|Home | About | Journals | Submit | Contact Us | Français|
Loss-of-function mutations in the NF1 tumor suppressor result in deregulated Ras signaling and drive tumorigenesis in the familial cancer syndrome neurofibromatosis type I. However, the extent to which NF1-inactivation promotes sporadic tumorigenesis is unknown. Here we report that NF1 is inactivated in sporadic gliomas via two mechanisms: excessive proteasomal degradation and genetic loss. NF1 protein destabilization is triggered by the hyperactivation of protein kinase C (PKC) and confers sensitivity to PKC inhibitors. However complete genetic loss, which only occurs when p53 is inactivated, mediates sensitivity to mTOR inhibitors. These studies reveal an expanding role for NF1-inactivation in sporadic gliomagenesis and illustrate how different mechanisms of inactivation are utilized in genetically distinct tumors, which consequently impacts therapeutic sensitivity.
Tumor suppressors are often mutated in human cancer; however, the excessive proteasomal destruction of tumor suppressor proteins also promotes tumorigenesis. Here we show that the NF1 protein is destabilized in sporadic GBMs as a consequence of the hyperactivation of PKC. Notably, this destabilization confers sensitivity to PKC inhibitors. In contrast, a separate subset of GBMs that possess NF1 mutations are insensitive to PKC inhibitors but are sensitive to mTOR inhibitors. These findings reveal a broad role for NF1-inactivation in gliomagenesis and illustrate how different mechanisms of inactivation are utilized in the same tumor-type. Moreover they highlight the importance of elucidating the molecular mechanisms that underlie tumorigenesis, as such knowledge may be essential for developing personalized therapies.
The Ras pathway is commonly deregulated in human cancer (Downward, 2003). Genetic alterations can occur in Ras genes, upstream regulators, or downstream effectors. One such regulator is the NF1 tumor suppressor, which encodes a Ras-GTPase activating protein (RasGAP), referred to as neurofibromin (Martin et al., 1990; Xu et al., 1990). Accordingly, NF1-inactivation triggers the aberrant activation of the Ras pathway and loss-of-function mutations in NF1 underlie the familial cancer syndrome neurofibromatosis type I (NF1) (Basu et al., 1992; DeClue et al., 1992).
NF1 patients develop a diverse set of tumor-types, including benign neurofibromas, malignant sarcomas, gliomas, pheochromocytomas, gastrointestinal stromal tumors, and myeloid leukemia (Riccardi, 1992). However, while neurofibromin critically regulates Ras in many tissues, it is unknown whether NF1-inactivation plays a major role in the development of sporadic tumors. NF1 mutations and genomic alterations have been reported in isolated tumor samples and cell lines of various origins (Andersen et al., 1993; Li et al., 1992; The et al., 1993, Thiel et al., 1995). More recently, heterozygous mutations in NF1 were observed in larger panels of glioblastomas (Parsons et al., 2008; (TCGA Research Network, 2008); however homozygous mutations were found to be relatively rare (TCGA Research Network, 2008). Thus, our understanding of how NF1 inactivation may contribute to sporadic tumor development still needs to be refined.
While tumor suppressors are often mutated in human cancer, precocious proteasomal degradation of several prominent tumor suppressors, including p53, PTEN and p27, also contributes to their functional inactivation (Honda et al., 1997; Pagano et al., 1995; Wang et al., 2007). Notably, neurofibromin has been shown to be a direct target of the ubiquitin-proteasome pathway and its regulated inactivation promotes proliferation in response to growth factors (Cichowski et al., 2003). Therefore, in this study we sought to identify the signals that trigger neurofibromin degradation and determine whether its destabilization might also play a role in sporadic tumorigenesis.
We have shown that serum and growth factors trigger the rapid ubiquitination and proteasomal destruction of neurofibromin in many cell types (Cichowski et al., 2003). Accordingly, the specific proteasome inhibitor bortezomib blocks the acute degradation of neurofibromin in NIH3T3 cells (Figure 1A). Antibodies recognizing distant regions of the protein confirm that neurofibromin is completely degraded and that the loss of immunoreactivity is not due to the masking of a specific epitope after stimulation (Figure 1A and 1B). Notably, neurofibromin is a direct and sensitive target of the ubiquitination machinery, which can be visualized by the accumulation of high-mobility ubiquitinated species both in vitro and in vivo (Figure 1C, Figure 2A, Cichowski et al., 2003). These observations highlight the dynamic and exquisite regulation of neurofibromin by the proteasome.
