Phosphorylation of Eukaryotic Translation Initiation Factor 2 Mediates Apoptosis in Response to Activation of the Double-stranded RNA-dependent Protein Kinase*

The interferon-inducible, double-stranded (ds) RNA-dependent serine/threonine protein kinase (PKR) plays a role in viral pathogenesis, cell growth, and differentiation and is implicated as a tumor suppressor gene. Expression of atrans-dominant negative, catalytically inactive mutant PKR protected NIH3T3 cells from apoptosis in response to either treatment with tumor necrosis factor α (TNFα), serum deprivation. In cells expressing mutant PKR, TNFα, but not dsRNA induced transcription from a nuclear factor κ B-dependent promoter, demonstrating specificity for dsRNA in signaling through the PKR pathway. Serum or platelet-derived growth factor addition to serum-deprived mutant PKR-expressing cells induced transcription of the early response genes c-fos and c-jun, indicating that the immediate early response signaling was intact. Overexpression of wild-type PKR in a transient DNA transfection system was sufficient to induce apoptosis. TNFα-induced apoptosis correlated with increased phosphorylation of the α subunit of eukaryotic translation initiation factor 2 (eIF-2α), the primary physiological substrate of the PKR. Furthermore, forced expression of a nonphosphorylatable S51A mutant eIF-2α partially protected cells from TNFα-induced apoptosis, and expression of a S51D mutant eIF-2α, a mutant that mimics phosphorylated eIF-2α, was sufficient to induce apoptosis. Taken together, these studies identify a novel requirement for PKR in stress-induced apoptosis that is mediated through eIF-2α phosphorylation.

The interferon-inducible, double-stranded (ds) 1 RNAdependent serine/threonine protein kinase (PKR) is a ubiquitously expressed protein in mammalian cells that was first identified as a mediator of the antiproliferative and antiviral actions of interferon (1,2). A variety of dsRNA molecules generated during viral infection, specific single-stranded RNAs, and various agents, such as heparin, activate PKR from its latent form. Upon activation, PKR displays two known activities: autophosphorylation and phosphorylation of its physiological substrate, the ␣ subunit of the heterotrimeric eukaryotic translational initiation factor 2 (eIF-2␣). eIF-2␣ phosphorylation impairs its ability to recycle and thus inhibits protein synthesis at the level of initiation (3,4). As a consequence, activated PKR inhibits viral replication (5,6) and cell growth (7,8) and may promote cell differentiation (9,10). In addition, expression of catalytically inactive mutants of PKR transforms NIH3T3 cells to induce tumors in nude mice. This observation is ascribed to a trans-dominant inhibitory effect of the mutant enzyme on the endogenous wild-type PKR and implicates PKR as a tumor suppressor gene (11)(12)(13)(14). PKR is also implicated in the transcriptional regulation of dsRNA-activated genes, such as interferon ␤, by activation of nuclear transcription factor B (NFB) by phosphorylating its inhibitor, IB, in response to dsRNA (15)(16)(17).
Polypeptide chain synthesis is initiated when the ternary complex of heterotrimeric eIF-2, GTP, and initiator met-tRNA bind the 40S ribosomal subunit to generate a 43S preinitiation complex (for reviews, see Refs. 3 and 4). Subsequently, mRNA binds and the 60S ribosomal subunit joins to form the 80S initiation complex with the concomitant hydrolysis of GTP to GDP. For eIF-2 to promote another round of initiation, GDP bound to eIF-2 must be exchanged for GTP, a reaction catalyzed by the guanine nucleotide exchange factor (eIF-2B). Phosphorylation of the ␣ subunit of eIF-2 (eIF-2␣) stabilizes the eIF-2⅐GDP⅐eIF-2B complex. Because eIF-2B is obligatory for the exchange of GTP for bound GDP and because eIF-2B exists in cells in relatively low molar quantities with respect to eIF-2, the exchange process can be inhibited when only a fraction (i.e. 10 -20%) of eIF-2␣ is phosphorylated (18). Thus, phosphorylation of eIF-2␣ is part of a key regulatory process that is proposed to result in the quantitative sequestering of eIF-2B that can virtually shut down protein synthesis initiation.
