Ribosome Targeting of PKR Is Mediated by Two Double-stranded RNA-binding Domains and Facilitates in Vivo Phosphorylation of Eukaryotic Initiation Factor-2*

Protein kinase PKR is activated in mammalian cells during viral infection, leading to phosphorylation of the α subunit of eukaryotic initiation factor-2 (eIF-2α) and inhibition of protein synthesis. This antiviral response is thought to be mediated by association of double-stranded RNA (ds-RNA), a by-product of viral replication, with two ds-RNA-binding domains (DRBDs) located in the amino terminus of PKR. Recent studies have observed that expression of mammalian PKR in yeast leads to a slow growth phenotype due to hyperphosphorylation of eIF-2α. In this report, we observed that while DRBD sequences are required for PKR to function in the yeast model system, these sequences are not required for in vitrophosphorylation of eIF-2α. To explain this apparent contradiction, we proposed that these sequences are required to target the kinase to the translation machinery. Using sucrose gradient sedimentation, we found that wild-type PKR was associated with ribosomes, specifically with 40 S particles. Deletions or residue substitutions in the DRBD sequences blocked kinase interaction with ribosomes. These results indicate that in addition to mediating ds-RNA control of PKR, the DRBD sequences facilitate PKR association with ribosomes. Targeting to ribosomes may enhance in vivo phosphorylation of eIF-2α, by providing PKR access to its substrate.

RNA-binding proteins are important participants in posttranscriptional regulation of gene expression. Sequence comparisons between these proteins has led to the discovery of several RNA-binding motifs, including the double-stranded RNA binding domain (DRBD) 1 (1)(2)(3)(4)(5). The DRBD is about 65 residues in length, with a lysine-rich sequence at the carboxylterminal end (1,2,4). More than 20 different proteins containing DRBD sequences have been identified (1). One of the best characterized DRBD-containing proteins is the doublestranded RNA-dependent protein kinase, PKR, that functions in a cellular antiviral response and is transcriptionally induced by interferon (6 -10). Upon viral infection in mammalian cells, PKR inhibits protein synthesis by phosphorylation of the ␣ subunit of eukaryotic initiation factor-2 (eIF-2) (6 -8). The eIF-2 bound to Met-tRNA i Met and GTP associates with 40 S ribosomal subunits and facilitates recognition of the start codon during translation initiation (11)(12)(13)(14). Upon completion of this process, the GTP complexed with eIF-2 is hydrolyzed to GDP. Phosphorylation of eIF-2␣ by PKR impedes the exchange of eIF-2-GDP to the GTP-bound form that is required for the next round of translation initiation. The resulting reduction in eIF-2-GTP levels inhibits cellular protein synthesis and blocks viral proliferation into neighboring cells (11)(12)(13).
Activation of PKR during viral infection is thought to be regulated directly by double-stranded RNA (ds-RNA) that is produced during viral replication. Evidence supporting this model is the observation that the addition of ds-RNA to a purified in vitro system stimulates both PKR autophosphorylation and phosphorylation of eIF-2␣ (7,8,(15)(16)(17). Regulation of PKR by ds-RNA is thought to be mediated by two DRBD sequences, termed motif 1 and motif 2, that are located in the amino-terminal portion of the kinase (Fig. 1). Numerous reports have shown that recombinant proteins containing motifs 1 and 2 can bind ds-RNA in vitro assays (18 -25). Motif 1 is a closer match to the DRBD consensus and has a higher affinity for ds-RNA than motif 2 (18 -22, 24, 26). However, both DRBD sequences are thought to be required for maximal ds-RNA binding to PKR. Accumulatively, these studies suggest a mechanism for activation of PKR whereby ds-RNA binds motifs 1 and 2 and alters the conformation of the kinase leading to stimulation of PKR function (7, 18 -22, 24, 25). Although the molecular details of this regulation are not yet clear, it is thought that ds-RNA-induced autophosphorylation of PKR participates in kinase activation, perhaps by enhancing the affinity of PKR for ATP (6,7,15,27). Autophosphorylation may occur between two PKR molecules that are linked by proteinprotein contacts or a ds-RNA bridge (7, 8, 28 -32).
