Phosphorylation of the α-Subunit of the Eukaryotic Initiation Factor-2 (eIF2α) Reduces Protein Synthesis and Enhances Apoptosis in Response to Proteasome Inhibition*

Protein ubiquitination and subsequent degradation by the proteasome are important mechanisms regulating cell cycle, growth and differentiation, and apoptosis. Recent studies in cancer therapy suggest that drugs that disrupt the ubiquitin/proteasome pathway induce apoptosis and sensitize malignant cells and tumors to conventional chemotherapy. In this study we addressed the role of phosphorylation of the α-subunit eukaryotic initiation factor-2 (eIF2), and its attendant regulation of gene expression, in the cellular stress response to proteasome inhibition. Phosphorylation of eIF2α in mouse embryo fibroblast (MEF) cells subjected to proteasome inhibition leads to a significant reduction in protein synthesis, concomitant with induced expression of the bZIP transcription regulator, ATF4, and its target gene CHOP/GADD153. The primary eIF2α kinase activated by exposure of these fibroblast cells to proteasome inhibition is GCN2 (EIF2AK4), which has a central role in the recognition of cytoplasmic stress signals. Endoplasmic reticulum (ER) stress is not effectively induced in MEF cells subjected to proteasome inhibition, with minimal activation of the ER stress sensory proteins, eIF2α kinase PEK (PERK/EIF2AK3), IRE1 protein kinase and the transcription regulator ATF6 following up to 6 h of proteasome inhibitor treatment. Loss of eIF2α phosphorylation thwarts caspase activation and delays apoptosis. Central to this pro-apoptotic function of eIF2α kinases during proteasome inhibition is the transcriptional regulator CHOP, as deletion of CHOP in MEF cells impedes apoptosis. We conclude that eIF2α kinases are integral to cellular stress pathways induced by proteasome inhibitors, and may be central to the efficacy of anticancer drugs that target the ubiquitin/proteasome pathway.

Precise control of protein turnover is central for the regulation of the cell cycle, differentiation, and apoptosis. An important mechanism for targeted degradation of regulatory proteins involves covalent linkage of ubiquitin. Ubiquitination of proteins targets their degradation by the 26 S proteasome, a large ATP-dependent protease. Recent advances in cancer therapy suggest that proteasome inhibitors induce apoptosis and sensitize malignant cells and tumors to conventional chemotherapy (1,2). Clinical trials indicate that proteasome inhibitors may be effective in treatment regimes for both solid and hematological malignancies.
We have been interested in the mechanisms by which eukaryotic cells recognize stress signals and regulate programs of gene expression designed to ameliorate cellular damage or induce apoptosis. Central to cellular stress responses is a family of protein kinases that phosphorylate the ␣-subunit of eukaryotic initiation factor-2 (eIF2) 1 (3). The eIF2␣ kinases include PEK/Perk important for remedying protein misfolding in the endoplasmic reticulum (ER stress) (4), GCN2, which is activated by amino acid deprivation and UV irradiation (5)(6)(7), HRI that is expressed predominantly in erythroid tissues and couples protein synthesis to heme availability (8), and PKR, which is central to an antiviral defense pathway that is regulated by interferon (9). Phosphorylation of eIF2␣ in response to cellular stress reduces the exchange of eIF2-GDP to eIF2-GTP required to bind and deliver initiator Met-tRNA i Met to the translation machinery (10). The resulting reduction in eIF2 activity lowers global protein synthesis along with induced translation of selected mRNAs. For example, eIF2␣ phosphorylation enhances the expression of ATF4, a basic zipper transcription activator, via a mechanism involving delayed translation reinitiation at upstream open reading frames located in the 5Ј-end of the ATF4 mRNA (11)(12)(13). In response to ER or nutrition stress, ATF4 increases the expression of additional bZIP regulators, CHOP/GADD153 and ATF3, that together function to regulate metabolism, the redox status of cells, and apoptosis (14 -17). Additionally, eIF2␣ phosphorylation facilitates activation of NF-B by mechanisms independent of IKK phosphorylation of IB␣ (18 -20). eIF2␣ kinases work in concert with additional stress-sensing pathways. For example, PEK functions in conjunction with IRE1 and ATF6 to monitor ER stress to elicit a program of gene expression referred to as the unfolded protein response (UPR) (4,21). IRE1 is a type 1 ER transmembrane protein whose protein kinase activity is regulated by a mechanism that shares many features with PEK. Release of the ER chaperone GRP78/ BiP from IRE1 leads to autophosphorylation that is thought to contribute to enhanced activity of its C-terminal RNase function that mediates splicing of XBP1 mRNA (22)(23)(24). The spliced XBP1 mRNA encodes a bZIP transcription activator of a subset of UPR genes required for folding, maturation, and degradation of secretory proteins (25). Activation of the membrane-associated ATF6 also involves dissociation from GRP78, leading to transport of ATF6 from the ER to Golgi (26). ATF6 is then proteolytically cleaved, releasing the transcription factor to be transported into the nucleus where it binds promoters of genes subject to the UPR. Together this collection of bZIP-related transcription factors, ATF4, XBP1, and ATF6, function to coordinate gene expression designed to alleviate ER stress. Given that eIF2␣ kinases are activated in response to a broad range of stress conditions, many genes linked with the UPR are not necessarily restricted to ER stress. A further source of confusion in the literature is the idea that certain gene products induced by eIF2␣ kinases can promote survival during cellular stress, whereas others promote cell death. For example, CHOP is linked to enhanced apoptosis and chromosomal translocations involving the structural rearrangement of the CHOP gene are characteristic of myxoid and round cell liposarcomas (27)(28)(29)(30). Alternatively, the transcriptional regulator NF-B, which can be activated by eIF2␣ phosphorylation, has well characterized roles in anti-apoptotic pathways (31)(32)(33). Aberrations in the ER stress response pathway can lead to disease. For example, mutations in PEK/EIF2AK3 lead to Wolcott-Rallison Syndrome (WRS), a autosomal recessive disease associated with neonatal insulin-dependent diabetes, epiphyseal dysplasia, and hepatic and renal complications (34,35). Finally, accumulation of misfolded protein in the secretory pathway has been implicated in a range of neuropathologies, including Alzheimer disease and Parkinson's disease (36 -40).
Recent reports suggest links between exposure to proteasome inhibitors and ER stress and induction of the UPR. Treatment of primary neurons with MG132 for extended times induced PEK and IRE1 activity as judged by autophosphorylation (41). Furthermore, expression of poly(Glu) fragments with pathogenic repeat lengths led to aggregates in the cytoplasm of neuronal cells that also activated these ER stress sensors (41). These results were suggested to indicate that impaired proteasome function through pharmacological inhibition, or by accumulation of malfolded protein in the cytoplasm, ultimately block ER-associated degradation (ERAD). The ERAD is important for eviction of misfolded proteins from the ER to the cytoplasm and clearance by the ubiquitin/proteasome pathway. Glimcher and co-workers (42) reported that MG132 treatment of NIH 3T3 fibroblast cells or J558 myeloma cells led to ER stress, as judged by induction of GRP78 and CHOP mRNA, and activation of ER-associated caspase 12. However, the apparent ER stress did not induce XBP1 mRNA splicing by IRE1, leading to the proposal that proteasome inhibitors suppressed certain portions of the UPR despite stress in this secretory organelle. Microarray analyses indicate that treatment of the breast cancer cell line MCF7 with proteasome inhibitors, lactacystin or MG132, induces a collection of genes that includes those subject to eIF2␣ kinase control (ATF3, CHOP, and asparagine synthase) (43).
In this study, we used mouse embryo fibroblasts (MEFs) deleted for eIF2␣ kinase genes or containing a homozygous mutation at the eIF2␣ phosphorylation site (S51A) to study the role of eIF2␣ phosphorylation in the cellular response to proteasome inhibition. Phosphorylation of eIF2␣ in response to treatment with MG132 leads to a significant reduction in protein synthesis, concomitant with induced expression of ATF4 and its target gene CHOP. The primary eIF2␣ kinase activated by exposure of these fibroblast cells to this proteasome inhibitor is GCN2 that recognizes cytoplasmic stress signals. Interestingly, ER stress is not a primary event in MEF cells subjected to proteasome inhibition, with minimal activation of PEK or IRE1 following extended treatments of proteasome inhibitor. Loss of eIF2␣ phosphorylation thwarts caspase activation and delays apoptosis. We conclude that eIF2␣ kinases are integral to cellular stress pathways induced by proteasome inhibitors.