TNF␣ is an inflammatory cytokine that, upon binding to its receptors TNFR1 and TNFR2, initiates signaling pathways that play important roles in cell activation, differentiation, and apoptosis, a physiologically controlled process important in development and pathogenesis of a variety of human diseases (for review, see Ref. 19). Upon binding of TNF␣, TNFR1 becomes trimerized and recruits the adapter protein, TNF receptor-associated death domain, which subsequently recruits Fasassociating protein with death domain to activate the interleukin-1 converting enzyme family of cysteine proteases that induce an apoptotic signal. TNF␣ provides an antiviral function for the host by targeting virus-infected cells for apoptotic cell death (20 -22). In addition, upon trimerization of TNFR1, TNFR-associated factor is recruited to lead to activation of NFB through phosphorylation of its inhibitor, IB. Activation of NFB induces the expression of survival genes. Because PKR is implicated in dsRNA-activation of NFB through phosphorylation of its inhibitor, IB (16,17), and because PKR induces apoptosis in the context of a vaccinia virus infection (6), we studied the ability of PKR to induce programmed cell death in response to TNF␣ by analyzing fibroblasts expressing a trans-dominant negative K296P mutant PKR to inhibit the PKR signaling pathway(s). Our results indicate that PKR can induce apoptosis through phosphorylation of eIF-2␣.

MATERIALS AND METHODS
Cell Culture-NIH3T3 cell lines that were transfected and selected for expression of the catalytically inactive K296P mutant PKR were previously described (23). Cells were co-transfected with the pEDmtx r VA Ϫ empty vector and pSV 2 Neo, selected for G418 resistance, pooled, and used as a control. NIH3T3 cells transfected and selected for wild-type and S51A mutant eIF-2␣ expression were previously described (24). Cells were cultured at 37°C with 10% CO 2 in Dulbecco's modified essential medium containing 10% fetal bovine serum (FBS) and penicillin/streptomycin in 100-mm tissue culture dishes (for cell cycle and DNA fragmentation analysis) or on coverslips (for fluorescence microscopy). Cells were incubated in the presence or absence of TNF␣ or in the presence of 10% FBS or 0.1% FBS for times indicated in the figure legends. Where indicated, cells were treated with interferon ␣ (1000 units/ml; Life Technologies, Inc.) for 20 h and then treated with poly(I)⅐poly(C) (100 g/ml; Pharmacia Biotech Inc.) for 6 h prior to analysis by phase contrast microscopy.
DNA Fragmentation Analysis-Cells were collected and harvested in 0.5 ml of lysis buffer (10 mM Tris-Cl, pH 7.4, 1 mM EDTA, 400 mM NaCl, 1% SDS). Lysates were incubated at 50°C for 2 h with 0.2 mg/ml proteinase K. After extraction with an equal volume of phenol:chloroform:isoamyl alcohol (25:24:1) and reextraction with chloroform: isoamyl alcohol, the DNA was precipitated by addition of 1 volume of isopropanol at Ϫ70°C overnight. Precipitated DNA was collected by centrifugation at 14,000 ϫ g for 30 min, washed with ice-cold 70% ethanol, resuspended in Tris-EDTA buffer (pH 8.0), and treated with RNase (DNase free, 20 g/ml; Boehringer Mannheim). DNA samples were fractionated by electrophoresis on a 1.6% agarose gel and visualized by ethidium bromide staining. Band intensities of DNA fragments (size ranging between 200 and 600 base pairs) were quantified using the National Institutes of Health Image 1.55b program.
Transient DNA Transfection and Analysis-NIH3T3 cell lines were co-transfected with the NFB-chloramphenicol acetyltransferase (CAT) expression plasmid (4xB-CAT; provided by Dr. Gary Nabel, Howard Hughes Medical Institute, Ann Arbor, MI; 25) and a Rous sarcoma virus promoter-driven luciferase construct (to measure transfection efficiency) by the DEAE-dextran method (26). At 48 h posttransfection, cells were treated either with poly(I)⅐poly(C) (100 g/ml; Pharmacia) or TNF␣ (10 ng/ml; Life Technologies, Inc.) for 20 h at 37°C. Cell extracts were harvested by freeze-thawing, and CAT and luciferase activities were measured essentially as described in the Promega Protocol and Applications Guide (Madison, WI).
COS-1 cells were transfected with expression plasmids encoding either wild-type PKR, K296P mutant PKR, the dsRNA binding domain of PKR (amino acid residues 1-243), wild-type eIF-2␣, S51A eIF-2␣, or S51D eIF-2␣ contained in the pMT 2 VA Ϫ expression vector (27,28). The Bcl-2 expression plasmid was kindly provided by Dr. S. J. Korsmeyer (Howard Huges Medical Institute, St. Louis, MO). Transfections were performed in the presence of an equal amount of DNA encoding the jellyfish green fluorescence protein (GFP) (expression vector psynGFPS65T (29)) to identify transfected cells. At 24 h posttransfection, cells were fixed and nuclei were stained with Hoechst 33258 (Molecular Probes, Eugene, OR). Stained cells were washed in cold phosphate-buffered saline, mounted on slides, and visualized using an Olympus BX60 fluorescence microscope.
