Inhibition of the ER Ca²⁺ pump forces multidrug-resistant cells deficient in Bak and Bax into necrosis

Although the role of mitochondria in apoptosis is well established, the endoplasmic reticulum (ER) is becoming increasingly recognized as an important cell death regulator (Boyce and Yuan, 2006; Xu et al., 2005; Szegezdi et al., 2006). In the ER, secretory and membrane proteins are synthesized and folded into their native conformation. Therefore, the ER has to provide a proper environment that can be perturbed, for instance, when post-translational protein modification fails or Ca²⁺ concentrations are altered (Schroder and Kaufman, 2005). Normally, a Ca²⁺ gradient exists across the ER membrane with millimolar levels in the ER lumen and about four orders of magnitude lower concentrations in the cytoplasm. This gradient is maintained by the sarco-endoplasmic reticulum calcium ATPase (SERCA) pump that transports Ca²⁺ into the ER. Application of the specific SERCA inhibitor thapsigargin depletes ER Ca²⁺ stores and causes ER stress (Dong et al., 2006).

ER stress activates signaling pathways, collectively termed the unfolded protein response (UPR). ER stress is sensed by three integral stress receptors including the pancreatic ER kinase-like ER kinase (PERK), activating transcription factor 6 (ATF6) and inositol-requiring enzyme 1α (IRE1α; also known as ERN1), which are kept inactive through their association with the chaperone BiP (also known as Grp78 or Hspa5) (Schroder and Kaufman, 2005). Misfolded proteins in the ER associate with BiP/Grp78, thereby releasing and activating PERK, ATF6 and IRE1α that trigger induction of UPR target genes. The UPR then drives a coordinated program to restore ER homeostasis after mild insults by increasing biosynthesis of factors involved in protein folding, secretion and ER-associated protein degradation. Severe stress, however, can trigger necrotic and apoptotic cell death through mechanisms largely unknown. Multiple mediators have been implicated in ER stress-induced death including caspases, Bcl-2 proteins, phosphatases, kinases and transcription factors (Boyce and Yuan, 2006; Szegezdi et al., 2006). ER stress also induces autophagy (Hoyer-Hansen and Jaattela, 2007), in which parts of the cytoplasm are sequestered within double-membrane vesicles. Normally, autophagy enables cells to endure low nutrient supply and assists the UPR in removing misfolded proteins. Enhanced autophagic vacuolization has been implicated in cell death, but whether autophagy contributes to cell death during ER stress remains unclear (Levine and Yuan, 2005; Maiuri et al., 2007).

Characteristic features of cell death are mitochondrial outer membrane permeabilization (MOMP) and the loss of the mitochondrial transmembrane potential ΔΨ_m (Kroemer et al., 2007). Stress such as mitochondrial Ca²⁺ overload can cause the breakdown of ΔΨ_m by activation of the mitochondrial permeability transition (MPT) pore, a large channel spanning the outer and inner mitochondrial membrane. MPT pore opening results in the swelling of mitochondria and subsequent outer membrane rupture that causes the leakage of proteins including cytochrome c (Green and Kroemer, 2004). Recent evidence suggests that MPT pore opening and subsequent loss of ΔΨ_m plays a role in necrotic but not apoptotic cell death. For example, animals deficient in cyclophilin D, an MPT component, are resistant to necrosis, but not apoptosis, induced during ischemia-reperfusion or elevation of cytosolic Ca²⁺ (Baines et al., 2005; Nakagawa et al., 2005).

During apoptosis, loss of ΔΨ_m seems to be a secondary event not required for cytochrome c release (Chipuk and Green, 2008). MOMP and the release of cytochrome c are instead controlled by Bcl-2 proteins acting on the outer mitochondrial membrane. Bcl-2 proteins are classified into two major groups: antiapoptotic (e.g. Bcl-2) and proapoptotic members, which are further subdivided into multidomain proteins (e.g. Bak, Bax) and BH3-only proteins (e.g. Bim, Bid). In response to apoptotic stimuli the multidomain proteins Bak and Bax undergo a conformational change, leading to their assembly into multimers with channel-forming properties in the outer mitochondrial membrane, which mediate cytochrome c release (Youle and Strasser, 2008). Conformational change of Bak or Bax is inducible by BH3-only proteins acting as allosteric sensors of apoptotic signaling. Once released into the cytosol, cytochrome c and cytosolic factors form the apoptosome, an activating platform for caspases that execute apoptosis by the cleavage of target proteins.

