Increased Fibroblast Growth Factor-Inducible 14 Expression Levels Promote Glioma Cell Invasion via Rac1 and Nuclear Factor-κB and Correlate with Poor Patient Outcome

Glial tumors progress to malignant grades by heightened proliferation and relentless dispersion throughout the central nervous system. Understanding genetic and biochemical processes that foster these behaviors is likely to reveal specific and effective targets for therapeutic intervention. Our current report shows that the fibroblast growth factor-inducible 14 (Fn14), a member of the tumor necrosis factor (TNF) receptor superfamily, is expressed at high levels in migrating glioma cells in vitro and invading glioma cells in vivo. Forced Fn14 overexpression stimulates glioma cell migration and invasion, and depletion of Rac1 by small interfering RNA inhibits this cellular response. Activation of Fn14 signaling by the ligand TNF-like weak inducer of apoptosis (TWEAK) stimulates migration and up-regulates expression of Fn14; this TWEAK effect requires Rac1 and nuclear factor-κB (NF-κB) activity. The Fn14 promoter region contains NF-κB binding sites, which mediate positive feedback causing sustained overexpression of Fn14 and enduring glioma cell invasion. Furthermore, Fn14 gene expression levels increase with glioma grade and inversely correlate with patient survival. These results show that the Fn14 cascade operates as a positive feedback mechanism for elevated and sustained Fn14 expression. Such a feedback loop argues for aggressive targeting of the Fn14 axis as a unique and specific driver of glioma malignant behavior. (Cancer Res 2006; 66(19): 9535-42)

Survival of patients with glioblastoma multiforme (GBM) is compromised by the proclivity of the tumor for local invasion into the surrounding normal brain, causing most patients with GBM to succumb to the disease in <1 year (1). An understanding of glioma oncogenesis has steadily improved, but the molecular mechanisms mediating glioma invasion are still nascent. Gene discovery strategies have been used with the anticipation of identifying molecular pathways that regulate glioma cell migration and invasion (2, 3).

Fibroblast growth factor inducible-14 (Fn14) is a type I transmembrane protein of 102 amino acids in length after removal of signal peptide, making it the smallest member of the tumor necrosis factor (TNF) superfamily of receptors (4–6). The 29-amino-acid cytoplasmic tail of the Fn14 receptor lacks a death domain but contains a single TNF receptor-associated factor (TRAF) binding site flanked by two conserved threonine residues (6, 7). This TRAF site links Fn14 to the nuclear factor-κB (NF-κB) and mitogen-activated protein kinase pathways to regulate endothelial cell proliferation (8) and glioma cell survival (9). We reported that Fn14 receptor is up-regulated in migrating glioma cells (10) and that activation of the Fn14 receptor by addition of TNF-like weak inducer of apoptosis (TWEAK), an exclusive ligand for Fn14, enhances glioma cell survival (9). Additionally, Fn14 expression is very low in normal brain tissue but elevated in GBM specimens (10). Furthermore, high expression of Fn14 has been reported in other human cancers, including hepatocellular (5), pancreatic (11), and breast carcinomas (12).

Dysregulation of cell migration contributes to malignant processes, such as tumor invasion, metastasis, and angiogenesis (13). Cell migration is dependent on continual reorganization of the actin cytoskeleton (14). Members of the Rho family of small GTPases, RhoA, Rac1, and Cdc42, are key mediators of such reorganization and regulate cell migration (15, 16). RhoA mediates the formation of stress fibers and focal adhesions (16), whereas Rac1 stimulates actin assembly, resulting in the formation of lamellipodia (16, 17).

Here, we describe a role for Rac1 in Fn14-mediated glioma migration and invasion and show that TWEAK can induce Fn14 expression levels via the Rac1/IKKβ/NF-κB pathway. This pathway serves as a positive feedback loop that drives sustained Fn14 expression, thereby stimulating invasion. High expression of Fn14 mRNA correlates with increasing glial tumor grade and poor clinical outcome. We show a critical role for Fn14 in the invasive phenotype of glioma cells and suggest that Fn14 overexpression imparts a poor prognosis to patients with GBM.

