Oppositional Regulation of Noxa by JNK1 and JNK2 during Apoptosis Induced by Proteasomal Inhibitors

Proteasome inhibitors (PIs) potently induce apoptosis in a variety of tumor cells, but the underlying mechanisms are not fully elucidated. Comparing PI-induced apoptosis susceptibilities of various mouse embryonic fibroblast (MEF) lines differing in their c-jun N-terminal kinase (JNK) 1 and 2 status, we show that several hallmarks of apoptosis were most rapidly detectable in JNK2−/− cells, whereas they appeared only delayed and severely reduced in their intensities in cells expressing JNK2. Consistent with our finding that PI-induced apoptosis requires de novo protein synthesis, the proteasomal inhibitor MG-132 induced expression of the BH3-only protein Noxa at the transcriptional level in a JNK1-dependent, but JNK2-opposing manner. As the knockdown of Noxa blocked only the rapid PI-induced apoptosis of JNK2−/− cells, but not the delayed death occurring in JNK1−/− and JNK1+/+ cells, our data uncover a novel PI-induced apoptosis pathway that is regulated by the JNK1/2-dependent expression of Noxa. Furthermore, several transcription factors known to modulate Noxa expression including ATF3, ATF4, c-Jun, c-Myc, HIF1α, and p53 were found upregulated following MG-132 exposure. From those, only knockdown of c-Myc rescued JNK2−/− cells from PI-induced apoptosis, however, without affecting expression of Noxa. Together, our data not only show that a rapid execution of PI-induced apoptosis requires JNK1 for upregulation of Noxa via an as yet unknown transcription factor, but also that JNK2 controls this event in an oppositional manner.


Introduction
In a plethora of in vitro studies it has been extensively demonstrated that inhibition of the proteasome for instance by the tripeptide aldehyd MG-132 or the dipeptide boronate bortezomib (Velcade TM ) selectively kills tumor cells of varying origin (reviewed in Ref. [1]). Proteasomal inhibitors (PIs) also sensitize cells to radio-and chemotherapy and even to apoptosis induced by death receptor ligands [2,3]. However, as the proteasome targets not only pro-, but also anti-apoptotic proteins, a successful combination therapy requires a successive application of first the apoptosis-inducing agent ensuring the breakdown of anti-apoptotic proteins followed by the PI treatment that then prevents degradation of the generated pro-apoptotic proteins [4]. Nevertheless, bortezomib was the first PI used in clinical trials and approved to treat patients suffering from multiple myeloma or mantle cell lymphoma [5].
Although the new generation of proteasome inhibitors such as salinosporamide and carfilzomib appear to exhibit somewhat different mechanisms of action than bortezomib, central to apoptosis induction by many PIs is certainly the mitochondrial or intrinsic death pathway, as their cytotoxic activity is almost completely abrogated in cells deficient for Bax and Bak [6].
Consistently, a number of studies strongly implicated certain proapoptotic BH3-only proteins in PI-induced apoptosis [7]. For instance, the pro-apoptotic cleavage product of Bid, t-Bid, is degraded by the proteasome and treatment of HeLa cells with MG-132 resulted in accumulation of t-Bid and sensitized the cells to death receptor-induced apoptosis [8]. Also Bik and Bim were found to be upregulated following PI treatment and cells deficient for both or cells in which Bik and Bim were down regulated by RNA interference were refractory to its cytotoxic action [9,10]. Likewise, different PIs including bortezomib and MG-132 were shown to induce expression of Noxa in several tumor models both at the protein and mRNA level and siRNA-mediated knockdown of Noxa partially rescued various tumor cells from PI-induced apoptosis [11][12][13]. Expression of other Bcl-2 family members such as Puma, Bax, Bak, Bcl-2, and Bcl-X L remained mostly unaffected following treatment of different cell lines with PIs.
