Basal Caspase Activity Promotes Migration and Invasiveness in Glioblastoma Cells

The successful treatment of glioblastomas, the most malignant of all brain tumors in adults, is severely hampered by their highly invasive phenotype. Because a subpopulation of glioblastoma cells migrates deep into the surrounding brain tissue, complete surgical resection is virtually impossible. Major research efforts have been undertaken to understand the molecular basis of glioma cell migration and invasion (reviewed in refs. 1, 2). Proteins and pathways that play a crucial role in these processes include metalloproteases, integrins, focal adhesion kinase, and Rho GTPases (3-6). It has been reported that various forms of cellular stress can activate tumor cell motility, for example, hypoxia or exposure to free oxygen radicals (7). Importantly, cellular stress through chemotherapy or radiotherapy can promote the proinvasive phenotype of cancer cells (8, 9). In this context, apoptosis-related proteins seem to play an important role in tumor cell migration. The activation of the death receptor CD95 results in increased migration and invasion of apoptosis-resistant tumor cells by mechanisms involving nuclear factor-κB, extracellular signal-regulated kinase (ERK) 1/2, and caspase-8 (10). There is also evidence that the activation of apoptosis-related signaling pathways in apoptosis-resistant tumor cells contributes to the migratory and invasive capacity of cancer cells. Most of these studies focused on the mechanisms involving the stress stimulus–dependent increase of cellular motility. However, little is known about the molecular basis for the constitutive migratory and invasive activity in cancer cells. Because it seems plausible that tumor cells do not only migrate and invade when confronted with cellular stress, the possible mechanisms involved in the basal motility of tumor cells are of great interest. Given the extensively invasive phenotype of glioblastoma cells, we studied the underlying causes of glioma cell migration and invasion in the absence of obvious cytotoxic stimuli or cellular stress. We found that glioblastoma cells exhibit a constitutive caspase activity that is too low to induce apoptosis but is sufficient to promote migratory and invasive activity. A continuing basal cleavage of the motility-promoting gelsolin protein may contribute to this proinvasive phenotype. Our findings are of relevance for apoptosis-inducing cancer treatments that could, in cases of insufficient caspase activation, result in enhanced motility of cancer cells. On the other hand, the presence of a basal caspase activity that is responsible for the invasive phenotype of glioma cells also opens up the possibility of therapeutic interventions by inhibiting caspase-mediated motility via treatment with low doses of caspase inhibitors.

Given the various nonapoptotic functions of caspases, we set out to investigate the presence of active caspase-3 in human brain tumors. First, immunohistochemical detection of activated caspase-3 was done in human glioblastoma tissue from 10 patients. We found strong immunoreactivity adjacent to necrotic areas, which are a histomorphologic hallmark in glioblastomas (Fig. 1A). Glioblastoma cells strongly positive for active caspase-3 were generally characterized by the typical apoptotic phenotype; that is, nuclear condensation and fragmentation as well as cytoplasmic blebbing. This finding suggests that cell death occurring in the context of necrotic areas is at least partly the result of the activation of apoptosis. Interestingly, in tumor areas unrelated to necrosis, a significant subset of glioblastoma cells exhibited a weak to moderate expression of active caspase-3 (Fig. 1B). These cells did not show any morphologic signs of apoptosis. Similar findings were obtained when human U251MG glioblastoma cells were grown subcutaneously in nude mice. Strong caspase-3 activity was observed next to necrotic areas with the typical apoptotic morphology of glioma cells (Fig. 1D). In tumor areas without necrotic change, weak to moderate caspase-3 activity was detected in vital glioblastoma cells lacking apoptotic morphology (Fig. 1E). These findings suggest that moderate levels of active caspase-3 in vivo may exert cellular functions other than execution of apoptotic cell death.

