Individuals with the tumor predisposition syndrome, neurofibromatosis 1 (NF1), are prone to development of nervous system tumors, including neurofibromas and pilocytic astrocytomas. Based on the ability of the NF1 gene product (neurofibromin) to function as a GTPase activating protein for RAS, initial biologically based therapies for NF1-associated tumors focused on the use of RAS inhibitors, but with limited clinical success. In an effort to identify additional targets for therapeutic drug design in NF1, we used an unbiased proteomic approach to uncover unanticipated intracellular signaling pathways dysregulated in Nf1-deficient astrocytes. We found that the expression of proteins involved in promoting ribosome biogenesis was increased in the absence of neurofibromin. In addition, Nf1-deficient astrocytes exhibit high levels of mammalian target of rapamycin (mTOR) pathway activation, which was inhibited by blocking K-RAS or phosphatidylinositol 3-kinase activation. This mTOR pathway hyperactivation was reflected by high levels of ribosomal S6 activation in both Nf1 mutant mouse optic nerve gliomas and in human NF1-associated pilocytic astrocytoma tumors. Moreover, inhibition of mTOR signaling in Nf1−/− astrocytes abrogated their growth advantage in culture, restoring normal proliferative rates. These results suggest that mTOR pathway inhibition may represent a logical and tractable biologically based therapy for brain tumors in NF1.
Neurofibromatosis 1 (NF1) is a common autosomal dominant tumor predisposition syndrome affecting 1:3,000 individuals worldwide (1). Although adolescents and adults with NF1 develop benign Schwann cell tumors affecting the peripheral nervous system (neurofibromas), children with NF1 are prone to the development of WHO grade 1 astrocytomas (gliomas) typically located in the optic pathway, hypothalamus, and brain stem (2). These low-grade gliomas (pilocytic astrocytomas) are composed of neoplastic astrocytes with characteristic histologic features (3).
Previous studies have shown that the NF1 gene encodes a protein termed neurofibromin, which contains a small domain with sequence similarity to a family of proteins that function to inhibit RAS activation (4). These GTPase activating protein (GAP) molecules associate with RAS and accelerate the conversion of RAS from an active, GTP-bound form to an inactive GDP-bound form (5). Loss of neurofibromin in NF1-associated benign and malignant tumors is associated with high levels of RAS activation (6–9). Moreover, expression of the NF1GAP domain in NF1-deficient cells is sufficient to reverse the hyperproliferative effects of neurofibromin loss (10, 11). This RAS-GAP function of neurofibromin provided the initial impetus for the use of anti-RAS therapies for NF1-associated tumors. RAS activation requires isoprenylation for membrane localization, which can be blocked by farnesyltransferase inhibitors in vitro and in vivo. In preclinical studies, treatment of both NF1-deficient human and Nf1−/− mouse cells with farnesyltransferase inhibitors results in marked reductions in cell proliferation (12, 13); however, the use of these agents in patients with NF1 has showed little effect on tumor growth (14).
The limited success of RAS inhibitors in clinical trials has raised the intriguing possibility that additional targets for NF1-associated tumor therapy might exist. Recent studies on Nf1−/− astrocytes have shown that the proliferative and cytoskeleton-associated cellular defects result from selective hyperactivation of K-RAS, and not H-RAS (15). In these experiments, K-RAS, but not H-RAS activation, is specifically associated with increased Rac1 pathway signaling in Nf1−/− astrocytes. In addition, other biochemical functions have been ascribed to neurofibromin, including cyclic AMP regulation, which also affects NF1 growth control (16, 17), underscoring the potential complex signaling networks involved in promoting proliferation in the absence of neurofibromin.
Given the various RAS and non-RAS effects of neurofibromin loss on growth regulation, we chose to use an unbiased proteomics-based approach to identify unanticipated signaling pathways dysregulated in Nf1-deficient astrocytes, which could serve as targets for biologically based therapies for NF1-associated glioma. Herein, we report that neurofibromin loss in astrocytes results in RAS- and phosphatidylinositol 3-kinase (PI3K)–dependent hyperactivation of the mammalian target of rapamycin (mTOR) signaling pathway. This mTOR hyperactivation is observed in both genetically engineered mouse and human optic nerve gliomas in vivo, and mTOR inhibition results in amelioration of the Nf1−/− astrocyte growth advantage in vitro.
