Gli1 Induces G2/M Arrest and Apoptosis in Hippocampal but Not Tumor-Derived Neural Stem Cells

Authors


Abstract

Sonic hedgehog (Shh) is necessary for sustaining the proliferation of neural stem cells (NSCs), yet little is known about its mechanisms. Whereas Gli1, Gli2, and Gli3, the primary mediators of Shh signaling, were all expressed in hippocampal neural progenitors, Shh treatment of NSCs induced only Gli1 expression. Acute depletion of Gli1 in postnatal NSCs by short-hairpin RNA decreased proliferation, whereas germline deletion of Gli1 did not affect NSC proliferation, suggesting a difference in mechanisms of Gli1 compensation that may be developmentally dependent. To determine whether Gli1 was sufficient to enhance NSC proliferation, we overexpressed this mitogen and were surprised to find that Gli1 resulted in decreased proliferation, accumulation of NSCs in the G2/M phase of cell cycle, and apoptosis. In contrast, Gli1-expressing lineage-restricted neural precursors demonstrated a 4.5-fold proliferation enhancement. Expression analyses of Gli1-expressing NSCs identified significant induction of Gadd45a and decreased cyclin A2 and Stag1 mRNA, genes involved in the G2-M transition and apoptosis. Furthermore, Gadd45a overexpression was sufficient to partially recapitulate the Gli1-induced G2/M accumulation and cell death of NSCs. In contrast to normal stem cells, tumor-derived stem cells had markedly higher basal Gli1 expression and did not undergo apoptosis with further elevation of Gli1. Our data suggest that Gli1-induced apoptosis may serve as a protective mechanism against premature mitosis and may give insight into mechanisms by which nonmalignant stem cells restrain hyperproliferation in the context of potentially transforming mitogenic signals. Tumor-derived stem cells apparently lack these mechanisms, which may contribute to their unrestrained proliferation and malignant potential.

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

Introduction

Neural stem cells (NSCs) are self-renewing, multipotent progenitor cells that reside in discrete areas within the nervous system and give rise to a diversity of neuronal and glial cells. The most well-characterized NSC niches are the subventricular zone (SVZ) of the lateral ventricle and subgranular zone (SGZ) of the dentate gyrus [1]. These newly generated neurons can integrate into the neural circuitry of mammalian brains and are thought to play roles in learning and memory, epilepsy, degenerative neurological disease, and the clinical response to antidepressant medications [2]. Although numerous mitogens, neurotrophins, and other factors modulate the self-renewal and cell fate of both lineage-restricted and multipotent progenitor cells, the mechanisms and regulation of these factors in stem cell populations are poorly understood.

Sonic hedgehog (Shh), a secreted morphogen of the hedgehog family with numerous developmental roles, promotes proliferation of lineage-restricted progenitor cells as well as multipotent stem cells [3]. Although downstream Shh activity has been well studied in neural precursors, little is known about the mechanisms of hedgehog transduction in NSCs. An extensive body of literature documents the role of the Gli family of zinc-finger transcription factors (Gli1, Gli2, and Gli3) in mediating Shh signaling in development and tumorigenesis [4, [5]–6]. Augmented proliferation and transformation can be initiated by Gli1 and Gli2 in multiple cell types, whereas Gli3 is believed to serve primarily repressive functions [7]. Within the nervous system, Gli1 increases the proliferation of neural precursors and plays an important role in transformation and tumorigenesis [8, [9]–10]. Gene expression analyses of medulloblastomas arising in Ptc+/− mice demonstrate that this mouse model closely resembles the “molecular fingerprint” of a subset of human medulloblastomas [11, 12]. Although hedgehog pathway activation in the SVZ and SGZ induces Gli1 expression [13, 14], it is not known whether Gli1 is necessary and/or sufficient to promote the proliferation of NSCs, nor has it been determined what level of Gli1 is expressed by brain tumor-derived stem cells (BTSCs) and whether Gli1 participates in the transformation of these cells.

To address the role of Shh and Gli family members in proliferation of postnatal stem cells, we analyzed expression of the Gli factors in hippocampal progenitors in vivo and in NSCs treated with Shh in vitro. Our work demonstrated that Gli1 was the only Gli factor to be transcriptionally induced in NSCs following Shh pathway activation. Although Gli1-null mice develop normally [15], we found that somatic postnatal depletion of Gli1 impaired NSC self-renewal. In addition, forced expression of Gli1 did not induce a proliferative response in NSCs, as noted in neural precursors, but, surprisingly, initiated cell cycle arrest and apoptosis. We have identified novel targets of Gli1 in NSCs, which suggest that NSCs have a cell-specific response to Gli1 modulation. Analysis of forced Gli1 expression in stem cells derived from murine medulloblastomas revealed that BTSCs are protected from Gli1-induced apoptosis. Embryonic stem cells have robust mechanisms to safeguard their genomes [16], and we propose that nonmalignant NSCs also maintain tight regulation over their proliferative response to potentially oncogenic Gli1 expression. Our data suggest that increased expression of this mitogen induces cell cycle arrest and apoptosis as a protective mechanism by which NSCs guard against hyperproliferation.

Methods and Materials

Mice

Experimental protocols were approved by the Institutional Animal Care and Use Committee of Mayo Clinic. NestinGFP transgenic mice were maintained on C57Bl6/129SvJ background [17], and C57Bl6/129SvJ mice were used for all in vitro analyses unless otherwise noted. Gli1−/− mice [18] and their wild-type controls were maintained on a Swiss Webster background (Charles River Laboratories, Wilmington, MA, http://www.criver.com). Mice carrying deletion of the first and second exons of Patched (Ptc+/) were previously described [19].