Knockout models provided evidence for the essential involvement of Bax and Bak in apoptotic cell death (Lindsten and Thompson, 2006). Deficiency of Bax and Bak therefore confers resistance to most conventional tumor therapies that employ apoptosis to achieve cell death (Wei et al., 2001; Meijerink et al., 1998). Single inactivation or downregulation of Bak (Kondo et al., 2000; Wang et al., 2001) or Bax (Brimmell et al., 1998; Gutierrez et al., 1999), frequently found in various solid and hematological neoplasias, can be sufficient to induce therapy resistance and poor prognosis (Friess et al., 1998; Sturm et al., 1999; Ionov et al., 2000; Wang et al., 2001; LeBlanc et al., 2002).

Lack of therapies for drug resistant tumors, as exemplified by cells lacking Bax or Bak, represents a severe problem. Since most current anticancer drugs signal through mitochondria, we speculated that the ER might be an alternative target for the induction of tumor cell death. In this report, we demonstrate that, even though deficiency in Bak and Bax (hereafter referred to as Bak/Bax deficiency) protects against apoptosis, it does not compromise necrosis induction by SERCA inhibitors. Our results suggest that interfering with ER function might be a therapeutic approach for certain multidrug-resistant tumor subsets.

Tumor cells deficient in both Bak and Bax are protected against mitochondria-mediated apoptosis that is exploited by most chemotherapeutic drugs to achieve tumor cell death (Fischer et al., 2007). Employing murine embryonic fibroblasts (MEFs) as a model system, our experiments confirmed that MEFs derived from Bak and Bax double knockout mice resist the cytotoxic effects of a wide variety of chemotherapeutic and apoptosis-inducing drugs (Fig. 1A,B and supplementary material Fig. S1). In a screen for new, alternative anti-tumor drugs that could overcome this resistance, we identified the SERCA inhibitor thapsigargin, as a potent inducer of death in Bak/Bax-deficient cells (Fig. 1B). Intriguingly, bak/bax^–/– (note that genes encoding Bak and Bax are referred to as Bak and Bax in the MGI database) cells were less sensitive to tunicamycin, an ER stress inducer that exerts a different mode of action by inhibiting protein N-glycosylation.

Cell death studies of up to 3 days (Fig. 1C) as well as dose-response studies (supplementary material Fig. S2) revealed that, although bak/bax^–/– MEFs were less sensitive to tunicamycin, they were killed by thapsigargin with a time and dose dependency similar to wild-type (WT) cells. To investigate effects on long-term survival, we performed clonogenicity assays (Fig. 1D). Unlike WT cells, Bak/Bax-deficient MEFs survived and retained proliferative capacity after treatment with low doses of tunicamycin (0.1 μg/ml, 48 hours). However, at higher doses (1 μg/ml) bak/bax^–/– cells also succumbed to tunicamycin-induced toxicity. By contrast, after treatment with even low doses of thapsigargin (0.1 μM) neither WT nor Bak/Bax-deficient MEFs retained clonogenicity. Compared with WT cells, even fewer bak/bax^–/– colonies grew, indicating that SERCA inhibitors could potentially circumvent therapy resistance of certain tumor subsets.

To ascertain whether WT and bak/bax^–/– cells died by apoptosis, we investigated activation of caspases (Fig. 2A,B). Cleavage of caspase-3 in response to ER stress was readily observed in WT MEFs, but almost absent in bak/bax^–/– MEFs. These results were confirmed by analyses of caspase substrate cleavage showing that ER stress induced high caspase activity in WT, but not bak/bax^–/– cells. The weak residual caspase activity observed in bak/bax^–/– MEFs might be attributed to caspases functioning upstream of mitochondria, such as caspase-2 or caspase-12, whose functional relevance for ER stress is controversial. Pharmacological caspase inhibition by q-VD-oPh (QVD) resulted in marked reduction of death in WT MEFs, but not bak/bax^–/– cells after 48 hours of treatment with thapsigargin (Fig. 2C). Therefore, SERCA inhibition triggers caspase-dependent death in Bak/Bax-expressing cells, but caspase-independent death in Bak/Bax-deficient cells.

Autophagy can exert both survival- and death-promoting effects (Hoyer-Hansen and Jaattela, 2007). Electron microscopy demonstrated that thapsigargin and tunicamycin induced formation of autophagic vacuoles in both WT and bak/bax^–/– MEFs (Fig. 3A). Moreover, treatment with both drugs resulted in conversion of the microtubule-associated protein light chain (LC3) from its cytosolic form to the lipidated form LC3-II, a marker for autophagic activity (Fig. 3A). Thus, in both cell types ER stress is associated with autophagic alterations.