Cell Cycle Analysis-Cells were incubated for the indicated times in the presence or absence of TNF␣ or in the presence of 10% FBS or 0.1% FBS. Cells were then washed with cold phosphate-buffered saline containing 0.1% bovine serum albumin and fixed in 70% ethanol in phosphate-buffered saline at 4°C overnight. Cells were rehydrated in phosphate-buffered saline, incubated with RNase (20 g/ml) for 30 min at 37°C, and stained with propidium iodide (50 g/ml) for 30 min. DNA content was analyzed using a FACStar Flow Cytometer (Immunocytometry System, Becton Dickinson, Mountain View, CA). Cells were analyzed for scatter-gated fluorescence to prevent scoring of dead cells and debris. Green fluorescence was measured using a 530/30-nm bandpass filter, and red fluorescence was measured using a 625/35-nm bandpass filter.
eIF-2␣ Phosphorylation-eIF-2␣ phosphorylation was measured by [ 32 P]phosphate incorporation as described (23) or by Western blot anal-ysis. Steady state levels of eIF-2␣ and phosphorylated eIF-2␣ were measured in cell extracts by SDS-polyacrylamide gel electrophoresis under reducing conditions and electroblotting to nitrocellulose. Filters were sequentially incubated and developed with rabbit anti-phosphopeptide antibody specific to phosphorylated eIF-2␣ (kindly provided by Gary Krause, Wayne State University) and then with anti-eIF-2␣ monoclonal antibody (provided by Dr. Henshaw). After incubation with antibody, the filters were washed and developed using the ECL kit (Amersham Corp.).
RNA Analysis-Total cellular RNA was isolated using TRIAzol (Life Technologies, Inc.). Total RNA samples (10 g) were analyzed by RNA Northern blot hybridization following electrophoresis on formaldehydeformamide 1% agarose gels (30,31). RNA was transferred to nitrocellulose membranes and baked at 80°C under vacuum. The blots were prehybridized, hybridized to either 32 P-labeled c-fos or 32 P-labeled c-jun cDNA (provided by Dr. Tom Kerpolla, Howard Hughes Medical Institute, Ann Arbor, MI) probes prepared by [␣-32 P]dCTP labeling using the Rediprime DNA labeling system (Amersham Corp.) and washed two times with 0.1ϫ SSC (1ϫ ϭ 150 mM sodium chloride, 15 mM sodium citrate) for 15 min. As a control for loading equal amounts of RNA, the blots were treated with 0.1% sodium dodecyl sulfate at 95°C and rehybridized to a 32 P-labeled ␤-actin probe. Band intensities were quantified using the National Institutes of Health Image 1.55b program.
Total RNA (150 ng) was reverse transcribed using oligo(dT) primer and superscript II reverse transcriptase (Life Technologies, Inc.) in a 20-l reaction volume essentially as described by the supplier. PCR was performed using 2 l from each cDNA reaction with primer sets for detecting Fas mRNA (5Ј-CATCTCCGAGAGTTTAAAGCTGAGG and 5Ј-GTTTCCTGCAGTTTGTATTGCTGGTTCC) or glyceraldehyde-3phosphate dehydrogenase mRNA (5Ј-CCATGGAGAAGGCTGGGG and 5Ј-CAAAGTTGTCATGGATGACC). Control experiments demonstrated that the PCR was performed in the linear range to detect Fas and glyceraldehyde-3-phosphate dehydrogenase mRNAs.

A Functional PKR Pathway Is Required for TNF␣-and Serum Deprivation-induced Apoptosis-
The requirement for a functional PKR pathway in apoptosis was studied in NIH3T3 cells stably transfected with empty vector alone or in NIH3T3 cells expressing mutant K296P PKR. The expression of K296P mutant PKR in these cells was previously characterized (23). TNF␣-and serum deprivation-induced apoptosis were studied by TNF␣ treatment or growth in 0.1% FBS. Both of these conditions induced characteristic ladders of DNA fragments in NIH3T3 cells stably transfected with empty vector alone that were not observed in a cells that overexpress mutant PKR (Fig.  1A). Similar results were obtained with three additional independently isolated clones expressing K296P PKR (data not shown), implying a requirement for a functional PKR pathway in both of these apoptotic responses. The resistance to TNF␣-or serum deprivation-induced apoptosis in mutant PKR-expressing cells was further confirmed by quantitating the percentage of apoptotic cells by flow cytometry. NIH3T3 cells transfected with vector alone contained a range of 2-4 N DNA representing cells in the G 0 /G 1 , S, and G 2 /M phases of the cell cycle. Following TNF␣ treatment or serum deprivation, the number of cells in the G 0 /G 1 phase of cell cycle decreased in parallel to the appearance of a peak of cells containing hypodiploid DNA (Fig.  1B). In contrast, cells that express mutant PKR showed a reduced number of cells containing hypodiploid DNA following TNF␣ (6%) or serum deprivation (8%). These results support the involvement of PKR in TNF␣-and serum deprivationinduced apoptosis.
