Mutations in the Double-stranded RNA-activated Protein Kinase Insert Region That Uncouple Catalysis from eIF2α Binding*

The interferon-induced, double-stranded RNA (dsRNA)-activated protein kinase, PKR, inhibits protein synthesis via phosphorylation of the alpha subunit of the translation initiation factor eIF2. A kinase insert region N-terminal of PKR kinase subdomain V, which is conserved among eIF2α kinases, has been proposed to determine substrate specificity of these kinases. To investigate the function of this kinase insert region, selective PKR mutants were generated, and kinase activities and eIF2α affinities were analyzedin vitro. The in vivo function was investigated by growth inhibitory assays in yeast and translational assays in COS cells. Among the 13 mutations, 5 lost kinase activity and 3 exhibited less than 30% of wild-type eIF2α binding activity. The deletion of the conserved sequence (amino acids 362–370) resulted in a protein that had no kinase activity and only about 25% of wild-type eIF2α binding, suggesting that this sequence is not only required for PKR kinase activity but also is important for substrate interaction. It was determined that the hydrophobicity of the conserved sequence of PKR is required for kinase activity but is not crucial for eIF2α binding. The amino acid residue Glu-367 in the conserved motif was shown to be directly involved in substrate binding but was not important for kinase activity. These results suggest that the activation of PKR is not a prerequisite for its binding to the substrate and that the conserved motif in subdomain V contributes to the interaction of PKR and eIF2α.

The interferon-induced, double-stranded RNA (dsRNA)-activated protein kinase, PKR, inhibits protein synthesis via phosphorylation of the alpha subunit of the translation initiation factor eIF2. A kinase insert region N-terminal of PKR kinase subdomain V, which is conserved among eIF2␣ kinases, has been proposed to determine substrate specificity of these kinases. To investigate the function of this kinase insert region, selective PKR mutants were generated, and kinase activities and eIF2␣ affinities were analyzed in vitro. The in vivo function was investigated by growth inhibitory assays in yeast and translational assays in COS cells. Among the 13 mutations, 5 lost kinase activity and 3 exhibited less than 30% of wild-type eIF2␣ binding activity. The deletion of the conserved sequence (amino acids 362-370) resulted in a protein that had no kinase activity and only about 25% of wild-type eIF2␣ binding, suggesting that this sequence is not only required for PKR kinase activity but also is important for substrate interaction. It was determined that the hydrophobicity of the conserved sequence of PKR is required for kinase activity but is not crucial for eIF2␣ binding. The amino acid residue Glu-367 in the conserved motif was shown to be directly involved in substrate binding but was not important for kinase activity. These results suggest that the activation of PKR is not a prerequisite for its binding to the substrate and that the conserved motif in subdomain V contributes to the interaction of PKR and eIF2␣.
Protein synthesis is divided into three phases: initiation, elongation, and termination. The reactions in each phase are promoted by soluble protein factors that transiently interact with the ribosome, mRNA, and aminoacyl-tRNA (1). The phosphorylation of the ␣ subunit of translation initiation factor 2 (eIF2␣) is a well characterized translational control mechanism. During the first step of translation initiation, a ternary complex eIF2-GTP-Met-tRNA is formed and subsequently binds to the 40 S ribosomal subunit of the ribosomes. After one round of initiation, eIF2 is released as an eIF2-GDP complex. To re-form the ternary complex, the GDP bound to eIF2 must be replaced with GTP, and this reaction is catalyzed by the guanine nucleotide exchange factor eIF2B. Phosphorylation of eIF2␣ sequesters the function of eIF2B, thus blocking eIF2 recycling, and results in the inhibition of protein synthesis (1).