Cell culture conditions. Human astrocytoma cell lines T98G, U118, U87 (American Type Culture Collection, Manassas, VA), G112 (18), and SF767 (19) were maintained in MEM supplemented with 10% heat-inactivated fetal bovine serum (FBS; Hyclone Laboratories, Inc., Logan, UT). For adenoviral infection, 1.5 × 10⁵ cells were plated into a six-well plate and cultured for 24 hours before infection. Before TWEAK addition, cells were cultured in reduced serum (0.5% FBS) for 16 hours.

Preparation of recombinant adenoviruses and infection. The cDNA encoding murine myc-tagged Fn14 wild-type protein (pSec/Tag2/Fn14wt; ref. 7) was excised and subcloned into the adenoviral shuttle vector pShuttle-CMV to prepare recombinant E1-deleted adenoviruses using the Ad-Easy system as previously described (9). Adenoviruses expressing the superrepressor IκB mutant (S32A/S36A) and signal deficient myc-Fn14tCT were previously described (9). Adenoviruses expressing Rac1N17 and Rac1V12 were previously described (20). Cells were infected at matched multiplicity of infections ranging from 5 to 20. Expression plasmids for Cdc42N17, Rac1N17, and Rac1V12 were previously described (21, 22).

Antibodies, reagents, Western blot analysis, and immunofluorescence. Monoclonal antibodies to Rac and RhoA were purchased from BD Transduction Labs (San Jose, CA). Monoclonal antibodies specific to phospho-IκBα (Ser^32/36) and polyclonal antibodies to Cdc42, IKKβ and IκBα, and phospho-IKKβ were obtained from Cell Signaling (Beverly, MA). A monoclonal antibody to α-tubulin was obtained from Upstate Biotechnology (Lake Placid, NY). Human recombinant TWEAK was purchased from PeproTech (Rock Hill, NJ), and laminin from human placenta was obtained from Sigma (St. Louis, MO). Anti-Fn14 polyclonal serum was precviously described (4). Rhodamine-phallodin was obtained from Invitrogen (Carlsbad, CA).

Immunoblotting and immunofluorescence microscopy experiments were done as previously described (23). Immunofluorescent samples were examined under LSM 5 Pascal Laser scanning confocal microscope (Zeiss, Thornwood, NY) using the appropriate filters.

Immunoprecipitation and Rac and Rho activity assays. Cells were infected with adenoviruses expressing control LacZ, myc-Fn14wt, or myc-Fn14tCT, then cultured for an additional 24 hours. Immunoprecipitation was carried out as previously described (23).

Activity assays for Rac1 and RhoA were done according to the protocol of the manufacturer (Pierce, Rockford, IL). Where indicated, 2.5 μg/mL soluble Fn14-Fc decoy receptor or control mouse IgG (Sigma) were preincubated with 100 ng/mL TWEAK at 37°C for 15 minutes before cell treatments as described previously (10).

Radial cell migration assay. Cellular migration was quantified as previously described (10).

Laser capture microdissection, RNA isolation, and quantitative reverse transcription-PCR. Laser capture microdissection of tumor core and invasive cells was done and total RNA was isolated as previously described (3). PCR analysis of Fn14 and histone H3.3 mRNA levels was conducted with LightCycler analysis software and quantified as previously described (10).

Small-interfering RNA preparation and transfection. Small interfering RNA (siRNA) oligonucleotides specific for Rac1 and GL2 Luciferase were previously described (24). siRNA sequences for Fn14 are as follows: Fn14i-475 (bp 475-495; 5′-CATCCATTCTAGAGCCAGTCT) and Fn14i-863 (bp 863-883; 5′-TAGGAGGGCTGGCCCTAAGAT). Transient transfection of siRNA was carried out as previously described (24). Fn14-directed siRNA was transfected at 40 nmol/L, and no cell toxicity was observed. Maximum inhibition of mRNA and protein levels was achieved 48 to 72 hours after transfection.

Organotypic brain slice invasion assay. Preparation and culture of adult rat brain slices were carried out as described previously (25). Glioma cells (1 × 10⁵; T98G, U118, and U87) stably expressing green fluorescence protein (GFP) were placed onto the putamen of the brain slice (0.5 μL transfer volume). Glioma cell invasion into rat brain slices was quantified using the LSM 5 confocal microscope and depth of invasion (z-axis stacks) was calculated as previously described (25).

