Apoptosis in malignant glioma cells triggered by the temozolomide ...

Oncogene (2007) 26, 186–197

& 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc

ORIGINAL ARTICLE

Apoptosis in malignant glioma cells triggered by the temozolomide-induced DNA lesion O6-methylguanine WP Roos1, LFZ Batista2, SC Naumann1, W Wick3, M Weller3, CFM Menck2 and B Kaina1 1 Department of Toxicology, University of Mainz, Mainz, Germany; 2Department of Microbiology, Institute of Biomedical Sciences, University of Sao Paulo, Sao Paulo, SP, Brazil and 3Department of General Neurology, Hertie Institute for Clinical Brain Research, University of Tu¨bingen, School of Medicine, Tu¨bingen, Germany

Methylating drugs such as temozolomide (TMZ) are widely used in the treatment of brain tumours (malignant gliomas). The mechanism of TMZ-induced glioma cell death is unknown. Here, we show that malignant glioma cells undergo apoptosis following treatment with the methylating agents N-methyl-N0 -nitro-N-nitrosoguanidine (MNNG) and TMZ. Cell death determined by colony formation and apoptosis following methylation is greatly stimulated by p53. Transfection experiments with O6methylguanine-DNA methyltransferase (MGMT) and depletion of MGMT by O6-benzylguanine showed that, in gliomas, the apoptotic signal originates from O6-methylguanine (O6MeG) and that repair of O6MeG by MGMT prevents apoptosis. We further demonstrate that O6MeGtriggered apoptosis requires Fas/CD95/Apo-1 receptor activation in p53 non-mutated glioma cells, whereas in p53 mutated gliomas the same DNA lesion triggers the mitochondrial apoptotic pathway. This occurs less effectively via Bcl-2 degradation and caspase-9, -2, -7 and -3 activation. O6MeG-triggered apoptosis in gliomas is a late response (occurring >120 h after treatment) that requires extensive cell proliferation. Stimulation of cell cycle progression by the Pasteurella multocida toxin promoted apoptosis whereas serum starvation attenuated it. O6MeGinduced apoptosis in glioma cells was preceded by the formation of DNA double-strand breaks (DSBs), as measured by cH2AX formation. Glioma cells mutated in DNA-PKcs, which is involved in non-homologous endjoining, were more sensitive to TMZ-induced apoptosis, supporting the involvement of DSBs as a downstream apoptosis triggering lesion. Overall, the data demonstrate that cell death induced by TMZ in gliomas is due to apoptosis and that determinants of sensitivity of gliomas to TMZ are MGMT, p53, proliferation rate and DSB repair. Oncogene (2007) 26, 186–197. doi:10.1038/sj.onc.1209785; published online 3 July 2006 Keywords: apoptosis; DNA damage; DNA repair; MGMT; glioblastoma; Fas; p53

Correspondence: Professor B Kaina, Department of Toxicology, University of Mainz, Obere Zahlbacher Str. 67, D-55131 Mainz, Germany. E-mail: [email protected] Received 7 March 2006; revised 22 May 2006; accepted 23 May 2006; published online 3 July 2006

Introduction During the last years, anticancer drugs with methylating properties (such as temozolomide [TMZ], procarbazine, dacarbazine, streptozotocine) have received much attention, notably in the therapy of malignant gliomas. These drugs target DNA, inducing about a dozen DNA methylation products (Beranek, 1990). Studies with O6methylguanine-DNA methyltransferase (MGMT)-deficient cells (Sklar and Strauss, 1981; Scudiero et al., 1984; Preuss et al., 1996) and MGMT-transfected isogenic cell lines (Kaina et al., 1991) unequivocally revealed that one of the lesions, the minor alkylation product O6methylguanine (O6MeG), is a most potent killing lesion. It acts as a powerful trigger of apoptosis (Kaina et al., 1997; Tominaga et al., 1997; Meikrantz et al., 1998). In the cell killing process driven by O6MeG mismatch repair (MMR) is essentially involved (Kat et al., 1993; Kaina et al., 1997; Hickman and Samson, 1999; Pepponi et al., 2003). If O6MeG is repaired, cells are either resistant to O6MeG-triggered apoptosis or die because of harmful N-alkylation lesions such as 3-methyladenine, 3-methylguanine and apurinic sites that have arisen from N-methylpurine hydrolysis (Lindahl, 2000). The contribution of O6MeG to cell death depends on the ratio of O6MeG to N-alkylations in the DNA, the capacity of the cell to repair O6MeG and the rate of repair of N-alkylations (Kaina et al., 1997). The repair of O6MeG also provokes protection against the mutagenic, clastogenic and carcinogenic effects of O6-alkylating agents, suggesting that O6MeG is not only an apoptotic but also a genotoxic (for a review see Margison and Santibanez-Koref, 2002), tumour-initiating (Dumenco et al., 1993; Becker et al., 1996) and tumour-converting (Becker et al., 2003) lesion. Repair of O6MeG is accomplished by the suicide repair protein MGMT via direct methyl group transfer from the oxygen in guanine to a cystein residue (Cys145) in the MGMT molecule. Guanine in DNA is thereby restored and MGMT gets inactivated (for a review see Pegg, 2000). Because of the stochiometry of the reaction the repair capacity strictly depends on the amount of pre-existing MGMT molecules in the cell. The level of MGMT in tumours is highly variable. Quite low amounts are expressed in brain tumours (Chen et al., 1992; Preuss et al., 1995; Silber et al., 1999; Bobola et al., 2001), presumably due to