To analyze the functional role of autophagy, we first used Bak/Bax-expressing atg-5^–/– (Atg5^–/–) MEFs that are deficient in autophagosome formation. Treatment with ER stress agents revealed no difference in cell death between WT and atg-5^–/– cells, suggesting that in Bak/Bax-proficient cells autophagy plays no critical role (Fig. 3B). Furthermore, treatment of Bak/Bax-deficient MEFs with the autophagy inhibitor 3-methyladenine (3-MA) resulted in only a minor reduction of cell death in response to the ER stressors (Fig. 3C). Moreover, siRNA-mediated knockdown of Beclin-1, a protein essentially involved in autophagy, did not provide protection against ER stress in either WT or bak/bax^–/– cells (data not shown). Thus, despite autophagic alterations, neither the death of Bak/Bax-proficient nor of Bak/Bax-deficient cells seems to be caused by autophagy.

ER stress triggers the UPR that has been implicated in some forms of cell death (Schroder and Kaufman, 2005). Therefore, we investigated a contribution of critical UPR components. Even in the absence of Bak/Bax, thapsigargin and tunicamycin induced a time-dependent increase of the ER chaperone BiP, enhanced phosphorylation of the PERK substrate eukaryotic initiation factor 2α (eIF2α) and the IRE1α substrates JNK1/2, indicating a functional ER stress response also in bak/bax^–/– cells (Fig. 3D). In addition, we observed induction of the transcription factor CHOP (Fig. 3E) that has been implicated in ER stress-induced cell death (McCullough et al., 2001). Concurrently, increased levels of the BH3-only protein Bim, a target gene of CHOP (Puthalakath et al., 2007) were detected. Although CHOP has also been reported to activate apoptosis by transcriptional repression of Bcl-2, we found no reduction in Bcl-2. Instead, cleavage and hence activation of the BH3-only protein Bid were evident (Fig. 3E). Similarly to a previous study (Cheng et al., 2001), we detected slightly elevated levels of Bim in bak/bax^–/– cells as compared to WT MEFs, which might be due to a missing survival constraint in the absence of Bax and Bak or other unknown mechanisms.

To clarify the impact of the UPR on cell death, we employed knockout MEFs of the major UPR mediators. Knockout of perk did not decrease cell death in response to thapsigargin or tunicamycin (supplementary material Fig. S3). The kinase IRE1α has also been implicated in ER stress-triggered death pathways and was proposed to form a complex with Bak and Bax that is essential for IRE1α activation (Hetz et al., 2006). Other studies implicated a ternary complex composed of IRE1α, TRAF-2 and the kinase ASK-1 in ER stress-mediated cell death and JNK activation (Nishitoh et al., 1998). However, the impact of such pathways is controversial and several lines of evidence argue against an essential role of IRE1α and associated components. First, we found no impairment of JNK activation in Bak/Bax-deficient cells upon ER stress (Fig. 3D and data not shown), arguing against a non-functional IRE1α pathway in the absence of Bak/Bax. Second, although bak/bax^–/– cells were protected against tunicamycin, cell death still occurred in ire1α^–/– cells to a similar extent as in WT cells (Fig. 3F). Moreover, the caspase inhibitor QVD prevented thapsigargin-induced death in WT and ire1α^–/–, but not in bak/bax^–/– cells. These results indicate that distinct mechanisms are involved in death signaling of ire1α^–/– and bak/bax^–/– cells. In the presence of Bak/Bax, cells die by a caspase-dependent pathway that is not impaired by lack of IRE1α, whereas in bak/bax^–/– cells, death does not require caspases.

To substantiate these results, we analyzed MEFs from jnk-1^–/–, jnk-2^–/– and jnk-1^–/–/jnk-2^–/– (also known as Mapk8) mice. JNK-deficient cells were even more sensitive to ER stress than their WT counterparts (Fig. 3G, supplementary material Fig. 4S). Moreover, no protection was found in ask-1^–/– MEFs (supplementary material Fig. S5). Finally, the JNK inhibitor SP-600125 did not affect cell death induced by ER stress at all, in either WT or in bak/bax^–/– MEFs (Fig. 3H). In conclusion, neither genetic ablation of the ER stress sensors PERK and Ire1α nor disturbance of JNK signaling diminished cell death in response to thapsigargin.

To elucidate how bak/bax^–/– MEFs died, we analyzed their morphology by electron microscopy. After 36 hours of treatment with thapsigargin WT cells showed typical signs of apoptosis including remodeling of mitochondrial cristae (Fig. 4A). By contrast, bak/bax^–/– mitochondria displayed massive swelling indicative of MPT and rupture of the outer membrane.

A permanent MPT can lead to a loss of ΔΨ_m and contribute to cytochrome c release in a Bak/Bax-independent manner. To explore whether the morphological alterations of bak/bax^–/– mitochondria were associated with a loss of ΔΨ_m, we stained cells with the potentiometric dye DiOC₆ (Fig. 4B,C). WT MEFs responded with a loss of ΔΨ_m and cytochrome c release to staurosporine as well as ER stress agents (Fig. 4C,D). By contrast, Bak/Bax-deficient MEFs maintained ΔΨ_m and retained cytochrome c after apoptosis induction by staurosporine and tunicamycin. Strikingly, however, ∼50% of bak/bax^–/– cells displayed loss of ΔΨ_m after 48 hours of thapsigargin treatment. Moreover, Bak/Bax-deficient cells also released cytochrome c. These findings indicate an MPT-mediated ΔΨ_m loss and cytochrome c release in thapsigargin-treated bak/bax^–/– MEFs.