Recently, dsRNA (poly(I)⅐poly(C)) was shown to induce apoptosis in a PKR-dependent manner using mouse embryo fibroblasts from PKR wild-type and PKR knockout mice (32). Therefore, we studied the requirement for PKR in apoptosis mediated by interferon and poly(I)⅐poly(C) treatment in NIH3T3 cells. NIH3T3 cells stably transfected with empty vector or NIH3T3 cells transfected with K296P mutant PKR were treated with interferon, poly(I)⅐poly(C), or poly(I)⅐poly(C) in combination with interferon. Under these experimental con-ditions, interferon or poly(I)⅐poly(C) alone did not induce significant morphological changes in either cell line (Fig. 2). However, poly(I)⅐poly(C) in combination with interferon induced significant cell death in NIH3T3 cells stably transfected with empty vector, whereas NIH3T3 cells that were overexpressing mutant PKR were protected under identical conditions. It is most likely that poly(I)⅐poly(C) alone did not affect NIH3T3 cells due to their low levels of PKR. These results are in agreement with previous observations (32,33) and support the idea that dsRNA is toxic to NIH3T3 cells that have a functional PKR pathway.
Signaling by dsRNA and not by TNF␣ or Platelet-derived Growth Factor (PDGF) Is Selectively Abrogated in Cells Expressing Mutant PKR-Whether PKR function was abrogated in the cells expressing the trans-dominant negative mutant PKR was tested by analyzing the dsRNA-dependent activation of NFB because previous studies support the idea that PKR is required in this signaling (16,34). Control and mutant PKR- eIF-2␣ Phosphorylation Mediates PKR-dependent Apoptosis expressing NIH3T3 cells were transiently transfected with an NFB-specific CAT reporter plasmid (25). In control cells, CAT activity was induced 5-fold in an NFB-dependent manner in response to poly(I)⅐poly(C). Mutation of the NFB binding sites within the vector prevented the induction (Fig. 3). Cells expressing the trans-dominant negative mutant PKR did not respond to poly(I)⅐poly(C), and in addition, the basal transcription of the NFB-dependent promoter in these cells was reduced compared with control NIH3T3 cells (Fig. 3). In this experiment, the transfection efficiency was monitored by cotransfection with a luciferase expression vector. The amount of luciferase activity varied less than 10% between different plates of transfected cells. The defective dsRNA response was specific, because both control and the trans-dominant negative mutant PKR-expressing cells responded to TNF␣ as indicated by induction of NFB-dependent CAT activity by 29-and 27fold, respectively (Fig. 3). The greater response of these cells to TNF␣ compared with poly(I)⅐poly(C) may reflect the inefficient delivery of poly(I)⅐poly(C) to the cells or differences in the general responsiveness of these two different signal transduction pathways. These results are consistent with a specific abrogation of PKR signaling by expression of mutant K296P PKR.
Results of recent studies suggest that PKR-mediated Fas mRNA induction may be one mechanism for PKR-mediated cell death in response to a variety of stresses (32,35). We therefore measured Fas mRNA levels in response to TNF␣, poly(I)⅐ poly(C), and interferon with poly(I)⅐poly(C) using reverse transcription-PCR with Fas mRNA-specific primers. After treatment with any of these agents, an increase in Fas mRNA level either in NIH3T3 cells transfected with vector or cells transfected with mutant PKR was not detected (data not shown). The results demonstrated that apoptosis in response to either TNF␣ or dsRNA with interferon is not secondary to an induction of Fas mRNA.