Constructs

AcGFP control and the open reading frame for Gli1, Gli1zfd [15], and Gadd45a were cloned into pIRES2-AcGFP1. In addition, enhanced green fluorescent protein (eGFP) and eGFP-Gli1 were cloned into Murine stem cell virus (pMSCV). For retrovirus production, pMSCV-green fluorescent protein (GFP) and pMSCV-GFP-Gli1 were transfected into GP2-293 cells, and their supernatants infected EcoPack2-293 cells (Clontech, Palo Alto, CA, http://www.clontech.com). The virus supernatants from stable virus-producing EcoPack2-293 cells were filtered and concentrated. For short-hairpin RNA (shRNA) constructs, oligonucleotides were ligated into pGSU6 and pGSU6-GFP (Genlantis, San Diego, http://www.genlantis.com). The control hairpin sequence comprised scrambled bases and did not match known murine genes [20]. The Gli1 hairpin targeted base pairs 2,842–2,863: top strand, 5′ AGCTCAGCTGG-TGTGTAATTACTTCAAGAGAGTAATTACACACCAGCTGAG-CTTTTTTTACGCGT 3′; bottom strand, 5′ ACGCGTAAAAA-AAGCTCAGCTGGTGTGTAATTACTCTCTTGAAGTAATTAC-ACACCAGCTGAGCT 3′. Recombinant Shh was produced as previously described [21].

Cell Culture and Analysis

Neurosphere cultures were derived from hippocampi of postnatal day 5–6 (P5-6) mice as previously described [21]. After at least three sequential passages to functionally select for a self-renewing NSC population, DNA was introduced to the NSCs by nucleofection following the Amaxa protocol for mouse NSCs (Amaxa Inc., Gaithersburg, MD, http://www.amaxa.com). For overexpression experiments, the cells were pulsed with 5-bromo-2′-deoxyuridine (BrdU) 30 hours after nucleofection and harvested at 16 hours. For short-hairpin knockdown experiments, the cells were pulsed with BrdU 48 hours after nucleofection and harvested at 6 hours. The BrdU epitope was opened by incubation with 75 U of DNase for 1 hour at 37°C. The cells were then stained with Alexa Fluor 647-conjugated BrdU antibody (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), and the number of GFP/BrdU-labeled cells was quantified by fluorescence-activated cell sorting (FACS). At 2 days after nucleofection, the NSCs were also analyzed for terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) staining (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com) and active-caspase-3 staining (BD Biosciences, San Diego, http://www.bdbiosciences.com). Cell cycle analyses were also performed at this time point, using propidium iodide to label DNA content.

Cerebellar granule cell precursors (CGCPs) were derived from P6-7 cerebellum, panned to remove astrocytes, and grown on poly-d-lysine-coated plates in Neurobasal (Invitrogen) with B27. Gli1 overexpression in CGCPs was modeled after the protocol by Oliver and colleagues [9]. Briefly, the neural precursors were infected with a GFP control or GFP-Gli1 retrovirus 3 hours after their initial plating and pulsed with BrdU 48 hours after infection for 16 hours. They were processed for GFP and BrdU labeling as described for NSC cultures.

The multipotentiality of the Gli1−/− neurospheres was assessed in clonal spheres (single cells initially plated at 2 × 104 cells per milliliter; [22]). Clonal spheres were differentiated for 7 days by plating on laminin-coated slides and removing growth factors. After fixation, the cells were incubated in 5% normal donkey serum and incubated overnight with primary antibodies and for 1 hour with secondary antibodies. Antibodies specific for markers of neurons (β-III tubulin; Promega, Madison, WI, http://www.promega.com), astrocytes (glial fibrillary acidic protein; Chemicon, Temecula, CA, http://www.chemicon.com), and oligodendrocytes (O4; Chemicon) were used. The clonal spheres were triple-labeled with all three mature markers and appropriate secondary antibodies (Jackson Immunoresearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com).

Tumorsphere Cultures

Ptc+/ mice were observed daily for signs of increased intracranial pressure, and when symptomatic, the animal was sacrificed, the brain was removed, and tumor was separated from adjacent cerebellum. The tissue bordering the tumor/cerebellum interface was excised and discarded to reduce the possibility of mixing normal cells with tumor cells. Tumor tissue was placed into NSC culture medium, triturated, and filtered to obtain a single cell suspension. The cells were initially plated at approximately 1 × 106 cells per milliliter and grown in NSC culture medium with 10 ng/ml epidermal growth factor and 10 ng/ml basic fibroblast growth factor. Cells were passaged approximately every 5 days through chemical and mechanical dissociation and replated at 2 × 105 cells per milliliter. BTSC proliferation was assessed following enforced expression of GFP and Gli1 as conducted with the NSCs.

