|Home | About | Journals | Submit | Contact Us | Français|
eNOS expression is elevated in human glioblastomas and correlated with increased tumor growth and aggressive character. We investigated the potential role of nitric oxide (NO) activity in the perivascular niche (PVN) using a genetic engineered mouse model of PDGF-induced gliomas. eNOS expression is highly elevated in tumor vascular endothelium adjacent to perivascular glioma cells expressing Nestin, Notch, and the NO receptor, sGC. In addition, the NO/cGMP/PKG pathway drives Notch signaling in PDGF-induced gliomas in vitro, and induces the side population phenotype in primary glioma cell cultures. NO also increases neurosphere forming capacity of PDGF-driven glioma primary cultures, and enhances their tumorigenic capacity in vivo. Loss of NO activity in these tumors suppresses Notch signaling in-vivo, and prolongs survival of mice. This mechanism is conserved in human PDGFR amplified gliomas. The NO/cGMP/PKG pathway’s promotion of stem cell-like character in the tumor PVN may identify therapeutic targets for this subset of gliomas.
In the normal brain, capillaries located in the subventricular zone (SVZ) and hippocampus form the major structural entity of the neural stem cell niche (Riquelme et al. 2008) and the perivascular region is the location for neural stem cells (Palmer et al. 2000; Louissaint et al. 2002). This proximiy between neural stem cells and the vasculature is believed to facilitate intercellular communication between neural stem cells and endothelia that release soluble factors critical for promoting stem cell renewal (Shen et al. 2004; Ramirez-Castillejo et al. 2006).
Glioblastoma multiforme (GBM) is the most malignant and aggressive type of central nervous system tumor (Legler et al. 1999) and are classified into groups based on gene expression profiles where approximately 30% are designated “proneural” and show evidence of PDGF signaling (Phillips et al. 2006; Brennan et al. 2009). As gliomas progress to higher grade they develop several histologic structures that define malignant behavior including microvascular proliferation or hypercellular vasculature. Microvascular proliferating regions of these tumors are grossly disorganized angiogenic vessels that are a hallmark of malignant behavior and are surrounded by a perivascular niche that is a habitat for brain tumor stem-like cells (Calabrese et al. 2007). Soluble factors released from endothelial cells promote the self-renewal and proliferation of brain tumor stem-like cells (Calabrese et al. 2007; Folkins et al. 2007). In medulloblastomas, stem-like cells in the perivascular niche are resistant to radiation and are believed to give rise to tumor recurrence (Hambardzumyan et al. 2008b). The underlying mechanism(s) responsible for the formation of vascular stem cell niches that maintain tumor cells in a stem-like state are not understood, moreover, the mechanisms responsible for driving stem-like character in the perivascular niche and the endothelia-derived factors that support cancer stem-like cells of the perivascular niche in brain tumors have not been identified.
Nitric oxide synthases are a family of enzymes that produce nitric oxide (NO) from their substrate Larginine. NO regulates many physiological processes through the NO/cGMP pathway as well as through protein S-nitrosylation (Fukumura et al. 2006). During NO/cGMP signaling, NO produced from one cell diffuses to neighboring cells where it binds to its receptor soluble guanylate cyclase (sGC). sGC converts GTP to cGMP to activate several downstream effectors including cGMP-dependent protein kinase (PKG). Many of the activities of NO signaling can be mimicked by cGMP analogs that activate PKG. The endothelial isoform of nitric oxide synthases, endothelial nitric oxide synthase (eNOS) is required for initiation and maintenance of human pancreatic tumor growth (Lim et al. 2008) and eNOS is elevated in various cancers (Fukumura et al. 2006) including human gliomas (Bakshi et al. 1998; Broholm et al. 2003) where its expression is correlated with glioma grade (Cobbs et al. 1995). Elevated levels of eNOS expression and activity in gliomas, is often localized to the tumor vascular endothelium (Iwata et al. 1999). However the specific role of eNOS in gliomagenesis has not been fully established.
We hypothesized that NO produced by eNOS in endothelial cells functions in a paracrine manner to activate signaling pathways in glioma cells in the perivascular niche and thereby promotes or reinforces stem-cell character. We used a genetically engineered mouse model of PDGF-induced gliomas to investigate the role of NO in gliomas. This model shows eNOS expression restricted to the tumor vascular endothelium and that a population of stem-like cells expressing Nestin and Notch1 are tightly apposed to the tumor endothelium. These Nestin-expressing stem-like cells also express sGC, the receptor for NO, and NO activates Notch in glioma stem-like cells through the NO/cGMP/PKG pathway. This NO-induced activation of Notch signaling in stem-like cells accelerates glioma initiation and tumor formation in mice. We further show that mice lacking eNOS have delayed gliomagenesis and subsequent enhanced survival correlating with decreased activation of the Notch pathway. We further demonstrate that this mechanism is conserved in human PDGFR-amplified gliomas. We also show that NO activates the Notch pathway to enhance the side population (SP) phenotype in cultured human glioma cells.
