The Interferon-induced Double-stranded RNA-activated Protein Kinase PKR Will Phosphorylate Serine, Threonine, or Tyrosine at Residue 51 in Eukaryotic Initiation Factor 2α*

The family of eukaryotic initiation factor 2α (eIF2α) protein kinases plays an important role in regulating cellular protein synthesis under stress conditions. The mammalian kinases PKR and HRI and the yeast kinase GCN2 specifically phosphorylate Ser-51 on the α subunit of the translation initiation factor eIF2. By using an in vivo assay in yeast, the substrate specificity of these three eIF2α kinases was examined by substituting Ser-51 in eIF2α with Thr or Tyr. In yeast, phosphorylation of eIF2 inhibits general translation but derepresses translation of the GCN4 mRNA. All three kinases phosphorylated Thr in place of Ser-51 and were able to regulate general and GCN4-specific translation. In addition, both PKR and HRI were found to phosphorylate eIF2α-S51Y and stimulateGCN4 expression. Isoelectric focusing analysis of eIF2α followed by detection using anti-eIF2α and anti-phosphotyrosine-specific antibodies demonstrated that PKR and HRI phosphorylated eIF2α-S51Y on Tyr in vivo. These results provide new insights into the substrate recognition properties of the eIF2α kinases, and they are intriguing considering the potential for alternate substrates for PKR in cellular signaling and growth control pathways.

The family of eukaryotic initiation factor 2␣ (eIF2␣) protein kinases plays an important role in regulating cellular protein synthesis under stress conditions. The mammalian kinases PKR and HRI and the yeast kinase GCN2 specifically phosphorylate Ser-51 on the ␣ subunit of the translation initiation factor eIF2. By using an in vivo assay in yeast, the substrate specificity of these three eIF2␣ kinases was examined by substituting Ser-51 in eIF2␣ with Thr or Tyr. In yeast, phosphorylation of eIF2 inhibits general translation but derepresses translation of the GCN4 mRNA. All three kinases phosphorylated Thr in place of Ser-51 and were able to regulate general and GCN4-specific translation. In addition, both PKR and HRI were found to phosphorylate eIF2␣-S51Y and stimulate GCN4 expression. Isoelectric focusing analysis of eIF2␣ followed by detection using anti-eIF2␣ and anti-phosphotyrosine-specific antibodies demonstrated that PKR and HRI phosphorylated eIF2␣-S51Y on Tyr in vivo. These results provide new insights into the substrate recognition properties of the eIF2␣ kinases, and they are intriguing considering the potential for alternate substrates for PKR in cellular signaling and growth control pathways.
The human interferon-induced double-stranded RNA-activated protein kinase PKR, which functions in the cellular antiviral defense mechanism, is a member of a family of structurally related Ser/Thr kinases that specifically phosphorylate Ser-51 on the ␣ subunit of the translation initiation factor eIF2 1 (1,2). The binding of double-stranded RNA, thought to be generated during viral infections, is proposed to alter the conformation of PKR and activate the kinase to autophosphorylate (1,2). The active, phosphorylated form of PKR can then phosphorylate eIF2␣ on Ser-51 and convert eIF2 into an inhibitor of its guanine nucleotide exchange factor eIF2B, resulting in the inhibition of translation initiation (1,2). The other members of the eIF2␣ kinase family are the mammalian heme-regulated inhibitor of translation (HRI) that is activated by heme deprivation, the apparently ubiquitous kinase GCN2, first identified in yeast but also found in flies and mammals, which is activated under conditions of amino acid or purine nucleotide dep-rivation (1)(2)(3)(4), and the newly identified mammalian kinase PERK or PEK, a transmembrane kinase located in the endoplasmic reticulum that is activated under conditions of endoplasmic reticulum stress (5,6). In the yeast Saccharomyces cerevisiae, low level phosphorylation of eIF2␣ by GCN2 alters the pattern of translation reinitiation on the GCN4 mRNA and induces GCN4 expression (2). Increased synthesis of GCN4, a transcriptional activator of amino acid biosynthetic genes, enables cells to withstand amino acid starvation conditions. The mammalian eIF2␣ kinases PKR and HRI can substitute for GCN2 in yeast to phosphorylate eIF2␣ and stimulate GCN4 translation (7). In addition, high level phosphorylation of eIF2␣ in yeast by mutationally hyperactivated alleles of GCN2 or by overexpression of PKR or HRI severely inhibits general translation initiation and impairs cell growth (7)(8)(9)(10).
