Inhibition of NF-kB Signaling Ablates the Invasive Phenotypeof Glioblastoma
Mike-AndrewWesthoff1, Shaoxia Zhou2, Lisa Nonnenmacher1, Georg Karpel-Massler1,3, Claudia Jennewein1,Matthias Schneider1,3, Marc-Eric Halatsch3, Neil O. Carragher7, Bernd Baumann4, Alexander Krause6,Thomas Simmet5, Max G. Bachem2, Christian R. Wirtz3, and Klaus-Michael Debatin1
AbstractGlioblastoma multiforme, the most common primary brain tumor, is highly refractory to therapy, mainly due to
its ability to form micrometastases, which are small clusters or individual cells that rapidly transverse the brain andmake full surgical resection impossible. Here, it is demonstrated that the invasive phenotype of glioblastomamultiforme is orchestrated by the transcription factor NF-kB which, via metalloproteinases (MMP), regulatesfibronectin processing. Both, cell lines and tumor stem cells from primary glioblastoma multiforme, secrete highlevels of fibronectin which when cleaved by MMPs forms an extracellular substrate. Subsequently, forming andinteracting with their own microenvironment, glioblastoma multiforme cells are licensed to invade their surround-ings. Mechanistic study revealed that NF-kB inhibition, either genetically or pharmacologically, by treatment withDisulfiram, significantly abolished the invasive phenotype in the chick chorioallantoic membrane assay. Further-more, having delineated the underlying molecular mechanism of glioblastoma multiforme invasion, the potential ofa disulfiram-based therapy was revealed in a highly invasive orthotrophic glioblastoma multiforme mouse model.
Implications: This study defines a novel therapeutic approach that inhibits micrometastases invasion and revertslethal glioblastoma into a less aggressive disease. Mol Cancer Res; 11(12); 1611–23. �2013 AACR.
IntroductionGlioblastomamultiforme is themost common tumorof the
central nervous system (1). The current standard of careconsists of tumor resection followed by radiotherapy and acourse of temozolomide (2), but despite intense efforts, glio-blastoma multiforme remains one of the most lethal tumorswith a mean patient survival of 14 months (3). The highmortality of patients is partially due to the particular growthpattern of this malignancy. Indeed, the presence of micro-metastases intheabsenceofadistincttumormassissufficienttocause progressive neurologic dysfunctions and even death (4).Glioblastoma multiforme grow diffusely and are highly
invasive, infiltrating the surrounding brain tissue, thus
making localized treatment, e.g., surgery, particularly inef-fective (1). Therapeutic interventions to ablate the invasivephenotype of glioblastoma multiforme have so far proved tobe insufficient, as systemic chemotherapy or whole brainirradiation have failed to eradicate invasive cells and micro-metastases (4). It is therefore imperative for any new ther-apeutic intervention to consider this highly invasive nature ofglioblastoma multiforme and ablate this aggressive pheno-type as efficiently as possible.Although the NF-kB pathway is frequently found aber-
rantly activated in glioblastoma multiforme (5), its role inglioblastoma multiforme biology remains elusive. Whilesome findings point to a role in mediating DNA-damagerepair (6) and modulating sensitivity to death receptor–mediated apoptosis (7), it has also been shown that inglioblastoma multiforme NF-kB does not play a role inproliferation and resistance toward therapeutic interventionbased on chemotherapy (8). Recent evidence points to apossible role of this transcription factor in the tumor-inva-sion potential of glioblastoma multiforme cells (9, 10).However, so far, neither the precise molecular mechanismsof this role nor the potential therapeutic benefits of thesefindings have been systematically addressed.
Materials and MethodsCell cultureU87MG cells from American Type Culture Collec-
tion (ATCC) were maintained in Dulbecco's ModifiedEagle Medium (DMEM; Gibco, Life Technologies, Inc.),
Authors' Affiliations: 1Departments of Pediatrics and Adolescent Med-icine, 2Clinical Chemistry, and 3Neurosurgery, University Medical CenterUlm; 4Institute of Physiological Chemistry; 5Institute of Pharmacology ofNatural Products & Clinical Pharmacology, Ulm University, Ulm, Germany;6AHF analysentechnik AG, T€ubingen, Germany; and 7Edinburgh CancerResearch UK Centre, Institute of Genetics and Molecular Medicine, Uni-versity of Edinburgh, Edinburgh, United Kingdom
Note: Supplementary data for this article are available at Molecular CancerResearch Online (http://mcr.aacrjournals.org/).
L. Nonnenmacher and G. Karpel-Massler contributed equally to this work.
Corresponding Author: Klaus-Michael Debatin, Department of Pediatricsand Adolescent Medicine, University Medical Center Ulm, Eythstrasse 24,D-89075 Ulm, Germany. Phone: 49-7315-000; E-mail:[email protected]
supplemented with 10% fetal calf serum (FCS; Biochrom),1 mmol/L glutamine (Biochrom), and 1% penicillin/strep-tomycin (Biochrom). Control and IkBa superrepressor (SR)cells were created as previously described (6).Patient-derived glioblastoma spheres were grown from
surgical specimens, after patients' consent was obtained.Specimens were minced and taken up in ice-cold PBS(Biochrom), followed by 5-minute centrifugation at 345�g/room temperature. After discharging the liquid, thetumor pellet was taken up in 5 mL TrypLE Express (Gibco,Life Technologies) and incubated for 5 minutes. Tumorswere then filtered through a sieve (pore size, 70 mm) andtaken up in DMEM/F-12 (HAM) medium (Gibco, LifeTechnologies), supplemented with L-glutamine, EGF (Bio-mol GmbH), basic fibroblast growth factor (bFGF;MiltenyiBiotec GmbH), and B27 (Gibco, Life Technologies).
Adhesion assayThe trypsin-based adhesion assay was performed as pre-
viously described (11).
Time-lapse photography and wind-rose analysisTime-lapse photography was performed as previously
described (11). Final analysis and wind-rose depiction wereperformed with ImageJ (Rasband, W.S., ImageJ; http://imagej.nih.gov/ij/, 1997–2011).
Migration assayCellular migration was determined after 12 hours by
analyzing cells migrating through either a collagen-coated(BD Biosciences) or a fibronectin-coated (Sigma-Aldrich)membrane in an 8-mm pore Costar Transwell insert (Corn-ing Inc.). Medium containing 20% FCS served as chemoat-tractant, while cells were seeded in medium containing 10%FCS.
Biologic invasion assayThe invasion of cells into a biologic structure was assayed
via the chorioallantoic membrane (CAM) assay (12). Briefly,1 � 106 cells were seeded in a 1:1 mixture of serum-freemedium andMatrigel (BDBiosciences) onto the CAMof 1-week-old fertilized eggs. Four days after seeding, the tumorand surrounding CAM were extracted, embedded in paraf-fin, and cut.
Tissue immunohistologyAvian, murine, and human tissue samples were stained for
hematoxylin and eosin. In addition, staining for Vimentin(Abcam plc) and Fibronectin (Dako Deutschland GmbH)were performed, according to the previously publishedprotocol (12).