EXPERIMENTAL PROCEDURES
Cell Culture and Stress Conditions-PEK (Perk)Ϫ/Ϫ, GCN2Ϫ/Ϫ, eIF2␣-A/A, ATF4Ϫ/Ϫ, ATF3Ϫ/Ϫ, CHOPϪ/Ϫ, and HSF1Ϫ/Ϫ MEF cells and their wild-type counterparts were previously described (15, 18, 29, 44 -47). PEKϪ/Ϫ PKRϪ/Ϫ MEF cells were obtained from Barbara McGrath and Douglas Cavener (Pennsylvania State University) and were derived from embryos from day 12 to day 14 of gestation from crosses between PEKϩ/Ϫ PKRϪ/Ϫ mice. Genotypes of MEF cells were confirmed by PCR analysis and by immunoblot measurements. MEF cells were cultured in Dulbecco's modified Eagle's medium (BioWhittaker), supplemented with 1 mM non-essential amino acids, 100 units/ml penicillin, 10% fetal bovine serum, and 100 g/ml streptomycin. ER stress in MEF cells was brought about by the addition of 1 M thapsigargin to the medium followed by incubation for up to 16 h as indicated. Proteasome inhibition was elicited by the addition of 0.1-2 M MG132 to the medium and incubation for the indicated times. Alternatively, the proteasome inhibitors N-acetyl-Leu-Leu-norleucinal (ALLN or MG101) and clasto-lactacystin ␤-lactone were added to the MEF cells at 25 and 5 M, respectively, for up to 6 h as indicated. Where specified 20 M SB203580, a p38 MAP kinase inhibitor, or 20 M Z-VAD-FMK, a caspase inhibitor, were added to MEF cells 30 min prior to treatment with 1 M MG132. MEF cells were subjected to amino acid starvation by culturing cells in Dulbecco's modified Eagle's medium without leucine (BioWhittaker). Heat shock was elicited by shifting MEF cells from 37 to 42°C for up to 6 h as indicated.
Preparation of Protein Lysates and Immunoblot Analyses-MEF cells were subjected to the indicated stress or no stress and washed twice with ice-cold phosphate-buffered solution. Cell lysates were prepared using lysis buffer solution (50 mM Tris-HCl, pH 7.9, 150 mM NaCl, 1% Nonidet P-40, 0.1% SDS, 100 mM NaF, 17.5 mM ␤-glycerolphosphate, 10% glycerol) supplemented with protease inhibitors (100 M phenylmethylsulfonyl fluoride, 0.15 M aprotinin, 1 M leupeptin, and 1 M pepstatin) and sonication for 30 s. Lysates were clarified by centrifugation, and the protein content was determined by the Bio-Rad protein quantitation kit for detergent lysis according to the manufacturer's directions. For immunoblots, equal amounts of each protein sample were separated in a SDS-polyacrylamide gel and transferred to nitrocellulose filters. Low and high range polypeptide markers (Bio-Rad) were used to measure the molecular weight of proteins. Immunoblot analyses were carried out by incubating protein bound filters in TBS-T solution containing 20 mM Tris-HCl (pH 7.9), 150 mM NaCl, and 0.2% Tween-20 supplemented with 4% nonfat milk followed by incubation in the same solution with antibody that specifically recognized the indicated protein. ATF3, ATF4, CHOP, HSP70, and GRP78/BiP antibodies were purchased from Santa Cruz Biotechnology, actin monoclonal antibody was obtained from Sigma, and PEK antibody was previously described (48). ATF3 and ATF4 studies were independently confirmed using ATF4 and ATF3 polyclonal antibody that was prepared from recombinant protein. Immunoblots measuring eIF2␣ phosphorylation were performed using polyclonal antibody that specifically recognizes phosphorylated eIF2␣ at Ser-51 (Research Genetics or StressGen). Monoclonal antibody that recognizes either phosphorylated or nonphosphorylated forms of eIF2␣ was provided by Dr. Scott Kimball (Pennsylvania State University, College of Medicine, Hershey, PA). Following incubation of the antibody with the filters, the blots were washed three times in TBS-T and the protein-antibody complexes were visualized using horseradish peroxidase-labeled secondary antibody and chemiluminescent substrate. Quantitation of visualized bands was carried out by densitometry. Autradiograms shown in the figures are representative of three independent experiments.
RNA Isolation and Analyses-Northern analyses were carried out as previously described (49). Total cellular RNA was isolated from MEF cell treated with 1 M thapsigargin or 1 M MG132 for the indicated number of hours using the TRIzol reagent (Invitrogen) according to the manufacturer's instructions. 20 g of total RNA from each sample preparation was separated by electrophoresis using a 1.4% agarose/6% formaldehyde gel and visualized using ethidium bromide staining and ultraviolet light. RNA was transferred onto GeneScreen Plus filters (PerkinElmer Life Sciences), hybridized to 32 P-labeled DNA probes specific for the indicated genes, and filters were washed using high stringency conditions and visualized by autoradiography. The probe for CHOP included a 540-bp DNA fragment from the coding region that was provided by Dr. Yanjun Ma (St Jude's Children's Research Hospital, Memphis, TN). RT-PCR analysis of XBP1 mRNA splicing by IRE1 was carried out as previously described (50).
Radiolabeling Assays-Measurements of translation inhibition in response to treatment with MG132 were carried out as previously described (11). Briefly, S/S and A/A MEF cells were grown to 50% confluency in Dulbecco's modified Eagle's medium and washed with phosphate-buffered saline and incubated in trans-labeling medium without methionine or cysteine supplemented with 10% dialyzed fetal bovine serum. Non-stressed MEF cells were then labeled with [ 35 S]Met/ Cys express labeling mixture (ICN) for 30 min and washed twice with ice-cold phosphate-buffered saline containing non-radiolabeled methionine and cysteine. To elicit proteasome inhibition, 1 M MG132 was added to the cells cultured in trans-labeling medium for between 0.5 and 6 h. Prior to harvesting and lysate preparation, cells were then labeled with [ 35 S]Met/Cys express labeling mixture for 30 min. Equal amounts of total protein from each lysate preparation were separated by SDS-PAGE, and radiolabeled proteins were visualized by autoradiography. In parallel, each cell lysate was subjected to trichloroacetic acid precipitation. The insoluble pellet was measured by scintillation counting to determine the incorporation of [ 35 S]Met/Cys into proteins.
Luciferase Assays-To measure ATF6 transcriptional activity, NIH 3T3 fibroblast cells were transiently transfected with ATF6-directed luciferase reporter plasmid provided by Jinshi Shen and Ron Prywes (Columbia University, New York) (51). Following 40 h of transfection, cells were treated with 1 M MG132 or 1 M thapsigargin for up to 8 h or to no stress (0). Transfected cells were analyzed in triplicate with the plasmid that encodes multiple ATF6 elements embedded in a promoter directing firefly luciferase (51). As a control NIH 3T3 cells were also co-transfected with the reporter plasmid and pCGNATF6, which expresses full-length ATF6 (residues 1-670) or pCGNATF6-(1-373) encoding the cleaved and activated version of ATF6 to confirm the ATF6 dependence for expression of the luciferase reporter as described (51). Luciferase assays were carried out using the Dual-Luciferase Reporter Assay System (Promega) as described by the manufacturer's instructions. NIH 3T3 cells were transfected using Lipofectamine (Invitrogen). Co-transfections were carried out using ATF6-luciferase reporter plasmid and a Renilla luciferase plasmid serving as an internal control (Promega). Light units of both luciferase activities were assayed for 10 s. Luciferase activity was measured as relative light units (RLU) (Monolight Luminometer model 2010). All of the data were presented as the mean relative luciferase activity as calculated by RLU for the ATF6-directed luciferase divided by the RLU for the internal control construct. Luciferase activity was then normalized to that activity determined for NIH 3T3 cells that were not subjected to stress treatment. Three independent experiments were carried out, and the error bars indicate the S.D. for each relative luciferase activity. ATF4 translation expression was analyzed by using a luciferase reporter containing the 5Ј-portion of the ATF4 mRNA inserted downstream of a TK promoter that was previously described (12). The ATF4-luciferase reporter was transfected into NIH 3T3 cells and luciferase activity measured as described above.