Materials and Methods
Mice. GFAP-Cre-IRES-LacZ transgenic (GFAPCre) mice were generated as previously described (18). Lox-stop-lox (LSL)-K-RASG12D mice, generously provided by Dr. Tyler Jacks (Massachusetts Institute of Technology, Cambridge, MA; ref. 19) were crossed with GFAPCre mice to obtain (LSL)-K-RASG12D; GFAPCre mice. Nf1flox/flox mice were a kind gift from Dr. Luis Parada (University of Texas Southwestern Medical Center, Dallas, TX). Mice were used in accordance with established animal studies protocols at the Washington University School of Medicine.
Primary astrocyte cultures. Murine neocortical astroglial cultures, containing >95% glial fibrillary acidic protein (GFAP) positive cells (astrocytes), were generated from postnatal day 2 Nf1flox/flox and (LSL)-K-RASG12D; GFAPCre transgenic pups as previously described (18). To inactivate Nf1 expression, Nf1flox/flox astrocytes were treated with Ad5Cre (University of Iowa Gene Transfer Vector core, Iowa City, IA) or Ad5-lacZ (control), and neurofibromin loss was confirmed by Western blot. Astrocyte proliferation assays were done by [3H]thymidine incorporation.
Proteomic analysis. Astrocytes were lysed in rehydration/sampling buffer (Bio-Rad, Hercules, CA), and protein concentrations determined using the DC assay (Bio-Rad). Two hundred micrograms of proteins and 12.5 μL of 200 mmol/L tributylphosphine (Bio-Rad) were mixed and incubated for 20 minutes at room temperature. Sample loading for the first dimension was done by passive in-gel rehydration. Immobilized pH-gradient strips (Bio-Rad) were focused according to standard protocol and stopped at 30,000 V h. After the first dimension, the strips were equilibrated for 20 minutes in buffer containing 6 mol/L urea, 0.375 mol/L Tris (pH 8.8), 20% glycerol, 2% Tween 20, 0.2% SDS, and 1 mmol/L EDTA with 2% dithiothreitol, washed for 10 minutes in the same buffer without dithiothreitol, and then equilibrated for 20 minutes in the same buffer with 5% iodoacetamide instead of dithiothreitol. Before the second dimension, the strips were rinsed in 1× SDS-PAGE running buffer, and placed on the top of 10% SDS-PAGE with 4% stacking gel. Separated proteins were stained with SYPRO-Ruby (Bio-Rad) according to the manufacturer's protocol. Gel images were analyzed using PDQuest software (Bio-Rad). Spots of interest were excised and processed for trypsin digestion. Tryptic peptides were calibrated with Sequazyme peptide mass standards kit (PE Biosystem, Norwalk, CT) and analyzed by matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (Voyager DE Pro, Applied Biosystems, Foster City, CA). Identification of proteins was done using MS-Fit software (http://prospector.ucsf.edu/ucsfhtml4.0/msfit.htm).
Western immunoblot analysis. Western blot analysis was done as described (16) using the following antibodies: anti-NF1GRP, nucleolin, and cyclin D1 (Santa Cruz Biotechnology, Santa Cruz, CA), phospho-S6 (Ser240/244; Cell Signaling Technology, Beverly, MA), α-tubulin and fibrillarin (Sigma, St. Louis, MO), anti-T7 (Novagen, Madison, WI), anti-KT3 (Babco, Richmond, CA), nucleophosmin (NPM, Zymed, San Francisco, CA), and L7 (Novus Biologicals, Littletown, CO).