Factors derived from the tumor vascular endothelium reinforce the self-renewal of stem-like cells residing in the brain tumor perivascular niche (PVN) (Calabrese et al. 2007). Since elevated eNOS expression and activity is restricted to the tumor vascular endothelium, we investigated whether NO might be one factor promoting stem-like activity within the niche. In order to investigate the potential role of NO on stem-like character, we analyzed its effect on pathways known to regulate stem cell character namely Notch, Shh and Wnt (Taipale and Beachy 2001; Radtke and Raj 2003). Luciferase reporters coupled to the promoters of Hes1, Gli1 or β-Catenin, were transiently transfected into U251 human glioma cells. As shown in (Figure 1A), S-nitrosoglutathione (GSNO) - an NO donor, induced a more than two-fold increase in luciferase expression in the Hes1-luciferase expressing cells relative to untreated controls (80.57 ± 0.55 versus 28.32 ± 0.13; P < 0.0001). This effect was specific to activation of the Hes1 promoter as there was no statistically significant difference in activation of Gli1 or β-Catenin promoters (41.15 ± 0.81 versus 39.91 ± 0.37 and 5.44 ± 0.23 versus 4.23 ± 0.19 respectively) following GSNO treatment. Nestin, a well-characterized marker of stem/progenitor cells in brain tumors, is highly expressed in stem-like cells of the glioma PVN and Notch signaling is known to activate the Nestin promoter in gliomas (Shih and Holland 2006). In addition, loss of eNOS was demonstrated to decrease Nestin expression in the brain in vivo (Chen et al. 2005). Therefore, we determined whether NO affects Nestin expression in human glioma cells. Transient transfection of the U251 cell line with a Nestin-luciferase reporter indicated that GSNO treatment led to an approximately two-fold induction of the Nestin reporter relative to controls (77.95 ± 2.55 versus 38.84 ± 0.66; P < 0.0001) (Figure 1A). We confirmed activation of the Notch pathway in U251 cells by Western blot for HES1 protein following GSNO treatment (Figure S1A). In addition we analyzed the mRNA transcripts encoding HES1, NESTIN, GLI1 and β-CATENIN in these cells following treatment with GSNO. The mRNAs Hes1 and Nestin were significantly elevated relative to controls (10.8 ± 2.45 versus 1 ± 0.26) and (5.2 ± 1.36 versus 1 ± 0.29) respectively while Gli1 and β-Catenin were unchanged (1.4 ± 0.56 versus 1 ± 0.39) and (0.97 ± 0.22 versus 1 ± 0.28) (Figure S1D) respectively. These data indicate that NO can specifically activate the Notch pathway in human glioma cells.
To further investigate the connection between NO and the Notch pathway in gliomas, we employed the RCAS/tv-a method for creating PDGF-induced gliomas in mice, because the well-characterized robust perivascular niche microenvironment and histological features of this model closely mimic those observed in human gliomas (Holland 2004). Western blot analysis demonstrated that both eNOS and cleaved Notch1 (Notch intracellular domain-NICD) were highly elevated in PDGF-induced mouse gliomas with respect to the contralateral side of the brain (P<0.0001) (Figure 1B). Using immunofluorescence we investigated their spatial relationship to one another within the glioma PVN. Immunostaining for total eNOS protein within the PDGF-induced gliomas indicated that eNOS co-localized with CD31-expressing endothelial cells (Figure 1C) surrounded by a population of Nestin-expressing cells that also co-express Notch1 (Figure 1D–E). These Nestin-expressing perivascular cells also express soluble guanylyl cyclase (sGC - the major receptor for NO) (Madhusoodanan and Murad 2007) whose staining is limited almost exclusively to the perivascular niche (Figure 1F), and which therefore may represent a population of cells within the niche that can respond to NO signaling.