Apoptotic Assays-S/S and A/A MEF cells were exposed to 1 M MG132 for up to 16 h, and the apoptosis was measured by Annexin V Apoptosis Detection Kit (BioVision) following the manufacturer's instructions. As indicated, 20 M SB203580 or 20 M Z-VAD-FMK were added to MEF cells 30 min prior to treatment with 1 M MG132. S/S and A/A MEF cells were also incubated with MG132 for 8 h, and then 10 g/ml cycloheximide were added and the cells were cultured for an additional 8 h. To address the role of CHOP in mediating apoptosis in response to proteasome inhibition, CHOPϩ/ϩ and Ϫ/Ϫ MEF cells were treated with 1 M MG132 or 1 M thapsigargin for up to 16 h. Average values and S.D. were derived from three independent experiments. Activation of caspases 3 and 9 was measured by immunoblot analysis using polyclonal antibodies specific to each caspase.

RESULTS
Proteasome Inhibition Induces eIF2␣ Phosphorylation-Proteasome inhibition elicits expression of stress-related genes and is linked with activation of the ER stress response (41,42). We wished to address whether eIF2␣ phosphorylation facilitates this program of stress gene expression in response to proteasome dysfunction. The well characterized proteasome inhibitor, MG132, was added to MEF cells containing a wildtype version of eIF2␣ (S/S) or a mutant version of eIF2␣ con-taining alanine substituted for the phosphorylatable residue serine 51 (A/A). As a control, we also treated the MEF cells with the standard ER stress agent, thapsigargin, that rapidly induces eIF2␣ phosphorylation by PEK. Phosphorylation of eIF2␣ at serine 51 was measured by immunoblot analysis using polyclonal antibodies specific to this phosphorylated epitope. Phosphorylation of eIF2␣ was induced in S/S MEF cells with as little as 0.1 M MG132 for 3 h (Fig. 1A). Moderate levels of eIF2␣ phosphorylation were measured following 1 h treatment with 1 M MG132 treatment with full induction following 3 h of this stress condition (Fig. 1, A and B). By comparison, within 1 h of exposure to 1 M thapsigargin, eIF2␣ phosphorylation was fully increased in S/S MEF cells and this level of phosphorylation was retained for up to 6 h of the ER stress (Fig. 1B). Levels of total eIF2␣ were similar among these different lysate preparations as judged by immunoblot using antibody that recognizes both phosphorylated and non-phosphorylated versions of the translation initiation factor (Fig. 1, A  and B). No eIF2␣ phosphorylation was detected in the A/A MEF cells, demonstrating the specificity of our immunoblot assay for eIF2␣ phosphorylated at serine 51. The well characterized proteasome inhibitors, ALLN and clasto-lactacystin ␤-lactone, were also found to effectively induce eIF2␣ phosphorylation in the S/S MEF cells, further supporting the idea that this stress response is activated in response to impairment of this protein degradative pathway (Fig. 1C).
Phosphorylation of eIF2␣ reduces the levels of eIF2-GTP required for translation initiation. To determine the effects of eIF2␣ phosphorylation during proteasome inhibition, we measured [ 35 S]Met/Cys incorporation following exposure of S/S and A/A MEF cells to MG132. There was a significant reduction in protein synthesis in S/S cells after 1 h of MG132 treatment with about 90% reduction in translation following 6 h of this stress treatment (Fig. 2). There was expression of a prominent protein about 70 kDa in size after 3 h of proteasome inhibition, suggestive of induced expression of a chaperone protein. By comparison, protein synthesis levels remained constant throughout the treatment of A/A MEF cells with MG132. Included among the proteins synthesized in the A/A cells was the 70 kDa protein, indicating that eIF2␣ phosphorylation was not required for its expression in response to proteasome inhibition. We conclude that eIF2␣ phosphorylation is required for reduced translation in response proteasome inhibition.
Proteasome Inhibition Enhances the Levels of eIF2␣ Kinasedirected Genes-Expression of transcriptional activators ATF4 and CHOP was reported to be increased in response to eIF2␣ phosphorylation during ER stress or amino acid starvation. Consistent with earlier reports (14 -16), we found that the levels of each of these bZIP transcriptional regulators were significantly increased in the S/S MEF cells in response to thapsigargin treatment, while no expression was measured in the mutant A/A cells (Fig. 1B). There were also increases in ATF4 and CHOP transcript levels in the S/S MEF cells as previously described (11,52) (Fig. 3). MG132 was also an effective inducer of ATF4 and CHOP over this time course (Fig.  1, A and B). Interestingly, while CHOP expression was absent in the A/A cells exposed to MG132 (Fig. 1, A and B), ATF4 levels were enhanced in these MEF cells devoid of eIF2␣ phosphorylation, albeit reduced compared with the S/S cells (Fig.  1B). CHOP mRNA levels were sharply increased within 3 h of MG132 treatment in S/S MEF cells but not in A/A cells (Fig. 3). By comparison, the amounts of ATF4 transcript were similar between non-stressed and MG132-treated S/S and A/A cells, suggesting regulatory roles for enhanced ATF4 translation and protein stabilization in response to this proteasome inhibitor. ATF4 and CHOP levels were also significantly increased in response to other proteasome inhibitors in the S/S cells, ALLN and clasto-lactacystin ␤-lactone, further demonstrating that the induction of these eIF2␣ kinase-targeted genes occurs in response to a block in the ubiquitin/proteasome pathway (Fig. 1C).
ATF3 is another bZIP transcriptional regulator that has been shown to be induced by eIF2␣ phosphorylation in response to ER or nutritional stress via a mechanism requiring ATF4 activity (15). We observed increased ATF3 levels in both S/S and A/A treated with MG132, ALLN or clasto-lactacystin ␤-lactone (Fig. 1, B and C), indicating that expression of this inhibitor, cell lysates were prepared, and the levels of phosphorylation of eIF2␣ were measured by immunoblot by using polyclonal antibody specific to eIF2␣ phosphorylated at serine 51. Equal amounts of proteins were analyzed in each lane. Levels of total eIF2␣ were assayed by using antibody that recognizes both phosphorylated and non-phosphorylated versions of the translation initiation factor. Levels of ATF3 and CHOP were measured by immunoblot analysis using antibody specific to each protein. B, S/S and A/A MEF cells were exposed to 1 M thapsigargin (TG) or 1 M MG132 (MG) for up to 6 h, as indicated, or no stress (0). Levels of phosphorylated eIF2␣, total eIF2␣, ATF4 ATF3, CHOP, GADD34, and HSP70 were measured by immunoblot analysis. C, S/S and A/A MEF cells were exposed to the following proteasome inhibitors: 1 M MG132, 25 M ALLN, 5 M clasto-lactacystin ␤-lactone (LAC) for up to 6 h, as indicated, or no stress (0). Levels of phosphorylated eIF2␣, ATF4, CHOP, ATF3, and actin were measured by immunoblot analysis. bZIP transcription factor during proteosome inhibition can occur independent of eIF2␣ phosphorylation. Recently, ATF3 expression was also shown to be induced independent of eIF2␣ phosphorylation in response to UV irradiation (20); therefore we conclude that the activities of eIF2␣ kinases are obligatory for ATF3 expression in response to only a subset of cellular stress conditions.
Another gene whose expression is enhanced by eIF2␣ phosphorylation and the ATF4 transcriptional regulator is GADD34. GADD34 is a regulator of Type 1 protein phosphatases that contributes to dephosphorylation of eIF2␣ in a mechanism of feedback control (53)(54)(55)(56). Analysis of GADD34 expression in response to thapsigargin and MG132 treatments showed a pattern similar to that described for ATF4 (Fig. 1B). GADD34 levels were elevated in S/S MEF cells treated with either stress agent. While expression of GADD34 was blocked in the A/A cells subjected to thapsigargin, there were significant levels of GADD34 in these mutant MEF cells exposed to MG132, albeit reduced compared with the similarly stressed S/S cells. Levels of the cytoplasmic chaperone, HSP70, were also shown to be increased in response to proteasome inhibition. We found similar increases in HSP70 mRNA and protein in the S/S and A/A MEF cells exposed to MG132 (Figs. 1B and 3). By comparison, ER stress did not induce HSP70 mRNA or protein expression in response to up to 6 h treatment with thapsigargin. Together these results indicate that eIF2␣ phosphorylation is required for full induction of stress-related gene products ATF4, CHOP, and GADD34, in response to proteasome inhibition.