Three distinct eIF2␣ kinases have been characterized in eukaryotic cells, including the hemin-regulated inhibitor (HRI, 1 or HCR), the double-stranded RNA activated protein kinase (PKR), and the yeast kinase GCN2 (general control nondepressible). All three kinases phosphorylate eIF2␣ at Ser-51 in response to cellular stress conditions (2,3). HRI mediates protein inhibition in heme-deficient reticulocyte lysate and is also activated by heat shock, sulfhydryl reagents, and heavy metal ions (4). PKR is up-regulated by interferon and becomes activated upon viral infections that produce dsRNA or single-stranded RNA with double-stranded secondary structure (5). GCN2 is activated by uncharged tRNA under amino acid starvation conditions. Phosphorylation of eIF2␣ by GCN2 mediates gene-specific translational control of GCN4, a transcriptional activator of amino acid biosynthetic genes in yeast. GCN2 is required for increased translation of GCN4 mRNA in amino acid-starved cells (3). Sequence analysis shows that PKR, HRI, and GCN2 share sequence and structural features distinct from other eukaryotic protein kinases. Most notable is the presence of an insert sequence located between kinase subdomains IV and V, consisting of 24 amino acid residues for human PKR, 11 amino acid residues for murine PKR, 128 amino acid residues for HRI, and 112 amino acid residues for GCN2 ( Fig. 1A) (6). Although subdomain V is less well conserved among eukaryotic protein kinases (6), the eIF2 kinases share significant homology in this domain, suggesting that this motif might contain a putative substrate-specific recognition domain (7,8).
PKR is a serine-threonine kinase that mediates the antiviral and antiproliferative effects of interferon through phosphorylation of eIF2␣ (5). Discovered as an enzyme responsible for the inhibition of translation in response to dsRNA in reticulocytes (9), PKR has two distinct kinase activities: autophosphorylation, requiring the presence of dsRNA, divalent cations (Mn 2ϩ and Mg 2ϩ ), and ATP (10,11), and phosphorylation of exogenous substrates. This second activity is independent of dsRNA but dependent on ATP and divalent cations (4). As phosphorylation of eIF2␣ by PKR provides a mechanism for cell defense against virus infection, many viruses have evolved specific mechanisms to inactivate PKR, including degradation of PKR (12), sequestering of dsRNA activator (13), binding to PKR to inhibit its activation (14), or activating a cellular PKR inhibitor (15). DNA transfection of certain plasmids can also activate PKR and result in poor translation of plasmid-derived mRNAs. Translation can be restored by coexpression of PKR inhibitors of a mutant form of eIF2␣ where serine residue 51 has been replaced by alanine (16). Because of the homology between human and yeast eIF2␣, PKR is also able to phosphorylate yeast eIF2␣ at Ser-51 and inhibit yeast growth (7).
Human PKR is a 551-amino acid protein consisting of an N-terminal regulatory domain and a C-terminal catalytic domain (17). The regulatory domain contains two dsRNA binding motifs that are homologous to those in other dsRNA-binding proteins (18). Motif I (amino acids 55-75) is essential for dsRNA binding, whereas motif II (amino acids 145-166) plays a less important role (19 -21). The catalytic domain contains 11 conserved kinase subdomains characterizing PKR as a serine/ threonine kinase (17). The invariant lysine (Lys-296) in subdomain II is essential for the phosphate transfer reaction. Mutation of this lysine to arginine or proline results in a catalytically inactive protein (7,22). As mentioned above, PKR has a 24amino acid kinase insert between subdomains IV and V and the highly conserved eIF2␣ kinase motif LFIQMEFCD in subdomain V. A full-length PKR containing a deletion of six amino acids (LFIQME) at this motif was unable to undergo autophosphorylation or to phosphorylate eIF2 (23), underlining the importance of this region to PKR activity.
To delineate the individual contribution of conserved individual amino acids in and around the kinase insert region of PKR to eIF2␣ binding and kinase activity, we generated 13 PKR mutants in this region, analyzed their in vitro auto-and substrate phosphorylation activities and eIF2␣ affinity, and examined the function of selected mutants in yeast and in mammalian cells. We found that the hydrophobicity of the conserved motif in subdomain V is essential for PKR kinase activity but is not necessarily required for the interaction with eIF2␣. We also identified a specific amino acid (Glu-367) that is involved in eIF2␣ binding but not in PKR auto-and substrate phosphorylation. We conclude that the conserved motif in subdomain V is required for PKR kinase activity and also contributes to its interaction with eIF2␣.

EXPERIMENTAL PROCEDURES
PKR Mutagenesis and Plasmid Constructions-PKR mutants were generated as described (24) using the Transformer site-directed mutagenesis kit (CLONTECH). PKR cDNA was cloned into the HindIII and PstI sites of pBluescript KS. After mutagenesis, transformants were selected with ampicillin and confirmed by DNA sequencing.
