Translational Repression of MCL-1 Couples Stress-induced eIF2α Phosphorylation to Mitochondrial Apoptosis Initiation*

The integrated stress response (ISR) integrates a broad range of environmental and endogenous stress signals to the phosphorylation of the α-subunit of eukaryotic translation initiation factor 2 (eIF2α). Although intense or prolonged activation of this pathway is known to induce apoptosis, the molecular mechanisms coupling stress-induced eIF2α phosphorylation to the cell death machinery have remained incompletely understood. In this study, we characterized apoptosis initiation in response to classical activators of the ISR (tunicamycin, UVC, elevated osmotic pressure, arsenite). We found that all applied stress stimuli activated a mitochondrial pathway of apoptosis initiation. Rapid and selective down-regulation of the anti-apoptotic BCL-2 family protein MCL-1 preceded the activation of BAX, BAK, and caspases. Stabilization of MCL-1 blocked apoptosis initiation, while cells with reduced MCL-1 protein content were strongly sensitized to stress-induced apoptosis. Stress-induced elimination of MCL-1 occurred with unchanged protein turnover and independently of MCL-1 mRNA levels. In contrast, stress-induced phosphorylation of eIF2α at Ser51 was both essential and sufficient for the down-regulation of MCL-1 protein in stressed cells. These findings indicate that stress-induced phosphorylation of eIF2α is directly coupled to mitochondrial apoptosis regulation via translational repression of MCL-1. Down-regulation of MCL-1 enables but not enforces apoptosis initiation in stressed cells.

The integrated stress response (ISR) integrates a broad range of environmental and endogenous stress signals to the phosphorylation of the ␣-subunit of eukaryotic translation initiation factor 2 (eIF2␣). Although intense or prolonged activation of this pathway is known to induce apoptosis, the molecular mechanisms coupling stress-induced eIF2␣ phosphorylation to the cell death machinery have remained incompletely understood. In this study, we characterized apoptosis initiation in response to classical activators of the ISR (tunicamycin, UVC, elevated osmotic pressure, arsenite). We found that all applied stress stimuli activated a mitochondrial pathway of apoptosis initiation. Rapid and selective down-regulation of the anti-apoptotic BCL-2 family protein MCL-1 preceded the activation of BAX, BAK, and caspases. Stabilization of MCL-1 blocked apoptosis initiation, while cells with reduced MCL-1 protein content were strongly sensitized to stress-induced apoptosis. Stress-induced elimination of MCL-1 occurred with unchanged protein turnover and independently of MCL-1 mRNA levels. In contrast, stress-induced phosphorylation of eIF2␣ at Ser 51 was both essential and sufficient for the down-regulation of MCL-1 protein in stressed cells. These findings indicate that stress-induced phosphorylation of eIF2␣ is directly coupled to mitochondrial apoptosis regulation via translational repression of MCL-1. Down-regulation of MCL-1 enables but not enforces apoptosis initiation in stressed cells.
The integrated stress response (ISR) 2 is a general stress response program conserved from yeast to mammals, that is known to integrate various types of environmental and endogenous stress signals, including endoplasmic reticulum stress, amino acid deprivation, infection with double-stranded RNA viruses, osmotic stress, UV light exposure, heme deficiency, and oxidative stress (1)(2)(3)(4). Those diverse signals activate specific stress kinases, each of which converges on the phosphorylation at Ser 51 on the ␣-subunit of eukaryotic translation initiation factor 2 (eIF2␣) (1). Phosphorylation of eIF2␣ at Ser 51 abrogates the function of eIF2␣, required for the transfer of the initiator Met-tRNA i Met to the small ribosomal subunit. This leads to a shutdown of global mRNA translation due to reduced AUG initiator codon recognition, along with increased translation of a few selected mRNAs including ATF-4, a basic zipper transcription activator (5)(6)(7). As a consequence of global translational arrest, the steady-state levels of most cellular proteins decrease with time, dependent on their respective protein halflife. This has recently been found to be essential for NF-B activation mediated by down-regulation of IB proteins in response to eIF2␣ phosphorylation (8 -10). These findings provide a first striking example of a signaling pathway activated by translational repression of a labile protein during eIF2␣ phosphorylation. Activation of the ISR also mobilizes stress-induced gene expression induced by ATF4 and its target genes, involved in cell growth, differentiation and apoptosis (11,12). The biological consequences of activation of the ISR largely