Previous studies suggested that induction of the early re-sponse genes requires a functional PKR pathway (15,36). For example, 2-aminopurine, an adenosine analog that is known to inhibit PKR, was shown to inhibit serum induction of early response genes (15). To elucidate whether the early response gene transcriptional induction is mediated through PKR, we studied induction of c-fos and c-jun in response to serum or PDGF addition to growth-arrested, serum-depleted control NIH3T3 cells or NIH3T3 cells expressing K296P mutant PKR. Analysis of total RNA by Northern blot hybridization demonstrated that c-fos and c-jun mRNAs were induced in response to the addition of either serum or PDGF to serum-starved control NIH3T3 cells (Fig. 4, compare lanes 1, 3, and 5), as well as in serum-starved NIH3T3 cells expressing K296P mutant PKR (Fig. 4, lanes 7, 9, and 11). Importantly, addition of 2-aminopurine inhibited c-fos and c-jun induction by serum and PDGF in both control (Fig. 4, lanes 2, 4, and 6) and mutant PKR-expressing cells (Fig. 4, lanes 8, 10, and 12). These results indicate that a functional PKR pathway is not required for the immediate early response and that 2-aminopurine inhibition of the immediate early response is likely a consequence of inhibiting a PKR-independent pathway. The observation that PKR is not required for the immediate early response is in contrast to the findings of Mundschau and Faller (36), who used antisense oligonucleotides to suppress PKR levels in BALB/c 3T3 cells to show a defect in the serum/PDGF-mediated transcriptional induction of c-fos and c-myc. Further studies are required to determine whether the observed differences in the apparent requirement for PKR in the immediate early response are due to the different cell culture systems or due to the different methods utilized to inhibit the PKR pathway. Our results, showing that c-fos and c-jun transcriptional induction were not inhibited by mutant PKR expression, demonstrate that apoptotic response and not the serum growth response requires a functional PKR pathway. This also indicates that protection from apoptosis in these mutant PKR-expressing cells is specific and not due to general dysfunction of signal transduction pathway(s).

PKR-dependent Phosphorylation of eIF-2␣ Is Necessary and Sufficient to Mediate
Apoptosis-To elucidate whether TNF␣induced apoptosis is mediated through phosphorylation of eIF-2␣ by PKR, in vivo analysis of eIF-2␣ phosphorylation was performed in NIH3T3 cells that overexpress wild-type eIF-2␣ or S51A mutant eIF-2␣, the site of PKR phosphorylation (27).

FIG. 2. Overexpression of K296P mutant PKR protects cells from death induced by poly(I)⅐poly(C) and interferon.
NIH3T3 cells stably transfected with empty vector or stably transfected with K296P mutant PKR were treated with interferon or with poly(I)⅐poly(C) or poly(I)⅐poly(C) in combination with interferon as described under "Materials and Methods." Following treatment, micrographs were taken at ϫ 200 magnification. The expression of wild-type and S51A mutant eIF-2␣ in these NIH3T3 cells was previously characterized (24). Immunoprecipitation of eIF-2␣ from [ 32 P]phosphate-labeled NIH3T3 cells that overexpress wild-type eIF-2␣ detected [ 32 P]phosphate incorporation into eIF-2␣ (Fig. 5A, lane 2). Treatment with TNF␣ increased the [ 32 P]phosphate incorporation into wild-type eIF-2␣ by 2.5 fold (Fig. 5A, lane 3). In contrast, cells that express the S51A mutant eIF-2␣ did not exhibit increased [ 32 P]phosphate incorporation in response to TNF␣ (Fig. 5A,  lanes 4 and 5). Cells treated with TNF␣ showed increased background labeling, as is evident by the use of an irrelevant monoclonal antibody (Fig. 5A, lanes 6 and 7). The increase in eIF-2␣ phosphorylation upon TNF␣ treatment was also demonstrated by Western blot analysis using antibodies against phosphorylated and nonphosphorylated eIF-2␣P (Fig. 5B). Previous detailed characterization of the anti-eIF-2␣ antibody demonstrated that it specifically reacts with phosphorylated eIF-2␣ and not with nonphosphorylated eIF-2␣ (37). Whereas TNF␣ treatment increased the relative amount of phosphorylated eIF-2␣ by approximately 2-fold in cells expressing wildtype eIF-2␣, there was no increase in cells expressing the S51A mutant eIF-2␣. These results support the idea that TNF␣ induces eIF-2␣ phosphorylation.
The requirement for eIF-2␣ phosphorylation for TNF␣-induced apoptosis was analyzed by TNF␣ treatment of NIH3T3 cells expressing either wild-type or S51A mutant eIF-2␣. Flow cytometry analysis of DNA content (Fig. 6A), as well as the DNA fragmentation analysis (Fig. 6B), demonstrated that expression of the phosphorylation-resistant S51A mutant eIF-2␣ partially protected cells from apoptosis upon TNF␣ treatment. We expect that the incomplete protection may be due to the presence of endogenous wild-type eIF-2␣, which may become phosphorylated and act in a dominant manner to inhibit protein synthesis.
