The Proteasome Is Required for Rapid Initiation of Death Receptor-Induced Apoptosis

ABSTRACT Due to their tremendous apoptosis-inducing potential, proteasomal inhibitors (PIs) have recently entered clinical trials. Here we show, however, that various PIs rescued proliferating tumor cells from death receptor-induced apoptosis. This protection correlated with the stabilization of X-linked IAP (XIAP) and c-FLIP and the inhibition of caspase activation. Together with the observation that PIs could not protect cells expressing XIAP or c-FLIP short interfering RNAs (siRNAs) from death receptor-induced apoptosis, our results demonstrate that PIs mediate their protective effect via the stabilization of these antiapoptotic proteins. Furthermore, we show that once these proteins were eliminated, either by long-term treatment with death receptor ligands or by siRNA-mediated suppression, active caspases accumulated to an even larger extent in the presence of PIs. Together, our data support a biphasic role for the proteasome in apoptosis, as they show that its constitutive activity is crucial for the rapid initiation of the death program by eliminating antiapoptotic proteins, whereas at later stages, the proteasome acts in an antiapoptotic manner due to the proteolysis of caspases. Thus, for a successful PI-based tumor therapy, it is crucial to carefully evaluate basal proteasomal activity and the status of antiapoptotic proteins, as their PI-mediated prolonged stability might even cause adverse effects, leading to the survival of a tumor.

Apoptosis is a fundamental process that is essential for normal tissue homeostasis and development (31,40,56). Hence, the dysregulation of apoptosis has been implicated in several human diseases,includingautoimmunedisordersandmalignancies.Apoptosis can be instigated via two principal pathways, both of which result in the activation of a class of aspartate-specific cysteinedependent proteases, called caspases (4), that lead to the demise of the cell via the limited proteolysis of a multitude of cellular substrates (10). In the extrinsic pathway, ligandinduced oligomerization of death receptors, such as TRAIL-R1, TRAIL-R2, or CD95, results in the direct recruitment of the adapter molecule Fas-associated death domain (FADD) and the initiator caspase -8 or -10 that, together, form the death-inducing signaling complex (DISC) (49). Signaling through tumor necrosis factor receptor 1 (TNF-R1), in contrast, appears to be more complex (59) and is proposed to proceed via two sequential complexes, in which the death-inducing signaling complex II containing FADD and an initiator caspase is formed only in the cytosol (39). Nevertheless, despite the fact that procaspase-8 can also be activated independently of DISC formation (51), receptortriggered dimerization of initiator caspases represents a crucial event in this pathway that results in their autocatalytic processing and activation. In most cell types, activated caspase-8 stimulates the second principal death route, the mitochondrial or intrinsic death pathway, by cleaving Bid, a BH3-only proapoptotic member of the Bcl-2 family (32).
The caspase-generated truncated Bid (tBid) fragment promotes the release of the proapoptotic factors cytochrome c and Smac from the mitochondria. Once released into the cytoplasm, cytochrome c associates with apoptotic proteaseactivating factor 1 and procaspase-9 to form the apoptosome, which leads to the activation of caspase-9 and the effector caspase-3 (17,28).
Both death pathways are tightly controlled by multiple mechanisms that efficiently prevent caspase activation (50). For example, the generation of active caspase-8 is blocked by the recruitment of the cellular FLICE-like inhibitory protein (c-FLIP) into the DISC (29). Two isoforms, c-FLIP-long and c-FLIP-short, are major detectable splice variants that have been the subjects of several studies so far. While c-FLIP-short consists of mainly two death effector domains that are required for the interaction with FADD, c-FLIP-long is structurally more closely related to caspase-8 but lacks key residues that are necessary to form an active caspase and is therefore completely devoid of any catalytic activity. Due to their homologies, both c-FLIP isoforms are able to form heterodimers with caspase-8 at the DISC and thereby block the processing and activation of this initiator caspase, which is crucial for the induction of death receptor-induced apoptosis (33,48). On the other hand, antiapoptotic members of the Bcl-2 family indirectly inhibit caspase-9 activation by preventing the release of proapoptotic factors from the mitochondria. Further complexity is added to the regulatory pathways involved in apoptosis signaling by the inhibitor-of-apoptosis proteins (IAPs) that include X-linked IAP (XIAP), cIAP1, and cIAP2, all of which are potent inhibitors of the active caspase-9, -7, and -3 (6,47). IAPs are characterized by two distinct motifs, the baculovirus IAP repeats that are crucial for caspase inhibition and a carboxyl-terminal RING domain that enables IAPs to catalyze the degradation of both themselves and selected target proteins via the ubiquitinproteasome pathway (57).
