Abstract
Neurofibromatosis 1 (NF1) is a common inherited disease in which affected children exhibit abnormalities in astrocyte growth regulation and are prone to the development of brain tumors (astrocytoma). Previous studies from our laboratory demonstrated that Nf1 mutant mouse astrocytomas contains populations of proliferating nestin+ progenitor cells, suggesting that immature astroglial progenitors may serve as a reservoir of proliferating tumor cells. Here, we directly examined the consequences of Nf1 inactivation on neural stem cell (NSC) proliferation in vitro and in vivo. We found dose-dependent effects of neurofibromin expression on NSC proliferation and survival in vitro, which reflected increased RAS pathway activation and increased bcl2 expression. In addition, unlike wild-type NSCs, Nf1-/- NSCs and, to a lesser extent, Nf1+/- NSCs survive as xenografts in naive recipient brains in vivo. Although Nf1-/- NSCs are multipotent, Nf1-/- and Nf1+/-, but not wild-type, NSCs generated increased numbers of morphologically abnormal, immature astroglial cells in vitro. Moreover, the Nf1-/- NSC growth and survival advantage as well as the astroglial cell differentiation defect were completely rescued by expression of the GAP (RAS-GTPase activating protein) domain of neurofibromin. Finally, the increase in astroglial progenitors and proliferating cells seen in vitro was also observed in Nf1-/- and Nf1+/- embryonic as well as Nf1+/- adult brains in vivo. Collectively, these findings support the hypothesis that alterations in neurofibromin expression in the developing brain have significant consequences for astrocyte growth and differentiation relevant to normal brain development and astrocytoma formation in children.
Introduction
Neurofibromatosis 1 (NF1) is a common genetic condition affecting the nervous system. Although the hallmark of the disease is the development of peripheral nervous system tumors (neurofibromas), the CNS is frequently involved (Gutmann et al., 1997). In this regard, 15-20% of children develop low-grade glial cell neoplasms (World Health Organization grade I pilocytic astrocytomas), typically located within the optic pathway (Listernick et al., 1994, 1999), termed optic pathway glioma (OPG).
In mice, reduced or absent Nf1 gene expression confers a proliferative advantage to astrocytes in vitro and in vivo (Bajenaru et al., 2001, 2002). In this regard, Nf1+/- mice with astrocyte Nf1 inactivation develop OPGs (Bajenaru et al., 2003). Consistent with the established function of the NF1 gene product (neurofibromin) in RAS inhibition (Ballester et al., 1990; Martin et al., 1990; Xu et al., 1990), OPGs also arise in Nf1+/- mice as a result of dysregulated K-RAS activity in astrocytes (Dasgupta et al., 2005).
The availability of mouse models of NF1-associated glioma provides a unique opportunity to study the molecular and cellular pathogenesis of these brain tumors. To this end, examination of OPG arising in these genetically engineered Nf1 mutant mice has revealed the presence of nests of proliferating nestin- and brain lipid-binding protein- (BLBP) immunoreactive cells within the evolving tumor (Bajenaru et al., 2005). The identification of cells expressing markers associated with neural stem/progenitor cells in these mouse brain tumors (Lendahl et al., 1990; Malatesta et al., 2000; Noctor et al., 2000; Hartfuss et al., 2001) raises the intriguing possibility that gliomas in children with NF1 might develop from proliferating NF1-/- progenitor cells.
In keeping with this notion, recent studies have identified a small fraction of highly proliferating cells within tumors that express markers of stem/progenitor cells. These “cancer stem cells” have been observed in a wide variety of diverse human cancers, including myeloid leukemia (Hope et al., 2004), breast cancer (Al-Hajj et al., 2003), and high-grade glioma (Hemmati et al., 2003; Singh et al., 2003; Galli et al., 2004). Moreover, cancer stem cells have also been isolated from pediatric brain tumors, including pilocytic astrocytoma (Singh et al., 2003), suggesting that aberrantly proliferating stem/progenitor cells might be involved in the formation of NF1-associated OPG.
Given both the clinical brain manifestations of NF1 and several lines of converging evidence that loss of neurofibromin in CNS cells may impair astroglial cell proliferation (Bajenaru et al., 2002; Bennett et al., 2003), we directly examined the consequence of neurofibromin loss on neural stem cell (NSC) proliferation in vitro and in vivo. We demonstrate that Nf1 inactivation profoundly affects NSC proliferation and survival as well as astroglial cell differentiation and that these functions of neurofibromin are mediated by residues within the NF1-RASGAP (RAS-GTPase activating protein) domain (GRD). Importantly, we show that Nf1 heterozygosity, as seen in NF1 patient brains, results in defective cell proliferation in vitro and the persistence of aberrantly proliferating and differentiating cells in the developing and adult brain in vivo.
Materials and Methods
Mice. Nf1+/- mice (a generous gift from Dr. Neal Copeland, National Institutes of Health, Bethesda, MD) (Brannan et al., 1994) were mated to generate embryos of each genotype. These mice were maintained as permanent colonies in the Department of Comparative Medicine small animal barrier facility at Washington University School of Medicine in accordance with approved Animal Studies Committee protocols.
