Background
Medulloblastoma is largely a cancer of children, with 75-80% of cases diagnosed in individuals younger than fifteen years; some are diagnosed in infancy [
1‐
3]. It is a very aggressive and invasive cancer which spreads primarily via cerebral spinal fluid to metastasize anywhere in the leptomeninges, or, in advanced disease, hematogenously to invade any body part. It is suspected to arise from cerebellar granule cell precursors [
1,
4] based on its primitive neuronal histology and location in the midline posterior fossa. Survival is achievable in many children, dependent on a number of factors, yet recurrence holds a dismal prognosis [
3]. Current understanding of the biology of medulloblastoma cannot fully provide an explanation for medulloblastoma occurrence, proliferative properties, migratory activity, or chemotherapy resistance.
Prognosis has improved over the last half century with the addition of radiation therapy and chemotherapy. In spite of these advances, there remains a considerable unmet need to increase survival rates, especially in high-risk disease. Further, targeted therapies need to be identified such that normal developing brain tissue will be spared, thereby avoiding the disabling sequelae which are, unfortunately, commonplace in survivors. These goals are more likely to be achieved through better understanding of the biology of the disease and exploiting features unique to the tumor rather than by the current strategy of damaging tumor cells more than normal cells.
Despite comprehensive studies to identify risk factors associated with medulloblastoma, no environmental risk factor has been linked to development of medulloblastoma. A variety of chromosomal abnormalities have been reported, and defects in signaling pathways such as Wingless (Wnt) and sonic hedgehog (SHH) have been identified in some sporadic and heritable forms of medulloblastoma [
1,
3,
5], but these represent a minority of cases.
Current research pertaining to cancer and immune response has revealed an association between nuclear factor kappa B (NFκB) signaling and tumorigenesis. Over the past 25 years, NFκB has been described and characterized through a wide range of normal and pathologic model systems [
6]. NFκB is a family of transcription factors that regulate genes involved in cell growth, apoptotic cell death, adhesion and angiogenesis. Although only the related viral oncogene v-rel is acutely transforming, growing evidence implicates nearly all members of the NFκB family in human malignancy [
7,
8]. Chromosomal abnormalities within the genes of these transcription factors are found in many solid and hematopoietic tumors. Also, many cancers have mutations affecting the activity of upstream regulators [
6]. Moreover, many forms of leukemia and a wide variety of solid tumors demonstrate constitutive activation of this otherwise tightly regulated pathway [
9] by increasing pathway stimulation or by inactivating negative feedback molecules [
9,
10]. Some of the most aggressive malignancies of childhood, including neuroblastoma, rhabdomyosarcoma, Wilms tumor, and retinoblastoma have also been reported to involve NFκB.
NFκB is normally quiescent in cells and only becomes activated in response to stress signals such as pro-inflammatory cytokines. In the canonical NFκB pathway, a heterodimer consisting of two subunits, p65 (RelA) and p50, is sequestered in the cytoplasm by an inhibitory protein, IκB-α, under normal physiological conditions. Upon induction of the pathway, the inhibitor IκB-α becomes phosphorylated by an IκB kinase (IKK), which leads to ubiquitination of IκB. The inhibitor is thus targeted to the ubiquitin-proteasome for degradation, and the NFκB heterodimer p65/p50 is free to translocate to the nucleus and begin altering gene expression [
10,
11]. Some specific downstream NFκB targets include survival genes Bcl-xl and XIAP; adhesion molecules ICAM-1, VCAM-1 and ELAM-1; metastasis promoting gene MMP-9; angiogenesis factors VEGF and TNFα; proliferation genes cyclin D1 and C-myc; and a host of other genes known for their association with proliferation, inflammation, and immortalization of cells [
6,
10,
12].
Because of the prominence NFκB signaling has gained in adult cancer research and evidence of its activity in some high grade pediatric malignancies, we have asked whether this pathway could be a dominant feature of medulloblastoma. Our studies have established that p65 is indeed over expressed in primary tumor samples and in tumor cell lines. Furthermore, multiple drugs that inhibit NFκB cause apoptotic cell death in all four cell lines tested. Finally, disruption of NFκB signaling using a dominant negative variant of the endogenous inhibitor of NFκB, dnIκB, resulted in reduced xenograft tumor growth. We have, therefore, established that NFκB plays a role in medulloblastoma and that it may be a target for therapeutic intervention.