depend on the context. While activation of the eIF2␣ phosphorylation pathway is primarily thought to protect cells from ongoing damage (11,13,14), prolonged or extensive activation of this pathway has been found to induce apoptosis (15)(16)(17)(18). However, the molecular mechanisms coupling activation of the ISR to the initiation of apoptosis have remained incompletely understood. In the vast majority of cell types, apoptosis depends on the mitochondrial pathway. Therein, disruption of mitochondrial integrity leads to the release of pro-apoptotic proteins such as cytochrome c, AIF, Smac/Diablo, EndoG, and HtrA2/Omi from the mitochondrial inter-membrane space into the cytosol (19). Cytochrome c contacts Apaf-1, thereby triggering the recruitment and auto-activation of procaspase-9. Caspase-9 then activates a cascade of specialized proteases, termed executioner caspases, which in turn execute cell death. Mitochondrial integrity is regulated by the members of the BCL-2 family of proteins (20). BAX and BAK, two pro-apoptotic multidomain BCL-2 proteins, directly or indirectly induce mitochondrial outer membrane permeabilization (MOMP) by oligomerization occurring upon their activation (21). MOMP is prevented by multidomain anti-apoptotic BCL-2 proteins such as BCL-2, BCL-x L , and MCL-1. Members of the BH 3 -only group of BCL-2 proteins undergo induction and/or activation by apoptotic *This work was supported in part by Technical University of Munich (KKF 8733166, to R. F.) and Deutsche Forschungsgemeinschaft (SFB 456, to G. S. and R. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. stimuli and either directly activate BAX-BAK or specifically neutralize anti-apoptotic multidomain BCL-2 proteins (22). In this study, we aimed to uncover the molecular cross-talk between cellular stress signaling and apoptosis. We characterized apoptosis initiation in response to classical activators of the ISR and found that stress-induced apoptosis essentially involves the rapid and selective down-regulation of MCL-1 protein, mediated by stress-induced phosphorylation of eIF2␣ at Ser 51 .

EXPERIMENTAL PROCEDURES
Materials-The following drugs were purchased as indicated: Coumermycin (Fluka); ZVAD-fmk (Biomol); dithiothreitol and actinomycin D (Sigma); arsenite (Merck); MG-132, tunicamycin, thapsigargin, and cycloheximide (Calbiochem). Drugs were dissolved in appropriate solvent and further diluted in cell culture medium. Appropriate solvent controls and mock treatment were performed throughout the experiments.
Cell Culture and Treatment-HeLa, HEK293, MiaPaCa-2, and EGI-1 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum, 2 mmol/l L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. DANG and TFK-1 cells were cultured in RPMI 1640 supplemented identically. All cells were grown to a subconfluent monolayer in a humidified atmosphere containing 5% CO 2 at 37°C. For UV irradiation, culture medium was removed and cells were covered with a small volume of prewarmed phosphate-buffered saline. The lid of the cell culture dish was removed, and cells were exposed to UVC (254 nm) using a cross-linker (Stratagene). To generate osmotic stress, cells were exposed to cell culture medium supplemented with 200 mM NaCl in excess for 30 min. MG-132 and ZVAD-fmk were added to cells 30 min prior to the exposure to cellular stress.
Transient and Stable Transfection-Plasmids were delivered into cells using Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer's recommendations. Cells were transfected at a density of 70 -80% and used for experiments 24 h after transfection. For stable expression of FLAG-MCL-1 in HeLa cells, the pcDNA3.2-FLAG-MCL-1 expression plasmid was linearized prior to transfection and clones were raised in the presence of 1 mg/ml Geneticin (Invitrogen). Clones were assayed for transgene expression by Western blot analysis.
Plasmids and Cloning-MCL-1 cDNA was amplified by RT-PCR using a cDNA library generated from HeLa cells. A FLAG tag sequence was introduced at the 5Ј-side of MCL-1 cDNA. pCI-neo-EGFP was generated by PCR subcloning of EGFP cDNA obtained from pEGFP-C1 (BD Biosciences Clontech) into pCI-neo (Stratagene). pCI-neo-FLAG-eIF2␣plasmids (wt, S51A, S51D) were generated by PCR subcloning of the respective eIF2␣ cDNAs obtained from expression plasmids kindly provided by David Ron. A FLAG tag was introduced at the 5Ј-side of the respective eIF2␣ cDNAs. pCI-neo-GyrB.PKR and the respective inactive mutant (pCIneo-GyrB.PKR-K296H) were generated by PCR subcloning of GyrB.PKR cDNAs obtained from expression plasmids kindly provided by Tom Dever and described elsewhere (23). Integrity of all subcloned sequences was verified by automated DNA sequencing. Detailed cloning procedures and primer sequences are available on request.
