|Home | About | Journals | Submit | Contact Us | Français|
Primary astrocytomas of World Health Organization grade 3 and grade 4 (HG-astrocytomas) are preponderant among adults and are almost invariably fatal despite multimodal therapy. Here, we show that the juvenile brain has an endogenous defense mechanism against HG-astrocytomas. Neural precursor cells (NPCs) migrate to HG-astrocytomas, reduce glioma expansion and prolong survival by releasing a group of fatty acid ethanolamides that have agonistic activity on the vanilloid receptor (transient receptor potential vanilloid subfamily member-1; TRPV1). TRPV1 expression is higher in HG-astrocytomas than in tumor-free brain and TRPV1 stimulation triggers tumor cell death via the activating transcription factor-3 (ATF3) controlled branch of the ER stress pathway. The anti-tumorigenic response of NPCs is lost with aging. NPC-mediated tumor suppression can be mimicked in the adult brain by systemic administration of the synthetic vanilloid Arvanil, suggesting that TRPV1 agonists hold potential as new HG-astrocytoma therapeutics.
Somatic mutant neural stem and precursor cells (NPCs) are thought to be the source for high-grade astrocytomas (HG-astrocytomas), one of the most aggressive forms of CNS tumors1. HG-astrocytomas and glioblastomas (GBM) are much more frequent in adults than in children2,3. However, adult neurogenesis, that is the presence and activity of NPCs in the postnatal and adult brain, is maintained at high rate only until puberty and declines thereafter4,5. Hence, the epidemiology of glioblastomas and the timing of adult neurogenesis are inversely correlated and glioblastomas are usually diagnosed several decades after the decline in brain stem cell activity.
It has been previously found that endogenous and exogenous NPCs have a strong tropism for primary brain tumors and that NPCs can release tumor-suppressive factors6–13. However, the molecular nature of the factors that mediate cell death in HG-astrocytoma cells has not been identified. We show that HG-astrocytoma associated NPCs induce tumor cell death via the release of endovanilloids. Endovanilloids14 like arachidonoyl-ethanolamide (AEA) and N-arachidonoyl-dopamine (NADA) directly stimulate the vanilloid receptor (TRPV1; transient receptor potential vanilloid-1)15. Synergistic TRPV1 activation by AEA in combination with other fatty acid ethanolamides such as oleoyl-ethanolamide (OEA) or palmitoyl-ethanolamide (PEA) is also observed15,16. TRPV1 is a non-selective cation channel that is best characterized in capsaicin-sensitive sensory neurons of the dorsal root and trigeminal ganglia16. The physiological role of the ion channel in non-neural tissues is largely unexplored16.
We investigated the signaling pathways that are activated in HG-astrocytoma cells after exposure to NPC conditioned medium (NPC-CM). Analysis of gene expression changes together with pharmacological and molecular studies (for microarray data, please refer to Gene expression omnibus repository; GSE37671) suggested a role for NPC-derived endovanilloids as tumor suppressors. Furthermore, we observed that TRPV1 expression positively correlates with grading in human primary brain tumors whereas little TRPV1 is detected in human tumor-free brain tissue; data from real-time PCR (Supplementary Fig. 1a) were supported by immunohistochemistry on tissue panels (not shown).
Next, we explored the role of the endovanilloid system in NPC-mediated HG-astrocytoma suppression in an established mouse model8,9,17. Here, we induced orthotopic brain tumors in Nestin-GFP mice, which are a model for the visualization of NPCs8,18 (Fig. 1a). Subventricular NPCs migrated to HG-astrocytomas that were located in the caudate putamen8,9,17. NPCs were identified by co-localization of GFP with established immunocytochemical markers such as PSA-Ncam (Fig. 1b) or Musashi8,19,20. PSA-NCAM is also a marker for tumor-associated NPCs in humans (S. Momma, personal communication). Importantly, we found that mouse HG-astrocytomas express high levels of TRPV1 (Fig. 1c); i.e. TRPV1 levels in tumors were higher than in tumor-free brain, while only a small fraction of tumor associated Nestin-GFP+ cells expressed TRPV1; TRPV1 was absent from subventricular NPCs (Supplementary Fig. 1b).