RNA Analysis

Comparative reverse transcription-polymerase chain reaction (rtPCR) was performed to determine gene expression in progenitor cells in vivo. The hippocampi from postnatal day 10–11 NestinGFP mice were separated into populations of nestin-GFP+ and GFP cells using FACS. Gene induction was also analyzed in NSCs following Shh treatment and Gli1 overexpression in vitro. RNA was harvested from vehicle and Shh-stimulated NSCs and CGCPs at 2 and 4 hours after the 48-hour Shh treatment [21]. By FACS, GFP-, Gli1zfd-, and Gli1-overexpressing NSCs were sorted for GFP+ and GFP cells 48 hours after nucleofection. The expression of the Gli factors was compared between NSCs and BTSCs from RNA harvested three days after passage. All RNA was extracted using Simply RNA (Stratagene, La Jolla, CA, http://www.stratagene.com), cDNA was generated using SuperScript III (Invitrogen) and comparative rtPCR was performed using Stratagene's MX3000p with Brilliant SYBR Green (Stratagene). Primers specific to each gene of interest were generated (supplemental online Table 2), and amplification efficiencies were optimized to 100% for all primer pairs.

Data Analysis

Results are graphed to demonstrate mean ± SEM. NSC cultures were, at minimum, performed in triplicate. Data comparing more than two experimental groups were first analyzed by analysis of variance. The data were compared using Student's t test, and a p value of <.05 was considered statistically significant.

Results

Nestin+ Hippocampal Neural Progenitors Have Strong Gli Expression

Shh, Ptc, Smoothened (Smo), and Gli1 mRNAs have been detected in the hippocampus and other proliferative regions of the postnatal nervous system [13, 23, 24]. However, the relative expression levels among neural progenitors and cells of the surrounding hippocampal parenchyma have not been analyzed previously. To compare the expression of Shh pathway components and target genes between these populations, we used a transgenic mouse expressing GFP under control of a nestin promoter (NestinGFP; Fig. 1A; [17]). Nestin, an intermediate filament protein expressed by immature cells, has been used extensively for the identification of neural progenitor cells [25, 26].

Figure Figure 1..

Expression of Shh pathway components and target genes in nestin+ neural progenitors and NSCs treated with Shh. (A): In the dentate gyrus of NestinGFP mice [17], GFP+ cells were localized mainly to the subgranular zone. (B): Hippocampi from postnatal day 10 NestinGFP mice were separated by fluorescence-activated cell sorting into populations of nestin-GFP (hippocampal parenchyma) and nestin-GFP+ (hippocampal progenitors) cells. Comparative reverse transcription-polymerase chain reaction of these two populations showed increased expression of Shh target genes (Gli1, Gli2, Gli3, and Nmyc) and decreased expression of cell cycle inhibitors (p21) in the nestin-GFP+ cells. Treatment of NSCs (C) and CGCPs (D) with exogenous Shh showed differential induction of Gli family members. RNA was harvested at 2 and 4 Hrs after the 48-Hr Shh treatment. Shh treatment of NSCs induced expression of only Gli1, whereas Shh treatment of CGCPs significantly induced expression of Gli1, Gli2, and Gli3. *, p < .05; **, p < .005. Error bars = SEM. Scale bar = 200 μm. Abbreviations: CGCP, cerebellar granule cell precursor; DAPI, 4,6-diamidino-2-phenylindole; Hrs, hours; NSC, neural stem cell; shh, sonic hedgehog; Veh, vehicle.

Nestin-GFP+ and GFP cells were sorted via FACS from freshly dissociated P10 and P11 NestinGFP hippocampi. Comparative rtPCR analysis demonstrated a significant enrichment in expression of hedgehog target genes in the nestin-GFP+ cells as compared with GFP cells, suggesting that the Shh pathway is activated in the neural progenitor cells to a much greater degree than in cells of the surrounding parenchyma (Fig. 1B). Expression of Gli1, Gli2, and Gli3 mRNA was 13–18-fold higher in the nestin-GFP+ cells (p < .005). In addition, Nmyc, an oncogene and target of Shh signaling in CGCP [9, 27], was expressed 180-fold higher in nestin-GFP+ than in GFP cells (p < .005). Further analysis revealed that the cyclin-dependent kinase inhibitor p21 was dramatically reduced in the nestin-GFP+ compared with the GFP cells (p < .01), which is to be expected in a population of actively proliferating progenitor cells. Our findings of prominent expression of Gli1 in hippocampal progenitor cells over surrounding parenchymal cells is consistent with previous reports of significantly higher Gli1 expression in SVZ stem cells compared with neuroblasts [23], suggesting that Gli1 is a hedgehog target common to multiple stem/progenitor populations from different brain regions.

Gli1 Induction in Shh-Treated NSCs

Given that all three Gli transcription factors were more highly expressed in nestin+ neural progenitors compared with the differentiated cells of the hippocampal parenchyma, we next asked whether Shh exposure would induce expression of each of the three Gli factors in NSCs. Shh treatment of NSCs elicited fivefold induction of Gli1, with no significant changes in expression of Gli2 or Gli3 mRNA (Fig. 1C). We also sought to determine whether Shh targets differed in multipotent NSCs from those already identified in CGCPs, a population of lineage-restricted neural precursors cells whose response to Shh has been well studied [3]. In contrast to NSCs, exposure of CGCPs to Shh resulted in significant induction of all three Glis (Fig. 1D). Gli1 mRNA expression was increased 30-fold in CGCPs—significantly higher than the induction of Gli2 and Gli3 in CGCPs and almost five times the induction of Gli1 observed in NSCs. Our expression analyses suggest that Shh exposure results in a differential gene induction in NSCs that is distinct from those induced in lineage-restricted neural precursors and that Shh-stimulated proliferation in NSCs may be mediated through induction of Gli1.