The bacterial expression vector pGEX-PKR was constructed by digestion of pBS-PKR with HindIII, end-filling of the PKR fragment, and then ligation into the Escherichia coli expression vector pGEX-1 (Pharmacia Biotech Inc.) at the SmaI site. The yeast expression vector pYex-PKR was created by inserting the 1.8-kilobase XbaI-PstI PKR cDNA fragment from pBS-PKR into pEMBLyex4 (40). The mammalian expression vectors were constructed by insertion of PKR/HindIII fragment from pBS-PKR into pRC/CMV (Invitrogen). The mutant yeast eIF2␣ expression plasmid (sui2M) was generated by subcloning a 2.7kilobase eIF2␣ gene fragment into a 2-micron vector (Yep13) carrying a LEU2 gene (20). The bacterial expression vector pQE-eIF2␣ expresses human eIF2␣ with a hexahistidine containing fusion peptide attached at its C terminus and was kindly provided by Dr. Richard Padgett (Lerner Research Institute).
Purification of Bacteria-expressed PKR-E. coli BL21(DE3)pLysS cells harboring the pGEX-PKR expression vector were grown overnight in 50 ml of Luria-Bertani broth containing 50 g/ml ampicillin. Following 1:10 dilution in fresh medium, cells were grown for 2 h to an optical density of approximately 0.9, induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside and incubated for an additional 3-4 h. Bacteria were pelleted and resuspended in modified NETN (20 mM Tris-HCl, pH7.6, 150 mM NaCl, 1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin). Cells were lysed by adding Triton X-100 to 1% (w/v) and sonicating, and the lysate was centrifuged at 27,000 ϫ g at 4°C for 20 min. The supernatant was incubated with glutathione-Sepharose beads (Pharmacia) for 30 min at 4°C with agitation to isolate the PKR fusion protein. After low speed centrifugation, the bound protein was washed four times with cold modified NETN. Bound protein was eluted with elution buffer (50 mM Tris, pH 8.0, 1 mM EDTA, 10 g/ml reduced glutathione).
In Vitro Phosphorylation Assays-The PKR autophosphorylation as-say was performed as follows. PKR protein (ϳ100 ng) was brought up to a volume of 15 l in buffer DBGA (10 mM Tris-HCl, pH 7.6, 50 mM KCl, 2 mM magnesium acetate, 20% glycerol, 7 mM 2-mercaptoethanol) before addition of 10 l of DBGB (DBGA containing 2.5 mM MnCl 2 ) and 5 Ci of [␥-32 P]ATP (50 Ci/mmol) in a final volume of 30 l. The reactions were incubated at room temperature for 20 min before the addition of SDS loading buffer and boiled for 3 min prior to loading on 10% SDS-PAGE gels and autoradiography. For the eIF2␣ phosphorylation assay, 200 ng of eIF2␣ was added into the above kinase reaction. PKR-eIF2␣ Binding Assays-To produce recombinant histidinetagged eIF2␣, an overnight culture of pQE-eIF2␣ containing E. coli M15[pREP4] cells (QIAGEN) was diluted 1:10 in fresh medium and grown for 1 h at 37°C. Isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 1 mM, and the culture was incubated for an additional 2 h at 37°C prior to harvesting by centrifugation. The cells were resuspend in phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.84 mM KH 2 PO4, pH 7.4) containing 0.1% (v/v) Triton X-100 and 10% (v/v) glycerol, lysed by sonication for 3 ϫ 20 s, and centrifuged. The supernatant was aliquoted and stored at Ϫ80°C.
For the binding assay, purified GST-PKR protein was mixed with 10 l of bacterial lysate containing eIF2␣. Binding buffer (phosphatebuffered saline containing 10% (v/v) glycerol) was added in a total volume of 200 l. After incubation on ice for 30 min, a further 800 l of binding buffer was added with 20 l of a slurry of glutathione-Sepharose beads (50% (v/v)), and the GST-PKR bound eIF2␣ pulled down by incubating at 4°C for 1 h with rotation. The beads were washed five times with phosphate-buffered saline containing 0.1% (v/v) Triton X-100, an equal volume of 2 ϫ SDS sample buffer was added, and samples were boiled for 5 min and electrophoresed on a 10% SDS-PAGE gel. Proteins were identified by Western blotting.