Biotinylated electrophoretic mobility shift assay. Glioma cells were plated at a density of 3 × 10⁶ in 100 mm² tissue culture dishes in normal growth medium. After 12 hours, cells were cultured under reduced serum (0.5% FBS) for an additional 16 hours before TWEAK (100 ng/mL) addition. In certain experiments, TWEAK was preincubated with Fn14-Fc decoy receptor (2.5 μg/mL) or control mouse IgG. Isolation of cell nuclear protein was carried out using the NE-PER kit (Pierce) according to the protocol of the manufacturer. Protein-DNA complexes were detected using biotin end-labeled double-stranded DNA 22-mer probes containing the NF-κB binding sites within the Fn14 promoter (NF-κB-Fn14wt sense: 5′-GAGATAAGGGAAATTCCTAGGC; antisense: 5′-GCCTAGGAATTTCCCTTATCTC; NF-κB-Fn14mt sense: 5′-GAGATAAGCGAGAATCGTAGGC; antisense: 5′-GCCTACGATTCTCGCTTATCTC). The binding reactions were done using the LightShift kit (Pierce) according to the protocol of the manufacturer. Where indicated, 200-fold molar excess of unlabeled NF-κB-Fn14wt oligonucleotides or anti-p65 antibody (Santa Cruz Biotechnology, Santa Cruz, CA; 1 mg/mL) was included. The reaction products were resolved by gel electrophoresis and detected by chemiluminescence according to the protocol of the manufacturer (Pierce).

Microarray data, survival analysis, and Fn14 expression. Snap-frozen nonneoplastic brain specimens from epileptogenic patients (n = 24) and tumor (n = 160) specimens with clinical information were collected at Hermelin Brain Tumor Center, Henry Ford Hospital (Detroit, MI). All specimens were collected under an Institutional Review Board–approved protocol and deidentified for patient confidentiality. Clinical information was provided for all samples (29 astrocytomas, 82 glioblastomas, and 49 oligodendrogliomas).

Gene expression profiles of these brain specimens were captured using Affymetrix U133 Plus 2 GeneChips according to the protocol of the manufacturer (Affymetrix, Santa Clara, CA) at the Neuro-Oncology Branch at the National Cancer Institute. Array data was processed according to Affymetrix MAS5 (Microarray Suite 5) algorithm implemented in Affymetrix GCOS (GeneChip Operating Software) and uploaded into GeneSpring 7.2 for data management (Silicon Genetics, Redwood City, CA).

Expression values were filtered for highly variable (differentially expressed) genes (coefficient of variation >30%) across samples producing a list of 7,322 genes. Principal component (PC) analysis was done to investigate the relationship between samples (i.e., to find clusters within the data). Components are sorted from most to least amount of variation. Two clusters were evident in a three-dimensional scatter plot of PC1, PC2, and PC3. The three components accumulatively account for 46% of the variation in the data set. Kaplan-Meier survival curves were developed for each cluster. One cluster had a median survival time of 401 days and the other cluster had a median survival time of 952 days. Box plots for Fn14 expression in each cluster derived from PC analysis were graphed. Significance between the two populations was tested with a two-sample t test assuming unequal variances.

Brain tumor tissue microarray and immunohistochemistry. A glioma invasion TMA was assembled by Dr. D.H. Friedrich as previously described (3). Immunohistochemical analysis for Fn14 was done using an Fn14 monoclonal antibody, P4A8 (Biogen Idec, Inc., Cambridge, MA). P4A8 was generated in an Fn14 knockout mouse (26) immunized with Fn14-myc-his recombinant protein. Anti-Fn14 monoclonal antibodies were screened and selected for binding to Fn14-positive cell lines. Immunohistochemistry was done as previously described using 2.5 μg/mL of the P4A8 antibody (10). A scoring system for chromophore was used to capture the outcome: 0, negative; 1, weak; 2, moderate; 3, strong staining.

Fn14 expression is increased in migrating glioma cells in vitro and invasive cells in vivo. We previously showed that Fn14 expression increased when glioma cell lines were plated on a migration-promoting substrate (10). Here, we examined if the up-regulated expression of Fn14 arises coincident to the migration behavior, irrespective of the specific substrate the cells encounter. RNA was isolated from two subpopulations of cells seeded on glioma-derived ECM: actively migrating cells at the dispersing rim and migration-restricted cells at the crowded cellular core. Fn14 expression was assessed using quantitative reverse transcription-PCR (QRT-PCR). In three glioma cell lines, cells migrating at the rim showed 4- to 19-fold induction compared with core cells (Fig. 1A). RNA was also isolated from tumor core and invasive rim glioma cells from human biopsies; QRT-PCR showed 4- to 6-fold induction of Fn14 in the GBM cells isolated at the rim compared with matched core (Fig. 1B).