This conclusion was substantiated by the finding that loss of ΔΨ_m was caspase-dependent in WT but not in bak/bax^–/– cells (Fig. 4E). In WT cells, caspases cleave the p75 subunit of complex I of the respiratory chain resulting in permanent loss of ΔΨ_m (Ricci et al., 2004). Provided that caspases are blocked, mitochondria have been shown to restore mitochondrial cytochrome c and ΔΨ_m and remain structurally and functionally intact after MOMP (Martinou et al., 1999). Our results show that caspases are not activated in bak/bax^–/– cells. Nevertheless, these cells are unable to regain respiratory function, further indicating an MPT-mediated loss of ΔΨ_m.

MPT pore opening can be induced by elevated mitochondrial calcium concentration [Ca²⁺]_Mi. Thapsigargin-mediated [Ca²⁺]_ER depletion has been reported to increase mitochondrial Ca²⁺ (Dong et al., 2006). As we found that Bak/Bax-deficient cells were not protected against death induced by a Ca²⁺ ionophore (Fig. 5A), we analyzed whether elevated [Ca²⁺] was involved in thapsigargin-mediated cell death. Indeed, Ca²⁺ chelation by BAPTA-AM protected Bak/Bax-deficient but not WT cells from death and reduced DNA fragmentation in a dose-dependent manner (Fig. 5B,C). Similar results were obtained with cyclopiazonic acid (CPA), another SERCA inhibitor. That BAPTA-AM did not completely prevent cell death is presumably due to close contact sites between the ER and mitochondria, which allow fast Ca²⁺ transfer (Szalai et al., 2000).

Mitochondrial Ca²⁺ uptake is driven through ΔΨ_m, generated by the respiratory chain. Consequently, inhibition of respiration should protect against thapsigargin. When we applied SERCA inhibitors in combination with rotenone, an inhibitor of complex I, bak/bax^–/– MEFs were rescued and even clonogenicity was preserved (Fig. 5D,E). By contrast, WT cells were not rescued by Ca²⁺ chelation or inhibition of respiration, either in short- or in long-term assays (Fig. 5B,D,E). This indicates that the death of apoptosis-competent MEFs does not require mitochondrial Ca²⁺ accumulation and MPT.

Our results demonstrate that ER stress eliminates bak/bax^–/– MEFs by a calcium-dependent, non-apoptotic mechanism. When we analyzed the morphology of bak/bax^–/– MEFs, we found that thapsigargin triggered signs of necrosis, including swelling of mitochondria, cell volume increase and rupture of the plasma membrane, alterations that were not observed in WT cells (Fig. 6A). Recently, the release of high-mobility group 1 protein (HMGB1) was shown to distinguish apoptotic from necrotic cell death (Scaffidi et al., 2002). This strict specificity has recently been questioned, because HMGB1 release was also detected in secondary necrotic, apoptotic cells (Bell et al., 2006). Therefore, to exclude secondary necrosis, we analyzed the release of HMGB1 in the presence of the caspase inhibitor QVD. We found that WT MEFs released only small amounts of HMGB1, and this was completely inhibitable by QVD (Fig. 6B). By contrast, bak/bax^–/– MEFs released substantial amounts of HMGB1 upon stimulation with thapsigargin. This release could not be blocked by QVD and was therefore not caused by secondary postapoptotic necrosis. Thus, these results indicate that the death of bak/bax^–/– cells was indeed due to necrosis.

Recently, it has been demonstrated that activation of poly(ADP-ribose)polymerase-1 (PARP-1) causes translocation of HMGB-1 from the nucleus into the cytoplasm of necrotic cells, from where it is released after plasma membrane rupture (Ditsworth et al., 2007). PARP-1 is activated upon DNA damage and oxidative stress and decorates several acceptor proteins and itself with ADP-ribose moieties that are synthesized by NAD⁺ hydrolysis (Herceg and Wang, 2001). Strong activation of PARP consumes large amounts of NAD⁺ and, in efforts to resynthesize NAD⁺, causes massive ATP depletion (Ha and Snyder, 1999). Depletion of cellular energy may switch apoptotic cell death that is energy-dependent to necrosis (Leist et al., 1997; Ferrari et al., 1998; Los et al., 2002). By contrast, during apoptosis, caspases cleave PARP-1 and thereby maintain ATP levels. Our results show that PARP-1 is cleaved in WT, but not in Bak/Bax-deficient cells upon treatment with thapsigargin (Fig. 6C). This finding indicates that ER stress might trigger PARP-dependent energy depletion in Bak/Bax-deficient MEFs.