PKR Expression in Mammalian Cells-PKR mutants were subcloned into the mammalian expression vector pRc/CMV. COS-1 cells were transiently transfected with 400 ng of pRc-PKR and 2 g of reporter plasmid pGL2-Control (Promega) by the lipofectamine procedure (Life Technologies, Inc.). Cells were harvested 48 h after transfection and assayed for luciferase activity after normalizing for transfection efficiency by measuring the total protein (27).

RESULTS
Site-directed Mutation of PKR-The eIF2␣ kinases (HRI, GCN2, and PKR) are distinguished in the family of the serine/ threonine protein kinases by possessing a kinase insert between subdomains IV and V. Sequence alignment shows that the amino acid residues flanking the insert regions are highly conserved, including two amino acids in subdomain IV (Tyr-232 and Trp-237 in PKR) and a nine-amino acid motif in subdomain V (LFIQMEFCD) (Fig. 1A). It has been proposed that these amino acids may contribute to substrate recognition (7,8). To investigate the function of these residues, we designed and constructed 13 mutated forms of PKR, including a deletion of the conserved motif (M12), a deletion of the subdomain V together with part of the kinase insert (M13), and single point mutations of different conserved residues (Fig. 1B). Site-directed mutagenesis was performed in pBluescript, confirmed by sequencing, and the mutated cDNAs were subcloned into expression vectors for in vitro and in vivo functional analysis.

Expression and Kinase Analysis of PKR Mutants in Vitro-
Wild-type and PKR mutant cDNAs were expressed in E. coli as GST fusion proteins in the bacterial expression vector pGEX (Fig. 2). The GST fusion proteins were purified by glutathione-Sepharose affinity chromatography, and their expression was monitored by SDS-PAGE with Coomassie Blue staining (data not shown). All mutants were well expressed except M5, which may be intrinsically unstable. The GST fusion proteins showed different mobilities on SDS-PAGE, which were likely due to different phosphorylation levels since recombinant PKR can be partially phosphorylated in bacteria (29).
To assess the autophosphorylation and substrate phosphorylation activities of PKR mutants, kinase assays were performed using the purified GST-PKR proteins. The levels of GST-PKR proteins used in the kinase assays were determined by Western blot using anti-PKR polyclonal antibody ( Fig. 2A,  upper panel). PKR autophosphorylation ( Fig. 2A, middle panel) and phosphorylation of eIF2␣ ( Fig. 2A, lower panel) were detected by autoradiography. As expected, wild-type PKR showed a high level of autophosphorylation and eIF2␣ phosphorylation, but no kinase activity was detected for the K296R mutant ( Fig. 2A, middle panel). The two deletion mutants (M12, M13) showed no auto-and substrate phosphorylation activities, whereas the single point mutants showed varying levels of auto-and substrate phosphorylation activities. By comparing the intensities of the phosphorylated protein bands with the PKR concentration, the specific kinase activities for each PKR mutant could be calculated (Fig. 2C). For most of the mutants, autophosphorylation and substrate phosphorylation activities were well correlated. An exception to this pattern is mutant M3, which shows autophosphorylation activity but no eIF2␣ phosphorylation. The single point mutants M3 (L362Q), M4 (I364N), and M8 (F363S), which lost all or most of their kinase activity, contain substitutions of hydrophobic to hydrophilic amino acids within the conserved sequence of subdomain V.