Immunoprecipitations for active BAK and BAX were conducted after lysis of cells in CHAPS buffer (5 mM MgCl 2 , 137 mM KCl, 1 mM EDTA, 1 mM EGTA, 1% CHAPS, 20 mM Tris-HCl, pH 7.5). 1 mg/ml protein lysate was precleared with 20 l of A-/G-Sepharose (Santa Cruz Biotechnology) for 1 h. 2 g of mouse anti-BAX 6A7 (Alexis) or mouse anti-BAK Ab-1 (Calbiochem), were added to 1 mg of cell lysate and incubated at 4°C overnight. Antibodies were precipitated with 20 l A-/G-Sepharose for 2 h and pellets were washed three times with phosphate-buffered saline before proteins were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes (Immobilon, Millipore). Precipitated BAX or BAK were detected by immunoblot as described above.
Quantitative RT-PCR-Total RNA was isolated from cells using the RNeasy kit (Qiagen). A DNase digestion step was included in all RNA isolation procedures to exclude DNA contamination. Specific mRNAs were quantified by real-time RT-PCR as described previously (24). Primer sequences are available on request. For detection of FLAG-tagged mRNAs, primer pairs specifically amplifying FLAG-tagged target sequences were chosen that did not amplify the corresponding untagged mRNAs or other FLAG-tagged sequences. Number of copies mRNA were determined by using standard curves generated by serial dilutions of plasmids containing the respective target sequence. Data are shown as relative number of copies mRNA. Data were obtained from at least three independent experiments.
Chromatin Condensation-Apoptotic nuclei were quantified by fluorescence microscopy after staining of chromatin with Hoechst 33342 according to the manufacturer's recommendations. For each experiment, a minimum of 200 nuclei from three different microscopic fields were counted under blinded conditions.

RESULTS
A Mitochondrial Pathway of Apoptosis Initiation Is Activated in Response to Classical Activators of the ISR-To study apoptosis initiation in response to activation of the ISR, HeLa cells were exposed to a number of classical ISR activators: the ERstress inductor tunicamycin (Fig. 1A), UVC (Fig. 1B), elevated osmotic pressure (Fig. 1C) and arsenite (Fig. 1D). All stress stimuli were applied at doses found to elicit a strong apoptotic response in HeLa cells. Activation of BAX and BAK was assayed by immunoprecipitation using conformation-specific antibodies (25,26). Cleavage of caspase-9 and caspase-3 was detected by Western blot. All applied stress stimuli synchronously induced the activation of BAX and BAK as well as cleavage of caspase-9 and caspase-3 ( Fig. 1, A-D, lanes 1-4). Biochemical signs of apoptosis were accompanied by characteristic morphological changes such as cytoplasmatic shrinkage and condensation of chromatin (data not shown). Parallel experiments were conducted in the presence of ZVAD-fmk, a broad-spectrum caspase inhibitor. Addition of ZVAD-fmk strongly inhibited cleavage of caspase-9 and completely prevented processing of caspase-3 ( Fig.  1, A-D, lanes [5][6][7][8] as well as condensation of chromatin (not shown). However, application of ZVAD-fmk did not affect the activation of BAX and BAK, indicating that BAX and BAK were activated upstream of caspases and that apoptosis was initiated at the mitochondria. Initiation of apoptosis was assayed by immunoprecipitation for active BAX and BAK and by immunoblotting for cleavage of caspase-9 and -3 at indicated time points. Protein levels of pro-(BAX, BAK) and anti-apoptotic (BCL-2, BCL-x L and MCL-1) multidomain BCL-2 proteins were determined by immunoblotting. Membranes were stripped and re-probed with ␤-actin as a loading control.

Down-regulation of MCL-1 Precedes Apoptosis Initiation in
Response to Cellular Stress-Mitochondrial integrity is preserved by the balance of pro-and anti-apoptotic members of the BCL-2 protein family. To uncover alterations within the levels of pro-and anti-apoptotic multidomain BCL-2 proteins during apoptosis initiation in response to the applied ISR activators, we performed Western blot analysis for BAX, BAK, BCL-2, BCL-x L , and MCL-1 (Fig. 1, A-D, lanes 1-4). Strikingly, all applied stress stimuli induced the rapid and selective downregulation of the anti-apoptotic BCL-2 protein MCL-1. Downregulation of MCL-1 preceded all further events of apoptosis initiation and was not affected by addition of ZVAD-fmk (Fig. 1, A-D, lanes [5][6][7][8]. However, although the initial decrease in MCL-1 levels remained unaffected by ZVAD-fmk, reappearance of MCL-1 occurred in response to tunicamycin, osmotic stress, and arsenite when caspase activity was blocked (Fig. 1, A, C, D, lanes [5][6][7][8].