Western blot analysis and integrin surface expressionWestern blot analysis and integrin surface expression were
performed as previously described (11), using the followingantibodies: mouse monoclonal anti-Vinculin (Sigma-Aldrich), mouse monoclonal anti-Src (Cell ApplicationsInc.), rabbit polyclonal anti-FAK (Millipore), mouse mono-
clonal anti-uPAR (American Diagnostica), mouse mono-clonal anti-a-tubulin (Millipore), mouse monoclonal anti-b-actin (Sigma-Aldrich), anti-mouse and anti-rabbit IgG-HRP (Santa Cruz Biotechonlogy) and APC-labeled anti-CD29 (BD Biosciences) for integrin b1.
Fluorescence microscopyFluorescence microscopy was performed as previously
described (11), using the following antibodies: rabbitpolyclonal anti-p65 (Santa Cruz Biotechnology) TexasRed anti-rabbit (Vector Laboratories), Texas Red anti-mouse (Vector Laboratories), and fluorescein isothiocya-nate (FITC) anti-rabbit (Milipore), as well as 40,6-diamidino-2-phenylindole (DAPI; Roche Diagnostics) orHoechst 33258 (Sigma-Aldrich) to counterstain the nucleiand phalloidin-TRITC (Sigma-Aldrich) to visualize theactin cytoskeleton. For fibronectin staining, rabbit anti-fibronectin (Siemens Healthcare Diagnostics GmbH) wasused in conjuncture with anti-rabbit immunoglobulin G(IgG) Biotin (Dako Deutschland GmbH) and SA-AlexaFluor R568 (Molecular Probes, Life Technologies).Pictures were taken with an AX70 "Provis" microscope(Olympus).To present pictures more clearly, the depiction of the actin
cytoskeleton in Fig. 2C was performed similarly as describedearlier, but using a laser scanning confocal microscope (LeicaDM IRB).
Analysis of the actin cytoskeletonThe amount of intact cytoskeleton was determined by
analyzing the ratio of G-actin to F-actin using the G-actin/F-actin In Vivo Assay Kit (Cytoskeleton, Co) according tothe manufacturer's instructions. The densitometric analysiswas performed with ImageJ.
Quantitative determination of Fibronectin synthesisThe amount of secreted Fibronectin was determined as
previously described (11).
Electrophoretic mobility shift assayNuclear extracts were prepared according to a previously
established protocol (8) using following g-[32P]-ATP-labeled oligomer: 50-AGTTGAGGGGGACTTTTCCCA-GGC-30.
Proteolytic cleavage of fibronectin by metalloproteinasesThe recombinant catalytic domains of metalloproteinase
(MMP)-2 and MMP-9 (ProSpec) were added upon seedingat a 1:10 dilution to 100 mL medium containing 6.0 � 104
cells. Cells were allowed to settle for approximately 1 hourbefore fixation.
MMP quantificationMMPs were quantified by Fluorokine Multianalyte Pro-
filing (MAP) kits following the manufacturer's instructions(R&D Systems), using following kits: Human FluorokineMAP Base Kit, MMP Panel, Human MMP-2 FluorokineMAP, and Human MMP-9 Fluorokine MAP.
Westhoff et al.
Mol Cancer Res; 11(12) December 2013 Molecular Cancer Research1612
Gelatin zymographyZymography was performed as previously described (13).
Apoptosis measurementThe apoptosis readout was DNA fragmentation as
assessed by fluorescence-activated cell-sorting (FACScan;Becton Dickinson) analysis of propidium iodide–stainednuclei as previously described (11).
Changes in cell numberCells were seeded and allowed to proliferate for indicated
times. This was followed by prolonged treatment with atrypsin/EDTA solution (Biochrom) until all cells were insuspension. The cell suspension was diluted 1:100 in CASY-ton solution (Innovatis) and cell numbers were then deter-mined using CASY1 DT (Innovatis).
Chemical and pharmacologic inhibitorsDisulfiram was obtained from Sigma-Aldrich, whereas
Antabuse (Actavis Group PTC) was used for the ortho-tropic mouse experiments. Cells were seeded at a concen-tration of 1 mmol/L GM1489 (Santa Cruz Biotechnolo-gy), 1 mmol/L RGD peptide (Enzo Life Science), and 200pmol/L VPLCK (D-Val-Phe-Lys Chloromethyl Ketone;Merck Chemicals).
Orthotrophic mouse modelHuman glioblastoma multiforme cells were transplanted
into mouse brains as previously described (14). Briefly, 1.0�105 cells were stereotactically implanted into the right stri-atum of NOD.Cg-PrkdcscidII2rgtm1WjI/SzJ mice. Animalexperiments were approved by the Regierungspr€asidiumT€ubingen, Germany.
Statistical analysisStatistical analysis was carried out by a two-sided Student t
test, unless stated otherwise.
ResultsNF-kB inhibition reduces of glioblastoma multiformecell/microenvironment interactionAfter recent suggestions (9) that in glioblastoma multi-
forme cells, NF-kB might contribute to cell invasion, wefurther investigated this intriguing possibility usingU87MGcells stably transfected with a previously described IkBasuperrepressor (SR) construct that inhibits NF-kB activity(refs. 6, 8; Supplementary Fig. S1A). Although proliferationand spontaneous apoptosis seemed unaffected (Supplemen-tary Fig. S1B and S1C), the SR cells displayed a strikingretardation in cell spreading (Fig. 1A) and detached morereadily after trypsin treatment (Fig. 1B). This diminishedadhesion also affects cell locomotion, i.e., undirected move-ment (Supplementary Fig. S1D): control cells travel in aroughly linear fashion, SR cells move more randomly,frequently altering direction, in particular after reattachmentupon transient loss of cell–substrate interaction (Supple-mentary Fig. S1D, arrowheads).
Although a scratch assay suggested that SR cells alsobehave differently from control cells in migration, i.e.,directed movement control cells (Supplementary Fig.S1E), we further confirmed this by ascertaining the migra-tory capacity of the cells through a collagen-coated mem-brane. As shown in Fig. 1C, control cells migrate signifi-cantly faster through the membrane, suggesting that theirincreased interaction with their substrate leads to increasedmotility, despite potentially reduced locomotion. To verifythat these findings are of physiologic relevance, we used theso-called chick CAM assay (15). Intriguingly, although bothcells remained viable and appeared to proliferate on theCAM, only the control cells exhibited an invasive pheno-type, whereas the SR cells did not grow into the CAM (Fig.1D). Taken together, these data strongly argue for a criticalrole of NF-kB–mediated signaling in glioblastoma multi-forme invasion.
