Inhibition of translational initiation by activators of the glucose-regulated stress protein and heat shock protein stress response systems. Role of the interferon-inducible double-stranded RNA-activated eukaryotic initiation factor 2alpha kinase.

Depletion of endoplasmic reticulum (ER) Ca2+ perturbs protein folding and processing within the organelle while inhibiting translational initiation through activation of the double-stranded RNA-activated eukaryotic initiation factor (eIF)-2α kinase (PKR) (Prostko, C. R., Dholakia, J. N., Brostrom, M. A., and Brostrom, C. O. (1995) J. Biol. Chem. 270, 6211-6215). The glucose-regulated stress protein (GRP) chaperones are subsequently induced. We now report that sodium arsenite, a prototype for stressors fostering cytoplasmic protein misfolding, also inhibits translational initiation through activation of PKR while subsequently inducing the heat shock protein (HSP) chaperones. Arsenite neither mobilized ER-associated Ca2+ nor slowed peptide chain elongation. Various HSP-inducing chemicals caused rapid phosphorylation of eIF-2α. When incubated with double-stranded RNA, extracts derived from arsenite-treated cells displayed greater degrees of phosphorylation of PKR and eIF-2α than did control extracts. Cells overexpressing a dominant negative PKR mutation resisted translational inhibition and eIF-2α phosphorylation in response to ER or cytoplasmic stressors. Induction of either the HSP or GRP chaperones was accompanied by development of translational tolerance to either Ca2+-mobilizing agents or arsenite. Following induction of the HSPs by arsenite, cells remained susceptible to induction of the GRPs by Ca2+-mobilizing agents. Conversely, cells possessing induced GRP contents in response to Ca2+-mobilizing agents readily induced the HSPs in response to arsenite. It is concluded that the two chaperone systems function independently except for their mutual suppression of PKR.

Chemicals and conditions that damage proteins, cause protein misfolding, or inhibit protein processing trigger the onset of protective homeostatic mechanisms resulting in "stress responses" in mammalian cells (reviewed in Refs. [1][2][3]. Included in these responses are an acute inhibition of mRNA translation, a subsequent induction of various protein chaperones, and the recovery of mRNA translation. Separate, but closely related, stress response systems exist for the endoplasmic retic-ulum (ER), 1 relating to the induction of specific glucose-regulated stress proteins (GRPs) and for the cytoplasm (cytosol and mitochondria), pertaining to the induction of the heat shock proteins (HSPs). The most prominent GRPs are chaperones of 78 and 94 kDa that share sequence homology with HSP70 and HSP90, respectively but localize to the ER lumen. GRP78, also termed BiP, is hypothesized to function in early ER protein folding and assembly, in the translocation of proteins from the cytosol to the ER for processing, and in the retention and accumulation of improperly folded proteins within the ER lumen when processing is distressed (4,5). GRP94 is a high capacity Ca 2ϩ -binding glycoprotein thought to chaperone partially oxidized or folded intermediates (6). The ER functions critically in the early processing of newly synthesized secretory, lysosomal, and integral membrane proteins. Ca 2ϩ sequestered by the ER supports the oligomerization, folding, and trimming of mannose residues of glycoproteins as well as the degradation of various incompletely assembled or abnormal proteins (7)(8)(9)(10)(11)(12)(13)(14). The oxidizing environment of the lumen is required for the processing of proteins with disulfide-bonded intermediates (15). Agents such as Ca 2ϩ ionophores, thapsigargin, and extracellular chelators that deplete ER Ca 2ϩ or those that act as luminal reductants induce GRP78 and GRP94 (reviewed in Refs. 4 and 5). Induction of these GRPs is also associated with viral infections, overproduction of incomplete secretory proteins, and inhibition of protein glycosylation from such agents as tunicamycin. GRP genes possess highly conserved promoter regions that confer ER stress inducibility and that bind specific nuclear factors during stress. An ER transmembrane kinase appears key for signaling of the ER stress response in yeast (16). Initiation on grp78 mRNA can occur by a cap-independent, internal ribosome-binding mechanism (17). GRP78 and GRP94 maintain viability, and induction appears necessary for survival during persistent ER stress (18).