Isolation and culture of neurospheres. CNS telencephalic lobes were removed from embryonic day 10.5 (E10.5) decidua of time-pregnant females and processed to obtain single-cell suspension of neural progenitors (neurospheres) as described previously (Molofsky et al., 2003), with minor modifications. Vesicles were digested with trypsin digest buffer containing 0.2% BSA (Sigma, St. Louis, MO), 0.5 mg/ml DNase I (Sigma), and 10% trypsin-EDTA stock (BioWhittaker, Walkersville, MD) in HBSS at 37°C for 10 min in a volume of 0.7 ml per vesicle. Equal volumes of 10% FCS medium containing 10% FCS (Life Technologies, Gaithersburg, MD), 2 mm l-glutamine (BioWhittaker), 0.1% glucose (Sigma), and 0.1 mm 2-mercaptoethanol (Sigma) in DMEM/F-12 (Sigma) were added, and vesicles were triturated with fire-polished Pasteur pipettes. Pelleted cells were washed with dissociation medium containing 0.1% sodium bicarbonate, 15 mm HEPES (Sigma), 0.5% glucose, and 0.2% BSA in HBSS. Cells were finally resuspended either in defined medium (Tropepe et al., 1999; Arsenijevic et al., 2001) or in NSC medium containing a 5:3 mixture of DMEM low glucose:Neurobasal medium (Life Technologies), 0.5 mm 2-mercaptoethanol, 2 mm L-glutamine, 5 IU of penicillin, and 5 μg/ml streptomycin (BioWhittaker) supplemented with 1% N2 supplement (Life Technologies), 2% B27 supplement (Life Technologies), 20 ng/ml epidermal growth factor (EGF) (Sigma), and 20 ng/ml basic fibroblast growth factor (FGF) (R & D Systems, Minneapolis, MN).
Measurements of in vitro cell proliferation and self-renewal. Ultra-low binding plates (Corning, Corning, NY) were used for all suspension cultures, proliferation, and self-renewal experiments. To assess proliferation, 104 cells of each genotype were seeded in triplicate. At each time point, resulting neurospheres were trypsinized and counted on a hemocytometer. For self-renewal assays, 10 single neurospheres of each genotype were triturated before plating, and the number of resulting neurospheres generated was counted after 7 d. For the clonogenic incidence experiments (Nunes et al., 2003), we used retroviral green fluorescent protein (MSCV-GFP) to generate GFP-labeled cells and seeded 2000 GFP+ cells from the fourth passage. The clonogenic incidence was calculated as the number of neurospheres formed (× 100) divided by the number of NSCs originally seeded. The number of resulting neurospheres generated after 7 d were counted.
For limiting dilution analysis (Tropepe et al., 1999), 5-1000 cells were seeded in five wells per dilution for each genotype, and the percentage of wells with at least one neurosphere was calculated. One hundred neurospheres were selected randomly, and their diameters were calculated using Analysis Imager software (Soft Imaging System, Lakewood, CO). For ectopic expression of NF1GRD, we used a retrovirus containing a KT3-tagged NF1 GAP (RAS-GTPase activating protein)-related domain (NF1GRD or NF1GRD containing the R1276P NF1 patient mutation) transgene (MSCV-NF1GRD or MSCV-NF1GRD-R1276P) and a puromycin resistance gene (Pac). Cells were infected with either MSCV-NF1GRD or MSCV-Pac (control), and puromycin-resistant cells were selected after 3 d. To determine the role of the cAMP, mitogen-activated protein (MAP) kinase, and phosphatidylinositol 3′-kinase (PI3K)-Akt pathways in NSC proliferation, self-renewal, survival, and differentiation, we used the cAMP analog dibutyril-cAMP (100 μm and 200 μm), the MAP kinase (MAPK) kinase (MEK kinase) inhibitor 2-(2-Amino-3-methoxyphenyl)-4H-1-benzopyran-4-one (PD98059; 20 μm), and the PI3K inhibitor 2-(4-Morpholinyl)-8-phenyl-1(4H-benzopyran-4-one hydrochloride (LY294002; 10 and 20 μm).
In vitro cell survival and death. Nf1+/+ and Nf1-/- NSCs were grown in defined medium containing 1.2 mg/ml sodium bicarbonate, 5 ng/ml insulin (Sigma), 0.1 mg/ml apotransferrin (Life Technologies), 5 ng/ml sodium selenite (Sigma), 6 ng/ml progesterone (Sigma), 0.25% glucose, 6 μg/ml putrescine (Sigma), 2 mm l-glutamine, and 0.1 mm 2-mercaptoethanol (Tropepe et al., 1999; Arsenijevic et al., 2001; Engstrom et al., 2002). This medium did not contain N2 and B27 supplement, EGF, or FGF. After 72 h of growth-factor deprivation and 1 h of 10 μm bromodeoxyuridine (BrdU; Sigma) exposure, cells were allowed to attach to poly-d-lysine-coated plates (Ciccolini and Svendsen, 1998). Cells were fixed in 100% methanol for 10 min at 4°C. DNA denaturation was accomplished using 2N HCl for 30 min at room temperature, followed by neutralization with two changes of 0.1 m sodium borate buffer, pH 8.5, for 10 min (Molofsky et al., 2003). Apoptotic and proliferating cells were detected using antibodies to cleaved caspase-3 (Cell Signaling Technology, Beverly, MA) and BrdU (Abcam, Cambridge, MA), respectively.
NSC differentiation. Individual neurospheres were selected and seeded individually onto poly-d-lysine (50 μg/ml)-coated and fibronectin (10 μg/ml; Life Technologies)-coated wells and allowed to differentiate in growth factor-free N2, B27 supplemented medium (Tropepe et al., 1999). Cells were stained with rabbit anti-GFAP (Abcam), mouse anti-Tuj1 (Covance, Berkeley, CA), and mouse anti-O4 IgM (Chemicon, Temecula, CA) primary antibodies, followed by incubation with appropriate Alexa Fluor-tagged secondary antibodies (Molecular Probes, Eugene, OR) to detect astrocytes, neurons, and oligodendrocytes, respectively. Additional immunocytochemistry also used rabbit anti-BLBP (a gift from Dr. Jeffrey E. DeClue, National Cancer Institute, Bethesda MD), mouse anti-vimentin (Sigma), mouse anti-RC2 IgM (Developmental Studies Hybridoma Bank, Department of Biological Studies, The University of Iowa, Iowa City, IA), and mouse anti-CD44 (Chemicon) antibodies. Undifferentiated neurospheres were characterized using mouse anti-nestin (Chemicon) and rat anti-CD133 (R & D Systems) antibodies.