Previous studies suggest that neurofibromin degradation is necessary for maximal Ras activation triggered by growth factors and that its re-expression is required for the appropriate attenuation of this signal (Cichowski et al., 2003). However, while growth factors trigger neurofibromin destruction, nothing is known about the specific signals that mediate this effect. To interrogate the role of individual signaling cascades in this process, NIH3T3 cells were exposed to pharmacological inhibitors of PI3K, MEK and PKC. Only the PKC inhibitor Bisindolylmaleimide I (Bis I) blocked neurofibromin degradation, and did so in response to multiple growth factors (Figure 1D and S1A). Bis I efficiently inhibited PKC activity, as it suppressed the phosphorylation of the MARCKS protein, a well-characterized PKC substrate. Two additional PKC inhibitors including Ro-31-8220 and Bisindolylmaleimide II had similar effects (Figure 1E and S1B). Conversely, activation of PKC via phorbol 12-myristate 13-acetate (PMA) rapidly induced neurofibromin degradation (Figure 1F) and did so in multiple cell types (Figure S1C). Neurofibromin levels were restored by 24 hours, at which time PKC was down-regulated by a negative feedback loop (Liu and Heckman, 1998). Finally, expression of a constitutively active form of PKCα also decreased neurofibromin protein levels (Figure 1G). Thus, these data indicate that PKC activation is both necessary and sufficient to promote neurofibromin degradation, thus identifying the critical signal that mediates its destruction.
PKC is known to promote Ras activation, although the mechanism by which this occurs has not been elucidated (Downward et al., 1990; Marais et al., 1998). To determine whether neurofibromin degradation mediates this effect, we utilized a degradation-resistant fragment of neurofibromin (DRNF1) that is resistant to ubiquitin-mediated degradation in vitro (Cichowski et al., 2003). While endogenous neurofibromin was degraded by PMA in vivo, the DRNF1 was not (Figure 2A). Importantly, DRNF1 attenuated Ras activation, indicating that neurofibromin must be degraded to permit normal levels of Ras activity (Figure 2B). In addition, a mutation that impaired RasGAP catalytic activity was impaired in its ability to suppress Ras-GTP levels (Figure S2). Conversely, inactivation of neurofibromin via RNAi was sufficient to activate Ras in the absence of PMA under these conditions (Figure 2D). Taken together, these results suggest that PKC-driven degradation of neurofibromin is an important Ras-regulatory event, dictating both the amplitude and duration of the Ras signal.
Having established the biochemical architecture of the growth factor receptor-PKC-neurofibromin-Ras signaling pathway, we next asked whether its excessive activation might contribute more generally to tumorigenesis. We investigated gliomas because 1) the aberrant activation of PKC has been implicated in gliomagenesis (Mackay and Twelves, 2007), 2) PKC inhibitors exhibit potent effects in pre-clinical studies in this tumor-type and are currently in clinical development (da Rocha et al., 2002; Mackay and Twelves, 2007), and 3) NF1 patients are predisposed to developing gliomas, indicating that NF1 can function as a tumor suppressor in astrocytic tumors (Riccardi, 1992).
We first investigated whether neurofibromin was constitutively destabilized by the proteasome in human glioblastoma (GBM) cell lines. The GBM-derived U87 cell line has been shown to possess aberrantly high levels of PKCα activity and is sensitive to PKC inhibitors (Yazaki et al., 1996). (Figure 3A, left). Notably, we found that proteasome inhibitors dramatically and rapidly stabilized neurofibromin protein (Figure 3A, left). As with most proteasomal substrates, these inhibitors promoted the initial appearance of higher mobility polyubiquitinated forms of neurofibromin, manifested as an immunoreactive smear, followed by a dramatic accumulation of unmodified protein, which increased by 5.4 fold (Figure 3A, right). Neurofibromin was rapidly stabilized by multiple PKC inhibitors (Bis I, Ro-31-8220, and Bis II) and stabilization was detected in 8 out of 12 GBM cell lines tested (Figure 3B, 3C and 3D). Notably, the 3 cell lines that were insensitive to the effects of PKC inhibitors exhibited the lowest levels of PKC activity (Figure 3D and S3). In addition, PKCα shRNA constructs promoted the accumulation of neurofibromin in U87 cells (Figure 3E). Thus, studies utilizing proteasome inhibitors, multiple PKC inhibitors, and PKC shRNA constructs indicate that neurofibromin is destabilized via PKC in a significant subset of human GBM cell lines.