Relative to all the death receptor ligands studied, TRAIL has attracted immense attention due to its ability to preferentially kill tumor cells while leaving most normal cells unharmed in vitro (61) and in vivo (1). However, clinical trials are hampered by the fact that more than 50% of all human cancers are resistant to the cytotoxic activity of TRAIL, demonstrating the necessity for alternative treatment modalities. Promising results were obtained from studies showing that the cytotoxic potential of TRAIL was synergistically enhanced by a combined treatment with chemotherapeutics (16) or radiation (36). Also, the simultaneous treatment with proteasomal inhibitors (PIs), such as bortezomib (PS-341), lactacystin, or MG-132, yielded encouraging results, as all of these compounds synergistically accelerated and enhanced TRAIL-or TNF-induced apoptosis in a variety of tumor cells (11,19,30,60). In view of this success, the proteasome inhibitor bortezomib has recently entered clinical practice as a treatment for multiple myeloma and is also undergoing clinical trials for other types of cancer (45,58).
The ubiquitin-proteasome pathway plays a central role in the regulation of cell cycle control, transcription, signal transduction, and apoptosis (27). In eukaryotes, it is the major machinery that mediates the targeted degradation of many key regulatory proteins, including p53, cyclins, and cyclin-dependent kinase inhibitors, as well as IB␣ (41). Although many diverse mechanisms have been proposed regarding which inhibition of the proteasome sensitizes cells for apoptosis, the precise mode of action remains elusive. The observations that many proapoptotic proteins, such as TRAIL-R2, Smac, Bax, tBid, Bik, and Bim, as well as various caspases, are degraded by the proteasome explain, at least partially, the apoptotic function of proteasomal inhibitors (35,41,42). However, several antiapoptotic proteins, such as c-FLIP, Bcl-2, and IAPs, also represent prominent targets of the proteasome (7,12,44,63). Hence, stabilization of these potent caspase antagonists by PIs should confer resistance to apoptotic stimuli; this is a hypothesis that, surprisingly, was not thoroughly investigated before. Therefore, we analyzed such a putative mechanism in more detail in HeLa cells and found that death receptor-induced apoptosis was indeed significantly blocked in the presence of various proteasome inhibitors. The inhibition of apoptosis induction correlated well with the stabilization of XIAP and c-FLIP, and the latter protein was efficiently recruited to the DISC (thereby preventing caspase-8 processing) only in the presence of PIs. The suppression of XIAP and c-FLIP expression by short interfering RNA (siRNA) technology not only abrogated the protective effect of proteasome inhibitors but also accelerated their cytotoxic potential in combination with death receptor ligands. Thus, our data demonstrate a biphasic role for the proteasome in apoptosis in which the degradation of antiapoptotic proteins, such as XIAP and c-FLIP, is necessary for a rapid initiation of the death receptor pathway.
Construction of small interfering RNAs and stable transfection. For the suppression of c-FLIP and XIAP expression, siRNAs were designed using the Dharmacon siDESIGN Center. The selected sense sequences were 5Ј-GCAGTC TGTTCAAGGAGCA-3Ј for c-FLIP and 5Ј-AATAGTGCCACGCAGTCTA-3Ј for XIAP. Complementary oligonucleotides consisting of sense, hairpin loop, and antisense sequences were annealed and ligated into the pSilencer siRNA expression vector according to the manufacturer's instructions (Ambion, United Kingdom). HeLa D98 cells were stably transfected by electroporation using the Bio-Rad gene pulser (500 F, 160 V). After hygromycin selection, several clones were obtained and the successful reduction of c-FLIP and XIAP expression was controlled by Western blot analyses.
Preparation of cell extracts and Western blotting. Cell extracts were prepared as described previously (26). Briefly, cells were lysed for 30 min at 4°C in lysis buffer A containing 50 mM Tris-HCl (pH 7.6), 150 mM NaCl, 1 mM dithiothreitol (DTT), 1% NP-40, and a protease inhibitor cocktail. After removing nuclei and cell debris by centrifugation (10,000 ϫ g) at 4°C, protein concentrations were determined with the Bio-Rad protein assay. Subsequently, proteins were separated under reducing conditions on a sodium dodecyl sulfate-polyacrylamide gel and electroblotted to a polyvinylidene difluoride membrane (Amersham, Braunschweig, Germany). Following incubation with the various antibodies, the proteins were visualized by enhanced chemiluminescent staining using ECL reagents (Amersham).
