. PTENexpression alone, however, was not a strong indicator of malig-nancy in our study, as there is no association between PTENexpression in GBM and patient survival. Finally, we have

shown that higher SHH signaling in PTEN-expressing GBM isassociated with reduced survival time, further supporting acritical role for SHH signaling in the PTEN-coexpressing subsetof GBM tumors.

CONCLUSION

In summary, we demonstrate that the distinct cell proliferationsignaling dependence of glioblastoma can be uncovered usinggene expression profiling with brain tumor stem-like cells. Ac-tivated SHH-GLI signaling is required for the growth of glio-blastomas with a PTEN-coexpression signature, whereas thePI3K-AKT-mTOR signaling pathway is generally critical forglioblastoma growth. These findings imply that future develop-ment of glioblastoma treatment based on signaling pathwayinhibitors may benefit from identification of tumor molecularsignatures and consideration of the genetic context.

ACKNOWLEDGMENTS

We thank Gretchen Duvall and Akop Seksenyan for critical read-ing of the manuscript and helpful advice. We also thank MarioCastro for providing the adenoviral vectors. This work was fundedin part by NIH grants NS048959 and NS048879 (to J.S.Y.).

DISCLOSURE OF POTENTIAL CONFLICTS

OF INTEREST

The authors indicate no potential conflicts of interest.

REFERENCES

1 Yuan X, Curtin J, Xiong Y et al. Isolation of cancer stem cells from adultglioblastoma multiforme. Oncogene 2004;23:9392–9400.

2 Singh SK, Hawkins C, Clarke ID et al. Identification of human braintumour initiating cells. Nature 2004;432:396–401.

3 Tang DG, Patrawala L, Calhoun T et al. Prostate cancer stem/progenitorcells: Identification, characterization, and implications. Mol Carcinog2007;46:1–14.

4 O’Brien CA, Pollett A, Gallinger S et al. A human colon cancer cellcapable of initiating tumour growth in immunodeficient mice. Nature2007;445:106–110.

5 Ricci-Vitiani L, Lombardi DG, Pilozzi E et al. Identification and expan-sion of human colon-cancer-initiating cells. Nature 2007;445:111–115.

6 Liao MJ, Zhang CC, Zhou B et al. Enrichment of a population ofmammary gland cells that form mammospheres and have in vivo repop-ulating activity. Cancer Res 2007;67:8131–8138.

7 Karnoub AE, Dash AB, Vo AP et al. Mesenchymal stem cells withintumour stroma promote breast cancer metastasis. Nature 2007;449:557–563.

8 Bao S, Wu Q, McLendon RE et al. Glioma stem cells promote radiore-sistance by preferential activation of the DNA damage response. Nature2006;444:756–760.

9 Lee J, Kotliarova S, Kotliarov Y et al. Tumor stem cells derived fromglioblastomas cultured in bFGF and EGF more closely mirror the phe-notype and genotype of primary tumors than do serum-cultured cell lines.Cancer Cell 2006;9:391–403.

10 Furnari FB, Fenton T, Bachoo RM et al. Malignant astrocytic glioma:Genetics, biology, and paths to treatment. Genes Dev 2007;21:2683–2710.

11 Piccirillo SG, Reynolds BA, Zanetti N et al. Bone morphogenetic pro-teins inhibit the tumorigenic potential of human brain tumour-initiatingcells. Nature 2006;444:761–765.

12 Phillips HS, Kharbanda S, Chen RH et al. Molecular subclasses ofhigh-grade glioma predict prognosis, delineate a pattern of diseaseprogression, and resemble stages in neurogenesis. Cancer Cell 2006;9:157–173.

13 Dahmane N, Sanchez P, Gitton Y et al. The Sonic Hedgehog-Gli path-

way regulates dorsal brain growth and tumorigenesis. Development2001;128:5201–5212.

14 Katayam M, Yoshida K, Ishimori H et al. Patched and smoothenedmRNA expression in human astrocytic tumors inversely correlates withhistological malignancy. J Neurooncol 2002;59:107–115.

15 Mellinghoff IK, Wang MY, Vivanco I et al. Molecular determinants ofthe response of glioblastomas to EGFR kinase inhibitors. N Engl J Med2005;353:2012–2024.

