Children with the tumor predisposition syndrome, neurofibromatosis 1 (NF1), develop optic pathway gliomas. The NF1 gene product, neurofibromin, functions as a negative regulator of RAS, such that NF1 inactivation results in RAS hyperactivation. Recent studies have highlighted the divergent biological and biochemical properties of the various RAS isoforms, which prompted us to examine the consequence of Nf1 inactivation in astrocytes on RAS isoform activation in vitro and in vivo. In this report, we show that only K-RAS is activated in Nf1−/− astrocytes and that activation of K-RAS, but not H-RAS, accounts for the proliferative advantage and abnormal actin cytoskeleton–mediated processes observed in Nf1−/− astrocytes in vitro. Moreover, dominant inhibitory K-RAS corrects these abnormalities in Nf1−/− astrocytes invitro. Lastly, we show that Nf1+/− mice with astrocyte-specific activated K-RAS expression in vivo develop optic pathway gliomas, similar to our previously reported Nf1+/− mice with astrocyte Nf1 inactivation. Collectively, our results show that K-RAS is the primary target for neurofibromin GTPase-activating protein activity in vitro and in vivo and that K-RAS activation in astrocytes recapitulates the biochemical, biological, and tumorigenic properties of neurofibromin loss.
Neurofibromatosis 1 (NF1) is a common autosomal-dominant inherited cancer syndrome in which affected individuals develop both benign and malignant tumors (1). The two most frequently observed tumors in NF1 arise from glia-like cell types that reside in the peripheral and central nervous systems. Most commonly, patients with NF1 develop Schwann cell tumors (neurofibroma) in peripheral nerves and pilocytic astrocytomas (WHO grade I) in the optic pathway [optic pathway glioma (OPG)]. Neurofibromas typically begin to appear around the time of puberty and exhibit a limited period of growth (2). In contrast, OPGs are almost exclusively found in young children and affect the optic nerves, chiasm, and hypothalamus (3).
The identification of the NF1 gene in 1990 provided the first clues to the molecular pathogenesis of these tumors in NF1. The NF1 gene product, neurofibromin, was found to contain a small region with structural and functional similarity to proteins that negatively regulate p21RAS, termed GTPase-activating proteins (GAP; ref. 4). GAP molecules accelerate the intrinsic hydrolysis of RAS-GTP to RAS-GDP, converting RAS from its active GTP-bound conformation to an inactive GDP-bound form. In this regard, the rate of RAS-GTP hydrolysis is accelerated 40-fold in the presence of neurofibromin or its GAP domain (5–7). Consistent with the role of neurofibromin as a RAS-GAP, loss of neurofibromin in Schwann cells (8–10), astrocytes (11), myeloid cells (12, 13), and fibroblasts (14) is associated with high levels of activated RAS and increased mitogen-activated protein kinase/Akt signaling in both mouse and human cells. Moreover, the introduction of the GAP domain of neurofibromin into myeloid cells (14) and malignant Schwann cells (9) results in normalization of RAS hyperactivation to wild-type levels and is associated with reduced cell growth.
In many cell types, activation of RAS is associated with increased proliferation (15, 16) and/or decreased cell death (17), suggesting that neurofibromin loss may result in tumor formation as a consequence of increased RAS-mediated mitogenesis and transformation. Therefore, an effective anticancer approach for NF1-related tumors would involve RAS inhibition. In order for RAS to signal to its downstream effectors, it needs to be localized to the plasma membrane, a process that is mediated by post-translational lipid modification. RAS proteins contain a canonical CAAX motif at their extreme carboxyl termini, which serves as a target for isoprenylation by lipid transferases (18). Blocking RAS isoprenylation with farnesyltransferase inhibitors (FTI) has been shown to have efficacy in reducing RAS activity and ameliorating the mitogenic effects of oncogenic RAS proteins (19, 20). These observations provided biological evidence for FTI clinical trials for NF1-associated plexiform neurofibroma. Unfortunately, these initial studies did not show a robust tumor response to FTI therapy (21), which may reflect the limited efficacy of these compounds asRAS inhibitors in vivo (22).