Several lines of research have supported the in vivo role of the DRBD in mediating ds-RNA activation of PKR. Expression of human PKR in yeast Saccharomyces cerevisiae deficient for its endogenous eIF-2␣ kinase encoded by GCN2 led to hyperphosphorylation of eIF-2␣ at serine 51, resulting in a slow growth defect due to a general inhibition of translation initiation (18,(33)(34)(35). When mutant forms of PKR with deletions in the motif 1 or 2 sequences were expressed in yeast, there was no phosphorylation of eIF-2␣ and near wild-type levels of growth (34). Experiments analyzing the activity of mutant forms of PKR in mammalian cells have relied on the observation that PKR specifically inhibits its own translation (29,36,37). From studies using biochemical fractionation and immunofluorescent staining, the majority of PKR in mammalian cells was found to be associated with ribosomes (7, 38 -40). The remaining portion, estimated at less than 20% of the total PKR, was found in the nucleus in proximity to the nucleoli (40). It was proposed that newly synthesized PKR associates with ribosomes in close vicinity to its own mRNA, resulting in a localized phosphorylation of eIF-2␣ that would preferentially diminish the synthesis of PKR protein. Transfection of mammalian cells with PKR cDNAs showed that alterations in an invariant kinase residue or in DRBD sequences decreased its ability to regulate its own synthesis (29,36,37).
Interestingly, expression of certain kinase-inactive mutant forms of PKR in NIH 3T3 cells conferred a malignant transformation phenotype, and subcutaneous injection of these transfected cells in nude mice gave rise to rapid tumor growth (41)(42)(43)(44). When similar experiments were carried out with wildtype PKR, no transformed phenotype was observed. These results suggest that the kinase functions as a tumor suppressor, since expression of a mutant PKR in NIH 3T3 cells appeared to reduce the activity of the endogenous wild-type PKR (42,43). Furthermore, it suggests that levels of eIF-2␣ phosphorylation in cells are important for control of cell proliferation. Consistent with this model, Donze et al. (45) showed that expression of a mutant form of eIF-2␣ that contains an alanine at the PKR phosphorylation site, serine 51, in NIH 3T3 cells resulted in malignant transformation. However, other mechanisms for cell transformation by PKR have not been ruled out. Recent studies have shown that PKR can also phosphorylate the IB family of inhibitors that regulate the transcription factor NF-B (46,47).
In this report, we observed that while motifs 1 and 2 are essential for PKR function in the yeast model system, these sequences are not required for in vitro phosphorylation of eIF-2␣. To explain this apparent contradiction, we proposed that the in vivo requirement for DRBD sequences is due to their role in targeting PKR to the translation machinery. Using sucrose gradient sedimentation, we did indeed find that wild-type PKR was associated with ribosomes, specifically with the 40 S particles. Deletions or residue substitutions in motif 1 or 2 blocked kinase interaction with ribosomes. Together, these results indicate that the DRBD sequences carry out a dual function in the regulation of PKR phosphorylation of eIF-2. First, motifs 1 and 2 are proposed to mediate ds-RNA stimulation of kinase activity. Our results are consistent with the idea that the DRBD sequences function together to repress PKR kinase activity. Binding of ds-RNA to these motifs would release their inhibitory affect, allowing for activation of the kinase. The second function of the DRBD sequences is appropriate targeting of PKR to the translational machinery. Localization of PKR to ribosomes may increase in vivo access to its eIF-2 substrate.