Several signaling pathways have been shown to play a role in PI-induced cytotoxicity including stabilization of the tumor suppressor protein p53, inhibition of the nuclear factor-kB (NF-kB), or induction of an ER-stress response [14][15][16]. As Noxa was first identified as a p53 target gene [17], the stabilization and activation of p53 would have been an attractive possibility for apoptosis induction by PIs. However, PI-mediated tumor cell killing was also observed in p53-deficient cells and independently of NF-kB inhibition suggesting that other signaling pathways targeted by the proteasome are even more crucial for cell death induction by PIs [15,18]. One of those might be instigated by members of the mitogen-activated protein (MAP) kinase family, the c-Jun N-terminal kinases (JNKs) that were reproducibly found to be activated in PI-treated cells [19,20]. More intriguingly, inhibition of JNK activity by either dominant-negative JNKs or by RNA interference rendered the cells resistant toward cell death induction by PIs [20,21]. Thus, it appears that JNKs, in addition to several other pathways in which they were shown to contribute to apoptosis signaling [22], are also crucial players in PI-induced apoptosis.
Three JNK isoforms (JNK1, 2 and 3) with different splice variants are expressed either ubiquitously (JNK1 and JNK2) or preferentially in neuronal and heart tissues (JNK3) [23]. They were originally identified by their ability to specifically phosphorylate and activate c-Jun, a constituent of the activator protein-1 (AP-1) transcription factor that is involved in the increased expression of several pro-apoptotic genes such as TNF-a, Fasligand, Bak and Bim [22]. Although silencing of the c-Jun/AP-1 pathway conferred resistance to JNK-mediated apoptosis in several cellular systems, the observed stimulus-and cell typedependent manner of protection suggests participation of other downstream effectors [24]. Indeed, JNKs appear to control apoptosis in quite a versatile manner as they not only phosphorylate and activate other pro-apoptotic transcriptions factors including p53 and c-Myc, but also several Bcl-2 family proteins causing inhibition of pro-survival members such as Bcl-2, Bcl-X L and Mcl-1 and activation of pro-apoptotic members such as Bim and Bad [22]. However, although these phosphorylation events are consistent with the observation that JNKs are required for stress-induced activation of the mitochondrial death pathway [25], their contributions to apoptosis are controversially discussed [23]. In addition, it is unknown whether JNK1 and JNK2 exhibit redundant or specific functions in PI-induced apoptosis and whether they are involved in the induction of Noxa.
To elucidate these questions in more detail, we compared PIinduced apoptosis signaling of immortalized mouse embryonic fibroblasts (MEF) that differ in their JNK1 and/or JNK2 status. In addition to our findings that JNK1 greatly accelerates de novo expression of Noxa and subsequent apoptosis, we also observed that both processes were strongly impaired in the presence of JNK2 implying oppositional roles for these isoforms in PI-induced apoptosis. Intriguingly, although either knockdown of the transcription factor c-myc or Noxa protected JNK22/2 cells from PIinduced apoptosis, they were found to function independently of each other.

Results
Proteasomal inhibitors require JNK1 and de novo protein synthesis to rapidly induce apoptosis We first compared JNK2-deficient MEFs with MEFs that are either heterozygously or homozygously deleted for JNK1 (Fig. 1E) [26] with regard to their sensitivities toward cell death induction by the proteasomal inhibitor MG-132. Whereas JNK2-deficient MEFs were efficiently and rapidly killed by MG-132 in a time-and dose-dependent manner, JNK1+/2 and JNK12/2 cells showed a significantly delayed and only moderate sensitivity toward this treatment (Figs. 1A, S1A; data not shown). Cell death was almost completely blocked in the presence of the pan-caspase inhibitory peptide Q-VD-OPh, indicating that MG-132 induces apoptosis in these cells (Fig. 1A). Further FACS analyses confirmed this mode of cell death, as MG-132-treated JNK2-deficient cells displayed extensive DNA fragmentation as evidenced by the time-dependent increase of the apoptotic sub-G1 peak and its susceptibility to caspase inhibition (Fig. 1C). Accordingly, processing and activity of caspase-3 peaked in JNK2-deficient cells after an 8 hour MG-132 treatment, whereas these events were completely undetectable or became only barely visible at this time point in similar treated JNK12/2 and JNK1+/2 MEFs, respectively (Figs. 1B, 1F, S1B). Although caspase-3 activities significantly increased when the latter two cell lines were exposed to MG-132 for 16 hours (Figs. 1B, 1F), our data suggest that JNK1 is required for a rapid apoptosis induction by MG-132. Furthermore, MG-132-induced cell death and caspase-3 activation were completely abrogated in the presence of actinomycin D (ActD) (Fig. 1D) or cycloheximide (Chx) (Fig. 1), suggesting the requirement for de novo protein synthesis in these JNK1-dependent processes.