The Transcription Factor HSF1 Is Not Required for Induction of ATF4, ATF3, and CHOP in Response to Proteasome Inhibition-HSF1 is a major transcription activator of heat shock proteins in response to accumulated misfolded protein in the cytoplasm (57). To address whether HSF1 contributes to expression of genes inducible by eIF2␣ phosphorylation, we exposed HSF1ϩ/ϩ and Ϫ/Ϫ MEF cells to MG132 or heat shock and measured expression of HSF70, ATF4, ATF3, and CHOP. While HSP70 levels were significantly elevated in response to either stress condition in the HSF1ϩ/ϩ MEF cells, there were only low amounts of this cytoplasmic chaperone in cells devoid of HSF1 (Fig. 4). By comparison, treatment with MG132 en-hanced the levels of ATF4, ATF3, and CHOP in both HSF1ϩ/ϩ and Ϫ/Ϫ MEF cells, although the cells devoid of HSF1 showed some reduction in the levels of these transcription factors early in the stress response (3 h). Actin levels were similar among the different lysate preparations showing that similar amounts of total protein were analyzed in these lanes. ATF3 levels were enhanced in response to heat shock, albeit lower than that measured in response to proteasome inhibition, and this elevation in the amount of ATF3 was largely absent in the HSF1Ϫ/Ϫ cells (Fig. 4). Minimal expression of ATF4 and CHOP were detected in response to heat shock. We conclude that HSF1 transcription factor is not obligate for induction of ATF4, ATF3, or CHOP in MEF cells subject to proteasome inhibition.
GCN2 Is the Primary eIF2␣ Kinase Activated by Proteasome Inhibition-Activation of PEK has been linked to dysfunction of proteasomes (41). To address whether PEK is the primary eIF2␣ kinase in MEF cells that is activated in response to MG132, we measured phosphorylation of eIF2␣ in PEKϩ/ϩ and Ϫ/Ϫ MEF cells subjected to this proteasome inhibitor or the ER stress agent thapsigargin. Phosphorylation of eIF2␣ was induced in the PEKϩ/ϩ MEF cells within 1 h of thapsigargin exposure, and this phosphorylation was substantially reduced in the PEKϪ/Ϫ cells (Fig. 5A). By comparison, enhanced phosphorylation of eIF2␣ in response to MG132 treatment was similar between the PEKϩ/ϩ and Ϫ/Ϫ cells. Activation of PEK during ER stress induces autophosphorylation of this eIF2␣ kinase, leading to a slow migration in SDS-PAGE (58,59). This retarded migration was detectable in the thapsigargin-treated PEKϩ/ϩ MEF cells but not in the cells treated for up to 6 h with MG132. Together, these experiments indicate that PEK is not the primary eIF2␣ kinase activated by proteasome inhibition in these fibroblast cells.
GCN2 has been reported to be activated by cytoplasmic stresses involving nutrient deprivation or exposure to UV irradiation (5-7,20,60 -63). To address the role of GCN2 in the stress response to proteasome inhibition, we carried out an analogous study using GCN2ϩ/ϩ and Ϫ/Ϫ MEF cells exposed to MG132 or subjected to leucine starvation. Consistent with earlier reports, eIF2␣ phosphorylation was enhanced in GCN2ϩ/ϩ cells deprived of leucine and this phosphorylation was greatly diminished in MEF cells devoid of GCN2 (Fig. 5B). GCN2 was also required for full phosphorylation of eIF2␣ in MEF cells exposed to MG132 with only modest phosphorylation of this translation initiation factor following 6 h of proteasome inhibition. CHOP expression required GCN2 in response to either amino acid limitation or MG132 exposure (Fig. 5B).
While ATF4 levels were induced in response to leucine limitation in the GCN2ϩ/ϩ cells, there was minimal expression of ATF4 in cells devoid of GCN2. Consistent with our earlier observations using S/S and A/A MEF cells, MG132 induced ATF4 levels in both GCN2ϩ/ϩ and Ϫ/Ϫ cells. In fact the levels of ATF4 in MG132-treated A/A cells exceeded that measured in S/S MEF cells deprived of leucine (Fig. 5B). As highlighted under "Discussion," earlier reports suggested that ATF4 alone is not sufficient for CHOP expression and this study combined with the prior publications support the idea that additional factors subject to regulation by eIF2␣ kinases contribute to CHOP induction in response to stress (11,15).
In addition to GCN2 and PEK, the eIF2␣ kinase PKR functions to regulate translation in a broad range of cell types. There was significant phosphorylation of eIF2␣ and subsequent expression of ATF4 and CHOP in response to proteasome inhibition in MEF cells deleted for PKR individually or in combination with PEKϪ/Ϫ ( Fig. 5C and data not shown). Taken together our analysis of MEF cells deleted for these eIF2␣ kinases demonstrate that GCN2 is the primary eIF2␣ kinase activated by proteasome dysfunction in MEF cells with one or more secondary eIF2␣ kinases being induced during extended treatments with MG132.
The ER Stress Response Is Secondary in the Cellular Regulation during Proteasome Inhibition-Given the previously noted association between proteasome inhibitors and induced ER stress, we wished to address whether the other ER stress sensors IRE1 and ATF6 are activated in response to proteasome inhibition. GRP78 is a well characterized UPR gene whose activation is linked to the ATF6 pathway. GRP78 mRNA was increased in the S/S cells treated with thapsigargin but did not change in response to up to 6 h of MG132 exposure (Fig. 3). However, there was a modest increase in GRP78 mRNA levels in A/A cells following 6 h of proteasome inhibition. As will be highlighted further below, loss of eIF2␣ phosphorylation and translation control appears to result in the mutant MEF cells eliciting a modest ER stress response following MG132 treatment. We propose that disruption of the eIF2␣ kinase pathway genetically predisposes these cells to ER stress. Loss of eIF2␣ phosphorylation may elicit an underlying stress that when combined with exposure to additional environmental toxins, such as proteasome inhibitors, induces a broader range of stress response pathways than detected in wild-type cells.
To directly test whether the proteasome inhibitor activates ATF6 transcriptional activity, NIH 3T3 fibroblast cells were transiently transfected with a plasmid encoding a luciferase reporter regulated by a promoter containing five ATF6 regulatory sites. As previously reported we found that co-transfection of an expression plasmid encoding the activated version of ATF6 enhanced luciferase activity even in the absence of stress, illustrating the dependence of this promoter on this transcriptional regulator (51) (data not shown). Following transfection, the NIH 3T3 cells containing only the regulated luciferase reporter gene were treated with MG132 or thapsigargin. In response to thapsigargin, there was a 15-and 24-fold increase in ATF6-regulated promoter activity following 6 and 8 h treatment, respectively (Fig. 6A). By comparison, the addition of 1 M MG132, a concentration that effectively induced ATF4, CHOP, ATF3, and GADD34 expression within 3 h of treatment, led to a more modest elevation in ATF6-directed luciferase activity. There was a 2-and 9-fold increase in luciferase activity following 6 and 8 h of proteasome inhibition, respectively.
We also measured ATF4 translational expression that is induced by eIF2␣ phosphorylation. Previous studies showed that a ATF4-luciferase fusion plasmid, which contained a con- stitutive thymidine kinase promoter, the 5Ј-leader of the encoded ATF4 mRNA, and an in-frame fusion between the ATF4 initiation codon and the luciferase coding region, was an effective measure of ATF4 mRNA translation control (12). This regulatory mechanism involves differential ribosome reinitiation and two upstream open reading frames located in the 5Ј-leader of the ATF4 transcript. The ATF4-luciferase plasmid was transfected into the NIH 3T3 fibroblast cells, and the cells were treated with thapsigargin or MG132. Luciferase activity was induced 8-fold within 1 h of MG132 with further increases following 3 and 6 h of proteasome inhibition (Fig. 6B). By comparison, thapsigargin, which induced a robust ATF6-directed promoter response, elicited a more modest increase in ATF4-Luc expression over this time course with an 8-fold increase in luciferase activity after 6 h of ER stress. We conclude that while MG132 is an effective inducer of eIF2␣ phosphorylation and ATF4 translation expression, this proteasome inhibitor is a more modest activator of the ER stress sensor ATF6.
As described above for PEK, autophosphorylation and activation of IRE1 occur in response to ER stress and this modification can be measured by retarded migration of IRE1 as judged by SDS-PAGE followed by immunoblot using antibodies specific to IRE1. We found that the IRE1 migration shift, as well as PEK, occurred during thapsigargin treatment in either S/S or A/A MEF cells (Fig. 7A). By comparison, there was minimal activation of IRE1 or PEK in response to MG132 in S/S cells. In the A/A cells, there was a progressive migration shift in IRE1 and PEK after 3 and 6 h of MG132 treatment (Fig.  7A). The fact that significant IRE1 and PEK autophosphorylation occurred only in the A/A cells further supports the idea that loss of eIF2␣ kinase stress response genetically predisposes cells to ER stress as outlined above.
