Novel function of MDA-9/Syntenin (SDCBP) as a regulator of survival and stemness in glioma stem cells

Glioblastoma multiforme (GBM) is an aggressive cancer with current therapies only marginally impacting on patient survival. Glioma stem cells (GSCs), a subpopulation of highly tumorigenic cells, are considered major contributors to glioma progression and play seminal roles in therapy resistance, immune evasion and increased invasion. Despite clinical relevance, effective/selective therapeutic targeting strategies for GSCs do not exist, potentially due to the lack of a definitive understanding of key regulators of GSCs. Consequently, there is a pressing need to identify therapeutic targets and novel options to effectively target this therapy-resistant cell population. The precise roles of GSCs in governing GBM development, progression and prognosis are under intense scrutiny, but key upstream regulatory genes remain speculative. MDA-9/Syntenin (SDCBP), a scaffold protein, regulates tumor pathogenesis in multiple cancers. Highly aggressive cancers like GBM express elevated levels of MDA-9 and contain increased populations of GSCs. We now uncover a unique function of MDA-9 as a facilitator and determinant of glioma stemness and survival. Mechanistically, MDA-9 regulates multiple stemness genes (Nanog, Oct4 and Sox2) through activation of STAT3. MDA-9 controls survival of GSCs by activating the NOTCH1 pathway through phospho-Src and DLL1. Once activated, cleaved NOTCH1 regulates C-Myc expression through RBPJK, thereby facilitating GSC growth and proliferation. Knockdown of MDA-9 affects the NOTCH1/C-Myc and p-STAT3/Nanog pathways causing a loss of stemness and initiation of apoptosis in GSCs. Our data uncover a previously unidentified relationship between MDA-9 and GSCs, reinforcing relevance of this gene as a potential therapeutic target in GBM.


INTRODUCTION
Cancer is multifactorial in its etiology and multistep in its evolution [1]. Glioblastoma multiforme (GBM), the most common form of primary brain tumor, is an aggressive cancer that causes high mortality and morbidity. GBM currently remains one of the most difficult cancers to treat, with less than a 5% 5-yr survival rate, despite multi-modality therapies including surgery, radiation therapy, and chemotherapy [2]. This is potentially due to a lack of well-defined understanding of the mechanism(s) underlying GBM's complex heterogeneity, plasticity and therapy resistance. Since their discovery and initial characterization in 1994, research on cancer stem cells has intensified, providing convincing evidence that these unique cells are major contributors to cancer growth, progression, tumor heterogeneity and resistance to therapeutic intervention [3]. The concept that cancer is comprised of nearly homogenous, ectopically growing cells has been replaced with a more complex heterogeneous model in which cancer cells have varied potential to metastasize, interact with the stroma and

Priority Research Paper
GSCs that would express elevated levels of MDA-9. GSCs are major contributors to cancer progression [9] and MDA-9 plays a seminal role in the progression of several cancer types including glioma [21][22][23][24][25][26]. Accordingly, we currently assessed the association between stemness and MDA-9 expression in GBM, as well as in normal astrocytes. Stemness is defined as the ability of stem cells to selfrenew and differentiate [27]. We studied this property using sphere formation assays, cell-surface based stem population assessment, monitoring genes regulating selfrenewal, and tumorigenicity. Additionally, the influence of MDA-9 on GSC survival, growth, angiogenesis and chemoresistance was also examined. Finally, we dissected the mechanisms contributing to MDA-9-mediated stem phenotypes and survival. Our studies now demonstrate for the first time that MDA-9 promotes glioma stem cell phenotypes and survival through regulation of NOTCH1, C-Myc, STAT3 and Nanog in GSCs.
Forced mda-9 overexpression in normal human astrocytes led to a significant increase in spheroid size ( Figure S1B) as reflected by assessment of putative GSC and NSGC populations as well as changes in genes involved in self-renewal. No change in tumorigenicity was observed, when assayed by mice xenograft studies (data not shown). Overexpression of MDA-9 in NSGCs, significantly increased the stem population and expression of canonical stem regulatory genes ( Figure 2B, 2C). Even though NSGC populations had elevated expression of mda-9, the GSC populations had significantly higher expression than the corresponding normal brain (Figure 2A, top right panel). To further confirm that MDA-9 regulates stem regulatory genes mda-9 was suppressed by kd in GBM (cell line and clinical samples). Silencing of mda-9 significantly decreased the recognized stem regulatory genes, and markers (Table 2). Overall, Nanog was decreased by ~33-, ~25-and ~11fold, Oct4 by ~7-, ~12-and ~2-fold, and Sox2 by ~10-, ~7-and ~4-fold in the mda-9 kd GSCs from VG2, VG9, and U-1242, respectively. Silencing of mda-9 also resulted in significant loss of self-renewal (Figure S1B) as defined by the self-renewal assays. In total, these data support the hypothesis that mda-9 can regulate stemness in both normal astrocyte stem cells and GSCs.