Down-regulation of MCL-1 in Response to Cellular Stress Is Not Restricted to a Single Cell Type-To ensure that the rapid down-regulation of MCL-1 in response to the applied stress stimuli is not restricted to a single cell type, we exposed several cancer cell lines to tunicamycin, UVC, elevated osmotic pressure, and arsenite, and studied MCL-1 protein levels. Cell lines examined were MiaPaCa-2 and DANG pancreatic cancer cells as well as TFK-1 and EGI-1 cholangiocarcinoma cells. As shown in Fig. S1, MCL-1 was rapidly lost in response to all of the applied stresses in all cell lines tested.
Stabilization of MCL-1 by Proteasome Inhibition Prevents Stress-induced Apoptosis-We next aimed to determine whether stabilizing MCL-1 under stressful conditions would alter stress-induced apoptosis. MCL-1 is a high turnover protein that is rapidly degraded by the proteasome (27). Consequently, addition of a proteasome inhibitor (MG-132), but not adequate solvent control, completely prevented the decrease in MCL-1 levels in response to UVC, osmotic stress and arsenite (Fig. 2, A-C, compare lanes 1-4 and 5-8). Moreover, MG-132 completely blocked mitochondrial apoptosis initiation as demonstrated by absent precipitation of active BAX and non-detectable cleavage of caspase-3. Because inhibition of the proteasome affects degradation and thereby increases concentration of a myriad of proteins, we next asked whether inhibition of apoptosis by MG-132 was dependent on the observed stabiliza- tion of MCL-1. To this end, we transfected cells either with control siRNA or siRNA specifically targeting MCL-1 (siMCL-1-1, see also Fig. 3) and exposed them to the respective stresses in the presence of MG-132. As shown in Fig. 2 (A-C, compare lanes 9 -12  and lanes 13-16), MG-132 blocked apoptosis initiation in controls (lanes 9 -12), but not in cells with strongly reduced MCL-1 protein content (lanes 13-16). These findings indicate that stabilization of MCL-1 at pre-existing levels accounted for the blockage of apo-ptosis initiation in the presence of MG-132. Of note, addition of MG-132 alone did not induce apoptosis during the time course of these experiments in both controls and cells with reduced MCL-1 abundance. However, significant toxicity occurred when cells were exposed to proteasome inhibitors for longer incubation times (ϳ12 h). Therefore, we could not perform parallel experiments with ER-stress-inducing agents that required longer incubation times to induce apoptosis.  1-1, siMCL-1-2). Protein lysates were assayed for MCL-1 protein content by immunoblotting 24 h after transfection. B, untreated HeLa cells (wt), and cells transfected with control siRNA or siRNA specifically targeting MCL-1 (siMCL-1-1, siMCL-1-2) were exposed to increasing doses of tunicamycin, UVC, osmotic stress, and arsenite as indicated. Cells were stained with H33342 after 4 h (UVC, osmotic stress, arsenite) or 24 h (tunicamycin). Chromatin condensation was assayed by fluorescence microscopy. C, HeLa cells transfected with nonspecific siRNA (control) or siMCL-1-1 were exposed to increasing doses of tunicamycin, UVC, osmotic stress, and arsenite as indicated. PARP cleavage was assayed by immunoblotting, ␤-actin served as loading control.

