Death-associated Protein 3 (Dap-3) Is Overexpressed in Invasive Glioblastoma Cells in Vivo and in Glioma Cell Lines with Induced Motility Phenotype in Vitro¹

GBM³ almost invariably recurs locally despite surgical resection, radiation therapy, and chemotherapy (1). The ability of gliomas in general and GBMs in particular to invade the surrounding tissue diffusely and eventually regrow the tumor bulk is unique among cancer types (2, 3, 4, 5). Although many data are available on cell motility from in vitro studies, the determinants of local and metastatic invasion in vivo are just beginning to emerge (6, 7). Genes that are critical for tumorigenesis are not only involved in cell-cycle regulation but also in migration/invasion and programmed cell death (7, 8, 9, 10, 11, 12, 13, 14). The discovery of a specific genetic profile responsible for the invasive behavior of GBM cells in vivo is therefore likely to lead to new therapeutic targets.

We used a gene discovery approach implementing LCM and differential display of cDNAs to compare the gene expression profile of cells in the core and at the invading edge of frozen human GBM specimens (15). Dap-3 was identified as being overexpressed in invasive GBM cells in vivo and in highly motile glioma cells in vitro. Dap-3 has been described mainly as a positive mediator of IFN-γ-, TNF-α-, and Fas-induced apoptosis (16, 17, 18); however, Dap-3 can also confer protection from Fas-, ionizing radiation-, and streptonigrin-induced cell death (18, 19). We found that glioma cells with an induced motility phenotype overexpressing Dap-3 were more resistant to apoptosis than the parental cells.

Cryopreserved GBM specimens from five patients were cut in serial 6–8 μm sections and mounted on uncoated slides treated with diethyl pyrocarbonate. The tumor core and adjacent invasive rim were identified on a coverslipped section stained with H&E. One specimen was selected for collection of 20,000 individual cells for mRNA isolation and differential display analysis; the other specimens were used for quantitative, differential RT-PCR analysis. Cryostat sections intended for LCM were transferred from −80°C storage, and immediately immersed in 75% ethanol at room temperature for 30 s. Slides were rinsed in H₂O, stained with filtered Meyer’s hematoxylin for 30 s, rinsed in H₂O, stained with blueing reagent for 20–30 s, washed in 70% and 95% ethanol for 1 min each, stained with eosin Y for 20–30 s, and dehydrated in 95% ethanol (2 × 1 min), 100% ethanol (stored over a molecular sieve; 3 × 1min), and xylene (3 × 10 min). Slides were air-dried under a laminar flow for 10–30 min and immediately processed for LCM. Diethyl pyrocarbonate-treated, autoclaved, distilled water was used to prepare every solution.

LCM was performed with a PixCell II Microscope (Arcturus Engineering, Mountain View, CA) using a 7.5 μm laser beam at 50–100 mW. GBM cells in the tumor core were readily identified and captured. Tumor cells immediately adjacent to necrotic or cortical areas, cells with a small regular nucleus, endothelial cells, and blood cells were avoided. Neoplastic astrocytes in the invasive rim ∼1 cm from the edge of the tumor core were identified according to the criteria of nuclear atypia (coarse chromatin, nuclear pleomorphism, multinucleation) and, when possible, according to their nuclear and/or cytoplasmic similarity with the GBM cells in the core.