Western blot analysis. Western blots were performed as described previously (Dasgupta et al., 2003). Active RAS (RAS-GTP) was detected by Raf1-RBD affinity chromatography using the RAS activation assay kit (Upstate Biotechnology, Lake Placid, NY) according to the manufacturer's recommendations (Dasgupta et al., 2005). The primary antibodies used were as follows: rabbit anti-NF1GRP-D (Santa Cruz Biotechnology, Santa Cruz, CA); mouse anti-KT3 (Covance, Richmond, CA); mouse anti-Bcl2 (BD Transduction Laboratories, Lexington KY); mouse anti-phospho-MAPK (Thr202/Tyr204), mouse anti-phospho-Akt (Ser473), rabbit anti-MAPK, and rabbit anti-Akt (all from Cell Signaling Technology); and mouse anti-α-tubulin (Sigma). Appropriate HRP-tagged secondary antibodies (Cell Signaling Technology) were used for detection by enhanced chemiluminescence (Amersham Biosciences, Piscataway, NJ).
5-(and-6)-Carboxyfluorescein diacetate, succinimidyl ester washout experiment. 5-(and-6)-Carboxyfluorescein diacetate, succinimidyl ester (CFSE; Molecular Probes) is a cell-permeable fluorescent dye that is metabolized by nonspecific esterases to result in a compound that gets trapped in the cytosol. Dividing daughter cells receive one-half the amount of dye and, with continued division, lose one-half of the fluorescence with each subsequent cell division. We pulse-labeled both Nf1+/+ and Nf1-/- NSCs with 5 μm CFSE at 37°C for 15 min in the dark. Cells were washed, and one-half of the cells were analyzed by flow cytometry. The remaining cells were allowed to grow for 5 d, and the fluorescence intensity was measured as above (Groszer et al., 2001).
Implantation of NSCs into nu/nu athymic mouse brains. NSCs for implantation were grown for 5-7 d as neurospheres in vitro as described above. During this time, neurospheres were labeled by adenoviral infection using Ad5LacZ [adenovirus expressing the β-galactosidase transgene (Yoon et al., 1996)]. Before injection, neurospheres were mildly trypsinized into single cells and smaller spheres and suspended in PBS. Six- to 8-week-old male mice (three to four mice per genotype per time point) were anesthetized with ketamine (60 μg/g body weight) and xylazine (7.5 μg/g body weight) and placed in a stereotactic frame. A total of ∼105 cells in a volume of 10 μl were injected with a Hamilton Microliter #170 syringe (Hamilton, Reno, NV), through a bore drilled 2 mm posterior, 2 mm lateral from bregma and 3 mm below the brain surface. The scalp was closed with a stapler clip, and the mice were allowed to recover.
In vivo BrdU labeling, tissue preparation, and immunohistological analysis. For in vivo cell proliferation experiments, adult Nf1+/+ and Nf1+/- mice were given injections of BrdU (50 μg/g body weight). One hour after injection, anesthetized animals were perfused transcardially with 0.1 m sodium phosphate buffer, pH 7.4, followed by 4% paraformaldehyde in 0.1 m sodium phosphate buffer. To detect the β-galactosidase activity of injected NSCs, mice were perfused similarly. For in vivo progenitor cell analysis, time-pregnant females were killed at E12.5, and Nf1+/+, Nf1+/-, and Nf1-/- embryos were collected. The brains of adult animals and entire embryos were fixed overnight with 4% paraformaldehyde at 4°C, cryoprotected in 30% sucrose in 0.1 m phosphate buffer at 4°C, embedded in OCT compound, and frozen in cryomolds in liquid nitrogen. Cryosections were collected on Superfrost glass slides. LacZ expression was detected by incubating the sections with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal) staining solution (Bajenaru et al., 2002).
For the BrdU incorporation experiments, frozen sections were postfixed in 100% methanol for 10 min at 4°C. DNA denaturation was accomplished using 2N HCl for 30 min at room temperature, followed by neutralization with two changes of 0.1 m sodium borate buffer, pH 8.5, for 10 min (Molofsky et al., 2003). All other sections were postfixed in 4% paraformaldehyde. The brains 4 and 6 months post-injection (p.i.) were fixed with Bouin's fixative and processed for paraffin embedding and sectioning (4 μm) in the Pharmacology Histology Core at Washington University School of Medicine. Sections were stained with hematoxylin and eosin (H&E) or used for immunohistochemistry using the microwave antigen-retrieval method (Bajenaru et al., 2003).
The following primary antibodies were used for in vivo immunohistochemical analysis: GFAP (Abcam); BLBP (a gift from Dr. Nathaniel Heintz, The Rockefeller University, New York, NY); Sox2, Olig1, and Nkx2.2 (all from Chemicon); Tuj1 (Covance, Berkeley CA); doublecortin (Chemicon); CD44 (Chemicon); MAP2 (Sigma); BrdU (Abcam); and Ki67 (Novocastra, Newcastle, UK). Appropriate Alexa Fluor-tagged secondary antibodies (all from Molecular Probes) were used for detection by immunofluorescence. For paraffin-embedded sections, detection was performed using the Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA) for GFAP expression, the TSA-Plus fluorescein kit (PerkinElmer, Boston, MA) for BLBP expression, and the VIP substrate kit (Vector Laboratories) for Ki67 immunostaining. All sections were photographed with a digital camera (Optronics, Goleta, CA) attached to an inverted microscope (Nikon, Melville, NY).