In the brain, there are three major RAS isoforms (K-RAS, N-RAS, and H-RAS) with different biological properties and sensitivities to isoprenylation inhibitors. Initial investigations of neurofibromin GAP activity showed that the NF1-GAP-related domain could inactivate both H-RAS and N-RAS in vitro (5–7), suggesting that neurofibromin was equally effective as a RAS-GAP for all RAS isoforms. However, elegant studies by Shannon et al. showed that FTIs completely blocked H-RAS activation but had no effect on Nf1−/− hematopoietic cells and did not ameliorate the Nf1−/− myeloproliferative disease in mice (20), suggesting that H-RAS might not be the physiologic target of neurofibromin RAS-GAP activity in hematopoietic cells. Further support for the possibility that activation of the different RAS isoforms has distinct physiologic effects is derived from experiments in which oncogenic H-RAS was expressed in astrocytes of transgenic mice. These mice developed multifocal regions of high-grade glioma (23), in striking contrast to astrocyte-specific Nf1 conditional knockout mice (Nf1GFAPCKO mice), which did not develop gliomas (24). These results prompted us to determine whether specific RAS isoforms are preferentially hyperactivated as a consequence of neurofibromin loss in astrocytes and to evaluate the functional and biochemical properties of H-RAS versus K-RAS activation in astrocytes relative to neurofibromin loss. In this report, we show that K-RAS is the preferential target of neurofibromin RAS-GAP activity in astrocytes and that hyperactivation of K-RAS in astrocytes mimics Nf1 loss in vitro and in vivo.
Materials and Methods
Mice. GFAP-Cre-IRES-LacZ transgenic (GFAP-Cre) mice and GFAP- H-RASG12V-IRES-LacZ (B8) mice were generated as described previously (23, 24). Lox-stop-lox (LSL)-K-RASG12D mice, generously provided by Dr. Tyler Jacks(Massachusetts Institute of Technology, Cambridge, MA; ref. 25) were crossed with GFAP-Cre mice (24) to obtain (LSL)-K-RASG12D; GFAP-Cre mice. Nf1flox/flox mice were a kind gift from Dr. Louis Parada (University of Texas-Southwestern Medical Center, Dallas, TX). All mice were genotyped by PCR amplification of tail genomic DNA. β-Galactosidase activity was confirmed in primary astrocyte cultures. Mice were used in accordance with established Animal Studies Protocols at the Washington University School of Medicine.
Dominant-Negative K-RAS. Dominant-negative K-RAS (dnK-RASN17) was generated as described previously (26, 27) by site-directed mutagenesis using the following primers: 5′-GCTGGTGGCGTAGGCAAGAATGCCTTGACGATACAG-3′ and 5′-CTGTATCGTCAAGGCATTCTTGCCTACGCCACCAGC-3′. The T7 epitope-tagged product was subcloned into the plasmid p5RαIRESGFP (MSCV retroviral vector) and verified by sequencing. Transduction of primary astrocytes was done using MSCV-dnK-RAS-IRES-GFP and MSCV-IRES-GFP (control) as described (28).
Primary Astrocyte Cultures. Murine neocortical astroglial cultures, containing >95% GFAP-positive cells (astrocytes), were generated from postnatal day 2 Nf1flox/flox, (LSL)-K-RASG12D; GFAP-Cre, and H-RAS transgenic pups as described previously (29). To eliminate neurofibromin expression, Nf1flox/flox astrocytes were treated with Ad5Cre (University of Iowa Gene Transfer Vector Core, Iowa City, IA) or Ad5-lacZ (control; ref. 24). Loss of neurofibromin was confirmed by Western blot. Astrocyte proliferation assays were done by [3H]thymidine incorporation (24).
Measurement of Intracellular Cyclic AMP. Intracellular cyclic AMP (cAMP) levels were measured in astrocytes using the cAMP enzyme immunoassay kit (Cayman Chemical, Ann Arbor, MI) according to the manufacturer's instructions. Serum-deprived astrocytes were stimulated with medium alone (control) or pituitary adenylyl cyclase–activating polypeptide (PACAP; 10−9 mol/L) and intracellular cAMP was measured (30).
RAS/Rac1 Binding Assay and Western Analysis. Active RAS (RAS-GTP) was detected by Raf1 binding assay using the RAS activation assay kit (Upstate, Lake Placid, NY). Active Rac1 (Rac1-GTP) was detected by Pac-1 binding assay using Rac activation assay kit (Upstate). Western blot analysis was done on serum-deprived astrocytes stimulated with epidermal growth factor (EGF) or 5% serum for 10 minutes (30). Active RAS isoforms were detected using RAS isoform-specific antibodies (Santa Cruz Biotechnology, Santa Cruz, CA). Phosphorylated cofilin (Ser3) and total cofilin antibodies were purchased from Cell Signaling Technology (Beverly, MA). Anti-T7 antibody was purchased from Novagen (Madison, WI).