Because the specificity of immunohistochemical detection of active caspase-3 at low staining intensity is difficult to prove, we did in vitro biochemical caspase activity assays with cellular lysates and fluorogenic assays to detect caspase activity. In the absence of proapoptotic stimulation, NCH89 ex vivo glioblastoma cells showed substantial DEVD-(caspase-3/7)–like and IETD-(caspase-8/10)–like caspase activity that was significantly blocked by the addition of the chemical caspase inhibitors, zDEVD and zIETD, respectively (Fig. 2A). For comparison, the treatment of cells with the proapoptotic staurosporine (800 nmol/L) induced much higher levels of caspase activation (22690 RFU, AFC-DEVDase; 27240 RFU, AFC-IETDase). Because we were interested in whether this basal caspase activity is specific to tumor cells, we did similar experiments with the human immortalized astrocytic cell line, SV-FHAS. SV-FHAS cells exhibited lower constitutive caspase activity than NCH89 glioblastoma cells, but the caspase activity could still be decreased by treating the cells with the respective caspase inhibitors (Fig. 2A, bottom panels). To exclude the possibility that the constitutive caspase activity observed was the result of unspecific cellular stress during the generation of cellular lysates, we did further experiments in which we directly detected the intracellular caspase-3 activity in living glioma cells. Glioma cells cultured under normal conditions and in the absence of proapoptotic stimuli were incubated with the fluorogenic caspase-3 substrate, PhiPhiLux G2D2, containing the peptide sequence GDEVDGI conjugated to two fluorophores. This experiment showed weak to moderate cytoplasmic fluorescence signals in ∼10% of glioma cells (Fig. 2B). Phase contrast microscopy confirmed that these cells did not show morphologic signs of apoptosis during the experiment or in the 24 h after taking the photographs (bottom, left). The addition of the caspase inhibitor zDEVD completely abolished the caspase-3–dependent fluorescence signal. As a positive control, the cells were treated with staurosporine, which resulted in strong fluorescence intensity in cells exhibiting apoptotic morphology. MCF-7 cells lacking caspase-3, used as a negative control, did not show any fluorescence signals.

To study the effect of caspase activity on the migration of glioma cells, we did scratch motility assays with monolayer cultures of primary (NCH89) or long-term (T98G) glioblastoma cells. The control cells rapidly migrated into the cell-free scratch area and almost completely filled the gap within 30 h (Fig. 3A, top panels). In contrast, the treatment of cells with chemical caspase inhibitors (zDEVD for caspase-3/7, zIETD for caspase-8/10, and zVAD for a broad spectrum of caspases) significantly slowed down the migration of the glioma cells (Fig. 3A-C). To exclude the possibility that caspase inhibitors negatively affected the proliferation of the glioma cells, the cells were counted before and after treatment with the inhibitors. The caspase inhibitors did not decrease cellular proliferation (Fig. 3B and C). The proliferation rates were unaffected or even slightly increased after treatment with the caspase inhibitors, especially in NCH89 cells after zDEVD treatment. Therefore, the blockage of migration by caspase inhibitors may rather be underestimated in the scrape motility assay. Next, we investigated the effect of the down-regulation of caspases on the migratory activity of glioma cells. Cellular migration was significantly impaired after down-regulating caspase-3 with two different small interfering RNA (siRNA) oligonucleotides (Fig. 3D). Similarly, the transfection of glioma cells with two different DNA antisense oligonucleotides targeting caspase-8 resulted in decreased migration, further supporting the hypothesis that caspase-3 as well as caspase-8 both contribute to glioma cell motility. The treatment with siRNA or DNA antisense oligonucleotides did not change the proliferation rates of glioma cells (data not shown).

In addition to the two-dimensional migration assays, we did cell motility assays with a more complex spacial, extracellular matrix environment (11). For this purpose, we implanted spontaneously formed multicellular spheroids derived from NCH89 cells into self-assembling collagen I gels. Whereas control cells deeply invaded the collagen matrix, the highly invasive phenotype of cells was lost after addition of the caspase inhibitors zVAD and zDEVD (Fig. 4A and B). When the control cells were allowed to invade the matrix for a longer period of time (25 days), the formation of numerous satellite tumors (up to 200 μm in diameter) that clustered around the primary central spheroid was observed. Strikingly, no satellite spheroids developed in caspase inhibitor–treated glioma cells (Fig. 4C and D).