Protein translation. Astrocytes were incubated with dialyzed serum (5%) in the absence of methionine for 30 minutes. [35S]Methionine was added to the culture medium and astrocytes were incubated at 37°C for 5, 10, 15, 30, and 45 minutes in the presence or absence of rapamycin. At the end of each incubation period, astrocytes were harvested in radioimmunoprecipitaion assay buffer [50 mmol/L Tris (pH 7.4), 150 mmol/L NaCl, 1% Triton X-100, 0.1% SDS, 0.5% deoxycholic acid]. Proteins were precipitated with trichloroacetic acid, and [35S]methionine-labeled proteins were detected by scintillation counting.
Retroviral transduction of astrocytes. Retroviral transduction was done as previously reported (10). Expression of NF1GRD was confirmed by Western blot. Dominant negative K-RAS (K-RASN17) was generated by site-directed mutagenesis as described previously (20, 21). Transduction of primary astrocytes was done using MSCV-K-RAS N17-IRES-GFP and MSCV-IRES-GFP (control; ref. 22).
Immunohistologic analysis. Immunohistochemical analysis was done on paraffin-embedded mouse optic nerve sections and human pilocytic astrocytomas with rabbit anti-phospho S6 antibodies (18). Human tumors were used in accordance with active human studies protocols at Washington University.
Results and Discussion
Proteomic analysis uncovers hyperactivation of the mRNA translation machinery in Nf1-deficient astrocytes. Mass spectrometry has emerged as a powerful tool to analyze entire cellular proteomes (23). Using a combination of two-dimensional gel separation and mass spectrometry, unique signaling cascades have been identified that result from differential activation of mitogen-activated protein-kinase (24). Based on our recent studies demonstrating that distinct RAS isoforms and downstream pathway effectors are dysregulated in response to Nf1 inactivation, we hypothesized that loss of neurofibromin expression would result in the activation of unique downstream signaling cascades. To examine this possibility, we did a large-scale proteomic analysis of primary astrocytes lacking neurofibromin. Proteins extracted from asynchronous wild-type (Nf1+/+) and Nf1−/− astrocytes were separated by two-dimensional electrophoresis using protein isoelectric points in the first dimension and protein molecular weights in the second dimension. When compared using PDQuest software, more than 100 visualized proteins exhibited significant expression deviation (>1.5-fold) between Nf1+/+ and Nf1−/− extracts (Fig. 1A). Differentially expressed proteins were excised and analyzed by MALDI-TOF for initial identification. For example, ribosomal protein L21 was positively identified by peptide matching (Fig. 1A) and exhibited a 1.5-fold increase in protein expression in Nf1−/− astrocytes compared with wild-type astrocytes. Although a number of differentially expressed proteins remained unclassified in the proteomic databases (listed as ESTs or no protein matches), several other proteins involved in protein synthesis were positively identified. As shown in Fig. 1B, NPM, S19, L7, and L10a displayed increased expression in Nf1−/− astrocytes and were subsequently positively identified by MALDI-TOF analysis.
To confirm the validity of our proteomic approach, protein lysates from Nf1+/+ and Nf1−/− astrocytes were analyzed by immunoblotting with antibodies recognizing initially identified proteins from the mass spectrometry screen. In agreement with our previous findings that Nf1−/− astrocytes proliferate at a greater rate than wild-type astrocytes, we found that cyclin D1 protein expression was increased in Nf1−/− astrocytes (Fig. 1C). Consistent with our proteomic analysis, NPM and L7 proteins were also significantly elevated in Nf1−/− astrocytes compared with wild-type astrocytes (Fig. 1C). Because both NPM and L7 are nucleolar proteins involved in rRNA processing, assembly, and nuclear export, we sought to determine whether the expression of other nucleolar proteins involved in protein synthesis might be dysregulated in Nf1−/− astrocytes. Indeed, nucleolin, a nucleolar protein involved in ribosome biogenesis, was increased in Nf1−/− astrocytes (Fig. 1C). However, fibrillarin, a nucleolar protein responsible for early processing of 47S rRNA molecules, was unchanged in Nf1−/− astrocytes, suggesting that increases in nucleolar ribosome components was selective following loss of neurofibromin expression. Consistent with this notion, both NPM and nucleolin are positive regulators of cell growth, whereas other nucleolar ribosome components, such as fibrillarin and L11, either exhibit little influence on cell cycle progression or act as putative tumor suppressors (25–27). Notably, NPM is a potent oncogene and its continued expression is required for efficient ribosome export and cell cycle progression (28), making it an attractive downstream target for increased proliferation in Nf1−/− astrocytes. In addition, our identification of numerous other ribosome-promoting factors in Nf1−/− astrocytes suggests that neurofibromin plays a more direct role in suppressing improper ribosome biogenesis and subsequent cell growth.