MDA-9 regulates stem cell survival and intracranial GSC growth
mda-9 kd increased apoptotic cell death in GSCs ( Figure 5A). The population of apoptotic cells in shmda-9 GSCs was 38 ± 3.3 %, 36 ± 5.1 % and 45 ± 4.9 % (in VG2, VG9 and U-1242, respectively) after 72 hours, which was ~5-fold of that observed in shcon GSCs. Suppression of mda-9 also resulted in a significant loss in GSC tumorigenicity and an increase in mouse survival in vivo ( Figure 5B, 5C, 5D; p < 0.05). The control mice showed spongioblastic tumors with rhythmic palisades, a constant feature of aggressive high grade glioblastoma. Tumors in mice injected with shmda-9 GSCs were extremely small, and did not display the distinguishing aggressive spongioblastic pattern ( Figure 5C).

MDA-9 regulates stem cell renewal, maintenance and survival through C-Myc
Considering C-Myc's influential role in stem cell renewal, maintenance, and survival [34,35], we investigated the role of MDA-9-mediated regulation of C-myc in HA SCs and GSCs. Suppression of mda-9 by kd and enhanced expression of mda-9 with an expression vector lead to a significant decrease (~3-, ~2-and ~5-fold protein, and ~3-, ~10-and ~12-fold mRNA in VG2, VG9 and U-1242 GSCs, respectively) or gain of C-Myc (~3fold protein in HA) expression, respectively ( Figure 7B; Figure S2; Table 2). The change in C-Myc was observed at both an RNA and protein level ( Figure 7B; Figure S2; Table 2). This loss of a c-myc expression phenotype in shmda-9 GSCs was reversed by c-myc overexpression ( Figure 7C). mda-9 regulation of c-myc occurred though RBPJK transcription ( Figure 7A); which is possibly regulated by NOTCH1 cleavage/activation ( Figure S2) via interaction with its ligand, DLL1 ( Figure 6A). These findings support the concept that MDA-9 plays a critical role in the regulation of C-Myc in GSCs, which is a major contributor of glioma stemness and GSC survival (34) via the activation of NOTCH1 and RBPJK.

MDA-9 regulates GSC survival through p27/Kip-1 and cIAP2
Kd of mda-9 led to increased expression of p27/ Kip-1 in GSCs, at both an RNA and protein level ( Figure  8A, Figure S2). The increased expression of p27 that culminated in cell death could be prevented by forced expression of c-myc, indicating that GSC survival is dependent on c-myc and p27/kip-1 expression ( Figure  8B). In shmda-9 GSCs, expression of miR-221 was also significantly decreased ( Figure 8A). These findings demonstrate that p27/kip-1 is regulated by mda-9 through c-myc and miR-221. mda-9 kd caused decreased cIAP2 expression ( Figure S2) and this combined with increased expression of p27/kip-1 in shmda-9 GSCs may amplify GSC death. To verify p27/kip-1's involvement in GSC survival, we overexpressed p27/kip-1 in GSCs and observed a loss of sphere integrity and viability, in both patient-derived GBM and the U-1242 GBM cell line  Figure 8C). We also observed that cell death in shmda-9 GSCs was mediated by Caspase activation ( Figure 8D).