Forced Fn14 overexpression stimulates glioma cell migration and alters cellular morphology. Fn14 signaling can be initiated by either TWEAK binding or forced Fn14 overexpression (7). Our finding of elevated expression of Fn14 in migrating glioma cells suggested that forced overexpression of Fn14 may stimulate glioma cell migration. Glioma cell lines were infected with adenoviruses expressing either a full-length Fn14 receptor (myc-Fn14wt) or a signaling-deficient truncated Fn14 receptor (myc-Fn14tCT; ref. 7). Overexpressed myc-Fn14wt in SF767 and T98G cells caused a 2-fold induction of cell motility (Fig. 1C). Overexpression of Fn14 in glioma cells also resulted in morphologic changes, including neurite-like extensions, membrane ruffling, and lamellipodia formation (Supplementary Data 1). Furthermore, dual staining for myc-Fn14wt and F-actin shows extensive colocalization near the cell perimeter and in the cellular extensions and lamellipodia (Supplementary Data 1). Cells expressing the signaling-deficient myc-Fn14tCT protein migrated poorly (Fig. 1C) and showed no morphologic changes (data not shown), indicating that Fn14 signaling is essential for Fn14-mediated motility.

Fn14 interacts with Rac1 and regulates Rac1 activity. Because Rac1 is essential for lamellipodia formation in glioma cells (15), our observation that Fn14 overexpression increased lamellipodia suggests that Fn14 signaling involves Rac1. To test this hypothesis, we examined whether TWEAK-mediated Fn14 signaling affects the activation of Rac1. T98G cells treated with TWEAK showed high Rac1 activation compared with untreated cells (Fig. 2A). Blocking TWEAK binding to Fn14 by an Fn14-Fc decoy receptor prevented TWEAK-induced Rac1 activation. Overexpression of myc-Fn14wt also induced Rac1 activation, whereas no Rac1 activation was detected in cells expressing the signaling-deficient receptor myc-Fn14tCT (Fig. 2A). The effect of Fn14 on Rac1 is specific because Fn14 signaling did not affect Cdc42 activation (data not shown).

Rac1 coimmunoprecipitates with myc-Fn14wt, but not with myc-Fn14tCT (Fig. 2B), indicating a physical interaction between Fn14 and Rac1. This interaction is independent of Rac1 activation because Fn14 interacts equally well with constitutively active and dominant negative Rac1 (data not shown).

TWEAK treatment caused a decrease in RhoA activation (Fig. 2C), and the Fn14-Fc decoy receptor antagonized TWEAK suppression of RhoA activation. Overexpression of myc-Fn14wt receptor also prevented RhoA activation, whereas signaling-deficient myc-Fn14tCT did not change RhoA activity (Fig. 2C). Immunoprecipitation of Fn14 failed to show a physical association with RhoA (Fig. 2D). These results show that Fn14 interacts and colocalizes with Rac1 and that the TWEAK-Fn14 signaling axis controls Rac1 and RhoA GTPase activity in opposite directions.

Depletion of Rac1 expression by siRNA suppresses Fn14-induced glioma cell migration and invasion. To determine the role Rac1 plays in Fn14-induced glioma migration, we inhibited Rac1 expression in two different glioma cell lines overexpressing the myc-Fn14wt receptor. siRNA-mediated depletion of Rac1 expression in T98G and SF767 cells was ∼90% effective (Fig. 3A) and caused a strong inhibition of Fn14-mediated cell migration (Fig. 3B,, lane d). This level of cell migration inhibition is similar to the level caused by suppression of Fn14 signaling using the Fn14tCT receptor (Fig. 3B , lane f).

The effect of either Fn14 or Rac1 knockdown on cell invasion was examined using an ex vivo organotypic rat brain slice model (24, 25). Invasion of GFP-expressing human glioma cells into the rat brain slices over 48 hours was quantified by confocal microscopy. Depletion of Fn14 (∼80%; Fig. 3C) resulted in a 40% to 50% decrease in cell invasion among the three glioma cell lines examined (Fig. 3D,; lanes d, h, and l) compared with LacZ controls (lanes a, e, and i). Conversely, overexpression of myc-Fn14wt receptor resulted in a 20% to 50% increase in cell invasion (Fig. 3D,; lanes b, f, and j); this invasion was suppressed by Rac1 depletion (Fig. 3D ; lanes c, g, and k). These data corroborate the cell migration results and further support a role for Fn14 in glioma cell invasion mediated through Rac1.