MCL-1 Is a Survival Factor in Response to Cellular Stress-In
contrast to other anti-apoptotic BCL-2 proteins like BCL-x L and BCL-2, MCL-1 is a short-lived protein that undergoes continuous regulation at the level of transcription, translation, and degradation (28). We therefore aimed to determine whether alterations in MCL-1 protein levels would affect the cellular response to the applied ISR activators. To this end, we modified MCL-1 protein expression through RNA interference and conducted dose-response experiments. In these experiments, apoptosis was assayed by quantification of chromatin condensation (Fig. 3B) and immunoblotting for PARP cleavage (Fig. 3C). Fig.  3A demonstrates efficient knockdown of MCL-1 protein after transfection of cells with two different siRNAs (siMCL-1-1, siMCL-1-2) specifically targeting MCL-1. In line with findings reported by others (29), down-regulation of MCL-1 protein alone was not sufficient to induce cell death (Fig. 3, B and C). Furthermore, cells with reduced MCL-1 protein levels proliferated normally and exhibited no gross morphological changes (data not shown). However, cells with reduced MCL-1 protein expression levels were strongly sensitized to the applied stress stimuli and underwent apoptosis in response to very mild stress conditions (Fig. 3, B and C). These findings indicate that MCL-1 protein levels determine apoptosis sensitivity under stressful conditions. Cellular Stress Does Not Affect MCL-1 Protein Turnover-We next asked for the molecular mechanisms mediating down-regulation of MCL-1 in response to the applied activators of the ISR. Throughout these experiments, the more rapidly acting ER-stress inducing drug thapsigargin was used instead of tunicamycin. As shown in Fig. 4A, elimination of MCL-1 in response to the applied stresses occurred very rapidly. To determine whether this was due to an increase in MCL-1 protein turnover, the kinetics of MCL-1 protein degradation were uncovered by treating cells with the translation inhibitor cycloheximide. At the same time, cells were exposed either to mock treatment (Fig. 4B, lanes 1-4) or to stress stimuli as indicated (Fig. 4B, lanes 5-8). Addition of cycloheximide uncovered a short constitutive MCL-1 protein half-life of ϳ30 min in HeLa cells (Fig. 4B, lanes 1-4). Notably, none of the applied stresses further accelerated the disappearance of MCL-1 protein when compared with mock treatment (Fig. 4B, compare lanes 5-8  and 1-4). These findings suggest that the rapid decrease in MCL-1 levels in response to cellular stress is not mediated by enhanced MCL-1 protein degradation.

Different Types of Stress Exhibit Different
Effects on MCL-1 mRNA-We next studied MCL-1 mRNA behavior in response to the applied stress stimuli. As shown in Fig. 4C, no uniform effect on MCL-1 mRNA was observed. Exposure of cells to UV light induced a rapid and sustained loss of MCL-1 mRNA, while

. UVC-induced MCL-1 mRNA degradation is dispensable for the down-regulation of MCL-1 protein.
A, HeLa cells were exposed to various doses of UVC. Total cellular RNA was isolated at indicated time points, and MCL-1 mRNA was quantified by real-time RT-PCR. B, HeLa cells were treated with actinomycin D (5 g/ml) in combination with mock treatment or exposure to UVC (100 J/m 2 ). Total RNA was isolated, and MCL-1 mRNA abundance was measured by real-time RT-PCR at the indicated time points. C, two different HeLa cell clones (37 and 9) stably expressing FLAG-tagged MCL-1 (FLAG-MCL-1) were exposed to UVC (100 J/m 2 ), and total mRNA was isolated indicated time points. FLAG-MCL-1 mRNA was quantified by real time-RT-PCR using primer pairs specifically amplifying FLAG-MCL-1. Endogenous MCL-1 mRNA abundance was determined by subtraction of FLAG-MCL-1 mRNA from total MCL-1 mRNA measured within the same experiment. D, two independent HeLa cell clones (37 and 9) stably expressing FLAG-MCL-1 were exposed to UVC (100 J/m 2 ), and immunoblotting for FLAG-MCL-1, total MCL-1, and ␤-actin protein levels was performed at the indicated time points.
osmotic stress elicited a more transient decline in MCL-1 mRNA levels, followed by an increase above baseline levels (Fig.  4C, left diagram). Arsenite did not significantly affect MCL-1 mRNA abundance, whereas thapsigargin induced a slight but steady increase in MCL-1 mRNA levels (Fig. 4C, right diagram). Virtually identical results were obtained in parallel experiments performed with MiaPaCa-2 cells (data not shown).