We next ascertained the biological consequences of preventing neurofibromin degradation. The non-degradable neurofibromin fragment (DRNF1) did not significantly affect the proliferation of GBM cells (Figure 3F, left), although it inhibited colony growth in soft agar (Figure 3F, top right) and dramatically suppressed the ability of such cells to form xenograft tumors (Figure 3F, bottom right). These results indicate that neurofibromin destabilization is essential for the transformed and tumorigenic properties of these GBM-derived cells in vitro and in vivo.
To extend these findings to primary tumor cells we utilized neurosphere cultures derived from a GBM that harbored an amplification of the PDGFRA gene (C.B. and I.K.M., unpublished observations), a well-established activator of PKC. Notably, both PKC and proteasome inhibitors resulted in a rapid stabilization of neurofibromin (Figure 3G). We next looked for evidence of proteasomal degradation of neurofibromin in primary human tumor tissue. Neurofibromin immunoblots were performed using tissue from nine Grade IV tumors that were resected prior to treatment. Notably, no neurofibromin was detected in 2 samples (Figure 3H, lanes 1 and 5), and minimal levels were detected in an additional 2 tumors (lanes 3 and 8). Moreover, in 4 out of 9 samples a high mobility neurofibromin smear was observed, consistent with the presence of polyubiquitinated species. Thus, these observations support our studies in GBM cell lines and primary cultures and are consistent with the model that neurofibromin is actively ubiquitinated and degraded in a significant subset of human GBMs. These findings differ from the genetic observation that only 2.9% of sporadic glioblastomas possess homozygous mutations in NF1 (TCGA Research Network, 2008), and provide a mechanistic explanation for the more frequent loss of expression observed by Western analysis.
In the course of our studies, we assessed the effects of RNAi-mediated neurofibromin loss in GBM cell lines. Unexpectedly, ablation of neurofibromin in U87 and Dbtrg-05mg cells triggered cellular senescence, as demonstrated by a potent growth arrest, a large flattened morphology, and the detection of senescence-associated β-galactosidase (SA-βgal) activity (Figure 4A and B). In contrast, senescence was not observed in Gli36, SF539, U251, T98G, M059J, and U138 cells (Figure 4B and data not shown). Although we have previously shown that NF1 loss triggers cellular senescence of normal human diploid fibroblasts (Courtois-Cox et al., 2006), it was nevertheless surprising that the inactivation of NF1 triggered such a response in human tumor cell lines that possess many additional genetic alterations. Notably however, all senescing cell lines expressed wild-type p53 and retained a functional p53 pathway, as illustrated by their induction of p21 (CDKN1A) in response to doxorubicin (Figure 4C). In contrast, all the GBM-derived cell lines that did not senesce in response to NF1-inactivation harbored p53 mutations and/or a defective p53-response (Figure 4C and (Ikediobi et al., 2006; Van Meir et al.). Moreover, when p53 was inactivated by shRNA prior to the ablation of NF1 expression, the senescence response was mitigated (Figure 4D and data not shown).