Approximately 20,000 individual cells from the core and the invasive rim of fresh-frozen sections of a GBM were harvested by LCM. RNA was isolated from these samples according to manufacturer’s directions (Stratagene, La Jolla, CA). One hundred ng of total RNA from each population was added to duplicate reverse transcriptase reactions, each containing the oligodeoxythymidylate primer (sequence: 5′-TTTTTTTTTTTA-3′), called H-T11A, anchored to the beginning of the poly(A) tail. PCR was used to amplify random primed segments of cDNA according to the manufacturer’s directions (Hunter RNAimage, Nashville, TN). Briefly, each reverse transcriptase mix was aliquotted; combined with one of eight different assorted primers (AP1–8), and tagged with ³³P[dATP]. A display of the cDNAs was generated in the form of bands on a 6% polyacrylamide/urea gel. Reproducible bands (determined by parallel processing mRNA extractions from the two cell populations) that were differentially expressed in either of the cell populations were excised from the gel. The cDNA was reamplified using the appropriate matching assorted primer (H-AP1) and H-T₁₁A primer, then cloned using a TA Cloning Kit (Invitrogen, San Diego, CA). Bacterial colonies were plated on agar containing 50 μg/ml ampicillin and 40 μl of 40 mg/ml 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside. Colonies carrying the plasmids with inserts (white in color) were harvested, expanded, verified using EcoRI restriction digestion, and sequenced using a CEQ2000 automated sequencer (Beckman). From these candidates two bands of interest were elected for in-depth analysis.

From four additional GBM biopsy specimens, approximately 1000 cells were captured from the tumor core and the invasive rim by LCM and further processed for RNA isolation (Stratagene). Isolates were processed into cDNA using reverse transcriptase according to the manufacturer’s protocol (Ambion). cDNA for histone H3.3 was used for quantitative normalization of harvested material in the matched rim-core analyses.

Real-time quantitative PCR was performed using the LightCycler (Roche), which exploits the ability of SYBR-green to fluoresce after hybridization with double-stranded DNA. SYBR-green fluorescence was detected in real time after each cycle of amplification (20, 21). The primer sequences for Dap-3 were as follows: forward, 5′-GGAATTAGCACGTGTTCACA-3′; and reverse 5′-CATAAAGCAGATGTCCACTG-3′ (amplicon size, 389 bp). PCR conditions for Dap-3 were as follows: denaturation at 95°C for 30 s and then 40 cycles of 95°C for 0 s.; 67°C for 7 s.; 72°C for 18 s.; and then the melting curve analysis was done. Primers for histone H3.3 were as follows: forward, 5′-CCACTGAACTTCTGATTCGC-3′; reverse, 5′-GCGTGCTAGCTGGATGTCTT-3′ (amplicon size, 215 bp). PCR conditions for histone H3.3 were as follows: 40 cycles of 95°C for 0 s.; 64°C for 6 s.; 72°C for 20 s.; and then the melting curve analysis was done (22).

Internal standards of templates for quantitative analysis of the gene of interest were prepared by cloning Dap-3 and Histone H3.3 sequences compatible with the specific primers into the pCR 2.1 TOPO TA vector (Invitrogen). After expansion in Escherichia coli, plasmids were extracted and linearized, and the concentration of DNA was determined by absorption at 260 nm. Serial dilutions of the 50 μg/ml stock template solutions were used as standards with a known number of copies of Dap-3 and Histone H3.3 cDNA. PCR was performed on two μl of DNA (template standard or clinical sample) in a final volume of 20 μl. Analysis of the melting curves (standards versus sample and negative control) ensured specificity of the amplification for the expected product. Additionally, agarose gel electrophoresis of the PCR products was followed by staining with ethidium bromide to confirm the specificity of the amplification based on the molecular weight of the amplicons.

Quantification was performed on the initial exponential phase of amplification above baseline according to the LightCycler software (20, 21). The calculated cDNA copy number in each sample was derived from an extrapolated crossing point of a mathematically derived line extending from the exponential phase of amplification in a plot of fluorescence intensity (SYBR-green) versus cycle number. Diluted known standards of templates provided quantitative standard curves for each gene of interest against which the unknown clinical specimens were compared. Histone H3.3 was used as a housekeeping gene to normalize the initial content of total cDNA in the paired samples from the tumor core and from the invasive rim. The rim-core ratio, r = X/Y, was calculated, where X is the Dap-3 copy number in the rim and Y is the Dap-3 copy number in the core, both normalized to equivalent amounts of histone H3.3.