Postnatal Gli1 Depletion Impairs NSC Self-Renewal

As Gli1 was the only Gli factor to be induced following Shh treatment of NSCs (Fig. 1C), we next asked whether Gli1 was necessary for NSC self-renewal. We designed and characterized several potential short-hairpin sequences targeting murine Gli1 and selected the construct with most efficient knockdown of Gli1 protein (Fig. 2A). To determine the effect of Gli1 depletion on NSC proliferation, we treated the nucleofected cells with G418 sulfate to select for NSCs containing the shRNA constructs, which contained a sequence encoding neomycin resistance. We found that the NSCs containing the control shRNA sequences continued to proliferate and began forming neurospheres and proliferating near normal NSC levels after five passages (Fig. 2B; supplemental online Fig. 1). In contrast, cells nucleofected with Gli1-shRNA constructs failed to expand in culture, remained as single viable cells, and did not generate neurospheres.

Figure Figure 2..

Short-hairpin KD of Gli1 in NSCs. The efficiency of Gli1 short hairpins was characterized in NIH3T3 cells overexpressing a GFP-Gli1 fusion protein. (A): The Gli1 hairpin targeting bases 2,842–2,863 significantly depleted GFP-Gli1 protein, whereas a control short hairpin did not affect Gli1 expression. The depletion was demonstrated by decreased expression of GFP-Gli1 in 3T3 cells and by Western analysis. (B): Control and Gli1 short-hairpin constructs expressing GFP were nucleofected into NSCs, and the cells were treated with G418 sulfate to select cells that had stably incorporated the KD construct. Although NSCs that stably incorporated the control hairpin retained the ability to self-renew and proliferate, the NSCs with stable Gli1 KD failed to self-renew, and the cultures were exhausted of cells. Transient KD of Gli1 demonstrated significant reduction in proliferation (C) but no change in the percentage of TUNEL+ cells (D). (E): Fluorescence-activated cell sorting analysis of cell cycle progression revealed that Gli1 KD decreased the percentage of cells in S and G2/M phase while significantly increasing the proportion of cells in G1/G0. *, p < .05. Error bars = SEM. Scale bar = 200 μm. Abbreviations: BrdU, 5-bromo-2′-deoxyuridine; GFP, green fluorescent protein; KD, knockdown; NSC, neural stem cell; TUNEL, terminal deoxynucleotidyl transferase dUTP nick-end labeling.

Given that it was not possible to establish stable NSC lines carrying knockdown depletion of Gli1 due to their marked decrease in proliferation, we sought to determine the effects of Gli1 knockdown on NSCs with transient expression of the depletion constructs. At 48 hours after nucleofection of the knockdown constructs, Gli1 depletion resulted in a 40% decrease in BrdU+ NSCs compared with the control (p < .01; Fig. 2C). The decrease in proliferation cannot be attributed to increased apoptosis, as there was no significant change in the percentage of TUNEL+ cells (Fig. 2D). We have found that the decrease in proliferation resulting from Gli1 depletion is likely a consequence of arrest in G0/G1 phase of cell cycle (Fig. 2E; p < .005). These data suggest that Gli1 is necessary for continued proliferation of postnatal NSCs and that acute depletion of Gli1 in these cells inhibits their ability to self-renew.

To further investigate the importance of Gli1 in NSCs, we turned to a previously reported Gli1-deficient mouse [15]. As Gli1-null mice develop normally without an overt phenotype, we were able to derive neurosphere cultures from hippocampal tissue to assay the self-renewal potential of NSCs derived from mice carrying germline deletion of Gli1. Neurospheres were generated from cells isolated from Gli1−/− hippocampal tissue (Fig. 3A, 3B) that retained the ability to self-renew over multiple passages (Fig. 3C) and were multipotent (Fig. 3D). Both Gli1+/+ and Gli1−/− cells demonstrated equivalent percentages of proliferating cells as quantified by incorporation of BrdU (Fig. 3C). Although germline deletion of Gli1 is likely compensated for to allow for normal embryonic development, acute depletion of Gli1 expression abrogates self-renewal and proliferation of postnatal hippocampal NSCs.

Figure Figure 3..

Proliferation of Gli1−/− neurospheres. Hippocampal neurospheres were derived from Gli1+/+(A) and Gli1−/−(B) postnatal day 5–6 mice and were maintained over several passages. (C): BrdU incorporation and detection demonstrated similar levels of proliferation between Gli1+/+ and Gli1−/− neurospheres. (D): Self-renewal and multipotentiality of Gli1−/− neurospheres was demonstrated through differentiation of cells grown at clonal density. Clonally generated Gli1−/− neurospheres contained cells that differentiated to display all neural phenotypes. Neuron is denoted by white arrow. Error bars = SEM. Scale bar = 100 μm. Abbreviations: BrdU, 5-bromo-2′-deoxyuridine; GFAP, glial fibrillary acidic protein.