In addition to regulating translation by phosphorylating eIF2␣, PKR has been proposed to play roles in cell signaling (11) and growth control (12). In addition, several reports have proposed additional substrates for PKR (13)(14)(15)(16). In biochemical analyses, PKR has been shown to phosphorylate intact eIF2␣ or a 12-residue peptide containing the Ser-51 phosphorylation site (17). As part of our studies aimed to define the in vivo substrate recognition properties of PKR, we chose to examine the ability of PKR, HRI, and GCN2 to phosphorylate Thr or Tyr in place of Ser at residue 51 in eIF2␣. In general, protein kinases phosphorylate either Ser/Thr or Tyr residues, and these two classes of protein kinases are mutually exclusive. In fact, when Ser or Thr was substituted for a Tyr autophosphorylation site in the p130 gag-fps Tyr kinase of fujinami sarcoma virus, neither hydroxyamino acid was phosphorylated, and the mutant kinases had reduced enzymatic and oncogenic activities (18). Similarly, structural analyses of Ser/Thr and Tyr kinases support the view that the active sites of these enzymes are designed to ensure substrate selectivity (19,20). Our analyses have revealed that the three eIF2␣ kinases can phosphorylate Thr in place of Ser-51 in eIF2␣ and, in addition, that PKR and HRI can phosphorylate Tyr in place of eIF2␣-Ser-51. These results suggest that PKR and HRI may have a rather flexible active site that can accommodate Tyr as well as Ser/ Thr, and these results have interesting implications regarding the substrate recognition properties of PKR.
Assays of GCN4-lacZ Expression and Isoelectric Focusing (IEF) Gel Electrophoresis-Identical cell growth conditions were used for GCN4-lacZ expression and IEF gel electrophoresis assays. For strains expressing HRI or HRI-K199R, pre-cultures were incubated 2 days in SR medium (2% raffinose in place of glucose in SD medium) and then inoculated 1:50 into fresh SR medium and harvested after overnight growth. For strains expressing PKR or PKR-K296R, pre-cultures were grown 2 days in SD medium, and then cells (ϳ0.1 A 600 units) were inoculated into SGR medium (10% galactose plus 2% raffinose in place of glucose in SD medium) and harvested after overnight growth. The methods for growing strains expressing GCN2 or GCN2 c kinases, cell harvesting and breaking, ␤-galactosidase assays, and IEF gel electrophoresis have been described previously (22).
PKR expression from the plasmid pET-PKR was induced in the Escherichia coli strain BL21(DE3)pLysS as described previously (24). Cells were pelleted, suspended in lysis buffer (20 mM Tris-HCl (pH 7.5), 50 mM KCl, 150 mM NaCl, 20% glycerol, 1% Triton X-100, 0.5 mM EDTA) containing inhibitors (1 mM phenylmethylsulfonyl fluoride, 7 g/ml pepstatin, 50 mM NaF, 35 mM ␤-glycerol phosphate), and then disrupted by sonication. PKR expression from the plasmid p1420 was induced in a derivative of the gcn2⌬ yeast strain H2507 expressing eIF2␣-S51A as described above for the IEF analysis. Cells were pelleted, suspended in lysis buffer, and broken with glass beads in a Braun homogenizer as described previously (22). To induce PKR expression, HeLa cells were treated overnight with interferon and then harvested, washed with cold phosphate-buffered saline, suspended in 0.5ϫ RIPA buffer (25 mM Tris-HCl (pH 7.5), 75 mM NaCl, 0.5% Triton X-100, 0.5% sodium deoxycholate, 0.05% SDS) containing inhibitors (see above), and broken by subjecting them to freeze-thaw cycles three times. Following breakage, the lysates from all three cell types were cleared by centrifugation. Whole cell extracts were incubated with anti-phosphotyrosine antibodies prebound to protein A-Sepharose beads or with beads alone in 0.2 ml of lysis buffer at 4°C for 1 h with rocking. The beads were washed three times each with 0.5 ml of lysis buffer and then boiled 5 min in SDS loading buffer.