In a series of in vitro experiments, we found that factors released from mouse NPCs (mNPC-CM), but not from their fully differentiated progeny (i.e. astrocytes, oligodendrocytes and neurons) or from fibroblasts (scrc), strongly reduced the viability of mouse HG-astrocytoma cells over a time-course of three days (Fig. 1d). In subsequent experiments HG-astrocytoma cells were always stimulated for three days, unless indicated otherwise. mNPC-CM reduced tumor cell viability by inducing cell death, as indicated by TUNEL- and cytotoxicity-assays (Supplementary Fig. 1c). Cytotoxicity values are given as percent of fully permeabilized cells (raw data are presented in Supplementary Fig. 2 and 3).
GL261 HG-astrocytomas express TRPV1 and contain specific binding sites for a selective TRPV1 ligand (Supplementary Fig. 1d,e). Importantly, mNPC-CM induced HG-astrocytoma cell death was greatly reduced by blocking TRPV1 with the selective antagonists16 iodo-resiniferatoxin (I-RTX, 10 nM; Fig. 1e) or capsazepine (CZP, 1 μM; Fig. 1e) and by TRPV1-knock down (TRPV1-KD; Fig. 1e). We verified TRPV1-KD efficiency (Supplementary Fig. 1d,e) and specificity by performing experiments with control shRNAs (scrambled) and over-expression of a knock-down resistant form of mouse-TRPV121 in TRPV1-KD tumor cells (rescue; Fig. 1e and and2f;,2f;, Supplementary Fig. 2c). mNPC-CM strongly induced cell death in different HG-astrocytoma cell lines (Fig. 1f) and a range of primary human glioblastomas (Fig. 1g), an effect that was always blocked by co-application of either I-RTX (Supplementary Fig. 2b, 3a and 3b) or CZP. Importantly, conditioned medium from human NPCs (human NPC-CM) also induced cell death in primary human glioblastomas after TRPV1 stimulation (Fig. 1h).
Overall, we observed that TRPV1 levels in HG-astrocytomas are much higher than in normal brain. NPCs migrate to brain tumors in vivo. In vitro experiments showed that human and mouse NPCs release endovanilloids (TRPV1 agonists) which induce HG-astrocytoma cell death.
We quantified the concentration of AEA, NADA, OEA, PEA and the endocannabinoid arachidonoyl-glycerol (2-AG) in samples from mNPCs, fully-differentiated progeny from mNPCs and mouse HG-astrocytoma cells: mNPCs contained considerably high amounts of AEA, PEA and OEA; the endovanilloid levels in differentiated mNPCs or mouse HG-astrocytoma cells were much lower (Fig. 2a); NADA was not detectable in any sample (not shown). In mNPC-CM we found high concentrations of AEA (Fig. 2b), whereas culture supernatants from differentiated mNPCs or HG-astrocytoma cells contained much less AEA; other lipids were at the detection limit. Importantly, we observed that combined application of synthetic AEA, PEA and OEA exerts a cooperative effect on HG-astrocytoma cell death22, which was blocked by I-RTX (Fig. 2c). Addition of fatty acid amide hydrolase (FAAH; which degrades ethanolamides23) fully abolished the cell death inducing effect of mNPC-CM (Fig. 2d) and of human NPC-CM (not shown).
The endocannabinoid 2-AG was detected in HG-astrocytoma cells at low levels, and was even less abundant in mNPCs (Fig. 2a) suggesting a role for endovanilloids (rather than endocannabinoids), as mediators of NPC-induced HG-astrocytoma cell death22. Consistently, addition of synthetic NADA to non-conditioned medium induced HG-astrocytoma cell death, whereas cannabinoid receptor blockade did not interfere with mNPC-CM induced cell death (Supplementary Figs. 2f, 2g, 3c and 3d). Next, we used FAAH-deficient mice (Faah−/−), which have largely increased amounts of endocannabinoids/endovanilloids in the CNS24, as a brain tumor model. Here, we investigated the effect of TRPV1 signaling on glioma growth (by implanting TRPV1-KD or control gliomas). We found that implantation of TRPV1 knock-down tumors resulted in much larger tumors compared to implantation of control tumors in Faah−/− animals (Fig. 2e). These data support our finding from in vitro experiments and show that even largely elevated levels of endogenous endocannabinoids/endovanilloids exert their tumor suppressive effects exclusively via TRPV1 receptors in our glioma model.