Forced Expression of Gli1 Induces Apoptosis and Cell Cycle Arrest in NSCs

Ectopic expression of Gli1 in neural progenitors results in induction of Shh targets [28, 29] and is sufficient to drive tumor formation in the epidermis [30, 31]. To determine whether Gli1 is sufficient to increase proliferation in NSCs and CGCPs, we forced expression of Gli1 and analyzed proliferation by BrdU incorporation. Overexpression of Gli1 in CGCPs significantly increased the proliferation of these neural precursors (p < .05; Fig. 4A), as also demonstrated by other investigators [9]. Surprisingly, increased expression of Gli1 had the opposite effect in NSCs and resulted in a 60% reduction in the number of BrdU+ cells (Fig. 4B). Forced expression of other transcription factors and cell cycle modulators did not negatively affect NSC proliferation [21], and a series of experiments was conducted to confirm that the decrease in NSC proliferation was specific to Gli1 expression and not due to nonspecific effects (supplemental online Figs. 2, 3). Forced expression of full-length Gli1 caused a decrease in proliferation that was of greater magnitude than overexpression of Gli1 lacking zinc fingers 2–5 (Gli1zfd), a truncated form of Gli1 that does not contain the DNA-binding domain and lacks the ability activate Shh targets [15]. We have found that increased expression of Gli1 induces a unique, cell type-specific phenotype and results in a paradoxical decrease in proliferation in NSCs, which is in contrast to the well-established role of Gli1-induced proliferation of precursors from the nervous system and other tissues.

Figure Figure 4..

Forced Gli1 expression in NSCs results in increased cell death and accumulation of cells in G2/M phase of the cell cycle. Expression of Gli1 or GFP (control) was enforced in NSCs and CGCPs, and the cells were assayed for BrdU incorporation, cell death, or cell cycle progression after 48 hours. (A): Gli1 overexpression increased CGCP proliferation approximately 4.5-fold. (B): In contrast, enforced expression of Gli1 in NSCs resulted in a dramatic decrease in proliferation, with only 4% of the Gli1+ cells incorporating BrdU compared with 12% of GFP+ cells. In NSCs, Gli1 overexpression increased both the number of TUNEL+(C) and active caspase-3+(D) cells. NSCs overexpressing GFP and Gli were stained with PI and analyzed for DNA content by flow cytometry. (E): Representative plots are displayed and are accompanied by the cell cycle phase percentages for three independent experiments. *, p < .05. Error bars = SEM. Abbreviations: BrdU, 5-bromo-2′-deoxyuridine; CGCP, cerebellar granule cell precursor; GFP, green fluorescent protein; NSC, neural stem cell; PI, propidium iodide; TUNEL, terminal deoxynucleotidyl transferase dUTP nick-end labeling.

We next sought to determine whether the diminished proliferation of Gli1-expressing NSCs could be attributed to impaired cell cycle progression and/or the induction of cell death. To identify dying cells, we used two independent markers of apoptosis: TUNEL and cleavage of caspase-3. After 48 hours, the Gli1-expressing NSCs had a 13-fold increase in the number of TUNEL+ cells and a 3-fold elevation in active capsase-3+ cells (p < .05; Fig. 4C, 4D). Gli1 expression in NSCs also disrupted cell cycle progression. Although forced expression of control GFP did not significantly alter cell cycle progression, Gli1 overexpression caused a dramatic accumulation of NSCs in the G2/M phase (nearly 40% of cells; Fig. 4E). Gli1-expressing NSCs had more than threefold accumulation of cells in G2/M compared with control cultures, suggesting that elevated Gli1 hinders progression of NSCs through the G2 or M phases of cell cycle. Thus, impaired progression through G2/M and increased apoptosis both contribute to the decrease in NSC proliferation resulting from enforced Gli1 expression. Whether cell cycle arrest and apoptosis are distinct events or points along the same continuum remains to be determined.

Novel Gli1 Targets Are Induced in NSCs

Analyses of transcriptional regulation by Gli1 in transformed kidney cells and keratinocytes have identified Gli1 target genes that propel the cells through the cell cycle, upregulate growth factor receptors, and reduce apoptosis [32, 33]. Given that Gli1 does not induce these phenotypes in NSCs, we sought to characterize Gli1-induced changes in gene expression that may contribute to cell cycle arrest and cell death. To analyze modulation in gene expression induced by Gli1 expression, we quantified mRNA levels between a population of cells transduced with Gli1 and two controls: the GFP-expressing vector and Gli1zfd, which lacks the zinc-finger DNA binding domain of Gli1. The mRNA from Gli1 and control vector-expressing cells was analyzed with the mouse cell cycle pathway array to generate a subset of genes of interest for further analysis (supplemental online Table 1).

Genes meeting our threshold criteria, as well as established Shh pathway genes, were further analyzed through comparative rtPCR. Ptc and Nmyc, known Shh-target genes, had, respectively, threefold and twofold induction in NSCs overexpressing Gli1 (p < .05; Fig. 5). Interestingly, we noted only slight induction (1.3-fold) of established Shh target genes known to serve antiapoptotic and/or proliferative functions (Bcl2 and cyclin D2; p < .05). We also identified induction of novel genes not previously reported in Gli1-expressing cells that are known to be involved in apoptosis and cell cycle arrest. Gadd45a (Growth Arrest and DNA Damage-Inducible gene), which has functions in arrest at G2/M phases of cell cycle and apoptosis [34], is upregulated greater than 2.5-fold in NSC overexpressing Gli1 (p < .0001; Fig. 5). In addition, cyclin A2, which plays a vital role in the transition between G2 and M phases of the cell cycle through formation of the mitosis-promoting factor [35], is decreased by greater than 50% in Gli1-expressing NSCs (p < .0001). Lastly, we also noted a specific and striking decrease in Stag1 (Stromal Antigen 1), a protein that complexes with the cohesins and is present during the early phases of mitosis (p < .0001; [36]). It is important to note that forced expression of Gli1zfd resulted in a significantly attenuated induction of Nmyc, cyclin D2, Gadd45a, and cyclin A2 compared with Gli1-expressing cells, yet the expression changes did also differ from GFP-expressing NSCs (p < .05). Overall, we have identified novel genes whose expression is modulated by Gli1 in NSCs. These genes likely contribute to the unique phenotype of cell cycle arrest and apoptosis induced by Gli1 in stem cells. We have detected two putative Gli1-binding sites in the Gadd45a promoter (K.E.G. and C.W., unpublished data), yet it remains to be determined whether Gadd45a is a direct or indirect target of Gli1.

