Death ligand/receptor-independent caspase activation ... - Nature

ã

Oncogene (1999) 18, 5044 ± 5053 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc

Death ligand/receptor-independent caspase activation mediates drug-induced cytotoxic cell death in human malignant glioma cells Tamara Glaser1, Bettina Wagenknecht1, Peter Groscurth2, Peter H Krammer3 and Michael Weller*,1 1

Laboratory of Molecular Neuro-Oncology, Department of Neurology, University of TuÈbingen, School of Medicine, TuÈbingen, Germany; 2Institute of Anatomy, University of ZuÈrich, ZuÈrich, Switzerland; 3German Cancer Research Center, Heidelberg, Germany

Death ligand/receptor interactions and caspase activation mediate drug-induced apoptosis in certain cancer cells. The molecular mechanisms responsible for the chemoresistance of human malignant gliomas are largely unknown. Here, we report that malignant glioma cells co-express CD95 and CD95L without undergoing suicidal or fratricidal apoptosis. Glioma cells do not commit CD95/CD95L-dependent suicide or fratricide even when RNA and protein synthesis are inhibited. This is because ectopic expression of the viral caspase inhibitor, crm-A, or exposure to a neutralizing CD95L antibody, block apoptosis induced by exogenous CD95L but not cell death induced by cytotoxic concentrations of inhibitors of RNA and protein synthesis. Although some cytotoxic drugs enhance the expression of CD95 or CD95L, crm-A fails to block drug-induced cytotoxic and clonogenic cell death, suggesting that the drug-induced changes in CD95 and CD95L expression are epiphenomenal. There is also no dierence in drug-induced apoptosis between crm-A-transfected and control cells as assessed by electron microscopy, in situ DNA end labeling and DNA fragmentation. Further, glioma cells selected for resistance to CD95L do not acquire crossresistance to chemotherapy. However, the broad spectrum caspase inhibitor, ZVAD-fmk, inhibits drug-induced cytotoxic cell death, suggesting a role of crm-Ainsensitive caspases in drug-induced apoptosis of glioma cells. Thus, drug resistance of malignant glioma cells may involve de®ciencies in two interrelated pathways that mediate death in order tumor cell types: (i) death ligand/receptor signalling; and (ii) caspase activation. Keywords: glioma; apoptosis; CD95; chemotherapy; caspases

Introduction Among the putative mechanisms of drug-induced cytotoxic and clonogenic cell death of human cancer cells, interactions of endogenous death ligand/receptor pairs and caspase activation have recently attracted a lot of interest. Thus, doxorubicin cytotoxicity of human leukemia T-cell lines (Friesen et al., 1996) and bleomycin

*Correspondence: M Weller Received 11 December 1998; revised 1 April 1999; accepted 1 April 1999

cytotoxicity of hepatoma cells (MuÈller et al., 1997) have been shown to depend critically on CD95/CD95L interactions. Similar data were obtained for neuroblastoma and other solid tumor cell lines (Fulda et al., 1997, 1998a,b). Conversely, cross-resistance to CD95mediated apoptosis and drug cytotoxicity has been linked to de®cient caspase activation (Los et al., 1997). However, the role of CD95/CD95L interactions for drug cytotoxicity of lymphoma cells has been questioned (Villunger et al., 1997; Eischen et al., 1997; Kataoka et al., 1998). Thus, the precise role of CD95/CD95L interactions for sensitivity or resistance to chemotherapy of human cancer is still a matter of controversy. In contrast, several studies indicate that caspases play a crucial role in the cytotoxicity induced by dierent cancer chemotherapy drugs (Martins et al., 1997; Droin et al., 1998; Benjamin et al., 1998; Ferrari et al., 1998). We have been interested in overcoming resistance to apoptosis in human malignant glioma cells, speci®cally by targeting the CD95/CD95L system (Weller et al., 1998a). Potentiation of apoptosis by inhibitors of RNA or protein synthesis is a classical feature of death ligand-induced apoptosis and has previously been characterized in CD95 antibody-treated glioma cells, too (Weller et al., 1994). One particularly striking observation in glioma cells is the apparently nonlethal co-expression of CD95 and CD95L in cells that are susceptible to apoptosis when exposed to exogenous CD95L (Weller et al., 1997). Here, we have examined the molecular mechanisms underlying this failed suicide or fratricide in human malignant glioma cells and the implications of this observation for their chemosensitivity. Further, we have asked whether death ligand/receptor interactions and caspase activation play a role in drug-induced apoptosis of human malignant glioma cells.