Analysis of eIF2␣ Binding-To study the interaction of PKR with eIF2␣, the bacterially expressed GST-PKR proteins were reacted with histidine-tagged eIF2␣. Glutathione-Sepharose beads were used to immobilize the GST-PKR/eIF2␣ complex, and PKR and eIF2␣ were detected in Western blots probed with monoclonal antibodies against PKR and eIF2␣, respectively (Fig. 2B). The interaction between PKR and eIF2␣ was specific because GST alone was not able to bind histidinetagged eIF2␣ (Fig. 2B, lower panel), and wild-type GST-PKR was not able to bind an unrelated histidine-tagged protein His-WT1 (data not shown). All the PKR mutants were able to bind to eIF2␣ but with different affinities (Fig. 2B). The relative affinity of different mutants for eIF2␣ were calculated by counting the ratios of how much eIF2␣ bound to certain amounts of PKR (Fig. 2D). Interestingly, there was no correlation between the affinity of the PKR mutants for eIF2␣ and their kinase activities (Fig. 2, C and D). For example, M4 has no kinase activity yet has almost twice the affinity of wild-type PKR for eIF2␣. In contrast, mutant M11 has only 20% of wild-type affinity for eIF2␣ but retains 90% of wild-type kinase activity. Furthermore, M11 can phosphorylate eIF2␣ very well, suggesting that the high affinity of PKR for eIF2␣ might not be a prerequisite for PKR function.
Effects of PKR Mutants on Yeast Cell Growth-To analyze the activities of the PKR mutants in vivo, we expressed PKR mutants in yeast cells (7,24,30). There is 80% homology between yeast and human eIF2␣ when conservative substitutions are considered (31), and yeast eIF2␣ can be phosphorylated by human PKR, leading to growth inhibition (7). Accordingly, the different PKR mutants were subcloned into the vector pEMBLyex4 and transformed into S. cerevisiae strain W303a. PKR protein expression was controlled via the GAL10- GST-PKR proteins were detected by Western blot with anti-PKR monoclonal antibody (upper panel). Bound eIF2␣ was detected by Western blot with anti-eIF2␣ antibody (lower panel). C and D, the intensity of the bands in A and B were measured by NIH image tools. The relative kinase activities were counted by the ratio of intensities of the phosphorylation bands and the immunoblotting bands of PKR (in A), and the relative binding activities were counted by the ratio of the intensities of the eIF2␣ bands and the corresponding PKR bands (in B). Assigning the activity of wild-type PKR as 100, the relative specific kinase activities and eIF2 affinity of PKR mutants were calculated. C, the specific auto-and substrate phosphorylation activities. D, the eIF2␣ affinities. WT, wild type. Cyc1 hybrid promoter, whereby growth on galactose as the sole carbon source activates expression. In glucose medium, yeast transformed with M1, M2, M6, M7, or wild-type PKR grew very slowly (Fig. 3A). This was likely caused by a low basal level expression of PKR. When PKR expression was induced in galactose medium (Fig. 3B), the growth of yeast transformed with M1, M2, M4, M6, M7, M9, M10, M11, and wild-type PKR was completely inhibited. Significantly, all of these mutants showed kinase activities in vitro except M4 ( Fig. 2A). Because M4 has a high affinity for eIF2␣ in vitro (Fig. 2), we speculated that the yeast slow growth phenotype might be caused by the direct binding of M4 to eIF2␣. To prove this hypothesis, we co-expressed M4 with sui2M, a yeast eIF2␣ allele that contains a mutation (Ser-51 3 Ala) that completely abolishes phosphorylation by PKR. We have previously shown that sui2M is able to rescue the slow growth phenotype when co-expressed with wild-type PKR (7). We found that sui2M also reverted the slow-growth phenotype caused by M4 (Fig. 4B). This suggests that M4 inhibits yeast growth by direct interaction with eIF2␣ independent of auto-or substrate phosphorylation.