Methods
Chemicals
All chemicals were obtained from Sigma-Aldrich, St. Louis, MO unless otherwise indicated. Stocks were prepared as follows: curcumin, 10 mM in ethanol; bortezomib (LC laboratories, Woburn, MA), 200 mg/mL in DMSO, then diluted to 10 mM in phosphate buffered saline (PBS) pH 7.4; pyrrolidine dithiocarbamate ammonium salt (PDTC), 10 mM in PBS; diethyldithiocarbamate sodium salt (DDTC), 10 mM in PBS; sulfasalazine (SAS) 10 mM in 0.1 M saline, pH7.4; doxycycline, 10 μg/mL in sterile water; TNFα (R&D Systems, Minneapolis, MD), 100 μg/mL in sterile PBS+0.1% bovine serum albumin.
Cell Culture
Experiments using cultured medulloblastoma cells were performed on two commercially available cell lines, Daoy and D283, (American Type Culture Collection, Manassas, VA) and two cell lines, D425 and D458 [
13,
14], established from primary medulloblastomas (kindly provided by Dr. D. Bigner, Duke University, Durham, NC). Cell cultures were maintained in MEMα supplemented with 2 mM
L-glutamine (Mediatech, Manassas, VA) and 10% characterized fetal bovine serum (FBS) (HyClone, Logan, UT), or in Richter's improved MEM Zinc Option containing 10 mM HEPES and 0.22% sodium bicarbonate (Life Technologies), 2 mM
L-glutamine, and 10% FBS. U-87MG Grade III glioma cell line (kindly provided by Dr. D. Bigner, Duke University, Durham, NC) was grown in Dulbecco's Modification of Eagle's Medium/F-12 supplemented with 7% FBS and 2 mM
L-glutamine (Mediatech). Mouse neurospheres were derived from embryonic day 14 mouse cortex (StemCell Technologies Inc, British Columbia, Canada) and were maintained in 90% NeuroCult NSC Basal Medium with 10% NeuroCult NSC Proliferation Supplements (StemCell Technologies); 20 ng/mL rhEGF was added just before use. All medulloblastoma cells and neurospheres were maintained at 37°C in a 95% O
2-5% CO
2 humidified atmosphere; U-87MG cells were incubated in 90%O
2-10%CO
2.
Proliferation assays
Cells were seeded in 24-well dishes at densities that had been determined to allow for exponential growth for the duration of the experiment. For the time course, cells were left untreated or were treated for 3 days at the IC90 for each line (325 nM PDTC for Daoy and 350 nM for D425). Cells were counted using a Beckman Multisizer 3 Coulter counter. For dose-response experiments, cells were grown for 3-4 days in the absence or presence of varying concentrations of each drug and counted as above.
Annexin-V staining
Apoptosis was measured with annexin V-Cy3 staining, following manufacturer's instructions (BioVision Inc, Mountain View, CA). At least 100 cells were counted for each experiment. Positively staining cells in each low power field were identified by fluorescence microscopy. The number of positive cells was divided by the total number of cells in those fields and reported as % positive for annexin V.
Tissue
Human autopsy tissue was obtained from the University of Alabama at Birmingham (UAB) Tissue Collection and Banking Facility, Cooperative Human Tissue Network Southern Division. Smo/Smo transgenic medulloblastoma mice with constitutive expression of
Smoothened in cerebellar granule neurons [
4] were generously provided by Dr. Jim Olson, Fred Hutchinson Cancer Research Center, Seattle, WA. All studies were performed in accordance with standards of the UAB Institutional Review Board (IRBN080731009, X070829010).
Whole cell or total tissue lysates
Cell pellets collected from 10 cm culture dishes were rinsed in PBS and lysed in RIPA (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% IGEPAL detergent, 0.1% SDS, 0.5% Na+ deoxycholate) containing 1X final concentration protease and phosphatase inhibitors (Sigma-Aldrich). Tissues (10-30 mg) were homogenized on ice in lysis buffer (100 mM Tris-HCl pH 7.4, 1% SDS) containing protease and phosphatase inhibitors. Samples were sonicated briefly, incubated at 4°C with occasional agitation for 1 hour, then clarified by centrifugation at 14,000 X g for 30 minutes at 4°C. Proteins were quantified using the modified detergent-compatible Lowry assay (BioRad Laboratories, Hercules, CA).