It has been proposed that CD95/CD95 ligand interactions play a role in the promotion of tumor cell migration (10). Moreover, CD95L secretion is often elevated in tumor tissue (12). Thus, we were interested whether the constitutive caspase activity observed in glioblastoma cells was dependent on a basal stimulation of CD95, as possibly caused by autocrine or paracrine CD95/CD95 ligand interactions. Because we had shown in our previous work that the decoy receptor 3 (DcR3) efficiently blocks CD95 signaling in malignant glioma cells (13), we examined the influence of DcR3 on glioma cell migration. The incubation of NCH89 glioblastoma cells with DcR3 did not significantly modulate the motility of the cells (Fig. 5A, left). In contrast, the same concentration of DcR3 was capable of substantially inhibiting CD95 ligand–mediated cell death, demonstrating that the DcR3 concentration used was functionally active (Fig. 5A, right). Because aberrant epidermal growth factor receptor signaling via the Ras/Raf/MEK (mitogen-activated protein/ERK kinase)/ERK pathway plays an important role in glioma development (14) and because the activation of ERK1/2 was shown to induce glioma cell migration and invasion (15), we tested whether caspase inhibition–mediated blockage of migration depends on a decreased phosphorylation of ERK1/2. Glioma cells were incubated with caspase inhibitors and subjected to a Western blot analysis of phosphorylated ERK1/2. No change in pERK1/2 expression levels were detected (Fig. 5B). Similarly, the phosphorylation of c-Jun-NH₂-kinase (JNK) 1/2 was unaffected by the inhibition of caspases. These findings suggest that the alterations of pERK1/2 levels are not a downstream event of caspase inhibition and therefore cannot explain the caspase inhibitor–mediated decrease in glioma cell motility. Additionally, scrape motility assays were done in the presence of caspase inhibitors and/or the MEK1/2 inhibitor, U0126. In accordance with earlier reports (10, 15), inhibition of ERK1/2 phosphorylation decreased the migration of cells (Fig. 5C). The extent of the inhibition of motility was similar after incubation with caspase inhibitors (10 μmol/L) and MEK1/2 inhibitor (10 μmol/L). When caspase inhibition was combined with MEK1/2 inhibition, the inhibitory effect on migration was not substantially enhanced. Apart from a minor proproliferative effect of zDEVD, no significant differences in proliferation were detected (data not shown).

Finally, we sought to identify possible caspase substrates that are involved in the regulation of migration and invasion and that might explain the constitutively caspase-driven motility of glioma cells. Because caspases are known to cleave gelsolin and because a carboxyl-terminal truncation product of gelsolin was shown to promote the migration, invasion, and metastasis of melanoma cells (16-18), we examined the basal expression levels of the cleaved and full-length gelsolin protein in various glioma cell lines. In the absence of proapoptotic stimuli, we found low, but reproducibly detectable levels of cleaved gelsolin in all cell lines examined (Fig. 6). When cells were treated with the caspase inhibitor zDEVD before the generation of cellular lysates, the cleavage of gelsolin was significantly decreased. This finding was accompanied by a decrease of caspase-3/7–like activity, as determined by the measurement of fluorogenic AFC-DEVD (data not shown). In contrast, the treatment of cells with the proapoptotic death ligand tumor necrosis factor–related apoptosis-inducing ligand–induced strong cleavage of gelsolin (right).