Measurement of cell death. Cell death determinations were done with the standard TNF cytotoxicity assay (crystal violet assay) that is based on the staining of viable cells (24). Briefly, cells (2 ϫ 10 5 /ml) were seeded into 96-well microtiter plates in 100 l culture medium. Cells were incubated with the death stimuli for the indicated times at 37°C, and viable cells were stained with 20% methanol containing 0.5% crystal violet and solubilized in 33% acetic acid. The absorbance was measured at an optical density of 590 nm (A 590 ). Percent specific cell death is defined as 100 Ϫ (A 590 of test well ϫ 100/A 590 of untreated well). Each experiment was performed independently at least three times, and an individual experiment was carried out in triplicate. The percent inhibition of cell death by MG-132 (see Immunoprecipitation. Immunoprecipitation reactions were performed as described previously (51). Briefly, cell extracts were prepared from 1 ϫ 10 7 cells in 500 l lysis buffer without DTT. For the precipitation, 1 g polyclonal caspase-8 antibody and 30 l protein G-Sepharose (Sigma) were added to the extracts and rotated for 4 h or overnight at 4°C. The Sepharose beads were extensively washed in lysis buffer and then analyzed by Western blotting.

Proteasome inhibition blocks TNF-induced caspase activation and apoptosis.
In an attempt to better understand the mode of action of PIs on death receptor-induced apoptosis, we treated two HeLa cell lines that display different apoptosis susceptibilities toward death receptor ligands (24) with increasing concentrations of TNF in the absence or presence of the peptide aldehyde inhibitor MG-132. Surprisingly, we found that MG-132 protected both cell lines very efficiently from TNFinduced apoptosis even when TNF concentrations as high as 100 ng/ml were used (Fig. 1A). This effect not only was determined using the crystal violet assay (Fig. 1A) but also was visualized by micrographs taken from HeLa D98 cells that were incubated for 2.5 or 5 h with TNF in the absence or presence of MG-132 (Fig. 1D). MG-132-mediated protection was caused by the inhibition of caspases, as processing and activity of both caspase-8 and -3 were significantly delayed by this compound in both cell lines ( Fig. 1B and C). Although this was clearly evidenced by the diminished generation of active caspase-8 and caspase-3 fragments, the inhibitory effect of MG-132 was most obvious with regard to the unprocessed proforms of both caspases. Depending on the apoptosis susceptibility of the cell line, TNF treatment resulted either in an almost complete loss of both caspase proforms, as evidenced in TNF-sensitive HeLa D98 cells (Fig. 1B, left panel), or in an at least significant reduction of the proforms, as was observed in the more-resistant H21 cell line (Fig. 1B, right panel). In the presence of MG-132, however, TNF induced only a marginal decrease of both proforms in both cell lines, clearly demonstrating that the inhibition of the proteasome blocks TNFinduced caspase processing. Note that protection by MG-132 was much more pronounced in the resistant HeLa H21 cells as the generation of active caspase fragments became almost completely inhibited even after a 20-h incubation with TNF.  1C) activities that were observed in both cell lines and also with the cleavage of ␣-fodrin into the caspase-3-dependent 120-kDa fragment (25) that was accordingly delayed when the cells were treated with TNF in the presence of MG-132 (data not shown). Interestingly, the incubation of the cells with and without MG-132 in the absence of TNF for up to 20 h did not reveal any significant differences in the amount of procaspase-8 and -3 (data not shown), suggesting that these procaspases are not direct targets of the proteasome. In contrast, after the initial delay in caspase activation, we even observed enhanced caspase-8 and caspase-3 ( Fig. 1C) cleavage activities when the cells were treated with TNF in the presence of MG-132. Thus, in contrast to the procaspases, active caspases are degraded by the proteasome, which so far was shown for only caspase-3 (53). Inhibition of the proteasome transiently blocks TNF-induced apoptosis in a dose-and time-dependent manner. The activity of the peptide aldehyde inhibitor MG-132 appears to be not exclusively specific for the proteasome (9). Therefore, we also performed the experiments in the presence of various other compounds, such as ALLN and ALLM, that are effective and weak inhibitors of the proteasome, respectively, as well as with CLC, which is among the most selective PIs known (9). Except for ALLM, which did not interfere with TNF signaling at any concentration tested, all the other compounds reproducibly blocked TNF-induced apoptosis of both HeLa cell lines in a dose-and time-dependent manner ( Fig. 2A and B). The protective effect was of a transient nature, as all of the various PIs efficiently delayed TNF-induced apoptosis of HeLa D98 cells up to 6 h, but they could not protect these cells at later time points (Fig. 2B, left panel). This is consistent with the caspase activation profiles shown in Fig. 1C. On the other hand, more than 80% of HeLa H21 cells survived even a 20-h TNF treatment in the presence of either of these compounds (Fig. 2B, right panel). However, these cells also eventually succumbed to apoptosis, even in the presence of PIs, when they were treated with TNF for up to 36 h (data not shown). Together, these data not only suggest that the protective effect of these inhibitors is indeed caused by their ability to block the 26S proteasome but also demonstrate that the efficiency of this effect critically depends on the strength of the individual apoptotic response. Furthermore, and more intriguingly, they indicate that proteasomal activity might be required for the efficient initiation of the death receptor pathway.