Statistical analysis. Student's t test was used to calculate statistical significance with p < 0.05 representing a statistically significant difference.
Results
Neurofibromin negatively regulates proliferation and self-renewal of NSCs in vitro
Previous studies have shown that neurofibromin loss in embryonic or postnatal astroglial cells results in modest increases in astrocyte proliferation (Bajenaru et al., 2002). To determine the role of the Nf1 gene in astroglial progenitor cell growth control, we used an in vitro NSC culture system. In contrast to Nf1-/- astroglial cells from either E12.5 or postnatal day 1 (P1) mice, which exhibit a twofold to threefold increase in proliferation (Bajenaru et al., 2002), Nf1-/- NSCs proliferated at a significantly higher rate at every time point studied. At 96 h, we observed a 12-fold increase in the number of Nf1-/- NSCs compared with Nf1+/+ NSCs (Fig. 1a).
To determine the consequence of neurofibromin loss on NSC self-renewal, we counted the number of secondary neurospheres generated by single primary neurospheres. Whereas only 2.25% of cells of wild-type neurospheres produced secondary neurospheres, 16.2% of cells of Nf1-/- neurospheres generated secondary neurospheres (Fig. 1b). In addition, we found that Nf1-/- NSCs had a clonogenic incidence of 15.37%, compared with 1.52% for Nf1+/+ and 4.36% for Nf1+/- NSCs, suggesting a high capacity for secondary neurosphere generation (Fig. 1c).
To determine whether neurospheres formed by Nf1-/- NSCs develop as a result of clonal expansion versus cell aggregation, we labeled single NSCs with retroviral GFP. The resulting neurospheres were composed entirely of either GFP+ or GFP-cells, suggesting a clonal origin of Nf1-/- neurospheres. Finally, we used limiting dilution analysis to determine the minimal number of cells required to form a neurosphere. Compared with Nf1+/+ NSCs, which required 20 ± 7.5 cells to form at least one sphere per well, the minimal number of Nf1-/- NSCs required to form at least one neurosphere per well was 5 ± 2.5. Although 10% of all wells seeded with 10 Nf1-/- NSCs contained at least one neurosphere, we did not find any wells containing neurospheres using either Nf1+/+ or Nf1+/- NSCs at this cell dilution (Fig. 1d).
In addition, as the Nf1-/- NSCs were allowed to propagate, the average diameter of their derivative neurospheres and the number of individual cells per sphere were significantly greater than that observed with wild-type NSCs (Fig. 1e). Together, these results demonstrate that loss of Nf1 results in a substantial increase in stem cell proliferation and self-renewal in vitro. Moreover, we observed a significant effect of reduced neurofibromin expression (Nf1+/- NSCs) on NSC self-renewal in all of these analyses (Fig. 1a-e), supporting the hypothesis that Nf1 heterozygosity may also influence NSC proliferation and renewal.
We next wanted to determine whether the increased Nf1-/- NSC self-renewal confers both a survival and proliferative advantage in vitro. To analyze proliferation, we performed BrdU incorporation analysis on growth factor-deprived NSCs. As shown in Figure 2a, 30% of Nf1-/- NSCs incorporated BrdU (Fig. 2a, bottom) compared with <1% of similarly treated Nf1+/+ cells (Fig. 2a, top). In addition, using the CFSE washout method, we observed an increase in the rate of cell division in Nf1-/- NSCs. Whereas 16.28% pulse-labeled Nf1+/+ NSCs demonstrated a shift in fluorescence, 70.83% Nf1-/- NSCs exhibited a fluorescence shift after 5 d in culture (Fig. 2b), suggesting that Nf1-/- NSCs had passed through a significantly greater number of cell divisions than Nf1+/+ NSCs.
To determine whether neurofibromin loss also conferred a survival advantage, we used activated caspase-3 immunostaining as a marker of apoptosis. We observed that growth factor-deprived Nf1-/- NSCs exhibited a 71.06 ± 18.35% decrease in activated caspase-3 immunoreactivity compared with Nf1+/+ NSCs (Fig. 2c). Collectively, these data indicate that the dramatic increase in Nf1-/- NSC self-renewal is attributable both to decreased cell death and increased cell proliferation.
Because neurofibromin functions as a negative regulator of RAS (Ballester et al., 1990; Martin et al., 1990; Xu et al., 1990), we next examined activation of RAS and its downstream effectors in Nf1-/- NSCs to provide a biochemical correlate for the increased survival and proliferative advantage observed. Growth factor-deprived Nf1-/- NSCs exhibited an approximate sixfold increased RAS activation (RAS-GTP), approximate threefold increased phospho-Akt expression, approximate eightfold increased phospho-MAPK expression, and approximate sixfold increased expression of the anti-apoptotic protein Bcl2 (Fig. 2d). These findings demonstrate that neurofibromin loss in NSCs results in impaired RAS regulation, as shown previously for Nf1-/- astrocytes (Bajenaru et al., 2002; Dasgupta et al., 2005).
