The direct binding of the catalytic subunit of protein phosphatase 1 to the PKR protein kinase is necessary but not sufficient for inactivation and disruption of enzyme dimer formation.

The PKR protein kinase is among the best-studied effectors of the host interferon (IFN)-induced antiviral and antiproliferative response system. In response to stress signals, including virus infection, the normally latent PKR becomes activated through autophosphorylation and dimerization and phosphorylates the eIF2alpha translation initiation factor subunit, leading to an inhibition of mRNA translation initiation. While numerous virally encoded or modulated proteins that bind and inhibit PKR during virus infection have been studied, little is known about the cellular proteins that counteract PKR activity in uninfected cells. Overexpression of PKR in yeast also leads to an inhibition of eIF2alpha-dependent protein synthesis, resulting in severe growth suppression. Screening of a human cDNA library for clones capable of counteracting the PKR-mediated growth defect in yeast led to the identification of the catalytic subunit (PP1(C)) of protein phosphatase 1alpha. PP1(C) reduced double-stranded RNA-mediated auto-activation of PKR and inhibited PKR transphosphorylation activities. A specific and direct interaction between PP1(C) and PKR was detected, with PP1(C) binding to the N-terminal regulatory region regardless of the double-stranded RNA-binding activity of PKR. Importantly, a consensus motif shared by many PP1(C)-interacting proteins was necessary for PKR binding to PP1(C). The PKR-interactive site was mapped to a C-terminal non-catalytic region that is conserved in the PP1(C)2 isoform. Indeed, co-expression of PP1(C) or PP1(C)2 inhibited PKR dimer formation in Escherichia coli. Interestingly, co-expression of a PP1(C) mutant lacking the catalytic domain, despite retaining its ability to bind PKR, did not prevent PKR dimerization. Our findings suggest that PP1(C) modulates PKR activity via protein dephosphorylation and subsequent disruption of PKR dimers.

Eukaryotic cells generally down-regulate protein synthesis in response to stress conditions presumably to protect against the harmful effects of toxic agents, to conserve resources that are needed to survive under adverse conditions, or to activate apoptosis (1). A major control mechanism for this cellular stress response involves protein phosphorylation of the ␣ subunit of the translation initiation factor 2 (eIF2␣) on serine 51 (reviewed in Ref. 2). When bound to GTP, eIF2 promotes the assembly of the translation initiation complex between Met-tRNA i and the 40 S ribosomal subunit, a process that results in GTP hydrolysis and an eIF2-GDP complex. Phosphorylation of eIF2␣ subverts the recycling step required for the formation of an active eIF2-GTP complex, thereby reducing the rate of mRNA translation initiation and, ultimately, an inhibition of global cellular protein synthesis. At least four structurally related serine/threonine protein kinases, each responding to specific stress stimuli, phosphorylate eIF2␣ (reviewed in Ref. 3): the yeast GCN2 kinase, activated by amino acid starvation; the reticulocyte-specific HRI kinase, activated by heme depletion; the endoplasmic reticulum-associated PERK/PEK kinase, activated by stresses that impair protein folding in the endoplasmic reticulum; and the interferon (IFN) 1 -inducible PKR serine/threonine kinase, activated primarily by virus infection.
The PKR protein kinase is one of the few well characterized IFN-induced gene products that directly mediate the antiviral effects of IFNs (reviewed in Ref. 4). PKR is ubiquitously expressed but is normally inactive, presumably because the ATPbinding site or the catalytic domain of PKR is masked by intramolecular interactions (5,6). Upon binding to dsRNA, or to RNA with secondary structures similar to viral replicative intermediates, PKR is autophosphorylated on multiple serine and threonine residues, which may induce a conformational change that leads to the disclosure of the ATP-binding site and/or the catalytic domain. This is followed by PKR dimerization, which is thought to promote the intermolecular autophosphorylation of PKR molecules, resulting in maximal activation of the enzyme (7)(8)(9)(10)(11). Binding to dsRNA may also serve to recruit PKR molecules to the ribosomes for localized action, where phosphorylation of eIF2␣ by PKR leads to a block in global protein synthesis, ultimately limiting virus replication within the infected cell (11,12). The important role of PKR in host innate immunity is underscored by the numerous strategies employed by different viruses to antagonize PKR (4). Furthermore, mice devoid of functional PKR display increased susceptibility to infection by some viruses (13)(14)(15)(16)(17)(18).
Activation of PKR can also lead to apoptosis (4). The translational inhibitory and pro-apoptotic properties of PKR have led to the suggestion that PKR may be a tumor suppressor. Indeed, overexpression of PKR is growth-suppressive in insect, yeast, and mammalian cells (19 -21), whereas overexpression of dominant-negative forms of PKR, or cellular or viral inhibitors of PKR, leads to malignant transformation of NIH 3T3 cells (22)(23)(24)(25)(26). Although the exact mechanisms are still not clear, PKR may function through its ability to regulate transcription factors NFB (27,28), STAT1 (29,30), and the tumor suppressor p53 (31,32). Accordingly, PKR activity should be tightly modulated in the cell. While viral studies have revealed different strategies for PKR countermeasures, including the inhibition of dsRNA-mediated activation of PKR, the interference with PKR dimerization process, and the degradation of PKR protein (4), the regulation of PKR by post-translational protein modifications in uninfected cells is poorly understood.
Of particular interest to this study is the modulation of PKR by reversible protein phosphorylation, which is widely used for rapid signal desensitization of enzymes and central to many signal transduction pathways.
