Double-stranded RNA-dependent Protein Kinase Phosphorylation of the α-Subunit of Eukaryotic Translation Initiation Factor 2 Mediates Apoptosis*

As the molecular processes of complex cell stress signaling pathways are defined, the subsequent challenge is to elucidate how each individual event influences the final biological outcome. Phosphorylation of the translation initiation factor 2 (eIF2α)atSer51 is a molecular signal that inhibits translation in response to activation of any of four diverse eIF2α stress kinases. We used gene targeting to replace the wild-type Ser51 allele with an Ala in the eIF2α gene to test the hypothesis that translational control through eIF2α phosphorylation is a central death stimulus in eukaryotic cells. Homozygous eIF2α mutant mouse embryo fibroblasts were resistant to the apoptotic effects of dsRNA, tumor necrosis factor-α, and serum deprivation. TNFα treatment induced eIF2α phosphorylation and activation of caspase 3 primarily through the dsRNA-activated eIF2α kinase PKR. In addition, expression of a phospho-mimetic Ser51 to Asp mutant eIF2α-activated caspase 3, indicating that eIF2α phosphorylation is sufficient to induce apoptosis. The proapoptotic effects of PKR-mediated eIF2α phosphorylation contrast with the anti-apoptotic response upon activation of the PKR-related endoplasmic reticulum eIF2α kinase, PERK. Therefore, divergent fates of death and survival can be mediated through phosphorylation at the same site within eIF2α. We propose that eIF2α phosphorylation is fundamentally a death signal, yet it may promote either death or survival, depending upon coincident signaling events.

One of the most perplexing problems in modern biology is to understand how the cell chooses between adaptation and apoptotic demise in response to stressful insults. Because there are multiple interacting anti-apoptotic and pro-apoptotic signaling pathways, it is assumed that the sum of these signaling cascades dictates the final outcome. When one pathway becomes predominant, a delicate balance is perturbed and either an adaptive or a lethal response ensues. Advances in our knowledge of how this commitment occurs will lead to a greater understanding of cell growth and differentiation as well as the etiology of various disease states.
Numerous phosphorylation events are known to regulate the overall rate of protein synthesis or translation of selective mRNAs. However, the most dominant influence is mediated through phosphorylation at Ser 51 on the ␣-subunit of heterotrimeric eukaryotic translation initiation factor 2 (eIF2␣) 3 (1). eIF2 is required to deliver Met tRNA i to the 40 S ribosomal subunit. Physiological conditions that induce eIF2␣ Ser 51 phosphorylation regulate global as well as specific mRNA translation. Phosphorylation of eIF2␣ at Ser 51 inactivates eIF2 and reduces the efficiency of AUG initiation codon recognition, thereby attenuating translation initiation. However, reduced AUG initiation codon recognition can increase the initiation efficiency at selective AUG codons, thereby altering initiation site utilization to regulate both the quantity and quality of proteins produced (2).
Four protein kinases phosphorylate eIF2␣ at Ser 51 in response to different stress stimuli: 1) the dsRNA-activated protein kinase PKR is a major component of the interferonmediated antiviral response and is activated by binding to dsRNA produced during viral infection (3); 2) the general control of nitrogen metabolism kinase GCN2 responds to amino acid depletion (4); 3) the heme-regulated inhibitor kinase HRI responds to heme deprivation to couple globin synthesis with available heme (5); and 4) the PKR-related endoplasmic reticulum (ER) kinase PERK responds to the accumulation of unfolded proteins in the ER in a subpathway of the unfolded protein response (6). Generally, eIF2␣ phosphorylation provides a fundamental mechanism to couple the rate of protein synthesis with the capacity to fold proteins under conditions of different physiological stress, such as nutrient deprivation or viral infection.
Although the mechanism by which phosphorylation of eIF2␣ inhibits protein synthesis is well characterized, the cellular responses to eIF2␣ phosphorylation remain elusive. Recent studies support the idea that eIF2␣ phosphorylation promotes survival under conditions of oxidative stress and accumulation of unfolded proteins in the lumen of the ER (7,8). In contrast, eIF2␣ phosphorylation was proposed to mediate apoptosis in response to PKR activation (9 -12). In this study, we addressed how eIF2␣ phosphorylation influences the balance between survival and apoptosis upon activation of PKR.
Treatment of cells with interferon and dsRNA is cytotoxic, and data support the hypothesis that this toxicity is mediated through PKR activation and induction of apoptosis (13). Although a number of different signal transduction and transcriptional programs are influenced through PKR activation (for review see Ref. 14), the most well characterized PKR substrate is eIF2␣. The growth suppressing activity mediated through eIF2␣ phosphorylation is an evolutionarily well conserved cell response. Either inactivation of the PKR pathway (12,(15)(16)(17)(18)(19)(20)(21)(22) or overexpression of a nonphosphorylatable S51A mutant eIF2␣ (10,12) protects from stress-mediated apoptosis. These studies provide compelling evidence for an anti-proliferative effect of PKR-mediated eIF2␣ phosphorylation in growth inhibition. However, the interpretation of these results is confounded because of the diverse effects that PKR activation has on multiple stress signaling pathways. Therefore, to date there is no direct evidence to support the hypothesis that eIF2␣ phosphorylation is necessary and/or sufficient for apoptosis.