Oncogene-induced senescence (OIS) is thought to restrain tumorigenesis by preventing progression of benign tumors to malignancy (Narita and Lowe, 2005). Notably, while NF1 patients are predisposed to developing gliomas, the majority are benign pilocytic astrocytomas that rarely, if ever, progress (Rodriguez et al., 2008). Moreover, the majority of tumors spontaneously stop growing and a subset regress. Therefore, we hypothesized that loss of NF1 in these astrocytic lesions might be restricting tumor development by triggering a senescence response. Two pilocytic astrocytomas from NF1 patients were obtained. p15INK4b, a reported marker of oncogene-induced senescence, and p16INK4a, which is also commonly detected in senescent tissue, were both highly expressed in these tumors (Figure 4E and data not shown) (Collado et al., 2005; Michaloglou et al., 2005). Tumors also express p53, suggesting that this pathway is activated in these lesions. SA-βgal activity could not be assessed in these archived samples; nevertheless, the expression of p15, p16 and p53 support the hypothesis that these tumors are restrained by oncogene-induced senescence. Interestingly, in previous studies Parada and colleagues had shown that inactivating mutations in Nf1 and p53 cooperate to promote the development of GBMs in mice, but only when p53- inactivation precedes, or is coincident with, Nf1-loss (Zhu et al., 2005). Our findings suggest a potential mechanistic explanation for these observations and provide insight into the constraints governing glioma evolution in NF1 patients.
The data shown in Figure 3 suggest that excessive PKC activity and subsequent proteasomal degradation is one mechanism by which neurofibromin can be inactivated in GBMs. However, we noted that 3/15 GBM-cell lines that we were studying did not express detectable levels of neurofibromin protein, even in the presence of PKC and proteasome inhibitors (data not shown). Notably, all three cell lines (U251, LN229 and LN319) harbored p53-inactivating mutations (Ikediobi et al., 2006; Tallen et al., 2008; Van Meir et al., 1994). Since p53 inactivation appeared to be permissive for total inactivation of NF1, we assessed the status of the NF1 gene in each cell line. All three cell lines harbored loss-of-function and/or recurrent mutations found in NF1 patients and no wild-type NF1 was present (Figure 5A).
In NF1 patients single point mutations, small insertions, and deletions are known to occur throughout the NF1 gene (Messiaen et al., 2000). As such, we reasoned that subtle NF1 mutations might be missed by standard genomic analysis. To examine changes in NF1 copy number we examined data from SNP analysis that was performed on archived tumor tissue from 141 spontaneous human tumors (Beroukhim et al., 2007). None of these tumors exhibited complete loss of the NF1 locus, consistent with other reported findings (Jensen et al., 1995). However, 23 (16%) exhibited single copy loss at the NF1 locus (Beroukhim et al., 2007). We then utilized these tissue samples in a directed sequencing effort. Based on our analysis of cell lines, human tumors from NF1 patients, and previous mouse modeling studies, we reasoned that null mutations in NF1 might only be tolerated in the absence of functional p53. Therefore we sequenced the NF1 gene in 14 of the 23 GBM tissue samples that also possessed p53 mutations (Beroukhim et al., 2007).
Three out of the 14 tumors with p53 alterations (21%) possessed homozygous null NF1 mutations (Figure 5B). Importantly, all of these mutations were either recurrent in NF1 patients and/or resulted in a truncation before or within the catalytic GAP-related domain (GRD). In tumor 217, a mutation in exon 37 (c.6789_6792delTTAC) was detected, resulting in a null NF1 allele that has been observed in NF1 patients (Pros et al., 2008). Tumor 311 possessed a mutation in exon 23-2 (c4103delT) resulting in a premature stop codon. In tumor 211, multiplex ligation-dependent analysis (MLPA) indicated that both NF1 alleles were present, however two independent, inactivating mutations were detected; one in exon 7 (c896-897dupTT) and the other in exon 20 (c3496G>T), also representing a recurrent patient mutation. Thus, the detection of null NF1 mutations in 3/15 GBM cell lines and 3/14 primary tumors indicates that genetic inactivation of both NF1 copies does occur in spontaneous human gliomas. While this manuscript was under consideration, heterozygous alterations of the NF1 gene (deletions and point mutations) were reported to occur in GBMs in two other studies (Parsons et al., 2008; TCGA Research Network, 2008). However, the latter study also reported that only a small fraction of tumors 6/203 (2.9%) possessed null mutations in both NF1 alleles. The data in the present study extend these observations by demonstrating that mutations in NF1 and p53 are coincidental in this tumor-type and provide a mechanistic explanation for why complete genetic loss of NF1 necessitates p53 inactivation. The observation that benign pilocytic astrocytomas from NF1 patients express markers of senescence and robust p53 expression, provides additional in vivo evidence to support the hypothesis that p53 plays a critical role in restricting the tumorigenic effects of NF1-inactivation.