Previous studies have suggested that a type 1-protein phosphatase (PP1) may be responsible for the inactivation of PKR (33), whereas a type 2 PP (PP2) is thought to regulate HRI activity (34). However, the exact protein phosphatases involved have not been determined. To gain insights into the cellular mechanisms of PKR regulation, we undertook experiments to identify negative regulators of PKR. To this end, we used a yeast-based functional assay for PKR to screen a human cDNA expression library for clones capable of repressing PKR activity. This screen yielded the catalytic subunit of type 1 protein phosphatase (PP1␣ or PP1 C ). Using various in vitro and in vivo assays, we verified the ability of PP1 C to inhibit PKR and further demonstrated a specific and direct interaction between the two proteins. Co-expression of PP1 C interfered with PKR dimerization, whereas a catalytically inactive mutant PP1 C did not. Our results suggest a potential mechanism for tight control of PKR activity and in turn points to a role for PP1 C in translational control.
Yeast Transformation and Library Screening-A human cDNA expression library in YES-R (a gift from Dr. S. J. Elledge) was used to infect Escherichia coli BNN132, and plasmid DNA was prepared as described by Elledge et al. (35). Purified library plasmid DNA (500 g) was transformed into RY1-1 cells by the LiAc method as described in the CLONTECH manual. An aliquot of the transformation mixture was plated on 2% raffinose-containing synthetic defined (SD) agar plates lacking uracil (Ura) to determine the transformation efficiency. Approximately 1.5 ϫ 10 6 cells were plated directly on SD-Ura agar plates containing 10% galactose and 2% raffinose. Colonies were picked after 3 to 5 days of incubation at 30°C.
Library plasmids were extracted from the colonies and retransformed into fresh RY1-1 cells to confirm the growth-suppressive rescue phenotype. cDNA inserts from positive clones were then PCR-amplified and categorized based on their restriction enzyme (HaeIII) digestion patterns. A representative cDNA insert from each group was subsequently sequenced using the oligonucleotide 5Ј-ACTTTAACGTCAAG-GAG-3Ј for reading from the GAL1 promoter.
GST-mediated Co-sedimentation Assay-The GST-PKR fusion construct was obtained from Dr. B. R. G. Williams. Dr. A. E. Koromilas kindly provided the GST-PKR K296R, GST-PKR 1-262, and GST-PKR 263-551 constructs. GST-PP1 C 2 and GST-PP1 C 2 181-342 were obtained from Dr. T. Durfee. GST-PP1 C was constructed by inserting a 0.9-kb PCR DNA fragment containing the entire coding sequence of human PP1 C , except the first methionine codon, into the EcoRI and SalI sites of pGEX-4T-3 (Amersham Biosciences). The PCR fragment was amplified from the plasmid pRB4891 (provided by Dr. B. He) using the oligonucleotide primers 5Ј-GCACTGAATTCTCCGACAGCGAGAAGC-TCAAC-3Ј and 5Ј-GCACTGTCGACATCTGGGGCACAGGGTGGTG-T-3Ј (restriction sites are underlined). Purification of GST fusion proteins, co-sedimentation, and immunoblotting were carried out as previously described (36). Bound antibodies were visualized using an enhanced chemiluminescence (ECL) detection system (Amersham Biosciences).
Yeast Two-hybrid System-The two-hybrid plasmids pGBT9 and pGAD424 (CLONTECH) were used for the expression of GAL4 DNAbinding domain (GAL4 BD ) and GAL4 transcriptional activation domain (GAL4 AD ) fusions, respectively. pGBT9 and pGAD424 contain the selectable auxotrophic markers, TRP1 and LEU2, respectively. GAL4 AD fusions containing different PKR fragments or point mutants were described earlier (37,38). The PKR mutant, Y167A, in which Tyr-167 was replaced with Ala, was generated by overlap extension PCR using GAL4 AD PKR K296R as a template. GAL4 AD PKR 1-220 (K64E) and GAL4 BD PP2A C were gifts from Dr. G. Sen and Dr. B. Hemmings, respectively. Dr. R. Jagus and Dr. J. Printen provided GAL4 BD K3L and GAL4 BD PP1 C , respectively. GAL4 AD SV40 T Ag and GAL4 BD P53 were purchased from CLONTECH. The yeast strain Hf7c, which carries the HIS3 reporter fused to a GAL4 promoter sequence, was used to assay for protein-protein interactions as described (37).
Repressor Fusion Assay-The assay was performed using the N-PKR-K296R construct as described previously (10,38). Plasmids encoding GST-PP1 C fusions were described above. PC168-derived plasmids encoding the repressor N-terminal DNA-binding domain (N) fused with PKR K296R and pGEX2T-derived plasmids encoding GST alone or GST fused with the indicated PP1 C proteins were co-transformed into E. coli AG1688 (obtained from Dr. J. C. Hu). Co-transformants were selected on LB plates containing 50 g of ampicillin and 20 g of chloramphenicol/ml. Cultures were grown overnight at 30°C in LB supplemented with antibiotics, 10 mM MgSO 4 , and 0.2% maltose and used to create bacterial lawns on agar containing 100 nM isopropylthio-␤-D-galactoside. Lawns were then spotted with 5-l aliquots of serial dilutions of a KH54 phage lysate (10 9 plaque-forming unit) at 10-fold intervals. Infected lawns were incubated overnight at 30°C, and the inhibition of dimerization mediated by N-PKR K296R fusion was scored by the appearance of dot plaques on the lawns.