To create a coating of cell-derived ECM proteins, culture flasks were seeded with SF767 glioma cells. The cells were grown in MEM supplemented with 10% FBS to postconfluence. They were removed by treatment with Triton X-100 0.5% for 30 min at room temperature, and then by NH₄OH 0.25 m for 3–5 min at room temperature, and thoroughly rinsed with PBS (23, 24). The flasks covered with cell-derived ECM proteins were stored at 4°C until used. Alternatively, culture flasks were coated with laminin (10 μg/ml) during 1 h at 37°C and then thoroughly rinsed with 1× PBS.

T98G and G112 glioma cells were seeded either on untreated flasks or on ECM- or laminin-coated flasks to induce a migratory phenotype. The cells were grown to 30–50% confluence, then trypsinized, counted, aliquotted, and additionally processed for QRT-PCR as described above or for protein analysis.

G112 glioma cells were seeded on plain roller bottles culture flasks or on flasks coated with SF767-derived ECM and grown until 40–60% confluency. Approximately 40 million cells/sample were harvested by trypsinization for isolation of total RNA (Qiagen RNeasy Maxi Kit). Total RNA was further purified with TRIzol (Life Technologies, Inc.); RNA concentrations were measured by spectrophotometry and normalized accordingly. Reverse transcription was performed using the SuperScript II reverse transcriptase (Life Technologies, Inc.), incorporating Cy3-dUTP in cDNA from 80 μg total RNA from cells grown on plastic and Cy5-dUTP in cDNA from 150 μg total RNA from cells grown on ECM. In a second reverse transcription reaction for a separate hybridization, the cDNA was labeled with the reciprocal fluorochrome. cDNAs probes were purified and concentrated with Microcon-30 columns (Amicon) mixed with a hybridization cocktail containing 15 μg of poly(dA), 6 μg of yeast tRNA, 15 μg of CoT 10 DNA, 2× Denhardt’s solution, 2.7× SSC-0.2% SDS, denatured, and hybridized at 65°C on a 6.712-k cDNA microarray chip (NIH) for16 h. Slides were washed in 0.1% SDS with 0.5× SSC-0.01% SDS with 0.5× SSC and 0.06× SSC for 2 min each at room temperature and scanned (6).

Fresh frozen tissue blocks were sectioned (6 μm thick) onto slides, then fixed in freshly prepared paraformaldehyde (2%) for 30 min, and rinsed with PBS. Sections were incubated at room temperature with 0.1% Triton X-100 for 10 min to permeabilize the membranes. To quench endogenous peroxidase activity, slides were incubated for 30 min in 0.3% hydrogen peroxide in water. To block nonspecific binding, slides were incubated in 10% goat serum in PBS at room temperature for 30 min. Slides were washed in PBS, then incubated with mouse antihuman Dap-3 IgG1,1:400 (Transduction Laboratories, Kentucky), or no primary antibody, for the negative control, for 1 h at room temperature. Slides were washed in PBS for 5 min and the secondary antibody, goat antimouse IgG-biotin, was added at a 1:500 dilution in PBS. Slides were incubated for 1 h at room temperature. After rinsing in PBS for 5 min, slides were incubated with VECTASTAIN Elite ABC reagent (Vector Laboratories) for 30 min at room temperature. The slides were rinsed with PBS for 5 min and developed in diaminobenzidine peroxidase substrate solution (Sigma Chemical Co.) for 2 min. The reaction was stopped with tap water. The slides were counterstained with Mayer’s hematoxylin and coverslipped.

Phosporothioate ODNs were designed from the cds of the Dap-3 mRNA as follows: antisense ODN, 5′-GCTGTGATCTTGGCTTAGAGGTA-3′ (complementary to bp no. 1262 until no. 1284 of the cds); mismatch ODN, 5′-CTGTGGCTATGTCGTGATGAGAT-3′. T98G and G112 glioma cells were grown to 40–60% confluence in 10-cm tissue culture dishes in MEM supplemented with 10% FBS. The calcium-phosphate precipitation method was used for transfection following a typical protocol (25). Cells were treated with either a calcium phosphate precipitate (1× HBSS, 0.125 mm CaCl2, and H₂O) alone or with the precipitate containing 7–18 μg of the phosphorothioate oligonucleotides for 12 h in MEM + 10% FBS in a standard incubator. The cells were then trypsinized, counted, and seeded for the migration assay. An aliquot of collected cells was used for QRT-PCR and for protein analysis.