Induction of the HSPs follows the production of damaged or misfolded cytoplasmic proteins in response to elevated temperature, oxidative free radicals, and heavy metals or from the synthesis of aberrant cytoplasmic proteins in response to amino acid analogs (reviewed in Refs. [1][2][3]. Proteins containing sulfhydryl groups are particularly sensitive to modification. Sodium arsenite, a prominent inducer of the HSPs that pro-duces minimal suppression of viability, is thought to act largely by inactivating sulfhydryl groups (19). Mammalian HSPs induced by these stressors include proteins of 110, 90, 72, 70, 60, and 30 kDa of differing subcellular distribution and chaperone function (3). The depletion of preexisting HSPs by binding to eccentric protein structures permits the trimerization of latent monomeric heat shock factor to a form that binds to heat shock response elements on DNA such that induction of heat shock mRNAs ensues (20,21). These mRNAs, like that for grp78, possess structural features that permit their selective translation during stress. The chaperone activity of the HSPs is exemplified by HSP70 and a closely related form of the protein expressed in nonstressed cells, HSC72. Both prevent incorrect folding of polypeptides during synthesis, permitting delivery to organelles in an unfolded state for translocation. These chaperones may also solubilize or refold denatured or aberrant proteins and/or deliver them to a degradative system.
The acute depression of mRNA translation occurring in response to ER stress results specifically from the phosphorylation of eIF-2␣ and inhibition of initiation (22)(23)(24)(25)(26)(27). Recently we reported that the dsRNA-activated, interferon-inducible protein kinase (PKR) mediates eIF-2␣ phosphorylation occurring in NIH-3T3 cells in response to mobilization of ER-sequestered Ca 2ϩ (27,28). While eIF-2␣ phosphorylation has been reported to occur in response to stressors that selectively induce HSPs (reviewed in Ref. 29), it was unclear whether translation was inhibited primarily at initiation or peptide chain elongation (30 -32). We now report that PKR mediates the phosphorylation of eIF-2␣ in response to sodium arsenite without evident effects on translational elongation. The potential role of PKR as a general eIF-2␣ kinase regulating translational initiation is discussed.
General Methodology-Murine NIH-3T3 cells were cultured in dishes or multiwell plates in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and were utilized at confluence (1.6 ϫ 10 5 /cm 2 ). NIH-3T3 cells expressing a dominant negative mutant PKR (K296P in catalytic subdomain II, clone KP 3 A) were derived as described previously (28). KP 3 A cells were cultured as above in the presence of 1 mg/ml G418 and utilized at confluence. Rat GH 3 pituitary cells were maintained in suspension and utilized as described (34). Prior to treatments cells were equilibrated for 5-15 min with serum-free Ham's F-10 modified to contain 25 M leucine and 0.2 mM Ca 2ϩ . Amino acid incorporation was measured as described (35) for 30-min incubations of 3 ϫ 10 5 cells/experimental condition. Incubations were conducted in triplicate, and results are presented as the average or average Ϯ range of values obtained. Findings were reproduced on at least two separate occasions. [ 35 S]methionine labeling (2 ϫ 10 6 cells/experimental condition), onedimensional 10% polyacrylamide gel electrophoresis (PAGE) of detergent-solubilized extracts of methionine-labeled cells, and autoradiography were conducted as described previously (25). Ribosomal and polyribosomal size distributions were measured by density gradient centrifugation as described previously (34). Cell-associated Ca 2ϩ was measured as described (36).