We propose that PKR activation with subsequent eIF2␣ phosphorylation is a primary mechanism that 1) inhibits initiation of protein synthesis, and 2) contributes to apoptosis in response to a variety of physiological and environmental stimuli. To test this hypothesis, we have studied apoptosis induced by dsRNA, TNF␣, or serum deprivation in cells that harbor a homozygous S51A knock-in mutation at the phosphorylation site in eIF2␣ (23). Here, we show that apoptosis induced by TNF␣, the interferon pathway, and serum deprivation requires PKR-mediated phosphorylation of eIF2␣. In addition, expression of a Ser 51 to Asp phospho-mimetic mutant of eIF2␣ was sufficient to activate caspase 3 in the absence of any apoptosisinducing stimuli. The results demonstrate that translational inhibition through eIF2␣ phosphorylation contributes to and can be sufficient to activate an apoptotic response.

MATERIALS AND METHODS
Plasmid DNA Transfection-The poly(ADP-ribose) polymerase (PARP) cDNA cloned in pCDNA3 was kindly provided by Dr. M. Keifer, (LXR Biotechnologies). The eIF2␣ expression vectors were previously described (24). Transfection of eIF2␣ expression plasmids into HeLa cells was performed as described (25). After transfection, the cells were washed twice with Dulbecco's modified essential medium (DMEM) and incubated at 5% CO 2 for 2 days in DMEM with 10% fetal bovine serum-containing antibiotics. The cells were washed twice with phosphate-buffered saline (PBS), harvested using Nonidet P-40 lysis buffer (1% Nonidet P-40, 150 mM NaCl, 50 mM Tris, pH 8.0) with complete protease inhibitors (Roche), and incubated on ice for 15 min followed by centrifugation at 10,000 rpm for 10 min. The supernatant was collected, and protein concentrations were determined by the Bradford method (26).
In Vitro Transcription and Translation of PARP-In vitro transcription and translation of PARP was performed in the presence of [ 35 S]methionine/cysteine (Redivue PRO-MIX, Amersham Biosciences.) using the TNT kit (Promega Biotech) following the manufacturer's instructions. Cleavage of in vitro translated PARP was previously described (27). The reaction products were analyzed by SDS-PAGE and autoradiography using EN 3 HANCE (Dupont). Briefly, a 50-l reaction containing 50 mM HEPES, pH 7.4, 100 mM NaCl, 0.1% CHAPS, 10% sucrose, 20 g of cell extract protein from transfected COS-1 cells, and 2 l of in vitro translated [ 35 S]methionine/cysteinelabeled PARP were mixed and incubated at 37°C for 2 h. Then, 25 l of 3ϫ SDS-PAGE sample buffer was added to each sample followed by heating at 90°C for 5 min. The reaction products were analyzed by SDS-PAGE under reducing conditions. After electrophoresis, the gels were fixed, soaked for 45 min in EN 3 HANCE (Dupont), dried, and subjected to autoradiography. ImageJ (Version 1.31 for Mac OS X, NIH) was used to quantitate band intensities.
Immunoblot Analysis-Cells were treated as described and harvested using Nonidet P-40 lysis buffer containing 150 mM NaFl, complete protease inhibitors (Roche), and 100 g/ml phenylmethylsulfonyl fluoride. Lysis buffer additionally included 500 mM ␤-glycerol phosphate, 50 mM sodium orthovanadate, and 1ϫ phosphatase inhibitor (Sigma P2850) (Fig. 3, E and F; supplemental Fig. S2). Samples were centrifuged at 10,000 rpm, and supernatants were collected for SDS-PAGE and transfer to nitrocellulose. The eIF2␣ Ser 51 phosphospecific antibody was obtained from BioSource (Camarillo, CA) and PKR antibody was kindly provided by Dr. Bryan Williams (Cleveland Clinic). Phosphospecific PKR antibody (3075) and PKR antibody (3072), (Fig. 3, E and F, supplemental Fig. S2), were obtained from Cell Signaling. The antibody that recognizes total eIF2␣ was previously described (23). Anti-TNFR1 antibody (SC-8436) was obtained from Santa Cruz Biotechnology. All Western blotting was performed with chemiluminescence detection and quantitation of film band intensities was performed with ImageJ (Version 1.31 for Mac OS X, NIH).
Measurement of Translation Rates-MEFs were cultured as described above. After overnight culture, subconfluent cultures were treated with culture medium containing TNF␣, okadaic acid (OA), or poly(rI-C) as described. Cells were washed two times with PBS and incubated in methionine/cysteine-free medium including 200 Ci/ml [ 35 S]methionine/cysteine (Redivue PRO-MIX, Amersham Biosciences) in the continued presence of the described stimulus for 15 min. Cells were washed two times with ice-cold PBS and cell lysates were prepared in Nonidet P-40 lysis buffer containing complete protease inhibitors as described above for immunoblot analysis. Protein concentration was determined by the Bradford method (26). Trichloroacetic acid precipitation was performed by spotting samples on Whatman filter paper with subsequent washing in ice-cold 20% trichloroacetic acid, 10% trichloroacetic acid, and 100% ethanol. Filters were dried and liquid scintillation counting was performed.