Regulation of Translation in Yeast Cells Expressing eIF2␣-
S51T or eIF2␣-S51Y in Place of Wild-type eIF2␣-To initiate an analysis of substrate recognition by the eIF2␣ kinases, we examined translational regulation in yeast strains in which Thr or Tyr was substituted for Ser-51 in eIF2␣. Plasmids that express wild-type or inactive forms of PKR or HRI or containing wild-type GCN2 or hyperactivated GCN2 c alleles were introduced into gcn2⌬ yeast strains expressing either eIF2␣-S51T (Thr-51), eIF2␣-S51Y (Tyr-51), eIF2␣-S51A (Ala-51), or wild-type eIF2␣ (Ser-51). Expression of the GCN2 c -513 kinase or PKR was toxic in strains expressing either the Ser-51 or Thr-51 forms of eIF2␣; however, the toxicity of the GCN2 c kinase was slightly reduced in the Thr-51 strain (Fig. 1, A and B, GCN3 sectors). In contrast, the eIF2␣-S51A and eIF2␣-S51Y mutations completely suppressed the toxic effects of the GCN2 c and PKR kinases. Translational regulation by the eIF2␣ kinases is dependent on both phosphorylation of eIF2␣ and the ability of phosphorylated eIF2 to inhibit its guanine nucleotide exchange factor eIF2B. The fact that expression of PKR and the other kinases showed no toxicity in the eIF2␣-S51Y strain suggests two possibilities as follows: 1) these kinases fail to phosphorylate eIF2␣-S51Y, or 2) the phosphorylated form of eIF2␣-S51Y is a weaker inhibitor of eIF2B than phosphorylated wild-type eIF2␣. This inhibition of eIF2B by phosphorylated eIF2 is dependent on the ␣ subunit of eIF2B, encoded by GCN3 in yeast (2,7). Deletion of GCN3 suppressed the toxicity resulting from expression of GCN2 c -513 or PKR in yeast strains expressing either the Ser-51 or Thr-51 forms of eIF2␣ (Fig. 1, A and B, gcn3⌬ sectors). These results suggest that the GCN2 c and PKR kinases efficiently phosphorylate eIF2␣ on Ser or Thr at residue 51 and that the phosphorylated forms of wild-type eIF2␣ and eIF2␣-S51T inhibit translation via the same mechanism.
A more sensitive assay of translational regulation in yeast is to monitor GCN4 expression. In wild-type strains phosphoryl-FIG. 1. Growth rate analysis and amino acid analog sensitivity of yeast strains expressing different eIF2␣ alleles and various eIF2␣ kinases. The indicated eIF2␣ proteins (S, wild-type eIF2␣; T, eIF2␣-S51T; Y, eIF2␣-S51Y; or A, eIF2␣-S51A) were expressed in the isogenic strains H2507 (gcn2⌬ GCN3) and J101 (gcn2⌬ gcn3⌬) as described under "Experimental Procedures." A, the resulting strains were then transformed with the plasmid p1052 carrying the GCN2 c -513 allele. The indicated transformants were streaked on SD plates and incubated 2 days at 30°C. B, the strains expressing the various eIF2␣ proteins were transformed with the plasmid p1420 that expresses the human PKR kinase under the control of a galactose-inducible promoter. The indicated transformants were streaked on an SGal plate (10% galactose) and incubated 6 days at 30°C. C, the strains described above, expressing the various eIF2␣ proteins, were transformed with plasmids encoding the indicated eIF2␣ kinases as follows: GCN2 (p722), PKR (p1420), PKR-K296R (p1421), HRI (p1246), and HRI-K199R (p1247). Expression of GCN2 was directed by the authentic GCN2 promoter, whereas the mammalian kinases were expressed under the control of a yeast galactose-inducible promoter. Patches of transformants were grown to confluence on SD medium and replica-plated to SGal medium and SGal plus 3-AT (10 mM) medium. Plates were incubated at 30°C for 3 days. ation of eIF2␣ by GCN2 stimulates GCN4 expression and enables cells to grow on medium containing 3-aminotriazole (3-AT), an inhibitor of histidine biosynthesis. Expression of GCN2 or HRI, but not the catalytically inactive HRI-K199R, conferred resistance to 3-AT in gcn2⌬ strains expressing either the Ser-51 or Thr-51 forms of eIF2␣ but not in strains expressing eIF2␣-S51A or eIF2␣-S51Y (Fig. 1C). High level expression of PKR was lethal in strains expressing wild-type eIF2␣ or eIF2␣-S51T, independent of 3-AT (Fig. 1C, left and right panels). In addition, expression of PKR, but not the catalytically inactive PKR-K296R, in eIF2␣-S51Y strains conferred resistance to 3-AT, suggesting that PKR can phosphorylate eIF2␣ on Tyr at residue 51 and that phosphorylated eIF2␣-S51Y can inhibit eIF2B and regulate GCN4 translation. Whereas the eIF2␣-S51Y strain expressing HRI failed to grow on SGal (10% galactose) medium containing 3-AT (Fig. 1C), it was partially resistant to 3-AT when grown on SR (2% raffinose) medium (data not shown).