Alkylating agent-induced apoptosis in glioma cells WP Roos et al

187

MGMT promoter methylation (Esteller et al., 1999, 2001). MGMT expression level (Chen et al., 1999; Anda et al., 2003) and, as recently shown, MGMT promoter methylation (Esteller et al., 2000; Hegi et al., 2004, 2005; Paz et al., 2004) were predictive for the clinical response to chemotherapy, stressing the critical role of MGMT in determining alkylating drug resistance (for a review see Gerson, 2004). Brain tumours, notably malignant gliomas, are highly therapy-refractory tumours. Therapy of gliomas includes tumour resection, followed by radiotherapy and chemotherapy that usually involves O6-alkylating agents. The methylating agent TMZ was shown to prolong survival of patients when administered during and after radiotherapy as part of the first-line treatment (Stupp et al., 2005). Despite the usefulness of O6-methylating agents in glioma therapy, the median survival times of patients suffering from the most severe form glioblastoma multiforme are still remarkably low (12–14 months). Obviously, there is an urgent need for improving glioma therapy. One goal that needs to be reached in achieving this would be to improve our knowledge on the mechanism of alkylating agent-induced death in glioma cells. Whereas the mechanism of apoptosis induced by O6-methylating agents has been elucidated in great detail in various experimental systems such as rodent cell lines (Ochs and Kaina, 2000), lymphoblastoid cells (Dunkern et al., 2003; Hickman and Samson, 2004) and peripheral human lymphocytes (Roos et al., 2004), it remains enigmatic in malignant gliomas and other tumour types. Thus, for human glioma cells, TMZ was reported to induce either a p53-associated G2/M arrest followed by senescence or p53-independent mitotic catastrophe (Hirose et al., 2001). In another study, TMZ was proposed to induce autophagy but failed to induce apoptosis in malignant glioma cells (Kanzawa et al., 2004), whereas glioma cells grown as spheriods have been shown to be able to undergo apoptosis following alkylating agent treatment (Gunther et al., 2003). These conflicting data prompted us to study the mechanism of glioma cell death upon treatment with methylating agents in detail. Here, we show that human malignant glioma cells undergo apoptosis dose and time dependently upon methylating agent treatment for which the specific DNA lesion O6MeG is the decisive trigger. We also demonstrate that MGMT protects against TMZ-induced apoptosis in gliomas. O6MeG-triggered apoptosis in glioma cells is dependent on the p53 status that determines whether cells undergo death receptor or mitochondrial apoptosis. Furthermore, we show the importance of DNA doublestrand break (DSB) formation and cell proliferation in O6MeG-triggered apoptosis of glioma cells. Clinical implications will also be discussed.

on U87MG (p53wt) and U138MG (p53mt) glioma cells was examined in colony-forming survival assays. As shown in Figure 1, U87MG (p53wt) cells are clearly more sensitive to the killing effect of MNNG than U138MG (p53mt) cells. Induction of apoptosis following MNNG or TMZ treatment and influence of the p53 status on apoptosis To show that the killing effect observed after methylating agent treatment in p53 wild-type and p53 mutant glioma cells is related to the induction of apoptosis, the apoptotic response was measured as a function of drug concentration. MNNG and TMZ induced apoptosis in a concentration dependent manner (Figure 2a and b). Importantly, the onset of apoptosis was observed at very late time points in cultures that were growing exponentially throughout the whole time of the experiment, starting at 96 h after MNNG treatment and increasing in frequency up to 144 h (Figure 2f). p53 wildtype glioma cells (U87MG) were clearly more sensitive than p53 mutated cells (U138MG) to the induction of apoptosis by MNNG and TMZ through the whole concentration range tested (Figure 2a and b), indicating p53 to be involved in triggering apoptosis upon methylation. To substantiate the role of p53 in methylating agent-induced apoptosis, p53 was inhibited in both cell lines using the specific inhibitor pifithrin-a. In U87MG (p53wt) cells, inhibition of p53 caused a significant decrease in the apoptotic response following MNNG and TMZ treatment whereas pifithrin-a had no effect in p53 mutant cells (Figure 2c and d). To determine whether MNNG can cause the stabilization and nuclear localization of p53 in glioma cells, we treated U87MG

Results Methylating agents reduce colony formation of glioma cells The cytotoxic effect of the powerful O6-methylating agent N-methyl-N0 -nitro-N-nitrosoguanidine (MNNG)

Figure 1 Colony formation after genotoxic treatment. Dose– response curves after MNNG treatment of U87MG (p53 wild-type ’) and U138MG (p53 mutant m) glioma cells. Data are the mean of three independent experiments. Oncogene


188

(p53wt) cells with 5 mM MNNG and determined the p53 protein levels in nuclear extracts using Western blot analysis. The p53 level increased in the nucleus of p53

Oncogene

wild-type glioma cells after MNNG treatment. This increase was observed at late time points (Figure 2e), corresponding to the induction of apoptosis (Figure 2f).


189

In time course experiments, an extra p53 mutant glioma cell line was included, namely LN308 (p53mt). Both p53 mutant cell lines, U138MG and LN308, showed a clearly lower apoptotic response than the p53 wild-type cell line U87MG (Figure 2f). An additional experiment supporting the influence of p53 on the induction of apoptosis was performed by using U87MG glioma cells stably transfected with siRNA targeted to p53 (Figure 2g). In these cells, p53 was clearly less expressed than in the parental line (Figure 2g, inset). With a clinically relevant concentration of TMZ (100 mM), the knockdown of p53 in U87MG cells significantly reduced apoptosis. Collectively, the data demonstrate that, upon methylation, p53 is involved in determining the cytotoxic and apoptotic response of glioma cells. Influence of MGMT on O6MeG-triggered apoptosis The question of whether O6MeG lesions are responsible for apoptosis upon DNA methylation in glioma cells had to be answered next. To this end, U87MG (p53wt) and U138MG (p53mt) cells were stably transfected with MGMT in order to compare the apoptotic response in cells not expressing and expressing MGMT (see Figure 3a and b for protein levels). Whereas U87MG (p53wt) and U138MG (p53mt) cells showed no MGMT activity, U87MGMT and U138MGMT showed MGMT activities of B600 and B800 fmol/mg protein, respectively (Figure 3c and d). MGMT overexpression decreased apoptosis in U87MG (p53wt) and U138MG (p53mt) cells by approximately 90% after MNNG (1 or 5 mM) treatment (Figure 3e and f). When inhibiting MGMT in U87MG and U138MG cells that were stably transfected with MGMT (designated as U87MGMT and U138MGMT respectively) with O6-benzylguanine (O6BG), the cells showed a dramatic increase in apoptosis upon treatment with MNNG (Figure 3e and f). As MGMT specifically repairs O6-methylations (the minor lesion O4-methylthymine can be neglected), the strong apoptotic response in MGMT deficient gliomas and the inhibition of apoptosis in MGMT overexpressing glioma cells clearly show that at least 90% of the apoptotic signalling upon TMZ treatment originates from O6MeG lesions. Replication dependence of O6MeG-triggered apoptosis in glioma cells The late onset of apoptosis upon pulse methylation of proliferating cells indicates that apoptosis occurs not in the treatment but in one of the post-treatment cell cycles. To elucidate the dependence of apoptosis on