Real-time Quantitative RT-PCR-Total RNA was isolated from MEFs at 12 h after cell plating using the TRIzol method (Invitrogen), and RNA was dissolved in diethylpyrocarbonatetreated water containing 1 unit/l RNase inhibitor (Roche). Reverse transcription reactions were performed with i-Script (Bio-Rad) and then diluted 25-fold with water for real-time PCR in an i-Cycler machine using 9 l of diluted reverse transcriptase product and iQ SYBR Green Supermix in a 20-l reaction (Bio-Rad). The amplification primers used for TNFR1 detection were forward (5Ј-CATCCCCAAGCAAGAGTC-ATG-3Ј) and reverse (5Ј-GCTACAGACGTTCACGATGC-3Ј) and the primers used for ␤-actin amplification were forward (5Ј-CCTCTATGCCAACACAGTGC-3Ј) and reverse (5Ј-GTACTT-GCGCTCAGGAGGAG-3Ј).
Cell Survival-Cells were cultured on 10-cm tissue culture dishes. At 24 h after plating, apoptosis was induced by treatment with culture medium containing 100 g/ml poly(rI-C) (Amersham Biosciences) for 16 -18 h or 1 ng/ml TNF␣ (Invitrogen) for 18 -21 h including 50 ng/ml actinomycin D (Act D) (Sigma) during both incubations. Serum deprivation was performed by washing cells three times with serum-free DMEM followed by 14 -23 h of incubation in DMEM containing 0.01% serum. In the morphological studies, cells were cultured on coverslips coated with 1% gelatin, treated as described, and fixed with 10% formalin (Sigma) prior to phase contrast microscopy. Cell viability was quantified by trypan blue dye exclusion. For nuclear staining, cells were plated, treated for induction of apoptosis, and then fixed with cold 70% ethanol at 4°C for 1 h. The cells were then washed with PBS and incubated in ice-cold PBS containing 0.5 mg/ml RNase and 50 g/ml propidium iodide for 15 min in the dark. Samples were mounted onto glass slides with ProLong Gold (Molecular Probes) and viewed using an Olympus BX51 microscope.
Caspase 3 Assay-Adherent and floating cells were washed three times with PBS, collected at 1,200 ϫ g and resuspended in 100 -200 l of 25 mM HEPES, pH 7.5, 5 mM MgCl 2 , 5 mM EDTA, 5 mM dithiothreitol, 2 mM phenylmethylsulfonyl fluoride, 10 g/ml pepstatin A, and 10 g/ml leupeptin. The samples were lysed by four freeze-thaw cycles, centrifuged at 10,000 rpm, and the supernatant was collected for caspase 3 fluorometric assay using 70 -80 g of protein extracts as described by the supplier (Promega CaspACE Fluorometric Assay System, Madison, WI). Addition of the supplied caspase 3 inhibitor peptide to the cell lysates inhibited all activity (supplemental Fig.  S1). A SPECTRAmax Gemini XS spectrofluorometer (Molecu-lar Devices, Sunnyvale, CA) was used with excitation, emission, and cutoff wavelengths of 368, 467, and 420 respectively. Protein concentrations were determined using a detergentcompatible assay (Bio-Rad).
IKK Assay-IKK activity was measured by an immune complex kinase assay as previously described (31,32). Briefly, cell lysates were immunoprecipitated with anti-IKK␣ antibody and the immune complexes used for phosphorylation of a GST-IB␣-(1-54) peptide substrate.

Phosphorylation of eIF2␣ Is Sufficient to Activate Caspase 3-
To elucidate whether PKR activation and/or eIF2␣ phosphorylation are sufficient to activate apoptosis through activation of caspase 3, wild-type, and mutant forms of eIF2␣ and PKR were transiently transfected into HeLa cells in the presence of a procaspase 3 expression vector. Because transiently co-transfected cells express both plasmid DNAs, using this approach it is possible to measure caspase 3 activation in the subpopulation of co-transfected cells that transiently express wild-type or mutant forms of PKR or eIF2␣. Western blot analysis demonstrated that significant levels of procaspase 3 were detected only in cells that received the procaspase 3 expression vector (Fig.  1A, lanes 1-3, 7-9 versus 4 -6). However, the total amount of procaspase 3 was 3-fold lower in cells co-transfected with either wild-type PKR or S51D phospho-mimetic mutant eIF2␣ expression vectors compared with the other transfectants ( Fig.  1A, lanes 2 and 9). This is consistent with cleavage and activation of procaspase 3 or with the translational inhibition observed upon overexpression of wild-type PKR or the S51D mutant eIF2␣ (33). At 48-h post-transfection, caspase 3 activa- tion was measured in cell lysates using a PARP cleavage assay. In the presence of vector alone or in the presence of vectors expressing K296P trans-dominant-negative mutant kinase PKR or S51A non-phosphorylatable mutant eIF2␣, a low level of PARP cleavage was detected (Fig. 1B, lanes 1 and 2 and 10  and 13). In contrast, co-transfection with vectors that express either wild-type PKR or S51D mutant eIF2␣, increased the amount of PARP cleavage 3-5-fold (Fig. 1B, lanes 3 and 12). These results show that expression of either wild-type PKR or a S51D mutant eIF2␣ activates caspase 3 compared with expression of K296P mutant PKR or S51A mutant eIF2␣. Therefore, we conclude that phosphorylation of eIF2␣ is sufficient to activate caspase 3.