Kinase reaction samples containing 5 g of protein were first prephosphorylated with 20 M nonradiolabeled ATP in 90 l of kinase buffer supplemented with protease inhibitors. This step reduced the incorporation of 32 P into endogenous yeast proteins and can be eliminated without affecting the measured specific activity relative to wildtype PKR. After incubating the samples at 30°C for 20 min, 2 g of recombinant eIF-2␣ substrate and 20 Ci of [␥-32 P]ATP in a final concentration of 40 M ATP were added to the kinase reactions. In experiments measuring PKR activation by ds-RNA, 0.1 or 1 g/ml poly(I)⅐poly(C) (Sigma) was also added to the assay. As described previously (53), the eIF-2␣ substrate is a modified form of yeast eIF-2␣ that is deleted for residues 200 -304 and contains a polyhistidine "tag" for rapid purification. The deletion of the carboxyl-terminal residues of eIF-2␣ removes three phosphorylation sites for casein kinase II (54,55). The final volume of the kinase reaction was 100 l, and 15-l aliquots were removed from the sample after 1, 2, 4, 10, and 20 min of incubation at 30°C. The aliquots were mixed with an equal volume of 2 ϫ SDS sample buffer, heated at 95°C for 5 min, and analyzed by electrophoresis in a 10% SDS-polyacrylamide gel. All in vitro kinase assays were carried out in at least three independent experiments. The eIF-2␣ in the kinase assays often showed some modest proteolytic degradation resulting in two bands of phosphorylated substrate. PKR immunoprecipitation kinase assays were carried out as described previously (56).
Levels of eIF-2␣ phosphorylation were quantitated by measuring 32 P incorporation using a Bio-Rad model GS-250 molecular imager. The 32 P levels/mg of total protein for each sample were plotted versus time, and the initial velocity, presented in arbitrary units and normalized to wild-type PKR, was determined by measuring the slope of the linear portion of the curve ( Table I). The specific activity for each PKR protein was determined by dividing the initial velocity by the level of kinase protein relative to wild-type PKR.
Another often used measure of PKR activity is autophosphorylation. In our in vitro assay we used 5 g of cell lysate in a 100-l reaction volume. This lysate concentration was in the linear range of the assay. Under these conditions, we detected very little autophosphorylation of PKR. When 100 g of lysate was used in similar reaction conditions, we observed incorporation of 32 P in wild-type PKR that was absent in the kinase-deficient mutant PKR-K296R. Under similar assay conditions, PKR-⌬1 showed a modest reduction in autophosphorylation compared with wild-type PKR when normalized for steady-state protein levels. Mutant PKR-⌬2 and PKR-⌬14 -257 showed larger reductions in the levels of kinase autophosphorylation. Correlating autophosphorylation and catalytic activity is difficult because these deleted sequences in the PKR mutants may remove potential autophosphorylation sites. Taylor et al. (57) identified three in vitro autophosphorylation sites in PKR, Ser-242, Thr-255, and Thr-258. Two of these sites are absent in PKR-⌬14 -257. Remaining sites of autophosphorylation are not currently characterized, and their role in activation of PKR is not yet fully understood. For these reasons we relied on the exogenous eIF-2 substrate as a more reliable measure of PKR catalytic activity.
Immunoblot Analysis of PKR-Samples containing 20 g of protein prepared for the in vitro kinase assays were analyzed by electrophoresis in a 10% SDS-polyacrylamide gel and then transferred to a nitrocellulose filter. Immunoblot filters were blocked in a TBS-T solution containing 20 mM Tris-HCl (pH 7.6), 150 mM NaCl, 0.1% Tween 20, and 5% nonfat dry milk and then incubated in TBS-T solution containing either PKR monoclonal antibody 70 -10 (58) or rabbit polyclonal antiserum PKR K-17 that was prepared against a carboxyl-terminal polypeptide of PKR (Santa Cruz Biotechnology, Inc.). Western blot results obtained with polyclonal antiserum PKR K-17 were independently confirmed with a monoclonal antibody recognizing an epitope in the carboxylterminal kinase domain of PKR. Filters were washed in TBS-T, and PKR-antibody complex was detected using horseradish peroxidase-labeled secondary antibody provided in the ECL Western blotting analysis system (Amersham Corp.). Relative amounts of PKR in the immunoblot experiments were quantitated by measuring band intensities using a Bio-Rad model GS-670 imaging densitometer from autoradiographs generated by different length exposures.