Determination of Average Ribosomal Transit Times-Transit times were determined for cells in monolayer culture by a modification of the procedure of Ledford and Davis (37). Growth media were removed by aspiration from cells grown to confluence in 76-cm 2 dishes (2 ϫ 10 7 cells). The cells were washed with modified Ham's F-10 adjusted to contain 100 M leucine, 33 M methionine, 200 nM Ca 2ϩ chloride, and 0.6 M phorbol-12-myristate 13-acetate (PMA) and then equilibrated with 38 ml of the medium for 30 min at 37°C with various experimental additives. Controls (blanks) were adjusted to 2 mM leucine and 100 M cycloheximide to prevent incorporation. Incorporations were initiated by the addition of 50 l [ 14 C]leucine (5 Ci). After 45 min the samples were adjusted with 50 l [ 3 H]leucine (20 Ci), and the incubation continued for 15 min. Samples (50 l) were taken for determination of slopes, the incubation media were removed, and the cultures were adjusted with 40 ml of ice-cold 125 M cycloheximide plus 2 mM leucine in isotonic saline to terminate incorporation. This medium was removed, and the cells were dislodged into 2.5 ml of lysis buffer containing 50 mM Tris, pH 7.6, 25 mM KCl, 1% Triton X100, 1% sodium deoxycholate, 10 g/ml cycloheximide, 2 mM leucine, 1 mg/ml heparin, and 5 mM MgCl 2 . The cells were homogenized with a Dounce glass homogenizer equipped with a tight fitting pestle (20 strokes), and the lysates were transferred to 15-ml Corex centrifuge tubes. The utensils were rinsed with 2 ml of additional homogenizing fluid, and the rinse was homogenized and combined with the lysate. The lysates were centrifuged at 10,000 ϫ g for 10 min, and the resulting postmitochondrial supernatant (PMS) fractions were recovered. Five 400-l samples were removed into 5-ml vials on ice. The remainder of each fraction was centrifuged at 100,000 ϫ g for 1 h, the postribosomal supernatant (PRS) fraction was decanted, and five 400-l samplings were taken as above. Both sets of samples were then precipitated with 10% trichloroacetic acid and analyzed for radioactivity by the procedure used for conventional incorporations (35). Samples taken for slope determination were adjusted to comparable 14 C and 3 H counting efficiencies to those for incorporation samples. Samples were analyzed for each isotope in a multichannel scintillation counter, and the data were corrected for channel crossover. Slopes were calculated from the relationship, Average ribosomal transit times were calculated by the relationship, Determination of the Phosphorylation State of the ␣-Subunit of eIF-2-Cells (6 ϫ 10 5 /treatment) were dissolved with sample buffer containing 3.5% ampholines (4 parts pH 4 -8 and 1 part pH 3.5-10), 2% ␤-mercaptoethanol, 0.4% Tween-20, and 12 M urea at 40°C. The preparations were then subjected to slab gel isoelectric focusing in 9.5 M urea at 950 V for 18 h to separate the phosphorylated and nonphosphorylated forms of eIF-2␣ in the pI range of 5-7 (31). Gels were treated with Tris buffer and blotted onto polyvinylidene difluoride membranes under basic conditions. The two species of eIF-2␣, which differ by approximately 0.1 pI unit, were detected by immunoblotting and chemiluminescence (22). Films were scanned by UMAX Magic Scan version 1.3.3. Analysis of the relative amounts of phosphorylated and nonphosphorylated subunit was performed on a Macintosh Quadra 700 computer using the public domain NIH Image program. 2 Measurement of eIF-2␣ Kinase Activity-The preparation of extracts of variously treated NIH-3T3 cells, the methodology for measuring the phosphorylation of endogenous PKR and of exogenous, purified eIF-2 in vitro, and the procedures for immunoprecipitation of PKR were each described previously (27).

Effects of ER and Heat Shock Stressors on Amino Acid
Incorporation and Ca 2ϩ Contents of NIH-3T3 Cells-Acute exposure of various cultured cells to either thiol-reducing or Ca 2ϩmobilizing agents inhibits amino acid incorporation in conjunction with eIF-2␣ phosphorylation and depressed eIF-2B activity, lowered 43 S preinitiation complex, and disappearance of polyribosomal contents (22,23). NIH-3T3 cells were chosen for the current investigation in view of their strong induction of PKR in response to interferon-␣ and the activation of the enzyme by ER stressors (27). On acute challenge, protein synthesis in NIH-3T3 cells was remarkably sensitive to inhibition by such ER stressors as the Ca 2ϩ ionophore ionomycin, to thapsigargin, an irreversible inhibitor of ER Ca 2ϩ accumulation, and to dithiothreitol, a sulfhydryl-reducing agent. All of these agents on longer term exposures (several hours) induce expression of GRP78 and GRP94 but not the HSPs. Protein synthesis in NIH-3T3 cells was also readily inhibited by sodium arsenite and cadmium ion, both of which induce the HSPs but not the GRPs.
During longer term exposures suitable for inducing either HSPs or GRPs, most cell types recover approximately 50 -100% of their original rates of amino acid incorporation. These recoveries depend on new mRNA synthesis. NIH-3T3 cells, however, were unable to restore amino acid incorporation when challenged by ER stressors such as thapsigargin unless fetal bovine serum, PMA, or epidermal growth factor was included in the incubation (not shown). Amino acid incorporation by NIH-3T3 cells also declined sharply (84%) during a 3-h exposure to actinomycin D. When either serum or PMA was added, actinomycin D blocked recovery from thapsigargin inhibition in the predicted fashion. Recovery of amino acid incorporation in cells challenged with arsenite, however, did not depend on the addition of promoters such as PMA. The addition of PMA was necessary, however, to demonstrate that actinomycin D prevented recovery of incorporation on longer-term arsenite treatment. The beneficial effects of PMA in stabilizing rates of amino acid incorporation in longer term incubations of NIH-3T3 cells prompted its judicial use in subsequent experiments.