Cell Adhesion, Spreading, and Motility Assays. Cell adhesion assays were done on fibronectin-coated plates and quantitated by spectrophotometric absorbance at 540 nm (29). Cell spreading was analyzed with BODIPY-conjugated phalloidin (Molecular Probes, Eugene, OR) on astrocytes seeded onto laminin-coated plates and examined on a Zeiss Axiophot microscope (29). Cell motility was determined in Transwell Boyden chambers containing Matrigel-coated (Collaborative Research, Bedford, MA) 8-μm membranes (29).
Morphologic and Immunohistologic Analysis. Mice were perfused transcardially with 4% paraformaldehyde. Optic nerves and brains were dissected, fixed, and imaged with a digital camera (Optronics, Muskogeeg, OK) attached to a dissection microscope (Nikon, Japan). Specimens were processed for paraffin embedding and sectioning. For histologic analysis, sections were stained with H&E. Nuclei were counted in the prechiasmatic area of H&E-stained optic nerve sections. Immunohistochemistry was done on adjacent paraffin sections with rat anti-GFAP (1:100, Zymed, San Francisco, CA) and rabbit anti-Ki-67 (1:1,000, Novocastra, Newcastle upon Tyne, United Kingdom) antibodies as described (24). Detection was done using a Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA) for GFAP expression and a VIP substrate kit (Vector Laboratories) for Ki-67 immunostaining. All sections were photographed with a digital camera (Optronics) attached to an inverted microscope (Nikon). GFAP-immunoreactive astrocytes were counted in the CA1 region of the hippocampus of brain sections from 3-month-old K-RASGFAP, H-RASGFAP, and control mice and from 12-month-old K-RASGFAP, Nf1+/−; K-RASGFAP, and control mice.
Nf1 Loss in Astrocytes Results in Preferential Activation of K-RAS. Previous studies have shown that the Nf1 gene product, neurofibromin, functions as a RAS GTPase in vitro and in vivo (6, 8, 13). Because the three major isoforms of RAS serve distinct downstream effector functions, we sought to determine whether neurofibromin selectively regulates RAS in an isoform-specific manner in astrocytes. In these experiments, we inactivated Nf1 inNf1flox/flox astrocytes in vitro by Ad5Cre infection, and neurofibromin loss was verified by Western blotting (Fig. 1A). We next assayed active RAS in astrocyte lysates using the Raf1-RBD affinity chromatography as reported previously (31). Whereas both H-RAS and N-RAS were abundantly expressed in astrocytes, the only RAS isoform hyperactivated as a consequence of neurofibromin loss was K-RAS (Fig. 1B). To confirm the specificity of the RAS isoform-specific antibodies, we transfected NIH-3T3 cells with T7-tagged H-RASG12V, N-RASG12V, and K-RASG12V expression plasmids. Active RAS from all three transfected cell lines was detected after Raf1-RBD affinity chromatography by Western blot using the T7 antibody as well as with RAS isoform-specific antibodies. In each case, the isoform-specific antibodies only recognized the RAS isoform against which they were directed (data not shown).
In light of our finding that H-RAS was not hyperactivated in Nf1−/− astrocytes in vitro, we sought to determine whether K-RAS activation in astrocytes might more closely mimic neurofibromin loss. To address this question, we generated primary astrocytes cultures from Nf1flox/flox mice and deleted Nf1 by Ad5Cre in vitro (Nf1−/− astrocytes), LSL-K-RASG12D; GFAP-Cre mice in which expression of oncogenic K-RAS is induced on the expression of Cre recombinase (K-RAS astrocytes), and B8 transgenic mice in which oncogenic H-RASG12V is expressed from the astrocyte-specific GFAP promoter (H-RAS astrocytes). To ensure that the levels of active RAS were not significantly different in these three types of astrocytes, we did Raf1-RBD affinity chromatography assays from astrocytes grown in serum-free medium for 48 hours. We did not observe any significant differences in the levels of activeRAS among Nf1−/−, K-RASG12D, and H-RASG12V astrocytes (Fig. 1C). No active RAS was observed in wild-type astrocytes grown under identical conditions.
EGF-stimulated Proliferation Is Similar in Nf1−/− and K-RASG12D Astrocytes, but the PACAP-Stimulated cAMP Response Discriminates Nf1−/− Astrocytes from K-RASG12D Astrocytes. After confirmingthat nearly equivalent levels of active RAS were detected in Nf1−/−, K-RASG12D, and H-RASG12V astrocytes, we next sought to determine whether the proliferative advantage conferred by loss of neurofibromin wassimilar to that seen with astrocyte-specific expression of oncogenic K-RASG12D and H-RASG12V. Earlier studies from our laboratory showed that loss of Nf1 in astrocytes results in a 2-fold increase in astrocyte proliferation invivo and in vitro (24). Serum-deprived Nf1−/− and K-RASG12D astrocytes exhibited a similar 1.5- to 2.2-fold increase in proliferation compared with their respective wild-type astrocyte littermate controls. In contrast, H-RASG12V astrocytes exhibited a 4.84-fold increase in proliferation relative to wild-type control astrocytes. Similar results were obtained using EGF. In these experiments, we observed a 3.6- and 4.48-fold increase in proliferation in Nf1−/− and K-RASG12D astrocytes, respectively, in response to 1 ng/mL EGF. In contrast, the H-RASG12V astrocytes exhibited a significantly higher proliferation rate (9.96-fold increase relative to wild-type controls; Fig. 1D).