Loss of Nf1 facilitates survival and engraftment of NSCs in vivo
To provide an in vivo correlate for the in vitro proliferative and survival advantage observed with Nf1-/- NSCs, we wanted to determine whether loss of neurofibromin enabled NSCs to survive and engraft in naive recipient brains in vivo. Previous studies have shown that Nf1-/- astrocytes were unable to survive or engraft into the cortices of athymic immunocompromised (nu/nu) mice in vivo, and completely disappeared within 1 month after injection (Bajenaru et al., 2003). To determine whether Nf1-/- NSCs were capable of surviving as explants in vivo, we injected β-galactosidase-expressing Nf1+/+, Nf1+/-, and Nf1-/- NSCs into the cortex of adult male nu/nu mice. As shown in Figure 3a, Nf1+/+, Nf1+/-, and Nf1-/- NSCs could be detected at the injection site 3 d p.i. However, wild-type NSCs quickly disappeared by 1 month p.i., and only very few LacZ+ cells were found at the injection site. We did not observe any LacZ+ Nf1+/+ NSCs at brain sites distant from the injection site (data not shown). In contrast, many LacZ+ Nf1-/- cells were detected in and around the injection site at 1 month. By 2 months p.i., LacZ+ Nf1+/+ cells were almost undetectable, whereas large numbers of Nf1-/- cells were clearly visible in and around the injection tract. Moreover, we detected isolated or occasional groups of four to five LacZ+ Nf1-/- cells distant from the injection site, with some Nf1-/- cells detected in the contralateral cortex (data not shown).
Analysis of the brains from mice implanted with Nf1+/- NSCs demonstrated the presence of a small number of scattered LacZ+ cells in the injection tract at 1 month p.i. but there were significantly fewer cells than observed in mice given injections of Nf1-/- NSCs.
Next, we wanted to determine whether the injected Nf1-/- NSCs were proliferating in the recipient brains. Injection tracts of Nf1-/- NSCs were identified by H&E staining of cortical sections (Fig. 3b). To identify proliferating cells in the injection tract, we performed Ki67 immunohistochemical analysis. Clusters of Ki67+ cells were detected in the injection tract at 4 and 6 months p.i. (Fig. 3b). No Ki67 immunoreactivity was detected in the cortex of nu/nu mice given injections of Nf1+/+ NSCs (data not shown).
To determine whether the engrafted NSCs were able to differentiate, cortical sections at 4 and 6 months p.i. were stained for expression of GFAP (astroglial cells), CD44 [astrocyte-restricted precursor (Liu et al., 2004)] and BLBP [astroglial progenitors (Feng et al., 1994)]. Robust GFAP and BLBP immunoreactivity was detected in and around the injection tract at 4 and 6 months p.i of Nf1-/- NSCs (Fig. 3c). In contrast, no CD44 immunoreactivity was observed (data not shown). We observed no change in the intensity of GFAP expression from 4 to 6 months p.i.; however, BLBP expression was reduced or absent in most GFAP+ cells at 6 months p.i., suggesting the possibility that these BLBP+, GFAP+ Nf1-/- immature astrocytes continue to differentiate in vivo. No intensely stained GFAP+ or BLBP+ cells were detected in the cortex of nu/nu mice given injections of Nf1+/+ NSCs. We did not observe any histological evidence for tumor formation in mice given injections of Nf1-/- NSCs even after 12 months (data not shown).
To determine whether these NSCs also differentiated into neurons, we stained the injected brain sections with antibodies to NeuN and MAP2 (data not shown). No neuronal nuclei or processes were identified in the injection tract, suggesting that the Nf1-/- NSCs differentiated into astrocytes and not neurons. Collectively, these results demonstrate that Nf1 loss results in the generation of a persistent population of GFAP+, BLBP+ proliferating cells, which could serve as a source for preneoplastic cells important for NF1-associated glioma formation.
Nf1 loss in NSCs does not impair multi-lineage differentiation but alters astroglial cell differentiation.
Like wild-type NSCs, undifferentiated Nf1-/- and Nf1+/- NSCs robustly expressed NSC markers, including nestin and CD133 (data not shown). To determine whether neurofibromin loss in NSCs affects the ability of these stem cells to differentiate into astroglial cells, we allowed EGF/FGF-expanded NSCs to differentiate into neurons and glia on growth factor withdrawal. After growth factor withdrawal, NSCs of all three genotypes differentiated into GFAP+ astrocytes, Tuj1+ neurons, and O4+ oligodendrocytes.
We observed no differences in the relative numbers of astrocytes or neurons between Nf1+/+, Nf1+/-, and Nf1-/- NSCs (Fig. 4a); however, we found a 24% increase in oligodendrocytes generated by Nf1-/- NSCs. This increase in oligodendrocyte production is consistent with recent results obtained using Nf1-/- spinal cord-derived progenitor cells (Bennett et al., 2003).
Although Nf1+/+ NSCs differentiated into large morphologically distinct GFAP+ cells characteristic of differentiated astrocytes, Nf1-/- NSCs gave rise to thin GFAP+ cells with long filamentous extensions (Fig. 4b). The GFAP+ cells that differentiated from Nf1+/- NSCs exhibited a phenotype intermediate between Nf1+/+ and Nf1+/- NSCs, although the filamentous appearance of Nf1-/- astrocytes was less apparent in these cells. No significant differences in oligodendrocyte morphology were observed between O4+ Nf1+/+ and Nf1-/- cells; however, a subpopulation of Nf1-/- O4+ cells were detected that also coexpressed GFAP (Fig. 4c).
Because Nf1-/- NSCs generated GFAP+ cells, which morphologically resembled immature glia, we performed immunohistochemical analyses using antibodies to BLBP, vimentin, CD44, and RC2, which represent markers of astroglial progenitors and immature astroglia. As shown in Figure 5, Nf1+/+ cells robustly expressed GFAP with little expression of BLBP (a, top), vimentin (b, top), or RC2 (c, top). In contrast, Nf1-/- GFAP+ cells robustly expressed all the three markers of immature astrocytes [BLBP (Fig. 5a, bottom); vimentin (Fig. 5b, bottom); RC2 (Fig. 5c, bottom)]. Neither Nf1+/+, Nf1+/- nor Nf1-/- GFAP+ cells expressed CD44 after in vitro differentiation (data not shown).