Figure Figure 5..

Changes in gene expression following enforced expression of Gli1 in NSCs. GFP, Gli1zfd, and Gli1 overexpression constructs were nucleofected into NSCs. Forty-eight hours after the nucleofections, the neurospheres were dissociated, and the cells overexpressing each construct were analyzed for GFP fluorescence through fluorescence-activated cell sorting. A mouse cell cycle pathway focused array (SuperArray) was performed to identify genes of interest. A subset of these genes was further analyzed through comparative reverse transcription-polymerase chain reaction. The graph represents three independent experiments, with the Gli1zfd and Gli1 groups calibrated to NSCs overexpressing GFP, while established Shh target genes (Ptc, Nmyc, CcnD2, Bcl2) were induced. In addition, our NSC analysis revealed changes in gene expression that have not previously been identified as Shh targets (Gadd45a, CcnA2, Stag1) and that correlate with the observed phenotype of cell cycle arrest and apoptosis. A p value <.05 was considered statistically significant. *, statistically different from GFP; #, statistically different from Gli1zfd. Error bars = SEM. Abbreviation: GFP, green fluorescent protein.

As Gli1 is believed to act primarily as a transcriptional activator and Gadd45a is strongly induced in Gli1-expressing NSCs, we sought to determine whether Gadd45a expression is sufficient to induce G2 arrest or apoptosis in NSCs. Overexpression of Gadd45a resulted in a decrease of more than 50% in the percentage of BrdU+ NSCs. Forced expression of Gadd45a in NSCs induced a twofold increase in the number of cells expressing activated caspase-3+ (Fig. 6A) and almost doubled the percentage of NSCs in G2/M phase of cell cycle (Fig. 6B). Thus, expression of Gadd45a is sufficient to largely recapitulate a phenotype similar to that induced by Gli1 in NSCs, yet other genes may also be involved.

Figure Figure 6..

Forced Gadd45a expression in NSCs is sufficient to induce a significant accumulation of cells in G2/M and apoptosis. NSCs were nucleofected with Gadd45a or GFP (control) and assayed for BrdU, cell death, or cell cycle progression after 48 hours. Gadd45a overexpression decreased the percentage of cells incorporating BrdU (A) and increased the number of active caspase-3+ cells (B). NSCs overexpressing GFP and Gadd45a were stained with PI and analyzed for DNA content by flow cytometry. Representative plots are displayed for (C) GFP and (D) Gadd45 and are accompanied by cell cycle phase percentages for three independent experiments. *, p < .05. Error bars = SEM. Abbreviations: BrdU, 5-bromo-2′-deoxyuridine; GFP, green fluorescent protein; NSC, neural stem cell; PI, propidium iodide.

Medulloblastoma Derived Stem Cells Resist Gli1-Induced Apoptosis

BTSCs have been isolated from multiple types of brain cancers [37] and are proposed to maintain and propagate malignant tumors [38]. Treatments that target malignant stem cells may help reduce metastatic spread and tumor reoccurrence. As Gli1 expression induced cell cycle arrest and cell death in nonmalignant NSCs, we sought to determine whether BTSCs expressed elevated Gli1 and whether further augmentation of Gli1 would also induce the death of BTSCs.

Mice haploinsuffienct for Ptc, the repressive component of the Shh receptor system, have endogenous activation of the Shh pathway and consequently develop spontaneous cerebellar and skin tumors [19, 39]. BTSCs were derived from tumors spontaneously formed in Ptc+/ and Ptc+/ NestinGFP mice, which to our knowledge is the first isolation of stem cells from a murine brain tumor. Tumors were grossly visible upon opening the calvarium, and strong expression of NestinGFP was apparent when viewed under a fluorescence-equipped dissecting microscope (Fig. 7A). Only a small subset of tumor cells survived passage and formed spheres that could be maintained under neurosphere-propagating conditions. The BTSCs continued to renew for more than 20 passages (Fig. 7B). Given that there is no marker exclusive to stem cells and not all stem cells express CD133, we defined the cells functionally through demonstration of self-renewal, differentiation of clonally propagated cells down oligodendrocytic, astrocytic, and neuronal lineages, and initiation of secondary tumors in naive mice (unpublished data). To determine differences in Gli levels between NSCs and BTSCs, comparative rtPCR was conducted on cells grown under identical conditions. All three of the Gli transcription factors were more highly expressed in BTSCs than in NSCs; however, Gli1 expression was markedly elevated in BTSCs (Fig. 7C). Gli1 mRNA expression was approximately 2,000-fold higher in BTSCs compared with NSCs. Expression of Gli2 and Gli3 was also moderately higher in BTSCs, approximately 11-fold and 2.5-fold, respectively.

Figure Figure 7..