cell cycle progression following O6MeG induction, two strategies were employed: (1) inhibiting cells from progressing through the cell cycle and (2) by pushing cells to divide faster. Firstly, slowing or stopping the proliferation of U87MG (p53wt) cells by serum starvation provoked the apoptotic response to decrease by approximately 55% (Figure 4a). Secondly, we proved the opposite argument that forcing glioma cells to progress through the cell cycle will give rise to a stronger apoptotic response. For this reason, U87MG (p53wt) cells were treated with the powerful mitogen PMT, thereby increasing the proliferation rate of the cells (Figure 4b, left panel). Treatment with PMT leads to more than double the apoptotic frequency (Figure 4b, right panel). Thus, manipulation of cell cycle progression after methylating agent exposure demonstrates that glioma cells require cellular proliferation in order for O6MeG to be able to trigger the apoptotic response. It is obvious that extensive proliferation stimulates the apoptotic response provoked by O6MeG lesions. Importance of the formation of DNA DSBs on apoptosis following O6-methylguanine induction It has been shown previously that unrepaired O6MeG lesions give rise to DSBs (Ochs and Kaina, 2000; Roos et al., 2004). It has also been shown that ‘clean’ DSBs are a powerful trigger of apoptosis (Lips and Kaina, 2001). To substantiate a role of DSBs in methylating agent-induced apoptosis for glioma cells, the formation of DSBs were assayed following O6-methylating agent treatment. In U87MG (p53wt) cells, MNNG treatment led to the formation of DSBs as visualized by an increase in the presence of phosphorylated H2AX (Figure 5a). This was again observed at late time points (48–96 h), preceding the onset of apoptosis. Next, we determined whether methylation-induced DSBs are also related to apoptosis in glioma cells. To this end, MO59K glioma DNA-PK wild-type and MO59J glioma DNAPK-deficient cells were compared following MNNG (Figure 5b) and TMZ (Figure 5c) treatment. To deplete MGMT completely, cells were pretreated with O6BG. In both cases, the DNA-PK-deficient cells were more sensitive to apoptosis induction than wild-type cells (Figure 5b and c). As DNA-PK is involved in the repair of DSBs by non-homologous end joining (NHEJ), these data support the hypothesis that in glioma cells DSBs are critically involved in O6MeG-triggered apoptosis. It has been suggested that inhibition of transcription may trigger apoptosis (Ljungman and Zhang, 1996). Therefore, we checked the level of transcriptional

Figure 2 Apoptosis induced by MNNG and TMZ in p53 wild-type and p53 mutant glioma cells. (a) Dose–response curve of MNNGtreated U87MG (J) and U138MG () glioma cells after 144 h incubation time. (b) Dose–response curve of TMZ-treated U87MG (J) and U138MG () glioma cells after 144 h incubation time. (c) Apoptosis of U87MG and U138MG cells after 10 mM MNNG treatment at 144 h in the presence and absence of the p53 inhibitor pifithrin-a (induced apoptosis was corrected for control). (d) Apoptosis of U87MG and U138MG cells after 0.1 mM TMZ treatment at 144 h in the presence and absence of the p53 inhibitor pifithrin-a. Data are the mean of at least three independent experiments (induced apoptosis was corrected for control). (e) p53 nuclear localization in U87MG p53 wild-type glioma cells after 5 mM MNNG treatments at indicated time points (the level of induced apoptosis is shown that is corrected for control). (f) Time response after 10 mM MNNG treatment in U87MG (’) p53 wild-type, U138MG (m) and LN-308 (E) p53 mutant glioma cells. (g) Apoptosis in U87MG (p53wt) cells transfected with siRNA targeting p53 following treatment with 100 mM TMZ (induced apoptosis means corrected for untreated cells). The appropriate controls are also shown. Oncogene


190

Figure 4 Replication dependence of O6MeG-triggered apoptosis. (a) Cell cycle distribution in the presence and absence of FCS (left panel). Relative apoptotic response in U87MG (p53wt) cells in the presence and absence of FCS after 0.1 mM TMZ treatment at 144 h (right panel). (b) Relative cell number after treatment with the wild-type (PMTwt) and mutant (PMTC1165s) mitogen PMT in U87MG (p53wt) cells at 120 h (left panel). Relative apoptosis induced with 5 mM MNNG in the presence of either PMTwt or PMTC1165s at 120 h (right panel). Data are the mean of three experiments. The histogram is a representative of one experiment.

Figure 3 Influence of MGMT on O6MeG-triggered apoptosis. (a) Western blot analysis of U87MG MGMT-transfected glioma cells. (b) Western blot analysis of U138MG MGMT-transfected glioma cells. (c) MGMT activity in U87MG MGMT-transfected cells. (d) MGMT activity in U138MG MGMT-transfected cells. (e) Relative apoptotic response of U87MG MGMT transfected cells after 1 and 5 mM MNNG treatment at 144 h in the presence and absence of the specific MGMT inhibitor O6BG. (f) Relative apoptotic response of U138MG MGMT-transfected cells after 1 and 5 mM MNNG treatment at 144 h in the presence and absence of the specific MGMT inhibitor O6BG. Representative Western blots are shown. Quantitative measurements and apoptosis experiments were repeated at least three times.

inhibition by TMZ. As shown in Figure 5d, TMZ (100 mM) did not attenuate the rate of transcription as measured immediately (2 h) and late (48 h) after treatment. It rather stimulated transcription at late postexposure times in U87MG (p53wt) cells. Therefore, Oncogene

transcriptional inhibition can be excluded as a mechanism of TMZ-induced apoptosis. Death receptor signalling in O6MeG-triggered apoptosis in glioma cells To determine whether glioma cells utilize the death receptor mediated apoptotic pathway triggered by O6MeG, the expression of the Fas receptor (Fas, CD95, Apo-1) was assayed. In membrane extracts of U87MG (p53wt) cells, a clear induction of Fas was observed following MNNG treatment (Figure 6a). This increase in Fas protein was again observed at late times (Figure 6a). U138MG (p53mt) cells did not express Fas (Figure 6a). By using a specific antagonizing antibody against Fas, thereby inactivating receptor signalling, a decrease in TMZ-induced apoptosis of approximately 55% was observed in U87MG (p53wt) cells, whereas no