Hyperproliferation and altered astroglial differentiation in Nf1-/- NSCs is rescued by expressing the GRD of neurofibromin
To determine whether the dramatic increases in self-renewal and proliferation of Nf1-/- NSCs were the direct result of loss of neurofibromin GTPase-activating domain function, we infected Nf1-/- NSCs with MSCV-Pac (vector control), MSCV- NF1GRD (NF1-GAP domain; residues 1172-1538), or MSCV- NF1GRD R1276P (NF1GRD with no RASGAP activity resulting from presence of an NF1 patient mutation). In these experiments, we found that NF1GRD expression reduced the proliferation (Fig. 6a) and self-renewal (Fig. 6b) observed in Nf1-/- NSCs to levels similar to wild-type NSCs (77.33 and 80.2%, respectively) at 96 h after seeding. No changes in proliferation or self-renewal were observed in Nf1-/- cells infected with MSCV-Pac or MSCV-NF1GRD R1276P or in Nf1+/+ cells infected with either MSCV-Pac or MSCV-NF1GRD.
To provide additional support for the role of RAS pathway activation in Nf1-/- NSC hyperproliferation, we wanted to determine whether Nf1-/- NSC hyperproliferation could be reversed by pharmacological inhibition of MEK kinase. In these experiments, Nf1-/- NSCs were treated with the MEK inhibitor PD98059 at pharmacological doses shown previously to restore Nf1-/- MEK hyperactivation to wild-type levels (data not shown). Treatment of Nf1-/- NSCs with 20 μm PD98059 reduced the proliferation (Fig. 6a) and self-renewal (Fig. 6b) by 84.5 and 86.2%, respectively. This level of inhibition was comparable with that observed after the expression of the NF1GRD. The PI3K inhibitor LY294002 (at both 10 and 20 μm concentrations) severely affected survival of NSCs of all genotypes to comparable levels (data not shown).
Because both increased self-renewal and proliferation of Nf1-/- NSCs could be reverted by introducing the NF1GRD, we tested whether NF1GRD expression could restore the normal activation or expression state of these RAS pathway effectors. After introduction of the NF1GRD, the expression of Bcl2, phospho-Akt and phospho-MAPK (Fig. 6c) reverted to wild-type levels.
To determine whether the defect in astroglial differentiation reflected loss of NF1GRD function, we allowed Nf1-/- NSCs expressing NF1GRD to undergo differentiation in vitro. As shown in Figure 6d, in contrast to NSCs infected with MSCV-Pac (control) virus, expression of the NF1GRD completely rescued the Nf1-/- NSC astroglial cell differentiation abnormalities. After NF1GRD expression, Nf1-/- GFAP+ cells became morphologically distinct and expressed almost exclusively GFAP with dramatic reductions in BLBP expression.
We have shown previously that Nf1-/- astrocytes have reduced G-protein-stimulated cAMP generation, which could be rescued by the addition of exogenous cAMP (Dasgupta et al., 2003). However, long-term exposure to cAMP at concentrations sufficient to restore reduced cAMP function in Nf1-/- astrocytes had no effect on the increased proliferation, self-renewal, or aberrant glial differentiation of Nf1-/- NSC (data not shown). Collectively, these results demonstrate that the ability of neurofibromin to modulate NSC proliferation, survival, and differentiation is regulated by sequences contained with the NF1GRD and reflects neurofibromin RAS regulation.
Nf1 heterozygosity in NSCs also results in defects in glial cell differentiation
Because the brains of children with NF1 develop from NF1+/- progenitor cells, we examined GFAP+ astrocytes differentiated from Nf1+/- NSCs from multiple embryos. Between 38 and 60% of GFAP+ Nf1+/- glial colonies exhibited long filamentous processes similar to Nf1-/- glial cells and expressed BLBP, in contrast to Nf1+/+ GFAP+ astrocytes, which showed little BLBP expression. Representative fields of differentiating GFAP+ Nf1+/+ (top) and GFAP+, BLBP+ Nf1+/- astrocytes (bottom) are shown in Figure 7. Collectively, these results raise the possibility that glial differentiation defects can result from Nf1 heterozygosity.
Neurofibromin regulates NSC number and proliferation in vivo
Based on the observed abnormalities in proliferation, survival, and astroglial cell differentiation in both Nf1+/- and Nf1-/- NSCs in vitro, we hypothesized that there would be increased cell proliferation and more neural stem/progenitor cells in Nf1-/- and Nf1+/- embryonic brains in vivo. To provide an in vivo correlate for our in vitro findings, we analyzed frozen sections of embryonic brains at day 12.5, because Nf1 loss results in embryonic lethality by E13.5. We used antibodies against Sox2, Olig1, NKx2.2, and BLBP to identify stem/progenitor cells. Immunohistochemistry revealed significantly increased numbers of Sox2+ cells in the neural tube and in the margins of lateral ventricles of Nf1-/- embryos, compared with wild-type embryos (Fig. 8a). Importantly, the developing brains of Nf1+/- mice also contained more Sox2+ cells than Nf1+/+ embryos. Although very few BLBP+ cells were observed in the neural tube of Nf1+/+ embryos, numerous BLBP+ cells were observed in the neural tubes of the Nf1+/- and Nf1-/- embryos (Fig. 8a). Similarly, we also observed considerably higher numbers of Nkx2.2+ and Olig1+ glial precursors in the neural tube from Nf1+/- and Nf1-/- embryos compared with Nf1+/+ embryos (Fig. 8a). To determine whether increased numbers of neuronal progenitors were found in Nf1-/- or Nf1+/- embryonic brains, we performed immunohistochemistry with Tuj1 and doublecortin antibodies. We did not observe any reproducible differences in the numbers of Tuj1+ or doublecortin+ cells in Nf1+/- or Nf1-/- embryo brains relative to Nf1+/+ embryos (data not shown).