To test the hypothesis that activation of PKR is sufficient to induce apoptosis, a wild-type PKR expression vector was intro-duced into cells by transient DNA transfection. In these studies, COS-1 monkey cells were co-transfected with an expression vector encoding the GFP (26) to identify the transfected cells. Co-transfection with wild-type PKR reduced expression of GFP, as indicated by the reduced fluorescence emitted by the co-transfected cells compared with control vector co-transfected cells (Fig. 7A, pETF.VA Ϫ versus PKRwt). This is consistent with an inhibition of protein synthesis mediated by PKR overexpression. Transfection of the wild-type PKR expression vector, but not that of expression vector alone, induced apoptosis in the majority of cells that received GFP as detected by nuclear fragmentation (Fig. 7A). To test whether phosphorylation of eIF-2␣ itself is sufficient to induce apoptosis, COS-1 monkey kidney cells were co-transfected with an expression vector encoding GFP and S51D mutant eIF-2␣, a mutant that mimics phosphorylated eIF-2␣ (27), into cells by transient DNA transfection. Transfection of S51D mutant eIF-2␣ reproducibly induced apoptosis in cells that received GFP compared with cells transfected with the vector alone ( Fig. 7A; pETF.VA Ϫ versus eIF2␣ S51D) or cells transfected with an expression vector encoding wild-type eIF-2␣ (data not shown). In addition, flow cytometric analysis of S51D mutant eIF-2␣ transfected cells demonstrated a higher percentage of hypodiploid cells (55%) compared with the cells transfected with the vector control (33%) (data not shown). Apoptosis induced by overexpression of wild-type PKR or S51D mutant eIF-2␣ was also confirmed by DNA fragmentation analysis (Fig. 7B). Expression of either wild-type PKR or S51D mutant eIF-2␣ increased the amount of degraded DNA by 3-fold compared with expression of K296P mutant PKR, S51A mutant eIF-2␣, the PKR dsRNA binding domain (amino acid residues 1-243), or Bcl-2. Under the transfection conditions, approximately 30% of the cells were transfected, and this may account for the small increase in the amount of degraded DNA. These results indicate that overexpression of either wild-type PKR or S51D mutant eIF-2␣ is sufficient to induce apoptosis. FIG. 4. Induction of early response genes is not inhibited by mutant K296P PKR. NIH3T3 cells stably transfected with empty vector or cells transfected with K296P mutant PKR were grown to confluency, and the medium was changed to 0.5% serum. After 48 h, cells were fed with medium containing 10% serum or 0.5% serum in the absence or presence of PDGF for 1 h. For 2-aminopurine experiments, 10 mM 2-aminopurine was added to the medium at the same time. Total cellular RNA was isolated and analyzed by Northern blot hybridization using 32 P-labeled c-fos or c-jun probes. As a control for loading equal amounts of RNA, the blots were treated with 0.1% SDS at 95°C and rehybridized to a 32 P-labeled ␤-actin probe. The fold induction in response to serum or PDGF was measured by densitometry analysis. In NIH3T3 cells transfected with empty vector alone, c-fos and c-jun mRNAs were induced 10-and 4-fold in response to serum and 12-and 6-fold in response to PDGF, respectively. In NIH3T3 cells expressing mutant PKR, c-fos and c-jun mRNAs were induced 6-and 8-fold in response to serum and 8-and 10-fold in response to PDGF, respectively.
FIG. 5. Expression of S51A mutant eIF-2␣ prevents TNF␣-induced eIF-2␣ phosphorylation. NIH3T3 cells that overexpress wildtype eIF-2␣ or S51A mutant eIF-2␣ were treated with TNF␣ (10 ng/ml for 20 h) and labeled with [ 32 P]orthophosphoric acid for 4 h as described previously (23). Cell extracts were prepared and immunoprecipitated with anti-eIF-2␣ monoclonal antibody and analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography (A). In parallel, cells were harvested and analyzed by SDS-polyacrylamide gel electrophoresis and by Western blot analysis using antibody that reacts with total eIF-2␣ or antibody that reacts with eIF-2␣-P (B). The fold induction of eIF-2␣ phosphorylation in response to TNF␣ was measured by densitometry analysis and is presented as a relative ratio of phosphorylated eIF-2␣ to nonphosphorylated eIF-2␣ (*).