191

Figure 5 DNA DSBs and transcription in O6MeG-triggered apoptosis. (a) Phosphorylation of histone H2AX in U87MG (p53wt) cells after 5 mM MNNG treatment at indicated time points. (b) Time response of induced apoptosis in MO59K DNA-PK wild-type and MO59J DNA-PK mutant glioma cells after 5 mM MNNG and 10 mM O6BG treatment. (c) Time response of induced apoptosis in MO59K DNA-PK wild-type and MO59J DNA-PK mutant glioma cells after 0.5 mM TMZ and 10 mM O6BG treatment. (d) Stimulation of transcription after TMZ treatment. Data of at least three experiments were pooled.

effect was seen in U138MG (p53mt) cells (Figure 6b). A second functional assay to demonstrate the contribution of death receptor signalling in glioma p53 wild-type cells was by generating U87MG (p53wt) and U138MG (p53mt) dominant-negative-FADD (DN-FADD) stable transfectants (for expression level see Figure 6c and d). Transfection of U87MG (p53wt) with DN-FADD protected cells against O6MeG-triggered apoptosis throughout the whole concentration range tested (Figure 6c). No protective effect was observed for U138MG (p53mt) cells (Figure 6d). Mitochondrial signalling in O6MeG-triggered apoptosis in U138MG (p53mt) cells It has been shown that in p53-mutated rodent cells, O6MeG triggers the mitochondrial apoptotic pathway characterized by a decline in Bcl-2 protein level (Ochs and Kaina, 2000). Therefore, we followed the hypothesis that

in p53 mutant glioma cells the mitochondrial pathway is activated in response to O6MeG lesions. Indeed, as shown in Figure 7a, after MNNG treatment U138MG (p53mt) cells exhibit the characteristic decline in Bcl-2. This was not observed in p53wt glioma cells (data not shown). Bcl-2 decline was accompanied by caspase-9 activation as well as the activation of the executive caspases-3 and -7 (Figure 7b–d). Also caspase-2 became activated (Figure 7e) whereas no caspase-8 activation was observed in U138MG (p53mt) cells (Figure 7f). These data are in agreement with the view that in p53-mutated glioma cells, O6MeG triggers the mitochondrial-mediated apoptotic pathway whereas in p53 wild-type cells the same lesion triggers the death receptor pathway. Caspase inhibition attenuates O6MeG-triggered apoptosis The activation of multiple caspases, as shown in Figure 7, prompted us to determine whether apoptosis Oncogene


192

Figure 6 Influence of death receptor signalling on O6MeG-triggered apoptosis. (a) Western blot analysis of Fas receptor induction after 5 mM MNNG treatment in U87MG (p53wt) and U138MG (p53mt) cells. ERK2 was used as the loading control. (b) Apoptosis in U87MG and U138MG cells after 0.1 mM TMZ treatment in the presence and absence of Fas neutralizing antibody at 144 h. (c) Dose– response curves of apoptosis in U87MG (p53wt) (’) and DN-FADD-transfected cells (m) at 144 h (experiments were performed at least 3 times). Western blot analysis for positive clone is included where ERK2 was used as the loading control. (d) Apoptosis response in U138MG (p53mt) (’) and DN-FADD-transfected cells (m) at 144 h. Western blot analysis for positive clone is included where ERK2 was used as the loading control.

Figure 7 Expression of Bcl-2 and caspases in U138MG cells treated with MNNG. (a) Western blot analysis of Bcl-2 and (b–f) caspases. For caspase-8, -2, -7 and -3 the antibody was directed against the active fragment. For caspase-9 both the inactive and the active protein is recognized. The activated form is indicated by arrow. Exponentially growing cells were treated with 5 mM MNNG and cells were harvested at indicated time points. Oncogene


193

triggered by O6MeG lesions can be blocked by caspase inhibition. It should be kept in mind that the triggering of apoptosis by O6MeG in glioma cells is dependent on cellular proliferation (Figure 4) and occurs at very late times following exposure (Figure 2f). Therefore, an inhibitor approach might exhibit only low effectiveness particularly if the inhibitor is unstable and the substrate gets re-synthesized. In this approach, we applied the broad-spectrum caspase inhibitor Boc-D-fmk. In U87MG (p53wt) and U138MG (p53mt) cells, the caspase inhibitor was able to decrease the apoptotic response by approximately 30% (Figure 8). This inhibition of apoptosis was statistically significant in both cell lines (Po0.001). Together with the demonstration of caspase activation (Figure 7 and data not shown), the attenuation of apoptosis by Boc-D-fmk demonstrates that caspases are involved in O6MeGtriggered apoptosis in glioma cells.

Discussion Patients suffering from malignant glioma, in particular glioblastoma multiforme, have a very poor prognosis. Standard therapy today, in addition to surgery and radiotherapy, includes treatment with alkylating agents, specifically TMZ and the chloroethylating drugs carmustine and nimustine. Currently and in the future, the use of TMZ and presumably also other methylating drugs will become more dominant than chloroethylating agents because of less severe side effects. The common use of O6-methylating agents in glioma therapy provokes the question of how these agents exert their killing effects and, based on the data obtained, is it possible to improve glioma chemotherapy? As the molecular action of O6-alkylating agents in gliomas (and other tumour types) is largely unknown, this study was aimed at elucidating the mode of death of glioma cells upon treatment with O6-methylating agents.

Figure 8 Influence of caspase inhibition on O6MeG-triggered apoptosis. U87MG (p53wt) and U138MG (p53mt) cells were treated with 5 mM MNNG and 5 mM O6BG in the presence and absence of the broad-spectrum caspase inhibitor Boc-D-fmk. BocD-fmk (50 mM) was added 72 h after MNNG treatment, before the onset of apoptosis, and then 25 mM every 24 h until samples were stopped. Data are the mean of three experiments7s.d.