We have shown previously that Nf1−/−, but not H-RASG12V, astrocytes exhibit reduced cAMP generation in response to the neuropeptide PACAP (30). Because K-RAS is preferentially activated in Nf1−/− astrocytes, we wished to determine whether K-RAS activation in astrocytes recapitulated the reduced cAMP response observed in Nf1−/− astrocytes. In these experiments, we found that wild-type astrocytes of all genotypes showed robust increases in cAMP generation in response to PACAP. In contrast to the Nf1−/− astrocytes, which exhibited a 74% reduction in cAMP production, both K-RASG12D and H-RASG12V astrocytes responded similarly as wild-type astrocytes to PACAP treatment (Fig. 1E). Collectively, these results suggest that K-RAS activation is more similar to neurofibromin loss than H-RAS activation with respect to astrocyte proliferation in vitro but that the ability of neurofibromin to regulate intracellular cAMP is independent of K-RAS activation.
Nf1−/− and K-RASG12D Astrocytes Share Similar Cytoskeleton-Associated Properties, Which Distinguish Them from H-RASG12V Astrocytes. We have shown previously that both heterozygosity andbiallelic inactivation of Nf1 results in abnormal cytoskeleton-associated processes in astrocytes, including increased cell motility, decreased cellattachment, and delayed cell spreading (29). We next wished to determine whether the cytoskeleton-associated defects seen in neurofibromin-deficient astrocytes could be recapitulated by K-RAS activation. In these experiments, we studied the effect of neurofibromin loss and K-RAS or H-RAS activation on astrocyte attachment, motility, and actin cytoskeleton organization during the initial phases of cell spreading.
We analyzed actin cytoskeleton organization during the initial phases of cell spreading in Nf1−/−, K-RASG12D, H-RASG12V, and their respective wild-type astrocyte controls. Phalloidin-BODIPY immunocytochemistry revealed that the majority of the Nf1−/− and K-RASG12D astrocytes failed to spread during the first hour after plating. In these experiments, most of the Nf1−/− and K-RASG12D astrocytes were rounded and displayed dense actin staining compared with distinct F-actin cytoskeleton staining in wild-type or H-RASG12V astrocytes (Fig. 2A).
Second, Nf1−/− astrocytes exhibited a 37% and 31% reduction in attachment at 1 and 4 hours, respectively, after plating onto fibronectin-coated plates compared with Nf1+/+ (wild-type) astrocytes (Fig. 2B). Similar to Nf1−/− astrocytes, K-RASG12D astrocytes exhibited reduced attachment at both 1 hour (49.89% reduction) and 4 hours (46.9% reduction) after plating. In contrast, H-RASG12V astrocytes were indistinguishable from wild-type astrocytes at both 1- and 4-hour time points.
Lastly, we studied the migratory properties of Nf1−/−, K-RASG12D, and H-RASG12V astrocytes using a Boyden chamber assay. Nf1−/− astrocytes showed a 1.91-fold increase in migration compared with Nf1+/+ astrocytes (Fig. 2C). Similar to Nf1−/− astrocytes, K-RASG12D astrocytes exhibited a 2.12-fold increase in motility compared with matched wild-type control littermate astrocyte cultures (P < 0.01). In contrast, H-RASG12V astrocytes showed only a 1.19-fold increase in motility compared with wild-type controls. Collectively, these results show that the abnormal cytoskeleton-associated defects resulting from neurofibromin loss in astrocytes can be mimicked by K-RAS, but not H-RAS, activation.