IkBa superrepressor–expressing cells do not efficientlyprocess FibronectinBefore investigating the underlying molecular mechan-
isms that link NF-kB signaling to migration/invasion, weindependently confirmed the role of NF-kB in a secondglioblastoma multiforme cell line, T98G, expressing thecontrol vector or the IkBa superrepressor (SupplementaryFig. S2).Next, we concentrated on integrin-mediated adhesion,
which we have previously shown to play a role in glioblas-toma multiforme (11). Although the protein expression ofseveral focal adhesion proteins, as well as the surface expres-sion of the b1 integrin subunit did not differ between controland SR cells (Fig. 2A and B, respectively), when visualizingthe actin cytoskeleton, which links focal adhesions to cellmorphology andmotility (16), we found that SR cells almostcompletely lacked organized stress fibers (Fig. 2C). Tofurther verify this, we separated filamentous F-actin fromglobular G-actin and compared the ratio of these two (Fig.2D). To assess whether focal adhesions formed and wereunable to connect to the cytoskeleton, or whether cellscannot form focal adhesions and thus do not provide atethering point for the actin stress fibers, we visualized thecellular localization of the focal adhesion. Interestingly, thetypical punctuate staining, which is indicative of focaladhesion (Fig. 2E, first column), is almost completely lack-ing in SR cells (Fig. 2E, second column). However, thisdifference does not seem to be due to the SR cells' intrinsicinability to form focal adhesions, because when providedwith an external fibronectin matrix, they are fully capable offorming focal adhesions (Fig. 2E, third column). Impor-tantly, the failure to interact with their substratum seemsto be a specific effect and not a general deficit of the SRcells to interact with their environment, as the SR cells'ability to form adherens junctions is unaffected (Supple-mentary Fig. S3A).To verify that the differences inmigration are due to altered
focal adhesion formation/actin organization in the absence offibronectin, we repeated the Transwell assay shown in Fig.1D with a fibronectin-coated membrane (Fig. 3A). When
Inhibition of Glioblastoma Invasion
www.aacrjournals.org Mol Cancer Res; 11(12) December 2013 1613
provided with an external source of fibronectin matrix, bothcontrol and SR cells migrate equally well through the mem-brane (Fig. 3A), suggesting that, indeed, the differences inenvironmental interaction of those cells are due to the SRcells' inability to provide their own fibronectin matrix. As aputative NF-kB response element was found in the fibro-nectin gene (17), we initially speculated that SR cells mightproduce and thus secrete less fibronectin; this, however, isclearly not the case (Fig. 3B). As the distinct SR phenotype islost upon providing the cells with an artificial fibronectinmatrix, SR cells might be incapable of processing intrinsicfibronectin efficiently and are thus unable to incorporate itinto a matrix. To analyze this hypothesis, we stained cells forfibronectin (Fig. 3C). Whereas control cells produce longstrands offibronectin (Fig. 3C, top, gray arrowheads) and alsolay down the beginning of an extracellular fibronectin matrix(Fig. 3C, top, white arrowheads), these features seem to beabsent in the SR cells (Fig. 3C, bottom).
AsMMPs have been implicated in the invasive phenotypeof glioblastoma multiforme (10) and are also regulated byNF-kB (18–20), we next investigated whether they areinvolved in NF-kB–dependent processing of fibronectin.Therefore, we added a recombinant catalytic MMP domainto the cells before seeding and analyzed fibronectin-proces-sing the next day (Fig. 3D). Upon addition of the recom-binant catalytic MMP domain, SR cells appear more spreadout and a reduction in fibronectin nodules is visible (Fig.3D). To further confirm a direct connection betweenMMPactivity and fibronectin processing, we again repeated theTranswell assay with a collagen-coated membrane and in thepresence of either the RGD peptide, that blocks the inter-action between fibronectin and integrins, or the pharmaco-logic inhibitor GM1489, which is highly specific forMMPs,or both (Fig. 3E). Each substance reducedmigration by 40%to 50%, but, importantly, combining both had no additiveeffect (Fig. 3E), suggesting that there is a linear relationship
Figure 1. NFkB is essential for glioblastoma invasion. A, equal numbers of either U87MGcontrol cells (control) or cells expressing an IkBa superrepressor (SR)were plated on cell culture–treated plastic and allowed to settle for 8 hours. B, equal concentrations of either U87MG control cells (control) or cellsexpressing an IkBa superrepressor (SR) were plated and allowed to settle overnight. Cells were treated for 5minuteswith a trypsin/EDTA solution, after whichsuspended cells were counted. C, equal numbers of either U87MG control cells (control) or cells expressing an IkBa superrepressor (SR) were seededon collagen-coated membranes and allowed to transmigrate for 24 hours, with medium containing 20% serum serving as chemoattractant. Cells wereDAPI-stained before counting. Right, representative pictures of transmigrated cells. D, equal numbers of either U87MG control cells (control) or cellsexpressing an IkBa superrepressor (SR) were seeded onto CAM as described in the Materials and Methods section, and allowed to grow for 4 days.Representative sections of paraffin-embedded tumors are shown, dotted lines indicate the interface CAM/tumor. Shown in A and D (right) are representativedata of at least two independent experiments, whereas in B andC themeanþSEMof three independent experiments performed in triplicate are shown. Scalebar in A and C, 0.2 mm; in D, 1 mm.
Westhoff et al.
Mol Cancer Res; 11(12) December 2013 Molecular Cancer Research1614
between both targets. As predicted by this emerging model,combining those two inhibitors has no additional effect onSR cells (Supplementary Fig. S3B).Next, we investigated whether MMP expression (Fig. 3F)
or activity (Fig. 3G) differed between control and SR cells,focusing on MMP-2, together with MMP-9, the dominantMMP in glioblastoma (10, 21). As we saw clear difference inMMP activity, but not expression level, we also looked atfurther molecules through which NF-kB might regulateMMP activation. Although MT1-MMP (22) and tissueinhibitor of metalloproteinases (TIMP; ref. 23) were notdynamically regulated when comparing control with SR cells(Supplementary Fig. S3C and S3D, respectively), we foundclear differences in uPAR expression (Fig. 3H). In particular,we found that SR cells expressed two high molecular weightbands. Although additional bands are often suggested to be
due to glycosylation, it has also been reported that uPARexists in different isoforms, which can be either membrane-bound or secreted (24). This is of particular interest, as NF-kBbinds to the uPARpromoter (25) andMMP-2 and -9 canbe activated by plasmin-activated uPAR-bound uPA (26),which has also been shown to play a role in glioblastomainvasion (27). To confirm the involvement of uPAR in NF-kB–mediatedMMP-2 activation, we used the highly specificplasmin inhibitor VPLCK (Fig. 3I). VPLCK strongly inhib-ited transmigration through a collagen-coated membrane,but combining VPLCK and GM1489 had no additionaleffect (Fig. 3I), again suggesting a linear relationship.Taken together, these data show that glioblastoma multi-
forme cells activateMMPs viaNF-kB–dependent regulationof the uPA/uPAR complex and thereby process secretedfibronectin. This allows the cells to form an extracellular
Figure 2. U87MG glioblastoma multiforme expressing IkBa superrepressor have a disorganized actin cytoskeleton and do not form focal adhesions. A,Westernblot analysis comparing the expressionof keyproteinsmediating focal adhesion in control (control) and cells expressing an IkBa superrepressor (SR),b-actin served as loading control. B, comparison by flow cytometry of surface expression of integrin subunits b1 on either control cells (control) or cellexpressing an IkBa superrepressor (SR). C, equal numbers of either U87MG control cells (control) or cells expressing an IkBa superrepressor (SR) wereseeded on glass slides and allowed to adhere overnight. This was followed by fixation and staining the actin cytoskeleton with phalloidin-TRITC andanalyzing the cells by confocal microscopy. D, Western blot analysis after separating F- and G-actin in control (control) or cells expressing an IkBasuperrepressor (SR), as described in the Materials & Methods section. Right, quantitative presentation of the percentage of F- and G-actin normalizedto total actin. E, equal numbers of either U87MG control cells (control) or cells expressing an IkBa superrepressor (SR) were seeded on glass slides orfibronectin (Fn)-coated glass slides and allowed to adhere overnight. This was followed by fixation, staining of the nuclei (blue), various indicated proteins(green), and the actin structures (red). Cells were analyzed by florescence microscopy. Bottom, quantitative analysis of cells expressing features of focaladhesions (FA). Shown are representative results of at least two independent experiments, quantification in E was performed on three independentexperiments performed in triplicate, approximately 100 cells were counted for each datum. Scale bar, 0.2 mm.