Interferon ␣and dsRNA-induced Apoptosis Requires eIF2␣ Phosphorylation-We then tested whether PKR-mediated apoptosis requires eIF2␣ phosphorylation by studying MEFs that harbor a knock-in replacement of Ser 51 for Ala in the endogenous eIF2␣ gene (23). Wild-type (S/S) and homozygous (A/A) eIF2␣ mutant MEFs were treated with interferon ␣ and poly(rI-C) to strongly activate the PKR pathway. Whereas treatment with interferon ␣ and poly(rI-C) increased the level of phosphorylated Ser 51 eIF2␣ in the wild-type S/S MEFs 1.9-fold ( Fig. 2A, lanes 1 and 2), phosphorylated eIF2␣ was not detected in the homozygous A/A mutant MEFs ( Fig. 2A, lanes 3 and 4), consistent with the presence of the homozygous eIF2␣ mutation in these cells. Where treatment of Pkr ϩ/ϩ MEFs with interferon ␣ and poly(rI-C) also increased levels of eIF2␣ phosphorylation 1.4-fold over that observed under control conditions, there was no increase in the Pkr Ϫ/Ϫ MEFs ( Fig. 2A, lanes 5-8).
Western blot analysis confirmed that the Pkr Ϫ/Ϫ MEFs did not express PKR (Fig. 2A). These results support the notion that PKR is the major eIF2␣ kinase activated under these conditions. Morphological analysis of cells treated with poly(rI-C) indicated a distinct difference in survival. Act D was included at a low concentration to prevent the anti-apoptotic response mediated by NF-B activation under these conditions (11). Act D is necessary to elicit an apoptotic response in cultured MEFs. Cycloheximide is also an apoptotic sensitizer that may be used in conjunction with TNF␣ (34). Compared with both wild-type Pkr ϩ/ϩ and wild-type eIF2␣ S/S MEFs that did not survive this treatment, the survival of homozygous eIF2␣A/A mutant MEFs was not compromised (Fig. 2B). Analysis of viability by trypan blue dye exclusion was consistent with the morphological observations (Fig. 2B, legend). These results demonstrate that inactivation of either the eIF2␣ kinase PKR or mutation at the Ser 51 phosphorylation site in eIF2␣ produced substantial resistance to poly(rI-C)-induced death.
To quantitatively monitor a direct marker of apoptosis, caspase 3 activity was measured in prepared cell lysates. Where a 16-h treatment with poly(rI-C) increased caspase 3 activity in wild-type MEFs, the activation of caspase 3 was significantly impaired in MEFs that harbor the homozygous S51A mutant eIF2␣ (A/A) (Fig. 2C). In addition, caspase 3 activation was reduced 50% in the Pkr Ϫ/Ϫ MEFs, consistent with earlier find- Wild-type and mutant MEFs were pretreated with 400 units/ml interferon-␣ overnight and then with poly(rI-C) for 8 h. Cell extracts were prepared for Western blot analysis with anti-phosphopeptide-specific eIF2␣, total anti-eIF2␣, or anti-PKR antibodies. The intensities of eIF2␣ phosphorylation relative to total eIF2␣ levels are indicated. B, Pkr Ϫ/Ϫ and eIF2␣ A/A MEFs are resistant to dsRNA-induced cell death. MEFs were treated with poly(rI-C) (0.1 mg/ml) and Act D (10 ng/ml) for 16 h and then analyzed by light microscopy. Act D treatment alone did not significantly affect cell morphology. Trypan blue dye exclusion indicated that the viable cell counts of treated wild-type Pkr ϩ/ϩ and eIF2␣ S/S MEFs were ϳ20% of control vehicle-treated cultures. In contrast, the viable cell counts were ϳ60 and 45% in the Pkr Ϫ/Ϫ and the eIF2␣ A/A mutant MEFs, respectively. C, procaspase 3 activation is reduced in eIF2␣ A/A and Pkr Ϫ/Ϫ MEFs. MEFs were treated with poly(rI-C) and Act D for 20 h, and then cell extracts were prepared for analysis of caspase 3 activity as described under "Materials and Methods." Caspase 3 activation was significantly reduced in eIF2␣ A/A MEFs and Pkr Ϫ/Ϫ MEFs compared with their respective controls. ***, p Ͻ 0.001. JULY 28, 2006 • VOLUME 281 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 21461
ings that Pkr Ϫ/Ϫ cells are resistant to apoptotic stimulation (9 -12, 18). These data suggest that eIF2␣ phosphorylation contributes to PKR-mediated cell death, although it is not absolutely required.