In recent years, detailed knowledge about the alternative, nonapoptotic functions of caspases has been obtained. Although data on these nonapoptotic activities of caspases in tumors are still limited, the importance of active caspases in normal tissue is well established (19, 20). Caspase-3 plays an important role in the development of the brain, and caspase-3–deficient mice exhibit hyperplasia of central nervous tissue and disorganized cell deployment (21). In the adult brain, active caspase-3 is found in a subpopulation of nonapoptotic astrocytes and neurons throughout the central nervous system (22, 23). Moreover, nonapoptotic neuronal progenitor cells exhibiting high levels of activated caspase-3 divide in proliferative zones in the rat forebrain and then migrate to the olfactory bulb while differentiating into neurons (24). Thus, the effector caspase-3 seems to play a crucial role in mammalian brain development and possibly also in astrocytic and neuronal differentiation, plasticity, and migration. In contrast, the functional relevance of nonapoptotic caspase activity in neoplastic brain cells, such as glioblastoma cells, remains obscure.

As a reaction to various stress factors, such as proapoptotic stimulation, active caspase-3 has been shown to induce the migration of diverse types of cancer cells (7, 10, 25). In this article, we show that in the absence of cellular stress, active caspases are constitutively present in glioblastoma cells and promote their motility. Moderate active caspase-3 levels are found in human glioblastoma samples, freshly isolated glioblastoma cells, and long-term cultured glioma cell lines. The amount of active caspase-3 and other caspases is not sufficient to induce apoptotic cell death, but it contributes substantially to the motility of glioblastoma cells. Glioblastoma cells might be especially prone to the nonapoptotic effects of caspases because they express high levels of antiapoptotic proteins, which may induce “tolerance” toward constantly elevated levels of active caspases. The inhibition of caspases by chemical peptide inhibitors or gene ablation experiments results in a significant inhibition of the migratory and invasive capacity of glioma cells. Our results are in line with a report about high basal levels of caspase-3 and caspase-8 activity in breast cancer cell lines and tissues in the absence of apoptotic stimuli (26). Obviously, caspase-3 is not the only caspase that promotes the motility of glioblastoma cells, because a caspase-8/10 inhibitor (zIETD) as well as the antisense-mediated down-regulation of caspase-8 also result in impaired migration. This finding is in accordance with a recent study describing a role for caspase-8 in the calpain-mediated motility of cells in the absence of apoptosis (27).

The CD95 ligand–triggered promotion of cellular motility was reported to involve mitogen-activated protein/ERK signaling, nuclear factor-κB, and caspase-8 (10). Glioblastomas often exhibit activation of Ras/ERK signaling, which is reported to promote glioma invasion (15). We found that the inhibition of the phosphorylation of ERK1/2 decreased glioblastoma cell migration to a similar extent as by caspase inhibition; however, treatment with caspase inhibitors did not alter the phosphorylation status of ERK1/2. Caspase and MEK1/2 inhibitors did not synergistically promote migration, suggesting that caspase-dependent and ERK-dependent motility are regulated by different signaling events. Because a negative relationship between the stress-activated JNK pathway and the mitogen-activated ERK pathway has been postulated (28), we examined the phosphorylation status of JNK after treatment of cells with caspase inhibitors. However, there was no change in the phosphorylation of JNK after the inhibition of caspases. Moreover, the constitutive caspase activation of glioblastoma cells was not mediated by autocrine or paracrine CD95 stimulation because incubation with the CD95 ligand antagonist DcR3 did not modulate the migration of glioma cells.

A possible explanation for our observations is that the constant cleavage of certain substrates by constitutively active caspases results in increased cellular migration and invasion. Proteins that are involved in the regulation of motility and that can be cleaved by caspase-3 and caspase-8 include focal adhesion kinase, fodrin, and gelsolin (18, 29-31). The gelsolin protein influences the migration of cells by its capacity to sever actin polymers (32). A caspase-3–cleaved gelsolin fragment cleaves F-actin more efficiently than the full-length variant (33). Neutrophils and fibroblasts in gelsolin-deficient mice showed decreased migration in vivo and in vitro (34). Our data suggest that gelsolin might contribute to the caspase-dependent promotion of glioblastoma cell motility, because all glioma cell lines tested showed basal levels of cleaved gelsolin, which were lowered by the addition of caspase inhibitors. Further functional investigations will address the molecular mechanisms of the gelsolin-mediated promotion of glioma cell migration and invasion.