Loss of Nf1 or K-RASG12D Expression Results in Activation of the Rac1 Pathway in Astrocytes. One of the downstream targets of RAS is the Rac1 pathway, important for regulating actin cytoskeleton dynamics (32). Because we observed defective cytoskeleton-mediated processes in Nf1−/− and K-RASG12D astrocytes, we sought to determine whether these abnormalities were reflected by abnormal Rac1 pathway activation. It has been shown previously in fibroblasts that K-RAS activates Rac1 more efficiently than H-RAS (26). In these experiments, Nf1−/− and K-RASG12D astrocytes exhibited equivalent amounts of active Rac1 in the absence of mitogenic stimulation (Fig. 3A and B). In contrast, no Rac1-GTP was detected in H-RASG12V or wild-type control astrocytes under these conditions (Fig. 3C). Astrocytes stimulated with either EGF or serum showed increased levels of Rac1-GTP relative to untreated astrocytes. No significant differences in active Rac1 levels were observed between EGF- and serum-stimulated Nf1−/− and K-RASG12D astrocytes. Controls for equal protein loading were determined by Western blotting (Fig. 3A–C, bottom panels).
To further explore the link between Rac1 activation and actin cytoskeleton–associated defects observed in Nf1−/− andK-RASG12D astrocytes, we examined phosphorylated cofilin levels by Western blot analysis. Activation of Rac1 results in activation and autophosphorylation of its downstream effector kinase, LIM kinase (33). Cofilin, a substrate of LIM kinase, is a potent regulator of actin filament dynamics (34), which binds and depolymerizes actin. Phosphorylation at Ser3 by LIM kinase inactivates cofilin, leads to the accumulation of actin filaments (35), and defines the direction of cell motility. We measured phosphorylated cofilin levels in astrocytes under serum-deprived conditions and after treatment with EGF. Although no phosphorylated cofilin was detected in Nf1+/+ or wild-type control astrocytes, both Nf1−/− and K-RASG12D astrocytes showed phosphorylated cofilin at baseline (Fig. 3D and E). In keeping with the absence of active Rac1 in H-RASG12V astrocytes under serum starvation conditions, we did not observe any phosphorylated cofilin in H-RASG12V astrocytes (Fig. 3F). EGF treatment resulted in increased phosphorylation of cofilin, which was similar among Nf1−/−, K-RASG12D, and H-RASG12V astrocytes. As before, the blots were stripped and reprobed with nonphosphorylated cofilin antibody to ensure equal loading. Taken together, these results provide additional evidence for a specific effect of neurofibromin loss on K-RAS-mediated signaling pathways relevant to actin cytoskeleton dynamics.
The Abnormal Biological and Biochemical Properties of Nf1-Null Astrocytes Are Reversed by Ectopic Expression of dnK-RASN17. Based on the observations described above, we next wished to determine whether the biological and biochemical properties of Nf1−/− astrocytes specifically reflected K-RAS hyperactivation. To address this, we ectopically expressed dnK-RAS in Nf1−/− astrocytes by retroviral transduction. N17-RAS mutants function as GDP-bound dominant-negative (inhibitory) moieties capable of blocking RAS-mediated signaling (36). The K-RASN17 mutant has been shown previously to inhibit only activated K-RAS, with minimal effects on H-RAS (37). To ensure that the dnK-RAS mutant used in these experiments specifically inhibited activated K-RAS, but not H-RAS, in astrocytes, we also transduced H-RASG12V astrocytes with dnK-RAS. In these experiments, Nf1−/− astrocytes expressing dnK-RAS (Fig. 4A; inset) showed significantly reduced proliferation (P < 0.001) compared with untreated or MSCV-GFP (vector control)–treated Nf1−/− astrocytes (Fig. 4A). H-RASG12V astrocytes continued to proliferate at a much higher rate than Nf1+/+ or Nf1−/− astrocytes regardless of retroviral transduction with MSCV-dnK-RAS or MSCV-GFP.
To determine whether the impaired attachment of Nf1−/− astrocytes could be reverted to normal by dnK-RAS inhibition, Nf1−/− astrocytes expressing dnK-RAS were analyzed. In these experiments, Nf1−/− astrocytes expressing dnK-RAS showed significant increases in cell attachment (P < 0.01) at 1 hour after initial plating compared with untransduced or MSCV-GFP transduced Nf1−/− astrocytes (Fig. 4B). Retroviral transduction did not affect attachment of Nf1+/+ or H-RASG12V astrocytes.
Finally, we analyzed the RAS downstream signaling abnormalities in Nf1−/− astrocytes. Expression of dnK-RAS completely inhibited the aberrant expression of activated Rac1 and phosphorylated cofilin (Fig. 4C and D) at baseline in Nf1−/− astrocytes. No activated Rac1 or phosphorylated cofilin was detected in untransduced or transduced Nf1+/+ or H-RASG12V astrocytes. Taken together, these results suggest that the proliferative advantage, abnormal cell attachment, and hyperactivation of the Rac1 pathway are most likely due to the increased K-RAS activity in Nf1−/− astrocytes.