Therefore, we analyzed cellular ATP content, which revealed that apoptosis induction by staurosporine led to complete ATP loss after 48 hours in WT, but not bak/bax^–/– MEFs. By contrast, ER stress caused complete ATP consumption in both WT and bak/bax^–/– cells (Fig. 6D). To confirm the functional relevance of PARP-1 in thapsigargin-induced death of bak/bax^–/– MEFs, we analyzed cell death, ATP content and PARP-1 cleavage in the presence or absence of the PARP-1 inhibitor 3-aminobenzamide (3-AB). These experiments revealed that bak/bax^–/– cells treated with 3-AB in addition to thapsigargin had significantly increased ATP levels (Fig. 6E), showed reduced poly(ADP-ribosylation) of PARP-1 (Fig. 6F) and were significantly rescued from necrotic death (Fig. 6E).

To test whether ER stress-induced necrosis of Bak/Bax-deficient cells is a general finding not limited to MEFs, we analyzed thapsigargin-induced cell death in different tumor cells (Fig. 7B,C). Similar to several other cell lines tested, HCT116 colon carcinoma cells were more sensitive to thapsigargin than to equimolar concentrations of etoposide (Fig. 7A). In response to thapsigargin, HCT116 cells expressing both Bak and Bax died at least partially by apoptosis, as assessed by the protective effect of the caspase inhibitor QVD (Fig. 7B). The shRNA-mediated knockdown of Bak alone did not abrogate apoptosis. Strikingly, however, HCT116 cells deficient in Bax or Bax and Bak were protected from caspase-dependent apoptosis, but not from necrotic death. Similarly, thapsigargin killed Bax-deficient DU145 prostate carcinoma cells by a caspase-independent mechanism (Fig. 7C), whereas re-expression of Bax switched the mode of cell death to apoptosis. Thus, by employing different cellular systems we demonstrate that thapsigargin can efficiently trigger non-apoptotic death of Bak/Bax-deficient cells.

Chemotherapeutic agents induce cell death mostly via the mitochondrial apoptotic pathway that is frequently compromised in tumors (Fischer et al., 2007). In this study, we demonstrate that multidrug-resistant Bak/Bax-deficient MEFs can still be efficiently eliminated by ER stress-inducing drugs, such as SERCA inhibitors, which trigger an MPT-dependent necrotic program (Fig. 8). We show that thapsigargin not only eliminates Bak/Bax-deficient MEFs, but also severely impairs their clonogenicity after short-term treatment. Similarly, tumor cells deficient in Bak/Bax expression succumb to a caspase-independent, non-apoptotic death after treatment with thapsigargin. Our results thus suggest that SERCA inhibition might be an attractive strategy for the treatment of multidrug-resistant tumors.

SERCA inhibitors and other ER stress inducers trigger the UPR that is launched primarily to maintain ER homeostasis and cell survival. Classical UPR components were, however, not critical in determining the cell death decision in response to thapsigargin. Distinct regulators of the UPR, such as PERK and ATF6, have been implicated in apoptosis upon ER stress (Szegezdi et al., 2006). However, our results show that thapsigargin-mediated cell death is not compromised in PERK-deficient cells. Similarly, genetic targeting of ATF6 was reported not to protect against ER stress-mediated cell death (Wu et al., 2007; Yamamoto et al., 2007). However, since both PERK and ATF6 independently regulate transcription of the UPR mediator CHOP (Szegezdi et al., 2006), these findings do not exclude a possible role of CHOP in thapsigargin-induced cell death. In both WT and bak/bax^–/– MEFs we observed induction of CHOP and its downstream target Bim that has been implicated in apoptosis (Puthalakath et al., 2007), but not necrosis during ER stress.

Another mechanism by which the UPR could impact on cell death is the IRE1α pathway. Bak/Bax were suggested to directly interact with IRE1α resulting in activation of JNK (Hetz et al., 2006). However, neither the knockout of IRE1α, JNK nor other components of this pathway or JNK inhibition conferred protection from ER stress. Our observation that death of ire1α^–/– cells was caspase dependent, whereas it was caspase independent in bak/bax^–/– cells, rather suggests that Bak/Bax and IRE1α are acting on separate pathways.