For specificity reasons we also analyzed cell death induction by other proteasomal inhibitors such as ALLN and ALLM that are effective and weak proteasomal inhibitors, respectively, or clastolactacystin b-lactone (CLC) and bortezomib that are among the most selective PIs known. Except for the weak inhibitor ALLM, all the compounds tested reproducibly induced apoptosis in a dosedependent manner almost exclusively in JNK22/2 cells, whereas these treatments triggered a significantly weaker response in JNK1+/2 and JNK12/2 cells. (Fig. S1C). These results strongly indicate that the observed apoptosis induction by these compounds is indeed caused by their ability to block the 26S proteasome. In addition, and similar to our results obtained with MG-132, apoptosis induction by these PIs was completely abrogated in the presence of Chx (Fig. S2). Thus, consistent with the fact that JNKs phosphorylate and activate numerous transcription factors involved in apoptosis signaling, our results demonstrate the absolute necessity of de novo protein synthesis in the JNK1-dependent apoptosis pathway induced by PIs.

MG-132 induces a JNK1-dependent upregulation of Noxa at the transcriptional level
Central to apoptosis induction by PIs and also JNKs is certainly the mitochondrial or intrinsic death pathway, as both are known to modulate expression and activity of several components involved in this pathway including transcription factors and members of the Bcl-2 protein family [6,22]. Therefore, we analysed the mitochondrial membrane potential (DY m ) of the individual MEF lines treated with MG-132. In agreement with our apoptosis data (Figs. 1, S1, S2), we found indeed that MG-132 induced a transcription/translation-dependent loss of DY m that was much earlier evident in JNK2-deficient cells than in JNK12/ 2 cells ( Fig. 2A).
We also assessed the influence of MG-132 on the expression of various members of the Bcl-2 protein family in these MEF lines. Whereas steady state levels of the pro-apoptotic multi domain proteins Bax and Bak as well as the anti-apoptotic Bcl-2 and Bcl-xL proteins were not altered by MG-132 in any of the MEF lines, Mcl-1 levels expectedly increased upon this treatment in accordance with the susceptibility of this protein to proteasomal degradation (Fig. 2B). Although this increase was prevented in the presence of Chx, it occurred in all cell lines to a comparable extent and thus in a JNK-independent manner. More interestingly, significant and JNK-dependent changes were observed with regard to expression of the BH3-only proteins Bim and Noxa (Fig. 2B). Both pro-apoptotic proteins were nearly undetectable in untreated cells, but were readily induced by MG-132 particularly in JNK22/2 cells that were most sensitive to PIs. In these cells, Bim and Noxa induction following MG-132 exposure was not only more pronounced, but proceeded also with much faster kinetics than in similar treated JNK1+/2 and JNK12/2 cells. In addition, MG-132-induced expression of both proteins was almost completely abolished in the presence of cycloheximide, an observation consistent with the thereby caused apoptosis resistance of JNK22/2 cells. Despite this, however, it appears that only Noxa fulfills the criteria of a PI-induced JNK1-dependent apoptosis mediator in an unconfined manner, as this BH3-only protein was most abundantly induced by MG-132 in JNK22/2 cells, but completely undetectable or only weakly expressed in similar treated JNK12/2 and JNK1+/2 cells, respectively. Bim on the other hand was induced by MG-132 in a JNK1independent manner, albeit most pronounced in JNK22/2 cells, and was, in contrast to Noxa, still detectable following cycloheximide treatment. Consistent with these findings and the observed dependency of PI-induced apoptosis on de novo protein synthesis, we found that MG-132 exerts a profound effect only on the transcriptional induction of Noxa mRNA, but not on Bim mRNA (Fig. 2C). Similar to the JNK1-dependent increase in Noxa protein (Fig. 2B), MG-132 induced a dramatic (approximately 40-fold) upregulation of Noxa mRNA in JNK22/2 cells (Fig. 2C). In JNK1+/2 cells in contrast, MG-132 increased Noxa mRNA expression only by approximately 8-fold and completely failed to do so in the complete absence of JNK1 in JNK12/2 cells. In contrast, expression of Bim mRNA was only slightly affected by MG-132 in either cell line, suggesting that Noxa, but not Bim, is responsible for the observed PI-induced and JNK1-dependent execution of apoptosis.