Ribosome Association-Transformants of H1817 containing different alleles of PKR were grown in synthetic medium containing 10% galactose and 2% raffinose for 15 h, and 50 g/ml cycloheximide was added to the culture 5 min before harvesting. Cells were chilled on ice, collected by centrifugation, and washed once with breaking solution (20 mM Tris-HCl (pH 7.9), 50 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 50 g/ml cycloheximide, and 200 g/ml heparin). Each of the subsequent steps was performed at 4°C, and protease inhibitors (1 M pepstatin, 1 M leupeptin, 0.15 M aprotinin, and 100 M phenylmethylsulfonyl fluoride) were added to all solutions. Cells were resuspended in breaking solution, lysed by glass beads using a Vortex mixer, and clarified by centrifugation at 12,000 ϫ g for 25 min. Supernatant samples containing 20 A 260 units were loaded onto a 5-47% sucrose gradient in breaking solution without heparin, and ultracentrifugation was performed using a Beckman rotor SW41 at 39,000 rpm for 3 h (59). Gradients were fractionated using an ISCO UA-6 absorbance monitor set at 254 nm, and 0.5-ml aliquots were collected. Proteins from each aliquot were precipitated by adding 5% trichloroacetic acid and separated by electrophoresis in a 10% SDS-polyacrylamide gel. PKR was detected by immunoblot analysis. To characterize PKR association with ribosomes in the absence of MgCl 2 , no cycloheximide was added before harvesting the cells, and MgCl 2 , cycloheximide, and heparin were omitted from the breaking solution. Sucrose gradient sedimentation and SDS-polyacrylamide gel electrophoresis were carried out as described for the fractionation studies performed in the presence of MgCl 2 .

DRBD Sequences Are Required for in Vivo Function in Yeast but Are Dispensable for in Vitro Phosphorylation of eIF-2␣-
Yeast is a useful model system to study the in vivo role of PKR sequences in the control of its kinase activity (18,(33)(34)(35). Dever et al. (33) showed that PKR expressed from a galactose-inducible promoter in yeast leads to high levels of phosphorylation of eIF-2␣ and a severely reduced rate of cellular growth. Consistent with these previous reports, we found this growth defect is substantially alleviated in yeast containing a kinase-defective PKR-K296R or mutants deleted in motif 1 or 2 ( Fig. 1 and Table  I). Romano et al. (34) found that these mutant PKR proteins were unable to phosphorylate eIF-2␣ in vivo. These observations were interpreted as supporting the model that ds-RNA endogenous in yeast interacted with the two DRBD sequences and stimulated PKR phosphorylation of eIF-2␣.
We next wished to measure the in vitro activity of the different mutant versions of PKR to determine whether these pro-teins were impaired for phosphorylation of eIF-2␣. As noted earlier, translational expression of PKR is autoregulated in mammalian cells. This appears to be also true in the yeast system, as illustrated by the observation by Romano et al. (34) that the protein levels of wild-type PKR were greatly reduced compared with the kinase-defective PKR-K296R mutant. We have confirmed this observation, with wild-type PKR expressed in strain H1816 (⌬gcn2) showing less than 5% of the protein levels of mutant PKR-⌬1, PKR-⌬14 -257, or PKR-K296R as judged by immunoblot analysis (data not shown). To avoid these large differences in PKR protein levels, we used strain H1817 (⌬gcn2) that contains an alanine substituted for the eIF-2␣ phosphorylation site serine 51. Expression of this nonphosphorylatable form of eIF-2␣ blocks both the slow growth phenotype associated with hyperphosphorylation of eIF-2␣ and autoregulation of PKR expression (34). Cell lysates were prepared from H1817 expressing the different PKR alleles, and the kinase protein levels were measured by immunoblotting (Fig.  2). Different PKR mutant protein levels ranged from 0.5 to 1.5 times that measured for wild-type PKR (Table I).