The data above suggests a regional correlation between eNOS expression and Notch1 activation in vivo. In order to determine whether there is a direct link between NO signaling and Notch signaling within PDGF-induced mouse gliomas, we investigated whether NO could upregulate the Notch signaling pathway in culture. Western blot analysis of GSNO treated PDGF-induced glioma primary cultures (PIGPCs) revealed a dose-dependent increase in Notch intracellular domain (NICD) indicating activation of the Notch pathway (Figure S1C). Cell viability was not adversely affected after 6 hrs of treatment (data not shown). We then examined the effect of NO on the expression of Notch ligand proteins and the downstream protein targets of Notch, Hes1 and Hey1. GSNO treatment of PIGPC indicated specific activation of the Notch pathway, as evidenced by a substantial increase in components of the activated Notch pathway (Figure 2A). The expression of the Notch ligand proteins, Delta-like1 and 4 (DLL1 and DLL4), increased within 30 minutes, which coincided with elevated NICD. Expression of the transcription factor targets of Notch signaling, Hes1 and Hey1 was subsequently elevated at 1 hour.
Activation of the Notch pathway plays a critical role in promoting stem-like character in brain tumors (Fan et al. 2006). Therefore, we investigated whether NO was involved in mediating this effect in PDGF-induced gliomas using side population (SP) analysis. SP analysis is used for the identification of stem cells via florescence activated cell sorting and is based on the capacity of stem cells to efflux Hoechst fluorescent dyes by the activity of ATP binding cassette transporters (ABC transporters) (Goodell et al. 1996). The SP cells in human glioma cell lines as well as other tumors are enriched in tumorigenic cells with stem cell properties and the SP cells of PDGF-induced gliomas are more capable of growing as tumor neurospheres and are more tumorigenic than non-SP cells when transplanted in mice (Bleau et al. 2009). PDGF-induced glioma primary cultures (PIGPCs) were incubated with Hoechst 33342 in the presence or absence of GSNO and assayed for their SP. These PIGPCs contained an SP that ranged from ~3% to 20% at baseline depending on the tumor. Following 2–2.5 hours of NO treatment a significant increase (2–5 fold) in the SP was observed relative to vehicle treated cells derived from the same tumor (15.50 ± 3.28 versus 4.25 ± 0.48; P = 0.015) (Figure 2B and S2A). Both control and GSNO-treated cultures were judged to be approximately 90% viable indicating that selection for viable cells was not occurring during the treatment (Figure S2B).
We hypothesized that NO might preferentially up-regulate Notch signaling in a subpopulation of PIGPC cells. To address this possibility, primary cultures were pretreated with vehicle or GSNO, sorted for side and main populations and analyzed by Western blot for NICD. The relative increase of NICD seen above, was greater in cells of the SP (Figure 2C) suggesting that NO activates the Notch pathway in a population of glioma cells which may promote their SP phenotype or stem cell-like characteristics.
ABCG2 is expressed in glioma stem-like cells and its expression was correlated with increasing glioma grade (Jin et al. 2009). Furthermore, abcg2 gene expression is specifically up-regulated in the cancer stem-like populations of mouse PDGF-induced gliomas (Bleau et al. 2009). We investigated whether NO might drive the expression of ABCG2 protein as an additional measure of NO activation of the Notch pathway. Therefore, we analyzed 4 PIGPCs treated with GSNO by Western blot for the expression of ABCG2 relative to vehicle treated controls. All four primary glioma cultures examined showed increased ABCG2 protein expression following GSNO treatment versus controls (69.67 ± 15.48 versus 22.72 ± 3.21; P = 0.041) (Figure 2D).
To further investigate whether Notch signaling drives the SP phenotype in gliomas as it does in medulloblastomas (Fan et al. 2006), we treated these PIGPCs for two hours with the gamma secretase inhibitor (GSI) MRK-003 (Lewis et al. 2007). The baseline SP in these primary glioma cultures was reduced by GSI treatment, suggesting that Notch signaling is critical for the maintenance of the SP phenotype in PDGF-induced gliomas (Figure S3A). We investigated whether the increase in the SP phenotype induced by NO is dependent on Notch activation. PIGPCs were incubated for two hours with GSI in the presence or absence of GSNO, then analyzed for their SP. Treatment of these primary glioma cultures with GSI abolished the GSNO-induced increase of the SP (13.88 ± 1.78 versus 0.33 ± 0.13; P = 0.003) (Figure 3A and S3B) suggesting that NO requires activation of the Notch pathway to drive the SP phenotype in PDGF-induced gliomas. Control, GSNO and GSI treated cultures were approximately 90% viable by PI staining, which confirms that cell viability was not adversely affected by the treatments (Figure S3C). We confirmed the specificity of GSI-induced Notch inhibition with RNAi. Notch1-shRNA (Notch1-SH) knock down of Notch1 mRNA in PIGPCs resulted in a 50% reduction of Notch1 protein and significantly decreased the SP relative to a nonspecific scramble control (P = 0.0109) (Figure 3B and C). To further verify that Notch signaling mediates the enhanced SP phenotype observed above, PIGPCs were infected with a vector expressing constitutively active Notch (NICD), and then analyzed for their SP phenotype. NICD over-expression induced a more than 2-fold increase in the SP when compared with the empty vector control (14.3 ± 2.70 versus 6.59 ± 1.41; P = 0.0226) (Figure 3D). This data indicates that Notch signaling is necessary and sufficient for the NO-induced elevation in SP phenotype in these glioma cells.