The ability of sodium arsenite to release NIH-3T3 cell-associated Ca 2ϩ was compared with that of two established releasers, ionomycin and thapsigargin (not shown). Both agents released approximately 40% of cell-associated Ca 2ϩ during incubation periods ranging from 90 min to 4 h. Arsenite did not release Ca 2ϩ or alter the release of Ca 2ϩ occurring in response to either thapsigargin or ionomycin. Cells that were pretreated with arsenite under conditions that would induce HSPs were not altered either in their Ca 2ϩ contents or in their responses to either thapsigargin or ionomycin.
Development by NIH-3T3 Cells of Translational Cross-tolerance to ER and Cytoplasmic Stressors-Cross-tolerance to translational inhibition by ionomycin was sought in NIH-3T3 cells that were accommodated to arsenite. Amino acid incorporation was compared for cells that were either pretreated for 90 min with 150 M sodium arsenite followed by a 2-h recovery period without drug or carried as untreated controls. Leucine pulse labeling during the recovery period indicated that protein synthesis was gradually returning in the arsenite-treated samples after a nearly complete inhibition ( Fig. 1, inset, filled circles). After 30 min in fresh medium, the two sets of samples were challenged with increasing concentrations of sodium arsenite or ionomycin (Fig. 1). Leucine incorporation by cells not previously exposed to arsenite was strongly inhibited by arsenite concentrations ranging from 25 to 50 M and by ionomycin ranging from 30 to 300 nM ( Fig. 1, open circles). Cells previously exposed to arsenite expressed nearly complete translational tolerance to either stressor across these respective concentration ranges as well as to severalfold higher concentrations ( Fig.  1, filled circles).
Cross-tolerance to arsenite was sought in cells that were accommodated to ionomycin. Amino acid incorporation was compared for cells that were either pretreated for 3.5 h with 0.5 M ionomycin or incubated as drug-free controls. Leucine pulse labeling conducted during the first 3 h of this period indicated gradual recovery of protein synthesis in the ionomycin-treated samples following almost complete initial inhibition (Fig. 2, inset, filled circles), whereas the untreated controls exhibited slight declines in activity (inset, open circles). After removal of the ionomycin with albumin washes, the two sets of samples were challenged with increasing concentrations of sodium arsenite or ionomycin (Fig. 2). Incorporation in cells not previously exposed to ionomycin was strongly inhibited by arsenite concentrations ranging from 25 to 100 M and by ionomycin ranging from 30 to 100 nM (Fig. 2, open circles). Cells previously exposed to ionomycin were almost completely tolerant to translational inhibition by either stressor at these respective concentration ranges and markedly tolerant at higher concentrations as well (filled circles). Both sets of cells exhibited comparable degrees of leucine incorporation when incubated without stressors.
It was of interest to ascertain whether NIH-3T3 cells retain responsiveness to ER stressors after the induction of the HSPs and, conversely, whether inducers of the HSPs are active following induction of GRPs. In an initial experiment cells were pretreated with sodium arsenite for 90 min (stage 1), washed and exposed to either thapsigargin or ionomycin for 3.5 h (stage 2), and subsequently exposed to pulse labeling with [ 35 S]methionine (Fig. 3A). Controls without drug were also included for each of the two stages of the incubation, and some samples during the second stage were incubated with actinomycin D to block potential induction of stress proteins during this period. Cells treated with arsenite (lanes g-l) during the first stage of the treatment displayed the typical induction of HSPs of 110, 90, 72, 70, 60, and 30 kDa. These inductions were equivalent for cells treated in the second stage either without Ca 2ϩ mobilizer (lane g), or with thapsigargin (lane h) or ionomycin (lane i) but were reduced sharply by actinomycin D during that period (lanes j, k, and l, respectively). Overall protein synthesis was lowered by actinomycin D in these samples but not abolished. Controls that were not treated with arsenite (lanes a-f) did not induce any detectable HSPs either without Ca mycin D under these respective conditions (lanes e, f, k, and l) or in cells that were never treated with a Ca 2ϩ mobilizer (lanes a, d, g, and j).