To provide a quantitative measure of the regulation of GCN4 expression in the strains expressing the various eIF2␣ proteins and kinases, expression of a GCN4-lacZ reporter was assayed. For the GCN2 and GCN2 c -513 kinases the induction of GCN4-lacZ expression was similar in strains expressing either the Ser-51 or Thr-51 forms of eIF2␣ (Fig. 2). However, GCN4 expression in the eIF2␣-S51Y strain was almost identical to that observed in a strain expressing the non-phosphorylatable Ala-51 form of eIF2␣. In strains expressing wild-type eIF2␣, eIF2␣-S51T, or eIF2␣-S51Y, PKR generated 2-5-fold higher GCN4-lacZ expression than inactive PKR-K296R (Fig. 2). In addition, GCN4-lacZ expression was 2-fold higher in eIF2␣-S51Y versus eIF2␣-S51A strains expressing PKR. These results are consistent with the 3-AT-resistant phenotype noted for eIF2␣-S51Y strains expressing PKR (Fig. 1C) and support the idea that PKR can phosphorylate eIF2␣ on Tyr at residue 51.
IEF Analysis Reveals Phosphorylation of eIF2␣ in Vivo on Ser, Thr, or Tyr at Residue 51-When analyzed by IEF-PAGE, eIF2␣ resolves as a doublet. The lower species of this doublet co-migrates with eIF2␣ from strains expressing mutant kinases or eIF2␣-S51A and represents basal eIF2␣, whereas the upper species of eIF2␣ is phosphorylated on residue 51 (7,22). In IEF-PAGE analyses GCN2, GCN2 c -516, and more active GCN2 c -513 kinases were found to phosphorylate eIF2␣ on Ser or Thr but not Tyr or Ala at residue 51 (Fig. 3, A and B), consistent with the results of the genetic tests. In addition, phosphorylation of eIF2␣ was readily detected in strains expressing wild-type eIF2␣, eIF2␣-S51T, or eIF2␣-S51Y and either PKR or HRI but not PKR-K296R or HRI-K199R (Fig. 3C, upper panel, and data not shown).