Results Absence of suicide or fratricide in human malignant glioma cells? The ®rst experiment was designed to examine whether malignant glioma cells can be forced to commit CD95/ CD95L-mediated suicide or fratricide. Previously published observations with relevance to this issue include the co-expression of CD95 and CD95L in these cells (Weller et al., 1997), their susceptibility to exogenous CD95L and abrogation of exogenous CD95L-induced apoptosis by crm-A (Wagenknecht et

Drug-induced caspase activation in glioma cells T Glaser et al

al., 1998). A further characterization of the coexpression of CD95 and CD95L in four dierent glioma cell lines, LN-18, LN-229, LN-308 and T98G, revealed that none of the cell lines down-regulated CD95 or CD95L expression as assessed by ¯ow cytometry when grown to tight con¯uency (data not shown). Therefore, we assumed that mechanisms other than regulation of CD95 and CD95L expression were responsible for the lack of suicidal or fratricidal cell death under conditions of tight con¯uency in vitro. We then tested the possibility that the presumptive cytoprotective proteins, de®ciency of which allows exogenous CD95L-induced apoptosis to ensue when RNA and protein synthesis are inhibited, might also prevent suicide or fratricide in cells co-expressing CD95 and CD95L. We exposed puro and crm-A LN-18 and LN-229 cells to CD95L or actinomycin D (ActD) or cycloheximide (CHX) at increasing concentrations (Figure 1a ± c) and for various lengths of time (Figure 1d ± f). Crm-A provided strong protection from CD95L-induced apoptosis (Figure 1a and d). No dierence in sensitivity to ActD or CHX between puro and crm-A cells became apparent, suggesting that CD95/CD95L-dependent suicide does not take place even in the presence of RNA and protein synthesis inhibitors. This is because crm-A is an eective inhibitor of exogenous CD95L-induced apoptosis and would also be predicted to block apoptosis triggered by endogenous CD95/CD95L interactions. To con®rm that the human glioma cells were unable to commit suicide or fratricide in the presence of ActD or CHX,

Figure 1 Human glioma cells are resistant to CD95/CD95Lmediated suicide. LN-18 (circles) or LN-229 (squares) puro (®lled symbols) or crm-A (open symbols) cells were exposed to CD95L (a and d), ActD (b and e) or CHX (c and f) at increasing concentrations for 24 h (a ± c) or for various time intervals at ®xed concentrations (d ± f). The concentrations were 150 U/ml (LN-18) and 300 U/ml (LN-229) for CD95L in (d), 0.5 mg/ml for ActD in (e) and 10 mg/ml for CHX in (f). Survival was assessed by crystal violet staining. Data are expressed as mean percentages of survival and s.e.m. (n=3)

we showed that a neutralizing anti-mouse CD95L antibody (Kay-10) blocked apoptosis induced by exogenous murine CD95L in a concentration-dependent manner (Figure 2a) whereas a neutralizing antibody to human CD95L (Nok-1) had no eect on the cytotoxicity of ActD and CHX (Table 1). Further, the neutralizing mouse CD95L antibody did not inhibit drug-induced cytotoxicity in the murine glioma cell lines, P497 and P560 (data not shown). In contrast to the failure to modulate the intrinsic toxicity of ActD or CHX (Figure 1b, c, e and f), the sensitization to cytokine-induced apoptosis mediated by these agents was eciently blocked by crm-A (Figure 2b and c). Note that under these conditions, ActD or CHX are devoid of signi®cant intrinsic toxicity. Finally, we also considered the possibility that death ligand/receptor interactions other than CD95L/CD95 mediated the intrinsic toxicity of ActD or CHX. Two other endogenous death ligand/receptor systems, TRAIL/DR4/5 or TNF/TNF receptor interactions, can also be excluded as mediators of ActD or CHX cytotoxicity when administered alone since the toxic