Cell lysates were prepared from H1817 expressing different PKR mutant proteins and analyzed for phosphorylation of recombinant eIF-2␣ substrate. As illustrated in the autoradiogram shown in Fig. 3, eIF-2␣ substrate was phosphorylated in the sample containing wild-type PKR, while no eIF-2␣ phosphorylation was detected in lysates prepared from cells expressing no PKR (⌬PKR) or the kinase-defective PKR-K296R and PKR-⌬K proteins (Fig. 3, Table I). To confirm that phosphorylation of eIF-2␣ was on serine 51, we carried out a control experiment using recombinant eIF-2␣ containing alanine for serine 51 (eIF-2␣-S51A). No phosphorylation of eIF-2␣-S51A was detected in the wild-type PKR reaction (Fig. 3). It is thought that PKR expressed in yeast is at least partially activated by endogenous ds-RNA (34,35). Consistent with the view that PKR in the cell lysates is in the activated conformation, we found that the addition of poly(I)⅐poly(C) to in vitro kinase reactions did not further increase phosphorylation of eIF-2␣ by PKR (data not shown). In a parallel experiment, we immuno-  The initial velocity of enzymatic activity for wild-type and mutant PKR expressed in H1817 (⌬gcn2 SUI2-S51A) was measured by plotting 32 P emission levels/mg of total protein for each sample versus time. Initial velocity, presented in arbitrary units and normalized to wildtype PKR, was determined by measuring the slope of the linear portion of the curve. Results shown are averages of 3-5 independent experiments, and individual measurements deviated from the average values shown by less than 10%.
b Molar ratio of different PKR proteins. Extracts used in the initial velocity experiments were used to measure the relative levels of PKR protein by immunoblot analysis. Values were derived from three independent experiments, and the representative results are shown in Fig. 2.
c Specific activities of wild-type and mutant PKR were calculated as initial velocity (footnote a) divided by the level of kinase protein relative to wild-type PKR (footnote b). d Transformants of strain H1816 (⌬gcn2) containing different PKR alleles were analyzed for growth in synthetic medium (SGAL) containing 10% galactose and 2% raffinose. ϩϩϩ, growth within 24 h; ϩ, growth after 2 days; Ϫ, no growth after 4 days. precipitated PKR from yeast extracts and found that the addition of poly(I)⅐poly(C) increased less than 2-fold the phosphorylation of eIF-2␣ by PKR compared with similar reactions without ds-RNA. These results are in general agreement with Romano et al. (34), who found that the activity of PKR immunoprecipitated from yeast lysates was not further stimulated by the addition of poly(I)⅐poly(C).
When PKR-⌬1 or PKR-⌬2, deleted for motif 1 or motif 2 sequences, respectively, were assayed for phosphorylation of eIF-2␣, we found kinase activities slightly elevated over wildtype PKR (Fig. 3 and Table I). An even more dramatic demonstration that the DRBD sequences are not essential for kinase activity was found when we assayed the PKR-⌬14 -257 protein. This mutant kinase, which contains a deletion of both DRBD sequences, had 9 times the eIF-2␣ kinase activity of wild-type PKR. These results suggest that the amino-terminal sequences of the kinase perform a dual function. First, motifs 1 and 2 mediate ds-RNA activation of PKR kinase catalytic activity. As further clarified under "Discussion," the dispensability of the DRBD sequences for eIF-2␣ kinase activity would argue against motifs 1 and 2 simply functioning as positively acting regulatory sequences in the yeast system. A second function of motifs 1 and 2 would not involve enhancing kinase catalytic activity per se but would facilitate phosphorylation of eIF-2␣ in vivo. For example, the DRBD sequences might target PKR to a cellular location required for its access to the eIF-2␣ substrate.
PKR Is Associated with 40 S Ribosomal Subunits in Yeast-Results from our in vitro kinase assays suggested that the DRBD sequences carry out a localization role in the cell that is required for phosphorylation of eIF-2␣. We hypothesized that the DRBD sequences target PKR to ribosomes. Several lines of evidence support the proposal that this kinase associates with the translational machinery. First, biochemical fractionation studies (7,38) and immunofluorescent experiments (39,40) indicate that PKR is associated with ribosomes in mammalian cells. Second, the kinase substrate, eIF-2␣, is localized to ribosomes during initiation of protein synthesis (59). Finally, Bycroft et al. (4) has determined the NMR solution structure of a DRBD sequence from Drosophila staufen and found a ␣␤␤␤␣ structure with many features common to the N-terminal domain of prokaryotic ribosomal protein S5. Many of the consensus residues shared among members of the S5 protein family are conserved in DRBD sequences, suggesting that the two domains share a common evolutionary origin. We determined the localization of PKR in yeast strain H1817 (⌬gcn2 SUI2-S51A) using sucrose gradient sedimentation. Strain H1817 was selected for this analysis because, as noted above, the wild-type PKR levels were not severely decreased due to autoregulation. Cycloheximide was added to the culture medium 5 min before harvesting and to the lysis buffer to arrest translation elongation and preserve the polysomes during preparation and analysis of the sample. Fractions from the gradient were characterized by immunoblot to determine the distribution of the eIF-2␣ kinase. PKR was broadly dispersed in the sucrose gradient with over 70% of the protein co-migrating in the ribosomal fractions, primarily with 40 and 60 S subunits and 80 S particles and to a lesser extent with polysomal fractions (Fig.  4). The remaining portion of PKR was found in the top of the gradient, in fractions free of ribosomes.