To assess the activity of NPC-released endovanilloids, we set-up a bioassay using dorsal root ganglion neurons (DRGs)25 from wild-type and Trpv1−/− mice. DRGs were stimulated with mNPC-CM and responses were measured with Fura-2 based calcium-imaging. In wild-type mice, mNPC-CM induced Ca2+ responses in 5.4% of DRGs, which were also capsaicin sensitive, whereas only 0.85% responded to mNPC-CM alone. Interestingly, in Trpv1−/− mice, only 0.91% of DRGs responded to mNPC-CM, which would suggest that TRPV1 is required for the majority of the responses to mNPC-CM observed in wild-type mouse DRG neurons (Supplementary Fig. 4a). Also, RT-PCR analysis of cultured mNPCs, their differentiated progeny and of whole brain extracts revealed that mNPCs express the major receptors and metabolic enzymes of the endo-vanilloid and -cannabinoid pathways (Supplementary Fig. 4b). These data support the view that AEA is a major constituent of mNPC-CM and that AEA is released from NPCs in physiologically relevant concentrations.
Furthermore, mNPC-CM stimulated TRPV1 mediated Ca2+ responses (single time-point ratiometric measurement of a bulk Ca2+ response in Fig. 2f; see also Supplementary Fig. 4c) in primary human glioblastoma cultures or human, rat and mouse HG-astrocytoma cell lines; capsaicin mediated Ca2+ signaling in mouse glioma cells (Supplementary Fig. 5a). We observed the ectopic expression of TRPV1 in the endoplasmic reticulum (ER; Supplementary Fig. 5b–e). The lower capsaicin-sensitivity of ER-located TRPV1 has been previously described26, which may explain why a higher concentrations of capsaicin was required to induce cell death in HG-astrocytoma cells in previous studies16.
Overall, we have shown that mNPCs constitutively synthesize and release endovanilloids, which induce Ca2+ responses and cell death in primary human glioblastomas and a range of HG-astrocytoma cell lines. Our data indicate that mouse and human NPCs use the same pathway for paracrine tumor suppression since addition of active FAAH or TRPV1 antagonists to NPC-CM of both species equally blocked HG-astrocytoma cell death.
We investigated the gene expression pattern in mouse tumor cells after incubation with non-conditioned medium (controls) or mNPC-CM by microarrays (GSE37671). We found that ER stress genes like the activating transcription factor-3 (ATF3) were robustly upregulated in mNPC-CM treated mouse HG-astrocytoma cells, compared to controls. Immunocytochemical labeling and reporter gene assays in mNPC-CM treated GL261 cells showed that ATF3-expression is increased both in the cytoplasm and nucleus (versus controls) and regulates ATF3-responsive genes (Fig. 3a,b). Forced expression of ATF3 reduced the number of GL261 cells in culture and increased the number of TUNEL+ tumor cells (Fig. 3c). Importantly, siRNA-mediated down-regulation of ATF3 expression (Supplementary Fig. 6a) in mouse HG-astrocytoma cells prevented mNPC-CM-induced tumor cell death (Fig. 3d). Hence, ATF3 is necessary and sufficient for mediating NPC-CM induced HG-astrocytoma cell death. The TRPV1 antagonist CZP blocked the mNPC-CM induced activation of the ATF3-dependent ER stress pathway in mouse HG-astrocytomas (Supplementary Fig. 6b,c).