Figure 2 Crm-A blocks cytotoxic cytokine-induced apoptosis in the absence or presence of inhibitors of RNA and protein synthesis. (a) LN-18 cells were exposed to CD95L (20 U/ml) in the absence or presence of increasing concentrations of neutralizing CD95L antibody (®lled squares) or an isotype control antibody (open squares) for 16 h. Data are expressed as means and s.e.m. of survival with speci®c versus control antibody. (b and c) LN-18 (b) or LN-229 (c) puro (®lled symbols) or crm-A (open symbols) cells were exposed to CD95L alone (10 U/ml for LN-18, 100 U/ml for LN-229) (circles), CD95L plus ActD (0.5 mg/ml) (squares) or CD95L plus CHX (10 mg/ml) (triangles). (d ± f) LN-18 or LN-229 puro or crm-A cells were exposed to CD95L (d; 10 U/ml for LN-18, 100 U/ml for LN-229), TRAIL-containing supernatant (e, 1 : 4) or TNF-a (f; 10 ng/ml) in the absence (black bars) or presence of ActD (0.5 mg/ml, grey bars) or CHX (10 mg/ml, open bars) for 16 h. Survival was assessed by crystal violet staining. Data are expressed as mean percentages of survival and s.e.m. compared with control cells that received either medium alone or ActD or CHX alone (n=3, *P50.05, t-test, puro versus crm-A; for the eects of ActD and CHX alone under these conditions, see Figure 1)

5045


5046

eects of exogenous TRAIL or TNF-a were blocked by crm-A (Figure 2d ± f), in contrast to those of ActD or CHX (Figure 1b, c, e and f). Drug toxicity of human malignant glioma cells: absence of a role for CD95/CD95L interactions? The data on ActD and CHX (Figure 1) predicted that glioma cells might not undergo CD95/CD95L-dependent suicidal apoptosis in response to cytotoxic drugs. First, we considered drug-induced changes in CD95 or CD95L expression. VM26 has previously been shown to induce CD95 expression in p53-heterozygous LN229 cells but not in p53-mutant T98G or p53-null LN308 cells (Winter et al., 1998). CD95 expression increased after exposure to various drugs in p53heterozygous LN-229 cells but not in p53-mutant LN-

18 cells (Table 2). Compared with untreated cells, CD95L protein expression increased moderately after exposure to vincristine, camptothecin and BCNU in LN-18 cells, and after exposure to BCNU in LN-229 cells. Next, puro or crm-A LN-18 or LN-229 cells were exposed to cytotoxic drugs in acute cytotoxicity and clonogenic cell death assays as described in Materials and methods. Crm-A did not block acute drug toxicity induced by BCNU, doxorubicin, VM26, vincristine, camptothecin, cytarabine or cisplatin (data not shown) and did not enhance clonogenic survival after drug exposure (Figure 3). To con®rm that the drugs were inducing apoptotic cell death in the glioma cells expressing crm-A, we performed electron microscopy (Figure 4a), in situ DNA end labeling (not shown) and quantitative DNA fragmentation analysis (Figure 5). These assays con®rmed that crm-A blocked apoptosis induced by exogenous CD95L but had no eect on apoptosis induced by VM26 or vincristine. Further, the neutralizing antihuman CD95L antibody did not block the cytotoxicity of VM26, vincristine or VP16 in four dierent glioma cell lines (Table 1). LN-18 cells selected for resistance to CD95L do not acquire cross-resistance to chemotherapy To further examine the relationship between the CD95/ CD95L pathway and drug-induced apoptosis, we generated CD95L-resistant LN-18 glioma cells (see Materials and methods). These cells had the same CD95 expression as the parental cells (Figure 6a) but did not exhibit caspase 8 cleavage (Figure 6b) or caspase 3-like activity (Figure 6c) in response to CD95L. These cells were cross-resistant to TNF and TRAIL (data not shown). However, the CD95Lresistant cells were as sensitive as parental LN-18 cells to various cytotoxic drugs (Figure 6d). Drug cytotoxicity of human malignant glioma cells: the role of caspases

Figure 3 CD95/CD95L interactions do not mediate druginduced clonogenic cell death of human malignant glioma cells. LN-18 (circles) or LN-229 (squares) puro (open symbols) or crmA (®lled symbols) cells were exposed to VM26, vincristine, cisplatin, camptothecin, cytarabine or BCNU for 24 h, washed, and then further cultured for 5 ± 10 generation times. Proliferation was assessed by crystal violet assay. Data are expressed as mean percentages and s.e.m. EC50 values derived from linear regression analysis showed no dierence between puro and crm-A cells (P40.05, t-test)

The data summarized in Figures 3 ± 6 exclude a role for crm-A-sensitive caspases in drug-induced apoptosis, e.g. caspase 8 (Zhou et al., 1997), but not other caspases less sensitive to inhibition by crm-A, e.g. caspase 3. Immunoblot analysis showed the formation of p43/caspase 8 in drug-treated cells, but the active p18/caspase 8 fragment was only detected in CD95Ltreated cells (Figure 7a). Accordingly, caspase 3