DISCUSSION
GSCs, also called glioma initiating cells, are considered defining elements in the carcinogenic process, hypothesized to represent critical constituents of invasion, angiogenesis, cancer cell resistance to therapy and escape of tumor cells from dormancy (tumor recurrence /relapse occurring after an initial therapeutic response) [36][37][38]. MDA-9 is a diagnostic marker of tumor aggression and grade, and a positive association has been reported between MDA-9 expression and glioma stage [23]. We now demonstrate a fundamental and central role of MDA-9/Syntenin (SDCBP) as an upstream regulator of stemness and GSC survival in GBM. MDA-9 contributes to GSC cell-cell/cell-matrix adhesion, angiogenesis and invasion. Stem cell-mediated cancer progression is a major clinical problem [6-9, 11, 18, 20] and is accentuated as a significant contributor to therapy-resistance and cancer relapse [4]. mda-9 expression positively correlated with stemness as confirmed by a direct association between expression of mda-9 and stem cell markers and genes, in both GBM patient samples and cell lines. Loss or gain of mda-9 expression led to a corresponding loss or gain of cell surface stem markers, respectively (Figure 2A, 2B, 2C, and Table S2), as well as changes in recognized selfrenewal/pluripotency genes including Nanog, Oct4, Sox2 and c-myc ( Figure 1A, 1B; Figure S1, S2; Table 2). mda- 9 expression was also significantly higher in GSCs than NSGCs (Table 1) and both were dramatically elevated as compared to corresponding normal astrocyte stem cells (Figure 2A). mda-9 also regulated STAT3 expression (Figure 3, Figure S2), which is a key contributor to cellular transformation and tumor maintenance in many cancer contexts, including GBM [16]. Activation of a STAT3-mediated transcriptional network correlates with mesenchymal GBM transformation and poor prognosis (30,31,39). STAT3 also regulates self-renewal in cancer stem cells by systematically regulating canonical stemness genes including Nanog, Sox2, Oct4 [16,29,30] and myc [40]. NANOG also acts as a master switch of the central stemness transcriptional network, as OCT4/SOX2 bind to the proximal region of the Nanog promoter stimulating Nanog expression [15]. NANOG, SOX2 and OCT4, also bind to their individual promoter's, thus sustaining a unified auto-regulatory network of cell pluripotency and self-renewal [15]. Our data reveal that mda-9 is a key regulator of this core GSC regulatory system through regulation of STAT3 (Figure 9). STAT3 can be regulated by SRC, IGF-1R, and p-44/42 [29][30][31][32][33]. Phosphorylated p-44/42 (T202/Y204) and SRC (T417, Y418) phosphorylate STAT3 at position Y705. Our data indicates that MDA-9 regulates STAT3 by controlling IGF-1R ( Figure 4B), p-44/42 ( Figure 4A) and Src ( Figure 6B) signaling. MDA-9 also regulates the activity of FAK [23,24], RAF and RKIP [25] and it ultimately controls the activation of p-44/42. MDA-9 physically interacts with c-SRC through its PDZ binding motifs and is essential for activation of SRC [23,41]. These data demonstrate that MDA-9 influences GSC stemness on multiple molecular levels. The higher expression of MDA-9 in GSCs than in normal astrocyte stem cells may indicate that GSCs are more dependent on mda-9 expression than their corresponding non-cancer stem cells. The potential "addiction" of GSCs to MDA-9 is an area of current investigation.
Another critical pathway in stem cell biology is the NOTCH pathway [16]. NOTCH signaling plays an important role in development by regulating cellfate determination, cell survival, and proliferation [17]. Activation of NOTCH receptors occurs through binding with a number of distinct ligands (including delta-like 1   -1, mda-9, and c-myc in shcon, mda-9 kd GSCs and mda-9 kd GSCs overexpressing c-myc. * indicates significance in expression of p27/kip-1 and c-myc between the shmda-9 and shmda-9 + c-myc groups. Relative expression indicates fold-change in expression. See also Figure S2. C. Image analysis of control and p27/Kip-1 overexpressing GSCs. D. Caspase 3/7 activation analysis in shcon and shmda-9 GSCs. Bars represent SEM. *P < 0.05, using student t-test and ANOVA. www.impactjournals.com/oncotarget / DLL1, jagged 1). Upon ligand binding, the intracellular NOTCH domain (ICD) is cleaved and translocates into the nucleus, where it regulates downstream target gene transcription. Aberrant NOTCH signaling promotes tumorigenesis [17]. Recently, a role of the NOTCH signaling pathway in promoting self-renewal of both normal and cancer stem cells has been demonstrated [17,42]. Our data indicates that MDA-9 regulated NOTCH1 activity on two levels. MDA-9 controlled the expression of NOTCH1 in cells by degradation ( Figure 6B), through NUMB, a NOTCH binding ubiquitin ligase [13,43]. In the presence of p-SRC, NUMB is phosphorylated and then degraded, preventing it from