Co-immunoprecipitation Studies-The human hepatoma Huh7 cells were lysed in buffer containing 20 mM Tris-HCl (pH 7.5), 50 mM KCl, 50 mM NaCl, 0.25% Triton X-100, 20% glycerol, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 1 mM sodium orthovanadate, and 1ϫ Complete Protease Inhibitor Mixture (Roche Molecular Biochemicals). Protein lysates (1 mg, as determined by Bradford Dye Assay; Bio-Rad) were incubated with 2 g of monoclonal anti-PKR antibody (provided by Dr. T. E. Dever) or 25 g of normal mouse serum (Jackson ImmunoResearch) at 4°C, rotating for 2 h in a final volume of 600 l of lysis buffer containing 0.1% Triton X-100. Protein-G-agarose beads (50 l; Roche Molecular Biochemicals) were added to the reaction, which was incubated for an additional 1 h at 4°C with rotation. The beads were pelleted by centrifugation and washed three times with lysis buffer containing 0.25% Triton X-100. The bead pellet was boiled in equal volume of 1ϫ Laemmli sample buffer for 5 min. Immunoprecipitated complexes were resolved by SDS-PAGE (12%). Proteins were transferred onto nitrocellulose membrane and immunoblotted with either anti-PP1 C (provided by Dr. K. Schlender) or anti-PKR polyclonal primary antibody (19), followed by donkey anti-rabbit secondary antibody (Jackson ImmunoResearch). Proteins were visualized by ECL (Amersham Biosciences) and autoradiography of the immunoblots.
Functional Assays for PKR Activity-For PKR in vitro kinase assays, native PKR protein, affinity-purified from IFN-treated human 293 cells, was used (39). GST-PP1 C and GST-Tat (obtained from Dr. M. Mathews) proteins were purified as described above. Quantitative SDS-PAGE with bovine serum albumin as a standard was used to determine the concentration of all purified proteins. In vitro kinase reactions were performed essentially as described (10), except that PKR (0.3 g) was incubated with increasing amounts (0.1, 0.2, and 0.4 g) of purified rabbit PP1 C (from Upstate Biotechnology) in a 30-l phosphatase reaction buffer (Upstate Biotechnology) at 30°C for 30 min. When indi-cated, rabbit PP1 C , the amino acid sequence of which is identical to human PP1 C , was inactivated by the addition of Inhibitor-2 (Upstate Biotechnology) according to the manufacturer's instructions. PKR was immunoprecipitated from the reaction mixture using a PKR-specific monoclonal antibody (71/10; Ribogene) and subjected to kinase reaction containing [␥-32 P]ATP (10 Ci) and histone H1 (10 g) in the presence or absence of poly(I⅐C) (3 g/ml). The reaction was terminated by the addition of 30-l 2ϫ sample buffer (100 mM Tris-HCl (pH 6.8), 4% SDS, 20% glycerol, 10% ␤-mercaptoethanol, and 0.2% bromphenol blue). One half of each sample (30 l) was subjected to SDS-PAGE (14%), followed by autoradiography. To determine PKR activity in vivo, eIF2␣ phosphorylation within RY1-1 yeast cells harboring various expression plasmids was determined by isoelectric gel focusing and immunoblot analyses as described by Gale et al. (40). A rabbit polyclonal antiserum specific to yeast eIF2␣ (a generous gift from Dr. T. E. Dever) was used to detect eIF2␣ by immunoblot analysis. Construction of pYX233-PP1 C was achieved by subcloning a 1-kb EcoRI-SalI DNA fragment from GAL4 BD PP1 C into pYX233 (Novagen) linearized with EcoRI and XhoI. Dr. P. R. Romano provided plasmid p1470, which expresses PKR K296R. The relative levels of protein phosphorylation were determined by quantifying the immunoblots using a Molecular Dynamics Phosphor-Imager and ImageQuant software (version 5.1).
In Vitro Phosphatase Assay-Affinity-purified PKR proteins (0.15 g) were subjected to autophosphorylation in 30 l of kinase reaction buffer as described above. The reaction mixture was subjected to chromatography using ProbeQuant G-50 Micro Columns according to the manufacturer's instructions, except that the columns were equilibrated with the phosphatase reaction buffer (Upstate Biotechnology) to desalt PKR proteins and to remove free [␥-32 P]ATP. Labeled PKR was incubated with 0.1 or 0.2 g of purified rabbit PP1 C (Upstate Biotechnology) in the phosphatase reaction buffer at 30°C for 30 min. The reaction was terminated, and the samples were subject to SDS-PAGE (14%) and autoradiography as described above.
DNA Sequence Analysis-All DNA constructs were sequenced by the fluorescent dye-terminator method using an Applied Biosystems Model 377 Automated Sequencer (University of Washington). DNA strider and the Wisconsin GCG package (Madison, WI) were used for DNA sequence analysis.

RESULTS
Identification of PP1 C as an Inhibitor of PKR-To identify novel cellular inhibitors of PKR, we adopted a genetic screening strategy using S. cerevisiae. Overexpression of human PKR protein in yeast cells is lethal because constitutive hyperphosphorylation of eIF2␣ by PKR leads to severe inhibition of mRNA translation (20). We reasoned that co-expression of mammalian genes that negatively regulate PKR would suppress the PKR-induced lethality. We used a gcn2⌬ yeast strain (RY1-1), which carries two copies of the human PKR allele under the control of a galactose-inducible promoter (9). We chose this strain because PKR protein expression is suppressed when the cells are grown in glucose-containing medium, thus allowing the cells to grow normally. Upon transfer to galactosecontaining medium, which induces the GAL promoter and hence PKR expression, these yeast cells cease to grow. We assumed that reversion to normal growth because of mutations would be minimal because the high expression levels of PKR should block cell division, a prerequisite for the generation of mutants. To this end, a galactose-inducible human cDNA expression library was introduced into RY1-1, and transformants that overcame the PKR-mediated growth-inhibitory effect were selected. Plasmids containing cDNAs were extracted from these transformants and retransformed into fresh RY1-1 to verify the reversal of growth arrest phenotype. One of the cDNA clones that reproducibly restored RY1-1 viability on galactose-containing medium encoded the catalytic subunit of type 1 protein phosphatase, PP1 C (41). As shown in Fig. 1A, RY1-1 cells grew normally on raffinose medium, on which PKR expression was suppressed (left panel). However, RY1-1 cells were unable to grow when PKR expression was induced on galactose medium (right panel, lane 1). Coexpression of PP1 C partially rescued the growth defect of RY1-1 (right panel, lane 2).