In WT and Bak/Bax-deficient MEFs ER stress rapidly induced autophagic alterations. This indicates that autophagic alterations per se are independent of proapoptotic Bcl-2 proteins. Autophagy has been implied as a death mechanism substituting for deficient apoptosis under certain conditions (Levine and Yuan, 2005; Maiuri et al., 2007). Interestingly, autophagy was more pronounced in bak/bax^–/– than in WT cells, as judged by an increased occurrence of autophagic vacuoles and faster lipidation of LC3. However, ER stress induced by thapsigargin and tunicamycin induced a similar autophagic response in Bak/Bax-deficient cells, but only thapsigargin treatment resulted in massive cell death. Furthermore, inhibition of autophagy did not rescue bak/bax^–/– or WT MEFs, suggesting that an autophagic response is associated but not relevant for thapsigargin-induced cell death.

Electron microscopy revealed that SERCA inhibitors strongly affected mitochondrial morphology displaying signs of MPT. Recent reports showed that MPT is not required for the release of proapoptotic factors from mitochondria (Baines et al., 2005; Nakagawa et al., 2005). However, the question of whether MPT contributes to apoptosis in response to classical MPT triggers, such as Ca²⁺, in WT cells or in situations where Bak/Bax-mediated MOMP is blocked, remained unsolved. Our results demonstrate that MPT is not involved in Ca²⁺-induced apoptosis in WT MEFs, and promotes necrosis rather than apoptosis in bak/bax^–/– MEFs. Therefore, the proapoptotic effect of Ca²⁺ probably relies on Bak/Bax-mediated MOMP. These results indicate that ER stress-induced death is primarily caused by activation of BH3-only proteins in Bak/Bax-proficient cells. Consistently, we observe an activation of Bid and Bim by SERCA inhibitors. ER stress activates Bim via protein stabilization and transcriptional induction, and the knockout of Bim partially protects thymocytes and macrophages against ER stress (Puthalakath et al., 2007). Nevertheless, we found no protection against ER stress in Bid or Bim single knockout MEFs (data not shown). ER Ca²⁺ release was also shown to activate calcineurin that enhances the proapoptotic potential of Bad (Wang et al., 1999) and loss of Puma protected motoneurons from ER-stress-induced apoptosis but not necrosis (Kieran et al., 2007). It is therefore likely that, depending on the cell type or ER stressor, different and eventually redundant BH3-only proteins are involved in ER stress-mediated cell death.

Bak/Bax-deficient MEFs are unable to trigger apoptosis and are thus highly resistant to chemotherapeutic agents. A major finding is therefore that SERCA inhibitors were still able to trigger cell death in those cells. We could exclude autophagy as a death mechanism employed by these agents. Our results suggest that cells die by necrosis upon ER stress. This is in line with the finding that conditional bak/bax^–/– mice react abnormally to ER stress and show strongly elevated liver transaminase levels and massive tissue damage, indicative of necrosis (Hetz et al., 2006). Here, we demonstrate clearly that ER Ca²⁺ release elicits MPT and necrotic death of bak/bax^–/– MEFs.

Interestingly, some Bcl-2 members also localize in ER membranes and impact on Ca²⁺ signaling (Pinton et al., 2008; Rong and Distelhorst, 2008). There is evidence that antiapoptotic Bcl-2 proteins regulate Ca²⁺ channels, such as SERCA or inositol trisphosphate receptors. Overexpression of Bcl-2 for instance, lowers [Ca²⁺]_ER, resulting in less cytosolic Ca²⁺ accumulation (Pinton et al., 2000). Less Ca²⁺ then enters mitochondria, which possibly explains the lack of cytochrome c release and apoptotic cell death. It has been suggested that Bak/Bax also directly influence basal ER Ca²⁺ load (Scorrano et al., 2003). However, as Bax is a cytosolic protein that needs to be activated in order to insert into membranes (Hsu and Youle, 1998; Suzuki et al., 2000; Zhou and Chang, 2008), this would not support a role for Bax on ER Ca²⁺ load under normal conditions. Furthermore, anti-apoptotic Bcl-2 proteins even reduce [Ca²⁺]_ER in cells lacking Bax and Bak, implying that Bcl-2 operates autonomously with respect to Ca²⁺ regulation (Xu et al., 2008).

In our experiments the death of Bak/Bax-deficient MEFs was causally linked to ER Ca²⁺ release, as evident by a protective effect of the Ca²⁺-chelator BAPTA-AM. Strikingly, bak/bax^–/– MEFs could be completely rescued and their clonogenicity sustained by inhibition of the respiratory chain that impairs mitochondrial Ca²⁺ uptake. This demonstrates that mitochondrial Ca²⁺ uptake regulated by ΔΨ_m is a determinant of cell death in Bak/Bax-deficient MEFs.