PKR extracts were next analyzed in gradients lacking Mg 2ϩ , and as expected, we detected only free 40 and 60 S particles (59). As previously observed (59,60), the ribosomal subunits sedimented more slowly than measured in gradients containing Mg 2ϩ because certain ribosome-associated factors are released upon withdrawal of the cation. Under these conditions, almost all of the PKR migrated with the 40 S subunit (Fig. 4B). To be certain that PKR was not simply associated with mRNA that might still remain coupled with 40 S subunits, we treated the lysates with micrococcal nuclease following a regimen used for preparation of RNA-dependent in vitro translation lysates. PKR was associated with the 40 S subunits after sucrose gradient centrifugation of the nuclease-treated lysate (data not shown). We conclude that PKR is associated with ribosomes in the yeast system. DRBD Sequences Mediate PKR Association with Ribosomes-To assess whether DRBD sequences mediate ribosomal association of PKR, we analyzed PKR mutants containing inframe deletions in either motif 1 or motif 2 sequences. Both PKR-⌬1 and PKR-⌬2 were found exclusively at the top portion of the gradient, indicating that these mutant proteins were not associated with ribosomes (Fig. 5). By comparison, PKR-⌬K, containing a deletion of the entire kinase catalytic domain from FIG. 2. Immunoblot analysis of wild-type and mutant versions of PKR. Protein extracts were prepared from strain H1817 (⌬gcn2 SUI2-S51A) expressing the indicated PKR protein, and equal amounts of total protein were separated by SDS-polyacrylamide gel electrophoresis. After proteins were transferred to nitrocellulose filters, the samples were incubated with PKR monoclonal antibody 70 -10 (A) or rabbit polyclonal antiserum PKR K-17 (B) and washed, and antibody-protein complexes were visualized using horseradish peroxidase-labeled secondary antibody. Two different antibodies were used due to the different epitopes available in the mutant proteins. Levels of mutant proteins relative to wild-type PKR were quantitated and are listed in Table I residue 271 to 551, was located in the ribosomal portion of the gradient, with about 85% of PKR-⌬K migrating in the 80 S particle and polysome fractions (Fig. 5A). Wild-type PKR was predominately found in the fractions containing 40, 60, and 80 S particles (Fig. 4), suggesting that kinase activity may participate in the ribosomal distribution of PKR. Together, these results indicate that sequences within the motif 1 and 2 regions are required for ribosomal association of PKR.
Sequences deleted in the PKR-⌬1 and PKR-⌬2 mutant proteins are required to mediate ribosomal targeting. To address whether these sequences are coincident with the DRBD regions, rather than flanking in or between these sequences, we characterized two mutant versions of PKR that each contain a single residue substitution. Both PKR-A68P and PKR-A158P contain a proline substituted for an alanine residue that is well conserved among DRBD-containing proteins (1,2,4,19,34). These alanine residues are located in the carboxyl-terminal portion of the DRBD sequences in a region presumed to be ␣-helical based on the resolved structures of DRBD sequences from Drosophila staufen and RNase III from Escherichia coli (1,4,61). Green and Mathews (19) showed that both mutant versions of PKR bind ds-RNA in vitro at 5% or less of the levels measured for the wild-type DRBD sequences. Cells containing PKR-A68P did not display the growth defect found in isogenic strains expressing wild-type PKR, and the PKR-A158P strain showed a partial relief of the slow growth phenotype (Table I).