Table 1 A neutralizing CD95L antibody fails to modulate drug-induced glioma cell death LN-18 Control CD95L antibody antibody CHX (10 mg/ml) ActD (0.5 mg/ml) VCR (0.4 mM) VM26 (6 mM) VP16 (100 mM)

57+2 65+3 44+3 41+2 22+1

61+3 65+2 42+4 40+4 20+2

LN229 Control CD95L antibody antibody 73+3 72+4 35+2 57+2 48+1

76+3 69+3 36+2 56+3 50+4

LN-308 Control CD95L antibody antibody 77+2 66+2 53+1 54+4 50+1

75+4 68+3 55+4 58+2 51+2

T98G Control CD95L antibody antibody 71+3 69+4 28+4 38+1 27+3

74+4 65+3 24+3 35+3 24+2

LN-18 or LN-229 cells were exposed to the drugs in the presence of CD95L antibody (50 mg/ml) or an isotype antibody (16 h for ActD of CHX, 48 h for vincristine (VCR), VM26 and VP16). Data are epxressed as mean percentages of survival and s.e.m. (n=3, P40.05)


5047

Figure 4 Morphology of VM26-mediated apoptosis of puro or crm-A LN-18 or LN-229 cells on transmission electron microscopy. LN-18 (a ± d) or LN-229 (e ± h) puro (a, c, e and g) or crm-A (b, d, f and h) or exposed to VM26 (6 mM) for 48 h (c, d, g and h) and then processed for electron microscopy as described (Weller et al., 1994). Note that untreated puro and crm-A cells exhibit the same morphology. There are no morphological dierences of drug-induced apoptosis between puro and crm-A cells (a and b:62500, c and d:64200, e:62000, f:62500, g and h:64200)


5048

Figure 5 Crm-A blocks DNA fragmentation induced by CD95L but not by VM26 or vincristine. Puro (open symbols) or crm-A (®lled symbols) LN-229 cells were exposed to increasing concentrations of CD95L (triangles) (a) or to VM26 (circles) or vincristine (squares) (b) for 48 h, or for various time intervals at ®xed concentrations (c and d) of 250 U/ml CD95L (c), 0.5 mM vincristine or 10 mM VM26 (d). DNA fragmentation was assessed as described in Materials and methods. Data are expressed as mean percentages and s.e.m. (n=3)

Table 2 Drug-induced changes in CD95 and CD95L expression in human malignant glioma cells CD95 LN-18 LN-229 Control VM26 Vincristine Cisplatin Camptothecin Cyarabine BCNU

(9 mM) (0.5 mM) (50 mM) (1.5 mM) (1 mM) (0.5 mM)

2.0+0.05 2.0+0.02 2.0+0.20 2.0+0.05 2.0+0.01 2.0+0.15 2.0+0.02

CD95L LN-19 LN-229

2.2+0.12 1.3+0.05 2.9+0.20* 1.4+0.10 2.6+0.10* 1.6+0.10* 3.1+0.15* 1.3+0.00 2.4+0.15 1.6+0.25* 2.7+0.20* 1.4+0.20 2.0+0.15 1.8+0.15*

1.2+0.00 1.3+0.20 1.3+0.00 1.3+0.10 1.3+0.10 1.3+0.10 1.8+0.25*

LN-18 or LN-229 cells were untreated or exposed to VM26, vincristine, cisplatin, camptothecin, cytarabine or BCNU for 24 h. CD95 and CD95L expression were assessed by ¯ow cytometry as described in Materials and methods. Data are expressed as mean SFI values and s.e.m. (n=3, *P50.05, t-test)

cleavage was induced only by CD95L but not by the drugs (Figure 7b), and there was no drug-induced cleavage of the caspase 3 substrate poly (ADP-ribose) polymerase (PARP) (data not shown). Next, we monitored caspase 3-like enzymatic activity assessed by DEVD-amc cleavage after exposure to CD95L or the drugs. Exposure of LN-18 and LN-229 cells to CD95L (Figure 8a and d) resulted in an early induction of caspase activity, preceding cell death assessed by detachment from the cell culture dish. In contrast, there was only a moderate and delayed induction of caspase activity after exposure to vincristine (Figure 8b and e), VM26 (Figure 8c and f), cisplatin and doxorubicin (data not shown). This drug-evoked caspase activity paralleled rather than preceded cell death. In contrast to DEVD-amc, the preferential caspase-1 substrate, YVAD-amc, was not