TWEAK regulation of Fn14 expression is dependent on the Rac1/IKKβ/NF-κB signaling pathway. The relationship between Fn14 expression and cell motility raises questions about the molecular and cellular upstream regulators of Fn14 transcription. It was intriguing to contemplate a possible positive feedback loop wherein activation of Fn14 signaling would lead to its own overexpression; therefore, we investigated whether TWEAK could induce Fn14 expression in glioma cells. Increased expression of Fn14 mRNA was detected 2 hours post-TWEAK addition and peaked between 4 and 8 hours (Fig. 4A). TWEAK induction of elevated Fn14 protein (Fig. 4B) was detected with similar kinetics to the changes in mRNA. Because Fn14 signals through Rac1 to increase cell motility, we queried whether TWEAK regulation of Fn14 expression was dependent upon Rac1. T98G cells were treated with Rac1 siRNA, placed in reduced serum, and then TWEAK was added. SiRNA-mediated depletion of Rac1 resulted in a strong reduction of both Rac1 and Fn14 protein expression (Fig. 4C). Depletion of Rac1 expression inhibited TWEAK-induced Fn14 expression (Fig. 4C), suggesting that Rac1 is essential for TWEAK stimulation of Fn14 expression.

Activated Rac1 influences diverse signaling pathways, including the activation of multiple intracellular signaling molecules, such as the c-Jun NH₂-terminal protein kinases (JNK; ref. 27), p38 kinases (28), and the NF-κB transcription factor family (29). Inhibition of JNK by SP-600125 or p38 by SB-239063 in TWEAK-treated glioma cells did not inhibit TWEAK induction of Fn14 (data not shown). However, inactivation of NF-κB in glioma cells by infection with an adenovirus expressing the superrepressor IκBα mutant (IκBαM) strongly diminished TWEAK-induced Fn14 mRNA and protein expression (Fig. 4D and E). Similarly, infection of glioma cells with the dominant-negative Rac1N17 adenovirus inhibited TWEAK-induced Fn14 mRNA and protein expression (Fig. 4D and E).

Because Rac1 has been shown to activate NF-κB via IKKβ phosphorylation (29), we examined the effect of Rac1 inhibition on both IKKβ and IκBα phosphorylation following TWEAK stimulation. T98G glioma cells treated with TWEAK showed a rapid increase in phosphorylation of IKKβ within 5 minutes, after which phosphorylation levels diminished by 1 hour (Fig. 5A). Likewise, IκBα serine phosphorylation was detected within 5 minutes of TWEAK treatment, but the phosphorylation level remained constant for over 2 hours, whereas total IκBα protein was degraded (Fig. 5A). Inhibition of Rac1 function with Rac1N17 protein suppressed TWEAK-induced IKKβ and IκBα phosphorylation (Fig. 5B). Neutralizing TWEAK with the Fn14-Fc decoy receptor also prevented TWEAK-induced IKKβ and IκBα phosphorylation (Fig. 5B). However, expression of dominant-negative Cdc42N17 did not affect Fn14-induced IKKβ phosphorylation (data not shown).

The human Fn14 gene sequence upstream of exon 1 was scanned using Ensemble⁷

and Motif Library Search.⁸ This region contains a NF-κB consensus binding site at −2167 to −2176. We examined NF-κB binding to this motif by analyzing nuclear extracts from untreated or TWEAK-treated T98G cells. No induction of NF-κB binding activity was observed in untreated cells (Fig. 5C,, lane a). Binding activity was robust 2 hours post-TWEAK addition (lane b), whereas, no NF-κB DNA binding activity was detected when TWEAK was first neutralized with the Fn14-Fc decoy receptor. The specificity of this complex was determined by competition assays using excess molar amounts of unbiotinylated wild type (lane d) or normal amounts of biotinylated mutated (lane e) NF-κB probe. In addition, an antibody to the p65 subunit of NF-κB shifted the NF-κB-DNA binding complex to a more slowly migrating form (a “supershift”), confirming the presence of the p65 subunit of NF-κB in the binding complex (Fig. 5C , lane f). These data indicate that TWEAK-induced Fn14 gene expression involves promoter activation by NF-κB.