Cell lysates containing PKR-A68P were analyzed by sucrose gradient sedimentation, and the mutant kinase was found free of ribosomes (Fig. 6). PKR-A158P was located in two peaks. The majority of PKR-A158P was found at the top of the sucrose gradient, and a second portion of the kinase was located between the free peak and the 40 S particle (Fig. 6). Analysis of in vitro kinase phosphorylation of eIF-2␣ by PKR-A68P and PKR-A1589P mutants revealed activities slightly elevated compared with wild-type PKR (Fig. 3 and Table I). These results support the hypothesis that the two DRBD sequences in PKR directly mediate targeting to the ribosome. Furthermore, the kinase activity results are consistent with those previously noted for the PKR-⌬1 and PKR-⌬2 (i.e. motif 1 and motif 2 sequences are not essential for PKR catalytic activity).

DISCUSSION
Association of DRBD sequences with ds-RNA is proposed to alter the protein configuration of PKR, leading to induced kinase catalytic activity. Consistent with this model, motifs 1 and 2 have been shown to be essential for PKR function in the yeast FIG. 4. Distribution of wild-type PKR in sucrose gradient sedimentation. Cellular lysates were prepared from strain H1817 (⌬gcn2 SUI2-S51A) expressing wild-type PKR, and 20 A 260 units were loaded onto a 5-47% sucrose gradient. A, the cell extract was prepared in the presence of cycloheximide and 10 mM MgCl 2 . After ultracentrifugation, the sucrose gradients were fractionated using an absorbance monitor set at 254 nm, and free 40, 60, and 80 S ribosomes and polysomes are indicated. B, both cycloheximide and MgCl 2 were omitted from the sucrose gradient. As discussed under "Results," 40 and 60 S subunits sedimented more slowly than in the analysis in A. To illustrate this point, the relative positions of 40, 60, and 80 S particles in a gradient prepared in the presence of 10 mM MgCl 2 are indicated by the arrows in the A 254 profile shown in B. Levels of PKR in each gradient sample were measured by immunoblot analysis that is shown in the bottom of each panel. The percentage of the total PKR protein found in each fraction was measured by densitometry and plotted as a histogram that is superimposed on the A 254 profiles. Lane M in the immunoblot assay is the cellular lysate used in the sucrose gradient.  (Table I). However, mutant versions of PKR containing deletions of DRBD sequences showed in vitro kinase activities greater than that measured for wild-type PKR (Fig. 3 and Table I). Given the apparent contradiction between the in vivo and in vitro assays, we proposed that in addition to mediating ds-RNA activation of PKR, the DRBD sequences direct the kinase to a cellular location important for in vivo phosphorylation of eIF-2␣. Indeed, characterization of PKR by sucrose gradient sedimentation revealed that PKR is associated with 40 S ribosomal subunits, and in-frame deletions or residue substitutions in either of the motif 1 or 2 regions blocked PKR interaction with ribosomes (Figs. 4, 5, and 6). Our results indicate that the DRBD-defective mutants are catalytically active but have impaired in vivo function due to a failure to target the mutant kinase to the translational machinery. The role of the ribosomal association in the regulation of PKR may be to provide enhanced access to its substrate, eIF-2, which associates with 40 S subunits during the process of translation initiation. In the in vitro assay, recombinant eIF-2␣ substrate would be readily available to the kinase, and thus targeting would not be a prerequisite for phosphorylation (62).