The data described thus far suggest that NF1 may be suppressed either by excessive proteasomal degradation of the NF1 protein triggered by excessive PKC activity or by genetic inactivation of the NF1 gene. To investigate whether these distinct mechanisms of inactivation might impact the sensitivity to specific therapeutic agents, the effects of PKC inhibitors on tumor cell lines were assessed. Importantly, the GBM cell lines in which we found neurofibromin to be destabilized by PKC were very sensitive to the PKC inhibitor Ro-31-8220 (Figure 5C), whereas cells that genetically lacked NF1 were relatively insensitive to PKC inhibition (Figure 5D). The difference in IC50 values between genotypes was found to be statistically significant as determined by a Mann-Whitney U test (p= .05). In addition, inactivation of NF1 via an shRNA-expressing lentivirus significantly decreased the response of normally sensitive cells to PKC inhibitors (Extra sum of squares F-test, 1.65 × 10-5) (Figure 5E). A similar differential sensitivity of genetically wild-type versus NF1-mutant cell lines was observed with a second PKC inhibitor (Figure S4). In addition, U87 cells were specifically sensitive to genetic ablation of PKCα (Figure S5).
The sensitivity to PKC inhibitors was also assessed in the primary GBM culture TS543, in which we found neurofibromin to be destabilized by PKC, in comparison to a second GBM culture (TS565), in which we detected 2 NF1 mutations (c2195G>T, c470G>T) and a complete loss of neurofibromin protein expression (Figure S6). Consistent with data from GBM cell lines, the TS543 cultures were exquisitely sensitive to PKC inhibitors, while the TS565 cultures were relatively insensitive (Extra sum of squares F-test, P=4.657.10-6) (Figure 5F). Taken together these data suggest that PKC may represent a potential therapeutic target in GBMs in which the NF1 gene is intact and the protein is destabilized.
However, we and others have previously shown that neurofibromin critically regulates the mTOR pathway and that NF1-inactivation confers sensitivity to mTOR inhibitors in other cell types (Dasgupta et al., 2005; Johannessen et al., 2005). Therefore, in an effort to identify a therapeutic agent that might be effective on GBMs that harbor null genetic mutations in NF1, the effect of the mTOR inhibitor rapamycin was assessed. While GBM cells that genetically lack NF1 exhibited a significantly decreased sensitivity to PKC inhibitors, we found that these cells were exquisitely sensitive to the mTOR (mTORC1) inhibitor rapamycin (Figure 5G). NF1-deficient cells were also sensitive to the inactivation of the mTOR pathway via RNAi-mediated suppression of raptor, which is exclusively a component of the mTORC1 complex (Figure S7). Conversely, the re-introduction of a neurofibromin fragment into NF1 null cells promoted rapamycin resistance, indicating that NF1-loss was responsible for the observed sensitivity (Figure 5G). Notably, in a recent Phase I clinical trial, rapamycin was shown to exhibit anti-tumor activity in a subset of PTEN-deficient glioblastomas (Cloughesy et al., 2008). These studies suggest that NF1 mutations may be an additional genetic event that may impact the sensitivity of sporadic glioblastomas to mTOR inhibitors.
The NF1 tumor suppressor is mutated in the familial cancer syndrome neurofibromatosis type I (Cawthon et al., 1990; Wallace et al., 1990). Nevertheless, gross genomic loss of NF1 does not appear to be a common event in spontaneous human tumors. The studies in this report reveal a critical role for NF1-inactivation in the development of sporadic human gliomas and provide a framework for understanding how proteasomal degradation and complete genetic inactivation of NF1 contribute to this process (Figure 6).