cleaved under any of the experimental conditions examined here (data not shown). We then asked whether a pseudosubstrate peptide inhibitor with preferential anity for caspase-3, ZDEVD-cmk, or a broad-spectrum peptid inhibitor of caspases, ZVAD-fmk, inhibited drug-induced cell death. Exposure to ZDEVD-cmk and ZVAD-fmk inhibited apoptosis induced by CD95L in a concentration-dependent manner (Figure 9a and b). The cytotoxicity of vincristine and VM26 was inhibited signi®cantly only by ZVAD-fmk but not ZDEVD-cmk. Similar results were obtained with cisplatin, camptothecin, cytarabine and BCNU (data not shown) and not only in LN-18 and LN-229, but also T98G and LN-308 cells (data not shown). We also performed similar experiments to assess the eects of the caspase inhibitor ZVAD-fmk in clonogenic cell death assays. Here, the caspase inhibitor had no eect on druginduced cell death whereas CD95L-induced apoptosis was still eectively blocked (Figure 9c). Discussion Most human malignant gliomas are resistant to chemotherapy. Transient responses are observed in less than 30% of the patients. Importantly, such responses are typically of a stable disease type and not a complete remission type, suggesting that cytostasis rather than cytotoxicity is achieved with the current modes of chemotherapy. The molecular basis for this resistance to chemotherapy in malignant gliomas is unknown. The present study shows that CD95/CD95L or other death ligand/receptor interactions do not mediate drug


5049

D

Figure 6 CD95L-resistant LN-18 cells remain susceptible to drug-induced apoptosis. (a) CD95 expression in LN-18-P (parental) or LN-18-R (CD95L-resistant) was assessed by ¯ow cytometry as described in Materials and methods. The thin line represents the isotype control antibody, the bold line the CD95 antibody. (b) LN-18-P or LN-18-R cells were exposed to CD95L (10 U/ml) for 8 h. Caspase 8 cleavage was assessed by immunoblot analysis. (c) LN-18-P (black bars) or LN-18-R (open bars) cells were exposed to CD95L (10 U/ml) for 6 h. DEVD-amc cleaving activity was measured as described in Materials and methods. (d) LN-18-P (open symbols) or LN-18-R (®lled symbols) were exposed to increasing concentrations of VM26, cisplatin or vincristine for 72 h. Survival was assessed by crystal violet staining. Data are expressed as mean percentages of survival and s.e.m. (n=3)


5050

cytotoxicity of malignant glioma cells. This is because: (i) exogenous CD95L-, TRAIL- and TNF-triggered apoptosis are blocked by crm-A (Figure 2) whereas drug cytotoxicity is not (Figures 1, 3 ± 5); (ii) a neutralizing CD95L antibody does not inhibit drug cytotoxicity (Figure 2a and Table 1); and (iii) glioma cells selected for resistance to cytotoxic cytokines do not acquire cross-resistance to cytotoxic drugs (Figure 6). Drug-induced apoptosis has been associated with enhanced expression of CD95 and CD95L and autocrine or paracrine suicide in some cell types (Friesen et al., 1996; MuÈller et al., 1997; Fulda et al., 1997, 1998a,b). The drug-induced changes in CD95 and CD95L expression in glioma cells (Table 2) were epiphenomenal to the death process, given the failure of crm-A to block drug cytotoxicity (Figure 3). However, we de®ne a role for death receptorindependent caspase activation in mediating acute drug cytotoxicity, even though not clonogenic cell