Fn14 mRNA expression levels correlate with brain tumor grade and poor patient outcome. In an earlier report, we evaluated Fn14 mRNA and protein expression levels in a limited panel of normal brain and brain tumor specimens (10). Here, the clinical relevance of Fn14 overexpression was investigated by analysis of Fn14 mRNA expression in 24 nonneoplastic brain and 160 brain tumor specimens (provided by Dr. Howard Fine, Neuro-Oncology Branch, National Cancer Institute, Bethesda, MD). In normal brain specimens, Fn14 expression is relatively low, but is significantly higher in GBM samples (n = 82) and seems to increase according to tumor grade (Fig. 6A). Levels of TWEAK mRNA did not vary across the clinical samples (data not shown). PC analysis was used to investigate the relationship of Fn14 expression across all tumor samples and patient outcome. Two separate clusters emerged in a three-dimensional scatter plot of PC1, PC2, and PC3 (Supplementary Data 2A). To discern any clinical relevance of the clusters, Kaplan-Meier survival curves were developed for the two clusters. Median survival time of cluster 1 was 952 days (long term), whereas cluster 2 had a median survival of 401 days (short term; Supplementary Data 2B). Specifically, analysis of the Affymetrix expression value for Fn14 in the GBM specimens for each cluster showed that patients with GBM in the short-term survival cluster had higher expression of Fn14 (11.6) than GBM patients in the long-term survival cluster (3.6; P < 0.001; Fig. 6B). These data suggests that high Fn14 expression levels correlate with poor patient outcome.

Immunohistochemical validation of Fn14 protein expression using a brain tumor tissue microarray. We next examined Fn14 protein expression in a series of brain tumor specimens assembled on a brain tissue microarray containing 9 control cases and 62 tumor cases. Examination of Fn14 levels in control nonneoplastic brain specimens from epileptogenic patients (Fig. 6C,, a and d) showed 66.7% negative staining and 33.3% weak staining (Supplementary Data 3), mainly found in normal vascular endothelium and some reactive astrocytes. In contrast, the majority of oligoastrocytomas showed negative to weak expression and 18.2% showed moderate staining (Supplementary Data 3). Similarly, in anaplastic astrocytoma, the majority of the cases showed weak (66.7%) to moderate (33.3%) expression of Fn14. High levels of Fn14 expression were observed in medulloblastoma cases, with 57% moderate and 42% strong positive staining. Similarly, of the 27 GBM cases, 1 tumor was negative (3.7%), 3 were weakly positive (11.1%), 20 were moderately positive (74.1%), and 3 were strongly positive (11.1%). Fn14 immunoreactivity was observed in both glioblastoma tumor core (Fig. 6C,, b and e; arrowhead) and invading cells (Fig. 6C,; e and f, arrow). Tumor vascular endothelium (Fig. 6C , f, V) and normal endothelium (data not shown) had moderate staining for Fn14.

This study shows that Fn14 signaling promotes glioma cell invasion and the small GTPase Rac1 is a key mediator of Fn14-induced cell migration and invasion. We also found that TWEAK binding to Fn14 stimulates transcription of Fn14 through a Rac1/NF-κB–dependent mechanism. Indeed, NF-κB binds to the highly conserved NF-κB consensus site upstream of the Fn14 transcriptional start site. A model of Fn14-mediated self-regulation is shown in Supplementary Data 4.

Elevated NF-κB activity is observed in various cancers including GBM (30, 31). NF-κB activation may contribute to cellular resistance to cytotoxic interventions and promote cell invasion by up-regulating genes involved in cell survival (i.e., BCL-X_L, A1), cell-cell adhesion (i.e., intercellular adhesion molecule-1) and cell-extracellular matrix (ECM) interactions (i.e., tenacin-C, laminin B2 chain; ref. 32). Our finding that NF-κB regulates Fn14 expression is quite intriguing because we have also shown that Fn14 promotes glioma cell survival via the NF-κB pathway that leads to activation of BCL-X_L and BCL-W (9). In glioma, genetic lesions that delete tumor suppressor genes, such as phosphatase and tensin homologue deleted on chromosome 10, or amplify oncogenes such as epidermal growth factor receptor, are likely to contribute to constitutive activation of NF-κB. It seems tenable that the trigger for this positive promotion cycle begins with elevated NF-κB activity driven by these well-described oncogenic changes that progressively lead to up-regulated transcription of Fn14. Activation of Fn14 receptor, in turn, signals to promote further activation of NF-κB and increased Fn14 levels, and consequently cell invasion and robust glioma cell survival. In pilot studies, we note that inhibition of NF-κB activity by either the superrepressor IκBα mutant or the pharmacologic inhibitor SN50 suppresses Fn14-induced cell migration and invasion (data not shown). We are currently probing the genetic signature of the Fn14-NF-κB pathway for points of vulnerability in its control of glioma cell invasion.