We next investigated whether the relationship between NO and Notch signaling seen in culture was conserved in PDGF-induced mouse gliomas in vivo. Ten glioma-bearing mice were treated with the NOS inhibitor L-NG-nitroarginine methyl ester (LNAME) for 24 hours. Tumor tissue and contralateral normal brain were analyzed by Western blot for Notch cleavage (NICD) and Notch ligands. Nine vehicle-treated PDGF glioma-bearing mice were used as controls. The amount of NICD in tumors was significantly elevated relative to the normal brain in all cases (P < 0.0001) (Figure 4A and Figure S4A). NICD was significantly diminished in LNAME treated mice relative to untreated controls, (79.35 ± 7.23 versus 130.4 ± 8.99; p < 0.0004) and the expression of the Notch ligand Jagged 2 was also significantly lower relative to the untreated controls (64.73 ± 12.57 versus 114.8 ± 3.51; p< 0.05) (Figure 4A). To determine if suppression of NO activity by LNAME would affect the SP in these PDGF-induced gliomas we analyzed the SP of gliomas in 6 mice treated for 3 days with LNAME relative to 6 vehicle-treated tumor bearing mice of the same age and background. We found that the SP for the treated group was significantly lower than the vehicle treated controls (2.8 ± 0.22 versus 4.2 ± 0.43; p < 0.0196) (Figure 4B and Figure S5).
To assess whether complete loss of eNOS would affect the development of PDGF-induced gliomas in vivo, we crossed mice carrying a homozygous disruption of eNOS (eNOS−/−), into Nestin tv-a (N-tva) mice and infected the progeny with RCAS-PDGF. We compared the survival of eNOS−/− mice with their respective wild type littermates (eNOS+/+). The survival of eNOS−/− mice (N = 43) was significantly longer than eNOS+/+ mice (N = 42) (P = 0.0042) (Figure 4C). Since NO activates the Notch pathway in these PDGF gliomas, we tested whether the increased survival observed with the loss of eNOS correlated with a decrease in activation of the Notch pathway during tumor progression or at the time of death. Using Western blot analysis, we compared the levels of protein expression of Notch pathway components between eNOS+/+ mice (approximately 54 days, the median age of death in eNOS+/+ mice) with respective eNOS−/− counterparts of the same age. eNOS+/+ mice expressed significantly higher levels of NICD and the Notch ligand DLL1 relative to their eNOS−/− counterparts. However, the expression levels of NICD and Notch ligands Jagged 1 and DLL 1 and 2 in the tumors at death were similar in both populations (54 days for eNOS+/+ and 84 days for eNOS−/−) (Figure S4B) implying that the development of Notch signaling was delayed in the absence of eNOS, but eventually reached a level similar to eNOS+/+ mice by the time the tumors were sufficiently aggressive to be lethal.
Downstream NO signaling can be cGMP dependent or cGMP independent. However, many physiological processes regulated by NO involve cGMP, which activates cGMP dependent protein kinase (PKG) (Blaise et al. 2005). To investigate whether the increase in the SP phenotype induced by NO involved PKG, we determined if PKG activation could increase the SP phenotype in PIGPCs similar to NO. We utilized the non-hydrolysable cGMP analog and potent activator of PKG, 8-Br-PET-cGMP (PET-cGMP) to address this question. PIGPCs were treated with vehicle or Pet-cGMP for 2 hours then analyzed for their SP. When compared with vehicle treated controls, PET-cGMP enhanced the percent of SP cells in PIGPCs to levels exceeding that achieved by GSNO alone (9.00 ± 5.03 versus 48.33 ± 7.67; 9.00 ± 5.03 versus 26.7 ± 7.84 respectively) (Figure 5A). This induction ranged from a 3 to 11-fold increase in the SP relative to vehicle treated controls (Figure S6). Furthermore, the increase in the SP observed with GSNO and Pet-cGMP combined was no more than that observed by either individual treatment (Figure 5A), suggesting that GSNO and Pet-cGMP act in the same pathway. To confirm activation of Notch signaling, PIGPCs were treated for 2 hrs with Pet-cGMP, then analyzed by Western blot for activation of the Notch pathway. Western blot analysis on these primary cultures revealed increased NICD and enhanced expression of the Notch downstream target Hes1 (Figure 5B). As an additional measure of the potential effect of PKG activity on Notch