Electron microscopy revealed that mNPC-CM treated GL261 cells had an enlarged ER as compared to controls (Fig. 4a; Supplementary Fig. 6d), which is a morphological hallmark of ER stress27. The effect of mNPC-CM (with or without CZP) on ER was quantified in primary human glioblastoma or human, rat and mouse HG-astrocytoma cell lines using ER-tracker (Fig. 4b,c). Likewise, the effect of human NPC-CM on ER-size in primary human glioblastomas was determined (Fig. 4d). In all HG-astrocytoma cells studied we detected a very robust increase in relative ER-size after stimulation with human or mouse NPC-CM, which was always attenuated by CZP (Fig. 4b–d; Supplementary Fig. 7). Additionally, we used synthetic AEA and ER stress inducers like tunicamycin or thapsigargin28 at concentrations that were sub-threshold for ER stress induction when applied alone. We found that the co-application of AEA plus tunicamycin or thapsigargin led to strong increases in relative ER-size in GL261 cells (Fig. 4e). Importantly, the combined substances had a clear cooperative effect on the rise in relative ER-size, confirming that vanilloid-induced signaling and ER stress are part of the same signal transduction pathway in HG-astrocytoma cells. In summary, these data show that human and mouse NPC-derived endovanilloids induce HG-astrocytoma cell death via the ER stress pathway (see cartoon in Fig. 4f).
To investigate if NPC-derived endovanilloids can suppress HG-astrocytomas in vivo, we performed orthotopic implantation of HG-astrocytomas into Nestin-GFP mice8,9,17. Inoculation of GL261 cells into young (30 day old) mice resulted in the association of many endogenous NPCs with the tumor8,9,17 (Fig. 5a). Nestin-GFP+ NPCs accumulated at GL261 controls and TRPV1-KD tumors in equal density. Strikingly, we found that young mice injected with TRPV1-KD cells had significantly larger tumors, compared to controls (70% bigger; Fig. 5b). No difference in tumor size was detected in adult mice (90 days old; Fig. 5b). Furthermore, we used a previously established mouse model to investigate the anti-tumor effect of endogenous NPCs. We orthotopically implanted GL261 cells (control or TRPV1-KD) into cyclin-D2 knockout mice (Ccnd2−/−, which have largely reduced adult neurogenesis9,29) or into their wild-type littermates. We found that tumor size in wild-type animals receiving HG-astrocytoma controls was at least 63% smaller as compared to wild-type mice receiving TRPV1-KD tumors cells or Ccnd2−/−receiving either HG-astrocytoma controls or TRPV1-KD tumor cells (Fig. 5c). Next, we measured cell death in vivo by systemically delivering propidium-iodide (Fig. 5d)30. We noted that TRPV1-KD largely protected HG-astrocytoma cells from death.
In another set of experiments we tested the impact of NPC-released endovanilloids on the overall survival of a cohort of mice with HG-astrocytomas. Firstly, we orthotopically implanted GL261 cells (control or TRPV1-KD) into young mice and compared the cumulative survival. We observed that young wild-type tumor bearing mice (Fig. 5e) significantly outlived the older mice, unless the young mice were implanted with TRPV1-KD cells. These data show that younger mice have an intrinsic protective mechanism against HG-astrocytomas, which is dependent on endovanilloid signaling. In a second study we investigated if the survival promoting effect could be attributed specifically to NPCs. Therefore, we co-implanted adult mice with exogenously cultivated NPCs and HG-astrocytoma controls or TRPV1-KD tumor cells (Fig. 5f). We found that co-implantation of NPCs together with HG-astrocytoma controls in adult mice promoted survival, compared to injection of HG-astrocytoma control cells alone (compare Fig. 5e and 5f). Importantly, we also noticed that the survival promoting effect of NPCs in adult mice was absent after co-implantation with TRPV1-KD cells (Fig. 5f).
In summary, our study suggests that NPCs release endovanilloids in vivo, in a similar way as demonstrated for NPCs in vitro. Consistently, the extent of the NPC-mediated anti-tumor response depended on the level of adult neurogenesis.