Nuclear and Cytoplasmic extracts
Frozen tissue was homogenized in cavitation buffer (5 mM HEPES, pH 7.4; 3 mM MgCl2; 1 mM EGTA; 250 mM sucrose) containing protease and phosphatase inhibitor cocktails (Sigma-Aldrich) and 0.1 μM okadaic acid (Alexis Biochemicals, San Diego, CA). Cell lysis was achieved by nitrogen cavitation (250 psi, 5 minutes on ice). The resulting lysate was centrifuged at 700 relative centrifugal force (rcf) for 10 minutes. Cytosolic fractions were obtained by centrifuging the low-speed supernatant at 16,000 rcf for one hour. Nuclei were purified from the low-speed pellet by washing twice in cavitation buffer at 2700 rcf for 5 minutes each. The washes were repeated in cavitation buffer containing 0.5% IGEPAL detergent (Sigma-Aldrich), and then the pellet was resuspended and loaded onto a 1 M continuous sucrose gradient prior to centrifugation at 4°C, 2700 rcf for 10 minutes. The pellet was recovered and the sucrose gradient repeated, followed by a wash in cavitation buffer containing detergent for 5 minutes at 2700 rcf, and then a final wash in cavitation buffer without detergent for 5 minutes at 16,000 rcf. Extracts were obtained from the purified nuclei by resuspending in lysis buffer (10 mM Tris pH 7.4, 150 mM NaCl, 1mM EGTA, 1 mM EDTA, 1 mM IGEPAL detergent) containing protease and phosphatase inhibitors (Sigma-Aldrich) and agitating at 4°C for 30 minutes. Nuclear extracts were clarified by centrifugation at 16,000 rcf for 30 minutes at 4°C.
Nuclear and cytoplasmic extracts of cultured cells or frozen tissue were also obtained using the Pierce NE-PER kit according to manufacturer's instructions (Thermo Scientific Pierce Protein Research Products, Rockford, IL). Cells were either untreated, or were treated for 30 minutes with 15 ng/mL TNFα. Proteins were quantified as above.
Western Blots
Fifteen to 20 micrograms total protein per lane were separated on 10%, 12%, or 15% SDS PAGE gels and transferred to PVDF membranes (Millipore, Billerica, MA). Membranes were blocked in TBS-T (20 mM Tris-HCl pH 7.6, 137 mM NaCl, 0.1% Tween-20) containing 5% nonfat dry milk for 1 hour at room temperature. Primary antibodies were applied overnight at 4°C in TBS-T+5% milk. Primary antibodies included: NFκB subunit p65, (Santa Cruz Biotechnology, Inc, Santa Cruz, CA); phosphorylated (S536 or S276) NFκB subunit p65, (Cell Signaling Technology, Danvers, MA); IκB-α (Santa Cruz); tubulin (Abcam Inc, Cambridge, MA); histones (Chemicon/Millipore) (MAb 052); actin (Sigma-Aldrich); cleaved caspase 3 (Cell Signaling). After extensive washing, membranes were incubated with the appropriate secondary antibody (goat anti-mouse or goat anti-rabbit-HRP conjugate, Santa Cruz) at room temperature for one hour. Detection was accomplished with ECL plus detection reagent (GE Healthcare Biosciences, Piscataway, NJ) and images were collected on a Kodak Image Station 4000MM. Membranes were stripped for 1 hour at 50°C in strip buffer (62.5 mM Tris-HCl pH 6.7, 2% SDS, 100 mM β-mercaptoethanol), rinsed extensively in TBS-T, and re-probed as needed.
Immunohistochemistry
Slides of human primary medulloblastoma were obtained from the Cooperative Human Tissue Network Pediatric Division (Columbus Children's Hospital, Columbus, OH). The sections were deparaffinized in Citrisolv (Fisher Scientific, Pittsburgh, PA) and a standard citrate buffer antigen retrieval technique [
15] was used to unmask protein epitopes. After blocking, the sections were incubated in phos(S276)p65 primary antibody (Cell Signaling, Boston, MA) overnight at °C. After washing, sections were incubated in an HRP-conjugated goat anti-rabbit secondary antibody at room temperature for 1 hour. Biotin tyramide (NEN Life Science Products, Perkin Elmer, Waltham, MA) signal amplification was used to enhance detection with the avidin biotin complex (ABC) method (Vector Laboratories, Burlingame, CA) and diaminobenzidine (Sigma-Aldrich) chromagen substrate. The sections were counterstained with hematoxylin. These studies were performed at the UAB Neuroscience Molecular Detection Core, supported by P30 NS47466.