Prior to this work neurofibromin was known to be a direct target of the proteasome, however the precise signals that mediate its destruction had not been identified (Cichowski et al., 2003). In this study we show that activation of PKC is both necessary and sufficient for neurofibromin degradation. This finding has important implications for normal signal transduction pathways as well as tumorigenesis. Specifically these studies demonstrate that the NF1 tumor suppressor is a critical mediator of Ras activation downstream of PKC and experimentally illustrate that robust Ras activation requires the coordinated regulation of GEFs and GAPs. Notably, the intensity of the Ras signal is tightly regulated and is critical for specifying biological responses to growth factors (Marshall, 1995). The observation that numerous receptors can trigger neurofibromin degradation suggests that in vivo the extent of neurofibromin degradation and consequential Ras activation may be dictated by the intensity of combined GFR/PKC signals.
We have also demonstrated that the PKC-neurofibromin signaling axis is aberrantly activated in spontaneous gliomas and that neurofibromin destabilization is required for the tumorigenic properties of tumor cells in vitro and in vivo. Notably, numerous GFRs have been implicated in glioma development (Furnari et al., 2007). Importantly, the majority if not all of these receptors activate PKC. PKC activity has also been independently shown to be hyperactivated in gliomas, presumably due to the aberrant activation of these receptors (da Rocha et al., 2002). However, it remains to be determined whether genetic alterations in PKC itself may also occur in some tumors (Mackay and Twelves, 2007). Interestingly, Ras mutations do not occur in gliomas, however Ras-GTP levels have been shown to be elevated (Feldkamp et al., 1999; Furnari et al., 2007). Our data suggest that the excessive destruction of neurofibromin is one mechanism that contributes to pathogenic levels of Ras activation in this tumor-type.
These findings have important clinical implications. Because multiple GFRs have been shown to be concomitantly activated in a single glioma, it has been suggested that combination therapies designed to target multiple receptors may be required for effective treatment (Stommel et al., 2007). By extension, these studies suggest that PKC inhibitors may provide a potential therapy for gliomas that genetically retain NF1 by simultaneously suppressing tumorigenic signals from multiple activated receptors. Notably, PKC inhibitors have been shown to be effective in pre-clinical studies of GBMs in vitro and in vivo and are currently in clinical development (da Rocha et al., 2002; Mackay and Twelves, 2007). Our data reveal one mechanism by which tumor cells are sensitized to these agents and indicate that PKC may represent an important therapeutic target in GBMs in which the NF1 gene is intact and the protein is destabilized.
While these data indicate that proteasomal inactivation of neurofibromin contributes to gliomagenesis, we have also detected NF1 mutations in cells and tumor tissue from sporadic human gliomas. Two recent studies have also observed mutations in the NF1 gene in GBMs (Parsons et al., 2008; TCGA Research Network, 2008); however, complete genetic loss was reported to be rare (TCGA Research Network, 2008). Thus, our findings suggest that NF1 inactivation may play an even broader role in glioma development than was previously recognized. These studies also provide a biological framework for understanding how different mechanisms of inactivation may be required during tumor evolution. While proteasomal reduction and perhaps even heterozygosity may promote tumorigenicity, this tissue-type appears to be very sensitive to complete inactivation of NF1, and in the latter context p53 plays an important role in restricting tumorigenesis. Genomic analysis of human tumor tissue, cellular studies, the senescent properties of pilocytic astrocytomas from NF1 patients, mouse modeling efforts (Zhu et al., 2005), all suggest that homozygous null NF1 mutations are only tolerated in the absence of an intact p53 pathway. Thus, the present study demonstrates that NF1 and p53 mutations are coincidental in sporadic human gliomas and provides mechanistic insight into why the order of mutational events influences glioma progression. This model also highlights why proteasomal inactivation (perhaps coupled with genetic heterozygosity) may represent a more prevalent mechanism of NF1-inactivation in gliomas.
It should be noted that NF1 shares many functional similarities with PTEN, which is also thought to promote tumorigenesis via genetic inactivation and enhanced proteasomal degradation. Moreover, like NF1, complete inactivation of PTEN early in the tumorigenic process (in the prostate) appears to limit tumor development by promoting cellular senescence (Chen et al., 2005). Such observations suggest that exquisitely sensitive protective mechanisms have evolved to respond to aberrant oncogenic signals above a certain threshold, resulting from complete, but not partial, inactivation of these tumor suppressors. Thus, studies aimed at further elucidating this tumor suppressive response, as well as defining additional proteins that contribute to the proteasomal regulation of these tumor suppressors, may identify new genes that restrict or promote human cancer. Finally, because such protective responses exist, different mechanisms of inactivation may be selected for during the course of tumor development, which may depend on the complement of genetic alterations already present. Moreover, our data suggest that defining the mechanism by which a tumor suppressor is inactivated may ultimately be important for choosing appropriate therapies.