The converse experiment was performed in which cells were treated with thapsigargin for 3 h to induce the GRPs (stage 1) and then exposed to sodium arsenite or CdCl 2 for 2.5 h (stage 2) and analyzed for pulse labeling with [ 35 S]methionine (Fig. 3B). As in Fig. 3A, drug-free samples for each stage and actinomycin D controls were included. Cells treated in stage 1 with thapsigargin (lanes g-l) displayed the strong induction of GRP78 and modest induction of GRP94, a later appearing protein. Pulse labeling of these proteins was not altered from the stage 2 control (lane g) by either arsenite (lane i) or Cd 2ϩ (lane k). Actinomycin D (lanes h, j, and l, respectively) reduced, but did not eliminate, pulse labeling of these proteins. No perceptible induction of the HSPs was observed in any sample in response to thapsigargin. Exposure of thapsigargin-pretreated cells (lane g) during stage 2 to either arsenite (lane i) or Cd 2ϩ (lane k) induced the full range of HSPs. Qualitatively comparable HSP inductions were noted for similarly treated control cells that had not received thapsigargin during stage 1 (lanes a, c,  and e, respectively). Actinomycin D blocked the stage 2 induction of the HSPs by either arsenite (lanes d and j) or by Cd 2ϩ (lanes f and l). It seems clear from these results (Fig. 3, A and  B) that, despite the cross-tolerance of protein synthesis developing in response to ER and heat shock stressors ( Figs. 1 and  2), the HSP and GRP stress proteins remain independently inducible.
Selective and Reversible Inhibition of Translational Initiation by ER Ca 2ϩ -mobilizing Drugs and Sodium Arsenite-The ability of sodium arsenite to inhibit translational initiation was compared with that of thapsigargin and ionomycin. Lysates derived from NIH-3T3 cells treated acutely for 45 min with 300 nM ionomycin (Fig. 4, upper panel, lane b), 30 nM thapsigargin (lane c), or 150 M sodium arsenite (lane d) displayed a comparable accumulation of ribosomal subunits and a disappearance of polyribosomal contents as compared with an untreated control (lane a). These alterations were completely reversible for each substance following the addition of an inhibitor of peptide chain elongation, cycloheximide, for an additional 15 min (Fig. 4, lower panel). Longer term incubations with either 150 M sodium arsenite or with 0.5 M ionomycin resulted in the development of cross-tolerance to polyribosome depletion by either agent (not shown). Cells incubated without stressor remained sensitive to polyribosome depletion when challenged with either ionomycin or arsenite, whereas the polyribosome contents of cells that had been pretreated with either arsenite or ionomycin were not affected upon challenge with either agent. While these results indicated that both arsenite and Ca 2ϩ -mobilizing agents exert a similar rate-limiting inhibition of translational initiation that was removed during the induction of either GRP78 or the HSPs, collateral inhibitions of high arsenite concentrations on peptide elongation were not completely excluded. Determinations of average ribosomal transit times were therefore conducted (Table I). This procedure, which measures the time to complete an average polypeptide, allows inhibitory actions at elongation to be assessed independently of inhibitory effects occurring at initiation (37). Exposure of cells to 75 M or 125 M sodium arsenite provided average ribosomal transit times of approximately 2 min in accord with both untreated control and ionomycin-treated cells. Cycloheximide (250 nM) approximately tripled this value.
The inhibition of translational initiation in response to Ca 2ϩmobilizing agents has been previously found to depend on the phosphorylation of eIF-2␣ (22). NIH-3T3 cells and GH 3 pituitary cells were therefore compared for phosphorylation of eIF-2␣ in response to ionomycin or to various agents that induce the HSPs, including arsenite, two heavy metal cations, and two agents that generate oxidative free radicals (Fig. 5). Each of these agents promoted eIF-2␣ phosphorylation to varying extents that were sufficient for substantial inhibition of protein synthesis. While fairly high degrees of eIF-2␣ phosphorylation can be achieved from one inhibitor to another, it should be noted that only 20 -30%, depending on cell type, is ordinarily required for complete inhibition of translational initiation (22,26,38,39). This relationship is highlighted by an experiment detailing the increasing degrees of phosphorylation of eIF-2␣ and of translational inhibition in NIH-3T3 cells in response to increasing concentrations of ionomycin or sodium arsenite (Table II). Maximal phosphorylation was achieved with a combination of ionomycin and arsenite. NIH-3T3 cells that were pretreated with either arsenite or ionomycin became refractory to eIF-2␣ phosphorylation upon rechallenge with either agent. For example, cells pretreated with 150 M sodium arsenite for 2 h followed by a 2-h recovery period were refractory to eIF-2␣ phosphorylation in response to the concentrations of ionomycin or arsenite employed above (Table II). It was clear that the recovery from the initial arsenite treatment was incomplete, since all of the samples retained some residual phosphorylation and displayed reduced protein synthesis with respect to the untreated control. The effect of ionomycin pretreatment on subsequent eIF-2␣ phosphorylation in response to rechallenge with arsenite or ionomycin was examined in a separate experiment (Table III). Acute treatment with ionomycin, arsenite, or both agents increased the phosphorylation of eIF-2␣ with respect to untreated controls. In contrast, cells that were pretreated with ionomycin, washed, and allowed to recover were resistant to eIF-2␣ phosphorylation and translational suppression upon comparable rechallenge.