DISCUSSION
PKR was first identified as a mediator of the antiproliferative and antiviral action of interferon (1,38). The antiproliferative action of PKR was previously demonstrated through expression of wild-type and catalytically inactive mutant PKR. Overexpression of wild-type PKR inhibited protein synthesis and cell growth (7,8), whereas overexpression of a catalytically inactive mutant PKR transformed NIH3T3 cells (11)(12)(13). The mechanism for mutant PKR-induced transformation is likely either through formation of inactive heterodimers with endogenous PKR or through competition for endogenous PKR activators. Both of these mechanisms invoke disruption of the proper regulation of endogenous PKR activity. Using this trans-dominant negative property of the catalytically inactive PKR, we have shown that a functional PKR pathway is re-quired to induce apoptosis in response to serum deprivation or TNF␣ treatment. Our results confirm and extend recent observations that implicate PKR as a general transducer of the apoptotic response (6,32,33,35). Although it was appreciated for many years that dsRNA in the presence of interferon is toxic to cells, it was only recently suggested that PKR mediates the apoptotic response to dsRNA upon vaccinia virus (6,35,39) and influenza virus (40) infection. Recent observations also support the idea that TNF␣-induced apoptosis requires a functional PKR pathway, whereas either expression of antisense PKR mRNA or deletion of the PKR gene reduced the apoptotic response to TNF␣ (32,33). We have demonstrated that apoptotic responses to TNF␣, serum deprivation, and interferon treatment with poly(I)⅐poly(C) require a functional PKR pathway by expression of a trans-dominant negative mutant of PKR. In eIF-2␣ Phosphorylation Mediates PKR-dependent Apoptosis addition, forced expression of wild-type PKR was sufficient to induce apoptosis; this is also consistent with previous observations (33). PKR activation was also previously demonstrated to occur in response to stress conditions, such as activation of the heat shock response, presence of unfolded protein in the endoplasmic reticulum, growth factor depletion, and increases in cytosolic calcium (23,(41)(42)(43). All of these stress responses are frequently accompanied by apoptosis. These results support the idea that PKR is a general transducer of apoptosis in response to a variety of different stimuli. It seems probable that the tumor suppressor activity previously attributed to PKR (11,12) is mediated through its proapoptotic function.
Two fundamental questions remain regarding the mechanism of PKR-induced apoptosis: 1) what mechanism(s) activates PKR in the absence of a viral infection? and 2) what are the immediate substrates downstream of PKR that lead to apoptosis? Although it is most likely that PKR activation upon viral infection occurs through expression of dsRNA from the viral genome, it is less obvious what activates PKR in the absence of a viral infection, and numerous possibilities exist. Ligand binding-induced trimerization of the TNF␣ receptor may elicit activation of an upstream protein kinase, such as the TNF␣ receptor-interacting protein RIP (44), that phosphorylates and activates PKR. Alternatively, activation of an interleukin-1 converting enzyme-like protease in response to TNF␣ or other stress inducers may cleave the dsRNA binding domain from PKR to generate a kinase domain fragment that is known to act as a constitutively active kinase (28). It is also possible that induced synthesis of a cellular dsRNA molecule or a change in secondary structure of a preexisting RNA molecule may elicit PKR activation. Previous studies demonstrated that PKR activation in response to increased cytosolic calcium required the dsRNA binding activity of PKR (23). The requirement for the dsRNA binding activity of PKR suggests that increased cytosolic calcium activates PKR through an RNA binding-dependent mechanism. Alternatively, the activation of PKR may be modulated by cellular inhibitors (45)(46)(47) that are regulated by external stimuli. For example, it was recently demonstrated that the ubiquitously expressed PKR inhibitor identified as p58 is inhibited by hsp40 (48), a molecular chaperone that interacts with the translational machinery and may play a role in nascent chain polypeptide folding (49).
The two best characterized substrates that mediate the downstream effects of PKR are eIF-2␣ and IB. Because TNF␣mediated activation of NFB elicits transcriptional induction of antiapoptotic genes (50 -52) and because PKR-deleted mouse embryo fibroblasts exhibit normal TNF␣ activation of NFB but are defective in their TNF␣ apoptotic response (17,32), a direct role for activation of NFB would seem unlikely. In these latter studies, the PKR-dependent apoptotic response to TNF␣ could be attributed to Fas mRNA induction, possibly through the transcription factor interferon regulatory factor 1. In contrast, the TNF␣ apoptotic response that we have studied in NIH3T3 cells did not induce Fas mRNA. However, it has been shown that PKR expression is induced by TNF␣ (33), and if TNF␣ is activated, this may increase the level of phosphorylated eIF-2␣. We have provided four observations that support the idea that eIF-2␣ is the downstream target for PKR-induced apoptosis. First, TNF␣ increased the phosphorylation status of eIF-2␣ in NIH3T3 cells. Second, expression of a trans-dominant negative mutant PKR prevented the TNF␣-mediated increase in eIF-2␣ phosphorylation, as well as the apoptotic response. Third, cells that express a nonphosphorylatable S51A mutant of eIF-2␣ were partially protected from TNF␣-induced apoptosis. It is possible that only partial protection was observed because these cells also express endogenous wild-type eIF-2␣ that could be phosphorylated and inhibit protein synthesis in a dominant manner. Alternatively, other mechanisms may also be involved in addition to phosphorylation of eIF-2␣. Fourth, expression of a S51D mutant eIF-2␣, which mimics a phosphorylated serine, was sufficient to induce apoptosis. These observations support a role for eIF-2␣ phosphorylation in the proapoptotic response and suggest a role for translation initiation in apoptosis. A requirement for inhibition of new protein synthesis is frequently observed for TNF␣-induced apoptosis (53), and inhibition of protein synthesis itself is sufficient to induce apoptosis in some systems (54,55). Upon TNF␣ treatment, it is possible that PKR-mediated phosphorylation of eIF-2␣ and subsequent translational inhibition counteracts the protective effect of NFB-dependent induced transcription of genes encoding proteins that prevent apoptosis.