Our findings are relevant for the understanding of glioblastoma biology and may have implications for the development of novel therapeutic strategies. Because conventional radiochemotherapy at least partly induces its cytotoxic effects through apoptotic mechanisms, this phenomenon could also result in the increased migration and invasiveness of a subpopulation of tumor cells in which caspase activation was too low to induce cell death. This might, in part, be an explanation for the overall treatment resistance of glioblastomas, which is often caused by the recurrence of novel tumor manifestations even shortly after radiochemotherapy. The development of satellite tumors in long-term spheroid assays as described in our study might be a model for the recurrence of glioblastomas in the infiltration zone, where single tumor cells invade the brain tissue and then resume their proliferation activity, resulting in the formation of a recurrent glioblastoma. Treatment with caspase inhibitors completely blocked the formation of satellite tumors, further suggesting that caspase inhibition may serve as a promising treatment option for glioblastoma. Constitutive caspase activity affecting the motility of glioma cells can be inhibited by relatively low concentrations of caspase inhibitors that are insufficient to block apoptosis (data not shown). Future studies will have to show if low doses of caspase inhibitors can limit glioma cell migration and invasiveness in vivo. Such a regimen could block the dissemination of glioblastoma cells and improve the efficacy of multimodal treatments for malignant gliomas.

Generation of the human glioblastoma cell line NCH89 has been described previously (35). Low passages (<10) of NCH89 cells, LN18 and T98G glioblastoma cells, and immortalized astrocytic SV-FHAS cells were maintained in DMEM (Life Technologies) containing 10% FCS, 4,5 g/L glucose, glutamine, 110 mg/L sodium pyruvate, and 1% penicillin-streptomycin (Life Technologies) at 37°C and 5% CO₂.

Ten WHO grade 4 glioblastomas were analyzed for active caspase-3 expression by immunohistochemistry. Surgically removed tissue was fixed in buffered 4% formalin (pH 7.4) solution and embedded in paraffin. All surgical glioma specimens were obtained from the Institute of Neuropathology, University of Bonn Medical Center (Germany). The use of human tissue for study purposes was approved by the local ethics committee at the Bonn University Hospital. Paraffin-embedded tissue sections were dewaxed and rehydrated using xylene and a series of graded alcohols, followed by heat-induced antigen retrieval using a target retrieval solution (S2031, DakoCytomation) in a pressure cooker for 15 min. Staining was done on an automated staining system (Techmate 500, DakoCytomation) with avidin-biotin-complex peroxidase technique using aminoethylcarbazole for visualization and hematoxylin for counterstaining. The sections were incubated with primary antibody for 30 min at room temperature [rabbit antiactive caspase-3, clone C92-605 (5 μg/mL), BD Biosciences] and processed according to the manufacturer's instructions for the following kits: ChemMate Detection Kit (K5003, DakoCytomation), ChemMate Buffer Kit (K5006, DakoCytomation), and Avidin/Biotin Blocking Kit (SP-2001, Vector Laboratories).

All animal work was carried out in accordance with the NIH guidelines Guide for the Care and Use of Laboratory Animals. U251MG glioma cells (7.5 × 10⁶) were subcutaneously injected into the flanks of 6-week-old male athymic CD1-nude mice (Charles River) using a 30-gauge needle. When the tumors reached a diameter of 0.8 cm, the animals were sacrificed and the tumors were subjected to immunohistochemical analysis as outlined above.