Here, we show that glioma cells undergo apoptosis following treatment with TMZ or MNNG (that acts in the same way as TMZ upon decomposition into reactive carbenium ions). Apoptosis in glioma cell lines is a very late response, occurring no earlier than 4 days after pulse treatment of exponentially growing cultures. This late response might explain the failure of detection of apoptosis in previous studies with glioma cells. Modulation of MGMT activity by pharmacological inhibition with O6BG (Dolan et al., 1990) or MGMT cDNA transfection (Kaina et al., 1991) revealed that O6MeG is the major proapoptotic DNA lesion in malignant glioma cells upon O6-methylating agent treatment. As MGMT strongly protects against O6MeG-triggered apoptosis, the data also confirm that MGMT is a decisive determinant of glioma resistance to O6-methylating agents. An intriguing observation was that p53 wild-typeexpressing glioma cells were more sensitive than their p53 mutant counterparts, and that pharmacological inhibition of p53 by pifithrin-a or knockdown of p53 by siRNA provoked O6-methylating agent resistance. This clearly illustrates that p53 is involved in O6MeGtriggered apoptosis in gliomas. As both p53 wild-type and p53 mutant glioma cells were able to undergo apoptosis in response to TMZ, we conclude that p53 is not absolutely required, but stimulates the apoptotic process. The p53 gene is often mutated in human cancers. Secondary glioblastomas that develop from grade II or III astrocytomas exhibit a p53 mutation frequency of 65%, whereas primary glioblastomas exhibit a quite low p53 mutation frequency of 10%. As p53 wild-type exerted a strong positive effect on O6-methylating agent sensitivity, p53 must be considered as a predictive factor in glioma therapy. As p53 mutant glioma cells are more resistant to TMZ, the data also implicate that p53-mutated tumour cells may be selected after treatment with TMZ or other methylating agents, resulting in recurrent gliomas that would not respond to the same treatment. How does p53 stimulate apoptosis triggered by O6MeG in glioma cells? In p53 wild-type, but not in p53 mutant glioma cells, O6-methylating agents provoked the induction of Fas (CD95, Apo-1) receptor expression and activation of the Fas-dependent apoptotic pathway (Figure 6), including activation of caspase-8 (data not shown). Transfection with DN-FADD attenuated apoptosis, which suggests that Fas is involved in O6MeG-triggered apoptosis in p53 wild-type glioma cells. p53 is a transcriptional activator of Fas (Muller et al., 1998; Pohl et al., 1999), which explains its upregulation upon treatment of glioma cells with O6methylating agents. Interestingly, in p53 mutant glioma cells Fas and caspase-8 were not induced, whereas Bcl-2 declined and caspase-9 and -3 were activated in response to MNNG. The decline of Bcl-2 is a hallmark of O6MeG-triggered apoptosis in p53-mutated cells (Ochs and Kaina, 2000). The data therefore show that in p53 mutant glioma cells, O6-methylating agents are able to trigger the intrinsic mitochondrial apoptotic pathway. It is important to note that the mitochondrial pathway is Oncogene


194

less efficiently evoked than the death receptor pathway. Consequently, at equimolar doses of TMZ p53 wildtype glioma cells deficient in MGMT exhibited a much greater response than p53 mutant cells. In both p53 wild-type and mutant glioma cells, transfection with MGMT blocked apoptosis almost to completion. This shows that O6MeG is the main trigger of apoptosis irrespective of the p53 status of cells. Given the data, p53 can be considered as a decision maker of whether the same DNA lesion, O6MeG, triggers the death receptor or the mitochondrial apoptotic pathway (see Figure 9). The potency of the lesion to trigger, in the absence of p53, the mitochondrial pathway is, however, clearly lower than to trigger the death receptor pathway. How does p53 get activated in response to O6MeG? The current model supported by a bulk of data is outlined in Figure 9. It proposes that O6MeG does not trigger apoptosis on its own. It rather becomes converted into a critical downstream lesion that is supposed to be DSBs. This occurs, according to a reasonable model, by mispairing of O6MeG lesions with thymine, which in turn is processed by MutSa-dependant mismatch repair provoking a futile MMR cycle (Karran and Bignami, 1994; Hampson et al., 1997). Secondary lesions will be formed that cause replication interference in the subsequent DNA replication cycle leading ultimately to DSBs (Ochs and Kaina, 2000). As shown recently, the glioma cells used in this study express similar levels of the MMR proteins MSH2 and MSH6 (Hermisson et al., 2006). They show DSB formation, as demonstrated by gH2AX induction that preceded apoptosis. A key repair pathway for DSBs is NHEJ that involves DNA-PKcs. Indeed, apoptosis induced by MNNG and TMZ was enhanced in DNAPKcs mutated glioma cells, which supports the view that DSBs are formed in response to O6MeG, contributing to apoptosis. The apoptotic response may occur by stabilization of p53 via the ATM/ATR-Chk1 pathway (O’Connell and Cimprich, 2005). Interestingly, however, cells deficient in ATM are hypersensitive to methylating agents (Debiak et al., 2004). Similarly, cells lacking XRCC2 are hypersensitive to TMZ and MNNG (Tsaryk et al., 2005). This supports a role of ATM and XRCC2 in the repair of DSBs originating from O6MeG lesions rather than apoptotic signalling. The apoptotic response of gliomas upon TMZ treatment can also be provoked by a p53-independent activity that degrades Bcl-2. The signalling involved in the Bcl-2-driven pathway still remains enigmatic. The model described above implicates DNA replication and cell proliferation as essential elements in O6MeG-triggered apoptosis in glioma cells. This is indeed the case, as substantiated by experiments in which proliferation was either blocked or stimulated. Inhibition of glioma cell proliferation clearly reduced the frequency of apoptosis upon O6-methylating agent treatment, whereas stimulation of proliferation by treatment with the powerful mitogen, Pasteurella multocida toxin (PMT) (Orth et al., 2003) significantly enhanced apoptosis. Therefore, proliferation rate appears to be an important predictive factor in O6MeGOncogene

Figure 9 A model of O6MeG-triggered apoptosis in p53 wild-type and p53 mutated glioma cells. For explanation see Discussion.

driven apoptosis in gliomas and very likely also in other tumour types. These data also show that any proliferation blocking treatment applied together or after TMZ might attenuate the therapeutic effect of TMZ. We should note that the concentration range of the methylating agents used throughout this work was relatively low, allowing full recovery and propagation of cells. For TMZ, the concentration range corresponds to serum levels achieved during treatment, namely 100 mM (Hammond et al., 2004). In summary, the data reported here show that glioma cells die after methylating agent treatment owing to the induction of apoptosis. This apoptotic response is primarily due to O6MeG lesions formed in the DNA. The apoptotic pathway employed by glioma cells depends on their p53 status. In p53 wild-type gliomas, O6MeG lesions trigger the death receptor pathway whereas in p53 mutant cells the same lesion activates


195

the mitochondrial pathway. The modulation of cell proliferation and DSB repair had a significant impact on O6MeG-triggered apoptosis. The data obtained suggest that predictive markers of TMZ sensitivity of gliomas are MGMT activity, proliferation rate, p53 status and the efficiency of DSB repair.