DNA construct
A dominant negative form of IκB-α (S32A and S36A) was amplified from the plasmid pCMV-dnIκBαM (Clontech, Mountain View, CA) and cloned into the pTRE-Tight-Bi-AcGFP1 vector (Clontech) at the KpnI and NotI sites to create pTRE-Tight-Bi-AcGFP-dnIκB (Additional File
1 A). This vector co-expresses AcGFP with dnIκB from a bidirectional promoter upon activation of the tet response element. Sequences were confirmed by automated sequencing.
Double-stable cell line 4H10
D425 cells were first transfected with the pTet-off plasmid (Clontech) to create a tet-responsive line expressing the tTA protein. Cells were transfected with Lipofectamine at a 1:5 DNA to Lipofectamine ratio according to the manufacturer's instructions (Invitrogen, Carlsbad, CA). The cells were incubated overnight with the Lipofectamine/DNA complexes, and the next day cells were washed to remove the DNA. Twenty four hours later cells were re-plated in a 6-well dish for selection with 0.4 mg/mL geneticin (Invitrogen). After 10 days, cells were expanded to a 6 cm dish, and a week later colonies were isolated using cloning disks (RPI) and expanded for testing. Clones were evaluated for expression of tTA using a luciferase assay (see below). Selection was maintained at 0.1 mg/mL geneticin.
The double stable cell line expressing dnIκB and AcGFP under control of the tet response element was created by transfecting the D425 tet off cell line with the pTRE-Tight-Bi-AcGFP-dnIκB construct and a linear hygromycin marker (Clontech) in a 10:1 ratio as above. Clones were selected by limiting dilution in 96-well dishes using 0.2 mg/mL hygromycin B (EMD Biosciences, Gibbstown, NJ). Doxycycline (DOX) (Sigma-Aldrich) was included at 100 ng-1 μg/mL to repress expression of dnIκB. Clones were expanded and tested by FACS (UAB CFAR Flow Cytometry Core) for expression of AcGFP by washing out DOX as follows. Cells were harvested, washed twice in PBS, and re-plated in fresh media containing no DOX. Three to 6 hours later the media was replaced again. The clone displaying the highest fluorescence, 4H10, was further evaluated; selection was maintained at 0.1 mg/mL hygromycin.
Luciferase assays
Tet off clonal cell lines were evaluated for expression of tTA by transiently transfecting with the reporter plasmid pTRE-HA-Luc (Clontech) using Lipofectamine as above. The next day cells were washed and split into 2 identical wells, and DOX was added to one well at 2 μg/mL. The following day, cells were washed and incubated with beetle luciferin (Promega, Madison, WI) and luminescence was measured on a LumiStar luminometer (BMG Labtech, Japan). Cells were then counted and luminescence was normalized to cell number. Fold induction was calculated as relative luciferase units (RLU): RLU (No DOX)/RLU (DOX). A cell line with greater than 20 fold induction and little to no background was selected for further work.
Double stable cell lines expressing dnIκB under control of the tet response element were evaluated for biological response by transiently transfecting with the reporter plasmid p5X NFκB-Luc using Lipofectamine as described. The next day, DNA-containing medium was removed and cells were cultured in 100 ng/mL DOX or washed as described above to remove DOX. Two days later, cells were collected and assayed for luciferase activity as above.
Xenografts in athymic mice
Twenty-four 6-8 week old female athymic mice were randomized to two arms, DOX or No DOX. Mice in the DOX arm received mouse pellets containing 200 mg/kg [
16] DOX (Bio-Serv, Frenchtown, NJ) for 1 week prior to tumor implantation.