Expression of PKR Mutants in Mammalian Cells-Transient transfection of COS-1 cells with expression plasmids encoding a reporter gene such as DHFR has been shown to result in inefficient translation due to PKR activation and eIF2␣ phosphorylation (16). A variety of agents can stimulate the synthesis of the reporter protein, including vaccinia virus gene products E3L and K3L, the catalytically inactive PKR mutant K296R, and the dsRNA binding-deficient mutant K64E (32,33). The luciferase reporter plasmid pGL2-control was also successfully used in this system (27). Therefore, we used the COS-1 expression system to analyze the function of some of our PKR mutants in mammalian cells. Three PKR mutants (M3, M4, M11) together with wild-type PKR and the K296R mutant were subcloned into the mammalian expression vector pRc/ CMV and transiently transfected into COS-1 cells together with the luciferase reporter vector. Transfection of wild-type PKR almost completely inhibited luciferase translation, whereas the K296R mutant had a dominant-negative effect, resulting in enhanced luciferase activity (Fig. 5A). Mutant M3, which has very little autophosphorylation activity and no eIF2␣ phosphorylation activity in vitro, was not able to inhibit translation in COS cells, consistent with its function in yeast. However, mutant M4, which inhibited yeast growth, also enhanced luciferase activity, suggesting that it is inactive in COS cells. As with wild-type PKR, mutant M11 inhibited luciferase gene translation, consistent with its function in yeast. As shown previously, although M11 has a low eIF2␣ affinity for eIF2␣, it can phosphorylate eIF2␣ almost as well as wild-type PKR in vitro (Fig. 2). In vivo experiments demonstrate that M11 is fully functional in both yeast and COS cells.
PKR has been shown to be autoregulatory in transfected COS cells (16,34). Therefore, we analyzed the expression of PKR and its mutants in transfected COS cells. PKR was immunoprecipitated from transfected cell lysate by anti-PKR monoclonal antibody and detected by Western blot with polyclonal antibody. Consistent with their kinase activities in vitro, wild-type PKR and mutant M11 were expressed at low levels in the transfected cells (Fig. 5B, lanes 3 and 7), whereas K296R mutant and M3 were expressed very well (Fig. 5B, lanes 4 and  5). Mutant M4 was expressed better than wild-type PKR (Fig.  5B, lane 6), whereas in another experiment, its expression was close to that of the K296R mutant (data not shown). Thus, the lack of inhibitory function of M4 in the COS cell assay is due to the loss of kinase activity rather than deficient protein expression.

DISCUSSION
The anti-viral effect of PKR is an important mechanism for host defense against viral infection and is mediated through the phosphorylation of eIF2␣ and inhibition of viral protein synthesis. In addition, PKR has been shown to have an antiproliferative function and to regulate cell growth (7,23,35), which is at least in part linked to eIF2␣ phosphorylation. In this study, we have examined the effects of mutations of the conserved amino acids around the PKR kinase insert region in vitro and in vivo. By using bacterially expressed PKR and eIF2␣, we were able to study their direct interaction in vitro. We found that amino acids in the conserved sequence LFIQMEFCD (amino acids 362-370) in subdomain V are important for PKR kinase activity, as the single point mutations L362Q (M3), I364N (M4), and F363S (M8) resulted in a loss of both kinase activity in vitro and growth and translation suppression function in vivo. However, these amino acids may not be directly involved in substrate recognition because all mutants were able to bind to eIF2␣, including the two deletion mutants, although some mutants bound eIF2␣ at a relatively low level compared with wild-type PKR. For example, M11 (E367R), which has kinase activity and is fully functional in vivo, showed only about 20% of normal (wild-type) eIF2␣ binding activity.
Because of the homologous nature of protein kinase domains, they all fold into topologically similar three-dimensional core structures and carry out phosphotransfer by a common mechanism (6). The crystal structures of more than ten eukaryotic protein kinases have been reported, and their kinase domains all fold into a two-lobed structure (6). The smaller lobe contains subdomains I-IV, and the larger lobe includes subdomains VI-XI. Subdomain V residues link the two lobes. The catalytic domain of PKR may also have the same general structure. According to sequence alignment with PKA-C␣, the N-terminal part of the conserved motif (LFIQM) in PKR kinase subdomain V is predicted to form a very hydrophobic ␤-strand (␤5) in the small lobe (Fig. 1A), and the hydrophobicity of these residues may be important for the maintenance of the two-lobed structure. This prediction is supported by our results showing that the deletion of this region (M12, M13) along with single point mutations that replace hydrophobic with hydrophilic residues (M3-L362Q, M4-I364N, M8-F363S) result in a loss of kinase activity. The C-terminal amino acids of the conserved motif (EFCD) form part of the extended chain that connects the two lobes. In PKA, Glu-127 on the extended chain participates in peptide binding by forming an ion pair with an arginine in the pseudosubstrate site of the PKA inhibitor peptide (39). The phosphorylation site in eIF2␣ also contains basic residues Cterminal to the serine phosphorylation site (Glu-Leu-Ser(P)-Arg-Arg) (2). PKR has no residues corresponding to PKA Glu-127 but has two acidic amino acids a few residues upstream, Glu-367 and Asp-370 (Fig. 1). However, the single point mutants M10 (E367Q) and M11 (E367R) have nearly full kinase activity, suggesting that Glu-367 has little effect on kinase structure. Nevertheless, in GST-pull down assays, M10 exhibits 50% eIF2␣ binding compared with wild-type PKR, and M11 has only 20% eIF2␣ binding activity, suggesting that Glu-367 may be involved in substrate anchoring through an ion pair with a basic residue in the substrate, possibly an arginine. This residue may not be crucial for substrate recognition, however, because the PKR mutants are still able to bind eIF2␣, and more important, they are fully functional in vivo (Figs. 3 and 5). Asp-370 may also contribute to eIF2␣ binding, but this remains to be tested by mutagenesis.