JNK1-dependent upregulation of Noxa is required for PIinduced apoptosis
To test the hypothesis that the JNK1-dependent transcriptional upregulation of Noxa represents a key event in PI-induced apoptosis, we analyzed the consequences of a siRNA-mediated knockdown of Noxa or Bim. Although both siRNAs caused a comparable reduction in their respective protein targets (Fig. 3A), the knockdown of Noxa was reproducibly more effective in preventing MG-132-induced caspase-3 activation and apoptosis than depletion of Bim (Figs. 3B, 3C). Whereas the Noxa siRNA reduced these events by approximately 64% and 50%, only 41% and 30% inhibition of caspase-3 activation and apoptosis were observed following knockdown of Bim, respectively. Thus, while Bim surely participates in PI-induced apoptosis signaling, JNK1induced Noxa appears to be the more crucial event. In support of this idea, we found that knockdown of residual Noxa had even after 16 h no effect on MG-132-induced caspase-3 activation and apoptosis in JNK12/2 cells (Fig. 3D-F), which is consistent with the delayed JNK1-independent death pathway induced by PIs in these cells (Fig. 1).

JNK1 and JNK2 oppositional regulate PI-induced Noxa expression and apoptosis
Although our data so far indicate that PI-induced apoptosis requires JNK1-dependent Noxa upregulation, they might additionally be interpreted that JNK2 functions in an inhibitory manner. In an attempt to clarify this issue, we examined PIinduced Noxa induction and apoptosis also in JNK1+/+ cells. Remarkably, these cells were clearly less sensitive than similar treated JNK22/2 cells (Fig. 4A). Similar to JNK1+/2 cells, this moderate MG-132 sensitivity correlated well not only with a relative weak activation of caspase-3 (data not shown), but also with an intermediate induction of Noxa protein and mRNA when compared to JNK12/2 and JNK22/2 cells (Figs. 4B, 4C). Consistent with our previous data (Fig. 2C), Noxa mRNA induction was most pronounced in JNK22/2 cells, but only barely detectable in JNK12/2 cells (Fig. 4C). In agreement with these findings, knockdown of Noxa had only a marginal effect on MG-132-induced caspase-3 activation and apoptosis in JNK1+/+ cells (Figs. 4D-F).
To unambigously clarify whether the two JNK isoforms exert oppositional roles in PI-induced apoptosis, we finally exposed cells completely lacking both JNK1 and JNK2 (JNK-DKO) to MG-132. Although JNK-DKO cells exhibited together with JNK1+/2 cells an intermediate death response when compared to JNK12/ 2 and JNK22/2 cells, their PI-induced Noxa protein levels were almost undetectable and thus similar to those observed in JNK12/2 cells (Fig. 5). As Noxa levels increased slightly and even more dramatically in JNK1+/2 and JNK22/2 cells, respectively, these data not only support our conclusion that JNK1 is required for PI-induced Noxa expression and apoptosis, but, in addition, provide strong evidence that JNK2 counteracts these responses.