Neurofibromin loss results in increased protein synthesis in astrocytes. Translational regulation and control of ribosome biogenesis are vital cellular processes that directly impact on cell proliferation and growth. Activated RAS and AKT controls protein translation (29, 30) and the combination of activated RAS and AKT has been shown to recruit specific growth-promoting mRNAs to polysomes in glial cells (31). Because we found that several proteins associated with protein translation were aberrantly expressed in Nf1−/− astrocytes, we sought to determine whether these protein expression changes were associated with an increase in the overall rate of protein synthesis. In these experiments, Nf1+/+ and Nf1−/− astrocytes were pulsed with [35S]methionine, and the amount of radioactivity incorporated into newly synthesized proteins was determined. At all time points examined, we observed significantly higher amounts of radioactivity incorporated in Nf1−/− astrocytes compared with wild-type astrocytes (Fig. 2A). At the 45-minute time point, there was an 8-fold increase in protein synthesis in Nf1−/− astrocytes relative to Nf1+/+ astrocytes. These results suggest that neurofibromin loss in astrocytes is associated with increased protein synthesis, likely reflecting increased protein translation.
Because neurofibromin functions as a negative regulator of RAS, we examined activation of a downstream RAS effector (ribosomal S6 protein) implicated in the regulation of mRNA translation (32, 33). Ribosomal S6 is activated by phosphorylation on specific residues (serine 235, 236, 240, and 244) primarily by its upstream kinase, ribosomal S6 kinase (33). To determine whether S6 is hyperactivated in Nf1−/− astrocytes, we measured the phosphorylation status of S6 in Nf1−/− and Nf1+/+ astrocytes. As shown in Fig. 2B, loss of neurofibromin in astrocytes resulted in a 6-fold increase in S6 phosphorylation on serine residues 240 and 244.
One of the downstream targets of activated RAS is PI3K (34). In this regard, loss of neurofibromin expression is associated with activation of the PI3K-AKT signaling pathway in human NF1-associated pilocytic astrocytoma and in Nf1−/− mouse myeloid cells (7, 35). Several studies have shown that activation of the PI3K-AKT axis leads to activation of mTOR, which leads to phosphorylation of S6 kinase and activation of ribosomal S6 (36, 37). Inhibition of S6 phosphorylation as a reflection of mTOR activity is achieved by blocking PI3K activation with the inhibitor LY294002, providing further evidence of PI3K-mediated mTOR activation (36, 38). To determine whether S6 hyperactivation in Nf1−/− astrocytes is downstream of PI3K and mTOR, we measured S6 phosphorylation in both Nf1 +/+ and Nf1−/− astrocytes in the presence of a PI3K inhibitor (LY294002) and the mTOR inhibitor, rapamycin. As shown in Fig. 2C, S6 phosphorylation was completely inhibited in wild-type and Nf1−/− astrocytes at 50 μmol/L LY294002 and at 5 μmol/L rapamycin, demonstrating the requirement for the PI3K/mTOR axis in promoting S6 activation in the absence of neurofibromin.