degrading NOTCH1 [44]. In the absence of MDA-9, SRC is not activated to p-SRC, and this leads to higher expression of NUMB resulting in degradation and a decrease in the levels of total NOTCH1. MDA-9 also controls NOTCH1 activity by regulating NOTCH1 activation through expression of DLL1, the ligand of the NOTCH1 receptor ( Figure 6A). The intracellular PDZ binding motif of DLL1 regulates DLL1 protein stability [45], DLL1 trafficking and signaling activity. DLL1 ubiquitination is not required for its internalization, but is necessary for its recycling back to the plasma membrane and efficient interaction with NOTCH1 [46]. MDA-9 can regulate the expression of DLL1 on the cell surface by regulating the interaction between DLL1 and ubiquitin. An effect of MDA-9 on DLL1 has been reported in zebrafish stem cells [47]. The c-terminal of MDA-9 binds to ubiquitin [48], and its PDZ domain may then bind to the PDZ binding motif of DLL1, and this interaction regulates the expression of DLL1 on the surface of GSCs. In the absence of MDA- 9 this interaction is altered, leading to decreased DLL1 surface expression. This further reduces the interaction of NOTCH1 with its ligand DLL1, leading to decreased activation of NOTCH1, reduced translocation of the intracellular domain (ICD) of Notch1 to the nucleus and decreased transcription of NOTCH1 target genes.
NOTCH1 directly regulates C-Myc expression [49]. The ICD of NOTCH1 translocates to the nucleus and binds to the promoter of the transcription factor RBPJK, which regulates c-myc expression [50]. The binding of NOTCH1 to the promoter region of RBPJK promotes expression of RBPJK, leading to expression of c-myc. In MDA-9 kd cells the ICD of NOCTH1 is unable to translocate to the nucleus, preventing transcription of RBPJK ( Figure 7A), thereby inhibiting elevated c-myc expression ( Figure 7B).
Elevated MYC proteins are associated with many cancers and correlate with cancer risk and poor patient survival [19,51]. Activation of MYC is linked to cellular growth, proliferation and metabolism. C-Myc controls the balance between stem cell self-renewal and differentiation in normal cells. In GBM, C-Myc is essential for GSC regulation and maintenance [34,35]. C-myc also controls the proliferation of cells by regulating cell cycle modulators including the cyclin-dependent kinase inhibitor, p27, which is a critical target of C-Myc [52]. SRC has also been shown to negatively regulate p27 and elevated levels of p27 cause arrest of tumor growth and apoptosis [53]. Additionally, p27 can suppress SOX-2 [54], which leads to apoptosis in stem cells [55]. Our data revealed that kd of mda-9 decreased SRC, Sox-2 and C-Myc activities, whereas p27/kip-1 expression was increased, culminating in apoptosis of GSCs ( Figures 3B,  5A, 6B, 7B). Another anti-apoptotic molecule cIAP2, was also regulated by MDA-9 in GSCs ( Figure S2). IAP family members, XIAP, cIAP1, cIAP2, NAIP and survivin, are expressed at higher levels in CD133 positive than in CD133 negative GBM [56], and these anti-apoptotic proteins contribute to GSC survival under adverse conditions. Kd of mda-9 expression decreased expression of cIAP2 ( Figure S2), which also participated in induction of apoptosis ( Figure 5A).
The current data suggests that MDA-9/syntenin (SDCBP) is part of a complex, tightly regulated connectivity network that confers self-renewal, survival and tumor progressive properties to GSCs [57]. Stemness, initially defined by the expression of cell surface markers and stem cell genes, is a property shared by normal stem cells and GSCs [58]. MDA-9 appears to regulate stemness through similar pathways in both normal astrocyte stem cells and GSCs. However, GSCs appear to be more dependent on ("addicted" to) MDA-9, with significantly elevated expression (Figure 2A), for maintenance and survival than normal stem cells. Forced elevated expression of MDA-9 in normal astrocytes increased their invasiveness, self-renewal and the overall proportion of stem cells, but it did not render these cells tumorigenic. The regulation of stemness by MDA-9 is not exclusive to GSCs, but elevated expression enhances GSC survival, invasion, angiogenesis, and self-renewal. MDA-9 is capable of regulating multiple aspects of stem cell phenotypes simultaneously, validating a critical role in determining GSC stemness. mda-9 can regulate the central transcriptional network of stem regulating genes, additional pluripotency genes, and affects interrelated pathways crucial for stem cell survival (Figure 9). Considering the pivotal role of MDA-9 in determining both GSC aggressiveness and survival, directly targeting MDA-9 expression or its interaction with effector interacting proteins using genetic or pharmacological approaches may provide a unique opportunity to develop targeted therapies for this important component of cancer pathogenesis.