signaling, we used an alternative approach to enhance the activity of PKG in our glioma primary cultures. PIGPCs were treated with the phosphodiesterase 5 (PDE5) inhibitor, sildenafil, which enhances steady state levels of cGMP by suppressing its degradation by PDE5 (Bender and Beavo 2006). As shown in Figure 5C PIGPCs treated with sildenafil for 1 hour show increased expression of NICD protein. The PKG target vasodilator-stimulated phosphoprotein (VASP), was phosphorylated at serine 239 (Butt 2009), indicating that PKG was activated by Pet-cGMP and sildenafil (Figure 5B–C). Sildenafil is a direct inhibitor of ABC transporters (Oesterheld 2009) therefore, SP analysis of primary glioma cultures treated with sildenafil could not be conducted. To confirm that PKG activity is required for NO-mediated enhancement of the SP phenotype, PIGPCs were treated with the PKG inhibitor KT5823 in the absence or presence of GSNO and analyzed for their SP characteristics. As shown in Figure 5D, in the presence of KT5823, induction of the SP by GSNO or Pet-cGMP was diminished relative to their respective individual treatments (6.11 ± 2.43 versus 17.30 ± 3.6 and 5.36 ± 0.87 versus 24.95 ± 0.55; P=0.0027). VASP phosphorylation was used to confirm inhibition of PKG by KT5823 (Figure 5E). Similar results were obtained using a second inhibitor of PKG, Rp-8-pCPT-cGMP, which inhibits PKG by a different mechanism than KT5823 (data not shown). Collectively, these data indicate that NO enhances Notch signaling and the SP phenotype in PDGF-induced gliomas through activation of PKG. Endothelial cells are known to express ABC transporters and are a component of the SP in PDGF-induced gliomas (Bleau et al. 2009). Moreover, they can activate the cGMP/PKG signaling pathway in response to NO (Fukumura et al. 2006). We verified that the NO-induced increase in the SP phenotype was not due to the presence of contaminating endothelial cells within these PIGPC by analyzing 6 independent PIGPCs by Western blot for the expression of the endothelial cell markers, eNOS and CD31 (Figure S1E).
Thus far, we have shown that in a population of PIGPC cells, NO activates the SP phenotype and Notch signaling, both consistent with stem-like characteristics. These characteristics are observed after transient (2 hours) activation of the NO/cGMP pathway with GSNO or the cGMP analog, Pet-cGMP. Using a neurosphere formation assay, we then investigated whether this transient activation might have long-term effects on the overall capacity of these PIGPC to behave in a stem-like manner after withdrawal of treatment. When cultured with the growth factors EGF and bFGF, in the absence of serum, neural stem cells form neurospheres. Neurosphere-forming glioma cells represent the more stem-like cell populations in brain tumors (Singh et al. 2003; Galli et al. 2004) and are enriched in the SP (Patrawala et al. 2005; Harris et al. 2008). In addition, SP cells generated from these PDGF-induced gliomas, were previously shown to be enriched with the self-renewing stem-like cell populations, which propagate the neurosphere forming capacity of these gliomas (Bleau et al. 2009). Therefore, we pretreated PIGPCs individually with GSNO or Pet-cGMP for 2 hours, after which treatment was withdrawn. The cells were re-suspended, plated at clonal density (1cell/µL) in neurosphere medium and then subjected to an in vitro tumor neurosphere-forming assay that was quantified by counting the number of neurospheres formed two weeks after treatment. Neurosphere formation in GSNO or Pet-cGMP treated glioma primary cells, was approximately twice as fast and more than double the number observed in vehicle treated plates (48.83 ± 8.167 versus 20.33 ± 2.171; p < 0.05 and 76.17 ± 5.845 versus 20.33 ± 2.171; p < 0.005) (Figure 6A). In addition, tumor neurospheres generated from each group could be serially passaged to generate secondary and tertiary neurospheres demonstrating their capacity for self-renewal. These neurospheres also stained with antibodies against the stem cell markers Nestin, Nanog, Musashi and Oct-4 (Figure 6B). These results indicate that short-term activation of the NO/cGMP pathway in a population of PIGPCs can induce a fundamental change in the behavior of these tumor cells to a more stem cell-like phenotype.