Regulation of mammalian target of rapamycin-dependent S6 activity in astrocytes is regulated by the neurofibromin GTPase activating protein-related domain. Previous studies have shown that neurofibromin growth regulation in multiple cell types requires residues within its RAS-GAP domain (10). To determine whether the hyperactivation of S6 in Nf1−/− astrocytes reflects dysregulated RAS activity resulting from loss of neurofibromin RAS-GAP function, we ectopically expressed the NF1 GAP-related domain (NF1GRD) in Nf1−/− astrocytes to restore neurofibromin GAP function. In Nf1−/− murine embryonic fibroblasts, expression of the NF1GRD was sufficient to restore normal growth (10). To examine the effect of NF1GRD expression on S6 hyperactivation in Nf1−/− astrocytes, we ectopically expressed the NF1GRD by MSCV-mediated retroviral transduction (Fig. 3A). Transduction of Nf1−/− astrocytes with MSCV-NF1GRD, but not with MSCV-Pac (vector control), completely abrogated the hyperphosphorylation of S6 seen in Nf1−/− astrocytes (Fig. 3B), demonstrating that inactivation of RAS by neurofibromin is a key event in suppressing S6 hyperactivation and astrocyte proliferation. Transduction of wild-type astrocytes with either MSCV-NF1GRD or vector control virus did not affect S6 phosphorylation (not shown).
Dysregulated K-RAS activity accounts for hyperactivation of ribosomal S6 in Nf1-deficient astrocytes. We have recently shown that although all four RAS isoforms are expressed in astrocytes, the only RAS isoform activated as a result of Nf1 loss in astrocytes is K-RAS (15). In this regard, activated K-RAS expression can substitute for Nf1 loss in astrocytes during optic nerve glioma formation in mice in vivo. To determine whether hyperactivated S6 in Nf1−/− astrocytes could also be mimicked by K-RAS activation in astrocytes, we examined the phosphorylation status of S6 in K-RASG12D transgenic astrocytes. In these experiments, we observed a 9.6-fold increase in phosphorylated S6 in K-RASG12D astrocytes compared with wild-type astrocytes (Fig. 4A).
In addition, we have shown that the growth advantage of Nf1−/− astrocytes could be reversed by expressing dominant inhibitory K-RAS (K-RASN17) in Nf1−/− astrocytes (15). To determine whether the S6 hyperactivation resulted from dysregulated K-RAS activity in Nf1−/− astrocytes, we ectopically expressed K-RASN17 in Nf1−/− astrocytes using MSCV-mediated retroviral transduction (Fig. 4B). As shown in Fig. 4C, dominant inhibitory K-RAS expression completely abrogated the increase in S6 phosphorylation observed in Nf1−/− astrocytes, whereas transduction of Nf1−/− astrocytes with MSCV-GFP (vector control), or Nf1+/+ astrocytes with either MSCV-GFP or MSCV-K-RASN17 (not shown) did not affect S6 phosphorylation. These observations are consistent with previous reports demonstrating that activation of K-RAS can induce PI3K-dependent activation of the mTOR-S6 kinase pathway and result in transformation of various cell types (39). Collectively, our data support a model in which neurofibromin loss results in K-RAS–dependent and PI3K-regulated hyperactivation of the mTOR/S6K pathway. These hyperproliferative signals are relayed to the ribosome machinery, stimulating increases in protein synthesis, and ultimately culminating in uncontrolled cell growth.
Genetically engineered Nf1 mouse optic pathway glioma and human neurofibromatosis 1–associated pilocytic astrocytoma exhibit increased ribosomal S6 activation in vivo. Because S6 was hyperactivated in Nf1−/− astrocytes in vitro, we next sought to determine whether astrocytic tumors arising in genetically engineered Nf1 murine glioma models exhibited S6 hyperactivation. For these experiments, we did immunohistochemistry on paraffin sections of tumors using phosphospecific S6 antibodies. Previous studies from our laboratory have shown that Nf1+/− mice either lacking Nf1 (40) or expressing activated K-RAS (15) in astrocytes develop low-grade optic nerve glioma. Analysis of optic nerve tumors from GFAPCre; Nf1flox/mut mice showed phosphorylated S6 expression in neoplastic astrocytes within the hypercellular optic chiasm (Fig. 5B). No phosphorylated S6 was observed in the optic chiasm of control mice (Fig. 5A). Similarly, in the Nf1+/−; K-RASGFAP mouse optic nerve glioma model, the neoplastic astrocytes with atypical nuclei in the optic chiasm also expressed phosphorylated S6 (Fig. 5D). No expression of phosphorylated S6 was observed in the optic chiasm of control mice (Fig. 5C).