To confirm that PKR and HRI were phosphorylating eIF2␣-S51Y on Tyr, a second IEF-PAGE analysis was performed. In yeast strains lacking the endogenous eIF2␣ kinase GCN2 and expressing either wild-type eIF2␣ or eIF2␣-S51Y, the expression of wild-type PKR or HRI resulted in eIF2␣ focusing as a doublet on IEF gels (Fig. 4, upper panel, lanes 2, 4, 5, and 7). When the same blot from Fig. 2 (upper panel) was probed with anti-phosphotyrosine antibodies (Fig. 4, lower panel), crossreactive bands were only detected in samples from strains expressing eIF2␣-S51Y and a wild-type kinase. When the two blots (Fig. 4, upper and lower panels) were overlaid, the antiphosphotyrosine cross-reactive species aligned perfectly with the upper, hyperphosphorylated form of eIF2␣. These results confirmed that PKR and HRI were phosphorylating eIF2␣-S51Y on Tyr and demonstrated that in vivo these proteins possess Tyr kinase activity. Comparison of the ratio of the hyperphosphorylated to the basal form of wild-type eIF2␣ and eIF2␣-S51Y in strains expressing PKR (see Fig. 4, lanes 4 and 5; also Fig. 3C, upper panel, 1st and 5th lanes) may suggest that eIF2␣-S51Y is a poorer substrate for PKR than is wild-type eIF2␣; however, alternative interpretations of these results are FIG. 2. Regulation of GCN4 expression by GCN2 and PKR in yeast strains expressing various eIF2␣ mutant proteins. The indicated kinases were expressed in derivatives of the gcn2⌬ strain H1925 containing the various eIF2␣ proteins. ␤-Galactosidase activities expressed from an integrated wild-type GCN4-lacZ fusion were measured in whole cell extracts and are the averages of 2-3 independent transformants; S.Es. were 32% or less. For strains expressing GCN2 or GCN2 c -513, cells were grown under either non-starvation conditions where GCN4 expression is repressed (R) or under amino acid starvation conditions imposed by the addition of 10 mM 3-AT, where GCN4 expression is derepressed (DR). The PKR and PKR-K296R proteins were expressed from a yeast GAL-CYC1 hybrid promoter. For assays, cells were grown exponentially in SD medium, where kinase expression is low, and then shifted to SGR-inducing medium. Cells were harvested after overnight growth in inducing medium. The GCN4-lacZ expression values obtained for GCN2 and PKR cannot be directly compared because the different media used for these cultures resulted in altered basal levels of GCN4-lacZ expression, as is apparent in the eIF2␣-S51A strain.
FIG. 3. IEF gel electrophoresis of eIF2␣ from yeast strains expressing various eIF2␣ proteins and the GCN2, GCN2 c , or PKR kinase. Plasmids that express the indicated eIF2␣ kinases were introduced into gcn2⌬ yeast strains expressing the indicated eIF2␣ proteins from low copy number plasmids. A and B, strains expressing GCN2 and GCN2 c kinases. Yeast strains expressing the indicated eIF2␣ proteins were transformed with plasmids that express wild-type GCN2 (p722), or the constitutively activated GCN2 c -516 (p1056), or GCN2 c -513 (p1052) kinases under the control of the natural GCN2 promoter. Cells were grown under nonstarvation conditions (R) or amino acid starvation conditions invoked by the addition of 10 mM 3-AT (DR), as indicated. C, strains expressing PKR. Plasmids expressing wild-type PKR (p1420) or the inactive mutant PKR-K296R (p1421) under the control of a GAL-CYC1 hybrid promoter were introduced into gcn2⌬ yeast strains expressing the indicated eIF2␣ proteins. Extracts were prepared from cells grown exponentially in SD medium and then shifted to SGR medium to induce PKR expression. Immunoblot analysis using PKR monoclonal antibodies on 50-g aliquots of the same extracts used for IEF-PAGE are aligned below the IEF data. provided below. The inability to detect Tyr phosphorylation by GCN2 may suggest that GCN2 is an inherently less active kinase than HRI or PKR or that the GCN2 active site cannot accommodate a Tyr residue. Alternatively, it is likely that HRI and PKR were expressed at higher levels than GCN2 in these experiments, so it may be possible to detect Tyr kinase activity if we express GCN2 at higher levels in yeast cells.
The expression of PKR is subject to negative translational autoregulation in both yeast (7,10) and mammalian (1) cells such that kinase expression is inversely related to its effects on cellular translational activity. For example, although PKR was expressed at lower levels than PKR-K296R in strains expressing wild-type eIF2␣ (Fig. 3C, lower panel, 1st 2 lanes), this autoregulation was relieved in eIF2␣-S51T strains and abolished in eIF2␣-S51A and eIF2␣-S51Y strains (Fig. 3C, lower panel, last 6 lanes). In addition, whereas the GCN2 c -513 kinase appeared to phosphorylate wild-type eIF2␣ and eIF2␣-S51T to the same extent (Fig. 3B), the eIF2␣-S51T strain grew significantly better (Fig. 1). This lack of correlation between eIF2␣ phosphorylation and growth rate was also observed in eIF2␣-S51Y strains expressing PKR as noted previously. These results are consistent with a model in which the phosphorylated forms of eIF2␣-S51T and eIF2␣-S51Y are weaker inhibitors of eIF2B than phosphorylated wild-type eIF2␣. Finally, whereas the eIF2␣, eIF2␣-S51T, and eIF2␣-S51Y proteins were phosphorylated to similar levels, it cannot be concluded that PKR and HRI are equally efficient at phosphorylating these three different amino acids at residue 51, because in vivo phosphorylation levels are dependent on the balance between kinase and phosphatase activities. As we do not know the identity or the efficiency of the phosphatases that dephosphorylate these three eIF2␣ proteins, we cannot at this time evaluate the relative efficiencies of PKR and HRI to phosphorylate eIF2␣ in vivo on Ser versus Thr or Tyr at residue 51.