The SP cells of PDGF-induced gliomas are enriched with the self-renewing stem cell-like populations, and are more tumorigenic than non-SP cells when injected into neonatal mice (Bleau et al. 2009). The data observed from the neurosphere formation assay above suggest that transient activation of the NO/cGMP pathway in vitro in a population of PIGPCs is sufficient to effect long-term changes toward a more stem cell-like phenotype within these gliomas. Therefore, we then determined if the pro-stem celllike characteristics achieved in these glioma cells by transient activation of the NO/cGMP pathway affected tumorigenesis in vivo. PIGPC cells were treated for 2 hrs with Pet-cGMP or GSNO as above and immediately injected into the cortex of neonatal mice (8 × 104 cells per pup, N=20 Pet-cGMP population; N=15 GSNO population). As a control, the same number of untreated PIGPC cells, were injected into the cortex of a second population of neonatal mice (N=31). Mice were closely monitored and sacrificed if they developed symptoms of hydrocephalus, lethargy or cachexia. Tumor formation was confirmed by histology, which showed tumors to be diffuse and some to exhibit histological features of high-grade gliomas such as the presence of microvascular proliferation and pseudopalisading necrosis (Figure 6C). As shown in Figure 6D, mice injected with PIGPCs pretreated with the cGMP analog Pet-cGMP developed tumors with both higher incidence and shorter latency than their control counterparts. Injection of Pet-cGMP treated primary glioma cells resulted in tumors at 45% incidence (9 of 20), compared with a 3% incidence (1 of 31) obtained from injections of untreated primary glioma cells (P = 0.0004). In addition, GSNO-treated PIGPCs pretreated for 2 hours and injected into a similar number of mice of the same background resulted in a tumor incidence of 27% (4 of 15) compared with the lower incidence seen in untreated controls (P = 0.033). Moreover the tumor incidence from GSNO treated PIGPCs was approximately 20% less than the 45% seen with PET-cGMP and consistent with the relative effects of these two treatments on the SP phenotype described above.
Thus far, using the RCAS/tv-a mouse model, we demonstrated that activation of NO/cGMP signaling in PDGF-induced gliomas promotes stem cell-like characteristics. In light of these data from this mouse model, along with previous findings showing elevated expression of eNOS in human gliomas, (Cobbs et al. 1995; Iwata et al. 1999), we wondered if NO might play a role in promoting stem cell-like activity within the PVN in the context of human gliomas. The T98G human glioma cell line was previously shown to contain a SP (Chua et al. 2008). We therefore treated these cells with GSNO and analyzed them for their SP phenotype relative to controls. GSNO induced an approximate 5-fold increase in the SP when compared with vehicle treated controls (13.87 ± 0.22 versus 3.43 ± 0.15; p< 0.0001) (Figure 7A). The induction of the SP by GSNO was blocked in the presence of the GSI (13.87 ± 0.22 versus 1.16 ± 0.04; p<0.0001) suggesting a role for Notch in mediating this effect. Confirmation of Notch pathway activation was performed by Western blot analysis (Figure S1B). As the PDGFR amplified subtype of human gliomas is accurately modeled in the PDGF-induced mouse model, we treated human primary cultured neurospheres from a PDGF receptor (PDGFR) amplified human glioblastoma sample with GSNO. Following NO treatment, there was an approximate 3-fold induction of the SP when compared with vehicle treated controls (18.87 ± 0.79 versus 6.87 ± 0.37; p<0.005). This GSNO-induced increase in the SP was blocked in the presence of the GSI (18.87 ± 0.79 versus 2.87 ± 1.14; p< 0.005) (Figure 7B), indicating that NO induction of the SP phenotype through activation of Notch is conserved in a subset of human gliomas. When combined, the percent increase in the SP by GSNO and Pet-cGMP was no more than with either individual treatment, as was seen in our PDGF-induced glioma mouse model. In addition, we analyzed 5 PDGFR amplified human glioma tissue samples by immunohistochemistry for the expression of ENOS, SGC, and NESTIN. In all 5 PDGFR-amplified human glioma samples, ENOS showed vivid expression in the tumor vasculature and in 3 of 5 samples, NESTIN expression was enhanced in cells surrounding the tumor vascular endothelium (Figure 7C), with the remaining 2 samples showing diffuse NESTIN staining throughout the tumor. The SGC staining paralleled the NESTIN expression in these tumors (data not shown). Identical immunohistochemical analysis of 6 human EGFR amplified human gliomas did not reveal similar ENOS and NESTIN expression limited to the PVN. (Figure S7).
To assess the tumorigenic potential of NO activated human glioma primary neurospheres, immunocompromised mice were injected with 2.5×105 GSNO treated primary glioma cells and compared to their age matched controls injected with an equal number of the same vehicle treated human primary glioma cells (10 mice per group). All mice developed tumors, however the survival difference between the two groups was statistically significant (p<0.005) with a shorter survival for mice receiving the NO activated cells. The median survival of vehicle and GSNO treated human primary glioma cultures was 36 and 29 days respectively (Figure 7D). These data suggest that within the PDGFR gene amplified subset of human gliomas, elevated nitric oxide activity within the tumor vasculature may play a role in promoting stem cell-like characteristics within some of these tumors.