Because Nf1-/- embryos die in utero, we next wanted to determine whether more proliferating cells persist in the brains of adult Nf1+/- mice. As shown in Figure 8b, increased numbers of BrdU+, proliferating cells were observed in the subventricular zone (twofold increase) and in the hippocampus (fivefold increase) of Nf1+/- mice compared with wild-type mice. In addition, we did not observe any reproducible differences in the numbers of Tuj1+ or MAP2+ cells in adult Nf1+/- brain compared with control wild-type littermates (data not shown). Together, these results demonstrate that increased numbers of neural stem/progenitor cells result from dose-dependent decreases in neurofibromin expression, which likely lead to the formation of a persistent population of proliferating cells in the brains of adult Nf1+/- mice.
Discussion
Much of our understanding of the function of neurofibromin derives from studies focused on NF1 growth regulation and its relationship to tumor formation. In this regard, NF1 loss in vivo is associated with neurofibroma formation (Cichowski et al., 1999), malignant peripheral nerve sheath tumor development (Vogel et al., 1999), and leukemogenesis (Bollag et al., 1996; Largaespada et al., 1996; Birnbaum et al., 2000). Nf1-/- Schwann cells (neurofibromas and malignant peripheral nerve sheath tumor) and myeloid cells (leukemia) exhibit increased cell proliferation in vitro and in vivo. In contrast, there is considerably less known about the role of neurofibromin in progenitor cell growth regulation within the CNS. Consistent with the function of neurofibromin as a negative growth regulator, we show that both reduced (Nf1+/-) and absent (Nf1-/-) neurofibromin expression in NSCs lead to increased stem cell proliferation. Although there is a clear effect of Nf1 haploinsufficiency (3-fold increase over wild-type cells), the most dramatic effects were seen in Nf1-deficient cells (12-fold increase). Similar analyses of NSCs from Pten-/- mice revealed a fivefold increase in self-renewal over wild-type NSCs (Groszer et al., 2001). Compared with Pten, Nf1-/- NSCs showed ∼10-fold increase in self-renewal over wild-type NSCs, suggesting that neurofibromin may play a more significant role in regulating stem cell self-renewal and proliferation in the developing brain.
Previous studies have suggested that neurofibromin exerts its primary effect on cell growth by modulating mitogenic signaling pathways downstream of RAS. Consistent with this finding, we observed hyperactivation of RAS and two of its downstream effectors, MAPK and Akt. These signaling abnormalities are reflected by both an increase in cell proliferation (Raf-MAPK) and a decrease in cell death (Akt). Surprisingly, we observed increased expression of the antiapoptotic Bcl2 protein in Nf1-/- NSCs, suggesting the possible contribution of both Akt and Bcl2 to increased cell survival. This is the first report of Bcl2 overexpression in a cell with Nf1 loss of function. The reversal of Bcl2 expression to wild-type levels after NF1GRD expression in Nf1-/- NSCs supports a RAS-dependent mechanism for regulating Bcl2 and modulating programmed cell death (apoptosis). These observations are consistent with studies in which inducible oncogenic RAS expression resulted in upregulation of Bcl2 expression, reduced apoptosis, and increased survival of hematopoietic cells (Kinoshita et al., 1995). Previous data from our laboratory on Nf1-/- postnatal astrocytes demonstrated that neurofibromin loss results in impaired cAMP generation (Dasgupta et al., 2003). Because cAMP treatment of Nf1-/- NSCs at concentrations sufficient to restore cAMP-dependent functions in Nf1-/- astrocytes did not reduce Nf1-/- NSC self-renewal, it is unlikely that the cAMP regulatory function of neurofibromin significantly contributes to the growth properties of Nf1-/- NSCs and that these biological effects are mediated by neurofibromin RAS regulation. These observations are consistent with our previous findings in which activated K-RAS (which does not impair intracellular cAMP generation in astrocytes) can substitute for Nf1 loss in the formation of optic glioma in vivo (Dasgupta et al., 2005).
In addition to abnormal growth control in Nf1-/- NSCs, we observed both morphological and immunochemical changes suggestive of an immature glial phenotype resulting from Nf1 inactivation. The glia that derive from Nf1-/- NSCs exhibit long filamentous processes, and many of the cells express markers typically associated with glial precursors, including BLBP, RC2, and vimentin. It is worth noting that human gliomas often express such glial precursor markers suggestive of the possible presence of immature progenitor cells within the tumors (Abaza et al., 1998). Similar to the growth defects, dysregulated RAS activity likely accounts for these abnormalities in glial differentiation, because inhibiting RAS activity using pharmacological inhibitors (data not shown) or by expressing the neurofibromin RAS-GAP domain (NF1GRD) corrected the Nf1-/- NSC astroglial differentiation defect.