Activation of PKR by Sodium Arsenite-We previously reported that depletion of ER Ca 2ϩ from NIH-3T3 cells by either Ca 2ϩ ionophore or thapsigargin activates an eIF-2␣ kinase identified as PKR (27). This activation occurs in intact cells but is retained upon subsequent homogenization; it is not generated by treatment of lysates with these agents. Activation is particularly prominent after pretreatment of the cells with interferon-␣, a well established inducer of PKR. Extracts of NIH-3T3 cells, when incubated with [␥-32 P]ATP, display autophosphorylation of PKR and phosphorylation of eIF-2␣, each of which is amplified by the addition of viral dsRNA and purified eIF-2 to the incubation. NIH-3T3 cells cultured with interferon were therefore used to test the possibility that sodium arsenite causes PKR to become activated (Fig. 6). Cells were exposed to 150 M sodium arsenite for 15 (lanes b and e) or 30 min (lanes c and f) or incubated for 30 min as untreated controls (lanes a and d). Extracts derived from arsenite-challenged cells were found to phosphorylate both PKR, as determined following immunoprecipitation with an antibody to the enzyme (Fig. 6A), and eIF-2␣ (Fig. 6B). The phosphorylation of each of the two proteins was greater in cells treated with arsenite for 30 min than for 15 min. While some phosphorylation of each protein was evident without the addition of dsRNA to the incubation (lanes a-c), phosphorylation was greatly increased by this addition (lanes d-f) and was evident to some extent even for lysates derived from cells that were not exposed to arsenite. When cells cultured without interferon were subjected to identical protocols, phosphorylations of PKR or eIF-2␣ in response to dsRNA addition or arsenite treatment were barely detectable (not shown).  or 1 mM diamide (lane h, upper panel only). After 30 min, lysates were prepared and subjected to slab gel isoelectric focusing followed by immunoblotting for eIF-2␣. Arrows indicate the migration positions of the phosphorylated (eIF-2␣(P)) and nonphosphorylated (eIF-2␣) subunits.

Effects of ER and Heat Shock Stressors on Amino Acid In
in catalytic subdomain II (28) were utilized to provide additional evidence that activation of PKR mediates translational suppression by sodium arsenite. Extracts of wild-type and mutant NIH-3T3 cells cultured in the absence or presence of interferon-␣ were subjected to SDS-PAGE and Western blot analysis using a PKR antibody recognizing both murine and human forms of the kinase (Fig. 7A). The endogenous kinase (64 kDa) was detectable in both cell lines only after interferon pretreatment. By contrast, a cross-reactive species migrating at the position of human PKR (68 kDa) was readily detected in the mutant line regardless of culture with interferon. The translational activities of wild-type and mutant NIH-3T3 cells challenged with increasing concentrations of ER and heat shock stressors were then compared (Fig. 8). Although culture of wild-type NIH-3T3 cells with interferon suppressed leucine incorporation in the absence of arsenite, sensitivity to arsenite was not altered by the cytokine (Fig. 8A). When compared with wild-type NIH-3T3 cells, amino acid incorporation in the mutant cells was generally two to three times greater and was significantly less sensitive to inhibition by arsenite, ionomycin, dithiothreitol, or t-butylhydroperoxide (Fig. 8, B-E). Culturing of mutant cells with interferon to increase expression of native PKR increased the sensitivity of protein synthesis to arsenite without suppressing basal activity (Fig. 8B). When compared with wild-type NIH-3T3 cells, the mutant cells also exhibited lesser phosphorylations of eIF-2␣ when challenged with two different concentrations each of ionomycin, dithiothreitol, arsenite, or t-butylhydroperoxide (Fig. 7, B and C). DISCUSSION A broad body of literature exists pertaining to stress-induced proteins and their respective roles as protein chaperones. Frequently the GRPs are considered to be members of the overall

FIG. 7. Induction of expression of endogenous PKR by interferon-␣ and phosphorylation of eIF-2␣ by ER and heat shock stressors in wild-type NIH-3T3 cells and cells expressing a dominant negative mutant PKR.