Astrocyte-Specific Activation of Oncogenic K-RAS Mimics Astrocyte-Specific Inactivation of Nf1 In vivo. We have shown previously that astrocyte-specific inactivation of Nf1 (Nf1flox/flox; GFAP-Cremice) resulted in increased astrocyte number in the brain but was insufficient for astrocytoma formation (24). Because some of the important properties of Nf1−/− astrocytes were mimicked by K-RAS astrocytes in vitro, we asked the question whether astrocyte-specific activation of K-RAS recapitulates the Nf1-null astrocyte phenotype in vivo. To address this question, we employed two complementary approaches.
First, we crossed the LSL-K-RASG12D mice with GFAP-Cre mice to specifically activate K-RAS in astrocytes. Histologic analysis of the brains from K-RAS (K-RAS Cre+; designated K-RASGFAP) mice at 3 months (Supplementary Fig. 1B) and 12 months (Supplementary Fig. 1H) did not show any evidence of tumor formation. However, immunohistochemistry with an antibody to GFAP (Supplementary Fig. 1D, E, J, and K) revealed that the K-RASGFAP mouse brains exhibited a 1.6-fold increase in astrocyte number compared with control mice (Supplementary Fig. 1M). No significant change in astrocyte number was observed between 3- and 12-month-old K-RASGFAP mice. In contrast, as reported previously (23), astrocyte-specific activation of H-RAS (H-RASGFAP mice) resulted in high-grade astrocytoma formation in the brain by 3 months of age. H&E staining (Supplementary Fig. 1C) and immunohistochemistry with an anti-GFAP antibody (Supplementary Fig. 1F) showed foci of hypercellularity in the cortex, which corresponded to nests of GFAP-immunoreactive cells. In addition, H-RASGFAP mice showed 2.5-fold increases in astrocytes in the brain (Supplementary Fig. 1M).
We have shown recently that astrocyte-specific biallelic inactivation of Nf1 in the context of Nf1 somatic heterozygosity (GFAP-Cre;Nf1flox/mut mice) is sufficient for OPG formation in vivo (38). If activated K-RAS mimics Nf1 loss in astrocytes, we hypothesized that astrocyte-specific activation of K-RAS in the Nf1+/− mice would also result in OPG. As a second approach to model Nf1 inactivation in astrocytes in vivo, we mated K-RASGFAP mice with Nf1+/− mice to generate Nf1+/−; K-RASGFAP mice. These mice are viable and did not show any obvious neurologic complications at 12 months of age. Immunohistologic analysis of the brains from these mice at 12 months of age showed a 1.8-fold increase in astrocyte numbers (Supplementary Fig. 1M), but there was no evidence of tumor formation (Supplementary Fig. 1I and L). This result is similar to what we reported previously for the GFAP-Cre;Nf1flox/mut mice (38).
However, six of six Nf1+/−; K-RASGFAP mice analyzed developed OPG. As shown in Fig. 5C, the chiasm and prechiasmatic optic nerves of these mice were distinctly enlarged compared with K-RASGFAP (Nf1+/+; K-RASGFAP; Fig. 5B) or control (Nf1+/+; K-RAS Cre−; Fig. 5B) mice. Pathologic analysis of optic nerves showed increased cellularity with grossly enlarged prechiasmatic areas (Fig. 5F), optic nerves (Fig. 6C), and chiasms (Fig. 6D) in Nf1+/−; K-RASGFAP mice compared with prechiasmatic areas and optic nerves from control (Figs. 5D and 6A) or K-RASGFAP (Figs. 5E and 6B) mice. The representative prechiasmatic region shown in Fig. 5F contains a nodular mass with clusters of enlarged atypical nuclei (Fig. 6I) and robust GFAP immunoreactivity (Fig. 5L), highlighting thin, long cellular processes (inset). Tumors in the optic nerves (Fig. 6G) and chiasm (Fig. 6H) contained scattered individual cells as well as clusters of large cells with distinct cellular atypia and nuclear pleiomorphism in contrast to similar areas of control (Fig. 6E) and K-RASGFAP (Fig. 6F) mice. A focal cluster of neoplastic cells in the chiasm of one Nf1+/−; K-RASGFAP mouse is also shown (Fig. 6D and H). A fibrillary pattern of intense GFAP staining was observed wherever clusters of atypical cells were encountered (Fig. 6K and L). We stained the sections with an antibody to Ki-67 (MIB-1) to identify proliferating cells. Ki-67+ cells were detected within the hypercellular prechiasmatic nodular masses (Fig. 5O) and throughout the optic nerves (Fig. 6O) and chiasms (Fig. 6P,, arrows) of Nf1+/−; K-RASGFAP mice. No Ki-67+ nuclei were found in the prechiasmatic areas or optic nerves of control (Figs. 5M and 6M) or K-RASGFAP (Figs. 5N and 6N) mice. Collectively, these results show that K-RAS activation can substitute for Nf1 loss in astrocytes in the genesis of OPG in Nf1+/− mice.