Next, we sought to determine whether S6 hyperactivation was also observed in human NF1-associated pilocytic astrocytoma. Four of six NF1-associated pilocytic astrocytomas showed expression of phosphorylated S6 (Fig. 5E and F). In one of these NF1-associated WHO grade 1 astrocytomas, phosphorylated S6 expression was observed in foci of hypercellular tumoral regions containing atypical nuclei (Fig. 5F , inset). Collectively, these results show that neurofibromin loss leads to hyperactivation of the mTOR/S6 pathway in astrocytes in vitro and in NF1-associated mouse and human astrocytomas in vivo, underscoring the physiologic relevance of neurofibromin-regulated S6 hyperactivation in NF1-associated tumor formation.
Rapamycin inhibition of ribosomal S6 abrogates the hyperproliferation observed in Nf1−/− astrocytes. Previous studies from our laboratory have shown that Nf1−/− astrocytes exhibit a 2- to 3-fold increase in proliferation compared with wild-type astrocytes (18). To determine whether the growth advantage of Nf1-deficient astrocytes reflected increased mTOR signaling, we measured Nf1−/− astrocyte proliferation after rapamycin treatment. First, we defined the minimal dose of rapamycin that could reduce Nf1−/− astrocyte S6 phosphorylation to levels seen in Nf1+/+ astrocytes without causing significant cell death. As shown in Fig. 6 (inset), 1 μmol/L rapamycin significantly inhibited S6 phosphorylation in Nf1−/− astrocytes but had no effect on S6 phosphorylation in wild-type astrocytes. Using trypan blue exclusion, we observed >98% cell survival at this dose of rapamycin. Using this minimal dose (1 μmol/L) of rapamycin, we found that rapamycin significantly inhibited the proliferation of Nf1−/− astrocytes (55 ± 9.5% reduction), with minimal effects on wild-type astrocyte basal proliferation (16 ± 4%). We also observed similar reduction in proliferation of Nf1−/− astrocytes in presence of serum (Fig. 6). These results show that the increased proliferation observed in Nf1−/− astrocytes can be inhibited by blocking hyperactivation of the mTOR/S6 pathway without deleterious effects on normal cells. In this regard, Nf1−/− astrocytes seem to be dependent on continued mTOR signaling and, as such, are much more susceptible to rapamycin treatment. Similar instances of drug sensitivity have recently been observed in cells lacking PTEN or in tumors overexpressing specific activating epidermal growth factor receptor (EGFR) mutants (41, 42). In these cases, selection of cells expressing EGFR mutants or cells lacking PTEN conferred dependence on these hyperactivated pathways for continued proliferation, such that inhibition of these pathways with selective inhibitors resulted in dramatically attenuated cell proliferation. In astrocytes lacking neurofibromin, there is hyperactivation of the mTOR/S6 pathway, which may serve as the predominant signal on which Nf1-deficient astrocytes depend for continued proliferation.
Rapamycin is an immunosuppressant drug that specifically inhibits mTOR activity and cellular hyperproliferation in many cell types resulting in G1 growth arrest. The rapamycin analogue CCI-779 inhibited hyperproliferation of PTEN-deficient prostate cancer cells (41) and is currently in phase III clinical trials (43). Based on our results, we suggest that rapamycin-based drugs might be valuable in the treatment of NF1-associated tumors. Preclinical mouse studies with rapamycin analogues for optic nerve glioma in genetically engineered mice are currently under way in our laboratory.
Grant support: Department of Defense grant DAMD17-03-1-0215 (D.H. Gutmann) and the Pew Scholars in Biomedicine (J.D. Weber).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Chrissie Kamp for contributions during the execution of these experiments.