Stem cells within the brain are localized to specific microenvironments that regulate their maintenance (Doetsch 2003). A perivascular location for stem cells has been described for normal (Shen et al. 2004; Ramirez-Castillejo et al. 2006) and brain tumor stem-like cells (Calabrese et al. 2007; Hambardzumyan et al. 2008a). In the context of brain tumors, regions of microvascular proliferation represent these vascular stem-like cell niches. As progression to more aggressive gliomas is associated with a prominent increase in the density of these vascular structures (known as microvascular proliferation), their presence in gliomas is a criteria for malignant histology by the WHO grading system (Kleihues et al. 1995). This association with increasing grade has led to proposals that the perivascular niche might directly contribute to progression of brain tumors. One proposal for the perivascular location of stem-like cells has been that factors released from endothelial cells intricately associated with the niche, activate one or more signaling pathways in adjacent cells that reinforces their stem-cell like character (Calabrese et al. 2007). Our finding suggests that one of these mechanisms is the activation of Notch signaling by NO produced from the endothelial cells. We propose that NO released from the tumor endothelium, diffuses to neighboring glioma stem-like cells, tightly associated with the tumor vasculature and activates the Notch pathway within these stem-like cells (Figure S8).
The tumor perivascular niche is a complex microenvironment with multiple cell types and signaling factors involved in the cross-talk between endothelial cells and stem-like cells residing within the niche. The contribution of aberrant NO signaling within the niche is not the only component involved in the cross-talk between the tumor endothelium and perivascular stem-like cells. In vivo, glioma stem-like cells are likely influenced by the convergence of signals from neighboring cells within the perivascular niche. In addition to glioma stem-like Nestin-expressing cells, there are other cell types known to reside within the brain tumor perivascular niche such as reactive astrocytes that mediate Shh signaling (Becher et al. 2008) known to promote stem cell renewal in gliomas (Clement et al. 2007), and pericytes (Hambardzumyan et al. 2008a) that likely contribute either through direct cellular interactions or in a paracrine fashion to regulate the activity of glioma stem-like cells.
It is worth noting that transient activation of the NO/cGMP pathway by GSNO or Pet-cGMP activates Notch signaling and increases the percentage of SP cells in PIGPC within 2–2.5 hours. These cells then show long lasting fundamental changes in their behavior as demonstrated by their increased capacity for neural stem cell formation 2 weeks after treatment withdrawal, and increased capacity for tumor formation more than 4 weeks after tumor cell transplantation. This effect might involve the conversion of a subpopulation of glioma cells within our primary cultures, with an inherent capacity to be primed by NO/cGMP signaling to a more stem cell-like state. The data strengthens the argument that stem cell character can be acquired by specific signaling from the tumor microenvironment. As the RCAS/tv-a PDGF mouse model mimics the PDGFR subtype of human gliomas, this observation may be limited to this subgroup of human gliomas. We have identified PKG as a critical mediator of NO-induced Notch pathway activation in PDGF-induced mouse gliomas, however, the mechanism through which PKG activates the Notch pathway remains to be determined.
This additional role for NO in promoting gliomagenesis through induction of stem-like activity complements its established role in tumor angiogenesis and underscores the emerging role of the glioma perivascular niche as a direct participant in the disease process. This work also highlights several potential therapeutic targets such as inhibitors of endothelial nitric oxide synthase (eNOS), PKG and Notch. Although these effects were ascribed to inhibition of eNOS induced angiogenesis, it is possible that these therapies additionally modified stem-like characteristics within these tumors. In addition, the use of anti-angiogenic agents that disrupt these aberrant glioma vascular stem cell-like niches could potentially sensitize tumors to conventional treatments by depriving the stem-like cells of niche signals, such as NO, required for their maintenance. Such synergy was previously demonstrated with bevacizumab, which after treatment of glioma-bearing mice, resulted in depletion of tumor blood vessels and a significant reduction of tumor stem cells (Calabrese et al. 2007). Since anti-angiogenics, such as, endostatin and thrombospondin exert their effects through inhibition of eNOS (Fukumura et al. 2006), the effect observed with bevacizumab, could also involve modifications to stem-like character within these tumors. Targeting of the Notch pathway was demonstrated to be an effective anti-tumor strategy in brain tumor preclinical studies (Fan et al. 2006). Pharmacologic suppression of the Notch pathway in medulloblastoma cells significantly decreased their cancer stem-cell populations, as indicated by SP analysis, CD133, and Nestin expression, and further diminished their capacity to form tumors following transplantation in mice. These cells were also more sensitized to apoptosis than more differentiated cells (Fan et al. 2006). This finding supports the notion that targeting the Notch pathway in gliomas could potentially sensitize the stem cell-like fractions to apoptosis. The prediction from the preponderance of the literature is that these approaches might be more effective when combined with radiation and chemotherapy where perivascular resistant cells contribute to tumor recurrence after conventional therapy.