NIH3T3 were cultured in DMEM supplemented with calf serum. IMR90, RT4, A172, U138, M059J, U87, U251, Gli36, SF539, Dbtrg-05mg, HUVECs and MEFs were cultured in DMEM supplemented with fetal calf serum.
The collection and analysis of human tissue samples was approved by the Institutional Review Boards of the University of California, Los Angeles, and Memorial Sloan-Kettering Cancer Center. Informed consent was obtained from all subjects. For tissues obtained in Figures 3H and and4E4E the Office of Research Protections at Brigham and Women’s Hospital ruled that this study does not represent human subjects research due exemption 32 CFR 219.101(b)4: the research involved collection or study of existing pathological specimens that are publicly provided without identifiers.
Tumor sphere (TS) cell cultures were isolated from primary human glioblastomas by disassociation into a single cell suspension with Accumax (Innovative Cell Technologies Inc). Cells were filtered through a 100um filter, then washed and plated in NeuroCult NS-A proliferation media (Stem Cell Technologies 05751) supplemented with EGF (20ng/mL; R&D Systems 236-EG-200), FGF (10ng/mL; R&D Systems 2099-FB), and heparin (2μg/mL).
NIH3T3 and IMR90 cells were plated in serum-free medium at 5 × 105 cells/10-cm plate. HUVEC, RT4 and GBM cell lines were plated in 1% serum and primary GBM cultures were plated in the media described above. After 18 h, cells were treated with either 1μM of BisI, 1μM BisII, 2.5-5μM Ro-31-8220 (Calbiochem), or 1μM Go 6976 (LC Laboratories) or vehicle. In Figure 1 cells were pre-treated with inhibitors 30 minutes before adding 6μM LPA, 20ng/mL PDGF or PMA 10ng/mL. Cells were lysed at specified time points with 1% SDS boiling lysis buffer or RIPA buffer.
The following antibodies were used: neurofibromin (C-terminal antibodies: UP69, a polyclonal antibody raised against a KLH conjugated peptide- RNSIKKIV [used for all experiments except Fig. 3G] (Courtois-Cox et al., 2006), and A300-140A (Bethyl Laboratories). N terminal antibody: Ab NF1-5.16 (a monoclonal antibody that was a gift of K. Scheffzek, unpublished); pMARCKs (Calbiochem); HA (12CA5, Roche), PKCα (BD Biosciences); p21 WAF1/cip-1 (c-19, Santa Cruz); phospho-PKC motif #2351 (Cell Signaling), GAPDH (Cell Signaling) (Johannessen et al., 2005).
Retroviral and lentiviral infections were performed and as described (Johannessen et al., 2005). PKCαCAT-HA and DRNF1-FLAG were introduced into the pbabe retroviral vector. The DRNF1-FLAG construct possesses a FLAG epitope tag and encodes amino acids 1175-1535. This protein fragment retains RasGAP activity and has been shown to be resistant to ubiquitination and degradation in vitro (Cichowski et al., 2003). The lentiviral pLKO vector containing the following shRNAs were used: NF1#2(H9), 5’-TTATAAATAGCCTGGAAAAGG-3’; PKCα1 (e1), 5’CCGTCTTAACACCACCTGA TCTCGAGATCAGGTGGTGTTAAGACGG-3’; PKCα #2(D12), 5’-ATGGAACTCAGGCAGAAATTCT CGAGAGAATTTCTGCCTGAGTTCCAT-3’. The retroviral pMSCV-pM vector containing the following shRNA was used p53shRNA, 5’-GGCCTGACTCAGACTGACATT-3’.