BTSCs derived from medulloblastomas fail to induce apoptosis following Gli1 expression. (A): Tumor cells from Ptc+/NestinGFP mice prominently expressed NestinGFP. (B): Tumorspheres developed under neurosphere propagating conditions. (C): Comparative reverse transcription-polymerase chain reaction analysis between hippocampal NSCs and BTSCs showed high expression of Gli1 and more moderately elevated expression of Gli2 and Gli3 in BTSCs. BTSCs were nucleofected with Gli1 or GFP (control) and assayed for BrdU incorporation, cell death, or cell cycle progression after 48 hours. (D): Forced expression of Gli1 in BTSCs resulted in decreased percentage of cells incorporating BrdU. (E, F): Gli1 overexpression did not alter the number of TUNEL+ or active caspase-3+ cells. (G): BTSCs overexpressing GFP and Gli were stained with PI and analyzed for DNA content by flow cytometry. Representative plots are displayed and are accompanied by the average cell cycle phase percentages for three independent experiments. *, p < .05. Error bars = SEM. Scale bar 200 μm (B). Abbreviations: BrdU, 5-bromo-2′-deoxyuridine; BTSC, brain tumor-derived stem cell; GFP, green fluorescent protein; NSC, neural stem cell; PI, propidium iodide; TUNEL, terminal deoxynucleotidyl transferase dUTP nick-end labeling.

We sought to determine whether further elevation in Gli1 expression would affect proliferation in BTSCs and whether increased expression of Gli1 would induce apoptosis in these malignant stem cells. Forty-eight hours after nucleofection, the Gli1-expressing BTSCs demonstrated a significant decrease in proliferation, similar to forced expression of Gli1 in NSCs. However, in contrast to NSCs, Gli1 overexpression in BTSCs did not cause a significant change in the percentage of TUNEL+ or active caspase-3+ cells (Fig. 7E, 7F), and Gli1-expressing BTSCs arrested in G0/G1 and not in G2/M as noted in NSCs (Fig. 7G). Thus, although forced expression of Gli1 resulted in decreased proliferation in both NSCs and BTSCs, the cell cycle phenotype and cellular response of these two populations appeared to be cell type-specific. In addition, in contrast to lineage-restricted CGCPs, which proliferate in response to augmented Gli1 expression (Fig. 4A), self-renewing stem-like cells derived from normal or neoplastic tissue do not further increase proliferation in response to enforced expression of Gli1. As NSCs have recently been identified in the cerebellum and postulated to be a potential cell of origin in medulloblastomas [40, 41], we have repeated our Gli1 expression work in cerebellar NSCs and have noted a similar induction in apoptosis as demonstrated in hippocampal NSCs (unpublished data). We hypothesize that the BTSCs have undergone changes, perhaps as a consequence of transformation, that alter their cellular response to elevated Gli1 expression and allow them to escape the Gli1-induced apoptosis exhibited by nontransformed stem cells.

Discussion

Ectopic expression of the Gli1 oncogene induces proliferation and transformation of numerous cell types [9, 30, 31]. Shh treatment of neocortical tissue solely induces Gli1 expression [42], and Gli1 was the only Gli family member to be transcriptionally induced following exposure of NSCs to Shh. These observations directed our research to address whether Gli1 was sufficient and/or necessary for NSC self-renewal.

Two Gli1 depletion model systems were used to determine whether Gli1 was necessary for the self-renewal of NSCs. Although chronic germline depletion did not affect NSC proliferation, acute knockdown of Gli1 prevented NSC self-renewal. We hypothesize that compensatory mechanisms fulfill the role served by Gli1 in the absence of Gli1 during embryonic development, and these compensatory changes are present in the postnatal Gli1−/− animal and allow for NSC self-renewal. However, in the context of acute Gli1 depletion, the NSCs are not able to modulate gene expression sufficiently to sustain self-renewal and proliferation. Our data complement prior work in zebrafish demonstrating the requirement for Gli1 during development and also substantiates the Gli1 amplifier model proposed by Karlstrom et al. [43]. Essentially, this model suggests that Gli2 and Gli3 initiate Shh signaling, and Gli1 acts as an amplifier of the signaling to sustain pathway activation. In support of this theory, Gli1-null mice, which develop normally, demonstrate decreased induction of Shh pathway genes when Gli2 expression is also diminished [18]. Our Gli1 depletion data suggest that there may be less redundancy in Gli1 signaling within the context of sustaining proliferation of postnatal stem cells than is exhibited by the overlapping functions of Gli family members during embryonic development [44].

To determine whether Gli1 is sufficient to induce NSC proliferation, this mitogen was overexpressed in NSCs. Surprisingly, forced expression of Gli1 induced cell cycle arrest and apoptosis in NSCs, which is in direct contrast to the proliferative effects of Gli1 on neural precursors and tumor cells [9, 45] and suggests that stem cells have a unique biologic response to elevation of Gli1 that serves to protect the cells against potentially mutagenic rates of proliferation. Given this novel Gli1-induced phenotype in NSCs, we sought to identify downstream mediators of the abrogation of cell cycle progression. We found that Gli1 expression in NSCs induced novel changes in gene expression, including proapoptotic genes (Gadd45a) and genes governing the G2-M transition (Gadd45a, cyclin A2, Stag 1) that have not been previously identified as genes linked to Gli1 expression [32, 33]. Gadd45a, normally induced by environmental stressors, plays an important role in the G2-M transition through regulation of Cdc2 kinase [46] and Cyclin B1 [47] and in cell death through translocation of proapoptotic Bim to the mitochondria [48]. Our analysis has shown that Gadd45a expression is sufficient to induce cell cycle arrest and apoptosis in NSCs. Of particular interest, medulloblastomas from Ptc1neo67/+ mice and Ptc1-null embryos have strong expression of Gadd45a, suggesting that Gadd45a may be regulated by Shh signaling in other tissues and may play a potential role in tumorigenesis [49].