Because relatively little is known about the cellular mechanisms controlling PKR activity, we chose to investigate the mode of action of PP1 C . PP1 C could conceivably restore RY1-1 growth on galactose-containing media through pathways that are independent of PKR. To confirm that the effect of PP1 C was at least in part specific to PKR, we took note of previous observations that the catalytically inactive PKR K296R is recessive to wild-type PKR in yeast (9). We thus reasoned that co-expression of PKR K296R in RY1-1 might reverse the inhibitory effect of PP1 C on PKR by sequestering PP1 C in nonfunctional PKR K296R-PP1 C complexes. Alternatively, PKR K296R might dimerize with PKR, and because these dimers would be functional in yeast, they could titrate out the PP1 C . As predicted, the PKR-mediated toxicity was partially restored when PKR K296R was coexpressed with PP1 C (Fig. 1A; right panel, lane 3). PKR proteins in these samples were expressed to comparable levels as shown by immunoblot analysis using a PKR-specific antibody (Fig. 1B). The more intense PKR band in lane 3 presumably represents the co-migrating wild-type PKR and PKR K296R. These results support our hypothesis that the PP1 C rescues RY1-1 from cell growth retardation, at least in part, via the PKR pathway.
PP1 C Inhibits PKR Auto-phosphorylation-To examine whether PP1 C directly dephosphorylates PKR, we took advantage of the fact that PKR is hyperphosphorylated when expressed in yeast (9). Treatment of protein extracts with excess protein phosphatase converted the slower-migrating PKR protein band to a faster-migrating band (lane 2), indicating that the difference of migration was due to differential phosphorylation of PKR ( Fig. 2A). Importantly, co-expression of PP1 C with PKR resulted in a similar hypophosphorylated form of PKR (lane 3). Moreover, as a control, expression of the catalytically inactive PKR K296R also produced a fastermigrating band, indicating that PP1 C dephosphorylates PKR (lane 4). Further support for this was obtained by the experimental results shown in Fig. 2B. Purified PKR proteins were first labeled by autophosphorylation in the presence of [␥-32 P]ATP and poly(I⅐C) and then purified and used as a substrate for PP1 C in a phosphatase assay as described under "Experimental Procedures." Indeed, PP1 C could dephosphorylate PKR in vitro in a dose-dependent manner (Fig. 2B, top  panel). Western blotting showed that the PKR dephosphoryl- ation was not due to degradation of PKR protein (Fig. 2B,  bottom panel).
Certain viruses have evolved pseudosubstrates that compete with eIF2␣ for phosphorylation by PKR (4). PP1 C could therefore conceivably function as a "substrate" inhibitor of PKR. We thus tested whether PP1 C is a potential substrate for PKR in the in vitro kinase assay. Consistent with previous findings that the HIV Tat protein is a substrate for PKR (42), we found that PKR phosphorylated GST-Tat (Fig. 2C, lane 1, bottom  arrow). This is not due to phosphorylation of the GST tag because GST alone was not phosphorylated by PKR (lane 2). In contrast, PKR did not phosphorylate GST-PP1 C (lane 3). The lack of phosphorylation of GST-PP1 C might be due to altered protein conformation of PP1 C resulting from fusion to GST. However, this scenario is unlikely because the recombinant GST-PP1 C , but not GST-Tat (compare lanes 1 and 3 in top panel; top arrow), was capable of dephosphorylating PKR, indicating that the PP1 C fusion was properly folded to be active. Furthermore, all GST fusion and PKR proteins were detected by immunoblot analysis using an anti-GST antibody (middle panel) and an anti-PKR antibody (bottom panel), respectively. These results collectively demonstrate that PP1 C antagonizes PKR function by directly dephosphorylating the protein kinase.
PP1 C Inhibits PKR Substrate Phosphorylation-To confirm the functional significance of the PKR inhibitory effect of PP1 C , we used isoelectric focusing to measure the phosphorylation level of eIF2␣, the physiological substrate of PKR, in RY1-1 cells (9). In this assay, PKR-phosphorylated eIF2␣ can be distinguished from the non-phosphorylated form by its gel mobility pattern. As previously reported, eIF2␣ was not phosphorylated in the parental gcn2⌬ yeast strain lacking PKR (Fig. 3A, no PKR). In contrast, induced expression of PKR led to hyperphosphorylation of eIF2␣ (Fig. 3A, PKR). When PP1 C was coexpressed, a reduction of eIF2␣ hyperphosphorylation was observed (PKR ϩ PP1 C ), which is comparable to that caused by coexpression of a PKR dominant-negative mutant (PKR⌬7). Although the majority of eIF2␣ remained in the hyperphosphorylated state, this is in agreement with previous results that relatively small changes (15-20%) in the overall level of eIF2␣ phosphorylation can have dramatic effects on cell growth (9,40). Taken together, these results support the notion that PP1 C is capable of antagonizing PKR function in vivo.