Cells were cultured in the presence of .25% serum and 10ng/ml of PMA was added for increasing amounts of time. Ras-GTP levels were detected using a Ras activation assay, following the manufacturer’s instructions (Pierce Biotechnology). Ras-GTP activation was quantified by using Photoshop to assess the amount of Ras that was pulled down with the RBD, divided by total levels of Ras in the lysates. Values at 5 and 15 minutes were then normalized to control to establish levels of fold Ras activation.
GBM cell lines and primary neurosphere cultures were treated with 1μM bortezomib (LC Labs) or 10μM MG132 (Boston Biochem) and 1μM bortezomib (LC Labs) as specified and lysed at designated time points.
Soft agar assays were performed as described (Johannessen et al., 2005).
Animal procedures were approved by the Center for Animal and Comparative Medicine in Harvard Medical School in accordance with the NIH Guild for the Care and Use of Laboratory Animals and the Animal Welfare Act. Subcutaneous implantations in athymic male nude mice (nu/nu; 6 weeks old) were carried out as previously described (Berger et al., 2004). Briefly, 2.5×106 of U87 cells expressing pbabe DRNF1-FLAG or pbabe vector in 100μl of PBS were injected subcutaneously (six injections per condition). Two weeks after injection (day 0) the width and length of the tumor were measured every 2 to 3 days by caliper and the volume was calculated with the following formula: Volume = (Length × Width2) π/6.
U87 cells expressing pbabe DRNF1-FLAG or pbabe vector were plated in triplicate at 5×105 in a 6-well dish and counted every other day. Dbtrg-05mg cells expressing p53shRNA and NF1shRNA or p53shRNA and vector were plated in at 2×105 and counted on day 1, 7, and 9. For the cumulative population doubling (PDL) assay 2×105 cells were plated in triplicate in 6-well dishes. Cells were counted and re-seeded at a density of 2×105 every three days for five passages. Population doublings were calculated as described (Courtois-Cox et al., 2006).
SA-β-Gal activity was performed as described and percentages were assessed by counting at least 300 cells (Dimri et al., 1995).
Immunohistochemistry for p16 was performed as described using the p16INK4a antibody (Ab-7, MS-1064-P, Neomarkers), the p53 antibody (D07, Signet Laboratories), or the p15 antibody (C-20, sc-612, Santa Cruz) (Courtois-Cox et al., 2006).
The entire NF1 coding region was amplified in 5 overlapping RT-PCR fragments and used as the template for direct sequencing as described (Messiaen et al., 2000). Copy number analysis by multiplex ligation-dependent probe amplification (MLPA) was performed as described (Wimmer et al., 2006). The nomenclature of the mutations is based on NF1 mRNA sequence NM_000267.1, with 1 being the first nucleotide of the ATG start codon. The NF1 exons are named according to the most widely used nomenclature adapted by the researchers and diagnostic labs, which does not use strictly consecutive numbers.
3×103 NF1+/+ and NF1-/- GBM cell lines were plated in replicates of 8 in a 96-well dish. Cells were treated with 0, 0.625, 1.25, 2.5, 4, 5, 7.5, 10 μM of Ro-31-8220 or vehicle for 18 hrs before assessing viability by adenosine triphosphate quantification using a Cell Titer-Glo Luminescent Cell viability assay kit (Promega). Gli36 cells were infected with NF1#2 shRNA or vector and selected in puromycin for 2 days. 3×103 cells were plated in 96-wells and IC50 values were determined as described above. To calculate IC50 values for GBM neurospheres TS-543 and TS-565, cells were plated in growth factor-free media and treated with the specified PKC inhibitor at the concentrations indicated for 24 hours. Cells were evaluated for viability and cell counts using Vi Cell XR (Beckman Coulter).
20,000 cells were seeded per well of a 6-well dish. Twenty-four hours later, cells were treated with various concentrations of rapamycin in triplicate. Log phase cells treated with rapamycin or vehicle were isolated after 3 doubling periods and viable cells were counted.
All numerical data including error bars represent the mean ± the standard deviation. Specific statistical tests for each experiment are described in the text.
We thank Wade Harper, Steve Elledge and Gerard Evan and Bill Hahn for helpful discussions. This work was supported by grants from the NCI (R01 CA111754-01) and the DOD (W81XWH-08-1-0136).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.