The ability of neurofibromin to negatively regulate RAS proteins has been shown clearly by several studies over the past decade and formed the rational basis for the use of anti-RAS chemotherapy in NF1-associated plexiform neurofibroma (21). In these studies, the GAP-related domain of neurofibromin accelerated GTP hydrolysis on mammalian H-RAS (5, 7, 39) and N-RAS (6) as well as substituted for yeast RAS-GAP proteins (6). Based on these data, inhibitors of RAS activity were shown to inhibit the growth of Nf1−/− rodent and NF1-associated tumor cell lines in vitro (40–42). Although neurofibromin can inactivate all RAS isoforms in vitro, the consequence of NF1 loss on RAS isoform activation has not been formally addressed. In this regard, it is highly likely that specific RAS isoforms are hyperactivated in the absence of neurofibromin and that this selective RAS isoform dysregulation has unique biological effects.
It is now appreciated that activation of the various RAS isoforms results in different biological consequences. In mice, K-RAS is essential for normal embryonic development, whereas both H-RAS and N-RAS do not play significant roles (43, 44). In cultured cells, expression of activated H-RAS, but not N-RAS or K-RAS, results in neoplastic transformation (45), and activated K-RAS, but not N-RAS or H-RAS, overexpression leads to increased cell motility, microtubule binding, and membrane ruffling (26, 27). In addition, the mechanisms of activation of the RAS isoforms differ due to their differential response to growth factors and sensitivity to RAS exchange factors (26, 46–48) Lastly, direct visualization of activated RAS isoforms in intact two-dimensional sheets of apical plasma membrane showed that the various RAS isoforms localize to distinct microdomains of the plasma membrane (49). Based on these observations, understanding which RAS isoforms are hyperactivated in neurofibromin-deficient cells would provide significant insight into the cellular and biochemical derangements that result from NF1 loss.
In this report, we show that all RAS isoforms are expressed in astrocytes; however, in the absence of neurofibromin, only K-RAS seems to be hyperactivated. Moreover, K-RAS (and not H-RAS) activation resulted in similar biochemical and biological effects as Nf1 loss in astrocytes, although equivalent levels of RAS-GTP levelswere observed in these different populations of astrocytes (Nf1−/−, K-RAS, and H-RAS). Although these data provide strong evidence for K-RAS as the primary target of neurofibromin RAS-GAP activity and growth control in astrocytes, these results do not exclude hyperactivation of other RAS isoforms in the pathogenesis of other NF1-associated tumors. It is possible that different cell types (e.g., neurons, myeloid cells, and Schwann cells) will exhibit hyperactivation of other RAS isoforms critical for their growth andtransformation. Similarly, it is possible that specific cellular conditions may trigger the hyperactivation of H-RAS or N-RAS in Nf1−/− astrocytes. Lastly, it has been suggested that expressing oncogenic RAS as a transgene may have different biological consequences than expressing oncogenic RAS from its endogenous promoter (50). Although the use of the endogenous promoter most accurately reflects the normal physiologic context of RAS expression, the observations detailed in this report using a variety of approaches strongly support the hypothesis that the biological effects of H-RAS and K-RAS activation in astrocytes invitro and invivo are distinct and that neurofibromin loss results in a selective hyperactivation of K-RAS.
Previous studies from our laboratory have shown that biallelic loss of Nf1 in astrocytes results in a 1.8- to 2.2-fold increase in proliferation in vitro and in vivo (24). Despite the growth advantage, Nf1−/− astrocytes do not transform with prolonged passage in vitro but rather undergo senescence, and astrocytomas do not develop in vivo. Similarly, astrocytes expressing oncogenic K-RAS exhibited a 1.6- to 2.0-fold growth advantage in vitro and didnot form astrocytomas in vivo. In sharp contrast, astrocytes expressing oncogenic H-RAS proliferated at a significantly higher rate than both Nf1−/− and K-RAS astrocytes at baseline, were extremely sensitive to mitogenic doses of EGF in vitro, and mice expressing oncogenic H-RAS in astrocytes developed high-grade astrocytoma.