(50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 2.5 mM CaCl2) at 201C and used at a concentration of 10 ng/ml (the mitogen was a kind gift of Dr Klaus Aktories, Freiburg). Caspases were inhibited using the broad-spectrum caspase inhibitor Boc-DFMK (Calbiochem, San Diego, USA). After treating U87MG p53 wild-type and U138MG p53 mutant glioma cells with 5 mM MNNG and O6BG, caspases were inhibited with adding 50 mM Boc-D-FMK 72 h after methylation and then 25 mM every 24 h until the samples were harvested at 144 h.

Materials and methods Cell lines and culture conditions The glioma cell lines U87MG, U138MG, LN-308, M059K and M059J were used in this study. U87MG is p53 wild-type whereas U138MG and LN-308 are p53 mutant (Wischhusen et al., 2003). M059K is DNA-PK wild type and M059J lack the p350 component of DNA-PK (Allalunis-Turner et al., 1995). All cell lines were cultured in Dulbecco’s modified Eagle’s medium containing 10% fetal bovine serum. Clonogenic survival assays Colony-forming assays were performed as previously described (Roos et al., 2000). Shortly, U87MG (p53wt) and U138MG (p53mt) glioma cells, growing in log phase were used. Cells were seeded in triplicate at appropriate cell numbers in 60 mm Petri dishes to yield approximately 100 surviving colonies after MNNG treatment. Cells were allowed to attach and then exposed to increasing concentrations of MNNG. After 2–3 weeks colonies were fixed (in acetic acid:methanol:H2O 1:1:8), stained (in 0.01% amido black) and colonies containing 50–100 cells were counted. The surviving fraction was plotted on a log scale and fitted to the linear-quadratic equation. Apoptosis determined by flow cytometry The apoptotic response after genotoxic drug treatment was measured using the flow cytometric method of sub-G1 determination. Treated and untreated cells were harvested, washed once with phosphate-buffered saline (PBS) and fixed in 70% ethanol. Ethanol-fixed cells were stained with propidium iodide (16.5 mg/ml) in PBS after RNase (0.03 mg/ml) digestion. Samples were analysed in a FACS Calibur (Becton Dickinson, Heidelberg, Germany). For each sample 10 000 cells were analysed. The number of apoptotic cells was calculated using the Cell Quest software (Heidelberg, Germany). Drugs and drug treatment MNNG (Sigma, Munich, Germany) stock solution was prepared by dissolving MNNG in dimethyl sulfoxide and then diluting it with sterile dH2O. TMZ (Schering-Plough, Kenilworth, NJ, USA) stocks were prepared by dissolving the drug in ethanol and diluting it with sterile dH2O. Stock solutions were filtered, aliquoted and stored at 801C. Glioma cell lines were treated with increasing concentrations of either MNNG or TMZ and then harvested after 144 h to determine apoptosis. For the time responses, the glioma cells were treated with MNNG and then harvested at the appropriated times. The p53 inhibitor pifithrin-a, which reversibly blocks p53 dependent transcriptional activation (Komarova and Gudkov, 2000), was added 72 h after treatment with either 10 mM MNNG or 100 mM TMZ and apoptosis was determined at time point 144 h. The specific MGMT inhibitor O6BG (Pegg et al., 1993) was added 1 h before MNNG treatment at a concentration of 10 mM. The powerful mitogen, Pasteurella multocida toxin (PMTwt), and the inactive mutant PMTC1165S (Orth et al., 2003) were kept in a 1.1 mg/ml stock solution

Preparation of protein extracts Fractionated cell extracts Cell pellets of treated and untreated samples were suspended in fractionation buffer A (10 mM HEPES–KOH, pH 7.4, 0.1 mM ethylene diaminetetraacetic acid (EDTA), 1 mM ethylene glycol-bis (b-aminoethyl ether), 250 mM sucrose, 1 mM Na3VO4, 0.5 mM phenylmethylsulfonyl fluoride (PMSF) and 10 mM dithiothreitol (DTT)). The cells were lysed by freeze/thaw/vortexing. The lysate was then centrifuged at 10 000 r.p.m. for 10 min and the supernatant containing the cytoplasmic proteins was isolated. The pellet, containing the nuclei, organelles and membranes, were then suspended in fractionation buffer B (20 mM Tris, 1 mM EDTA, 1 mM b-mercaptoethanol, 5% glycerine, 1 mM Na3VO4, 0.5 mM PMSF, 10 mM DTT, pH 8.5). This suspension was homogenized by sonication. After centrifugation at 10 000 r.p.m. for 10 min the supernatant contains the nuclear proteins and the pellet the membrane fragments. This membrane pellet was suspended in fractionation buffer B containing 1% Triton X-100. The protein concentration was determined by the method of Bradford (Bradford, 1976). Cell extracts for MGMT activity assay Cells were harvested and homogenized by sonication in buffer containing 20 mM Tris-HCl, pH 8.5, 1 mM EDTA, 1 mM b-mercaptoethanol, 5% glycerol and the protease inhibitor PMSF (0.1 mM). The extract was centrifuged at 10 000 r.p.m. (10 min) in the cold in order to remove debris and the supernatant was snap-frozen in aliquots using liquid nitrogen and stored at 801C until use. Western blot analysis The method used here is based on the method described by Renart et al. (1979). Protein (30 mg) of cell extracts was separated in a 12% SDS polyacrylamide gel. Thereafter, proteins were blotted onto a nitrocellulose transfer membrane (Protran; Schleicher & Schuell, Dassel, Germany) for 3 h. Membranes were blocked for 2 h at room temperature in 5% (wt/vol) fat-free milk powder in TBS containing 0.1% Tween 20, incubated overnight at 41C with the primary antibody (1:500–1000 dilution), washed three times with 0.1% Tween 20 in TBS, and incubated for 2 h with a horseradish peroxidasecoupled secondary antibody 1:3000 (Amersham Biosciences AB, Uppsala, Sweden). Antibodies used were anti-p53 (BD PharMingen, San Diego, USA), anti-Fas (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA), anti-Bcl-2 and anti-Bax (Santa Cruz Biotechnology Inc.), anti-GAPDH (Ambion Inc.), anti-MGMT (Chemicon International Inc.), anti-ERK2 (Santa Cruz Biotechnology Inc.), anti-gH2AX (Upstate, Dundee, Scotland), anti-caspase-8 (Cell Signalling, Danvers, USA), anti-caspase-9 (Cell Signalling), anti-caspase2 (Neomarkers, Fremont, USA), anti-caspase-7 (Cell signalling) and anti-caspase-3 (Cell Signalling). After final washing with 0.1% Tween 20 in TBS (3 times for 10 min each) blots were developed by using a chemiluminescence detection system (Amersham Biosciences AB). Oncogene