Immunodetection of Human PKR Using Anti-phosphotyrosine Antibodies-Previously it has been reported that mouse PKR, also known as TIK, can be detected in immunoblot assays using anti-phosphotyrosine antibodies (25). To determine if human PKR is also immunoreactive with anti-phosphotyrosine antibodies, wild-type human PKR and the inactive PKR-K296R proteins were expressed in E. coli, and crude protein extracts were separated by SDS-PAGE followed by immunoblotting with anti-PKR or anti-phosphotyrosine antibodies. As shown in Fig. 5A (right panel), both wild-type PKR and the PKR-K296R mutant proteins were expressed in E. coli. When the same membrane was probed with affinity purified anti-phosphotyrosine antibodies, cross-reactive species were detected in extracts prepared from cells expressing wild-type PKR but not PKR-K296R (Fig. 5A, left panel). Similar results demonstrating that recombinant wild-type human and mouse PKR (TIK), but not catalytic mutants, cross-react with anti-phosphotyrosine antibodies were recently published during the course of these experiments (26). In addition to the prominent anti-phospho- FIG. 4. IEF gel electrophoresis of eIF2␣ and eIF2␣-S51Y from yeast strains expressing PKR and HRI. Plasmids expressing wildtype HRI (p1246) and PKR (p1420) as well as the inactive mutants HRI-K199R (p1247) and PKR-K296R (p1421) under the control of a GAL-CYC1 hybrid promoter were introduced into gcn2⌬ yeast strains expressing either wild-type eIF2␣ (lanes 1-4) or eIF2␣-S51Y (lanes 5-8) from high copy number plasmids. For strains expressing HRI, extracts were prepared from cells grown overnight under inducing conditions in SR medium. For strains expressing PKR, extracts were prepared from cells grown exponentially in SD medium and then shifted to SGR medium to induce PKR expression. In the upper panel the blot was probed with anti-eIF2␣ antiserum, and in the lower panel the same blot was probed with affinity purified anti-phosphotyrosine antibodies.

FIG. 5. Immunodetection of human PKR using anti-phosphotyrosine antibodies.
A, immunoblot analysis of PKR expressed in E. coli using anti-PKR and anti-phosphotyrosine antibodies. Lysates of E. coli BL21(DE3)pLysS cells expressing wild-type PKR, PKR-K296R, or no PKR (vector) were subjected to 10% SDS-PAGE followed by immunoblotting with polyclonal anti-phosphotyrosine antibodies (left panel). Immune complexes were detected by chemiluminescence. The blot was then stripped according to the vendor's instructions, probed with anti-PKR monoclonal antibodies (right panel), and immune complexes were again detected by chemiluminescence. The migration of molecular mass markers (kDa) is indicated on the left. B, immunoprecipitation of PKR using anti-phosphotyrosine antibodies. Crude protein extracts from the indicated cells, described below, were incubated with anti-phosphotyrosine antibodies prebound to protein A-Sepharose beads or with beads alone (no antibody, no Ab). Immunoprecipitated proteins were eluted in SDS sample buffer and subjected to SDS-PAGE followed by immunoblotting with anti-PKR monoclonal antibodies. Lanes 1-3, a derivative of the gcn2⌬ yeast strain H2507 expressing eIF2␣-S51A and transformed with the PKR expression vector p1420 was grown in SGR medium to induce PKR expression. For immunoprecipitation reactions 200 g of whole cell extract was used; the loading control was 4 g of crude extract. Lanes 4 -6, expression of wild-type PKR in the E. coli strain BL21(DE3)pLysS was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to the culture medium. For immunoprecipitation (IP) reactions 200 g of extract was used; the loading control was 8 g of crude extract. Lanes 7-9, HeLa cells were treated overnight with interferon (␣) to induce PKR expression. For immunoprecipitation reactions 2.5 mg of whole cell extract was used; the loading control was 50 g of crude extract. C, immunoprecipitation of PKR, but not PKR-K296R, using antiphosphotyrosine antibodies. Crude protein extracts from E. coli cells expressing PKR or PKR-K296R were incubated with anti-phosphotyrosine antibodies prebound to protein A-Sepharose beads. Immunoprecipitated proteins were eluted and analyzed by SDS-PAGE as described above. For the immunoprecipitation reactions 500 g of extract was used; the loading control was 2 g of crude extract. tyrosine antibody cross-reactive species co-migrating with PKR at Ϸ70 kDa, several larger proteins were detected (Fig. 5A, left  panel). This result suggests that PKR expressed in E. coli may autophosphorylate on Tyr and can also phosphorylate certain bacterial proteins on Tyr.