The microliter scale migration assay has been detailed previously (26, 27) and recently has been used to verify in vitro the invasive phenotype observed in vivo (6). Briefly, 10-well slides were coated with BSA 1% or laminin 10 μg/ml at 37°C for 1 h. Approximately 2500 cells/well were seeded into a cell sedimentation manifold (CSM, Inc., Phoenix, AZ) to establish a confluent monolayer circle within a 1-mm diameter. Cells were allowed to migrate during 24 h in MEM supplemented with 10% FBS at 37°C. The migration rate was calculated as the radius increase of the entire cell population. Experiments were performed as five replicates. At the end of the migration, the slides were fixed in 2% paraformaldehyde or 70% ethanol and processed for Dap-3 immunostaining and quantitative analysis of immunofluorescence by laser scanner cytometry.

The AlamarBlue assay (28) was used to test for the potential toxic effect of ODNs on glioma cell lines. Alamar blue incorporates an oxidation-reduction indicator that fluoresces and changes color in response to chemical reduction of growth medium resulting from cell growth. Briefly, cells were grown to 60% confluency before treatment with either 2.5 μm antisense or 2.5 μm mismatched ODNs overnight. After 6 h of treatment, 4000 cells of each population were seeded in quadruplicate wells of three 96-well flat-bottomed plates in 200 μl of culture medium supplemented with 10% FBS. The plates were incubated for 4, 20, and 32 h, respectively. Alamar blue was added in a volume of 20 μl (10% of total volume) to the cells at the various time points and incubated for 2 h. The plates were read on a fluorescence plate reader (excitation, 530 nm; emission, 590 nm). Averages of the absorbance values were calculated and plotted against a standard curve of untreated cells to assess increases in cell number over time.

After multiple washes with PBS, the cells were lysed in 100 μl/million cells of protein lysis buffer containing protease inhibitors.

A protein assay was performed to normalize the protein concentration among samples (BCA Protein Assay Reagent; Pierce). An identical volume of electrophoresis buffer 2× [24.8 mm Tris base, 192 mm Glycine, 0.1% w/v SDS, bromphenol blue in 1 liter dH₂O (pH 8.3)] was added to the sample, which was boiled for 10 min. Twenty- to 50-μl aliquots of the sample and 10 μl of the molecular weight marker were separated on a 12.5% acrylamide gel in an electrophoretic chamber at 150–200 mA current for ∼1 h. The proteins were transferred to a nitrocellulose membrane (Protran) using a protein transfer buffer [24.8 mm Tris base, 192 mm glycine, and 10% vol/vol methanol in 1 liter dH₂O (pH 8.3)] in a transfer chamber at 300 mA for ∼1 h. The nitrocellulose membrane was blocked in 5% nonfat dry milk in dH₂O overnight at 4°C and washed 3× 10 min with 0.05% Tween 20 in PBS. The membrane was then incubated with the anti-Dap-3 primary antibody (catalog no. D56620; mouse, IgG1; Transduction Laboratories; 1:1000 dilution), washed 3× 10 min with 0.05% Tween 20 in PBS and then incubated with a goat antimouse IgG (H+L) biotin-conjugated secondary antibody (Pierce) in ddH₂O for 1 h and washed 3× 10 min with 0.05% Tween 20 in PBS. Finally, the membrane was incubated with streptavidin-horse radish peroxidase (Amersham Life Science) for 30 min and with a mixture of the same amount of Solutions No. 1 and No. 2 of the ECL Detection reagent (Amersham Pharmacia Biotech) at the final volume of 0.125 ml/cm² membrane. The membrane was exposed to a photosensitive film (Kodak Scientific Imaging Films) and developed. The film was visualized with a digital camera and the intensity of the bands was quantified using GelExpert (NucleoTech) software.