4H10 cells (D425 medulloblastoma cells containing dnIκB construct) were cultured in 100 ng/mL DOX until 24 hours prior to injection of flank tumors. DOX was washed out of half the cells as described above for injection into mice in the No DOX arm. One million cells were resuspended in 200 μL of a solution containing 50% serum-free media and 50% Matrigel (BD Biosciences, San Jose, CA) and injected subcutaneously into the right flank of the mice, one injection per animal, on Day 1. Tumors were measured with digital calipers and tumor volume was calculated using the formula A
2xB/2, where A is the short diameter of the tumor, and B is the long diameter perpendicular to A. Mice in all studies were weighed and assessed for general health once a week, and tumors were measured twice a week. Animal experiments were performed according to the National Institutes of Health Guide for the Care and Use of Experimental Animals and approved by the Institutional Animal Care and Use Committee (IACUC APN 08607). Animals were sacrificed according to humane guidelines provided by the IACUC. The study endpoint was Day 37, when several tumor volumes were estimated at nearly 10% of the animals' weight. Tumors were removed and measured in three dimensions, and weighed to validate the accuracy of the
in situ measurements.
Statistics
The xenograft study used one tail Student's t-test assuming equal variance. In figures, all error bars represent standard error of the mean (SEM).
Discussion
Our results indicate that NFκB is over-active in medulloblastoma, the most common malignant brain tumor in children. This was demonstrated in all six primary tumors examined by either IHC or western, as well as in xenografts of three human medulloblastoma cell lines. Further, limiting NFκB activity using a variety of pharmacological agents that target different steps of the pathway has a negative effect on tumor cell line viability and growth. These results implicate NFκB signaling as a necessary component of medulloblastoma tumor maintenance and tumor progression. We did not investigate the role of NFκB in tumorigenesis.
The five pharmacologic NFκB inhibitors we tested have different targets for disrupting NFκB signaling. Curcumin acts on IKK [
37,
38], a signal-dependent kinase, to block phosphorylation of p65 and IκBα by generation of reactive oxygen species [
39]. Sulfasalazine inhibits phosphorylation of IκBα by blocking IKK activity [
40,
41]. Another activity of sulfasalazine is inhibition of system Xc(-), an amino acid transporter essential for cystine uptake and survival of glioblastoma [
20]. We ruled out system Xc(-) as the mechanism of action in medulloblastoma cell lines by treating with S-4-CPG, a specific system Xc(-) inhibitor, with no effect on cell survival (data not shown). Bortezomib inhibits the ubiquitin-proteasome, thereby preventing IκB degradation [
32]. Dithiocarbamates are metal chelators, antioxidants, proteasome inhibitors (under certain conditions) [
42,
43], and potent NFκB inhibitors [
44], functioning to block IκBα release from NFκB in the cytoplasm. PDTC causes oxidation of NFκB, thereby decreasing DNA binding [
45]. None of these agents is specific to NFκB but strike different targets within the pathway. Because these agents have many activities, yet have NFκB inhibition in common, we conclude that the commonality of tumor cell sensitivity to all of them lies in the NFκB pathway [
10].
Because of the lack of specific pharmacologic inhibitors, we investigated a genetic blockade of the canonical NFkB pathway. The inducible dnIkB construct reduced growth of cells in culture by only about 25% of parent D425 cells, on average, as did a second dominant negative cell line established in Daoy parent cells (data not shown). Furthermore, the dnIκB in D425 cell xenograft tumors had a modest impact on tumor growth but did not completely eliminate it in most animals. It is likely, given these results, that pharmacologic inhibitors are more effective because they target NFκB-dependent as well as NFkB-independent pathways.
The role of NFκB has been extensively studied in neurons. p65/p50 heterodimers have been isolated from synapses [
21]; neurotransmitters are among the many activators of NFκB [
22]. Despite their neuronal lineage, medulloblastoma cells do not fire action potentials [
46], so the role neurotransmitters play in NFκB signaling in medulloblastoma is not known. NFκB is activated by Ca
2+ influx via L-type Ca
2+ channels, but medulloblastomas do not express these channels, at least
in vitro [
46]. Therefore, neuronal physiology has not been helpful in clarifying the role of NFκB in medulloblastoma. However, it is clear that NFκB is activated in neurons in a variety of pathologic brain conditions. Neurons become more vulnerable to injury when NFκB is inhibited [
23‐
25]. Thus, NFκB may be one mechanism used by neurons to survive insult [
23]. We demonstrate that medulloblastoma cells are much more sensitive to NFκB inhibitors than normal immature neurons (Figure
2B), which would suggest that medulloblastoma cells rely on downstream targets, such as antiapoptotic genes, when they are physiologically challenged. In contrast, neuroprotection has also been reported by blocking NFκB in cerebellar granule neurons [
25]. Indeed, inhibition of NFκB activity has been shown to promote as well as protect against neurotoxicity [
25,
26]. NFκB plays a role in cerebellar development during the first 1-2 weeks of life in mice and rats, and probably the first several months in humans [
27,
28]. It is suspected that NFκB signaling is involved in early granule neuron migration [
27]. More studies are needed to determine whether NFκB inhibition could prevent medulloblastoma cell migration or metastatic spread, especially because this is a feature of medulloblastoma that is not very well understood.