Knockdown of c-Myc impairs PI-induced apoptosis of JNK22/2 cells without affecting expression of Noxa
Having demonstrated that the JNK1-dependent transcriptional upregulation of Noxa is crucial for PI-induced apoptosis of JNK22/2 MEFs, we aimed to identify the transcription factor responsible for this event. Searching the Genomatix software suite (www.genomatix.de), 85 putative candidates were found to associate with the Noxa gene PMAIP1. From these, we closely examined the role of several transcription factors including ATF3, ATF4, c-Jun, c-Myc, HIF1a, p53, and the glucocorticoid receptor (GR), as most of them were not only shown to transcriptionally regulate Noxa expression [17,[27][28][29][30][31], and known to be involved in JNK signaling [32][33][34][35][36][37], but also because they represent prominent targets of the proteasome [38][39][40][41][42][43][44]. In addition, several of these transcription factors were found upregulated in various screens for modulators of cell death induced by PIs [45][46][47]. In agreement with these data we observed that MG-132 induced in all three MEF lines an upregulation of ATF3, ATF4, c-Jun, c-Myc and HIF1a that was abrogated in the presence of cycloheximide (  (Fig. 7A). Together, these results provide strong evidence that none of the herein examined transcription factors including c-Myc is required for the upregulation of Noxa following PI exposure.

Discussion
Inhibition of the proteasome either on its own or in combination with other apoptotic stimuli is a powerful means to specifically eradicate tumor cells, but the underlying molecular pathways are only incompletely deciphered [1]. JNKs and the BH3-only protein Noxa were reproducibly demonstrated in many diverse systems to be essential constituents of this process as inhibition of their activity and/or expression substantially protected cells from PI-induced apoptosis [11][12][13]20,21]. However, these two pathways have never been connected before and the contributions (if any) of individual JNK isoforms to PI-mediated induction of Noxa and apoptosis were completely unknown.
Using various JNK1-and/or JNK2-deficient fibroblasts, we demonstrate here for the first time that JNK1 is required for PIinduced Noxa expression and subsequent rapid induction of apoptosis. This is evidenced by our findings that Noxa mRNA and protein levels were almost undetectable in PI-treated JNK12/2 cells, but were induced in similar treated cells expressing at least one allele of JNK1. Intriguingly, as the JNK1-dependent Noxa expression was most dramatically induced in the absence of JNK2 (, 40fold in JNK22/2 cells compared to , 6-8fold in either JNK1+/2 or JNK1+/+ cells), these findings additionally imply a negative regulatory role for JNK2 in this event. Finally, both JNK1-deficient MEF lines (JNK-DKO and JNK12/2) failed to upregulate Noxa following PI exposure, further suggesting that withdrawing the inhibitory effect of JNK2 is by itself not sufficient to launch expression of Noxa in the concurrent absence of JNK1. Consistently, JNK22/2 cells succumbed rapidly to apoptosis following PI treatment, whereas several hallmarks of apoptosis appeared only delayed and severely reduced in their intensities in similar treated cells expressing JNK2. Based on these results and our finding that knockdown of Noxa severely compromised PIinduced apoptosis of JNK22/2 cells, we postulate that the opposing functions of JNK1 and JNK2 in the regulation of Noxa expression represent crucial events in this cell death pathway. Although so far, JNK1 and JNK2 have mostly been considered to exert overlapping or even redundant functions, a few studies have recently described opposing effects for these kinases that are in line with our observations. Particularly with regard to their involvement in numerous cell death systems, JNK1 and JNK2 were shown to differentially regulate expression and/or function of their targets p53, c-jun and Elk-1 resulting in an oppositional modulation of stress-and basal (non-stress)-induced apoptosis and RNA polymerase III-dependent transcription [48][49][50]. The mechanisms involved, however, were not well characterized leaving the question open whether JNK1 and JNK2 mediate these oppositional effects directly by phosphorylating diverse sites in these targets, or indirectly by modulating additional components that function in an inhibitory or stimulatory manner. In any case, individual JNKs may harbour intrinsic target site specificities or their activities may be regulated differentially by other factors. With regard to the latter scenario we have recently shown that the cyclin-dependent kinase (CDK) inhibitor p21 can differentially modulate the activities of certain kinases including those of several JNK1/2 isoforms in a remarkable substrate-dependent manner  [51]. Whether this or other mechanisms are involved in the oppositional regulation of Noxa expression by JNK1 and JNK2 remains to be elucidated.