LSC was used to quantify the level of Dap-3 protein at the end of the migration assay on different substrates and after treatment with Dap-3-antisense or mismatched oligonucleotides. LSC is based on the same principles as flow cytometry, but is done on a slide. After Dap-3-immunofluorescent staining of the migration assays, the slides were processed using a laser scanning cytometer (CompuCyte; Cambridge, MA). The amount of fluorescence (FITC, green) was recorded for each single cell of a migration assay. The mean peak channel fluorescence of every migration assay, reflecting the amount of Dap-3-staining, was used as a parameter for quantification. The mean peak channel fluorescence of migration assays on BSA or ECM treated with either antisense-ODNs or mismatch-ODNs was quantified.

Camptothecin-induced apoptosis in the T98G cells was performed in a 96-well plate. One-half of the plate was coated with 10 μg/ml laminin and another region was left uncoated. Approximately 2 × 10⁵ cells were seeded in every well the day before apoptosis induction. The cells were treated for 24 h with 0.125, 0.25, 0.5, and 1 μm of camptothecin. The DePsipher assay (R&D Systems) was then used to assess apoptosis in T98G cells. DePsipher is a lipophylic cation (5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolyl carbocyanine iodide) that aggregates upon membrane polarization to form an orange-red fluorescent compound. Failure of the mitochondrial transmembrane potential, called ψ, is one of the first intracellular changes in apoptosis (29). If the potential is disturbed, the dye cannot access the transmembrane space and remains or reverts to its green monomeric form. After treatment with camptothecin, cells were incubated with a DePsipher solution for 20 min. The red fluorescence representing live cells was measured at 485/590 nm using a microtiter-plate reader. A standard curve plotting a known number of live cells and the corresponding amount of red fluorescence without treatment was used to assess the number of living cells after every experiment.

Approximately 20,000 individual cells were harvested from both the tumor core and the invasive rim of frozen sections of one GBM specimen using LCM (see Fig. 1). The isolated RNA was reverse transcribed into cDNA in duplicate reactions. Differential display between the two cDNA populations revealed 50–60 bands of reproducible, dissimilar intensities on autoradiograms. Discrete cDNA bands differentially expressed using the primer set H-T11A and H-AP1 (clone R1.3) were consistently overexpressed in the rim cell population (Fig. 2). A sequence homology of 97% was found with the 3′-end of the cds of Dap-3 mRNA and the IRRC protein mRNA (498 of 509 bp and 433 of 444 bp, respectively).

The Dap-3 mRNA (WWW Blast at NCBI; NM 004632) has 1608 bp; the IRRC mRNA has 1648 bp (WWW Blast at NCBI; U18321). The sequences of Dap-3 and IRRC are 99% homologous (1485/1489 bp). Accordingly, we considered them as the same cDNA sequence discovered using different approaches and reported by two different authors (17, 19).

To confirm overexpression of Dap-3 in invasive cells, ∼1000 cells from both the invasive rim and core of four GBM specimens from different patients were harvested by LCM. RNA was isolated from the cells and processed for QRT-PCR analysis. Real-time monitoring of increasing amplicon abundance allowed quantification of template copy number in the original samples. Quantification was achieved using standards of the targeted PCR template for the gene of interest at known copy numbers run with the samples of unknown concentration, as exemplified in Fig. 3, A–D. Verification of amplicon identity was done by analysis of the melting curve of the DNA (Fig. 3, E–F), and molecular weight determination of the product in agarose gel electrophoresis of the product (Fig. 3,G). The samples from the tumor core and the invasive rim were normalized according to the housekeeping gene. The ratio or Dap-3 copy number rim/core varied between 3.1 and 20 (summarized in Table 1). The ratios from two separate reverse transcription reactions (RT 1 and RT 2) in the four specimens were as follows: 8.2 (RT 1) and 2.9 (RT 2) for specimen 7; 3.4 and 3.1 for specimen 15; 20 and 3.4 for specimen 16; and 10.4 and 10.8 for specimen 17. We compared all of the ratio values with the theoretical value of 1 (null hypothesis) using the unpaired Student’s t test and obtained a P of 0.006. When the same test was used for each separate specimen, P = 0.22 for specimen 7, P = 0.0044 for specimen 15, P = 0.326 for specimen 16, and P = 0.00043 for specimen 17. Overall these results confirmed overexpression of the Dap-3 mRNA in the invasive GBM cells compared with cells residing in the tumor core.