We next used purified components in an in vitro kinase assay (43) to determine whether PP1 C could directly inhibit the ability of PKR to phosphorylate H1 histones. We chose to use H1 histones in this assay because PP1 C is capable of dephosphorylating eIF2␣ (44), but not H1 histones in vitro. PKR proteins were preincubated with different amounts of recombinant PP1 C prior to activation by poly(I⅐C) as described under "Experimental Procedures." A PKR-specific monoclonal antibody was then used to precipitate PKR under stringent washing conditions. The purified PKR proteins, which were free of detectable PP1 C (as judged by immunoblot analysis; data not shown) were subjected to the in vitro kinase assay in the presence of [␥-32 P]ATP and H1 histones as substrate. As shown in Fig. 3B (lanes 1 and 2), PKR efficiently phosphorylated H1 histones in a dsRNA-dependent manner when the preincubation step did not include PP1 C . Preincubation with PP1 C , how- FIG. 2. PP1 C dephosphorylates PKR in vivo and in vitro. A, PP1 C expression reduces hyperphosphorylation of PKR. RY1-1 cells expressing PKR, PKR K296R, and/or PP1 C were grown for 5 h in galactose-containing liquid media, and extracts prepared as described previously (9). Proteins (25 g) were separated by 7.5% SDS-PAGE and subjected to immunoblot analysis using an antibody to PKR. PP1 denotes protein phosphatase 1; PKR* indicates hyperphosphorylated PKR. B, PP1 C inhibits PKR autophosphorylation. Purified PKR proteins were first labeled by autophosphorylation in a kinase buffer in the presence of [␥-32 P]ATP and poly(I⅐C), then purified by chromatography (ProbeQuant G-50 Micro Column), and used as a substrate for different amounts of PP1 C in a phosphatase assay as described under "Experimental Procedures." Samples were subjected to 14% SDS-PAGE and autoradiography (top panel) or immunoblot analysis using an anti-PKR antibody (␣-PKR; bottom panel). C, PP1 C is not a substrate for PKR in vitro. The kinase activity of affinity-purified PKR protein (0.15 g) was assayed in the presence of GST or the indicated GST fusion (0.2 g) as described above. Protein phosphorylation was detected by autoradiography (top panel), and protein expression was confirmed by Western blot analysis using either a GST-specific or PKR-specific antibody. GST-Tat was used as a positive control for phosphorylation by PKR.

FIG. 3. PP1 C inactivates PKR function in vivo and in vitro.
A, eIF2␣ phosphorylation analysis. Extracts from yeast RY1-1 cells expressing the indicated proteins were separated by isoelectric focusing-PAGE and subjected to immunoblot analysis with an antiserum to yeast eIF2␣ as described under "Experimental Procedures" (40). Arrows indicate the positions of yeast eIF2␣ phosphorylated on basal sites only (lower band), and yeast eIF2␣ phosphorylated on Ser-51 (eIF2␣-P), the site of phosphorylation by PKR (upper band). The relative levels of protein phosphorylation were determined by quantifying the immunoblots using a Molecular Dynamics PhosphorImager and ImageQuant software (version 5.1), and the percentage of unphosphorylated eIF2␣ was shown. B, in vitro kinase assay. Affinity-purified PKR proteins (0.3 g) were preincubated with increasing concentrations (0.1, 0.2, and 0.4 g) of purified PP1 C , then activated by poly(I⅐C) at 30°C for 30 min as described under "Experimental Procedures." PKR was then purified by immunoprecipitation and tested for its ability to phosphorylate histone H1 in the presence of [␥-32 P]ATP in a kinase reaction. The reactions were stopped by boiling in 2ϫ Laemmli sample buffer. Proteins were separated on SDS-PAGE, the gels were stained with Coomassie Blue, and dried, and phosphorylated histones were detected by autoradiography. ever, significantly inhibited the ability of PKR to phosphorylate histones (lanes 3 and 4). As a further control in this experiment, PKR preincubated with inactive PP1 C (inhibited by the PP1 inhibitor I-2) retained its ability to phosphorylate histones (lane 5), indicating that the activity of PP1 C is required to inhibit PKR in vitro. These results are consistent with the notion that PP1 C directly inactivates PKR function. PP1 C Forms a Physical Complex with PKR-An emerging theme in cell signaling is the ability of protein kinases to form stable complexes with their corresponding phosphatases to ensure rapid and transient signal transduction mediated through reversible protein phosphorylation. Furthermore, the results shown thus far suggested that PP1 C likely interacts with PKR. To test this possibility, we used a GST fusion protein-mediated co-sedimentation and immunoblotting assay. GST-PP1 C -containing agarose beads were incubated with human HeLa cell extracts. Bound proteins were pulled down by centrifugation, washed, and subjected to SDS-PAGE and immunoblotting analyses using an anti-PKR antibody. As predicted, we found that endogenous PKR from HeLa cell extracts co-sedimented with recombinant GST-PP1 C , but not with the GST control (Fig. 4A, left panel). The PKR-PP1 C interaction was verified by reciprocal experiments, which showed that GST-PKR was able to pull down endogenous PP1 C from HeLa cell lysates, although to a lesser extent (right panel). It is not clear why GST-PP1 C is more effective than GST-PKR in pulling down the interacting partner. A possible explanation for this is that the fused GST tag is partially masking a region or affecting the protein conformation that is required for PKR interaction with PP1 C .