To further analyze the conditions required for a maximal protective effect, we have also investigated the necessity to preincubate the cells with these PIs. Relative to a 1-h preincubation time, the simultaneous addition of the proteasome inhibitors, together with TNF, resulted in a weaker protection, whereas preincubation for more than 1 h did not further increase their protective effect. In contrast, no protection was observed when the PIs were added 1 h or 2 h after TNF treatment (data not shown). Thus, the preincubation of HeLa D98 cells with either MG-132 or CLC for 1 h before the addition of TNF was both necessary and sufficient to achieve optimal protection against the cytotoxic activity of this cytokine (Fig. 2C). Hence, subsequent experiments were conducted following a 1-h preincubation period with the PIs.
Inhibition of the proteasome also blocks TRAIL-and anti-CD95-induced apoptosis and caspase activation. Next we investigated whether the observed effect is restricted to the TNF system or whether PIs would generally protect cells from death receptor-induced apoptosis. To this end, we analyzed TRAIL and CD95 sensitivities of HeLa D98 cells that either were left untreated or were preincubated for 1 h with MG-132. Similar to the results obtained so far with TNF, inhibition of the proteasome by MG-132 also transiently protected the cells against the cytotoxic activities of these two death receptor ligands (Fig. 3A). However, the protection was clearly more pronounced in the CD95 signaling pathway, which induces a significantly slower apoptotic response in these cells than TRAIL. This is reminiscent of the scenario observed with the less sensitive HeLa H21 cells, which are also more efficiently protected from TNF-induced apoptosis by PIs than are the TNF-sensitive HeLa D98 cells. Nevertheless, depending on the dose and time that the two death receptor ligands were ap-plied, the cells also eventually succumbed to apoptosis in the presence of MG-132 ( Fig. 3A and data not shown). Similar to TNF-induced apoptosis, MG-132-mediated protection from TRAIL-and anti-CD95-induced killing correlated closely with a significant delay of processing caspase-8 and caspase-3 (Fig.  3B) and the corresponding activities of these caspases ( Fig. 3C and data not shown). In addition, MG-132 also significantly delayed TRAIL-induced apoptosis and caspases-8 and -3 activation in HeLa D98 cells in the absence of cycloheximide (data not shown), indicating that the PI-mediated protection also occurs under more physiological conditions, which might be especially relevant for TRAIL-mediated signaling in vivo. Transient inhibition of death receptor-induced apoptosis and caspase activation by PIs is a common event observed in several cell lines. Next we analyzed whether our findings can also be applied to other experimental cell systems. To this end, we compared caspase-3 (DEVD) activities in several tumor cell lines of various origins that were treated with TNF in the absence or presence of MG-132. Similar to the data obtained with HeLa D98 cells, the inhibition of the proteasome by MG-132 also transiently blocked TNF-induced caspase-3 activation of MCF-7/casp-3 breast carcinoma cells, KB cervical carcinoma cells, and HCT116 colon carcinoma cells (Fig. 4A). Consistent with the transient nature of this effect, we also found that MG-132 efficiently protected the various tumor cell lines from death receptor-induced apoptosis in a short-term (4 h to 8 h) assay but failed to do so following a longer (24 h) stimulation time (Fig. 4B). Similar results were obtained with the lung carcinoma cell line A549, albeit not to the same extent (data not shown). Although MCF-7/casp-3 cells that do not express the CD95 receptor were treated with only TNF and TRAIL, our data clearly show the general protective impact of PIs on death receptor-induced apoptosis.