UV-induced MCL-1 mRNA Degradation Is Dispensable for UV-induced Down-regulation of MCL-1 Protein-
The rapid decrease of MCL-1 mRNA in response to UVC paralleled the kinetics of MCL-1 protein disappearance. To determine whether UV-induced down-regulation of MCL-1 mRNA was required for the down-regulation of MCL-1 protein, we further characterized UV-induced loss of MCL-1 mRNA. As shown in Fig. 5A, UVC induced a dose-dependent decline in MCL-1 mRNA levels. Treatment of cells with actinomycin D to block novel mRNA synthesis uncovered a MCL-1 mRNA half-life of ϳ2 h (Fig. 5B). Exposure of actinomycin D-treated cells to UVC strongly enhanced MCL-1 mRNA decay, followed by a stabilization of the remaining pool (Fig. 5B), indicating that UV light interfered with MCL-1 mRNA stability. Regulation of mRNA stability during cellular stress has been found to depend on regulatory elements located within the 5Ј-and 3Ј-untranslated regions of eukaryotic mRNAs (30). Therefore, we tested whether transgenic MCL-1 mRNA devoid of 5Ј-and 3Ј-untranslated regions would undergo UV-induced degradation. Two independent HeLa cell clones (37 and 9) raised after stable transfection with pcDNA-3.2-FLAG-MCL-1 were exposed to UV light and both mRNA (Fig. 5C) and protein (Fig. 5D) levels of FLAG-tagged MCL-1 and endogenous MCL-1 were studied. As shown in Fig. 5C, endogenous MCL-1 mRNA was rapidly down-regulated in response to UVC, while FLAG-MCL-1 mRNA remained largely unaffected. However, at protein level, both FLAG-tagged MCL-1 and endogenous MCL-1 protein were eliminated with identical kinetics (Fig. 5D), indicating that the down-regulation of MCL-1 protein occurred independently of MCL-1 mRNA behavior. In summary, all of the applied stresses affected MCL-1 protein abundance downstream of MCL-1 mRNA levels, and upstream of MCL-1 protein degradation, suggesting blockage at the level of mRNA translation.
Phosphorylation of eIF2␣ Parallels the Suppression of MCL-1 Protein-Manifold stresses have been shown to induce global translational arrest by inducing the phosphorylation of the ␣-subunit of eukaryotic translation initiation factor 2 (eIF2␣) at Ser 51 , leading to decreased formation of the ternary complex required for the binding of Met-tRNA i Met to the 40 S ribosomal subunit (5). To determine whether stress-induced down-regulation of MCL-1 was related to eIF2␣ phosphorylation, the kinetics of eIF2␣ phosphorylation and MCL-1 down-regulation were studied in parallel. As shown in Fig. 6, all of the applied stresses induced a strong phosphorylation of eIF2␣ that was temporally paralleled by the down-regulation of MCL-1. In detail, exposure of cells to UVC induced a rapid, strong and sustained eIF2␣ phosphorylation that inversely correlated with MCL-1 protein levels (Fig. 6A). Osmotic stress induced a strong but transient phosphorylation of eIF2␣, which was paralleled by a sudden but transient decrease of MCL-1 protein (Fig. 6B). Similarly, the reducing agent and ISR activator DTT, arsenite (Fig. 6C), thapsigargin and tunicamycin (Fig. 6D) induced eIF2␣ phosphorylation and down-regulation of MCL-1 protein in parallel.

FIGURE 6. Stress-induced down-regulation of MCL-1 is paralleled by phosphorylation of eIF2␣.
A, HeLa cells were exposed to UVC (100 J/m 2 ), and cells were harvested in whole cell lysis buffer supplemented with phosphatase inhibitors as described under "Experimental Procedures." Phosphorylation of eIF2␣ at Ser 51 was detected by immunoblotting using a phosphospecific antibody. Total eIF2␣, MCL-1, and ␤-actin were visualized after stripping of the membrane. B, HeLa cells were exposed to elevated osmotic pressure (400 mosm/liter NaCl in excess for 30 min), and immunoblotting was performed as indicated. C, treatment of HeLa cells with the reducing agent and ISR activator dithiothreitol (DTT, 1 mM) or arsenite (200 M) as indicated. D, HeLa cells were exposed to thapsigargin (1 M) or tunicamycin (2 g/ml) and immunoblotting was performed as indicated.
Activation of PKR Induces the Elimination of MCL-1-To test whether phosphorylation of eIF2␣ by one of its upstream kinases results in elimination of MCL-1, we used a chemically activatable fusion construct containing the catalytic domain of the dsRNA-activated protein kinase (PKR) and Escherichia coli gyrase B (23). PKR is one of four kinases known to phosphorylate eIF2␣ at Ser 51 (33). The GyrB.PKR fusion protein undergoes rapid dimerization and activation upon addition of coumermycin. As shown in Fig. 8C, HEK293 cells were transfected with pCI-neo expression plasmids encoding for EGFP (lanes 1 and 2), GyrB.PKR-wt (lanes 3 and 4) or the catalytically inactive mutant GyrB.PKR-K296H (lanes 5 and 6). Addition of coumermycin induced an increase in eIF2␣ phosphorylation as well as a partial loss of MCL-1 protein in cells expressing  (Fig. 8A, lane 3). Again, transfection efficiency in HEK293 cells was 70 -80%. Parallel experiments were conducted with HeLa cells cotransfected with pcDNA3.2-FLAG-MCL-1 and the respective PKR constructs. As shown in Fig. 8D, addition of coumermycin induced a virtually complete elimination of FLAG-MCL-1 paralleled by increased eIF2␣ phosphorylation specifically in cells expressing GyrB.PKR-wt (Fig. 8D,  compare lanes 3 and 4 to lanes 1 and 2 and 5 and 6). Taken together, these data show that phosphorylation of eIF2␣ is sufficient to rapidly and profoundly down-regulate MCL-1 protein.