Another property of Nf1-/- NSCs was their ability to persist as focal collections of hyperproliferating cells in the injection tract for >6 months. In contrast to wild-type NSCs, Nf1-/- embryonic (E12) astrocytes, or postnatal (P1-P2) Nf1-/- astrocytes, Nf1-/- NSCs exhibited long-term survival in vivo. These results raise the possibility that Nf1 inactivation in NSCs or astroglial progenitors leads to the generation of a small population of atypical stem cells with the ability to aberrantly proliferate and survive for long periods of time in the adult brain and potentially serve as a reservoir of “preneoplastic” cells. Despite the lack of obvious tumor formation after 1 year, these progenitor cells may represent targets for additional genetic or cellular changes that might lead to glioma development. Finally, we noted that Nf1-/- NSCs were found at sites distal from the original injection, including the contralateral hemisphere. These observations suggest that a small number of Nf1-/- progenitor cells are capable of widely disseminating throughout the brain and establishing themselves as stem cells with dramatic self-renewal, proliferative, and survival advantages over wild-type cells.
The expansion of Sox2+, BLBP+, Olig1+, and Nkx2.2+ stem/progenitor cells in the developing Nf1-/- brain suggests that partial or complete neurofibromin loss in the neural progenitor cells contributes to the abnormal proliferation and differentiation of these cells in vivo. Expression of the Sox2 transcription factor is a molecular signature of proliferating stem/progenitor cells, and Sox2 expression is necessary for maintaining neural stem/progenitor cell identity (Uwanogho et al., 1995; Graham et al., 2003). Similarly, BLBP is expressed as early as E10 in multipotent proliferating cells within the CNS, which are thought to represent progenitors of astroglial cells (Feng et al., 1994; Li et al., 2004). Finally, both Olig1 and Nkx2.2 are expressed in cells within the developing neural tube (Lu et al., 2000; Zhou et al., 2001; Liu et al., 2004). Based on our findings, we propose that loss of neurofibromin in NSCs results in the expansion of progenitors that contribute to the abnormal CNS features seen in individuals with NF1. As has been proposed previously for other genes involved in astroglial cell differentiation, such as p27kip1 (Casaccia-Bonnefil et al., 1999) or Pten (Groszer et al., 2001), reduced or absent expression of the neurofibromin tumor suppressor results in abnormal brain development. In this regard, 12.5% of Nf1-/- embryos exhibited gross defects of cranial neural tube closure, including exencephaly (Lakkis et al., 1999).
We observed a moderate, yet significant increase in proliferation and self-renewal of Nf1+/- NSCs and aberrant expression of the immature glial marker BLBP in GFAP+ Nf1+/- astrocytes in vitro. Our in vitro observations are corroborated by the presence of increased numbers of neural progenitors in the embryonic Nf1+/- brain. We hypothesize that these Nf1+/- cells also have the potential to proliferate aberrantly during Nf1+/- brain development. Consistent with this notion, we observed more proliferating cells in the adult Nf1+/- brain. There are two potential scenarios in which increased numbers of NF1+/- progenitors might result in NF1-associated clinical brain abnormalities. First, the retention of increased numbers of proliferating stem/progenitor cells in NF1+/- human brains during development and their persistence in the adult brain would increase the statistical likelihood of undergoing inactivation of the one remaining NF1 allele. Homozygous loss of NF1 expression would then result in elevated RAS pathway activation and neoplastic transformation, culminating in glioma formation. Although it is not clear that cancer stem cells are equivalent to NSCs, these cancer-associated progenitor cells are multipotent and have the ability to self-renew like NSCs. Moreover, these cancer stem cells, when re-implanted into the brains of naive recipient rodents, can recapitulate the original tumor, suggesting that these cells might possess all of the required cellular and genetic changes sufficient for tumorigenesis and might be a logical target for anticancer therapy. Given the similarities between NSCs and cancer stem cells, future experiments might use Nf1-/- NSCs as a manipulable system for NF1 tumor modeling studies.
Second, because the brains of children with NF1 develop from NF1+/- stem cells, small impairments in stem/progenitor cell function might have profound implications for normal brain development. In this regard, over one-half of children with NF1 have cognitive deficits and learning disabilities (North et al., 1997), and most have significantly enlarged heads (Clementi et al., 1999). Although we did not notice any gross morphological abnormalities in the neurons from either embryonic Nf1-/- or adult Nf1+/- brains in vivo, we did observe shorter neurites in Tuj1+ neurons that differentiated from Nf1+/- and Nf1-/- neurospheres in vitro (data not shown). Moreover, we did not observe any reproducible differences in the numbers of Tuj1+ or MAP2+ cells in adult Nf1+/- mouse brains (data not shown). In light of the elegant studies by Costa et al. (2002) demonstrating that Nf1+/- neurons in situ have increased GABA-mediated inhibition and that Nf1+/- mice exhibit defects in long-term potentiation, future experiments will be needed to analyze the functional properties of the Nf1+/- neurons with shortened neuritic processes detected in the present study.
In summary, the present report demonstrates a dose-dependent effect of Nf1 loss on NSC proliferation, survival, self-renewal, and differentiation in vitro and in vivo. These findings suggest that NF1 heterozygosity in progenitor cells during brain development and even in the adult CNS (Sanai et al., 2004) may result in a predisposition to glioma formation. The ability to model these phenotypes in vitro and in vivo provides a unique opportunity to define the molecular basis for the nervous system abnormalities seen in patients with NF1.
Footnotes
This work was supported by a grant from the United States Army (DAMD-17-03-1-0215 to D.H.G.). We thank Katherine Gold for advice with neurosphere generation, Christine Kamp for technical assistance, and Dr. Jason Weber for the MSCV-GFP viral construct.
Correspondence should be addressed to Dr. David H. Gutmann, Department of Neurology, Washington University School of Medicine, Box 8111, 660 South Euclid Avenue, St. Louis, MO 63110. E-mail: gutmannd{at}neuro.wustl.edu.
Copyright © 2005 Society for Neuroscience 0270-6474/05/255584-11$15.00/0