We next used the yeast two-hybrid system to confirm the interaction between PP1 C and PKR. Because wild-type PKR is toxic to S. cerevisiae, we used the catalytically inactive PKR protein, PKR K296R. As shown in Fig. 4B, yeast strain Hf7c co-transformed with GAL4 AD PKR K296R and the GAL4 BD expression vector, or a fusion control to GAL4 BD (GAL4 BD P53), was unable to activate the HIS3 reporter genes and was therefore unable to grow in the absence of histidine (His). In addition, yeast cells co-transformed with GAL4 BD PP1 C and the GAL4 AD expression vector, or a control fusion to GAL4 BD (GAL4 BD SV40 T ag), were unable to transactivate the reporter construct. However, when the PP1 C and PKR hybrid proteins were coexpressed in Hf7c, transactivation of HIS3 occurred, allowing the cells to grow in the absence of His. A similar effect was observed using the positive control proteins P53 and SV40 T Ag (45). The specificity of PKR-PP1 C interaction was further demonstrated by the observation that the p53 protein did not bind PKR in this system, nor did we detect an interaction between PP1 C and eIF2␣, which is consistent with published observations (44,46). It is interesting to note that the catalytic subunit of PP2A (PP2A C ), the other major protein phosphatase, interacted with PKR very weakly in light of recent findings that PKR interacts with the regulatory subunit of PP2A (47).
To examine whether PKR and PP1 C interact in living cells, we immunoprecipitated endogenous PKR from protein lysates prepared from Huh7 cells with an antibody to PKR. The precipitates were then analyzed by Western blotting using a PP1 Cspecific antibody that also reacts with an unknown 100-kDa protein ( Fig. 4C; lysate). A significantly large portion of endogenous PP1 C , but not the 100-kDa protein, could be detected in PKR immunoprecipitates (Fig. 4C; ␣-PKR). Furthermore, we did not detect PP1 C in control immunoprecipitates obtained using normal mouse serum. Taken together, these results strongly support the notion that PP1 C specifically interacts with PKR in intact cells.
PP1 C Binds to the Regulatory Domain of PKR via a PP1 Cbinding Consensus Motif-We next performed GST pulldown assays to identify the region of PKR that interacts with PP1 C . Lysates from HeLa cells were incubated with recombinant proteins consisting of GST fused to deletion or point mutants of PKR. As shown in Fig. 5A, PP1 C bound to the N-terminal regulatory domain, but not to the C-terminal catalytic domain of PKR (top panel, lanes 1 and 2). Consistent with this result, the catalytic activity of PKR was not required for the interaction with PP1 C because the catalytically attenuated PKR K296R retained its ability to bind PP1 C (lane 3). The amount of the various GST fusion proteins that were cosedimented in these experiments was revealed by Western blot analysis using an antibody against GST (bottom panel; indicated by asterisks).
To identify the PKR domain participating in interacting with PP1 C in an in vivo environment, we used the two-hybrid sys-

FIG. 4. PP1 C interacts with PKR in vitro and in vivo.
A, GST pulldown analysis. Crude lysates (500 g) from HeLa cells were incubated with GST or the indicated GST fusion protein (0.2 or 0.5 g) immobilized on glutathione-agarose beads as described under "Experimental Procedures." The beads were washed, and bound proteins were resolved by SDS-PAGE (12.5% acrylamide). After transfer to nitrocellulose, the blots were probed with a PKR-specific monoclonal antibody (␣-PKR; left panel A) or a PP1 C -specific polyclonal antibody purchased from Upstate Biotechnology (␣-PP1 C ; right panel). The PKR and PP1 C proteins are indicated by arrows. B, summary of two-hybrid analysis. The Hf7c reporter strain was transformed with the indicated plasmids. The interaction between the two-hybrid proteins was scored by the induction of HIS3 expression (growth on SD agar plates lacking histidine). ϩ indicates growth on SD medium-His (indicative of interaction) and Ϫ denotes no growth. C, coimmunoprecipitation of endogenous PKR and PP1 C . Immunoprecipitation was performed using either PKR-specific monoclonal antibody or normal mouse serum (NMS), and the blots were probed with a PP1 C -specific polyclonal antibody obtained from Dr. K. Schlender (␣-PP1 C ; top panel) or a PKR-specific polyclonal antibody (␣-PKR; bottom panel).
tem. We found that PP1 C bound to the N-terminal 242 residues of PKR (Fig. 5B), consistent with the in vitro results above. The undetectable interaction between PP1 C and PKR 244 -551 cannot be explained by the lack of protein expression or incorrect protein folding because the latter interacted effectively with the vaccinia virus K3L protein positive control, consistent with published results (37). Because PKR binds dsRNA, it is possible that the interaction between PP1 C and PKR is tethered via an RNA bridge. To test this possibility, we used a truncated mutant PKR (amino acids 1-220) containing a mutation at lysine 64 (K64E), which abrogates its ability to bind dsRNA (48). The results show that this mutant retained its ability to bind efficiently to PP1 C , supporting our contention that PP1 C binds PKR via a direct protein-protein contact mechanism. However, we cannot rule out completely the presence of resid-ual dsRNA mediating the interaction between PKR 1-220 (K64E) and PP1 C in this assay.
Many PP1 C -interacting proteins share a short PP1 C -binding consensus motif, defined as (R/K)(V/I/L)X(F/W/Y) (49). Sequence analysis revealed that PKR has two potential motifs that are analogous to the PP1 C -binding consensus sequence: the first is located at position 164 -167 and the second at position 297-300 of PKR. Because removal of the C-terminal part of PKR, which includes the second PP1 C -binding consensus motif, did not abrogate PP1 C binding (Fig. 5B), we reasoned that the first PP1 C -binding consensus motif might be important for PP1 C binding. To validate this, we mutated the Tyr residue to Ala in the motif and tested the ability of the mutant (PKR Y167A) to bind PP1 C . As predicted, the mutant PKR was unable to interact with PP1 C in the two-hybrid assay (Fig. 5B). As a control, PKR Y167A retained the ability to bind K3L, indicating that the mutant protein was properly expressed and translocated to the nucleus.