Constitutive NFκB activity has been reported in adult malignant brain tumors and a number of other solid tumors in adulthood. There has been a focus on evidence relating chronic inflammation to cancer in recent years [
10,
29], including the role that NFκB plays in each. More specifically, neural stem cell inflammation and dysregulation of NFκB signaling could be a source of tumorigenesis in the brain [
29]. However, the link between inflammation and cancer will be much harder to make in pediatric cases, because most tumors present in children who have no history of illness and no clinical inflammation. In fact, H&E staining of the three primary tumors presented here showed no inflammatory cell infiltration (data not shown). Yet, NFκB still seems to be significant in pediatric cancers. Retinoblastoma, a cancer of primitive neuroectodermal cells of the retina that occurs in very young children, is dependent on NFκB activity for survival [
30]. Neuroblastoma, a tumor of children arising from neural crest cells, displays a spectrum of outcomes from spontaneous regression to rapid metastatic spread resistant to the most aggressive therapeutic interventions. Some neuroblastoma cell lines show high NFκB activity with dramatic response to its inhibition [
31,
32]. Additionally, high NFκB levels were found in rhabdomyosarcoma and Ewing's family of tumors [
33,
34]. There are two reported investigations of NFκB in medulloblastoma. The first examined methionine stress in brain tumors, and saw that methionine deprivation resulted in NFκB activation in one medulloblastoma cell line, causing rapid cell death [
35]. This is consistent with our finding that NFκB is active and can be upregulated in cell lines. Specifically, we showed a large increase in detectable p65 when cultured cells are implanted as xenografts, compared to the same cells in culture. The other report indicates that 2-methylestrodiol, an inhibitor of NFκB, is pro-apoptotic in three medulloblastoma cell lines [
36]. This result is very consistent with our data using five inhibitors of NFκB in multiple cell lines.
The clinical relevance of associating NFκB activity with malignancy may be substantial. First, there are a number of US Food and Drug Administration-approved or experimental drugs available that act to inhibit the NFκB signaling pathway. There are other anti-apoptotic signals in play in solid tumors, such that inhibition of NFκB is unlikely to change the course of the tumor when used as the sole anticancer therapy. Yet one must keep in mind that multi-drug therapy has long been the approach to cancer care because of this very issue. Second, blocking NFκB generally enhances the responsiveness of tumors to traditional chemotherapy [
47,
48]. This, however, is also not straightforward. NFκB activity is necessary for paclitaxel and doxorubicin cytotoxicity [
49,
50], and some traditional chemotherapies activate NFκB signaling [
51,
52]. Conversely, apoptosis induced by irinotecan, daunorubicin and cisplatin therapies is reportedly enhanced by blocking NFκB [
47,
49,
53]. Once again the complexity of NFκB -modulated apoptosis is revealed, highlighting the importance of the proper balance of NFκB activity in individual tumors treated with specific chemotherapeutic agents. Cisplatin is an important anti-medulloblastoma drug; studies need to be performed to evaluate the effect of NFκB blockade on the chemosensitivity of medulloblastoma to cisplatin. Finally, radiation is the most effective therapy for medulloblastoma and is known to activate NFκB. Yet inhibition of NFκB with curcumin or other pharmacological and molecular inhibitors increases radiosensitivity in glioblastoma [
54,
55] and fibrosarcoma cells [
49]. Increasing medulloblastoma sensitivity to radiation, especially if normal neuron sensitivity is unchanged, could dramatically change current treatment and improve outcomes for patients with medulloblastoma.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SES helped to conceive of the study, participated in study design and coordination, prepared the manuscript, performed annexin V staining, xenograft studies, and some dose-response testing. NJL carried out the cloning studies, subcellular fractionation, participated in study design, neurosphere experiments, dose-response testing and Western analysis. LAD participated in immunostaining, and Western analysis. HS helped to conceive of the study, and participated in its design and coordination. All authors read and approved the final manuscript.