We hypothesize that potently mitogenic Gli1 signal is interpreted differently within the intracellular context of self-renewing, multipotent NSCs than it is within the population of differentiated, lineage-restricted neural precursor cells. Our data suggest that NSCs maintain tight regulation over cell cycle progression and restrain proliferation in response to elevated and potentially transforming levels of Gli1. Interestingly, the Gli1 induction in neural precursors treated with Shh was almost sixfold higher than the Gli1 induction in Shh-treated NSCs. Thus, under ligand stimulation, NSCs have a relatively low-level of Gli1 induction, suggesting that stem cells are exquisitely sensitive to subtle changes in Gli1 expression and maintain tight regulation of their biologic response to this mitogen.

We propose that elevated expression of Gli1 serves as an intracellular stressor to NSCs and that the resulting cell cycle arrest and apoptosis are defensive mechanisms to guard against potentially detrimental rates of hyperproliferation in this self-renewing population. Intracellular stress induces activation of p53, a transcription factor that is a potent inhibitor of cell cycle progression and a primary mechanism of inducing apoptosis. p53 is a negative regulator of NSC self-renewal [50], and the loss of p53 accelerates Shh-initiated brain tumors [51]. Altered expression of p53 targets Gadd45a and cyclin A2 suggests that the p53 pathway may mediate the Gli1-induced cell cycle arrest and apoptosis in NSCs [52, 53]. Analysis of forced expression of Gli1 in p53−/− NSCs would determine whether the Gli1-induced cell death and cell cycle arrest are dependent on the p53 pathway.

Although the observed phenotype of cell cycle arrest and apoptosis in Gli1-expressing NSCs was unexpected, it is not unprecedented. Ectopic expression of Gli1 in transgenic mice blocks spermatogenesis and leads to apoptosis of germ cells [54]. Interestingly, supplemental expression of Gli1 from the Gli2 locus results in transgenic mice with defects such as smaller size and reduced survival [55]. Thus, prior reports by other investigators suggest that populations of somatic and/or germ cells may be sensitive to elevated expression of the Gli1 mitogen in vivo. An emerging body of work demonstrates that stem cells are uniquely sensitive to somatic mutations and have multiple mechanisms to protect the integrity of their genome. Embryonic stem cells have low mutation rates, and the absence of a G1 checkpoint following exposure to ionizing radiation provides a mechanism by which these multipotent stem cells protect their genome by undergoing apoptosis rather than initiating DNA damage repair [16]. Thus, our NSC data suggest that tissue-specific stem cells may have genome stability protective mechanisms similar to those of embryonic stem cells.

Similar potentially genome-protective mechanisms were not present in malignant BTSCs with augmented expression of Gli1, and such loss may contribute to the malignant potential of these cells. Although Gli1 expression in nonmalignant NSCs induced apoptotic cascades, BTSCs overexpressing Gli1 showed no changes in cell death. Furthermore, in contrast to the accumulation of cells in G2/M noted in NSCs after Gli1 expression, BTSCs maintained a G1 arrest. The absence of Gli1-induced cell death and G2/M arrest demonstrate that the transformed stem cells have a different intracellular response to elevated mitogen levels than NSCs. The marked arrest of Gli1-expressing BTSCs in G1 suggests that the checkpoint mediating DNA damage may have been activated by the presence of strong mitogenic signal. In relation to the work previously discussed regarding the absence of G1 arrest in irradiated embryonic stem cells [56], one could hypothesize that neoplastic stem cells have lost this protective mechanism. The lack of cell death induced by Gli1 expression in BTSC also agrees with a defective damage protection system. Further work is needed to determine whether Gli1 expression induces different subsets of genes in nonmalignant and malignant stem cells.

Summary

Overall, our work shows that NSCs retain tight regulation of Gli1 expression and the consequent cell cycle progression. Although Gli1 expression is necessary for self-renewal of postnatal NSCs, elevated expression of this mitogen is detrimental to cell survival and induces cell cycle arrest and apoptosis. This unique response to Gli1 expression in NSCs highlights the importance of maintaining strict regulation of the proliferative capacity and genomic integrity of these multipotent stem cells. Aberrant activation of the Shh/Gli1 pathway is transforming and contributes to tumorigenesis in several tissues, and thus, the ability of NSCs to maintain tight regulation over cell cycle progression is likely an important mechanism to defend against cellular transformation. Additional experimentation is needed to discern the potential mechanism by which stem cells may escape Gli1-induced apoptosis and undergo transformation or whether abrogation of apoptosis is a later event in the spectrum of malignant transformation.

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.

Acknowledgements

We thank Gregori Enikopolov for the NestinGFP mice and Alexandra Joyner for the Gli1lacz mice and Gli1zfd construct. We are also grateful to Charles Howe and Eric Buenz for helpful suggestions and to Jan van Deursen and Liviu Malureanu for assistance on retrovirus production. Research was supported by the Sontag Foundation Distinguished Scientist Award, Hope Street Kids Foundation, Pediatric Brain Tumor Foundation of the United States, and the Bernie and Edith Waterman Center for Cancer Genetics.

Ancillary