A, effect of interferon on expression of PKR. Wild-type cells and those that overexpress a dominant negative human PKR mutant (K296P in catalytic subdomain II, clone KP 3 A) were cultured for 24 h with interferon-␣ (IFN␣, 1000 units/ml). Lysates were subjected to SDS-PAGE (12.5%) followed by immunoblotting with a polyclonal antibody recognizing both murine and human PKR. Arrows indicate the migration positions of the murine and human kinases. The asterisk indicates an unidentified species present in all samples and that serves as substrate for the enzyme conjugated to the secondary antibody. B and C, eIF-2␣ phosphorylation in NIH-3T3 and KP 3 A cells following treatments with ER and heat shock stressors. NIH-3T3 (B) and KP 3 A (C) cells were treated for 30 min without drug (1), with 0.1 or 0.3 M ionomycin (2 and 3), with 0.2 or 0.6 mM dithiothreitol (4 and 5), with 50 or 150 M sodium arsenite (6 and 7), or with 2 or 6 mM t-butylhydroperoxide (8 and 9). Lysates were subjected to slab gel isoelectric focusing followed by immunoblotting for eIF-2␣. Arrows indicate the migration positions of the phosphorylated (eIF-2␣(P)) and nonphosphorylated (eIF-2␣) subunits.
HSP family in view of the close sequence homologies that exist between GRP78 and HSP70/HSC72 and between GRP94 and HSP90 (3)(4)(5). The two sets of stress proteins, however, are induced in response to different chemicals and conditions in mammalian cells and localize to different subcellular compartments. In this report we have chosen to view ER stressors as those perturbants that inhibit translation and ER protein folding or processing while subsequently inducing the ER resident chaperones, GRP78 and GRP94. The ER stress response system is activated by Ca 2ϩ -mobilizing or thiol-reducing agents. In contrast, those perturbants that inhibit translation and protein folding in the cytoplasm while inducing the HSPs are viewed as cytoplasmic or "heat shock" stressors. The heat shock stress response system is activated by oxidative chemicals and heavy metal ions. Interdigitation of these two closely related systems is of considerable interest. In this report we have utilized Ca 2ϩ ionophore and thapsigargin as representative activators of the ER stress system. As activators of the cytoplasmic stress system we employed the sulfhydryl-inactivating agents sodium arsenite and, to a lesser extent, cadmium ion. Sodium arsenite offered the advantage of producing reproducible, strong heat shock responses without the widespread collateral actions and cell death developed by higher temperatures or strong oxidants.
Previous reports have emphasized that eIF-2␣ phosphorylation is increased by a variety of HSP inducers to uneven degrees, ranging from pronounced phosphorylations with arsenite to low to high phosphorylations with heat shock to marginal phosphorylations with iodoacetamide and various amino acid analogs (29 -32). The relative contributions of initiation and elongation blockade to the overall degree of translational inhibition by these chemicals has been unclear. Translation is also suppressed in some cell types by serum removal in conjunction with eIF-2␣ phosphorylation (29). Serum depletion almost certainly also slows translational elongation, since this condition is associated with dephosphorylation of various elongation factors (40).