PKR Binds to a Conserved C-terminal Non-catalytic Region of PP1 C Isoforms-We also examined whether PKR could bind to PP1␣2 (or PP1 C 2), an isoform that differs from PP1 C by an N-terminal 11-amino acid insert (50). GST or GST fusions containing PP1 C or PP1 C 2 immobilized on agarose beads were incubated with HeLa cell extracts, and the bound proteins were eluted and analyzed by SDS-PAGE and immunoblotting using an antibody against PKR or GST. As shown in Fig. 6A (top panel), PKR interacted with PP1 C 2 (lane 4). GST and GST-PP1 C were used as a negative and positive control, respectively (lane 2 and 3). Because PP1 C and PP1 C 2 share a common C-terminal non-catalytic region, we suspected that this region might be sufficient to mediate PKR interaction. To test this, we used a GST fusion containing the C-terminal 161 residues of PP1 C . As predicted, PKR was still capable of interacting with this truncated PP1 C protein (lane 5). Importantly, all GST fusions were expressed efficiently (bottom panel). Thus the PKR-interactive region appears to be localized within the C-terminal non-catalytic region conserved among PP1 C isoforms.
The Catalytic Domain of PP1 C Is Required for Inhibition of PKR Dimerization-To obtain corroborative evidence that PP1 C binds to the N-terminal region of PKR, which is critical for enzyme dimerization, we turned to the repressor fusion dimerization assay (51). In this system, full-length inactive PKR is expressed in E. coli as a fusion to the N-terminal domain of cI repressor (N), which contains the DNA-binding domain but lacks the dimerization domain of cI. Dimerization of PKR reconstitutes the DNA-binding activity of N fusion, leading to repression of the P R promoter that can be scored by the resistance of the E. coli cells to lysis by the phage (10,36). As summarized in Fig. 6B, we found that co-expressing fulllength GST-PP1 C or GST-PP1 C 2, but not GST or GST-Tat, blocked dimerization by the N-PKR K296R fusion. Interestingly, a GST fusion containing the C-terminal 161 residues of PP1 C , despite retaining the ability to bind PKR (Fig. 6A), did not block PKR dimerization in this assay. Taken together, these results suggest that PP1 C does not inhibit PKR by merely binding to the kinase and that the catalytic domain of PP1 C is required to disrupt the dimerization process of PKR. DISCUSSION Compared with the widely studied mechanisms for the phosphorylation and activation of PKR, the mechanisms underlying the inactivation of the enzyme are largely uncharacterized. Although viral studies have led to the identification of many virus-encoded or -directed factors that bind PKR or inhibit its activity (4), little progress has made in understanding the mechanisms that normally control PKR activity during cellular homeostasis. Identification of protein phosphatases that reverse the activating phosphorylation of PKR will provide a better understanding of its regulation and function. Previous two-hybrid screens for PKR-interacting proteins have yielded only activators or substrates of PKR (47,(52)(53)(54). We thus chose to use a functional screen to identify novel cellular PKR antagonists. Expression of PKR in yeast inhibits growth by phosphorylating eIF2␣, which leads to the disruption of the cellular translational apparatus (20). This PKR-mediated toxicity can be partially reversed by the co-expression of viral protein inhibitors (39,41,45). Using this functional assay, we screened a human cDNA expression library for clones capable of counteracting the growth-suppressive effect of PKR, which led to the identification of PP1 C (Fig. 1).
Both PP1 and PP2 are capable of dephosphorylating eIF2␣ kinases (33,34). However, the exact protein phosphatase for PKR has not been identified, nor is the mechanism of action known. In this report, we present evidence that PP1 C is a bona fide antagonist of PKR. Co-expression of PP1 C leads to reduced phosphorylation of PKR and its physiological substrate, eIF2␣ (Figs. 2 and 3). The interaction of PP1 C with PKR appears to be specific and functional, because the dephosphorylation of PKR in vitro by purified PP1 C was concentration-dependent and could be blocked by a PP1 inhibitor. Furthermore, an in vitro binding assay, the two-hybrid system, and co-immunoprecipitation experiments collectively demonstrated a specific and direct interaction between PP1 C and PKR (Figs. 4 and 5). The finding that PKR phosphorylation may be transiently modulated by PP1 C is consistent with the proposed mechanism of action of viral RNA inhibitors of PKR (55). The RNA binding affinity of PKR is regulated by its phosphorylation states; autophosphorylated PKR molecules display low affinity for RNA and high eIF2␣ kinase activity. Thus, viral RNA inhibitors would not be able to effectively inhibit autophosphorylated PKR unless the kinase is dephosphorylated by a cellular phosphatase(s). It remains to be seen whether viruses have evolved a mechanism to activate PP1 C or recruit a PP1 C -like phosphatase to dephosphorylate PKR, resulting in PKR forms that are susceptible to viral RNA-mediated inhibition.