IRE1 facilitates removal of a 26-nucleotide segment of XBP1 mRNA, leading to translation of a longer open reading frame encoding an activated version of this bZIP transcription factor. Using an RT-PCR analysis to delineate between these two forms of XBP1 mRNA, we found accumulation of the smaller, spliced sXBP1 transcript during thapsigargin treatment in both the S/S and A/A MEF cells (Fig. 7B). Consistent with the failure to detect activation of IRE1, there was no IRE1-mediated splicing of the XBP1 mRNA in the S/S MEF cells treated with MG132. These results further support the idea that proteasome inhibition in MEF cells over the time courses described is not an effective inducer of IRE1 or PEK and has a modest regulatory effect on ATF6. Interestingly, there was XBP1 mRNA splicing in the A/A cells subject to proteasome inhibition, albeit less than thapsigargin treatment. This XBP1 mRNA splicing suggests that proteasome inhibition combined with the absence of eIF2␣ phosphorylation can sensitize cells to modest ER stress and IRE1 activation.
ATF4 Is Required for Full Induction of ATF3 and CHOP in Response to Proteasome Inhibition-To address whether ATF4 FIG. 6. Phosphorylation of eIF2␣ during proteasome inhibition enhances ATF4 translation, but elicits lower levels of ATF6directed transcription. A, an ATF6-directed-firefly luciferase reporter gene was co-transfected with a Renilla luciferase control plasmid into NIH 3T3 fibroblast cells. Transfected cells were treated with from 1 M thapsigargin (black bar) or 1 M MG132 (white bar) for up to 8 h, and the dual luciferase assay was carried out as described in the "Experimental Procedures." Relative luciferase activity is presented in the histogram normalized for NIH 3T3 cells not subjected to stress treatment. B, ATF4 translational expression was measured using an ATF4-Luciferase reporter construct that contains the upstream open reading frames located in the 5Ј-portion of the ATF4 mRNA that were shown to be important for translation reinitiation in response to eIF2␣ phosphorylation (12). White bars represent values obtained from thapsigargin-treated NIH 3T3 cells, and black bars represent values from cells subjected to MG132 exposure.

FIG. 7. Proteasome inhibition is not an effective inducer of IRE1 activity or of XBP1 mRNA splicing in MEF cells. A, S/S and
A/A MEF cells were exposed to 1 M thapsigargin (Tg) or 1 M MG132 (MG) for up to 6 h, as indicated, or no stress (0). Levels of IRE1, PEK, ATF4, NRF2, and actin were measured by immunoblot analysis. Arrows indicate phosphorylated IRE1 (upper band) and non-phosphorylated IRE1 or PEK (lower band). Equal amounts of proteins were analyzed in each lane. B, S/S and A/A MEF cells were exposed to 1 M thapsigargin or 1 M MG132 for up to 6 h, or no stress (0). Total RNA was isolated from the MEF cells and spliced (sXBP1) and unspliced (uXBP1) versions of XBP1 transcripts were assayed by RT-PCR analysis.
is required for increased ATF3 and CHOP levels in response to proteasome inhibitor, we measured the amounts of these bZIP transcription factors in ATF4ϩ/ϩ and Ϫ/Ϫ MEF cells subjected to MG132 for up to 6 h. ATF4ϩ/ϩ cells showed a significant increase in ATF3 and CHOP levels in response to proteasome inhibition, while the amounts of these transcription factors were partially diminished in the absence of ATF4 function (Fig.  8A). Expression of GADD34, also shown to be regulated by ATF4 during ER or nutrition stress, appears to be induced largely independent of ATF4 in response to MG132. To determine whether ATF3 is important for induction of ATF4 and CHOP, we measured the levels of these transcription factors in ATF4ϩ/ϩ and Ϫ/Ϫ MEF cells subjected to proteasome inhibition. Deletion of ATF3 did not appreciably affect induction of CHOP and ATF4 expression in response to proteasome inhibition (Fig. 8B). We conclude that ATF4 is required for full induction of ATF3 and CHOP levels in response to MG132, although additional transcription factors other than ATF3 appear to be important facilitators for expression of these stressrelated genes in response to proteasome inhibition.
Post-transcriptional mechanisms are important for induced XBP1 and ATF4 expression in response to ER stress. We wished to address whether ATF4 participates in the expression of XBP1. Levels of the larger version of the XBP1 protein derived from the spliced mRNA (sXBP1) were measured in ATF4ϩ/ϩ and Ϫ/Ϫ MEF cells following exposure to thapsigargin or MG132. In the ATF4ϩ/ϩ MEF cells, there were high levels of sXBP1 within 1 h of thapsigargin exposure (Fig. 8C). In response to proteasome inhibition, there were measurable amounts of both the longer sXBP1 and the shorter form of XBP1 derived from the unspliced mRNA (uXBP1). Following 6 h of MG132 treatment, the uXBP1 was the predominant form in the ATF4ϩ/ϩ cells, consistent with the idea that there is minimal activation of IRE1 in response to proteasome inhibition. By comparison, in ATF4Ϫ/Ϫ MEF cells there was appreciable expression of XBP1 only following 6 h of thapsigargin exposure (Fig. 8C). There was also minimal expression of XBP1 in ATF4Ϫ/Ϫ cells with MG132 with measurable levels of uXBP1 and minimal expression of sXBP1 after 6 h of treatment with the proteasome inhibitor. Actin levels were similar among the different lanes, demonstrating that the amounts of total protein analyzed were comparable between cellular lysate preparations. These experiments indicate that ATF4 functions in conjunction with ATF6 and IRE1 in the expression of XBP1. Furthermore, these results support the idea that proteasome inhibition is not an efficient inducer of IRE1 and XBP1 mRNA splicing.
NRF2 is a member of the cap 'n' collar (CNC) subfamily of bZIP transcription factors and plays a critical role in the maintenance of glutathione levels that serves to alleviate the accumulation of reactive oxygen species during ER stress (64). Moreover, NRF2 has been suggested to be involved in PEK-mediated cell survival via direct PEK phosphorylation of this transcription factor (65). Treatment with proteasome inhibitors, lactacystin or MG132, has been reported to lead to NRF2 accumulation as well as its target reporter genes (66 -68). We addressed whether accumulation of NRF2 depends on ATF4 in response to thapsigargin or MG132 treatments. There was minimal accumulation of NRF2 in either ATF4ϩ/ϩ or ATF4Ϫ/Ϫ MEF cells following 6 h of thapsigargin treatment (Fig. 8C). Proteasome inhibition enhanced the amounts of NRF2 in ATF4ϩ/ϩ cells within 1 h of MG132 treatment, and this level of accumulation was further increased in the ATF4Ϫ/Ϫ cells. We carried out an analogous experiment using S/S and A/A cells and also observed a similar accumulation of NRF2 in response to MG132 during the 6-h time course with the greatest levels in the A/A cells (Fig. 7A). These experiments suggest that eIF2␣ phosphorylation and the accompanying induction of ATF4 are not required for NRF2 accumulation in response to proteasome inhibition.
Phosphorylation of eIF2␣ Accelerates Cell Apoptosis Induced by Proteasome Inhibitor-Proteasome inhibition induces a cascade of caspase cleavages that facilitates apoptosis (1,69). We wished to determine whether eIF2␣ phosphorylation and its attendant stress-related gene expression are regulators of the events leading to apoptosis. First, we treated S/S and A/A MEF cells with MG132 and measured activation of caspases 3 and 9, both of which were previously linked with proteasome inhibition. We found that caspases 3 and 9 were activated within 4 h of MG132 treatment in S/S MEF cells and that their cleavage were largely absent in similarly treated A/A cells (Fig. 9A). As expected, the activation of both caspases was blocked by pretreatment with 20 M of the caspase inhibitor, Z-VAD-FMK. The MAP kinase p38 pathway regulates apoptosis in response to diverse stress conditions, including proteasome inhibition (70 -72). We tested this premise by pretreating the MEF cells with the p38 inhibitor, SB203580, and did not find any reproducible change in the activation of caspase 3 or 9 in the S/S MEF cells in response to MG132. CHOP is suggested to facilitate apoptosis and, as described above, increased levels of CHOP in response to proteasome inhibition require eIF2␣ phosphorylation (Fig. 9A). Induction of CHOP levels occurred even in the presence of the p38 inhibitor and was not adversely impacted by Z-VAD-FMK. As observed above, no CHOP expression was observed in the A/A MEF cells exposed to the MG132. These experiments suggest that eIF2␣ phosphorylation is important for activation of caspases in MEF cells subject to proteasome inhibition.