16 Bachoo RM, Maher EA, Ligon KL et al. Epidermal growth factorreceptor and Ink4a/Arf: Convergent mechanisms governing terminaldifferentiation and transformation along the neural stem cell to astrocyteaxis. Cancer Cell 2002;1:269–277.

17 Fan QW, Knight ZA, Goldenberg DD et al. A dual PI3 kinase/mTORinhibitor reveals emergent efficacy in glioma. Cancer Cell 2006;9:341–349.

18 Jackson EL, Garcia-Verdugo JM, Gil-Perotin S et al. PDGFR alpha-positive B cells are neural stem cells in the adult SVZ that formglioma-like growths in response to increased PDGF signaling. Neuron2006;51:187–199.

19 Ligon KL, Huillard E, Mehta S et al. Olig2-regulated lineage-restrictedpathway controls replication competence in neural stem cells and malig-nant glioma. Neuron 2007;53:503–517.

20 Clement V, Sanchez P, de Tribolet N et al. HEDGEHOG-GLI1 signalingregulates human glioma growth, cancer stem cell self-renewal, andtumorigenicity. Curr Biol 2007;17:165–172.

21 Sanchez P, Hernandez AM, Stecca B et al. Inhibition of prostate cancerproliferation by interference with SONIC HEDGEHOG-GLI1 signaling.Proc Natl Acad Sci U S A 2004;101:12561–12566.

22 Guertin DA, Sabatini DM. Defining the role of mTOR in cancer. CancerCell 2007;12:9–22.

23 Thomas GV, Tran C, Mellinghoff IK et al. Hypoxia-inducible factordetermines sensitivity to inhibitors of mTOR in kidney cancer. Nat Med2006;12:122–127.

24 Ito D, Fujimoto K, Mori T et al. In vivo antitumor effect of the mTORinhibitor CCI-779 and gemcitabine in xenograft models of human pan-creatic cancer. Int J Cancer 2006;118:2337–2343.

25 Mungamuri SK, Yang X, Thor AD et al. Survival signaling by Notch1:Mammalian target of rapamycin (mTOR)-dependent inhibition of p53.Cancer Res 2006;66:4715–4724.

26 Dragowska WH, Verreault M, Yapp DT et al. Decreased levels of

3025Xu, Yuan, Liu et al.

www.StemCells.com

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

hypoxic cells in gefitinib treated ER(�) HER-2 overexpressing MCF-7breast cancer tumors are associated with hyperactivation of the mTORpathway: Therapeutic implications for combination therapy with rapa-mycin. Breast Cancer Res Treat 2007;106:319–331.

27 Creighton CJ. A gene transcription signature of the Akt/mTOR pathwayin clinical breast tumors. Oncogene 2007;26:4648–4655.

28 Masiello D, Mohi MG, McKnight NC et al. Combining an mTORantagonist and receptor tyrosine kinase inhibitors for the treatment ofprostate cancer. Cancer Biol Ther 2007;6:195–201.

29 Li D, Shimamura T, Ji H et al. Bronchial and peripheral murine lungcarcinomas induced by T790M-L858R mutant EGFR respond to HKI-272 and rapamycin combination therapy. Cancer Cell 2007;12:81–93.

30 Horvath S, Zhang B, Carlson M et al. Analysis of oncogenic signalingnetworks in glioblastoma identifies ASPM as a molecular target. ProcNatl Acad Sci U S A 2006;103:17402–17407.

31 Liang Y, Diehn M, Watson N et al. Gene expression profiling revealsmolecularly and clinically distinct subtypes of glioblastoma multiforme.Proc Natl Acad Sci U S A 2005;102:5814–5819.

32 Mischel PS, Shai R, Shi T et al. Identification of molecular subtypesof glioblastoma by gene expression profiling. Oncogene 2003;22:2361–2373.

33 Leung C, Lingbeek M, Shakhova O et al. Bmi1 is essential for cerebellardevelopment and is overexpressed in human medulloblastomas. Nature2004;428:337–341.

34 Rao G, Pedone CA, Valle LD et al. Sonic hedgehog and insulin-likegrowth factor signaling synergize to induce medulloblastoma formationfrom nestin-expressing neural progenitors in mice. Oncogene 2004;23:6156–6162.