196 Transfection of glioma cells with MGMT, siRNA (p53) and DN-FADD MGMT transfectants were generated by co-transfection of U87MG (p53wt) and U138MG (p53mt) cells with the mammalian expression vector (pSV2MGMT) harbouring the MGMT gene described previously (Kaina et al., 1991) and the pSV2neo plasmid for selection. The transfection method employed was the Effectene transfection kit (Qiagen, Hilden, Germany). In brief, B1.5 mg of pSV2MGMT and 0.2 mg pSV2neo were transfected and then the transfected cells were selected for with firstly 1.5 mg/ml G418 for 72 h and then 0.5 mg/ml until clones formed. G418-resistant clones were picked in 24-well plates and tested for MGMT expression using Western blot and MGMT activity assay. Transfectant clones were routinely cultured in a medium containing 0.5 mg/ml G418 (Sigma-Aldrich, Munich, Germany) that was omitted during the experiments. DN-FADD transfectants were generated in U87MG (p53wt) and U138MG (p53mt) cells by transfecting B1.5 mg pcDNA3-FADD-DN (Tewari and Dixit, 1995) in the same way as MGMT except that this plasmid also contained the neo gene so co-transfection was unnecessary. FADD-DNpositive clones were determined by Western blotting. siRNA transfectants with siRNA targeted towards p53 have already been described (Wischhusen et al., 2003). RNA synthesis RNA synthesis was determined based on a method described previously (Balajee et al., 1997). Approximately 4.0 104 cells were plated in 35 mm Petri dishes and 24 h after plating they were treated with TMZ (0.1 mM) and maintained with the drug for the indicated times. After this period, cells were incubated in a medium containing 3% dialysed FCS and (5-3H)-uridine (3H-Udi 4.0 mCi/ml, Amersham-Pharmacia Biotech, Uppsala, Sweden) for 30 min. Cells were then harvested and separated into two samples. In one of the samples, cells were lysed (NaCl 0.3 M; Tris-HCl pH 8.0 20 mM; EDTA 2 mM; SDS 1% and K proteinase 200 mg/ml) and then transferred to Whatman 17 paper and washed twice with 15% trichloroacetic acid and hydrated ethanol for 30 min, for radioactivity measurement. The second sample was used to determine the

absorbance at 260 nm, for data normalization. The ratio between radioactivity and absorbance expresses the RNA synthesis in these cells. Fas receptor neutralization Fas receptor (CD95/Apo1) was inhibited by adding 1 mg/ml anti-Fas neutralizing antibody (ZB4) (Biozol Diagnostica Vertrieb GmbH, Eching, Germany) 72 h after MNNG treatment and then every 24 h till samples were harvested. The percentage of population undergoing apoptosis was determined after 144 h. MGMT activity assay The MGMT activity in U87MG (p53wt), U138MG (p53mt) and MGMT co-transfected cells were determined using a method based on the radioactive assay where tritium-labelled methyl groups are transferred from the O6-position of guanine to protein in the cell extract (Preuss et al., 1995). HeLa S3 cells expressing MGMT (588786 fmol/mg protein) and HeLa MR cells deficient in MGMT served as positive and negative controls, respectively. The radioactivity of the protein was then measured. For each assay cell extracts containing 200 mg protein, as determined by the method of Bradford (Bradford, 1976), was incubated with [3H]methyl-nitrosourea-labelled calf thymus DNA containing O6-MeG (total 80 000 c.p.m./sample) in 700 mM HEPES–KOH (pH 7.8), 10 mM DTT, 50 mM EDTA for 90 min. Data are expressed as femtomoles of radioactivity transferred from 3H-labelled DNA to protein per milligram of protein within the sample. Acknowledgements This Work was supported by Deutsche Forschungsgemeinschaft, DFG KA 724/13-1 and 13-2 and SFB 432/B7 (Mainz), NGFN-2 (Tu¨bingen) as well as FAPESP (Sao Paulo, Brazil). We gratefully acknowledge Georg Nagel and Andrea Piee-Staffa for technical assistance. We also acknowledge a generous gift of PMT from Professor Klaus Aktories, Freiburg and TMZ from Schering-Plough.