Remarkably, despite the fact that all three JNK isoforms are able to promote or induce apoptosis, similar to the findings in our study, it was particularly JNK1 that has been implicated in different apoptosis pathways including those instigated by TNFa, UV irradiation and nitric oxide [25,26,52]. Employing JNKdeficient cells it was for instance shown that JNK1 promotes TNFa killing by phosphorylation and activation of the ubiquitin ligase Itch that mediates the proteasome-dependent degradation of the caspase-8 inhibitor FLIP [53]. During UV-and nitric oxideinduced apoptosis on the other hand, it was demonstrated that the JNK1-dependent phosphorylation of the anti-apoptotic myeloid cell leukemia-1 (Mcl-1) protein results in its proteasomal degradation [54,55]. Interestingly, as Mcl-1 is the major counterpart of Noxa and its loss is a critical event that leads to activation of Bax and Bak [56,57], its JNK1-dependent elimination shifts the delicate Mcl-1-Noxa balance toward apoptosis induction as observed in many systems including those instigated by UV irradiation [58,59]. Although for obvious reasons proteasomal degradation of Mcl-1 does not contribute to PI-induced apoptosis, the herein described massive upregulation of Noxa (40-fold) is surely able to efficiently bypass this shortcoming by directly counteracting anti-apoptotic Bcl-2 proteins including Mcl-1. This view is supported by our finding that MG-132 induced similar Mcl-1 levels independently of JNK1/2, further emphasizing that the JNK1/2-dependent regulation of Noxa represents the most crucial event in the herein uncovered PI-induced apoptosis pathway. Furthermore, once activated, the mitochondrial death cascade then causes elimination of Mcl-1, as this anti-apoptotic Bcl-2 protein was shown to become proteolytically inactivated during PI-induced apoptosis in a caspase-3-dependent manner [60,61]. Intriguingly, JNK12/2 mice are highly susceptible to DMBA/PMA-induced skin tumor formation when compared to similar treated wild type mice [62]. As both JNKs and the human Noxa orthologue PMAIP1 (PMA-inducible protein 1) can be strongly activated by phorbol esters, it is tempting to speculate that this tumor suppressive function of JNK1 depends on the induction of Noxa. As, however, JNKs including JNK1 are able to exert also oncogenic functions, that were particularly evident in the development of human hepatocellular carcinoma [63], it is highly likely that JNK1 mediates its versatile functions strictly in a cell type-dependent manner.
Besides being reported as a PMA-inducible protein, Noxa was originally identified as a p53-induced stress response gene, but is now known to be regulated by an array of different transcription factors independently of p53 [17,64]. However, although several transcription factors known to participate in JNK signaling and diverse apoptosis pathways, and that were even shown to constitute prominent targets of the proteasome, none of those examined here including c-Jun, c-Myc, Hif1a, ATF3, ATF4 and GR appear to be involved in the herein observed JNK1/2dependent opposite regulation of Noxa. A participation of p53 cannot be completely ruled out, as a few MEF lines studied here may harbor p53 mutations. Perhaps most unexpected was our finding that even the knockdown of c-Myc, a transcription factor that was recently demonstrated to transcriptionally upregulate Noxa during bortezomib-induced apoptosis [29], did not affect Noxa expression despite the observed protection of JNK22/2 cells from PI-induced apoptosis. Particularly with regard to the fact that JNKs phosphorylate and thereby regulate the apoptotic function of c-Myc [34], its previous identification as a potent Noxa inducer upon PI treatment provided strong evidence for the existence of a JNK-Myc-Noxa axis, at least in melanoma cells [29]. As the thereby identified c-Myc binding site in the Noxa promoter is conserved among human, mouse and rat, it is presently unknown why c-Myc induces expression of Noxa only in human melanoma cells, but not in the MEF lines studied here.