Inhibition of Glioblastoma Invasion
www.aacrjournals.org Mol Cancer Res; 11(12) December 2013 1615
matrix to which the cells then adhere strongly via focaladhesions. These focal adhesions also form tethering pointsfor the actin cytoskeleton, allowing the cells to spreadrapidly. Our data suggest that glioblastoma multiforme cellsactivate NF-kB upon encountering an unfavorable micro-environment, experimentally simulated by keeping cells insuspension (Supplementary Fig. S3E).
Therapeutic potential of NF-kB inhibition inglioblastoma multiformeTo assess whether NF-kB inhibition might be of thera-
peutic benefit in glioblastoma multiforme treatment, weused disulfiram, the active component of Antabuse, whichwas originally described as a well-tolerated inhibitor ofacetaldehyde metabolism (28–30). It can cross the blood-brain-barrier (30–33) and has been shown to inhibit inva-sion via modulation of MMPs (34, 35), although this effecthas so far not been linked to NF-kB signaling and fibro-nectin, and is already in clinical use.First, we assessed whether disulfiram inhibits NF-kB
signaling in glioblastoma multiforme cells. Treatment withDisulfiram leads to inhibition of both basal and TNF-a–induced NF-kB signaling, as shown by localization ofthe p65 subunit (Fig. 4A) and the transcription factor'sbinding toDNA (Fig. 4B). Disulfiram has a significant effecton cell numbers (Fig. 4C), presumably due to its strongintrinsic capacity to induce apoptosis (Fig. 4D). Apoptosis islikely to be initiated by anoikis, as cells detach (Supplemen-tary Fig. S4A). This effect of disulfiram is specifically due toNF-kB inhibition, as it is much less marked in SR cells(Supplementary Fig. S4B) and transient, as cells that havebeen seeded for longer and thus, presumably, formed theirown extracellular matrix do not detach as readily as freshlyseeded cells (Supplementary Fig. S4C). In addition, thedistinct punctuate staining of the focal adhesion and thefibrous actin cytoskeleton are also lost upon disulfiramtreatment (Supplementary Fig. S4D). This set of dataindicates that the SR construct and disulfiram have similar,but not identical mechanisms of action upon cellular behav-
ior. This notion is further confirmed by antibody-basedproteomic screening (Supplementary Fig. S5).Next, we confirmed whether NF-kB inhibition by disul-
firam has the desired function effects on transmigration (Fig.4E) and invasion (Fig. 4F). As our data so far were obtainedfrom cell lines thatmight not fully reflect the genomic profileof glioblastoma multiforme in vivo (36), we turned to threeprimary-cultured tumor initiating cells (Supplementary Fig.S6A), which have been cultured as spheres, thus retaining theoriginal tumor-expression profile (36). Two of these cellpopulations, G38 and G40, were obtained from patientswith glioblastoma multiforme, whereas G55 has been iden-tified as gliosarcoma. As G40 cells secreted the highestamount of fibronectin (Supplementary Fig. S6B), we inves-tigated whether we could adapt the previously describedorthotrophic mouse model (14) for G40 cells and treat themice from day 16 postoperation, every second day withdisulfiram. Interestingly, tumor bulk could not be found inbrains of mice harboring G40 cells, although abnormal cellswere clearly present (Supplementary Fig. S6C) throughoutthe whole murine brain, to no lesser extent than G38 cells,although G40 appear to micrometastasize in smaller groupsor individual cells (Supplementary Fig. S6D). Disulfiramtreatment led to an increase in cell-clusters' size (Supple-mentary Fig. S6E and S6F), suggesting the possibility that itcan efficiently inhibit further migration after cell division. Itseems unlikely that disulfiram has increased proliferation ofglioblastoma multiforme as mice appeared healthy whensacrificed. As cell morphology and growth rates of G38 andG40 cells are similar in vitro (Supplementary Fig. S6G–S6J),it is tempting to speculate that this marked difference ingrowth pattern in vivo is due to the striking difference infibronectin secretion between those cells (SupplementaryFig. S6B).Next, we repeated the in vivo experiment using the G38
cells (Fig. 5A). Twenty-seven days postoperation, the micewere sacrificed, as controlmice began to show symptoms andthe brain regions parietal of the tumor bulk were analyzed formicrometastasis (Fig. 5B). Number, size, and position of cell
Figure 3. Fibronectin processing is essential for the interaction of U87MG glioblastomamultiforme with their environment. A, equal numbers of either U87MGcontrol cells (control) or cells expressing an IkBa superrepressor (SR) were seeded on fibronectin-coated membranes and allowed to transmigrate for 24hours, with medium containing 20% serum serving as chemoattractant. B, equal numbers of either U87MG control cells (control) or cells expressing an IkBasuperrepressor (SR) were plated on cell culture–treated plastic. After 24 and 48 hours, both supernatant and adherent cells were collected. The secretedfibronectin was quantified via ELISA and put into relation to total DNA content. C, equal numbers of either U87MG control cells (control) or cells expressing anIkBa superrepressor (SR) were plated on glass slides and allowed to settle overnight. This was followed by fixation and staining of fibronectin. Grayarrowheads indicate strong fibronectin strands, white arrowheads show extracellular fibronectin. D, equal numbers of either U87MG control cells (control) orcells expressing an IkBa superrepressor (SR) were incubated for 1 hour withMMP-2 andMMP-9. Cells were then fixed and their nucleus (blue) and fibronectin(red) stained. E, U87MG cells were seeded on collagen-coated membranes, either in the absence of pharmacologic inhibitors, in the presence ofeither the integrin/fibronectin interaction-blocking RGD peptide, or GM1489, which inhibits MMP activity, or both. Cells were allowed to transmigrate for 24hours, withmedium containing 20%serum serving as chemoattractant, then fixed, and their nuclei stainedwith DAPI before counting. F, the secretedMMP-2concentration was determined for either control U87MGcells (control) or cells expressing an IkBa superrepressor (SR) seeded for 24 hours on either plastic orfibronectin (Fn). G, zymograph depictingMMP-2 and -9 of either control U87MG cells (control) or cells expressing an IkBa superrepressor (SR) seeded for 24hours. Adjusted to DNA content of the samples, equal amounts of supernatant were loaded. H, Western blot analysis comparing the expression andmodification of uPAR in control (control) and cells expressing an IkBa superrepressor (SR), glyceraldehyde—3—phosphate dehydrogenase (GAPDH) servedas loading control. I, U87MG cells were seeded on collagen-coated membranes, either in the absence of pharmacologic inhibitors, in the presenceof the plasmin inhibitor VPLCK, GM1489, or both. Cells were allowed to transmigrate for 24 hours, with medium containing 20% serum serving aschemoattractant. Shown in A are the mean þ SEM of three independent experiments performed in triplicate; in B and F, the meanþ SD of two independentexperiments performed in triplicate; in C, D, G, and H, representative pictures of at least two independent experiments. E and I, the mean þ SD ofthree independent experiments is depicted. Scale bar in C and D, 0.02 mm.