Rac1 is essential for various aspects of malignant transformation, including anchorage-independent growth, survival, invasion, and metastasis (21, 33, 34). In glioma cells, depletion of Rac1 expression by siRNA oligonucleotides results in a decrease in cell migration and invasion (33). Additionally, suppression of Rac1 activity via a dominant-negative form of Rac1 induces death in glioma cell lines and primary GBM but not normal human adult astrocytes (35). We report a critical role for Rac1 in Fn14-stimulated migration and invasion of glioma cells. We show that Fn14 activates Rac1 and inactivates RhoA. Rac1 can inhibit RhoA activity (36), suggesting that the suppression of RhoA activity by Fn14 signaling may be a consequence of Fn14-mediated activation of Rac1. In glioma cells, Fn14 binds equally well to active and inactive Rac1, which is in agreement with a previous study by Tanabe et al. (37) using PC12 cells. Our immunoprecipitation studies do not distinguish whether Rac1 associates with Fn14 directly or indirectly as a complex with other proteins. However, the observation that Rac1 does not associate with the Fn14tCT protein suggests either that Rac1 associates with Fn14 through TRAF proteins or via signaling proteins that act downstream of TRAF; the coimmunoprecipitation of Rac1 with Fn14 argues for the former. A study by Min and Pober (38) showed that TRAF2 interacts with Rac1 to promote downstream JNK activation. Additionally, yeast two-hybrid studies have shown that TRAF1, TRAF2, TRAF3, and TRAF5 interact with Fn14 at the TRAF binding region, which links the Fn14 receptor to transcription factors such as NF-κB (7). Current studies are ongoing to explore the mechanism of interaction between Fn14 and Rac1.

Although local dissemination of glioma cells is a principal contributor to the poor clinical outcome for these patients, a therapeutic approach targeting these invading glioma cells remains elusive. The results of an extended investigation into the clinical relevance of Fn14 point to Fn14 as an attractive candidate for such targeted therapy. We report that the levels of expression of Fn14 in glial tumors correlate with tumor grade; advanced GBM showed the highest expression of Fn14 mRNA and protein. Examination of Fn14 protein expression in situ portrays strong Fn14 staining in cells at the site of active invasion compared with tumor core. This notion of microregional changes in Fn14 expression driving invasion is supported by the 4- to 6-fold increase in Fn14 mRNA in invading cells collected by laser capture microdissection from GBM specimens. This corroborates the in vivo immunohistochemistry observations. More compelling is the finding that Fn14 expression is an indicator of poor outcome, because GBM patients in short-term survival expressed significantly increased Fn14 levels compared with GBM patients with long-term survival (Fig. 6B). Thus, the level of Fn14 expression may be a good predictor of survival outcome, complimenting other prognostic indicators.

This study suggests that Fn14 may be a key driver of the malignant invasive behavior for glial tumors. Fn14 activation promotes dispersion of the tumor, whereas also engaging biochemical mechanisms leading to enhanced cell survival. Our discovery that Fn14 signaling includes the downstream mediators Rac1 and NF-κB, whose activation by Fn14 subsequently triggers transcription of Fn14, is among the first positive feedback signaling events described for human malignancies, and offers a model for the biological basis of spontaneous malignant progression. We argue that interference with Fn14 activation and its signaling are rational therapeutic targets for glioma.

Grant support: NIH grants NS-42262 (M.E. Berens), HL-39727 (J.A. Winkles), CA87567 (M. Symons), and CA103956 (J.C. Loftus); Ruth L. Kirschstein National Research Service Award F32 CA112986-01 (N.L. Tran); and a Russell Becker-American Brain Tumor Association grant (N.L. Tran).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Howard Fine for providing access to the gene expression profiles of human nonneoplasmic and brain tumor specimens.