A

B

death, in glioma cells (Figure 9). The failure of the caspase 8 inhibitor, crm-A, to block drug-induced cell death (Figure 3) is consistent with the unaltered chemosensitivity of caspase 87/7 cells (Varfolomeev et al., 1998). Previous data on the role of caspase 8 in drug-induced cell death, as assessed by the antiapoptotic action of crm-A, are controversial in that crm-A failed to protect U937 myeloid leukemia from druginduced apoptosis (Datta et al., 1996) but was protective in Jurkat T lymphoma cells (Antoku et al., 1997). Drug-induced caspase 3 activation as assessed by immunoblot analysis that is not blocked by crm-A has been observed in leukemia cells (Datta et al., 1996) and is consistent with preferential caspase 8 rather than 3 inhibition mediated by crmA (Zhou et al., 1997). In glioma cells, we did not detect caspase 3 cleavage during drug-induced apoptosis by immunoblot analysis (Figure 7b). More sensitive techniques such as anity labeling may be required to rule out a role for caspase 3 in druginduced apoptosis of glioma cells (Faleiro et al., 1997). Caspases are not dispensable for drug-induced apoptosis in glioma cells since the broad spectrum caspase inhibitor, ZVAD-fmk, attenuates drug cytotoxicity (Figure 9). Our data con®rm a role for caspases in cisplatin cytotoxicity of glioma cells even though caspase 1 may not be the critical caspase in that process (Kondo et al., 1996), given the lack of caspase 1 processing and YVAD-amc-cleaving activity in drug-treated glioma cells (data not shown). The drug concentrations required to induce caspase activation in glioma cells in vitro are unlikely to be achieved in vivo, consistent with the failure of current chemotherapy to induce tumor regression, as opposed to stable disease, in a relevant proportion of glioma patients. The lack of rapid and strong drug-induced caspase activation in glioma cells together with their inherent drug resistance raises the possibility that the failure to activate caspases in response to cytotoxic drugs may be a major reason for the failure of gliomas to respond to cytotoxic drug therapy. In support of this, forced expression of caspases has been shown to induce ecient tumor cell killing in rodent glioma models (Yu et al., 1996; Kondo et al., 1998). Since CD95L kills glioma cells in a caspase-dependent manner (Figure 1b), caspase pathways are not generally defective in these cells (Wagenknecht et al., 1998). We conclude that the lack of death ligand/receptordependent caspase activation and the insucient death ligand/receptor-independent caspase activation in response to cytotoxic drugs are major reasons for the chemoresistance of malignant glioma.

Materials and methods Materials

Figure 7 No evidence for drug-induced caspase 8 and 3 processing. (a and b) LN-18 or LN-229 cells were untreated or exposed to vincristine (0.2 mM) or VM26 (6 mM) for 16, 24 or 48 h. Immunoblot analyses for caspases 8 (a) and 3 (b) were performed as described in Materials and methods. CD95L-treated (16 h) cells are shown as a positive control in the left outer lane

Cisplatin, cytarabine, vincristine, doxorubicin and camptothecin were purchased from Sigma (St Louis, MO, USA). BCNU and VM26 were obtained from Bristol (Syracuse, NY, USA). The preferential caspase 3 inhibitor, ZDEVD-cmk, the broad spectrum caspase inhibitor, ZVAD-fmk and the ¯uorescent caspase substrates, DEVD-amc and YVAD-amc were from Bachem (Heidelberg, Germany). Soluble CD95L was obtained from CD95L cDNA-transfected murine N2A


5051

LN-18

Figure 8 DEVD-amc cleaving caspase activity during CD95L- and drug-induced apoptosis. LN-18 (a ± c) or LN-229 (d ± f) cells were exposed to CD95L (20 U/ml for LN-18, 100 U/ml for LN-229, (a and d), vincristine (0.2 mM, b and e) or VM26 (6 mM, c and f) for various time intervals. DEVD-amc cleavage (right axis, open symbols) and viability (left axis, closed symbols) were assessed as detailed in Materials and methods

neuroblastoma cells (Zipp et al., 1997). The caspase 8 antibody was prepared as described (Scadi et al., 1998). Mouse monoclonal caspase-3 antibody was from Transduction Laboratories (Lexington, KY, USA), a polyclonal rabbit PARP antibody from Boehringer (Mannheim, Germany). The rabbit polyclonal CD95L antibody (C-20) was from Santa Cruz (Santa Cruz, CA, USA). Neutralizing antibodies, anti-mouse CD95L antibody (Kay-10) and anti-human CD95L antibody (Nok-1), were obtained from Pharmingen (San Diego, CA, USA).

Cell lines T98G cells were from ATCC. LN-18, LN-229 and LN-308 cells were kindly provided by N de Tribolet (Lausanne, Switzerland). P497 and P560 mouse glioma cells were kindly provided by DD Bigner (Durham, NC, USA). The cells were maintained in DMEM containing 10% FCS, 1% glutamine and antibiotics (Weller et al., 1994). CD95L-resistent LN-18 glioma cells were generated by continuous culture in medium containing soluble CD95L in increasing concentrations (10 ± 100 U/ml) for 3 months. Transfections The cells were transfected with a crm-A expression plasmid or a puro control plasmid, kindly provided by Dr A Strasser (Victoria, Australia) (Strasser et al., 1995) by electroporation (Biorad Gene Pulser, 250 V, 950 mF). Puromycin (2 mg/ml) was added 48 h later. Pooled transfectants selected for 6 weeks were used for the experiments (Wagenknecht et al., 1998).