Inhibition of Glioblastoma Invasion
www.aacrjournals.org Mol Cancer Res; 11(12) December 2013 1617
clusters were then compared between untreated and disul-firam-treated tumors (Fig. 5C and D), until cell clustersexceeded at least 100 cells, as these clusters cannot be safelyreferred to as micrometastasis. Over comparable brain areas,untreated tumors formed 448 clusters consisting of 2,051cells, whereas disulfiram treatment reduced the amount ofclusters to 19, consisting of 247 in total. Furthermore, tumorcells in untreated mice were found more parietal, i.e.,invaded further, than in treated mice (Fig. 5D). Althoughtumor bulk (Fig. 5E), as well as range and amount ofmicrometastasis (Fig. 5D and F, respectively) were clearlyreduced by disulfiram treatment, the size of individual cellclusters was actually significantly increased by disulfiram,from 4.6 to 13 cells per cluster (P value ¼ 0.04, one-sidedMann–Whitney U test). These data suggested that disulfi-ram can successfully inhibit glioblastomamultiformemicro-metastasis and tumor growth.
Finally, we analyzed human brain sections of 2 patientsafflicted with glioblastoma multiforme (Fig. 6) to assessthe relevance of our findings in a clinical setting. Impor-tantly, the four key features necessary for veracity of ourmodel were also present in these sections: (i) fibronectinexpression is enriched at the leading/invasive edge of thetumor compared with tumor bulk (Fig. 6A). (ii) Cells ofthe vascular system expressing fibronectin, the only area inhealthy brain that has been reported to express fibronectin(37), are clearly distinguishable from tumor cells (Fig. 6B).(iii) Fibronectin was not uniformly expressed/secreted inmany of those cells, suggestive of directional incorporationof this matrix compound into the microenvironment (Fig.6B and C). (iv) Several fibronectin-positive cells seemed tobe associated with "trails" of this substance. These "trails"probably indicate the route taken by the invasive cells (Fig.6B and C).
Figure 4. Disulfiram inhibits NF-kB signaling in U87MG glioblastoma multiforme cells. A, U87MG cells were seeded on glass slides and allowed to adhereovernight. Cells were then stimulated with 0 [dimethyl sulfoxide (DMSO) control], 20, and 40 mmol/L disulfiram for 2 hours, 50 ng/mL TNF-a for 1 hour, or acombination of the two substances. This was followed by fixation and staining for the p65 subunit of NF-kB. B, U87MGcells were treated as described earlierand the DNA-binding capacity of NF-kB was assessed via the electrophoretic mobility shift assay. C, U87MG cells were seeded and treated with 0 (DMSOcontrol), 20, and 40 mmol/L disulfiram 24 hours later. Cells then were allowed to grow for additional 24 (i.e., 48 hours after seeding) or 72 (i.e., 96 hoursafter seeding) hours, followed by counting. D, percentage of apoptosis was determined by fluorescence-activated cell-sorting analysis of propidium iodide–stainednuclei after 24 and48hours treatmentwith 0 (DMSOcontrol), 20, and40mmol/L disulfiram.E,U87MGcellswere treatedwithwith 0 (DMSOcontrol), 20,and 40 mmol/L disulfiram and seeded on collagen-coated membranes and allowed to transmigrate for 24 hours, with medium containing 20% serumserving as chemoattractant. F, U87MGcells, either pretreated for 2 hours before seedingwithDMSO (DMSO) or 40mmol/L disulfiram (disulfiram),were seededonto CAM and allowed to grow for 4 days. Treatment during this time was twice a day DMSO or 40 mmol/L disulfiram. Representative sections of paraffin-embedded tumors are shown, dotted lines indicate the interface CAM/tumor. Shown are representative pictures of two independent experiments. Scale bar,1 mm. Shown in A, B, and F are representative pictures of at least two independent experiments, in C and D, the mean þ SEM of three independentexperiments performed in triplicate are depicted, in E the mean þ SD of three independent experiments. Scale bar in A and E, 0.2 mm; in F, 1 mm.
Westhoff et al.
Mol Cancer Res; 11(12) December 2013 Molecular Cancer Research1618
DiscussionIn this study, we delineate the molecular mechanisms that
allow glioblastoma multiforme cells to alter their microen-vironment so as to increase their invasive potential (Fig. 7).Furthermore, we suggest a possible therapeutic interventionthat is able to prevent glioblastoma multiforme invasion/micrometastasis, the most difficult to treat and deadliestfeature of this malignancy (3). In detail, we found that thesuggested link between NF-kB activity and glioblastomamultiforme invasiveness is due to the processing of fibro-nectin by MMPs, which thus permits incorporation of thismatrix component directly into the surrounding of theinvading tumor cells.MMP-2 is processed and thus activated
in an uPAR-dependent manner, which in turn is known tobe regulated by NF-kB (38–40). Although the individualsteps of this signaling cascade were previously proposed, theyhad so far not been demonstrated in glioblastoma multi-forme. More importantly, however, the involvement ofMMPs in fibronectin processing has so far only been viewedin the context of invasion facilitated by extracellular matrixdestruction. Here, we present compelling evidence that theopposite can also occur, invasion mediated by fibronectinprocessing that leads to novel matrix formation.Interestingly, these findings might also shed some light
onto themore generalmechanisms offibronectin processing,about which only little is known. Soluble Fibronectin exists
Figure 5. Disulfiram also inhibits NF-kB in primary tumor-initiating glioblastoma multiforme cells. A, treatment scheme for therapeutic intervention in mouseharboring a G38-induced glioblastomamultiforme. A total of 0.5� 105 cells were orthotopically injected into mouse brains. After 16 days of tumor growth, 20mgdisulfiram are injected on day 16 and 17, then every second day until the experiment is terminated on day 27 (due to symptoms in the controlmice). B, rightlateral representation of amouse brain, with the approximate position of the tumor depicted in blue. Ten 4 mmaxial sections were analyzed, covering the areaparietal of themain tumor bulk (shown in red). Individual sectionswere spaced at least 50 mmapart. C, frequency of cell cluster occurrences in ten comparablebrain sections of either control mice or disulfiram-treated mice. Clusters were grouped into three categories: small (1–10 cells/cluster), intermediate(11–100 cells/cluster), and large (more than 100 cells/cluster). D, frequency and relative position of cell clusters (independent of size) in ten comparable brainsections of either control mice or disulfiram-treated mice. E, maximal tumor bulk of either untreated mice (control) or mice treated with disulfiram, G38cells are visualized with Vimentin (red). F, three sections of apical mouse brain 28 days after injection, untreated (control) or treated with disulfiram.Tissue was stained with hematoxylin (blue), G38 cells are visualized with Vimentin (red). In C and D, 10 consecutive and comparable brain sections of twomice per group were analyzed, whereas in E and F, exemplary results of two independent analyses are shown. Scale bar in E, 1 mm; in F, 0.2 mm.