Also cells lacking JNK1 eventually succumb to apoptosis in almost the complete absence of Noxa, albeit in a delayed manner. Together with the observation that knockdown of Noxa had no effect on the delayed JNK1/2-independent cell death pathway occurring most likely in all here examined MEF lines following their exposure to PIs, our results strongly support the existence of alternative PI-induced pathways that kill cells independently of JNKs and Noxa. The BH3-only protein Bim might be part of such a pathway, as it was found by us and others upregulated in response to PIs in a JNK-independent manner, and its knockdown partially protected JNK22/2 cells from PI-induced apoptosis [9]. Furthermore, the PI-mediated upregulation of Bim was, unlike the induction of Noxa, not entirely blocked in the presence of cycloheximide. This suggests that the increase in Bim protein expression is probably a direct effect of proteasome inhibition that prevents degradation of this pro-apoptotic BH3-only molecule [9]. Thus, Bim most likely represents an alternative route to cell death in cases in which PIs are unable to mediate the JNK1-dependent upregulation of Noxa. In summary, we have shown here that a rapid PI-induced apoptosis pathway critically depends on the induction of Noxa that is controlled by JNK1 and JNK2 in an opposing manner. Although we were unable so far to identiy the transcription factor(s) involved, our results might help to further improve future anticancer strategies that are based on proteasomal inhibitors. Thereby, one should keep in mind that our observations are solely based on the use of immortalized MEFs. To exclude possible phenotypical changes acquired during their immortalization, it will be necessary to confirm these findigs using primary MEFs or lymphocytes from JNK1/2 knockout mice.

Treatment of cells and measurement of cell death
Cells were usually treated with 10 mM MG-132 in the absence or presence of cycloheximide (Chx; 10 mg/ml), or the pan caspase inhibitor Q-VD-OPh (10 mM). Treatment with other proteasomal inhibitors is specified. Cell death was analysed cytometrically either by the uptake of propidium iodide (2 mg/ml in phosphatebuffered saline) to determine the percentage of cells with a loss of membrane integrity, or by quantifying the proportion of nuclei

Preparation of cell extracts and Western blotting
Total cell extracts were prepared in lysis buffer containing 1% NP-40, 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM DTT, and protease and phosphatase inhibitors. Protein concentrations were determined with the BioRad protein assay. Subsequently, proteins were separated on SDS-polyacrylamide gels and electroblotted to polyvinylidene difluoride membranes (Millipore, Schwalbach, Germany). Following antibody incubation, the proteins were visualized by enhanced chemiluminescent staining using ECL reagents (Amersham Biosciences). Fluorometric determination of caspase activity Caspase-3 activities were assessed by using 50 mg of the cell extracts that were directly dissolved in 200 ml substrate buffer (50 mM HEPES pH 7.3, 100 mM NaCl, 10% sucrose, 0.1% CHAPS, 10 mM DTT) supplemented with 50 mM DEVD-AMC (Biomol). The reaction was incubated at 37uC for 3 h and the release of fluorogenic AMC was measured at an excitation wavelength of 346 nm and an emission wavelength of 442 nm using an measured in a Infinite M200 microplate reader (Tecan, Langenfeld, Germany). The detected fluorometric signal that directly correlates to the caspase activity in the cell extracts is expressed in arbitrary units (AU).
Determination of Noxa and Bim mRNA expression by semi-quantitative PCR analysis Cells were exposed to MG-132 for 4 hours and total mRNA was isolated with the RNeasy mini kit (Qiagen, Hilden, Germany) according to the instructions of the manufacturer. 2 mg of the isolated RNA was transcribed into cDNA using random hexamers (Applied Biosystems, Darmstadt, Germany) as primers and MMLV reverse transcriptase (Invitrogen, Karlsruhe, Germany). 100 ng of the resulting cDNA was subjected to semi-quantitative PCR for 25 cycles using Taq Polymerase (5 Prime, Hamburg, Germany) and specific exon-spanning primers for the mRNAs of Noxa and Bim as well as for GAPDH as an endogenous control. One half of each PCR reaction was separated on an ethidium bromide-stained agarose gel and the corresponding bands were densitometrically analyzed using a CCD camera (LAS-4000) together with the Multi Gauge software, both from Fujifilm Life Science (Düsseldorf, Germany). The signal of each band was normalized to the corresponding GAPDH band.