Down-regulation of MCL-1 Enables but Not Enforces Apoptosis Initiation in Stressed
Cells-In this study, we show that apoptosis initiation in cells challenged with stimuli activating the integrated stress response involves the rapid down-regulation of the anti-apoptotic BCL-2 family protein MCL-1. Elimination of MCL-1 has been shown to be required for apoptosis induction in response to a number of stimuli including DNAdamaging agents (29), adenovirus infection (34), anoxia (35), FIGURE 8. Phosphorylation of eIF2␣ is sufficient to eliminate MCL-1. A, HEK293 cells were transiently transfected with control vector (pCI-neo-EGFP) or expression plasmids for wild-type eIF2␣ (pCI-neo-FLAG-eIF2␣-wt), the phosphomimetic mutant eIF2␣-S51D (pCI-neo-FLAG-eIF2␣-S51D) or the non-phosphorylatable mutant eIF2␣-S51A (pCI-neo-FLAG-eIF2␣-S51A). Cells were harvested 16 h later, and immunoblot analysis of MCL-1, FLAG-eIF2␣, and ␤-actin protein content were performed. In parallel experiments, MCL-1 mRNA (middle panel) and FLAG-eIF2␣ mRNA (lower panel) expression levels were determined by real time RT-PCR using primer pairs amplifying MCL-1 and specifically FLAG-tagged eIF2␣. B, HeLa cells were co-transfected with expression plasmids encoding for FLAG-tagged MCL-1 (pcDNA3.2-FLAG-MCL-1) and either wild-type eIF2␣ (pCI-neo-FLAG-eIF2␣-wt), the phosphomimetic mutant eIF2␣-S51D (pCI-neo-FLAG-eIF2␣-S51D) or the non-phosphorylatable mutant eIF2␣-S51A (pCI-neo-FLAG-eIF2␣-S51A). Cells were harvested 16 h after transfection. Protein levels of FLAG-MCL-1, FLAG-eIF2␣, and ␤-actin were determined by Western blot. FLAG-MCL-1 (middle panel) and FLAG-eIF2␣ mRNA levels (lower panel) were determined by real-time RT-PCR in parallel experiments. C, HEK293 cells were transiently transfected with expression plasmids encoding for EGFP (pCI-neo-EGFP; lanes 1 and 2), a fusion protein containing the catalytic domain of the dsRNA-activated protein kinase (PKR) and E. coli gyrase B (pCI-neo-GyrB.PKR; lanes 3 and 4) or a catalytically inactive mutant (pCI-neo-GyrB.PKR-K296H; lanes 5 and 6). The GyrB.PKR constructs undergo dimerization and activation upon addition of coumermycin. Coumermycin (100 ng/ml) was added 24 h after transfection and cells were harvested 4 h later. Immunoblotting for MCL-1, PKR, P-eIF2␣, eIF2␣, and ␤-actin proteins was performed. D, HeLa cells were co-transfected with expression plasmids encoding for FLAG-tagged MCL-1 (pcDNA3.2-FLAG-MCL-1) and either EGFP (pCI-neo-EGFP; lanes 1 and 2), GyrB.PKR (pCI-neo-GyrB.PKR; lanes 3 and 4), or GyrB.PKR-K296H (pCI-neo-GyrB.PKR-K296H; lanes 5 and 6). 24 h after transfection, coumermycin or solvent control was added, and cells were harvested for Western blot 4 h later. and IL-3 withdrawal (36). Our findings point to a crucial role of MCL-1 in regulating the onset of apoptosis in stressed cells. In line, we found that artificial down-regulation of MCL-1 was sufficient to strongly sensitize cells to very mild stress conditions, suggesting that MCL-1 protein levels play a crucial role in determining cell fate under stressful conditions. This is of special importance since MCL-1 is unique among the group of anti-apoptotic members of the Bcl-2 family in that it is a highly regulated short lived protein responsive to multiple intracellular and extracellular signaling events (28). In line with other reports (29,37), we found that down-regulation of MCL-1 alone was not sufficient to induce apoptosis. These results implicate that additional apoptotic events must coincide with down-regulation of MCL-1 to disrupt mitochondrial integrity in stressed cells. Activation of BIM has been found to be involved in UV-induced apoptosis (38) and BIM-deficient lymphocytes showed resistance toward apoptosis induction in response to ionomycin (39), a calcium ionophore known to induce ER-stress. Others have found induction of NOXA and PUMA to be involved in ER-stress-mediated apoptosis (40,41). Further studies will be required to identify additional apoptotic events involved and to determine whether these signals are elicited in a stimulus-specific manner or generally occur with activation of the integrated stress response.