eIF2␣ Phosphorylation Is Required for TNF␣-induced Apoptosis-To determine the role of eIF2␣ phosphorylation in response to another inducer of apoptosis, the response to TNF␣ was measured. Although previous studies suggest that TNF␣ induces eIF2␣ phosphorylation (9 -11), this has not been directly demonstrated nor has the role of PKR and possibly other eIF2␣ kinases in this process been established. Treatment of subconfluent wild-type MEFs with TNF␣ induced apoptotic bodies in 36% of the cells analyzed by propidium iodide staining (41/271 fragmented apoptotic nuclei, 57/271 picnotic nuclei). In contrast, the same treatment produced no apoptotic nuclei in cells that harbor the homozygous S51A mutation in eIF2␣ (A/A) (Fig.  3A). TNF␣ treatment increased caspase 3 activity 6-fold and 3-fold in wild-type eIF2␣ S/S MEFs and wild-type Pkr ϩ/ϩ MEFs (Fig. 3B). In contrast, TNF␣ treatment did not markedly increase caspase 3 activity in the Pkr Ϫ/Ϫ and homozygous S51A eIF2␣ mutant MEFs (Fig. 3B). The activity measured was specifically caspase 3 as inclusion of the aldehydenoncleavable DEVD peptide fully blocked the activities measured in the lysates. Although the induced caspase 3 activity was primarily dependent upon TNF␣, there was detectable caspase 3 induction upon treatment with Act D alone (supplemental Fig.  S1). In every experiment using TNF␣ as an apoptotic stimulus, caspase 3 activity was significantly increased 2-5-fold in lysates from Pkr ϩ/ϩ cells compared with Pkr Ϫ/Ϫ cells. However, the maximal level of activity in Pkr ϩ/ϩ cells was reproducibly lower than that measured in lysates from TNF␣-treated eIF2␣ S/S cells (Figs. 3B and 4, A-C). The difference may result from the different genetic backgrounds for the two strains of mice. The Pkr ϩ/ϩ and Pkr Ϫ/Ϫ MEFs were derived from C57Bl/6J, and the eIF2␣ S/S and eIF2␣ A/A MEFs were derived from C57Bl/6J X 129/Sv.
To verify the validity of the protective effect of the eIF2␣ mutation, two additional wild-type MEF isolates were derived from independent litters of the eIF2␣ mouse strain (S/S-2 and S/S-3) and analyzed. Both lines displayed remarkably similar degrees of caspase 3 activation upon treatment with TNF␣ compared with the original control MEFs (eIF2␣ S/S, S/S-1) (Fig. 3C). Additionally, an alternative preparation of homozygous eIF2␣ mutant MEFs (A/A-2) was also markedly impaired in caspase 3 activation in response to TNF␣.
To evaluate the requirement for the three additional eIF2␣ kinases in TNF␣ signaling of eIF2␣ phosphorylation and apoptosis, caspase 3 activity was analyzed in Perk Ϫ/Ϫ (29), Hri Ϫ/Ϫ (5), and Gcn2 Ϫ/Ϫ (35) MEFs and their respective wild-type con-   A and B). Act D is not required for induction of caspase 3 activity under these conditions. In comparison to wild-type control MEFs, caspase 3 activation was significantly reduced in Pkr Ϫ/Ϫ , and eIF2␣ A/A MEFs ***, p Ͻ 0.001. C, proteasome inhibition protects from TNF␣-induced apoptosis. Cells were treated same as above (Fig. 3, A and B), except lactacystin (LC) was present during the TNF␣ treatment. Lactacystin treatment decreased caspase 3 activation in wild-type MEFs treated with TNF␣ and Act D. ***, p Ͻ 0.001. trol MEFs. TNF␣ treatment significantly elevated caspase 3 activity in both the wild-type and all knock-out MEFs, indicating that the TNF␣-dependent caspase 3 activity does not require PERK, HRI, or GCN2 eIF2␣ kinases (Fig. 3D).