Transfection of siRNAs
ON-TARGETplus SMARTpool and non-targeting control siRNAs were purchased from Dharmacon RNA technologies (Lafayette, CO, USA), and the knockdown was carried out according to the manufacturer's instructions. Fourty-eight to seventy-two hours after transfection, cells were divided equally to receive either no treatment or exposure to proteasomal inhibitors. Cells were harvested at the indicated time points and directly analysed by Western blotting, by the fluorometric caspase substrate assay, or cytometrically for a successful knockdown of the target protein, DEVDase activity and cell death, respectively.

Statistical Analyses
Data, expressed as means 6 SD, were analyzed statistically using SigmaPlot 12.3. Shapiro-Wilk test was performed for normality testing. Equal variance was checked by Spearman rank correlation. One-way analysis of variance (ANOVA) as a parametric test or Kruskal-Wallis one-way analysis of variance on ranks as a non-parametric test for comparison of differences between measurements was used. p values of less than 0.05 were considered statistically significant. For post-hoc comparison Dunns, Tukey or Student-Newman-Keuls test were employed. Figure S1 (A and B) Rapid induction of apoptosis by MG-132 only in JNK22/2 cells. Flow cytometric (propidium iodide uptake) (A) and fluorometric (caspase-3-like DEVDase activity) (B) cell death determination of JNK1+/2, JNK12/2 and JNK22/ 2 MEFs that were exposed to MG-132 for the indicated times.

Supporting Information
Values represent the mean of three independent experiments +/2 SD. (C) Proteasomal inhibitors preferentially induce cell death in JNK22/2 MEFs. Flow cytometric determination of cell death (propidium iodide uptake) in JNK1+/2, JNK12/2 and JNK22/2 cells that were either left untreated or exposed for 24 hours to the indicated compounds. Values are the mean of three independent experiments +/2 SD. For all panels, * p,0.05; ** p,0.01; *** p,0.005 according to ANOVA or ANOVA on ranks Tukeys test in a time-(A and B) and dose-matched (C) comparison of JNK22/2 vs. JNK1+/2 and JNK12/2, respectively. Please note that panels A to C are alternative presentations of some of the data of Fig. 1A, 1B and Suppl. Fig. 2, respectively. (TIF) Figure S2 Various proteasomal inhibitors including bortezomib, ALLN and clasto-lactocystine b-lactone (CLC) rapidly induce cell death preferentially in JNK22/2 MEFs. Flow cytometric determination of cell death (propidium iodide uptake) in JNK1+/2, JNK12/2 and JNK22/2 cells that were either left untreated or exposed for 24 hours to the indicated compounds in the absence and presence of cycloheximide. Values are the mean of three independent experiments +/2 SD. For all panels, * p,0.05; ** p,0.01; *** p,0.005 according to ANOVA Tukeys test in a dose-matched comparison vs. Chx treatment. (TIF) Figure S3 Knockdown of ATF3 and ATF4 has no effect on apoptosis, caspase-3 activation or expression of Noxa and Bim in MG-132-treated JNK22/2 cells. (A) Western blots showing the status of the indicated proteins in JNK22/2 cells that were either left untreated or exposed for the indicated times to MG-132 72 hours post transfection with control, ATF3 or ATF4 siRNAs. One representative experiment out of three is shown. (B and C) Fluorometric and flow cytometric determination of caspase-3 (DEVDase) activities and cell death (propidium iodide uptake), respectively, in JNK22/2 cells that were treated as described in A. Values are the mean of three independent experiments +/2 SD. (TIF) Figure S4 Knockdown of Hif1a and the glucocorticoid receptor (GR) has no effect on apoptosis, caspase-3 activation or expression of Noxa and Bim in MG-132-treated JNK22/2 cells. (A and D) Western blots showing the status of the indicated proteins in JNK22/2 cells that were either left untreated or exposed for the indicated times to MG-132 72 hours post transfection with control, Hif1a or GR siRNAs. One representative experiment out of three is shown. (B, C, E, F) Fluorometric and flow cytometric determination of caspase-3 (DEVDase) activities and cell death (propidium iodide uptake), respectively, in JNK22/2 cells that were treated as described in A and D. Values are the mean of three independent experiments +/2 SD. (TIF)