Inhibition of Glioblastoma Invasion
www.aacrjournals.org Mol Cancer Res; 11(12) December 2013 1619
as a compact, tightly packed protein, leaving many proteindomains inaccessible for potential interactions (41). Activa-tion of fibronectin is believed to occur via extending theprotein, i.e., integrins on the cell surfacemechanically stretchthe fibronectin molecules into an open conformation thatallows their incorporation into a mesh (42). Interestingly,fibronectin consists mainly of homologous type III repeats,which contain cryptic binding sites and are highly sensitiveto proteolysis (43). Our data indicate that NF-kB–depen-dent activation of MMPs can mediate increased incorpo-ration of fibronectin into a matrix.Furthermore, our finding also shed light on the rather
controversially discussed role of fibronectin in glioblastomamultiforme, which had been frequently dismissed as a cellculture phenomenon, as fibronectin is, vasculature aside, notgenerally expressed in brain tissue. Two of two humanpathology sections of glioblastoma multiforme/normal braintissue intersection exhibited cells staining positive for fibro-nectin, which was previously predicted by gene expressionprofiling (44). This is in contrast to a previous study, whichfound Fibronectin expression in the minority of samplesinvestigated (37). However, it should be pointed out thatwhereas we focused on invading cells at the tumor periphery,the latter group investigated the glioblastoma multiformetumor bulk. Because we propose a role for fibronectin ininvasion/micrometastasis and have previously shown thathomotypic cell–cell interactions, as found in the tumor bulk,can compensate for cell–substrate interactions in manyrespects (11), we would not necessarily expect high levels of
fibronectin within the tumor. We found, as did others (37),an increase in fibronectin positivity in human glioblastomamultiforme cells grown as mouse xenographs, suggesting thepossibility that fibronectin-expressing cells engraft better in aforeign environment, or that glioblastoma multiforme tumorstem cells themselves also produce high amounts of fibro-nectin (as shown here for G38 and G40) and these relativesmall tumors grown in mouse are enriched in stem cells.Obviously, these possibilities are not mutually exclusive.However, should additional research reenforce the suggestionthat fibronectin production can also be a stem-cell charac-teristic; this would further strengthen the rationale for usingdisulfiram in glioblastoma multiforme therapy. Althoughdisulfiram, a drug commonly associated with treatment ofalcoholism (28, 29), is currently being clinically evaluated fortreatment of brain tumors (44), it has so far only been studiedas an inhibitor of aldehyde dehydrogenase and thus prevent-ing cancer repopulation after chemotherapy (31). Our datasuggest that disulfiram's ability to prevent fibronectin's incor-poration into the matrix might also further inhibit stem-cellexpansion, which has been shown to be facilitated by cell/fibronectin interaction (45). To establish a tool to use forinvestigating the potential role of disulfiram in future ther-apeutic regiments, we showed that G38 glioblastoma multi-forme–initiating cells can form a highly invasive tumor whenimplanted orthotopically into mouse brains. Importantly,disulfiram seems to prevent both, micrometastatic spreadand tumor bulk expansion, making it an ideal candidate fornovel therapeutic approaches in glioblastoma multiforme.
Figure 6. Fibronectin is highlyexpressed in a subpopulation ofglioblastomamultiforme cells in thehuman brain. Two sections of theintersection glioblastomamultiforme/human brain fromindividual patients (#1 and #2) werestained for hematoxylin (blue) andfibronectin (red) and analyzed atdifferent magnifications: A, �10(left) and �40 (right); B, �20; andC, �40. B indicates normal braintissue; T, tumor tissue, asterisks,vasculature, the only commonsource of fibronectin in normalbrain; square, area of invasivetumor cells that expressfibronectin; rings have beenplaced around exemplary cellsthat that express fibronectinheterogeneously and arrowsrun parallel to secreted fibronectintracks. Scale bar in A, 0.025 mm(left), 0.1 mm (right); B, 0.05 mm;in C, 0.1 mm.
Westhoff et al.
Mol Cancer Res; 11(12) December 2013 Molecular Cancer Research1620
Although this study focuses on inhibition of gliobla-stoma multiforme cell invasion, it also has further impli-cations for other aspects of glioblastoma multiformetherapy, such as adhesion-mediated apoptosis resistance,or AMAR, a major obstacle in modern cancer therapy(46). Cell–fibronectin adhesion is not the only means bywhich glioblastoma multiforme cells interact with theirmicroenvironment, as glioblastoma multiforme–astrocyteinteraction (47) and homotypic glioblastoma multiforme–glioblastoma multiforme cell interaction (11) also playan important role in glioblastoma multiforme biology.Importantly, both these cell–cell interactions are mediatedby gap junctions (47, 11), therefore, it is conceivable toenvision a glioblastoma multiforme therapy that combines
disulfiram with an inhibitor of gap junction, such ascarbenoxolone (11). This could fully isolate glioblastomamultiforme cells from their surroundings and although itmight not suffice to induce anoikis in glioblastoma multi-forme, it would significantly reduce AMAR and thussensitize glioblastoma multiforme cells for chemothera-py-induced apoptosis.In summary, this work identifies a mechanism by which
glioblastoma multiforme cells alter their microenviron-ment so as to enable their brain invasion and the estab-lishment of micrometastasis, the most lethal hallmark ofthis particular malignancy. Furthermore, we have delin-eated the underlying molecular steps that facilitate thisprocess, which also sheds new light on the processing of
Figure 7. Scheme of NF-kB–dependent invasion inglioblastoma multiforme. Aglioblastoma multiforme cellneither interacting with substratumnor other cells, reacts to isolationby activation of NF-kB. This leads,via uPA/uPAR to activation ofMMPs, MMP-2 in particular.Glioblastoma multiforme alreadyexpress inactiveMMP-2and tightlypacked Fn on their surface,therefore a more permissivemicroenvironment can berapidly created.