We next explored the relationship between TNF␣ receptor signaling, PKR activation, and eIF2␣ phosphorylation. We tested whether TNF␣ treatment leads to PKR activation, eIF2␣ phosphorylation, and translational inhibition. Western blot analysis using a phosphopeptide-specific antibody detected an ϳ2.5-fold increase in activated PKR-Thr 451 -P in wild-type MEFs treated with TNF␣for 2 or 4 h (Fig. 3E, lanes 3 and 4). These treatments did not alter the steady state level of PKR (data not shown and supplemental Fig. S2). PKR-Thr 451 -P was not detected in Pkr Ϫ/Ϫ MEFs indicating specificity of the antibody (Fig. 3E, lane 1). In addition, TNF␣ stimulation induced an ϳ2.5-fold increase in eIF2␣-Ser 51 -P (Fig. 3E). The TNF␣mediated increases in PKR-Thr 451 -P and eIF2␣-Ser 51 -P were similar to those observed upon treatment of MEFs with poly(rI-C), which is a very strong stimulus for PKR activation (Fig. 3E,  lane 6). These increases in phospho-PKR and phospho-eIF2␣, were reproducibly detected in independent experiments (supplemental Fig. S2). When TNF␣ stimulation was performed in the presence of the phosphatase inhibitor okadaic acid, slightly larger increases in PKR phosphorylation and eIF2␣ phosphorylation were observed. In contrast, poly(rI-C), and TNF␣ did not increase eIF2␣ phosphorylation in Pkr Ϫ/Ϫ MEFs (Fig. 3F). These results demonstrate that TNF␣ signaling activates PKR and is required to elicit eIF2␣ phosphorylation under these conditions.
Because TNF␣ induces PKR activation and eIF2␣ phosphorylation, we asked whether protein synthesis is inhibited upon TNF␣ treatment. After 4 h, TNF␣ inhibited protein synthesis to ϳ40% in wild-type MEFs, but had little effect in Pkr Ϫ/Ϫ MEFs (Fig. 3G). Treatment with okadaic acid alone or okadaic acid with TNF␣ reduced protein synthesis to ϳ25% in wild-type MEFs, consistent with the increased eIF2␣ phosphorylation observed in the presence of okadaic acid. In contrast, TNF␣ treatment only modestly reduced protein synthesis to ϳ90% in the Pkr Ϫ/Ϫ MEFs, consistent with the reduced level of eIF2␣ phosphorylation measured (Fig. 3F). Overall, the same conditions that elicit eIF2␣ phosphorylation measured by Western blot analysis also inhibit translation in a PKR-dependent manner. Therefore, TNF␣ treatment inhibits protein synthesis through the PKR-eIF2␣ pathway.
TNF␣-induced Apoptosis Requires Protein Synthesis Inhibition-Phosphorylation of eIF2␣ inhibits protein synthesis at the level of initiation. Our results support the hypothesis that phosphorylation of eIF2␣ is required for apoptosis induced by poly(rI-C) and TNF␣. To test the requirement for protein synthesis in the TNF␣ apoptotic response, we measured the effect of protein synthesis elongation inhibition on caspase 3 activation induced by TNF␣. Increasing time of cycloheximide (CHX) treatment in the presence of TNF␣ very rapidly increased caspase 3 activation in the eIF2␣ wildtype and Pkr ϩ/ϩ MEFs (Fig. 4, A and B). In contrast, caspase 3 activation was significantly reduced in the eIF2␣ A/A and Pkr Ϫ/Ϫ mutant MEFs. However, in these mutant MEFs, 6 h of CHX treatment restored activation of caspase 3. Therefore, the protective effect of the S51A mutant eIF2␣ allele or PKR deletion could be partially reversed by general inhibition of protein synthesis.
These results suggest that eIF2␣ phosphorylation may inhibit the translation of a short lived inhibitor of apoptosis, such as p53 (36) or inhibitors of caspase activation (IAPs). Most IAPs contain a C-terminal RING-Zinc finger domain that has ubiquitin ligase (E3) activity and is responsible for their rapid degradation mediated by the proteasome (37). Therefore, inhibition of proteasome activity to prevent p53 and/or IAP degradation may also protect cells from the caspase activation. Indeed, treatment with the proteasome inhibitor lactacystin did partially prevent caspase activation in the wild-type cells (Fig. 4C), as previously described (12).
eIF2␣ Phosphorylation Is Not Required for TNF␣ Signaling and Activation of IKK-Under conditions of ultraviolet light or ER stress, eIF2␣ phosphorylation facilitates activation of NF-B by decreasing translation of IB (38,39). Because PKR is also known to activate IKK in response to dsRNA (40), we determined the requirement for eIF2␣ phosphorylation in this response. Treatment with poly(rI-C), dsRNA, TNF␣, or IL-1, activated IKK to a similar degree in the wild-type and homozygous eIF2␣ mutant A/A MEFs (Fig. 5A). These results demonstrate that eIF2␣ phosphorylation is not required for IKK activation by these stimuli and are consistent with reports that the catalytic activity of PKR is not required to signal NF-B activation and target gene activation (11, 40 -42). In addition, transcriptional activation of a luciferase reporter gene under control of three NF-B binding sites was not altered in the homozygous eIF2␣ A/A mutant MEFs (data not shown). These studies demonstrate that impaired receptor signaling is not the FIGURE 5. eIF2␣ phosphorylation is not required for PKR-mediated activation of IKK. Cells expressing wild-type or designated mutants of PKR or eIF2␣ were assayed for IKK activity after treatment with poly(rI-C), TNF␣, or IL-1 (A). Real-time quantitative RT-PCR for TNFR1 mRNA (B) and Western blot analysis (C ) for TNFR1 and eIF2␣ was performed using lysates from logarithmically growing cells as described under "Materials and Methods." reason eIF2␣ mutant A/A MEFs are resistant to TNF␣-mediated apoptosis.