Proteasomal activity correlates with apoptosis susceptibility. Our experiments performed so far indicate that the proteasome might be required for the initiation of the apoptotic cascade induced by death receptors. Thus, diminished proteasomal activity might be one mechanism contributing to the apoptosis-resistant phenotype of HeLa H21 cells. To investigate this hypothesis, we compared proteasomal activities in cellular extracts of untreated and TNF-stimulated HeLa D98 and H21 cells over a period of 20 h by determining their capabilities to cleave the fluorogenic peptide LLVY-AMC that represents a substrate for the 20S proteasome. Indeed, extracts of untreated HeLa D98 cells reproducibly displayed a proteasomal activity that was on average approximately 1.5fold higher than that observed in extracts of untreated HeLa H21 cells, a difference that was even further enhanced when the cells were stimulated with TNF (Fig. 5A, left panel). When the cell extracts were analyzed in the presence of MG-132, no LLVY-AMC cleavage could be observed, demonstrating that the cleavage was mediated by the proteasome (data not shown). In contrast, regardless of whether the cells were stimulated with TNF, both cell lines displayed similar ubiquitin C-terminal hydrolase activities, as determined by cleavage of the fluorogenic substrate LRGG-AMC (Fig. 5A, right panel). These results show that apoptosis-sensitive HeLa D98 cells exhibit a higher constitutive proteasomal activity than HeLa H21 cells that are less sensitive toward death receptor-induced cell death. Although the reasons for the differential proteasomal activities in these two cell lines are unknown, this finding is in agreement with our data, demonstrating an important role for the proteasome in the initiation of the death receptor program. In the search for the cellular target(s) that is responsible for the observed PI-mediated resistance toward death receptorinduced apoptosis, we analyzed the expression of several cell cycle and/or apoptosis regulatory proteins. As expected, MG-132 prevented the TNF-induced degradation of the proteasomal targets p21, p27, and cyclin A (41) but had no influence on unrelated proteins, such as RIP and FADD (data not shown). For our study, however, we analyzed the status of XIAP and c-FLIP in more detail because they represent two proteasomal targets that are well-known antiapoptotic proteins which are interfering with the initial stages of caspase activation (12,44,47,63). In addition to their degradation by the proteasome, both proteins are also specifically proteolysed during death receptor-induced apoptosis by caspase-3 and caspase-8, respectively (5,33). Interestingly, although TNF treatment resulted in comparable amounts of caspase-generated XIAP fragments in both cell lines, a significant reduction of fulllength XIAP protein was observed only in HeLa D98 cells (Fig.  5B, left panel). In contrast, the levels of uncleaved XIAP remained almost unchanged during apoptosis of HeLa H21 cells (Fig. 5B, right panel), a finding that is consistent with their lower proteasomal activity (Fig. 5A). Similar results were obtained when we analyzed the status of c-FLIP during apoptosis in these cells. In both cell lines, treatment with TNF induced a rapid proteasome-mediated degradation of c-FLIP-short that was efficiently prevented in the presence of MG-132 (Fig. 5B). In addition to being a proteasomal target, c-FLIP-long was also cleaved within 1 h in both cell lines by DISC-bound caspase-8 generating a p43 fragment. Also, this process was inhibited by MG-132 much more efficiently in HeLa H21 cells, as uncleaved c-FLIP-long remained detectable in these cells up to 8 h, whereas it was cleaved in HeLa D98 cells even in the presence of MG-132 following a 3-h TNF treatment (Fig. 5B). As it is known that c-FLIP-long is only cleaved by caspase-8 when both proteins heterodimerize at the DISC (33,48), this finding strongly indicates substantially reduced caspase-8 activation at the DISC in both cell lines treated with TNF in the presence of MG-132. Other antiapoptotic proteins, including IAP1, IAP2, and survivin, were not differentially expressed in the two HeLa cell lines, nor were they affected by the individual treatments (TNF in the absence or presence of various PIs) (Fig. 5C) as was observed for XIAP and c-FLIP (Fig. 5B). Together, these data strongly support our hypothesis that the proteasome is an important component required for the rapid initiation of death receptor-induced cell death via the degradation of antiapoptotic proteins, such as XIAP and c-FLIP.

Inhibition of the proteasome fails to protect XIAP knockdown cells from death receptor-induced apoptosis.