Caspase-3 activity in NCH89 glioblastoma cells or MCF-7 breast cancer cells was assessed with the Caspase-3 Intracellular Activity Assay Kit II (PhiPhiLux G2D2) from Calbiochem-Novalbiochem according to the manufacturer's instructions. Briefly, cells were incubated with a caspase-3–specific peptide (sequence GDEVDGI, conjugated to two fluorophores) in DMEM supplemented with 10% (v/v) FCS. Caspase-3–like activity was visualized with a Leica DM institutional review board inverted microscope at 552 and 580 nm wavelengths after 60 min of incubation at 37°C. Cells were incubated for 24 h in culture medium containing zDEVD-fmk peptide (20 μmol/L) or, as a negative control, DMSO (1:500) before the evaluation of caspase-3–like activity. Treatment with staurosporine (250 nmol/L) for 3 h served as a positive control.

Caspase-3/7– and caspase-8/10–like activities were examined in cellular lysates. NCH89 glioblastoma cells and the SV-FHAS immortalized astrocytic cell line were cultured in six-well plates, washed with ice-cold PBS, and lysed on ice with lysis buffer containing 20 mmol/L Tris-HCl (pH 7.4), 137 mmol/L NaCl, 10% (v/v) glycerine, 1% Triton X-100, and 2 mmol/L EDTA in the absence of protease inhibitors. The lysates were centrifuged for 15 min at 13,000 rpm at 4°C and the supernatant was instantly used for the assay. Caspase activity was determined using synthetic tetrapeptide fluorogenic substrates specific for caspase-3/7 (Ac-DEVD-AFC) or caspase-8/10 (Ac-IETD-AFC). The reaction buffer (pH 7.5; 50 mmol/L HEPES, 50 mmol/L NaCl, 0.1% Chaps, 10 mmol/L EDTA, 5% glycerol, 10 mmol/L 1,4-dithio-dl-threitol) contained equal amounts of protein from the lysates and 50 μmol/L of specific substrates. Caspase activity was measured with a fluorescence microplate reader at 495 nm wavelength. For each experiment, control samples were incubated with specific caspase inhibitors, including caspase-3/7 inhibitor (zDEVD-fmk), caspase-8/10 inhibitor (zIETD-fmk), and a pancaspase inhibitor (zVAD-fmk). All caspase substrates and inhibitors were purchased from Axxora Deutschland GmbH.

Human tumor cells were lysed in the lysis buffer containing 20 mmol/L Tris-HCl (pH 7.4), 137 mmol/L NaCl, 10% (v/v) glycerine, 1% Triton X-100, 2 mmol/L EDTA, 100 mmol/L phenylmethylsulfonyl fluoride, and protease inhibitors (Complete mini; Roche Molecular Biochemicals). For the detection of phosphoprotein levels, 10 mmol/L NaPP_i, 50 mmol/L NaF, and 20 mmol/L NaVO₄ were added to the lysis buffer. After 15-min incubation on ice, the lysates were centrifuged at 14,000 × g for 10 min. The total protein concentration of the lysates was measured using the Bradford assay (Bio-Rad). Soluble protein (25-600 μg per lane) was separated on 10% to 15% polyacrylamide gels and blotted onto nitrocellulose by standard procedures. The membranes were washed, incubated with primary antibody as indicated: mouse anti-human gelsolin (BD Biosciences); mouse anti-human caspase-3 (Imgenex); mouse anti-human caspase-8 (kindly provided by P.H. Krammer); polyclonal rabbit anti-human phosphorylated p44/42 mitogen-activated protein kinase and polyclonal rabbit anti-human phosphorylated stress-activated protein kinase/JNK (Cell Signaling Technology, Inc.); and mouse monoclonal antibodies to β-tubulin or β-actin (Sigma Chemical Co.), washed, and incubated with the appropriate secondary antibody (1:3,000; horseradish peroxidase–conjugated, Bio-Rad). Bound antibodies were visualized using an enhanced chemiluminescence (19) detection system (GE-Healthcare).