Cell lines and tissue samples
Specimens of human primary normal and malignant brain tumors (n = 50) were collected from subjects who underwent surgical removal of their brain tumors. All subjects were informed of the nature and requirements of the study and provided written consent to donate their tissues for research purposes. Informed consent was obtained according to Origene and the research proposals approved by the Institutional Review Board at the VCU TDAAC. The human glioma cell line U-1242-luc-GFP was kindly provided by Dr. Kristofer Valerie (VCU). U-1242/ luc-GFP (U-1242-Luc), cells were cultured in DMEM medium supplemented with 10% fetal bovine serum and antibiotics. Isolated non-stem glioma cells (NSGCs), based on lack of expression of CD133 and CD44, were cultured similarly in monolayer culture. Normal human astrocytes (HA) were obtained from Clonetics, USA and grown in Astrocyte Basal Media (ScienCell, CA) supplemented with Astrocyte growth supplement (ScienCell, CA). The cumulative culture length of the cells was less than 6 months after resuscitation. Early passage cells were used for all experiments and authenticated. All the cell lines were frequently tested for mycoplasma contamination using a mycoplasma detection kit from Sigma. All primary GBM cells were authenticated by IDEXX Bioresearch (Columbia, MO).

Isolation and culture of human GBM, putative GSCs and NSGCs
Human GBM GSCs and NSGCs were isolated from GBM tissue from surgical specimens and from established U-1242/luc-GFP GBM cells. GBM tissue samples were dissociated and digested samples were filtered with 70 µm nylon cell strainer (BD) and resuspended in stem cell medium comprised of DMEM/F-12 50:50 containing K27 supplements, glutamine 2 μmol/L (Invitrogen), basic fibroblast and epidermal growth factors (PeproTech, 20 ng/mL each) for continuous culturing [59]. All primary cells were cultured as suspended spheres in ultra-low attachment T25 or T75 culture disks (Corning) in Essential 8 medium (Invitrogen) and analyzed prior to 5 passages to form GSC enriched neurospheres. Neurospheres were dissociated and then labeled with CD44 and CD133 antibody (Miltenyi Biotech) to isolate both pro-neural and mesenchymal GSC subtypes. Stained cells were sorted through a BD Aria II sorting station. Antibody negative and positive cell populations were counted and collected for further culturing. GSC and NSGC populations were counted and collected for further culturing. Xenografted human GSCs were isolated from mice and analyzed for cell surface and intracellular proteins by FACS. The GSCs were cultured in ultra-low attachment plates and flasks with Essential 8 medium (Invitrogen), unless indicated. Isolated NSGCs were cultured in monolayer with complete DMEM medium.

Isolation and culture of primary human astrocyte stem cells
Primary normal human astrocytes were cultured in ultra-low attachment plates and flasks (Corning) in Astrocyte media (ScienCell, CA) . The cells were stained with CD44 and CD133 antibodies, sorted and cultured further under ultra-low attachment conditions.

Self-renewal assay
Sphere-forming assays were used to determine clonogenic growth potential in vitro of both normal and neoplastic cells [60]. Sorted GSCs and NSGC populations were diluted to a density of 500 cells/ml. 2 μl of the diluted cell suspension was plated per well in a 96 well ultra low attachment plate (Corning Inc., Corning, NY, USA), and 150 μl of serum-free medium was added, cultures were then observed daily (n = 96). Additionally, flow cytometry with CD44 and CD133 antibody (Miltenyi Biotech) was performed to assess the stem populations.