To confirm that PKR was the major species cross-reacting with anti-phosphotyrosine antibodies, we performed immunoprecipitation reactions. As shown in Fig. 5B (lanes 3 and 6), recombinant PKR could be immunoprecipitated from both yeast and bacterial cell extracts using affinity purified antiphosphotyrosine antibodies. PKR was not precipitated when the anti-phosphotyrosine antibodies were omitted from the reactions (Fig. 5B, compare lanes 3 versus 2 and 6 versus 5). In addition, the precipitation was specific for functional PKR because the catalytically inactive PKR-K296R protein expressed in E. coli could not be precipitated using anti-phosphotyrosine antibodies (Fig. 5C, lanes 3 and 4). As shown in Fig. 5B (lanes  7-9), the endogenous PKR expressed in interferon-treated HeLa cells could also be immunoprecipitated using the antiphosphotyrosine antibodies. Whereas these results suggest that PKR can autophosphorylate on Tyr, phosphoamino acid analyses of PKR isolated from in vivo labeled yeast and bacterial cells revealed phosphoserine and phosphothreonine, but not phosphotyrosine (data not shown). This latter result is consistent with the results of Icely et al. (25), who could only find phosphoserine and phosphothreonine in mouse PKR despite the fact that mouse PKR also cross-reacted with antiphosphotyrosine antibodies. Icely et al. (25) speculated that the anti-phosphotyrosine antibodies may have recognized an unusual epitope on mouse PKR (25); however, negative results in phosphoamino acid analysis may reflect a low phosphorylation stoichiometry or a labile phosphotyrosine residue (27). In addition, it has been proposed that anti-phosphotyrosine antibodies may be over 100-fold more efficient at detecting phosphotyrosine than is phosphoamino acid analysis (28). Due to these conflicting results and the plausible explanations for failure to detect phosphotyrosine in the phosphoamino acid analyses, we are unable to conclude whether human PKR autophosphorylates on Tyr. DISCUSSION We have shown that the kinases PKR, HRI, and GCN2 can phosphorylate eIF2␣ on Ser or Thr at residue 51. In addition both PKR and HRI can phosphorylate eIF2␣ on Tyr at residue 51 in vivo. It is generally accepted that Ser and Thr kinases are structurally similar, and many kinases are known to phosphorylate both residues, so the finding that the eIF2␣ kinases could phosphorylate Thr in place of Ser-51 is not surprising. However, the phosphorylation of Tyr at residue 51 in eIF2␣ by PKR and HRI is unexpected.