Previous studies suggested that TNFR1 mRNA is down regulated in cells that express a trans-dominant-negative PKR mutant (15). Indeed, mRNA analysis by real-time quantitative RT-PCR demonstrated that homozygous eIF2␣ A/A and Pkr Ϫ/Ϫ MEFs did express lower levels of TNFR1 mRNA (Fig.  5B). However, Western blot analysis of TNFR1 protein demonstrated similar levels of expressed TNFR1 protein, relative to the loading control eIF2␣ (Fig. 5C). In conclusion, these results demonstrate that although TNFR1 mRNA was reduced in the mutant MEFs, the levels of TNFR1 protein were not altered and that signaling from the TNF␣ receptor to NF-B activation was functional in the mutant MEFs.
PKR-mediated eIF2␣ Phosphorylation Is Required for Serum Deprivation-induced Apoptosis-Because previous studies suggested that serum deprivation induces apoptosis through PKR-mediated phosphorylation of eIF2␣ (10), we analyzed the response to serum deprivation in the wild-type, Pkr Ϫ/Ϫ , and eIF2␣ A/A MEFs. Where serum deprivation induced eIF2␣ phosphorylation by greater than 5-fold in wild-type MEFs (Fig. 6A), significantly less eIF2␣ phosphorylation occurred (1.8-fold) in Pkr Ϫ/Ϫ MEFs. Serum deprivation activated caspase 3 by 5-7-fold in wildtype MEFs. In contrast, serum deprivation did not activate caspase 3 in the Pkr Ϫ/Ϫ or eIF2␣ A/A MEFs (Fig.  6, B and D). Therefore, PKR-mediated phosphorylation is required for apoptosis induced by a different stimulus in the absence of the transcriptional blockade with Act D.

JOURNAL OF BIOLOGICAL CHEMISTRY 21465
We have studied the role of eIF2␣ phosphorylation by analysis of cells that express S51A or S51D mutants of eIF2␣. The following results support the hypothesis that eIF2␣ phosphorylation is alone sufficient to activate apoptosis and in addition, is required for the apoptotic response to PKR activation. First, transient overexpression of wild-type PKR or S51D mutant eIF2␣-induced caspase 3 activation (Fig. 1). Second, caspase 3 activation in cells that harbor a knock-in S51A mutation in eIF2␣ was significantly reduced in response to TNF␣, poly(rI-C), as well as serum deprivation (Figs. 2 and 3). Our studies directly demonstrate that TNF␣ activates PKR to phosphorylate eIF2␣ and inhibit translation. TNF␣-mediated apoptosis required eIF2␣ phosphorylation, whereas TNF␣-dependent activation of IKK did not require eIF2␣ phosphorylation. This is consistent with findings that demonstrate PKR signals activation of IKK in a manner that does not require PKR kinase activity (11, 40 -42). Furthermore, TNF␣-induced eIF2␣ phosphorylation was exclusively dependent on PKR, and not any of the other known eIF2␣ kinases. Finally, treatment with CHX partially restored caspase activation in the S51A eIF2␣ A/A mutant cells, suggesting that eIF2␣ phosphorylation mediates its apoptotic effects through translational inhibition (Fig. 4). These findings support the idea that apoptosis does not require new protein synthesis and that all the machinery required for cell death preexists in the cell. This finding is consistent with a requirement for continued protein synthesis to maintain a pool of short-lived anti-apoptotic factors, such as p53 or IAPs (Fig. 7) (36). The latter was also supported by the protective effect observed by proteasomal inhibition, conditions that should stabilize short-lived protective molecules.
Our results indicate that eIF2␣ phosphorylation is necessary and sufficient for the PKR apoptotic response. These findings are in contrast to conclusions recently derived from observations using an inducible overexpression system to produce S51A and S51D mutants of eIF2␣ (56). These studies did not detect a complete reduction in protein synthesis that would be expected with S51D eIF2␣ expression (33). In addition, although cell number was significantly reduced at 24 h postinduction, apoptosis was not measured at this time. When analyzed at 6 days after induction of the S51D mutant eIF2␣, apoptosis was not detected. These results suggest that the robust apoptotic effect of eIF2␣ phosphorylation may be transient, with subsequent survival of a subpopulation of cells that activate adaptive mechanisms. In contrast, our apoptosis studies using transient DNA transfection of S51D mutant eIF2␣ in HeLa cells were performed at 24-h post-transfection, early after expression of S51D eIF2␣ commenced. Previous studies support the idea that apoptosis through PKR activation is rapid, occurring within 24 h (9, 57). Our studies are consistent with additional findings that support the conclusion that PKR-mediated phosphorylation of eIF2␣ promotes apoptosis. First, overexpression of S51A mutant eIF2␣ protected from vaccinia virus-, TNF␣-, and serum deprivation-induced apoptosis (10,12). In addition, macrophages from S51A eIF2␣ homozygous mutant mice were resistant to apoptosis induced by lipopolysaccharide treatment in the presence of p38 MAPK inhibition (9). This apoptotic response is mediated through toll-like receptor 4 and PKR.