If the protective effect of PIs is mediated at least partially by the stabilization of XIAP and c-FLIP, then the suppression of either of these proteins should render TNF-, TRAIL-, or anti-CD95treated cells unresponsive toward PIs. To test this hypothesis, we first established HeLa D98 clones in which the expression of XIAP was suppressed due to the stable expression of an XIAP siRNA. Several individual clones were obtained, and four that displayed various degrees of XIAP suppression in the absence of any detectable alterations in surface expression of the three death receptors (Fig. 6A and data not shown) were chosen for further analyses. Cytotoxicity assays revealed that, indeed, the inhibition of the proteasome by MG-132 protected only parental HeLa D98 cells from death receptor-induced apoptosis and completely failed to protect the four XIAP siRNA clones from TNF-or TRAIL-induced cell death ( Fig.  6B and C). The protective effect of MG-132 was also dramatically reduced when the XIAP siRNA clones were treated with the anti-CD95 antibody, although this protection was never abrogated as efficiently as that in TNF-or in TRAIL-treated cells (Fig. 6D). This is consistent with our observation that the level of PI-mediated protection closely correlates with apoptosis susceptibility of the cells, as it is most pronounced in apoptosis-resistant HeLa H21 cells (Fig. 1) or even in anti-CD95treated HeLa D98 cells (Fig. 3), which represent the least efficient apoptosis inducer among the three death receptor ligands investigated.
Our finding that suppression of XIAP expression abrogated the protective effect of MG-132 was further verified by Western blot analyses demonstrating that MG-132 was not capable of preventing TNF-induced caspase processing in XIAP siRNA clones as efficiently as in parental HeLa D98 cells (Fig.  7). Whereas the decrease of both caspase-8 and caspase-3 proforms, as well as the generation of the active caspase fragments, was strongly inhibited by MG-132 in D98 cells, no such effect was observed in similarly treated XIAP siRNA clones. Hence, our results suggest that the observed PI-induced protection against death receptor-induced apoptosis is brought about at least partially by stabilizing XIAP protein levels.

Inhibition of the proteasome fails to protect c-FLIP knockdown cells from death receptor-induced apoptosis.
With regard to the role of c-FLIP in this process, we observed that MG-132 prevented TNF-induced cleavage of c-FLIP-long in both cell lines (Fig. 5B). As c-FLIP-long is predominantly cleaved by DISC-bound caspase-8, this result suggested that PIs prevent the recruitment of this initiator caspase into the DISC. This hypothesis would also be consistent with our findings that treatment with MG-132 not only led to a massive accumulation of c-FLIP-short (that actively prevents caspase-8 recruitment into the DISC) but also resulted in a substantial inhibition of death receptor ligand-induced caspase-8 processing and activation ( Fig. 1 and 3). To verify this scenario, we analyzed caspase-8 immunoprecipitates of HeLa D98 cells that were treated with or without MG-132 for up to 3 h with TNF for the presence of both forms of c-FLIP. Western blot analysis of these precipitates not only verified our previous findings that MG-132 efficiently prevents activation of caspase-8 but also confirmed a successful and efficient caspase-8 precipitation (Fig. 8A). More importantly, however, whereas the adaptor molecule FADD as well as the cleaved p43 fragment of c-FLIP-long was coprecipitated with caspase-8 regardless of whether the TNF treatment occurred in the presence of MG-132, c-FLIP-short was only found in these immunoprecipitates when MG-132 was present during TNF stimulation (Fig. 8B). As both c-FLIP isoforms bind to caspase-8 only within the DISC via interaction with FADD (23), these findings imply that the levels of both c-FLIP isoforms found in the caspase-8 precipitates are constituents of the TNF-R signaling complex FIG. 5. Higher proteasomal activity in apoptosis-sensitive HeLa D98 cells correlates with accelerated degradation of the antiapoptotic proteins XIAP and c-FLIP-short. (A) HeLa D98 and HeLa H21 cells either were left untreated or were treated with TNF (10 ng/ml) and Chx. After the indicated times, cell extracts were prepared and analyzed for proteasomal and ubiquitin C-terminal hydrolase activities using the LLVY-AMC and LRGG-AMC substrates, respectively. AU, arbitrary units. (B) Following preincubation for 1 h with or without MG-132 (10 M), HeLa D98 and H21 cells either were left untreated or were treated with TNF (10 ng/ml) and Chx. After the indicated times, cell extracts were prepared and analyzed by Western blotting for the status of XIAP and c-FLIP expression. The asterisks denote an unspecific band. (C) Following preincubation for 1 h with or without (none) the indicated protease inhibitors, HeLa D98 and H21 cells either were left untreated or were treated with TNF (10 ng/ml) and Chx. After the indicated times, cell extracts were prepared and analyzed by Western blotting for the status of the indicated proteins.
VOL. 26,2006 PROTEASOME INHIBITORS CONFER APOPTOSIS RESISTANCE 1973 II (39) and thus explain how MG-132 is able to inhibit caspase-8 activation. MG-132-mediated inhibition of caspase-8 recruitment (due to stabilization and accumulation of c-FLIP-short at the DISC) was also observed with regard to the CD95 DISC or when clasto-lactacystin was used to block TNF-induced cell death (data not shown).