References Allalunis-Turner MJ, Zia PK, Barron GM, Mirzayans R, Day III RS. (1995). Radiat Res 144: 288–293. Anda T, Shabani HK, Tsunoda K, Tokunaga Y, Kaminogo M, Shibata S et al. (2003). Neurol Res 25: 241–248. Balajee AS, May A, Dianov GL, Friedberg EC, Bohr VA. (1997). Proc Natl Acad Sci USA 94: 4306–4311. Becker K, Dosch J, Gregel CM, Martin BA, Kaina B. (1996). Cancer Res 56: 3244–3249. Becker K, Gregel C, Fricke C, Komitowski D, Dosch J, Kaina B. (2003). Carcinogenesis 24: 541–546. Beranek DT. (1990). Mutat Res 231: 11–30. Bobola MS, Berger MS, Ellenbogen RG, Roberts TS, Geyer JR, Silber JR. (2001). Clin Cancer Res 7: 613–619. Bradford MM. (1976). Anal Biochem 72: 248–254. Chen JM, Zhang YP, Wang C, Sun Y, Fujimoto J, Ikenaga M. (1992). Carcinogenesis 13: 1503–1507. Chen ZP, Yarosh D, Garcia Y, Tampieri D, Mohr G, Malapetsa A et al. (1999). Can J Neurol Sci 26: 104–109. Debiak M, Nikolova T, Kaina B. (2004). DNA Repair (Amsterdam) 3: 359–368. Dolan ME, Moschel RC, Pegg AE. (1990). Proc Natl Acad Sci USA 87: 5368–5372. Oncogene

Dumenco LL, Allay E, Norton K, Gerson SL. (1993). Science 259: 219–222. Dunkern T, Roos W, Kaina B. (2003). Mutat Res 544: 167–172. Esteller M, Corn PG, Baylin SB, Herman JG. (2001). Cancer Res 61: 3225–3229. Esteller M, Garcia-Foncillas J, Andion E, Goodman SN, Hidalgo OF, Vanaclocha V et al. (2000). N Engl J Med 343: 1350–1354. Esteller M, Hamilton SR, Burger PC, Baylin SB, Herman JG. (1999). Cancer Res 59: 793–797. Gerson SL. (2004). Nat Rev Cancer 4: 296–307. Gunther W, Pawlak E, Damasceno R, Arnold H, Terzis AJ. (2003). Br J Cancer 88: 463–469. Hammond LA, Eckardt JR, Kuhn JG, Gerson SL, Johnson T, Smith L et al. (2004). Clin Cancer Res 10: 1645–1656. Hampson R, Humbert P, Macpherson P, Aquilina G, Karran P. (1997). J Biol Chem 272: 28596–28606. Hegi ME, Diserens AC, Godard S, Dietrich PY, Regli L, Ostermann S et al. (2004). Clin Cancer Res 10: 1871–1874. Hegi ME, Diserens AC, Gorlia T, Hamou MF, de Tribolet N, Weller M et al. (2005). N Engl J Med 352: 997–1003.


197 Hermisson M, Klumpp A, Wick W, Wischhusen J, Nagel G, Roos W et al. (2006). J Neurochem 96: 766–776. Hickman MJ, Samson LD. (1999). Proc Natl Acad Sci USA 96: 10764–10769. Hickman MJ, Samson LD. (2004). Mol Cell 14: 105–116. Hirose Y, Berger MS, Pieper RO. (2001). Cancer Res 61: 5843–5849. Kaina B, Fritz G, Mitra S, Coquerelle T. (1991). Carcinogenesis 12: 1857–1867. Kaina B, Ziouta A, Ochs K, Coquerelle T. (1997). Mutat Res 381: 227–241. Kanzawa T, Germano IM, Komata T, Ito H, Kondo Y, Kondo S. (2004). Cell Death Differ 11: 448–457. Karran P, Bignami M. (1994). Bioessays 16: 833–839. Kat A, Thilly WG, Fang WH, Longley MJ, Li GM, Modrich P. (1993). Proc Natl Acad Sci USA 90: 6424–6428. Komarova EA, Gudkov AV. (2000). Biochemistry (Moscow) 65: 41–48. Lindahl T. (2000). Mutat Res 462: 129–135. Lips J, Kaina B. (2001). Carcinogenesis 22: 579–585. Ljungman M, Zhang F. (1996). Oncogene 13: 823–831. Margison GP, Santibanez-Koref MF. (2002). Bioessays 24: 255–266. Meikrantz W, Bergom MA, Memisoglu A, Samson L. (1998). Carcinogenesis 19: 369–372. Muller M, Wilder S, Bannasch D, Israeli D, Lehlbach K, Li-Weber M et al. (1998). J Exp Med 188: 2033–2045. O’Connell MJ, Cimprich KA. (2005). J Cell Sci 118: 1–6. Ochs K, Kaina B. (2000). Cancer Res 60: 5815–5824. Orth JH, Blocker D, Aktories K. (2003). Biochemistry 42: 4971–4977. Paz MF, Yaya-Tur R, Rojas-Marcos I, Reynes G, Pollan M, Aguirre-Cruz L et al. (2004). Clin Cancer Res 10: 4933–4938.

Pegg AE. (2000). Mutat Res 462: 83–100. Pegg AE, Boosalis M, Samson L, Moschel RC, Byers TL, Swenn K et al. (1993). Biochemistry 32: 11998–12006. Pepponi R, Marra G, Fuggetta MP, Falcinelli S, Pagani E, Bonmassar E et al. (2003). J Pharmacol Exp Ther 304: 661–668. Pohl U, Wagenknecht B, Naumann U, Weller M. (1999). Cell Physiol Biochem 9: 29–37. Preuss I, Eberhagen I, Haas S, Eibl RH, Kaufmann M, von Minckwitz G et al. (1995). Int J Cancer 61: 321–326. Preuss I, Thust R, Kaina B. (1996). Int J Cancer 65: 506–512. Renart J, Reiser J, Stark GR. (1979). Proc Natl Acad Sci USA 76: 3116–3120. Roos W, Baumgartner M, Kaina B. (2004). Oncogene 23: 359–367. Roos WP, Binder A, Bohm L. (2000). Int J Radiat Biol 76: 1493–1500. Scudiero DA, Meyer SA, Clatterbuck BE, Mattern MR, Ziolkowski CH, Day III RS. (1984). Cancer Res 44: 2467–2474. Silber JR, Blank A, Bobola MS, Ghatan S, Kolstoe DD, Berger MS. (1999). Clin Cancer Res 5: 807–814. Sklar RM, Strauss B. (1981). Nature 289: 417–420. Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ et al. (2005). N Engl J Med 352: 987–996. Tewari M, Dixit VM. (1995). J Biol Chem 270: 3255–3260. Tominaga Y, Tsuzuki T, Shiraishi A, Kawate H, Sekiguchi M. (1997). Carcinogenesis 18: 889–896. Tsaryk R, Fabian K, Thacker J, Kaina B. (2005). Cancer Lett (in press). Wischhusen J, Naumann U, Ohgaki H, Rastinejad F, Weller M. (2003). Oncogene 22: 8233–8245.

Oncogene