To directly address the role of eIF2␣ phosphorylation in cell apoptosis induced by MG132, we measured the programmed cell death by the annexin assay. S/S and A/A MEF cells were treated for up to 16 h with 1 M MG132. We found increasing levels of apoptosis in S/S cells treated with MG132 over this time course (Fig. 9B). While 16 h of treatment with this proteasome inhibitor elicited similar levels of apoptosis, about 60%, in S/S and A/A cells, there was a significant reduction at the earlier time points in the absence of eIF2␣ phosphorylation (Fig. 9, B and C). For example after 8 h of MG132 exposure, 32% S/S cells elicited apoptosis, while only 12% A/A cells had undergone this programmed cell death. As expected, the pretreatment of S/S cells with Z-VAD-FMK had a marked reduction in apoptosis levels in S/S with only 22% apoptosis in the cells stressed for 16 h (Fig. 9B). Interestingly, pretreatment of the A/A MEF cells with the caspase inhibitor did not have a significant affect on apoptotic levels. There was about 60% apoptosis in the A/A MEF cells treated for 16 h with MG132, independent of pretreatment with Z-VAD-FMK, suggesting that there are differences in the signals eliciting apoptosis between S/S and A/A cells (Fig. 9C). Prior treatment with the p38 inhibitor, SB203580, led to a modest enhancement of apoptosis that was observed in both S/S and A/A cells (Fig. 9,  B and C).
The absence of eIF2␣ phosphorylation in A/A MEF cells led to aberrantly high levels of protein synthesis following treatment with MG132. Previously it was reported that pretreatment of PEK-deficient cells with cycloheximide improved cell survival following ER stress (73). It was rationalized that high levels of general translation further exacerbated accumulation of misfolded protein in the ER organelle triggering cell apoptosis. The apoptotic signal could be ablated by exposure to cycloheximide, which would sharply dampen general translation in PEKϪ/Ϫ cells. We wished to address whether such elevated translation contributed to the enhanced apoptosis in A/A cells following 16 h of MG132 treatment. The MEF cells were exposed to MG132 for 8 h, and then cycloheximide was added and the cells were incubated for an additional 8 h. Treatment of the A/A MEF cells with cycloheximide reduced apoptosis by almost 3-fold (Fig. 9D). By comparison, the combined MG132 and cycloheximide treatment of S/S cells led to further apoptosis Apoptosis was monitored using the annexin V apoptosis detection kit. Apoptotic measurements were derived from three independent experiments. D, apoptosis was measured in S/S and A/A MEF cells with 1 M MG132 for 16 h as indicated by the ϩ symbol. Alternatively, the MEF cells were incubated with MG132 for 8 h, and then 10 g/ml cycloheximide (CHX) was added to the MEF cells, as indicated by the ϩ, and the cells were cultured for an additional 8 h prior to the apoptosis measurements. compared with exposure to only MG132.
CHOP was reported to be important for implementation of apoptosis in response to ER stress (29). To address the requirement for CHOP in programmed cell death during proteasome inhibition, we treated CHOPϩ/ϩ and Ϫ/Ϫ MEF cells with MG132 for up to 16 h. As a control, we also treated these cells with the ER stress agent thapsigargin. Consistent with the previous report, deletion of CHOP reduced apoptosis by 50% following 16 h of thapsigargin treatment (Fig. 10A). Within 4 h of MG132 treatment, 22% CHOPϩ/ϩ cells had undergone apoptosis, while only 7% CHOP-deficient cells had elicited the programmed cell death (Fig. 10). With longer durations of proteasome inhibition, elimination of CHOP also lowered apoptosis. There was a 44% reduction in apoptosis in CHOPϪ/Ϫ cells following 8 h of MG132 treatment and a 25% decrease in the CHOP-deficient cells after 16 h of proteasome inhibition.
Immunoblot analysis of cleaved and activated caspases 9 and 3 also supported the involvement of CHOP. Cleaved caspase 9 was detected after 8 h of MG132 treatment in the CHOPϩ/ϩ MEF cells, and caspase 3 activation occurred following 8 and 16 h of proteasome inhibition (Fig. 10B). By comparison, minimal caspase 3 and 9 activation was detected in CHOPϪ/Ϫ cells following MG132 exposure. Furthermore, minimal caspase 3 and 9 activation was detected in either the CHOPϩ/ϩ or Ϫ/Ϫ cells following the ER stress. These results support the idea that eIF2␣ kinase induction of CHOP activity is at least one important reason for the pro-apoptotic function of eIF2␣ phosphorylation in response to proteasome inhibitors. DISCUSSION Proteasome inhibitors constitute a promising new class of anticancer drugs, with clinically proven efficacy against multiple myeloma (1,2). Gene expression profiling has revealed that a large number of stress-related genes are induced upon blockage of the ubiquitin/proteasome pathway (43). In this study we show that eIF2␣ phosphorylation and its attendant reduction in translation is central to cellular stress responses against proteasome inhibition. Reduced translation would allow cells additional time to repair cell damage induced by proteasome inhibition prior to synthesizing additional proteins (Fig. 11). Furthermore, phosphorylation of eIF2␣ is required for full induction of transcription factors ATF4 and CHOP that are important for regulation of stress response genes, including those involved in metabolism, the redox status of cells, and regulation of apoptosis. We show that the primary eIF2␣ kinase activated by proteasome inhibition in MEF cells is GCN2. Loss of eIF2␣ phosphorylation contributes to reduced levels of caspase activation and diminished apoptosis in response to MG132 treatment. Induction of CHOP function is one important reason for the pro-apoptotic function of eIF2␣ kinases during proteasome inhibition.
GCN2 Is a Primary eIF2␣ Kinase in Response to Proteasome Inhibition-Loss of GCN2 in MEF cells significantly reduced eIF2␣ phosphorylation in response to proteasome inhibition with appreciable eIF2␣ phosphorylation detected only following an extended treatment (6 h) with MG132 (Fig. 5B). By contrast deletion of PEK individually or in combination with PKR, the third major eIF2␣ kinase in MEF cells, did not diminish eIF2␣ phosphorylation or expression of its target genes ATF4 and CHOP in response to MG132 (Fig. 5, A and C). These results indicate that GCN2 is the primary eIF2␣ kinase activated by proteasome inhibition (Fig. 11). The observation that there was measurable eIF2␣ phosphorylation in GCN2Ϫ/Ϫ MEF cells following extended exposure to MG132 suggests that PEK and/or PKR can serve as secondary eIF2␣ kinases.