Gene expression arrays, and analyses
TissueScan Brain Cancer Tissue cDNA array I, containing 46 malignant (covering four stages) and 2 tumor-adjacent normal tissue cDNAs, were obtained from Origene Technologies, (Rockville, MD, USA). This array was analyzed for mda-9, c-myc, Oct4 and Sox2 expression using taqman probes (Invitrogen) according to the manufacturer's protocol. A human GSC array (Qiagen) was used according to the manufacturer's protocol to analyze a clinical GBM sample VG2 (shcon) and an mda-9 kd clone of VG2 (shmda-9). Eighty-four genes were studied. The data was analyzed using the Qiagen webbased PCR array data analysis software.

Promoter reporter assays
Luciferase reporter assays were performed using 2×10 5 cells infected with either Ad.5/3.shcon or Ad.5/3. shmda-9. Twenty-four hours post-infection, cells were transfected with an RBPJK-responsive luciferase reporter construct with Lipofectamine 2000 as described [23]. Cell lysates were harvested and luciferase activity was measured using a Dual-Luciferase Reporter Assay system (Promega) according to the manufacturer's instructions. Luciferase activity was normalized to Renilla activity, and data represent the average of triplicates ± S.D.

Reverse transcription polymerase chain reaction
Total RNA was isolated by TRIzol extraction (Invitrogen) and purified using the RNeasy kit (Qiagen). First-strand cDNA was synthesized with SuperScript III reverse transcriptase (Invitrogen). Quantitative PCR studies were carried out by using the TaqMan Gene expression assays (Invitrogen), and were normalized to 18S expression (Invitrogen). Probe details are provided in supplemental information.

Tumorigenicity studies
All experiments and procedures involving mice were approved by the Institutional Animal Care and Use Committee of Virginia Commonwealth University. For the intracranial brain tumor model, athymic female NCrnu/nu mice (National Cancer Institute-Frederick) were used (n = 10 per group). Mice were anesthetized through i.p. administration of ketamine (40 mg/kg) and xylazine (3 mg/kg) and immobilized in a stereotactic frame (Stoelting). Intracerebral injections of 1.5 × 10 4 cells in 2 μL per mouse were done using an automated injector (Stoelting) as described earlier [23,59]. Tumor burden was determined by bioluminescent imaging [63]. Animals of each group were monitored until they reached the point of euthanization according to the VCU-IACUC approved protocol and survival data was analyzed.
Peptide blocking/activation studies 1 x 10 5 control and treated GSCs were cultured in 6-well ultra-low attachment plates. NOTCH1 blocking peptide (Biovision) and DLL1 peptide (Abcam) were used at a concentration of 10 μg/ml and incubated with cells for 48 hours. After incubation the cells were stained and analyzed for viability, spheroid size and structure.

Live/dead cell assay
Live/Dead cell staining was performed according to the manufacturer's instructions (Invitrogen), followed by imaging by laser confocal microscopy (Leica). The images were analyzed by Zen software.

Statistical analysis
For all in vitro and ex vivo experiments, statistical analyses were conducted using Student's t test and ANOVA (Microsoft Excel). For in vivo studies, statistical analyses were performed using the Kaplan-Meier method (survival studies). Pearson's correlation coefficient (R) and coefficient of determination (R 2 ) were calculated for correlation analysis. The data from clinical samples were analyzed using Microsoft Excel's multiple regression analysis tool. All statistical tests were two-sided, and p values ≤ 0.05 and ≤ 0.01 were considered to be significant and highly significant, respectively. Confidence interval (CI) of 95% was considered statistically significant. Patient data was analyzed using correlation heatmap and cluster analysis tools (Plotly Technologies Inc. Montreal, QC). The RT 2 Profiler PCR Array Data Analysis software was used to study the statistical significance of cancer stem cell array data, and a minimum of an ~4-fold decrease was analyzed by selecting the statistical boundary for Log 10 shmda-9 del del CT/ Log 10 shcon del del CT as 4.

ACKNOWLEDGMENTS
The present study was supported in part by NIH