It is tempting to speculate about why WT cells undergo apoptosis, whereas Bax/Bak-deficient die by Ca²⁺- and MPT-mediated necrosis. ER stress usually triggers apoptosis by BH3-only protein-mediated Bak and Bax activation. As a result, activated caspases shut down respiration by the proteolysis of electron chain components and amplify MOMP by cleavage of Bid. In this scenario, the loss of ΔΨ_m prevents further uptake of Ca²⁺. In the absence of active caspases loss of cytochrome c still downregulates ΔΨ_m, which could protect mitochondria from the deleterious effects of Ca²⁺ accumulation. Mitochondria are then able to restore cytochrome c and respiration. In cells lacking Bak and Bax, cytochrome c is not quickly released via Bak/Bax pores and caspases are not activated. This probably enhances ΔΨ_m and Ca²⁺ uptake until MPT induces swelling and final rupture of bak/bax^–/– mitochondria.

It is also possible that mitochondria of Bak/Bax-deficient MEFs have a higher sensitivity to MPT-induced membrane rupture. Unlike elongated mitochondria of WT cells, mitochondria from bak/bax^–/– MEFs are small and spherical, probably because of deficient mitochondrial fusion (Karbowski et al., 2006). Therefore, Bak/Bax-deficient mitochondria can be assumed to have a maximal volume to surface ratio, resulting in higher susceptibility to swelling-induced outer membrane rupture. By contrast, elongated WT mitochondria could presumably withstand a substantial volume increase without loosing outer membrane integrity.

Apart from mitochondrial disruption, we find that PARP-1 was activated in bak/bax^–/– cells, whereas inhibition of PARP-1 prevented ER stress-induced death. PARP-1 is a major regulator of necrotic cell death and persistent PARP-1 activation is known to cause deleterious ATP depletion in necrotic cells (Ha and Snyder, 1999). As we show here, PARP-1 is cleaved and inactivated by caspases in WT cells, but remained active in Bak/Bax-deficient MEFs under ER stress. ATP consumption triggered by PARP-1 under ER stress is probably crucial for Bak/Bax-deficient cells, as mitochondrial disruption compromises their main source of ATP synthesis. It is tempting to speculate that the pronounced activation of autophagy in bak/bax^–/– MEFs treated with ER stress drugs might be a mechanism to compensate for the loss of ATP to enable survival. PARP-1 activation is readily triggered by DNA damage and oxidative stress. Since calcium overload, PT pore opening and mitochondrial rupture are accompanied by increased oxidative stress (Green and Reed, 1998), similar mechanisms are presumably responsible for PARP-1 activation in bak/bax^–/– MEFs.

In conclusion, our study demonstrates that induction of necrosis by ER stress might be an attractive approach to overcome the resistance of Bax- or Bak/Bax-deficient tumor cells. Compared with agents that affect other ER functions, SERCA inhibitors might be particularly promising candidate drugs. Experiments performed within the National Cancer Institute's anticancer screens already demonstrated that thapsigargin possesses a broad growth-inhibitory activity in hematopoietic and non-hematopoietic tumor cells with a 10- to 100-fold greater efficiency as compared with traditional anticancer agents (Denmeade et al., 2003). The well-known antimalarial artemisinin derivatives are potent inhibitors of SERCA that occupy the same binding site as thapsigargin (Uhlemann et al., 2005). Interestingly, it has been found that artemisinins not only exert antimalarial activity but also profound cytotoxic activity in otherwise drug-resistant tumor cells (Efferth et al., 2006). Strategies employing SERCA inhibitor-derived prodrugs that are activated specifically in tumor cells (Denmeade et al., 2005; Nakase et al., 2009) might, therefore, be promising tools for the treatment of certain highly malignant tumor entities.

Knockout and respective wild-type (WT) MEFs were kindly provided by Randal J. Kaufman (University of Michigan Medical Center, Ann Arbor, MI), Hans-Uwe Simon (University of Bern, Bern, Switzerland), Andreas Strasser (Walter and Eliza Hall Institute, Melbourne, Australia), Hidenori Ichijo (Tokyo Medical University, Tokyo, Japan), Erwin F. Wagner (Institute of Molecular Pathology, Vienna, Austria) and Stanley J. Korsmeyer (Harvard Medical School, Boston, MA) and maintained in DMEM supplemented with 10% FCS and antibiotics. U2OS, Hey and DU145 cells were grown in the same medium. T98, HT1080 and HeLa cells were kept in complete RPMI-1640. HCT116 WT and Bax-deficient cells were provided by Bert Vogelstein (Johns Hopkins Cancer Center, Baltimore, MD) and cultured in McCoy's 5A medium. Stable knockdown of Bak was performed in HCT116 WT and HCT116-Bax knockout cells as described previously (Hemmati et al., 2006).