The idea of primary and secondary eIF2␣ kinases being activated in response to a given stress condition is central to our understanding of the regulation of the eIF2␣ kinase stress response and its integration into programs of gene expression directed toward prevention or remediation of stress-induced cell damage. This model is well illustrated by a recent study by Zhan et al. (74) that characterized stress-induced eIF2␣ phosphorylation in the yeast Schizosaccharomyces pombe. In addition to GCN2, this yeast expresses two eIF2␣ kinases, designated HRI1 and HRI2, which are related to mammalian HRI. Analysis of mutant S. pombe strains that express only one of the three eIF2␣ kinases revealed that GCN2 is rapidly activated in response to nutrient deprivation or upon exposure to high levels of sodium chloride (74). By contrast, HRI2 is the primary eIF2␣ kinase induced by exposure to heat shock, arsenite, or cadmium. Loss of the primary eIF2␣ kinase, for example HRI2 in response to cadmium exposure, significantly delays eIF2␣ phosphorylation. Only after 6 h of cadmium exposure is there activation of secondary eIF2␣ kinases GCN2 and HRI1. In some cases such as arsenite treatment, a secondary eIF2␣ kinase GCN2 can be induced rapidly but with an FIG. 10. CHOP facilitates apoptosis in response to proteasome inhibition. A, apoptosis was measured in CHOPϩ/ϩ and Ϫ/Ϫ MEF cells that were exposed to 1 M MG132 (MG) or 1 M thapsigargin (Tg) for 4, 8, or 16 h as indicated, or to no stress (0). Apoptosis was monitored in three independent experiments by using the annexin V apoptosis detection kit. B, CHOPϩ/ϩ and Ϫ/Ϫ MEF cells were treated with 1 M MG132 or 1 M thapsigargin for 4, 8, or 16 h as indicated, or to no stress (0). Cleavage and activation of caspases 3 and 9 were measured by immunoblot analysis using polyclonal antibody specific to each caspase. An arrow indicates the cleaved/activated caspase 9. CHOP levels were measured by immunoblot. Equal amounts of proteins were analyzed in each lane, as highlighted by the immunoblot measurements of actin. activity much diminished compared with the primary eIF2␣ kinase, HRI2. A third scenario was tandem primary eIF2␣ kinases as highlighted by cells expressing only GCN2 or HRI2 displaying levels of eIF2␣ phosphorylation following 1 h exposure to hydrogen peroxide that were similar to that measured in wild-type cells. These studies suggest a more complex picture than a single eIF2␣ kinase being activated by one or two well defined stress conditions. In this multifaceted model, eukaryotic stress responses have integrated mechanisms for coordinate activation of eIF2␣ kinases in response to a multitude of stress conditions. GCN2 was originally identified as a regulator of translation control in response to starvation for one of many different amino acids (5,6,75). Uncharged tRNA that accumulates during amino acid depletion binds to a GCN2 regulatory domain homologous to histidyl-tRNA synthetase (HisRS) enzyme, triggering enhanced eIF2␣ kinase activity. Studies of mammalian and yeast cells suggest that GCN2 is activated by many different cellular stresses that are not directly linked to amino acid starvation. For example, in the yeast Saccharomyces cerevisiae, phosphorylation of eIF2␣ by GCN2 can be enhanced by purine or glucose deprivation, exposure to high concentrations of sodium chloride or MMS, and treatment with rapamycin or hydroxyurea (6). We do not yet understand the molecular basis for activation of GCN2 in response to these different environmental toxins, but many of these stress conditions in yeast have been shown to require a functional HisRS-related domain of GCN2, suggesting that uncharged tRNA may be a direct or indirect signal regulating this eIF2␣ kinase (63). A study by Fribley et al. (76) that was published following the completion of this study suggested that treatment of cultured human head and neck squamous cell carcinoma (HNSCC) cells with the proteasome inhibitor PS-431 can reduce general translation and enhance apoptosis. It was suggested that reactive oxygen species (ROS) induced by PS-431, combined with low expression of NF-B, may be an important reason for the potent anti-cancer function of this proteasome inhibitor. The basis for induced ROS by PS-341 was not fully understood, although it was suggested that activation of PEK and ER stress in HNSCC cells were contributors (76). The link between ER stress and this proteasome inhibitor will be discussed further below.
ER Stress Is Secondary in MEF Cells Responding to Proteasome Inhibition-The UPR directs gene expression important for remediating accumulation of malfolded protein in the ER. Part of this coordinated response is the ERAD, which facilitates transport of the misfolded protein from the ER to the cytoplasm, where the evicted proteins are ubiquitinated and degraded via the proteasome. Recent reports have suggested that inhibition of the proteasome function can induce the UPR via thwarting the ERAD-directed turnover of malfolded secretory proteins (41,42). Our study suggests that the ER stress response pathway is not a primary event in MEF cells subjected to proteasome inhibition. First, consistent with earlier studies, we find that levels of CHOP mRNA and protein are enhanced by proteasome inhibition (Figs. 1B and 3) (42,76); however, in this stress condition, our study suggests that GCN2 phosphorylation of eIF2␣ is the primary regulator for CHOP expression (Fig. 5B). In fact PEK and PKR are fully dispensable for eIF2␣ phosphorylation in response to proteasome inhibition and we find that there is minimal activation of PEK as judged by autophosphorylation in response to this cellular stress (Figs. 5, A and C and 7A). Second, autophosphorylation accompanying activation of IRE1 is largely absent in MEF cells treated with MG132 (Fig. 7A). This is consistent with an earlier report that that MG132 treatment adversely affects the activity of this ER transmembrane protein (42). Third, expression of XBP1 is only weakly induced by MG132 treatment and XBP1 mRNA splicing is largely absent in S/S MEF cells subjected to this proteasome inhibitor (Fig. 7B). It is noted that a portion of XBP1 mRNA are spliced in A/A cells, suggesting that loss of eIF2␣ phosphorylation may render cells more susceptible to induction of ER stress. With regard to XBP1 expression, analysis of ATF4ϩ/ϩ FIG. 11. Inhibition of the ubiquitin/proteasome pathway induces eIF2␣ phosphorylation and translational regulation. Proteasome inhibition increases GCN2 phosphorylation of eIF2␣ and translational control. Lowered general translation would allow cells additional time to ameliorate cell damage induced by proteasome inhibition prior to synthesizing additional proteins. Concomitant with this reduction in total protein synthesis, eIF2␣ phosphorylation induces expression of ATF4, a transcriptional activator of genes important for the repair of cellular damage elicited by stress or alternatively for apoptosis. An important target gene of ATF4 is CHOP/GADD153, which is involved in modulating the redox status of cells and promoting apoptosis. It is noted that there are some differences between proteasome inhibition and ER stress regarding the stress-related genes whose expression require eIF2␣ phosphorylation. For example, eIF2␣ kinase function is required for induction of ATF3 and GADD34 during ER stress, but not in MEF cells treated with proteasome inhibitors. This suggests that additional stress-sensing pathways can functionally replace portions of the eIF2␣ kinase pathway in response to certain environmental stress conditions. Proteasome inhibition also enhances transcriptional activation of HSP70 by HSF1 by a mechanism that does not require eIF2␣ phosphorylation. The HSF1 transcriptional regulator is not essential for induced expression of eIF2␣ kinase target genes, ATF4 and CHOP. Finally, inhibition of the ubiquitin-mediated degradative pathway enhances the level of key regulatory proteins, including NRF2 and IB. NRF2 can dimerize with ATF4 and enhance transcriptional regulation of the gene encoding heme oxygenase-1, an important antioxidant enzyme, and IB serves as a repressor of NF-B (31,80,81). NF-B activity is also subject to regulation by eIF2␣ phosphorylation. and Ϫ/Ϫ MEF cells indicates that the ATF4 transcription factor facilitates XBP1 expression in response to thapsigargin treatment and more modestly following proteasome inhibition (Fig. 8C). Finally, we found that transcription of the classic UPR gene, GRP78, is not increased in wild-type MEF cells following up to 6 h of MG132 treatment (Fig. 3). It is noted that there was a moderate increase in ATF6 transcription activity in NIH 3T3 fibroblast cells following 6 h of MG132 treatment (Fig. 6A). This suggests that ER stress can be elicited in MEF cells with longer durations of proteasome inhibition.
Together these results suggest that ER stress is a secondary event in MEF cells stemming from proteasome dysfunction. This conclusion is not necessarily in disagreement with earlier reports regarding proteasome inhibitors and induction of the UPR. First as discussed earlier, some UPR genes, such as CHOP, are not regulated only by ER stress. For example, GCN2 phosphorylation of eIF2␣ during amino acid starvation increases the expression of CHOP with no apparent ER stress (11,15). Therefore, induction of CHOP can be incorrectly interpreted as a marker of accumulation of malfolded protein in the ER organelle. Second, our studies have utilized MEF cells because of their well characterized stress responses and the availability of relevant gene knockouts. Many of the earlier studies have incorporated other cell types, including primary and immortalized neuronal cells and myeloma cell lines (41,42). Differences between cell physiology could alter induction of ER stress in response to proteasome inhibitors. For example, the increased secretory load of immune cells may enhance the amount of proteins directed to the ERAD for clearance by the ubiquitin/proteasome pathway. Therefore, proteasome inhibitors could more rapidly lead to accumulation of malfolded proteins in the ER. In the case of HNSCC cells, the proteasome inhibitor PS-341 was reported to lower protein synthesis following 1 h of treatment with further reductions after 3 h of exposure to the proteasome inhibitor (76). Although the role of eIF2␣ phosphorylation was not addressed in this earlier study, it is of interest to note that PEK was observed to be phosphorylated following 4 h of PS-341 exposure. Autophosphorylation of PEK is linked to its activation state, suggesting that proteasome inhibition in HNSCC cells may trigger ER stress. In MEF cells, PEK phosphorylation and ATF4 and CHOP expression were enhanced after an extended 16 h exposure to PS-341 (76). Following our discussion above of stress activation of primary and secondary eIF2␣ kinases, the present study would support the idea that ER stress and activation of PEK are a secondary event in fibroblast cells that is elicited after extended proteasome inhibition.