Sequences of siRNA were as follows: caspase-3 A, 5′-UGAGGUAGCUUCAUAGUGGtt-3′; caspase-3 B, 5′-AUCAAUGGACUCUGGAAUAtt-3′; and scrambled control, 5′-CCUGCAGUACUUCAAGCGGtt-3′. Sequences of antisense DNA were as follows: caspase-8 A, 5′-CTGCTCAGACAGCAGATGCT-3′; caspase-8 B, 5′-ATCCAGCAGGTTCATGTCAT-3′; scrambled control, 5′-GAGGTCTCGACTTACCCGCT-3′. siRNA was purchased from MWG-BIOTECH AG. Antisense oligonucleotides were kindly provided by Y. Kim (Isis Pharmaceuticals). Cells were transiently transfected with 100 nmol/L siRNA or 100 nmol/L antisense DNA by lipofection (Lipofectamine 2000, Invitrogen). Forty-eight hours posttransfection, the expression levels of caspase-3 and caspase-8 were detected by Western blot analysis.

Conventional scrape motility assays (wound healing assays) were done according to the protocol described by Valster et al. (36). Briefly, cells were grown to confluence on polylysine-coated culture dishes. The monolayer was “wounded” with a yellow or a blue pipette tip in calcium-free PBS. The wounds of 300 μm or, alternatively, 400 μm were photographed at time zero and then at regular intervals over the course of 12 to 48 h. Images were analyzed by digitally drawn lines using Adobe Photo Shop 7.0 software. For the scrape motility assays with cells transiently transfected with siRNA or antisense DNA oligonucleotides, confluent monolayers were wounded 48 h after transfection. The MEK1/2 inhibitor U0126 (Cell Signaling Technology) was used in some of the motility assays.

NCH89 glioblastoma cells (25,000) were resuspended in 20 μL medium and suspended on the lid of a 100-mm Petri dish. Within 24 to 72 h, the cells started to form aggregates that were placed in cell culture dishes base coated with 10 mL of 2% sterile agar (Merck)/PBS (solidified) and 10 mL of growth medium. Spheroids with optimal size and shape were harvested and placed into the collagen gels. Collagen gels were made by mixing Vitrogen (97% collagen type I; Nutacon BV) with 10-fold concentrated minimal essential medium (Life Technologies) and sterile 0.1 mol/L sodium hydroxide, reaching a final concentration of 2.4 mg/mL collagen. After adjusting the pH, the collagen solution was distributed into a 24- or 48-well plate (0.5 or 0.4 mL/well, respectively). The spheroids were placed into the wells, and after solidification the gels were overlaid with 0.5 mL DMEM/10% FCS/1% penicillin-streptomycin. When testing the effects of caspase inhibitors on invasion, the medium was supplemented with inhibitors (10 μmol/L) or 1‰ (v/v) DMSO as a control. Medium with inhibitors or DMSO was changed daily. The cell migration away from the spheroid was monitored over a period of 2 to 25 days by photographing the midplane of the spheroids at time zero and at intervals of 24 h with an inverted Olympus IMT2-RFA/340 phase contrast microscope. For the invasion assays, the distance from the spheroid to the 10 cells that showed maximum invasion into the collagen matrix was measured using Adobe Photo Shop 7.0 software. For detecting satellite tumor formation, the newly formed spheroids were counted after 25 days under the phase contrast microscope and representative photos of the invasion borders were taken.

LN18 or NCH89 glioblastoma cells were plated into 96-well plates and after 24 h were treated with leucine zipper-tagged CD95 ligand (LZ-CD95L)–containing supernatants (kindly provided by P.H. Krammer) and/or with DcR3-containing supernatants derived from cells stably transfected with a plasmid encoding human DcR3 cDNA (kindly provided by A. Ashkenazi, Genentech, San Francisco). After 24 h, cell survival was measured by crystal violet staining as described previously (37).

All data are mean ± SD. The significance of differences between groups was determined by Student's t test. P values <0.05 were considered statistically significant.

We thank T. Pietsch (Institute of Neuropathology, University of Bonn Medical Center, Bonn, Germany) for providing the surgical glioma specimens, and P.H. Krammer and R. Arnold (German Cancer Research Center, Heidelberg, Germany) for providing anti-caspase-8 antibody and CD95 ligand.