To unambiguously determine the role of c-FLIP in the PImediated protection from death receptor-induced apoptosis, we finally generated HeLa D98 transfectants in which c-FLIP expression is suppressed due to the stable expression of a c-FLIP siRNA. As both c-FLIP forms are expressed only very weakly in HeLa D98 cells but accumulate significantly following proteasome inhibition, we analyzed the efficiency of the siRNA-mediated c-FLIP knockdown in the absence and presence of MG-132. As shown in Fig. 9A, the expression of both c-FLIP-long (upper panel) and c-FLIP-short (lower panel) was almost completely suppressed in all three clones examined, whereas a representative D98 clone that was transfected with the empty vector displayed c-FLIP levels comparable to those observed in wild-type D98 cells. All clones, including the vector cells, displayed no obvious alterations in death receptor surface expression (data not shown). Similar to the results obtained with the XIAP siRNA clones (Fig. 6), MG-132 protected only the parental HeLa D98 and vector control cells from death receptor-induced apoptosis but completely failed to do so with regard to the c-FLIP siRNA clones (Fig. 9B  through D). In fact, MG-132 treatment even further accelerated and enhanced TNF-and TRAIL-induced apoptosis of the three c-FLIP siRNA clones ( Fig. 9B and C). This intriguing finding was substantiated even further when we analyzed TNFinduced caspase processing in these clones. Whereas process- ing and thereby activation of caspase-8 and caspase-3 were significantly blocked when HeLa D98 cells were treated with TNF in the presence of MG-132, both caspases were processed much more efficiently and with an accelerated kinetic in similarly treated c-FLIP siRNA clones (Fig. 10). Similar results were obtained when the c-FLIP siRNA clones were treated with TRAIL (data not shown). Interestingly, compared to parental HeLa D98 cells, TNF-induced caspase processing was slightly delayed in the three c-FLIP siRNA clones, which might be explained by a recently proposed hypothesis that c-FLIPlong is required for caspase-8 activation at the DISC (3). Collectively, our data clearly demonstrate that proteasomal inhibitors transiently protect cells from death receptor-induced apoptosis and that they exert this protective effect via stabilization of XIAP and c-FLIP.

DISCUSSION
Two complicated proteolytic systems are involved in the control of cell death: the caspase family of cysteine proteases and the ubiquitin-proteasome degradation system. Whereas the function of caspases is well defined, the exact role of the ubiquitin-proteasome system in apoptosis is far from being elucidated. Several studies showed that the inhibition of the proteasome blocked apoptosis in various settings (18,20,46), whereas others demonstrated efficient induction of apoptosis when cells were treated with PIs either alone or in combination with various death-inducing stimuli (8,11,19,30,60). As the antiapoptotic activities of PIs were described for resting thymocytes and differentiated neuronal cells and the proapoptotic effects were described for various tumor cell lines, it was postulated that the requirement of proteasomal activity for the progression or inhibition of apoptosis critically depends on the proliferative status of the cells (9,41). In an alternative model, it was suggested that the PI concentrations used determine whether such peptide aldehyde inhibitors function in a pro-or antiapoptotic manner (34). Here, however, we describe an antiapoptotic role of PIs in several proliferating tumor cells that was observed not only at low but also at high PI concentrations. In addition, although at first glance our study appears to be controversial to many other reports, it provides strong evidence that both the pro-and antiapoptotic functions of the proteasome have to be taken into account for efficient PIbased tumor therapy. Using death receptor-induced apoptosis as a model system, we demonstrated that specifically proteasomal inhibitors, such as clasto-lactacystin, MG-132, or ALLN, but not the calpain and cathepsin inhibitor ALLM efficiently protected cells against TNF-, TRAIL-or CD95-induced apoptosis. This protection correlated well not only with the inhibition of the executioner caspase-3 but also with a marked delay in the processing and activation of caspase-8, the most apical initiator caspase in death receptor signaling. Hence, our results suggest that the PI-mediated protection is caused by inhibition of an early initiation event that most likely involves the stabilization of pre-existing antiapoptotic proteins, such as c-FLIP and XIAP, which are both known to efficiently counteract the initial stages of caspase activation (29,47). Indeed, both proteins were stabilized by PIs in the two cell lines used. In addition, c-FLIPshort was found to be associated with the DISC only in the presence of PIs, clearly explaining the lack and delay of caspase-8 processing under these conditions. Finally, siRNAmediated knockdown of either c-FLIP or XIAP expression completely abrogated the protective function of PIs and, in some instances, even reversed this effect, resulting in an increased apoptosis rate. Thus, from these results, it is obvious that PIs exert their antiapoptotic effect via the stabilization of XIAP and c-FLIP. This result is especially important with regard to the fact that many human tumors express high levels of these antiapoptotic proteins, rendering them rather resistant toward apoptosis induction by various agents (21,37,38,47,55). Based on our results, it is, however, also unlikely that a combined treatment of death receptor ligands (or perhaps drugs; see below) with PIs would result in a more beneficial outcome, as such a treatment might even cause antagonistic effects due to the PI-mediated prolonged survival of these tumors. For an efficient therapy, it is therefore important to thoroughly analyze every individual tumor for the expression levels of XIAP and c-FLIP as well as for the status of other antiapoptotic proteins, such as Bcl-2, that are known to be targets of the proteasome (7,12,44,63). In addition, it is inevitable to also carefully evaluate the biochemical pathways involved, as PIs were recently shown to also inhibit apoptosis induced by retinoic acid, geldanamycin, and staurosporine and even the combined treatment of melanoma cells with TNF and PIs did not show the expected synergism of action (2,43,52,62). The underlying mechanisms causing the resistance to these drugs remained, however, unknown.