We have found that there is a contradiction in M4 function in yeast and in COS-1 cells. M4 is able to inhibit yeast growth but does not inhibit translation in COS-1 cells. The mechanism by which M4 inhibits yeast growth is unclear, although our data suggest that it may result from an interaction with yeast eIF2␣. In yeast, the level of phosphorylation of eIF2 by GCN2 during amino acid starvation is regulated precisely to lower the efficiency of reinitiation without affecting primary translation (35). It has been shown that low level expression of the mammalian kinases (PKR or HRI) or activation of GCN2 under amino acid starvation conditions leads to increased GCN4 expression, whereas yeast cells continue to grow and divide (36). In contrast, high level expression of the mammalian kinases or genetic activation of GCN2 severely reduces the growth rate of yeast cells (37,38). In our yeast system, GCN4 expression or low level eIF2␣ phosphorylation may be required for yeast growth because we used synthetic medium with only essential amino acids, requiring that the yeast synthesize nonessential amino acids. When M4 overexpression is induced by galactose, it might sequester eIF2␣ from phosphorylation by GCN2 due to its high affinity for eIF2␣, thus blocking the induction of GCN4 expression and inhibiting yeast growth. The reversal of this effect by the eIF2␣ mutant sui2M could then be explained by competition with endogenous eIF2␣ for binding to M4, freeing eIF2␣ to release the suppression of GCN4 expression. The mechanism of action of M4 in COS-1 cells is likely independent of eIF2␣ binding and may involve the formation of inactive heterodimers with endogenous PKR.
Finally, our data show that phosphorylation of PKR is not required for binding to its substrate, eIF2␣. The catalytically inactive mutant K296R has no kinase activity but is able to bind eIF2␣. Moreover, mutant M4 can bind eIF2␣ more readily than wild-type PKR. This increased affinity of M4 for eIF2␣ may result from a conformational change due to the replacement of the hydrophobic isoleucine residue with a hydrophilic asparagine. In any case, high affinity of the PKR mutants for eIF2␣ is not required for kinase function. M11 has only 20% of eIF2␣ binding activity versus wild-type PKR but is fully functional in vitro and in vivo. Mutant M1 has less than half of the normal eIF2␣ binding activity but has a higher kinase activity in vitro than wild-type PKR (Fig. 2).
The identification of the exact eIF2␣ recognition sites within PKR requires further experimentation. However, as the substrate binding sites may not be contiguous in primary sequence but scattered through the catalytic domain and close in the ternary structure, it will be essential to solve the structure of PKR to address this question directly. The results of luciferase activity assays were normalized by total protein concentration. For the CTRL, pGL2-control was cotransfected with pRc/CMV. Others were cotransfected with PKR wild-type (WT) or mutants. B, immunoprecipitation of PKR protein. COS-1 cell lysates containing 2 mg of total protein was immunoprecipitated with anti-PKR monoclonal antibody. Precipitated PKR proteins were separated by 10% SDS-PAGE and detected by Western blotting with anti-PKR polyclonal antibody. Lane 1 shows human PKR in interferon-treated Daudi cells. Lane 2 shows the endogenous monkey PKR of COS-1 cells. Lanes 3-7 show human PKR wild-type and mutant expression in COS-1 cells.