Human glioma cell lines G112 and T98G were grown either in standard culture flasks or in flasks in duplicates or triplicate. The flasks were either precoated with glioma-derived ECM or laminin or left untreated. ECM and laminin both enhance the motility behavior of these cells (30, 31; Fig. 4,A). Total RNA was isolated from each flask to perform QRT-PCR analysis of Dap-3 expression in duplicate. The samples were normalized according to the number of copies of histone h3.3 and lamin B. The whole experiment was performed three times for every cell line, and the results were consistent. Fig. 4,B shows the results of one of these experiments. There was a clear and statistically significant overexpression of the Dap-3 mRNA in T98G cells seeded on ECM and on laminin (∼3-fold) and a marginal overexpression in G112 cells. A discrete up-regulation by 1.63-fold of Dap-3 in G112 cells on ECM was confirmed by cDNA microarray analysis (not shown). Analysis of the band intensity of a Western blot using a monoclonal antibody directed against Dap-3 qualitatively confirmed overexpression of Dap-3 at the protein level (Fig. 4,C). Quantitative analysis of the level Dap-3 protein using laser scanner cytometry further confirmed overexpression in T98G cells on ECM (Fig. 4 D).

The immunochemistry studies on one frozen GBM specimen (specimen 16) showed the presence of specific immunostaining for Dap-3 in the cytoplasm of tumor cells (not shown). The fluorescent immunocytochemistry of T98G cells showed mainly perinuclear staining (not shown).

Expression of Dap-3 in glioma cells is amenable to manipulation by treatment with antisense ODNs designed against the 3′ end of the Dap-3 mRNA. Human glioma cell line T98G showed specific reduction in Dap-3 mRNA and protein levels after treatment with antisense-Dap-3 ODNs compared with treatment with mismatch ODNs (Fig. 5, A and B). The dose-dependent reduction in levels of Dap-3 protein with antisense treatment was confirmed by laser scanner cytometry (not shown). The migration rate of T98G cells treated with antisense ODNs was moderately but significantly decreased in a dose-dependent fashion (Fig. 5 C).

T98G gliomas cells growing on a motility-enhancing substrate like laminin or cell-derived ECM were more resistant to apoptosis induced by camptothecin than were the cells growing on uncoated culture flasks (Fig. 6, A–D, and 7A). Exposure to camptothecin 1 μm for 22 h resulted in extensive cell detachment and death in uncoated culture flasks compared with flasks coated with laminin or ECM. Using the DePsipher assay on a microtiter plate with subsequent automated fluorescence reading allowed for the analysis of many replicates and thus for statistical significance. When T98G cells growing on plain plastic were treated with 0.5 μm camptothecin, about 70% of live cells were consistently detected after 24 h. When T98G cells were seeded on laminin 10 μg/ml and treated with 0.5 μm camptothecin, >95% of the cells remained alive after 24 h (Fig. 7,A). This resistance to apoptosis was at least partially reversed when the cells were pretreated with Dap-3-antisense ODNs (Figs. 6 and 7 B). This finding suggests that the survival advantage while the cells are growing on a motility-promoting substrate is at least in part related to a higher level of Dap-3.