It is evident from the present data and those of previous reports (27,28) that perturbants of either the ER/GRP or cytoplasmic/HSP chaperone systems produce an immediate activation of the eIF-2␣ kinase, PKR, and the phosphorylation of eIF-2␣ in conjunction with the inhibition of translational initiation. Phosphorylation of eIF-2␣ is well established to inhibit eIF-2B, the GTP/GDP exchange factor required for recycling of eIF-2 during initiation (29). Once sufficient eIF-2 is phosphorylated to complex eIF-2B, additional degrees of phosphorylation are superfluous, although in extreme circumstances almost the entire eIF-2␣ pool can be phosphorylated (24). Sodium arsenite inhibited translational initiation without affecting peptide chain elongation. Polyribosomal contents almost totally disappeared from NIH-3T3 cells, and ribosomal subunits were correspondingly increased (Fig. 4). The addition of cycloheximide such that elongation once again became rate-limiting restored polysomal contents. This reversibility pointed to a slowing of initiation without the occurrence of physical damage to the translational apparatus. More importantly, average ribosomal transit times were not lengthened detectably by sodium arsenite (Table I), allowing us to conclude that elongation was not affected even by relatively high arsenite concentrations that abolished amino acid incorporation. All of these effects were faithfully reproduced by ionomycin and thapsigargin, which have been found repeatedly to inhibit translational initiation selectively in a variety of cell types (23,25).
The induction of either the GRPs or the HSPs over several hours in NIH-3T3 cells was accompanied by a reduced phosphorylation of eIF-2 (Tables II and III) and a partial resumption of mRNA translation (Figs. 1 and 2). These data provided further evidence that the acute inhibition of initiation derived from the phosphorylation of eIF-2␣. Induction of either class of stress proteins was associated with the development of translational tolerance to subsequent rechallenge of the cells with either ER or heat shock stressors. Tolerance was observed in terms of continued amino acid incorporation (Figs. 1 and 2), maintenance of polyribosomal contents, and the lack of increased eIF-2␣ phosphorylation (Tables II and III). Translational recovery from inhibition by Ca 2ϩ ionophores is partly overturned by antisense oligonucleotides directed against grp78 mRNA (41). Presumably, various GRP and HSP chaperones inhibit PKR through complexing with a critical component(s) of the enzyme or with other protein(s), which affect its activity. If so, the sequence homology of GRP78 and HSP70 may be important to the putative dual input. PKR is thought to be a complex enzyme that interacts with various inhibitory proteins, responds to various activators and inducers, and possesses at least one other substrate, I-B (42,43). A closely related eIF-2␣ kinase of erythroid cells (heme-regulated eIF-2␣ kinase) is reported to exist in association with HSP70 and HSP90 (44). Not all stresses affect PKR comparably. PKR is found in the soluble fraction of lysates derived from cells treated with sodium arsenite but in the insoluble fraction of heat-shocked cells (45). In our experiments with arsenite-and ionomycin-treated cells, PKR was found in the soluble fraction of cell lysates containing Triton X-100.
With the exception of their mutual abilities to influence the activity of PKR and translational initiation, the two stress systems appear to operate independently. For example, sodium arsenite did not induce detectable amounts of the GRPs (Fig.  3), nor did it affect ER function in NIH-3T3 cells. The cells maintained their Ca 2ϩ contents during extended treatment with arsenite and, upon Ca 2ϩ depletion with either ionomycin or thapsigargin, readily induced GRP78 superimposed upon the preexisting HSP induction by arsenite (Fig. 3). Similarly, following the induction of GRP78 with Ca 2ϩ -mobilizing agents, the cells remained responsive to HSP induction by subsequent arsenite treatment. It should be recognized that the phosphorylation of eIF-2␣ and inhibition of translation do not appear mandatory for the subsequent induction of stress proteins even in unstressed cells. Low concentrations of ER stressors clearly induce GRP78 in GH 3 cells in the absence of eIF-2 phosphorylation or inhibition of translation (26).
The body of available information pertaining to eIF-2␣ kinase activities suggests that PKR may function as a common focal point for governing rates of translational initiation in response to a variety of stimuli including, but not limited to, viral infections, ER stress, and cytoplasmic proteotoxic stress. Currently only two mammalian eIF-2␣ kinases are known to exist: heme-regulated eIF-2␣ kinase, which is expressed selectively by erythroid cells (46), and PKR, which is ubiquitous to all mammalian cell types. The lack of other eIF-2␣ kinase activities and the activation of PKR by stressors suggests that the enzyme may mediate most, if not all, eIF-2␣-dependent inhibitions of translational initiation in higher eukaryotes. Putative additional stimuli for PKR activation could include hormonal or nutritional alterations or treatments that damage the plasmalemma. The possibility that PKR possesses multiple substrates and/or serves more broadly in cellular control mechanisms is also supported by various recent findings. For example, PKR activates specific gene transcription through NF-Bdependent (41,42) and -independent (47) mechanisms and is strongly implicated in the control of differentiation and growth (48). The emerging picture suggests that the biochemical structure and regulation of the enzyme should prove both interesting and informative.