Previous studies also suggest that eIF-2␣ is the downstream target for PKR-mediated apoptosis. PKR mediates phosphorylation of eIF-2␣ in response to calcium depletion from the endoplasmic reticulum mediated by ionophore (23,41), a treatment used to induce apoptosis. Furthermore, removal of the growth factor interleukin 3 from an interleukin 3-dependent cell line induces the autophosphorylation of PKR, which in turn FIG. 7. Overexpression of wild-type PKR or S51D mutant eIF-2␣ induces apoptosis. COS-1 monkey kidney cells were co-transfected with expression vectors encoding the indicated products with an expression vector encoding the jellyfish GFP to identify transfected cells. After 48 h, nuclei were stained with Hoechst 33258 and were analyzed for nuclear fragmentation by immunofluorescence analysis (A). In addition, DNA from transfected cells was prepared and analyzed by electrophoresis on an agarose gel as described under "Material and Methods" (B).

eIF-2␣ Phosphorylation Mediates PKR-dependent Apoptosis
phosphorylates eIF-2␣ with subsequent inhibition of protein synthesis and cell death (43). Finally, overexpression of a nonphosphorylatable S51A mutant eIF-2␣ was reported to transform NIH3T3 cells (56), suggesting that wild-type eIF-2␣ may act as a tumor suppressor, possibly through its phosphorylation and subsequent inhibition of translation initiation and induction of apoptosis. Although phosphorylation of eIF-2␣ is generally thought to be a general inhibitor of global protein synthesis, it is possible that translation of mRNAs encoding proapoptotic functions is selectively enhanced, whereas translation of mRNAs encoding antiapoptotic functions is suppressed when levels of eIF-2 become limited, i.e. when eIF-2␣ becomes phosphorylated. The best characterized paradigm for specific mRNA translational control mediated by eIF-2␣ phosphorylation is the translation of the mRNA encoding the transcriptional activator GCN4, which is selectively translated in response to amino acid deprivation through activation of the eIF-2␣ kinase GCN2 in Saccharomyces cerevisaie (57).
Numerous viruses have evolved gene products that inhibit apoptosis (58). Among these antideath mechanisms are the modulation of Bcl-2 (59), inactivation of the tumor suppressor p53 (60), and inhibition of the interleukin-1 converting enzymelike proteases (61). PKR was originally identified as an antiviral response mechanism of the host. The protective response is mediated by inhibition of protein synthesis to prohibit viral replication. However, the study presented here indicates that activation of PKR in the context of a viral infection may provide additional antiviral activity through induction of apoptosis and subsequent autodigestion to prevent viral spread. Viruses have evolved numerous mechanisms to circumvent PKR activation (62), suggesting that PKR is an additional target for which viruses have evolved antiapoptotic responses. The best characterized viral inhibitor of PKR is adenovirus VAI RNA, which specifically binds the dsRNA binding site on PKR (63). Many viruses, such as reovirus (64,65), rotavirus (66), and vaccinia virus (6,39), have evolved specific dsRNA-binding proteins that can sequester dsRNA and prevent PKR activation. Vaccinia virus actually produces two gene products, E3L and K3L, that block PKR activity through different mechanisms (67,68), and vaccinia virus deletion mutants lacking E3L activate apoptosis (6). In contrast, influenza virus activates the cellular inhibitor of PKR, p58 (13). Interestingly, the herpes simplex virus gene product ␥34.5 inhibits virus-induced apoptosis by restoring cellular protein synthesis, possibly through dephosphorylation of eIF-2␣ (69,70). This implies that viral PKR antagonists will prove to be general inhibitors of apoptosis. It is possible that pathogenesis for those viruses that do not have effective means of preventing PKR activation is attributed to apoptotic cell death.