Inhibition of Glioblastoma Invasion
www.aacrjournals.org Mol Cancer Res; 11(12) December 2013 1621
fibronectin and thus the formation of the extracellularmatrix in general. Our data and the resulting model (Fig.7) offer a unifying platform for previously unconnectedfindings about glioblastoma multiforme invasion. Forexample, although the role of uPAR and MMPs inglioblastoma multiforme invasion has been known forsome time (48) and a putative role for NF-kB has alsobeen suggested (9, 10), so far, no one had been able toconnect these individual findings. Furthermore, our mod-el also predicts additional points of interest/targets fortherapeutic intervention, for example as the interactionbetween glioblastoma multiforme cell and microenviron-ment is mediated via fibronectin and focal adhesion, wewould expect that alterations in focal adhesion proteinswould also affect invasion, which has been recently con-firmed (49).Finally, we also show how these findings might be trans-
lated into clinical use and strongly encourage further researchinto the incorporation of disulfiram, or related substances,into the current clinical standards of glioblastoma multi-forme therapy.
Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.
Authors' ContributionsConception and design: M.-A. WesthoffDevelopment of methodology:M.-A. Westhoff, L. Nonnenmacher, C. Jennewein,T. Simmet, M.G. BachemAcquisition of data (provided animals, acquired and managed patients, providedfacilities, etc.): M.-A. Westhoff, S. Zhou, M. Schneider, N.O. Carragher,B. Baumann, A. Krause, T. Simmet, M.G. BachemAnalysis and interpretation of data (e.g., statistical analysis, biostatistics, compu-tational analysis): M.-A. Westhoff, S. Zhou, G. Karpel-Massler, N.O. Carragher,K.-M. DebatinWriting, review, and/or revision of the manuscript: M.-A. Westhoff, S. Zhou,L. Nonnenmacher, G. Karpel-Massler, M.-E. Halatsch, N.O. Carragher, T. Simmet,M.G. Bachem, C.R. Wirtz, K.-M. DebatinAdministrative, technical, or material support (i.e., reporting or organizing data,constructing databases): M.-E. Halatsch, K.-M. DebatinStudy supervision: M.-A. Westhoff, C.R. Wirtz
AcknowledgmentsThe authors are grateful to S. Baumgart, A. Dittrich, E. Scheidhauer, and F. Genze
for providing expert technical assistance, as well as Melanie Rall and Reem KhaledFathy Fathallah. The authors also thank P. Kirsch, C. Sand, and B.Mayer for their helpregarding the xenotransplant experiments and A. Scheuerle for kindly providing theneuropathologic human material. Figure 7 was designed by Bruehl Studio Grafica(www.oliverbruehl.de), with K. La Ferla-Bruehl serving as scientific advisor. Finally,the authors also thank S. E. Barry for critically reading early drafts of the manuscriptand S. Fulda for her long and lasting cooperation.
The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be herebymarked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.
Received August 19, 2013; revised September 19, 2013; accepted September 20,2013; published OnlineFirst October 21, 2013.
References1. Huse JT, Holland EC. Targeting brain cancer: advances in the molec-
ular pathology of malignant glioma and medulloblastoma. Nat RevCancer 2010;10:319–31.
2. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, TaphoornMJ, et al. Radiotherapy plus concomitant and adjuvant temozolomidefor glioblastoma. N Engl J Med 2005;352:987–96.
3. Wen PY, Kesari S. Malignant gliomas in adults. N Engl J Med 2008;359:492–507.
4. Giese A. Glioma invasion–pattern of dissemination by mechanisms ofinvasion and surgical intervention, pattern of gene expression and itsregulatory control by tumorsuppressor p53 and proto-oncogene ETS-1. Acta Neurochir Suppl 2003;88:153–62.
5. Idbaih A,Marie Y, SansonM.NFKBIA deletion in glioblastomas. NEnglJ Med 2011;365:277–278.
6. Karl S, Pritschow Y, Volcic M, Hacker S, Baumann B, Wiesmuller L,et al. Identification of a novel pro-apoptotic function of NF-kappaB inthe DNA damage response. J Cell Mol Med 2009;13:4239–56.
7. Jennewein C, Karl S, Baumann B, Micheau O, Debatin KM, Fulda S.Identification of a novel pro-apoptotic role of NF-kappaB in theregulation of TRAIL- and CD95-mediated apoptosis of glioblastomacells. Oncogene 2011;31:1468–74.
8. La Ferla-Bruhl K, Westhoff MA, Karl S, Kasperczyk H, Zwacka RM,Debatin KM, et al. NF-kappaB-independent sensitization of glioblas-toma cells for TRAIL-induced apoptosis by proteasome inhibition.Oncogene 2007;26:571–82.
9. Raychaudhuri B, Vogelbaum MA. IL-8 is a mediator of NF-kappaBinduced invasion by gliomas. J Neurooncol 2010;101:227–35.
10. Badiga AV, Chetty C, Kesanakurti D, Are D, Gujrati M, Klopfenstein JD,et al.MMP-2 siRNA inhibits radiation-enhanced invasiveness in gliomacells. PLoS ONE 2011;6:e20614.
11. WesthoffMA,ZhouS,BachemMG,DebatinKM,FuldaS. Identificationof a novel switch in the dominant formsof cell adhesion-mediated drugresistance in glioblastoma cells. Oncogene 2008;27:5169–81.
12. Syrovets T, Gschwend JE, Buchele B, Laumonnier Y, Zugmaier W,Genze F, et al. Inhibition of IkappaB kinase activity by acetyl-boswellicacids promotes apoptosis in androgen-independent PC-3 prostatecancer cells in vitro and in vivo. J Biol Chem 2005;280:6170–80.
13. Wei J, Zhou S, Bachem MG, Debatin KM, Beltinger C. Infiltration ofblood outgrowth endothelial cells into tumor spheroids: role ofmatrix metalloproteinases and irradiation. Anticancer Res 2007;27:1415–21.
14. Fulda S, Wick W, Weller M, Debatin KM. Smac agonists sensitizefor Apo2L/TRAIL- or anticancer drug-induced apoptosis andinduce regression of malignant glioma in vivo. Nat Med 2002;8:808–15.
15. Cimpean AM, Ribatti D, Raica M. The chick embryo chorioallantoicmembrane as a model to study tumor metastasis. Angiogenesis2008;11:311–9.
16. Mitra SK, Hanson DA, Schlaepfer DD. Focal adhesion kinase: incommand and control of cell motility. Nat Rev Mol Cell Biol 2005;6:56–68.
17. Lee BH, Park SY, Kang KB, Park RW, Kim IS. NF-kappaB activatesfibronectin gene expression in rat hepatocytes. Biochem Biophys ResCommun 2002;297:1218–24.
18. Bond M, Chase AJ, Baker AH, Newby AC. Inhibition of transcriptionfactor NF-kappaB reduces matrix metalloproteinase-1, -3 and -9production by vascular smoothmuscle cells. CardiovascRes 2001;50:556–65.
19. Fukuyama R, Ng KP, Cicek M, Kelleher C, Niculaita R, Casey G, et al.Role of IKK and oscillatory NFkappaB kinetics in MMP-9 gene expres-sion and chemoresistance to 5-fluorouracil in RKO colorectal cancercells. Mol Carcinog 2007;46:402–13.