Down-regulation of MCL-1 Is the Consequence of Stress-induced Phosphorylation of eIF2␣-Our experiments show that down-regulation of MCL-1 in stressed cells is the consequence of stress-induced phosphorylation of eIF2␣ at Ser 51 . This conclusion is supported by the following observations: 1) None of the applied stress stimuli significantly affected MCL-1 protein turnover. 2) A profound and persistent decrease of MCL-1 mRNA occurred exclusively in response to UVC and was dispensable for UV-induced down-regulation of MCL-1 protein.
3) Phosphorylation of eIF2␣ at Ser 51 temporally paralleled the decrease in MCL-1 and abrogation of this pathway prevented decrease of MCL-1 in stressed cells. 4) Mimicking eIF2␣ phosphorylation and phosphorylation of eIF2␣ by PKR in the absence of external stress stimuli was sufficient to down-regulate MCL-1. Various mechanisms have been shown to lead to a down-regulation of MCL-1 during apoptosis. In response to DNA damaging agents, elimination of MCL-1 occurred with unchanged protein turnover due to a stop of novel MCL-1 protein synthesis (29). Others have found that interaction with NOXA enhanced MCL-1 degradation (37,42). Upon IL-3 withdrawal, accelerated degradation of MCL-1 occurred in response to phosphorylation by glycogen synthase kinase-3 (36). Our study presents a rather conclusive pathway, explaining why MCL-1 protein is rapidly lost in stressed cells. Manifold stresses converge on the phosphorylation of eIF2␣ at Ser 51 , leading to a global shutdown of mRNA translation, affecting virtually all mRNAs with a few exceptions (5). Consequently, expression levels of most cellular proteins will decrease with time, according to their respective protein half-lives. The consequential rapid decrease in the steady-state level of high-turnover proteins has recently been found to lead to the activation of NF-B in response to eIF2␣ phosphorylation, mediated by the down-regulation of IB proteins (8 -10). The down-regulation of MCL-1 protein in response to eIF2␣ phosphorylation follows the same basic model. The decrease of MCL-1 occurs so rapidly, because its constitutively high protein turnover (27) demands a high rate of protein synthesis to maintain a high steady state level of MCL-1 protein. Thus, the global inhibition of mRNA translation arising from eIF2␣ phosphorylation in stressed cells exhibits specific effects on mitochondrial apoptosis signaling by rapidly removing MCL-1, acting as essential inhibitor of apoptosis initiation in stressed cells.
Down-regulation of MCL-1 Couples Cellular Stress Signaling to Apoptosis-Activation of the ISR has been found to promote both, death or survival, depending on the physiological context (13,15). While the transcriptional branch of the ISR has been found to recruit death pathways requiring induction of the ATF4 target CHOP (43,44), the translational branch, centered by eIF2␣ phosphorylation, has mainly been considered a protective pathway (11). This credo has been challenged by recent studies showing that eIF2␣ phosphorylation may also act as an important cellular death stimulus (15)(16)(17)(18), raising the question of how eIF2␣ phosphorylation contributes to cell death. Our findings show that phosphorylation of eIF2␣ is coupled to the rapid down-regulation of the highly unstable anti-apoptotic protein MCL-1. We further found that this decrease of MCL-1 is required for apoptosis initiation in stressed cells. However, because down-regulation of MCL-1 alone is not sufficient to disrupt mitochondrial integrity, phosphorylation of eIF2␣ does not automatically induce cell death, making this pathway compatible with both, a pro-death and a pro-survival role of eIF2␣ phosphorylation, depending on concomitantly occurring signaling events. Further studies will be required to better dissect the interwoven pathways of death and survival activated by eIF2␣ phosphorylation in stressed cells, improving our understanding of how the life or death decision is made under stressful conditions.