While MG132 treatment of wild-type MEF cells do not elicit a primary ER stress response, several lines of evidence support the idea that mutant cells deficient in eIF2␣ phosphorylation are predisposed to eliciting ER stress following proteasome inhibition. We observed that in A/A MEF cells, MG132 treatment led to autophosphorylation of PEK and IRE1, splicing of XBP1, and a modest increase in GRP78 expression (Figs. 3 and  7). We suggested that a defective eIF2␣ kinase response elicits an underlying stress that, when coupled with proteasome inhibition, enhances accumulation of misfolded protein in the ER organelle. Together these results indicate that certain genetic deficiencies can alter cellular stress responses such that new pathways are activated to alleviate cellular damage. Therefore, genetic polymorphisms can contribute specifically to the combinations of stress response pathways that are activated in response to a given environmental stress. Variations in stress response programs may be important contributors to disease prevention as these pathways may alter the ability of the cells to avert and remediate cellular damage.
Role of eIF2␣ Phosphorylation in the Cascade of Gene Expression Induced by Cellular Stress-Proteasome inhibition induces a cascade of gene expression important for preventing or repairing cell damage or alternatively triggering apoptosis. This study supports the idea that, while many of the core features of gene expression induced by eIF2␣ phosphorylation during proteasome inhibition are conserved among different stress conditions, there are clear differences regarding the contribution of key transcription factors. Phosphorylation of eIF2␣ and the resulting increase in ATF4 levels are integral to this program of stress-related gene expression. We show that eIF2␣ phosphorylation is required for full ATF4 accumulation in response to proteasome inhibition. ATF4 translation is induced by eIF2␣ phosphorylation, and our ATF4-luciferase reporter assay and ATF4 mRNA and protein analysis support the idea that proteasome inhibition enhances ATF4 translation expression (Figs. 3, 6B, and 11). However, there were significantly elevated ATF4 levels in the A/A MEF cells subjected to MG132, while those cells treated with thapsigargin showed minimal amounts of this transcription factor (Fig. 1B). ATF4 is a labile protein, and its degradation relies on ubiquitin via the E3 ubiquitin ligase SCF ␤TrCP (77). MG132 was shown previously to significantly enhance the half-life of ATF4, suggesting that stabilization of ATF4 is an important reason for the accumulation of ATF4 levels in A/A cells treated with this proteasome inhibitor (77). Furthermore, Yang and Karsenty (78) used RNA interference to reduce expression of SCF ␤TrCP in NIH3T3 cells and found enhanced levels of ATF4 independent of environmental stress. Together these results suggest that mechanisms stabilizing ATF4 can lead to its accumulation even in the absence of eIF2␣ phosphorylation and translation control.
Despite the high levels of ATF4 in A/A cells subjected to proteasome inhibition, there was minimal CHOP expression in cells devoid of eIF2␣ phosphorylation (Figs. 1, A and B and 5B). Analysis of ATF4Ϫ/Ϫ MEF cells indicated that this bZIP transcription factor was required for full induction of CHOP and ATF3 in response to proteasome inhibition (Fig. 8A), a pattern that was previously reported in MEF cells exposed to the ER stress ( Fig. 1B) (14,15). These results suggest that ATF4 alone is not sufficient for induction of CHOP. One or more additional factors that are induced by eIF2␣ phosphorylation appear to work in conjunction with ATF4 to increase CHOP expression. This model is consistent with earlier studies that found that transient transfection of an ATF4 expression plasmid in MEF cells did not elicit CHOP expression in the absence of stress (11,15).
Analysis of ATF3 expression in MEF cells also revealed pattern variations among different stress conditions. Similar to that described for CHOP, expression of ATF3 in response to ER or nutritional stress was previously shown to require eIF2␣ phosphorylation and ATF4 (15). In the example of proteasome inhibition, there was accumulation of ATF3 levels even in the absence of eIF2␣ phosphorylation (Fig. 1B). However, full expression of ATF3 in response to MG132 treatment required ATF4 (Fig. 8A). These results suggest that the increase in ATF4 levels, combined with one or more stress-related factors induced independent of eIF2␣ kinases, contributes to ATF3 expression. Finally, it is noted that there are similar levels of GADD34 present in ATF4ϩ/ϩ and ATF4Ϫ/Ϫ MEF cells following MG132 treatment (Fig. 8A). This observation appears to be at variance with previous studies using ER or nutritional stresses that found the eIF2␣ kinase stress response and ATF4 activity to be required for transcriptional expression of GADD34 (54,79). Interestingly, arsenite-induced expression of GADD34 occurs independent of eIF2␣ phosphorylation (54), suggesting that there may be common mechanistic features between this stress and proteasome inhibition.
In response to ER stress, PEK works in concert with IRE1 and ATF6 to direct the UPR expression program. We sought to determine whether other stress-responsive transcription regulators, such as HSF1, are integrated into the eIF2␣ kinase stress response during proteasome inhibition. HSF1 is required for expression of heat shock proteins and molecular chaperones, such as HSP70, in response to diverse stress conditions, including proteasome inhibition. However, loss of HSF1 resulted in only a modest reduction in the expression of CHOP, ATF4, and ATF3 in response to MG132 (Fig. 4). Furthermore, increased expression of HSP70 in response to MG132 occurred independent of eIF2␣ phosphorylation. These results suggest that eIF2␣ kinases regulate gene expression independent of HSF1 transcriptional control (Fig. 11).
Role of eIF2␣ Phosphorylation in Apoptosis-Analysis of S/S and A/A MEF cells revealed that loss of eIF2␣ phosphorylation in response to proteasome inhibition reduces activation of caspases 3 and 9 and delays apoptosis (Fig. 9). Differences in apoptosis were detected following 8 h of MG132 exposure with about 32% apoptosis in S/S MEF cells and 12% in A/A cells. Following 16 h of proteasome inhibition, there were similar levels of apoptosis, about 60%, among the S/S and A/A cells. However, in the A/A cells treated for this extended stress condition, there was no measurable activation of caspase 3 and 9, nor was apoptosis reduced by pretreatment with the caspase inhibitor Z-VAD-FMK. These results suggest that different mechanisms direct apoptosis in the A/A cells compared with the S/S cells that contain functional eIF2␣ kinase stress pathway. Central to induction of apoptosis in S/S cells is enhanced expression of CHOP in response to eIF2␣ phosphorylation ( Figs. 10 and 11).
The pro-apoptotic function of eIF2␣ phosphorylation in response to proteasome inhibition contrasts with the anti-apoptotic activities of eIF2␣ kinases described for other stress conditions. For example, treatment of PEKϪ/Ϫ MEF cells with the ER stress agent thapsigargin for 24 h led to increased apoptosis compared with its wild-type counterpart (73). The basis for the lowered survival of PEKϪ/Ϫ cells is thought to be due at least in part to the abnormally elevated levels of protein synthesis in the PEK-deficient cells subjected to ER stress. Continued synthesis of proteins slated for the secretory pathway is proposed to exacerbate the ER stress by further overloading the folding and assembly apparatus of the ER organelle (73). Consistent with this premise, we found that by combining cycloheximide with MG132 there was almost a 3-fold reduction in apoptosis in the A/A MEF cells compared with those exposed only to the proteasome inhibitor (Fig. 9D). This observation supports the idea that there are different molecular signals eliciting apoptosis in the A/A and S/S MEF cells.
Loss of GCN2 contributes to elevated apoptosis in MEF cells following exposure to UV irradiation (20). GCN2 phosphorylation of eIF2␣ in response to UV irradiation facilitates activation of NF-B but does not lead to detectable accumulation of the pro-apoptotic factor CHOP. It was proposed that this regulatory pattern contributes to the anti-apoptotic function of GCN2 in response to UV stress (20). There is no induction of NF-B by proteasome inhibition. In fact pretreatment with proteasome inhibitors can block NF-B activation in response to exposure to cytokines or UV irradiation by stabilizing the labile IB␣ inhibitory protein (Fig. 11). Perhaps, the failure of eIF2␣ phosphorylation to activate NF-B in response to proteasome inhibition, combined with induced expression of CHOP, is an important underlying reason for the pro-apoptotic function of GCN2 in response to proteasome dysfunction.