To identify potential genetic determinants of glioma invasion in vivo, we implemented LCM, differential display of mRNA, and QRT-PCR. We compared the gene expression profile of GBM cells in the tumor core with those at the invasive rim. The major caveat of LCM at the invasive edge of a GBM specimen resides in the difficulty of reliably identifying tumor cells and differentiating them from normal or reactive astrocytes and other glial or neuronal cells on a frozen section stained with H&E (32, 33). To reduce the risk of harvesting normal cells, we captured cells in the immediate vicinity of the tumoral edge in the white matter. We strictly selected cells with large, dysplastic nuclei and cells similar to those in the frank tumor tissue. LCM of large cell numbers for expression profiling is very time-consuming. Interestingly, it was possible to provide a sufficient amount of useful RNA to perform gene differential display from 20,000 cells and to perform analysis by QRT-PCR from ∼1,000 cells. That >99% of the cDNA sequences of the differential display were similarly expressed in the two cell populations was expected. Tumor cells at the invasive edge are hypothesized to differ from those in the tumor core because of the particular local environment to which they are exposed and because of their invasive behavior.

Dap-3 was overexpressed in invasive cells by a factor of 3–10 in five GBM specimens. Because of the potential problems with LCM and because of the conceptual and technical limitations related to performing QRT-PCR on small amounts of RNA, we focused our attention on an in vitro strategy to further validate Dap-3 as a potential glioma invasion gene. Exposure of glioma cells to laminin or to cell-derived ECM enhances their motility behavior, which can be tested with a migration assay. The migration rate in vitro strongly correlates with invasiveness in vivo (6, 34). overexpression of Dap-3 mRNA and protein on laminin and ECM was confirmed in vitro, particularly in T98G cells.

Dap-3 has been described as a proapoptotic protein downstream of the receptor/ligand complexes related to IFN-γ, TNF-α, and Fas by Kissil et al. (17, 18, 35, 36). They found that HeLa cells expressing antisense RNA against the Dap-3 gene were provided with a growth advantage in a selective environment containing the cell-death inducer IFN-γ (17). They subsequently localized this gene on the chromosome 1q21 by in situ hybridization (36). The gene encodes a M_r 46,000 protein containing a ATP/GTP-binding domain (P-loop). The full-length protein was shown to be proapoptotic on cells exposed to TNF-α or to the agonistic Fas-antibody, whereas a P-loop mutant was less effective in inducing apoptosis, and the COOH-terminal deleted protein (230 amino acids) acted in a dominant negative fashion, protecting cells from induced apoptosis (18). More recently, Berger et al. (37) have shown that Dap-3 is a member of a well-conserved family of mitochondrial proteins. Hulkko et al. (38) showed that Dap-3 can interact with the glucocorticoid receptor, with other nuclear receptors, and with some basic helix-loop-helix/Per-Arnt-Sim proteins. Interestingly, the same cDNA encoding for Dap-3 was isolated by Henning (19) for its ability to complement functionally streptonigrin and ionizing radiation sensitivity of ataxia telangiectasia cells (AT-D) in culture. Transfection of cells with this IRRP cDNA conferred resistance to ionizing radiation- and streptonigrin-mediated apoptosis (19). Also, Kissil et al. (18) have demonstrated that the NH₂ terminus of Dap-3 has dominant negative effects on apoptosis. The protein has an ATP/GTP-binding (P-loop) domain and other potential binding sites (37, 38), which suggest that it might have multiple functions and/or might be implicated in different pathways. From our data, we infer that Dap-3 can sustain cell migration, that cells exposed to motility-promoting substrates become more resistant to apoptosis, and that, at least to a certain extent, the Dap-3 protein protects the cells from undergoing apoptosis. We postulate that integrin-activation increases the level of Dap-3 to sustain migration, and that this secondary activation causes a switch in function that decreases the proapoptotic activity of Dap-3. Given the overexpression of Dap-3 by invasive GBM cells in vivo, we can speculate that this protein can render these cells more resistant to radiation therapy and chemotherapy.

Our findings corroborate the emerging body of data linking invasiveness (7, 14) and suppressed apoptosis. Moreover, various tumor suppressor genes, e.g., p53 (11, 39) and PTEN (8, 9, 10), or oncogenes, e.g., vSrc (12) and Ras (13), are linked to invasion-promoting and antiapoptotic pathway events.

We are indebted to Jeffrey M. Trent (NIH, Bethesda, MD) for providing us with the microarray data and to Jim Borree from CompuCyte, Cambridge, MA, for generating the laser scanner cytometry data.