We investigated the therapeutic potential of a synthetic, non-pungent, blood-brain-barrier permeable vanilloid named Arvanil31,32. In organotypic brain slice cultures HG-astrocytomas were allowed to develop for 5 days. Addition of Temozolomide33 (200 μM; the current standard of care for the treatment of glioblastoma patients) or Arvanil (50 nM) strongly reduced HG-astrocytoma size as compared to controls (Fig. 6a,b). Furthermore, Arvanil induced a TRPV1 dependent Ca2+ signal and TRPV1 dependent cell death in HG-astrocytomas (Supplementary Fig. 8). In further experiments we implanted TRPV1-KD or HG-astrocytoma controls and treated both groups with Arvanil as described above. We observed a significantly improved survival time in the control group compared to mice receiving TRPV1-KD tumor cells (Fig. 6c), suggesting that Arvanil elicits its therapeutic effect as a TRPV1 agonist. To determine if Arvanil would also increase survival in other HG-astrocytoma models, we implanted primary human glioblastoma cells (GBM1 and GBM2) into immune-compromised (Scid) mice. After one week, we examined tumor development and administered Arvanil31 (a total of four i.p. injections with 1mg kg−1) or vehicle. Strikingly, Arvanil-treatment robustly prolonged survival as compared to the vehicle-treated controls (Fig. 6d–f). Finally, we compared the effects of application of Arvanil and Temozolomide on survival after implantation of a third primary human HG-astrocytoma culture used in the present study (GBM3). We found that Arvanil prolonged survival in a cohort of immune-deficient mice that received HG-astrocytoma cells that did not respond to Temozolomide (given34 once daily for 5 days at 100 mg kg−1; Fig. 6f). These data show the potential clinical value of an experimental HG-astrocytoma therapy using vanilloids, which may also offer a new therapeutic option for Temozolomide resistant HG-astrocytomas35.
We have shown that HG-astrocytomas express high levels of TRPV1 and that TRPV1 stimulation induces tumor cell death. Neural stem and precursor cells home in on HG-astrocytomas and release anti-tumorigenic TRPV1 agonists (i.e. fatty acid ethanolamides). Endogenous and exogenous NPCs show extensive tropism for brain tumors6–10. However, the number of endogenous NPCs accumulating at HG-astrocytomas depends on the proliferative activity in the stem cell niche and declines before the onset of adulthood9. Hence, the recruitment of large numbers of NPCs to a tumor, and concomitantly the anti-tumorigenic release of endovanilloids, is restricted to the young brain. Additionally, other age-related changes in neural stem cell physiology may also impinge on the capacity for NPC-mediated tumor suppression36,37.
We demonstrated that NPCs are a primary source of endogenous TRPV1 and cannabinoid receptor agonists like AEA15,38. This is substantiated by the detection of high amounts of AEA and related acyl-ethanolamides in undifferentiated NPCs, by the finding that NPC-released factors evoke TRPV1 dependent Ca2+ responses in DRGs and HG-astrocytomas, that the tumor suppressive effect of NPC-CM is lost after addition of FAAH and that NPC-induced HG-astrocytoma cell death is TRPV1 dependent in vitro and in vivo. These data are in agreement with previous reports indicating that synthetic AEA induces HG-astrocytoma cell death39.
A role for TRP-channels in tumor suppression was previously suggested by us and others40–44, but the present study is the first to identify NPCs as a cellular source for tumor suppressive endovanilloids and to uncover the role of NPC-released TRPV1-agonists and -modulators on HG-astrocytoma cell death. Overall, our study suggests that endovanilloids are intrinsic tumor suppressors in the brain and that synthetic vanilloid compounds may have clinical potential for brain tumor treatment.
Animal experiments were carried out in compliance with the German laws on animal welfare, and the animal protocols were approved by the Landesamt für Gesundheit und Soziales (LaGeSo) in Berlin. Wild type C57BL/6 mice, Nestin-GFP 18 mice, Trpv1−/−mice 45, Ccnd2−/− 29, Faah−/− 24 and Scid mice (B6.CB17-Prkdcscid/SzJ; Charles River Breeding Laboratories; Schöneiche, Germany) were housed with a 12 h light/dark cycle and received food ad libitum.
Normal SVZ specimens were obtained as part of planned resections during anterior temporal lobectomy for the treatment of intractable epilepsy from mesial temporal sclerosis. We obtained the ethical approval (given by the ethics committee of Charité university clinics; license numbers EA112/2001, EA3/023/06 and EA2/101/08. Tumor samples were obtained from (otherwise untreated) primary glioblastomas, according to governmental and internal (Charité) rules and regulations; cDNA samples and tissue arrays from human brain tumors and from tumor-free brain were obtained from OriGene.