20. Lin ML, Lu YC, Chung JG, Wang SG, Lin HT, Kang SE, et al. Down-regulation of MMP-2 through the p38 MAPK-NF-kappaB-dependentpathway by aloe-emodin leads to inhibition of nasopharyngeal carci-noma cell invasion. Mol Carcinog 2010;49:783–97.
21. Nakada M, Okada Y, Yamashita J. The role of matrix metalloprotei-nases in glioma invasion. Front Biosci 2003;8:e261–9.
22. Annabi B, Laflamme C, Sina A, Lachambre MP, Beliveau R. A MT1-MMP/NF-kappaB signaling axis as a checkpoint controller of COX-2expression in CD133þ U87 glioblastoma cells. J Neuroinflammation2009;6:8.
23. Wisniewski P, Ellert-Miklaszewska A, Kwiatkowska A, Kaminska B.Non-apoptotic Fas signaling regulates invasiveness of glioma cells
Westhoff et al.
Mol Cancer Res; 11(12) December 2013 Molecular Cancer Research1622
and modulates MMP-2 activity via NFkappaB-TIMP-2 pathway. CellSignal 2010;22:212–20.
24. Stewart CE, Sayers I. Characterisation of urokinase plasminogenactivator receptor variants in human airway and peripheral cells. BMCMol Biol 2009;10:75.
25. Wang Y. The role and regulation of urokinase-type plasminogenactivator receptor gene expression in cancer invasion and metastasis.Med Res Rev 2001;21:146–70.
26. Mazzieri R, Masiero L, Zanetta L, Monea S, Onisto M, Garbisa S, et al.Control of type IV collagenase activity by components of the uroki-nase-plasmin system: a regulatory mechanism with cell-bound reac-tants. EMBO J 1997;16:2319–32.
27. Raghu H, Gondi CS, Dinh DH, Gujrati M, Rao JS. Specific knockdownof uPA/uPAR attenuates invasion in glioblastoma cells and xenograftsby inhibition of cleavage and trafficking of Notch -1 receptor. MolCancer 2011;10:130.
28. Jacobsen E, Martensen-Larsen O. Treatment of alcoholism withtetraethylthiuram disulfide (antabuse). J Am Med Assoc 1949;139:918–22.
29. Martensen-Larsen O. Treatment of alcoholism with a sensitizing drug.Lancet 1948;2:1004.
30. Richert DA, VanderlindeR,WesterfeldWW.The composition of rat liverxanthine oxidase and its inhibition by antabuse. J Biol Chem 1950;186:261–74.
31. Kast RE, Belda-Iniesta C. Suppressing glioblastoma stem cell functionby aldehyde dehydrogenase inhibition with chloramphenicol or disul-firam as a new treatment adjunct: an hypothesis. Curr Stem Cell ResTher 2009;4:314–7.
32. Guo X, Xu B, Pandey S, Goessl E, Brown J, Armesilla AL, et al.Disulfiram/copper complex inhibiting NFkappaB activity and potenti-ating cytotoxic effect of gemcitabine on colon and breast cancer celllines. Cancer Lett 2010;290:104–13.
33. WangW,McLeodHL, Cassidy J. Disulfiram-mediated inhibition of NF-kappaB activity enhances cytotoxicity of 5-fluorouracil in humancolorectal cancer cell lines. Int J Cancer 2003;104:504–11.
34. Cho HJ, Lee TS, Park JB, Park KK, Choe JY, Sin DI, et al. Disulfiramsuppresses invasive ability of osteosarcoma cells via the inhibitionof MMP-2 and MMP-9 expression. J Biochem Mol Biol 2007;40:1069–76.
35. Shian SG, Kao YR, Wu FY, Wu CW. Inhibition of invasion and angio-genesis by zinc-chelating agent disulfiram. Mol Pharmacol 2003;64:1076–84.
36. De Witt Hamer PC, Van Tilborg AA, Eijk PP, Sminia P, Troost D, VanNoorden CJ, et al. The genomic profile of human malignant glioma is
altered early in primary cell culture and preserved in spheroids. Onco-gene 2008;27:2091–6.
37. Jones TR, Ruoslahti E, Schold SC, Bigner DD. Fibronectin and glialfibrillary acidic protein expression in normal human brain and ana-plastic human gliomas. Cancer Res 1982,42:168–77.
38. Chang HJ, Kim MH, Baek MK, Park JS, Chung IJ, Shin BA, et al.Triptolide inhibits tumor promoter-induced uPAR expression viablocking NF-kappaB signaling in human gastric AGS cells. AnticancerRes 2007;27:3411–7.
39. WangY,DangJ,WangH, AllgayerH,Murrell GA,BoydD. Identificationof a novel nuclear factor-kappaB sequence involved in expression ofurokinase-type plasminogen activator receptor. Eur J Biochem2000;267:3248–54.
40. Moreau M, Mourah S, Dosquet C. Beta-Catenin and NF-kappaBcooperate to regulate the uPA/uPAR system in cancer cells. Int JCancer 2011;128:1280–92.
41. Faralli JA, Schwinn MK, Gonzalez JM Jr, Filla MS, Peters DM. Func-tional properties of fibronectin in the trabecular meshwork. Exp EyeRes 2009;88:689–93.
42. Wierzbicka-Patynowski I, Schwarzbauer JE. The ins and outs offibronectin matrix assembly. J Cell Sci 2003;116:3269–76.
43. Colin C, BaezaN,Bartoli C, Fina F, EudesN,Nanni I, et al. Identificationof genes differentially expressed in glioblastoma versus pilocyticastrocytoma using Suppression Subtractive Hybridization. Oncogene2006;25:2818–26.
44. TrialX Web Site. [updated 2013 Jul 22; cited 2013 Feb]. Available from:http://trialx.com/curetalk/2012/12/disulfiram-anti-alcoholism-drugs-brain-cancer-study-gets-2-5-m-grant/ [accessed Feb 27, 2013].
46. Westhoff MA, Fulda S. Adhesion-mediated apoptosis resistance incancer. Drug Resist Updat 2009;12:127–36.
47. Oliveira R, Christov C, Guillamo JS, de Bouard S, Palfi S, Venance L,et al. Contribution of gap junctional communication between tumorcells and astroglia to the invasion of the brain parenchyma by humanglioblastomas. BMC Cell Biol 2005;6:7.
48. Gondi CS, Lakka SS, Dinh DH, Olivero WC, Gujrati M, Rao JS. Down-regulation of uPA, uPAR and MMP-9 using small, interfering, hairpinRNA (siRNA) inhibits glioma cell invasion, angiogenesis and tumorgrowth. Neuron Glia Biol 2004;1:165–76.
49. Brockschmidt A, Trost D, Peterziel H, Zimmermann K, Ehrler M,GrassmannH, et al. KIAA1797/FOCADencodes anovel focal adhesionprotein with tumour suppressor function in gliomas. Brain 2012;135:1027–41.
Inhibition of Glioblastoma Invasion
www.aacrjournals.org Mol Cancer Res; 11(12) December 2013 1623