Viability and apoptosis assays For acute cytotoxicity assays, the cells were seeded at 16104 cells per well in 96-well plates, adhered for 24 h, and exposed to the drugs for 72 h. Survival was assessed by crystal violet staining. For clonogenic survival assays, the cells were seeded at 16103 cells per well, adhered for 24 h, pulse-treated for 24 h with the drugs, and maintained drug-free for 5 ± 10 generation times in complete medium (Weller et al., 1998b). Data derived from these simpli®ed clonogenic cell death assays were found to correlate well with classical colony formation assays performed with fewer cells (250 ± 500) in 6well plates in our hands. Apoptotic cell death was also characterized by electron microscopy, in situ DNA end labeling and quantitative assessment of DNA fragmentation as described (Weller et al., 1994). Caspase activity was assessed using the ¯uorescent substrates, DEVD-amc and YVAD-amc (Wagenknecht et al., 1998). Immunoblot analysis For immunoblot analysis, 20 mg protein per lane were separated on 12 ± 15% SDS ± PAGE gels and eletroblotted onto nitrocellulose. Equal loading was ascertained by Ponceau S staining. The blots were probed with the respective primary antibodies. Secondary horseradish peroxidase conjugate antibodies (1 : 3000) were detected using enhanced chemiluminescence (ECL) (Amersham, Braunschweig, Germany). Flow cytometry To assess CD95 expression, 106 cells were trypsinized, harvested into ice-cold complete medium containing 10%


5052

FCS and centrifuged for 10 min at 48C. The cells were resuspended in 100 ml ¯ow cytometry buer (1% BSA/PBS/ 0.01% sodium azide) containing 10% sheep serum, incubated for 20 min at 48C, centrifuged and labeled with mouse IgG monoclonal CD95 antibody (UB2, Immunotech, Krefeld, Germany) or mouse IgG1k (Sigma) as a control for 1 h at 48C. Thereafter, the cells were washed twice in ¯ow cytometry buer and stained with sheep anti-mouse IgGFITC (Sigma). After incubation (1 h at 48C) and washing, samples were resuspended in 300 ml of PBS containing 1% formaldehyde and stored light-protected at 48C before analysis in a Becton Dickinson FACScalibur cytometer. CD95L expression was assessed accordingly, using goat serum to block unspeci®c binding, rabbit polyclonal antihuman CD95L antibody (C-20, Santa Cruz) and rabbit IgG (Sigma) as an isotype control. The antibodies were labeled with FITC-conjugated goat anti-rabbit IgG (Sigma). The SFI was calculated as the ratio of mean ¯uorescence values obtained with the speci®c antibody and isotype control antibodies.

Figure 9 Inhibition of drug-induced apoptosis by ZVAD-fmk. LN-18 (a) or LN-229 (b) cells were exposed to CD95L (20 U/ml for LN-18, 100 U/ml for LN-229), vincristine (0.2 mM) or VM26 (6 mM) in the absence (black bars) or presence of 100 mM (gray bars) or 300 mM (white bars) of ZVAD-fmk (left panel) or ZDEVD-cmk (right panel) for 48 h. Survival was assessed by crystal violet staining. (c) LN-18 (left) or LN-229 (right) cells were exposed to CD95L (6 U/ml for LN-18 and 60 U/ml for LN-229), vincristine (0.15 mM) or VM26 (0.6 mM) in the absence (black bars) or presence of 100 mM (grey bars) or 300 mM (white bars) of ZVAD-fmk for 24 h, washed, and then further cultured for 5 ± 10 generation times in complete medium containing the respective concentrations of ZVAD-fmk. Proliferation was assessed by crystal violet staining. Data are expressed as mean percentages of survival and s.e.m. (n=3, *P50.05, t-test)

Acknowledgements This study was supported by the German Research Foundation (We 1502/3-2) and the IKFZ TuÈbingen.

References Antoku K, Liu Z and Johnson DE. (1997). Leukemia, 11, 1665 ± 1672. Benjamin CW, Hiebsch RR and Jones DA. (1998). Mol. Pharmacol., 53, 446 ± 450. Datta R, Banach D, Kojima H, Talanian RV, Alnenmri ES, Wong WW and Kufe DW. (1996). Blood, 88, 1936 ± 1943. Droin N, Dubrez L, RenvoizeÂ C, BreÂard J, Dimanche-Boitrel MT and Solary E. (1998). Oncogene, 16, 2885 ± 2894. Eischen CM, Kottke TJ, Martins LM, Basi GS, Tung JS, Earnshaw WC, Leibson PJ and Kaufmann SH. (1997). Blood, 90, 935 ± 943.