35 Romer JT, Kimura H, Magdaleno S et al. Suppression of the Shhpathway using a small molecule inhibitor eliminates medulloblastoma inPtc1(�/-)p53(-/-) mice. Cancer Cell 2004;6:229–240.

36 Kimura H, Stephen D, Joyner A et al. Gli1 is important for medullo-blastoma formation in Ptc1�/- mice. Oncogene 2005;24:4026–4036.

37 Bar EE, Chaudhry A, Lin A et al. Cyclopamine-mediated hedgehogpathway inhibition depletes stem-like cancer cells in glioblastoma.STEM CELLS 2007;25:2524–2533.

38 Nery S, Wichterle H, Fishell G. Sonic hedgehog contributes to oligo-dendrocyte specification in the mammalian forebrain. Development2001;128:527–540.

39 Lu QR, Yuk D, Alberta JA et al. Sonic hedgehog-regulated oligoden-drocyte lineage genes encoding bHLH proteins in the mammalian centralnervous system. Neuron 2000;25:317–329.

40 Ehtesham M, Sarangi A, Valadez JG et al. Ligand-dependent activationof the hedgehog pathway in glioma progenitor cells. Oncogene 2007;26:5752–5761.

41 Choe G, Horvath S, Cloughesy TF et al. Analysis of the phosphatidyl-inositol 3�-kinase signaling pathway in glioblastoma patients in vivo.Cancer Res 2003;63:2742–2746.

42 Backman SA, Ghazarian D, So K et al. Early onset of neoplasia in theprostate and skin of mice with tissue-specific deletion of Pten. Proc NatlAcad Sci U S A 2004;101:1725–1730.

43 Cully M, Elia A, Ong SH et al. grb2 heterozygosity rescues embryoniclethality but not tumorigenesis in pten�/- mice. Proc Natl Acad SciU S A 2004;101:15358–15363.

44 Okumura K, Zhao M, Depinho RA et al. Cellular transformation by theMSP58 oncogene is inhibited by its physical interaction with the PTENtumor suppressor. Proc Natl Acad Sci U S A 2005;102:2703–2706.

45 Chang SM, Kuhn J, Wen P et al. Phase I/pharmacokinetic study ofCCI-779 in patients with recurrent malignant glioma on enzyme-induc-ing antiepileptic drugs. Invest New Drugs 2004;22:427–435.

46 Chang SM, Wen P, Cloughesy T et al. Phase II study of CCI-779 inpatients with recurrent glioblastoma multiforme. Invest New Drugs2005;23:357–361.

47 Galanis E, Buckner JC, Maurer MJ et al. Phase II trial of temsirolimus(CCI-779) in recurrent glioblastoma multiforme: A North Central CancerTreatment Group Study. J Clin Oncol 2005;23:5294–5304.

48 Goudar RK, Shi Q, Hjelmeland MD et al. Combination therapy ofinhibitors of epidermal growth factor receptor/vascular endothelialgrowth factor receptor 2 (AEE788) and the mammalian target of rapa-mycin (RAD001) offers improved glioblastoma tumor growth inhibition.Mol Cancer Ther 2005;4:101–112.

See www.StemCells.com for supplemental material available online.

3026 Hedgehog Signaling in Glioblastomas

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from

DOI: 10.1634/stemcells.2008-0459 2008;26;3018-3026; originally published online Sep 11, 2008; Stem Cells

Qijin Xu, Xiangpeng Yuan, Gentao Liu, Keith L. Black and John S. Yu Portends Shorter Survival for Patients with PTEN-Coexpressing Glioblastomas

Hedgehog Signaling Regulates Brain Tumor-Initiating Cell Proliferation and

This information is current as of December 19, 2008

& ServicesUpdated Information

http://www.StemCells.com/cgi/content/full/26/12/3018including high-resolution figures, can be found at:

Supplementary Material http://www.StemCells.com/cgi/content/full/2008-0459/DC1

Supplementary material can be found at:

at University of C

alifornia on Decem

ber 19, 2008 w

ww

.StemC

ells.comD

ownloaded from