Of note, however, is also our observation that although PI treatment significantly delayed caspase activation and cell death of various tumor cell lines, it did not result in a permanent protection from death receptor-induced apoptosis. In fact, when the initial PI-mediated roadblock of caspase processing was eventually overcome, active caspases accumulated to an even larger extent in cells treated with the individual death receptor ligands in the presence rather than in the absence of PIs. Such a scenario was found not only in cells that have been exposed to this treatment for longer terms but also in cells with suppressed c-FLIP or XIAP expressions. These observations clearly demonstrate a biphasic role for the proteasome in death receptor-induced apoptosis in which at first its activity is absolutely essential for the initiation phase by eliminating various roadblocks, including c-FLIP and XIAP. During later stages, however, the proteasome acts in an antiapoptotic manner, as it also promotes the degradation of proapoptotic proteins such as caspases. The proposed biphasic role of the proteasome would also imply that the outcome of a PI-based tumor therapy depends not only on the expression levels of antiapoptotic proteins but also on the time point at which the PIs are applied. With regard to this, we found that the protection was most pronounced when the cells were exposed to MG-132 1 h prior to the death stimulus, whereas the protection effect was abolished when the inhibitor was added 1 h or 2 h after TNF or TRAIL addition. A similar finding was also reported recently with regard to DNA damage-induced apoptosis (54). In that study, the authors showed that preincubation with MG-132 increased the resistance of human lung carcinoma cells toward etoposide, whereas they observed an increased apoptosis rate when the cells were first exposed to this drug for 30 min, followed by the addition of MG-132. Besides the expression levels of antiapoptotic proteins and the time point at which the PIs might be applied, our results also provide strong evidence that the level of the intrinsic proteasomal activity represents a crucial factor determining the outcome of a PI-based therapy. In fact, HeLa H21 cells that are less susceptible to death receptor-induced apoptosis displayed a reproducibly lower proteasomal activity than did TNF-sensitive HeLa D98 cells. Together with our observation that TNF did not induce degradation of XIAP as efficiently in HeLa H21 as in HeLa D98 cells, these results are in accordance with our hypothesis that the proteasome is required for an efficient initiation of the death program. In further support of our assumption are also recent findings demonstrating, in vitro and in vivo, that tumors acquired apoptosis resistance by down regulating their proteasomal activity (14,22). As the proteolytic activity is essential for every proliferating cell, these tumors adapted to the low proteasomal activity by switching to another newly identified proteolytic system, the tripeptidyl peptidase II (15). Due to the hereby altered specificity of cytosolic proteolysis, however, these tumors were unable to efficiently degrade the various IAPs, including XIAP, resulting in rapidly growing tumors in vivo (22).
In summary, our data clearly support a model in which the proteasome plays an important biphasic role in apoptosis. At first, it is essential for the initiation of the cell death program by degrading antiapoptotic proteins but it acts at later stages in an antiapoptotic manner as it also proteolyses proapoptotic proteins. Hence, our data should be taken into serious account when a PI-based tumor treatment is considered in combination with various apoptosis-inducing drugs or death receptor ligands. However, they also encourage attempts to combine such treatments with small molecule inhibitors targeting expression and/or activity of antiapoptotic proteins (13).