In addition to modest increases in astrocyte proliferation, we previously observed delayed attachment and spreading, as well as increased motility in Nf1−/− astrocytes (29). Similar to Nf1−/− astrocytes, astrocytes expressing oncogenic K-RAS showed a significant delay in the initial phases of cell spreading and showed reduced cell attachment. These abnormalities were not observed in astrocytes expressing oncogenic H-RAS. In addition, both Nf1−/− and K-RAS astrocytes exhibited increased motility compared with H-RAS astrocytes. Moreover, the biochemical signaling abnormalities associated with cytoskeleton maintenance (Rac1 hyperactivation and cofilin phosphorylation) seen in Nf1−/− astrocytes were also observed in K-RASG12D, but not H-RASG12V, astrocytes. Rac1 has been implicated in the transduction of signals that modulate actin polymerization during lamellipodia and filopodia formation (51) by activating its downstream effector serine/threonine kinase Pak1 and resulting in cofilin phosphorylation (33–35). It is likely that the preferential activation of K-RAS in neurofibromin-deficient astrocytes results in increased signaling through the Rac1 pathway, which leads to changes in cytoskeleton-associated properties important in NF1-associated tumorigenesis.
To provide additional evidence for K-RAS as the target of neurofibromin GAP activity in vitro, we ectopically expressed dnK-RASN17 in Nf1−/− astrocytes and analyzed cell proliferation and cytoskeleton-associated abnormalities. In these experiments, we observed a significant reduction in proliferation in Nf1−/−, but not H-RASG12V, astrocytes expressing K-RASN17. Similarly, dnK-RAS expression in Nf1−/− astrocytes also reversed the attachment defect, the increase in Rac1 activity and cofilin phosphorylation. We chose to use this genetic approach based on a previously published report in which dnK-RAS selectively inhibited K-RAS, with minimal or no effects on H-RAS or N-RAS (37). We did not use dominant-negative H-RAS (H-RASN17) in Nf1−/− astrocytes, because H-RASN17 is known to effectively inhibit all isoforms of RAS.
Previous studies by our laboratory have shown that somatic heterozygosity of Nf1 is necessary for OPG formation in astrocyte-specific Nf1 conditional knockout mice (GFAPCre; Nf1flox/mut mice). In all GFAPCre; Nf1flox/mut mice, gliomas formed in the prechiasmatic optic nerves and chiasm (38). Based on these findings, we reasoned that Nf1+/−; K-RASGFAP mice would also develop OPG if activation of K-RAS is the biological equivalent of Nf1 loss in astrocytes. In this study, we report that Nf1+/−; K-RASGFAP mice also developed OPG composed of atypical GFAP-immunoreactive proliferating cells with nuclear atypia, similar to the gliomas observed in GFAPCre; Nf1flox/mut mice. These results support the model in which activated K-RAS is the functional equivalent of Nf1 loss in astrocytes. Because K-RAS does not modulate cAMP activation, it is unlikely that the abnormal cAMP regulation resulting from neurofibromin loss in astrocytes contributes to tumorigenesis. It is possible, however, that cAMP regulation is important in the Nf1+/− brain cells that facilitate Nf1−/− transformation in vivo. Experiments are under way to address this possibility.
Our finding that there is preferential activation of K-RAS as a result of neurofibromin loss has important pharmacologic and therapeutic implications. Although all RAS isoforms are farnesylated by farnesyltransferases in vivo, K-RAS is also geranylgeranylated by geranylgeranyltransferases (52) and, following FTI treatment, retains full biological and transforming activity (53, 54). Consistent with our findings, the myeloproliferative disorder in mice caused by Nf1−/− hematopoietic cells was refractory to FTI treatment and could be mimicked in vivo by conditional K-RAS activation, although the Nf1-related myeloproliferative disorder was characterized by similar but milder phenotypes (55–57). As we enter into the age of molecularly targeted therapies, it will be important to consider the details of neurofibromin RAS regulation. The observation that the RAS isoform preferentially activated as aconsequence of Nf1 loss is K-RAS suggests that the use of geranylgeranyltransferase inhibitors alone or in combination with FTIs, but not FTIs alone, may be the most appropriate choice in thetreatment of NF1-associated tumors. In light of the toxicity associated with GGTI therapy in vivo (58), future therapeutic development could also exploit unique downstream effectors of K-RAS as biologically based targets for NF1 tumor therapy.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org).
Grant support: NIH grant NS36996 and U.S. Department of Defense grant DAMD-17-03-1-0215 (D.H. Gutmann).
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 expert technical assistance, Dr. M. Livia Bajenaru forhelpful suggestions, Dr. Luis Parada for the generous gift of Nf1flox/flox mice, and Dr.Tyler Jacks for the generous gift of LSL-K-RAS mice.