Cells were seeded at 5×10⁴ cells per well in six-well plates and treated for 24-96 hours with thapsigargin (1.5 μM, or as indicated), tunicamycin (1 μg/ml, or as indicated) or staurosporine (1 μM; all from Calbiochem, Darmstadt, Germany) in the presence or absence of QVD (20 μM; MP-Biomedicals, Eschwege, Germany), BAPTA-AM (Alexis Biochemicals, Lörrach, Germany), rotenone, 3-aminobenzamide, 3-methyladenine (Sigma-Aldrich, Steinheim, Germany) or SP-600125, cyclopiazonic acid and A23187 (Calbiochem). Taxol, etoposide, cisplatin, actinomycin D and camptothecin were obtained from Calbiochem.

Cells were lysed in buffer containing 1% NP-40, 50 mM Tris, pH 7.5, 350 mM NaCl, 0.5 mM EDTA, 2 mM DTT (dithiothreitol), 10 nM NaF, 0.5 mM sodium vanadate and protease inhibitor cocktail. Proteins were separated on 8-15% polyacrylamide gels, transferred to polyvinylidene fluoride membranes and detected by chemiluminescence.

The following primary antibodies were used: β-actin (clone AC-74) and α-tubulin (clone DM1A) from Sigma-Aldrich, Bim (clone 3C5) from Alexis Biochemicals, active Bak and Bax (rabbit polyclonal) from Upstate (Lake Placid, NY), Bcl-2 (clone C-2), CHOP/GADD153 (clone B-3), GAPDH (clone 0411) from Santa Cruz Biotechnology (Santa Cruz, CA), LC3 (clone 2G6) from Nanotools (Teningen, Germany), BiP/Grp78 (clone 40), caspase-3 (clone 46), cytochrome c (clone 6H.B4), PARP-1 (clone 7D3-6) from BD Biosciences, Bid (goat polyclonal) from R&D Systems (Wiesbaden, Germany), and peIF2-Ser51, JNK, pJNK-Thr183/Tyr185, pJNK-Thr221/Tyr223 (all rabbit polyclonal) from Cell Signaling Technology (Danvers, MA).

ATP content was measured by bioluminescent detection (ApoSENSOR, Biovision, Mountain View, CA) according to the manufacturer's instructions. Triplicates of 10⁴ cells per condition were transferred into white 96-well plates and incubated for 5 minutes in nucleotide-releasing buffer. After addition of luciferase, light emission was measured in a luminometer (Berthold Technologies, Bad Wildbad, Germany).

HMGB1 release was assessed by immunoblot analysis. Cells were seeded at 10⁵ per well into six-well plates. After 24 hours the supernatant was replaced with serum-free medium containing the indicated drugs. After a further 24 hours the medium was collected and cleared by centrifugation. A 1/100 volume of 2% sodium deoxycholate was added and samples were incubated at 4°C for 30 minutes. Proteins were precipitated with trichloroacetic acid at 4°C overnight. Immunoblot analysis was performed using a rabbit anti-HMGB1 antibody (BD Biosciences).

Caspase activity was measured using a luminescence assay (Caspase-Glo 3/7, Promega) according to the manufacturer's protocol. Cell lysates (25 μg protein) were added to a proluminescent caspase substrate containing the tetrapeptide sequence DEVD. Light production indicative of substrate cleavage was monitored for up to 180 minutes.

For electron microscopy, MEFs were plated at a density of 10⁶ cells in 175 cm² cell culture flasks. After 24 hours the cells were stimulated with either thapsigargin or tunicamycin or left untreated and harvested 36 hours later. Sections (50 nm) were prepared following a standard protocol (Viemann et al., 2007) and examined on an electron microscope (Zeiss EM 902A, Zeiss, Jena, Germany).

Cell death was determined by the uptake of propidium iodide (PI, 2 μg/ml) and analyses of DNA fragmentation were performed as described previously (Nicoletti et al., 1991). Loss of ΔΨ_m was analyzed using the cationic dye DiOC₆ (Fischer et al., 2005). Cytochrome c release was analyzed as previously described (Waterhouse and Trapani, 2003). GFP-Bax expression was controlled by measuring fluorescence in Fl-1. Each experiment was performed independently at least three times. An individual experiment analyzing 10,000 cells on a FACSCalibur flow cytometer (Becton Dickinson) was carried out in duplicate.

The colony-forming potential of adherent cells was measured as described previously (Janssen et al., 2007). In brief, 3×10⁴ cells per well were seeded onto six-well plates and treated with the indicated drug concentrations on the following day. After 24 or 48 hours cells were washed and maintained in normal medium for another 24 hours until the untreated controls reached 90% confluency. Cells were split 1:10 and reseeded onto new plates. After 6 days viable cells were stained with 0.5% Crystal Violet and solubilized in 33% acetic acid. Absorbance was measured at 590 nm (A590). Relative proliferation was defined as A590 of test well × 100 divided by the A590 of untreated well. Experiments were performed independently three times in duplicate.

Statistical analyses were performed using two-tailed t-tests for equality of means (SPSS Statistics 17.0 software). A P-value of less than 0.05 was considered significant.