Faleiro L, Kobayashi R, Fearnhead H and Lazebnik Y. (1997). EMBO J., 16, 2271 ± 2281. Ferrari D, Stepczynska A, Los M, Wesselborg S, SchulzeOstho K. (1998). J. Exp. Med., 188, 979 ± 984. Friesen C, Herr I, Krammer PH and Debatin KM. (1996). Nature Med., 2, 574 ± 577. Fulda S, Sieverts H, Friesen C, Herr I and Debatin KM. (1997). Cancer Res., 57, 3823 ± 3829. Fulda S, Los M, Friesen C and Debatin KM. (1998a). Int. J. Cancer, 76, 105 ± 114.


Fulda S, Scadi C, Pietsch T, Krammer PH, Peter ME and Debatin KM. (1998b). Cell Death Dier., 5, 884 ± 893. Kataoka T, SchroÈter M, Hahne M, Schneider P, Irmler M, Thome M, Froelich CJ, Tschopp J. (1998). J. Immunol., 161, 3936 ± 3942. Kondo S, Barna BP, Morimura T, Takeuchi J, Yuan J, Akbasak A and Barnett GH. (1996). Cancer Res., 55, 6166 ± 6171. Kondo S, Tanaka Y, Kondo Y, Ishizaka Y, Hitomi M, Haqqi T, Liu J, Barnett GH, Alnemri ES and Barna BP. (1998). Cancer Res., 58, 962 ± 967. Los M, Herr I, Friesen C, Fulda S, Schulze-Ostho K and Debatin KM. (1997). Blood, 90, 3118 ± 3129. Martins LM, Mesner PW, Kottke TJ, Basi GS, Sinha S, Tung JS, Svingen PA, Madden BJ, Takahashi A, McCormick D, Earnshaw WC and Kaufmann SH. (1997). Blood, 90, 4283 ± 4296. MuÈller M, Strand S, Hug H, Heinemann EM, Walczak H, Homann WJ, Stremmel W, Krammer PH and Galle R. (1997). J. Clin. Invest., 99, 403 ± 413. Scadi C, Fulda S, Srinivasan A, Friesen C, Li F, Tomaselli KJ, Debatin KM, Krammer PH and Peter ME. (1998). EMBO J., 17, 1675 ± 1687. Strasser A, Harris AW, Huang DC, Krammer PH and Cory S. (1995). EMBO J., 14, 6136 ± 6147. Varfolomeev EE, Schuchmann M, Luria V, Chiannilkulchai N, Beckman JS, Mett IL, Rebrikov D, Brodianski VM, Kemper OC, Kollet O, Lapidot T, Soer D, Sobe T, Avraham KB, Goncharov T, Holtmann H, Lonai P and Wallach D. (1998). Immunity, 9, 267 ± 276.

Villunger A, Egle A, Kos M, Hartmann BL, Geley S, Ko¯er R and Greil R. (1997). Cancer Res., 57, 3331 ± 3334. Wagenknecht B, Schulz JB, Gulbins E and Weller M. (1998). Cell Death Dier., 5, 894 ± 900. Weller M, Frei K, Groscurth P, Krammer PH, Yonekawa Y and Fontana A. (1994). J. Clin. Invest., 94, 954 ± 964. Weller M, Weinstock C, Will C, Wagenknecht B, Dichgans J, Lang F and Gulbins E. (1997). Cell. Physiol. Biochem., 7, 282 ± 288. Weller M, Kleihues P, Dichgans J and Ohgaki H. (1998a). Brain Pathol., 8, 285 ± 293. Weller M, Rieger J, Grimmel C, Van Meir EG, De Tribolet N, Krajewski S, Reed JC, Von Deimling A and Dichgans J. (1998b). Int. J. Cancer, 79, 640 ± 644. Winter S, Roth W, Dichgans J and Weller M. (1998). Eur. J. Pharmacol., 341, 323 ± 328. Yu JS, Sena-Esteves M, Paulus W, Breake®eld XO and Reeves SA. (1996). Cancer Res., 56, 5423 ± 5427. Zhou Q, Snipas S, Orth K, Muzio M, Dixit VM and Salvesen GS. (1997). J. Biol. Chem., 272, 7797 ± 7800. Zipp F, Martin R, Lichtenfels R, Roth W, Dichgans J, Krammer PH and Weller M. (1997). J. Immunol., 159, 2108 ± 2115.

5053