PDGF-induced glioma primary cultures (PIGPC) were suspended at 1×106 cells/ml in 15ml falcon tubes in DMEM+10FBS medium, then preincubated at 37°C for 35 minutes in the presence of vehicle (water), GSNO (100µM) (Sigma Aldrich, USA) Pet-Br-cGMP (200µM) (Axxora Platform, San Deigo, CA), Fumitremorgin C (FTC), an ABCG2 ihibitor (5µM) (Axxora Platform, San Deigo, CA) or the gamma secretase inhibitor (GSI)-MK-003 (3µM) a kind donation from Merck, USA. Cells were then incubated with Hoechst 33342 (5µg/ml) for 90 minutes at 37°C with periodic shaking. Following Hoechst staining, cells were incubated on ice for 10 minutes then washed 2x in ice cold PBS. Hoechst dye was excited at 407nm using a trigon violet laser and dual wavelength detection performed using 450/40 (Hoechst 33342-Blue) and 695/40 (Hoechst 33342-Red) filters. Dead cell exclusion in control, GSNO, pet-cGMP and GSI treated glioma primary cultures occurred by forward and side scatter gating and the exclusion of PI positive populations. Data were analyzed using FlowJo (Ashland, OR)
Neonatal Nestin tv-a (N-tva) INK4A/Arf−/−mice described previously (Uhrbom et al. 2005) were injected intracranially with DF1 chicken fibroblasts producing RCAS-PDGF retroviral particles to generate gliomas as described previously (Shih et al. 2004). PDGF-induced gliomas were dissected and enzymatically digested for 15 minutes in 1x Earles balanced salt solution containing 12% papain (Worthington, Lakewood, NJ) and 10µg/ml DNase at 37°C. The digestion was stopped with 1mg/ml ovomucoid (Worthington, Lakewood, NJ). Cells were washed and resuspended 3x in basal medium. Single cell suspensions were plated in DMEM+10% FBS and cells were grown as a monolayer. The medium was replaced 24–48hrs later. Primary culture experiments were performed using PDGF-induced tumors from different mice, however control and treated groups compared in all experiments, were generated from the same tumor. Primary cultures used in all experiments were performed using 10cm cultures dishes at 50–60% confluence. In addition all PIGPC were used between passage numbers p0-p4.
To assess the tumorigenic capacity of PIGPC following GSNO treatment approximately 8 × 104 vehicle or GSNO treated (2 hours) PIGPC were injected into the cortex of newborn pups. Tumor development occurred between 8–14 weeks.
Injection of human cells, were done by stereotactic injection. NOD-SCID mice between 5–10 weeks old were anaesthetized by i.p. injection of ketamine (0.1mg/g) and xylazine (0.02mg/g). 1µL of 2.5 × 105 cells were injected into the cortex. Stereotactic coordinates were taken relative to bregma, right hemisphere: AP-0mm from bregma, Lat-3.0mm (left or right) depth-1.0mm from dural surface. Locations were determined using mouse atlas (Franklin K 2007). Tumor cells were injected using a Hamilton syringe. Mice were closely monitored for tumor development as assessed by the presence of hydrocephalus, lethargy or cachexia.
Comparisons between two groups were made using two sided t-tests. Chi square test was use to compare groups in Kaplan Meier graphs. Two-way ANOVA was used to analyze data from luciferase assays. Fisher’s Exact Test was used to compare tumorigenicity between vehicle and pet-cGMP treated primary glioma cells. Data represents the mean of three experiments unless otherwise noted. P values of < 0.05 were considered statistically significant.
We thank Jim Finney and Qunchao Zhang for technical assistance; Alicia Pedraza-Fernandez for providing human tumor neurospheres for analyses. We extend thanks to Rebecca Bish for critical reading and feedback of this manuscript. Sincere thanks to the flow cytometry core facility for acquisition of flow cytometry data. This work was supported by NIH grant U54 CA126518.
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.
The supplemental data includes six supplemental figures and all additional experimental procedures and can be found with this article online at