In contrast to the requirement for eIF2␣ phosphorylation for apoptosis mediated through PKR activation, S51A mutation in eIF2␣ or deletion of the eIF2␣ kinase PERK in MEFs dramatically increased sensitivity to agents that disrupt protein folding and produce stress in the ER (23,29). Therefore, it was surprising that the same S51A eIF2␣ A/A mutant MEFs were resistant to apoptotic stimuli that signal through PKR activation. It is unknown how two different stress stimuli that signal through phosphorylation at the same site in eIF2␣ result in opposing responses. We propose that upon ER stress, cells stimulate the death-inducing property of eIF2␣ phosphorylation. However, the outcome is survival because eIF2␣ phosphorylation decreases the protein-folding burden on the ER to relieve the stress. In addition, ER stress induces auxiliary pathways to reverse eIF2␣ phosphorylation so that the eIF2␣ phosphorylation is transient (58). Thus, the ER-stressed cell may benefit from acute reduction of biosynthetic load, while escaping apoptosis in the long term through eIF2␣ dephosphorylation.
We hypothesize that the delicate balance between cell survival and death upon a stress stimuli is determined by the strength of the primary death-inducing stimulus and the input of auxiliary and compensatory pathways that are coordinately activated. Some of these secondary signals may assist or be required for death while others may be protective. We propose that eIF2␣ phosphorylation is fundamentally a death-promoting signal. Our data show that TNF␣-induced eIF2␣ phosphorylation inhibits translation and possibly mediates apoptosis by inhibiting the synthesis of anti-apoptotic cellular factors, such as IAPs (Fig. 7). Under these conditions, the primary death signal to elicit caspase activation is increased. In addition, eIF2␣ phosphorylation also thwarts the adaptive transcriptional FIGURE 7. Phosphorylation of eIF2␣ contributes toward dsRNA-, TNF␣-, and serum deprivation-induced apoptosis. TNFR1 activation initiates apoptosis through caspase 8 leading to caspase 3 activation. TNFR1 occupancy also activates PKR, leading to phosphorylation of eIF2␣. Translation of antiapoptotic factors is inhibited to promote apoptosis. Inhibition of protein synthesis by CHX can complement the requirement for eIF2␣ phosphorylation to promote apoptosis in TNF␣-treated cells. Lactacystin inhibits apoptosis by preventing degradation of anti-apoptotic factors. In parallel, PKR mediates activation of IKK in a manner that does not require kinase catalytic activity. response through translational inhibition to prevent synthesis of protective factors.
Given the observations of the importance of PKR and eIF2␣ phosphorylation in apoptosis, it was interesting and curious that homozygous mutation of S51A in eIF2␣, Pkr deletion, or expression of a trans-dominant-negative mutant PKR did not have an obvious developmental phenotype in the mouse (23,28,59,60). These findings would support the idea that eIF2␣ phosphorylation is not an essential apoptotic signal in mammalian embryonic development, a process where apoptosis plays a central role. This is in contrast to dramatic phenotypes observed in mice harboring deletions in essential caspase genes or key modulators of apoptosis (61,62). However, it is consistent with absence of embryonic lethality in mice lacking the known TNF␣ receptor family members (63). Thus, death receptor signaling has a lesser role in development than essential apoptosis effectors.
Although TNF␣ receptor signaling is not required for embryonic apoptosis, there are a number of circumstances where TNF␣-induced apoptosis is physiologically important including lipopolysaccharide-mediated apoptosis in the liver (64), hepatotoxicant-induced apoptosis (65), suppression of acute HSV-1 viral infection (66), limitation of T cell number during chronic LCMV viral infection (67), infarction induced myocardial rupture and ventricular dysfunction (68), and death of malformed embryos (69). Adenovirus delivery of TNF␣ induced apoptosis in esophageal cancer cells in a manner that required PKR (70), suggesting the utility of this approach to promote apoptosis in transformed cells. The delineation of TNF␣ signaling to eIF2␣ phosphorylation and apoptosis established by our studies suggests eIF2␣ phosphorylation may be an important death signal in these physiologically important apoptotic events. Conditional homozygous eIF2␣ A/A mice with tissue specific expression will provide important tools for future studies on TNF␣-induced apoptosis.
Regulation of eIF2␣ phosphorylation could provide an attractive target for therapeutic intervention. Agents that inhibit eIF2␣ phosphorylation could promote survival under desired conditions, for example to inhibit macrophage apoptosis upon viral infection (9) or prevent ischemic cell injury (71). Alternatively, direct targeting of therapeutic agents to induce eIF2␣ phosphorylation may accentuate apoptosis of virus-infected cells (72). Future analysis of potential therapeutics to increase eIF2␣ phosphorylation likely will lead to death promoting applications in anti-tumor or anti-microbial targeted therapeutics as well as to protective functions in ER stressrelated disease.