Two proposals can account for the Tyr phosphorylation activity by PKR and HRI. In the first proposal PKR and HRI would recognize eIF2␣ with high affinity and simply phosphorylate any hydroxyl group present at residue 51. Based on a mutational analysis of the vaccinia virus K3L protein, a pseudosubstrate inhibitor of PKR with homology to eIF2␣, we have proposed that PKR utilizes a sequence element over 30 residues from the site of phosphorylation to recognize eIF2␣ (29). According to this model, the ability of PKR to phosphorylate Tyr in place of Ser-51 in eIF2␣ simply reflects the strong contribution of this remote sequence for substrate recognition by PKR and the lack of specificity determinants around residue 51. Consistent with this idea, it has ben reported that the PKR and HRI phosphorylation of intact eIF2 is roughly 3 orders of magnitude more efficient than phosphorylation of a 12-residue synthetic peptide containing the Ser-51 phosphorylation site (30). However, regardless of how PKR initially recognizes eIF2␣, it is important to note that the kinase active site must be able to accommodate both the alkyl hydroxyl groups of Ser and Thr and the phenolic hydroxyl of Tyr. The crystal structures solved to date for both Ser/Thr and Tyr kinases suggest that these enzymes would be unable to phosphorylate substrates on the alternate phospho-accepting residue due principally to steric limitations (20). Indeed, it has previously been shown that the p130 gag-fps Tyr kinase could not phosphorylate Ser or Thr in place of an authentic Tyr phosphorylation site (18). Therefore, it is reasonable to expect that PKR and HRI possess a unique and more flexible structure as compared with the traditional Ser/Thr or Tyr kinases.
The second proposal to account for the Tyr kinase activity of PKR and HRI is that these proteins are members of the class of dual specificity protein kinases. A number of kinases have been proposed to have dual specificity; however, the criteria used in making this assignment has not been standardized (27,31). Several kinases autophosphorylate on both Tyr and Ser/Thr residues (27,31), and the mitogen-activated protein kinase/ extracellular signal-regulated kinase kinase kinase (32) phosphorylates the extracellular signal-regulated kinases on both Tyr and Thr. The wee1 kinase, although structurally related to the Ser/Thr class of protein kinases, phosphorylates the cdc2 kinase on Tyr (27). The Myt1 kinase, identified in both Xenopus and humans and a member of the wee1 family of kinases, phosphorylates Cdc2 on both Thr and Tyr (33); and recently, Myt1 has also been shown to autophosphorylate in vitro on Ser, Thr, and Tyr (34). In contrast to these kinases in which phosphotyrosine and phosphoserine or phosphothreonine was readily detected in substrate phosphorylation or autophosphorylation reactions, other proposed dual specificity kinases, including TIK (25) and PYT/ESK/TTK (28), were identified based primarily on cross-reactivity with anti-phosphotyrosine antibodies. These kinases structurally resemble Ser/Thr kinases; however, when expressed in bacteria they cross-reacted with anti-phosphotyrosine antibodies. In this report we have demonstrated that human PKR (the TIK homolog) will also crossreact with anti-phosphotyrosine antibodies. Whereas these results may suggest that PKR is a dual specificity kinase, it will be necessary to identify an in vivo substrate that PKR phosphorylates on Tyr to conclude convincingly that PKR is a member of the class of dual specificity protein kinases.
The identification of Tyr and Thr kinase activity by PKR is very interesting in regards to alternative substrates and biological roles proposed for PKR. Although eIF2␣ is the only well characterized PKR substrate, and the regulation of translation is thought to be the primary function of PKR, recent reports suggest additional substrates and roles for PKR. PKR has been reported to phosphorylate IB (13), HIV Tat (14,15), and NF90 (16) in vitro. In addition, overexpression of catalytically inactive mutants of PKR leads to heightened viral sensitivity (35), altered gene regulation (Ref. 36 and references therein) and malignant transformation (12). At least some of the effects of PKR on gene regulation do not appear to be mediated by changes in eIF2␣ phosphorylation, suggesting that phosphorylation of other proteins may mediate these effects (36). Finally, mice deficient in PKR are impaired in cell signaling pathways including their interferon-␥ and double-stranded RNA-induced antiviral response (11), although these effects are not observed in all PKR null mice (37). Our studies on eIF2␣ phosphorylation by PKR raise the possibility that PKR may have other substrates that it naturally phosphorylates on Thr or Tyr. As Tyr phosphorylation is a common step in cellular signal transduction pathways, it is tempting to speculate that the defects in signal transduction pathways and transcriptional regulation of gene expression associated with reduced PKR function are due to a loss of the PKR Tyr kinase activity. It will be interesting to identify these alternative PKR substrates and to determine if they are phosphorylated on Tyr, and thereby provide a physiological role for this unexpected Tyr kinase activity of PKR.