All glioblastoma cells were maintained as described for neurospheres46. Mouse, rat, human HG-astrocytoma cell lines and 293T cells were obtained from the National Cancer Institute, NCI-Frederick and from ATCC. Mouse NPCs were gained from SVZ; dorsal root ganglia (DRG) neurons were prepared from both wild-type and Trpv1−/− adult mice as described previously25.
The pLKO.1 shRNA vector was from BioCat. The validity of the shRNA mediated knockdown was affirmed on the protein level, by Western-blotting and FACS analysis, as described47 and on the functional level. The TRPV1 rescue construct was mutated in the seed-region of the shRNA knock-down construct21.
CytoTox-Fluor™ cytotoxicity assays (Promega) were measured (485nm/520nm) with the fluorometer (TECAN).
TUNEL+ cells were quantified using the DELFIA cell-based fragmentation assay (PerkinElmer).
cDNA microarrays48 were generated using ~20,000 murine cDNA clones (arrayTAG clone collection) from LION Bioscience, six arrays were used in total. Image acquisition and data analysis was done as described48.
Lipids were purified by open-bed chromatography on silica gel and AEA, 2 AG, PEA, OEA and NADA were analyzed by isotope dilution-liquid chromatography/atmospheric pressure chemical ionization/mass spectrometry49–51.
Peptides for the development of a SRM method (selected reaction monitoring) were selected. Cells were lysed, protease digested, purified, separated by HPLC and electrosprayed into the mass spectrometer (ABSciex Q-TRAP 4000). For the data analysis the MultiQuant (ABSciex) and the R-software packages (www.R-project.org) were used52.
Cells were loaded with Fura-2 acetoxymethyl ester (TEF-Lab), excited at 340 and 380 nm and imaged with a 510 nm long-pass filter; the results are presented as the ratio between the emission signals acquired using the two excitation wavelengths.
qRT-PCR was performed on the iCycler IQ 5 multicolor real-time detection system (Bio-Rad), using absolute SYBR green fluorescein (ABgene). Oligonucleotides were purchased from Invitrogen.
Membranes were incubated with specific antibodies and Western blots were developed using the chemiluminescence method (GE-Healthcare).
All stainings and microscopy for NPC- and HG-astrocytomas markers was carried out as described previously53.
For ER- visualization, ultrathin cryosections (70 nm) of fixed HG-astrocytoma cells were contrasted, stabilized54 and examined with a Zeiss 910 electron microscope. For preembedding immunogold labeling, HG-astrocytoma cells were fixed in 4% paraformaldehyde and 1% glutaraldehyde and incubated with TRPV1 antibody.
In every 12th axial section we sampled the area that was primarily infiltrated by the tumor in an unbiased approach using the optical fractionator procedure (StereoInvestigator, MicroBrightField Inc.). Tumor volume was quantified according to the Cavalieri principle.
Survival statistics were analyzed using MatLab software (Natick). Bar diagrams are shown as mean values ± standard deviation of the mean. Comparisons among the groups were performed with the Student’s t test, Fishers exact test and the Wilcoxon rank test (as indicated).
We thank G. Gargiulo, O. Daumke, and J. Kurreck for discussion of the manuscript, S. Kitajima for ATF3 constructs, L. Kaczmarek and M. Szymanska for help with the Ccnd2/− mice. Funding by Helios-Clinics (HeFoFö-ID1148) and by US National Institute of Health (DA-009789 to VDM) is gratefully acknowledged.
Author contributions:SP, RI, AL, LDP, UG and SRC designed the experiments, conducted the experiments and interpreted the data; KS, JK, EStJS, PW, BP, UAN, VDM, JHW, GD and LC additionally contributed to manuscript preparation. GRL, VDM and HK designed the experiments, supervised the project, interpreted the data and contributed to manuscript preparation. MS performed brain tumor resections and provided tumor samples. MS and RG designed the experiments, conducted the experiments, supervised the project, interpreted the data and prepared the manuscript.