To begin to delineate the mechanisms of PP1 C action, we FIG. 6. A conserved C-terminal region of PP1 isoforms interacts with PKR. A, GST co-sedimentation experiments were performed as described in the legend of Fig. 4A. E. coli-purified GST and GST fusions with full-length PP1␣2 and the C-terminal 161 amino acids of PP1 C /PP1␣2 were tested for their ability to interact with PKR from HeLa cell lysates. Bound proteins were resolved by SDS-PAGE (12.5% acrylamide), and immunoblots were probed with anti-PKR (␣-PKR; top panel) or anti-GST (␣-GST; bottom panel) antibody. The PKR protein is indicated by arrow; GST fusion proteins are indicated by an asterisk. B, PP1 C isoforms prevent PKR dimerization in E. coli. The repressor fusion assay was performed, and inhibition of PKR dimmer formation was scored as described under "Experimental Procedures" (10,36). A summary of the results is shown on the right. ϩ, inhibition of PKR dimerization; Ϫ, no inhibition.
performed deletion analysis and found that PP1 C bound to the N-terminal regulatory region of PKR independently of the dsRNA-binding capability of the kinase (Fig. 5B). Importantly, we found that the N terminus of PKR contains a PP1 C -binding motif, which is present in most PP1 C -interacting proteins, and which was required for the interaction of PKR with PP1 C . PKR is phosphorylated at multiple serine and threonine residues, including those in the N-terminal regulatory domain (56). While we do not know whether PP1 C directly binds to those sites, there is increasing precedence for both kinases and phosphatases interacting with sites other than the phosphorylation sites in their substrates. We mapped the PKR-interacting region to the C-terminal non-catalytic region, which is conserved between PP1 C and PP1 C 2 isoforms. Consistent with this observation, we found that both PP1 C and PP1 C 2 were capable of disrupting PKR dimer formation in the repressor fusion assay (Fig. 6). Interestingly, a truncated PP1 C lacking the catalytic domain, but retaining its ability to interact with PKR, did not prevent PKR dimerization. Based on these results, we propose that PP1 C and PKR interact directly through their respective non-catalytic regions. However, the catalytic domain of PP1 C is required to dephosphorylate PKR, resulting in monomeric PKR forms due to their higher affinities for RNA (55). Alternatively, PP1 C -mediated dephosphorylation of PKR may produce a protein conformation that is unable to dimerize independently of RNA binding. These complex interactions should be more apparent when a three-dimensional structure of the PKR-PP1 C complex is solved. PP1 C modulates an enormous variety of cellular functions and normally exists as a heterodimer consisting of a core catalytic subunit and one of a number of different regulatory subunits. It has been suggested that the substrate specificity of PP1 C is dictated by the interaction of PP1 C with different regulatory subunits, which may target the catalytic subunit to specific subcellular locations (57). Regulation of PP1 C in response to extracellular and intracellular signals occurs mostly through changes in the levels, conformation, or phosphorylation status of targeting subunits. Most of these bind to a small hydrophobic groove on the surface of PP1 C through a short conserved binding motif-the (R/K)(V/I/L)X(F/W/Y) motif, which is often preceded by further basic residues, although several putative targeting subunits do not possess an (R/K)(V/I/L)X(F/ W/Y) motif but nevertheless interact with the same region of PP1 C . In this regard, the herpes simplex virus type 1 (HSV-1)encoded ␥1 34.5 protein contains such a motif, which interacts with PP1 C to redirect the phosphatase to dephosphorylate eIF2␣ (44,46). Selective dephosphorylation of eIF2␣ may be a clever strategy used by HSV-1 to circumvent the PKR-induced shut-off of protein synthesis while maintaining PKR activity for other biological functions that are essential to the virus life cycle. Here, we demonstrated that PKR also contains the (R/ K)(V/I/L)X(F/W/Y) motif, which is required for its binding to PP1 C . However, we cannot exclude the possibility that a cellular regulatory subunit mediates PP1 C specificity toward PKR. One candidate is the glycogen-targeting subunit of PP1, termed PP1 GL (10). PP1 GL , which is expressed in heart and skeletal muscle, plays a pivotal role in rat skeletal muscle cell myogenesis via its regulation of PP1 C activity (58). PKR also plays an important regulatory role in murine myogenic processes (59,60), prompting the speculation of a possible localized role for PKR in skeletal muscle via its association with PP1 C and PP1 GL . Finally, PP1 C -catalyzed dephosphorylation of PKR may be implicated in insulin signaling; both PP1 C activity (61) and PP1 GL phosphorylation (62) are stimulated by insulin. Interestingly, insulin induces a decrease in eIF2␣ phosphorylation in chondrocytes (63), although it is not known whether this decrease is an insulin-mediated increase in PP1 activity toward PKR and/or eIF2␣. Our findings suggest an updated model for PKR regulation within and outside the context of virus infection (Fig. 7). This model should provide the basis for future studies to examine whether a regulatory subunit is involved in PP1 C interaction with and/or inhibition of PKR under specific conditions. Such studies may begin to ascribe the consequences of the PP1 C dephosphorylation of PKR to specific biological effects.
Acknowledgments-We thank Dr. A. G. Hinnebusch for yeast strains, Dr. J. C. Hu for the repressor system, and Dr. S. J. Elledge for FIG. 7. Model for PKR regulation by PP1. A, antiviral and antiproliferative effects of PKR resulting from eIF2␣ phosphorylation. B, neutralization of PKR-mediated effects by direct dephosphorylation and monomerization of PKR by PP1 C during normal cell physiology or by PP1 C -mediated eIF2␣ dephosphorylation during HSV-1 infection. It is not clear why HSV-1 does not target PKR directly, but it appears that the virus also encodes two additional gene products, Us11 and Us12, to deliberately activate PKR while encoding a separate function that selectively prevents the phosphorylation of eIF2␣. Presumably, this represents a mechanism by which the virus maintains other biological functions of PKR, such as cell differentiation or apoptosis, that are important during different stages of the viral life cycle. See "Discussion" for details.