Role of DRBD Sequences in the Regulation of PKR Kinase Catalytic Activity-Deletion of motif 1 or 2 sequences that bind ds-RNA, the presumed activator in the yeast system, reduced the growth defect associated with PKR. When cell lysates containing these mutant versions of PKR were assayed for the ability to phosphorylate eIF-2␣, we found that the mutant kinases displayed activities that were actually greater than those measured for wild-type PKR (Table I). A previous report (34) used the yeast model system and also found that the kinase activity of immunoprecipitated PKR mutants containing residue substitutions in the DRBD sequences was similar to wild-type PKR. It was suggested that the concentration of ds-RNA in the immunoprecipitated kinase preparations may have been high enough to compensate for the reduced ds-RNA in the DRBD sequences. Alternatively, it was speculated that binding of PKR to antibodies may have induced the kinase activity independently of activating ligand. In the studies presented in this report, PKR activity was measured directly in lysates, without the aid of antibodies; therefore, the elevated kinase activity associated with PKR-⌬1 and PKR-⌬2 could not be the result of antibody-directed activation. The possibility that elevated concentrations of ds-RNA may activate these mutant kinases also seems less likely given that the kinase activity was measured using cell lysates. Even more definitively, we found that PKR-⌬14 -257, deleted for both ds-RNA binding motifs, displayed 9 times the kinase activity measured for wild-type PKR (Table I).
What role do the DRBD sequences play in the regulation of PKR in the yeast model system? The dramatic increase in kinase activity measured for PKR-⌬14 -257 compared with wild-type kinase indicates that the amino-terminal sequences of PKR have a negatively acting effect on PKR. Binding of ds-RNA to both DRBD regions may release this inhibitory effect on kinase catalytic activity, resulting in increased PKR autophosphorylation and phosphorylation of eIF-2␣. This proposed negatively acting function of motif 1 and motif 2 could also be relieved by directly deleting either DRBD sequences. Consistent with this model, PKR-⌬1 or PKR-⌬2 exhibited eIF-2␣ kinase activity slightly higher than wild-type PKR, which is thought to be in the induced conformation (Table I).
The fact that kinase activity of mutant PKR-⌬14 -257 was 9 times greater than wild-type PKR suggests that the two DRBD sequences may function coordinately to repress kinase or that sequences between motif 2 and the protein kinase domain also contribute in conjunction with DRBD sequences to inhibit PKR.
Several reports using transfected cultured monkey kidney cells have also discussed the proposal that the DRBD sequences function as negatively acting regulators of PKR (29,63,64). Most recently, Wu and Kaufman (64), observed that expression of the kinase domain from residues 228 -551 was found to have elevated in vitro PKR activity. A model was proposed that the dsRNA binding domains inhibit PKR activity, and dsRNA binding induces a conformation change facilitating activation kinase function. A point of caution, as discussed in these cited reports, was the difficulty delinating the degree to which endogenous wild-type PKR contributed to the measured kinase activity. Concerning the role of sequences between motif 2 and the kinase domain in the regulation of PKR, Lee et al. (63) reported that these sequences, described as the third basic region, or motif 3, are critical for in vitro kinase activity as measured by autophosphorylation. We observed that the sequences extending to residue 257 were dispensable for in vitro kinase activity (Table I). In fact, PKR-⌬14 -257 phosphorylated eIF-2␣ to levels greatly exceeding wild-type PKR (Fig. 3, Table I). One explanation for this apparent contradiction could be that the PKR mutant deleted for the third basic region removed residues important for autophosphorylation and would erroneously indicate a loss of kinase activity (57). Another important difference between these in vitro kinase studies is that deletion of the third region described in Lee et al. (63) extended to residue 271, which would appear to remove sequences within subdomain I of the PKR kinase catalytic region (65). These kinase domain sequences are retained in PKR-⌬14 -257.
Targeting of PKR to Ribosomes Is Facilitated by DRBD Sequences-Regardless of the molecular rationale for why the PKR mutants depleted of DRBD sequences are catalytically active, it is striking that these mutant proteins do not phosphorylate eIF-2␣ in yeast (34) or cause the accompanying slow growth phenotype (Table I) (34). To explain this in vivo deficiency, we proposed that DRBD sequences are required to target PKR to the translation machinery, and this localization step is a prerequisite for phosphorylation of eIF-2␣. Consistent with this hypothesis, we found that about 70% of PKR migrated with ribosomal particles and polysomes fractionated by sucrose gradient sedimentation (Fig. 4). When Mg 2ϩ and cycloheximide were omitted from the experiment, resulting in dissociation of ribosomes into 40 and 60 S subunits, PKR was almost entirely associated with 40 S particles. These results indicate that PKR is a ribosome-associated protein with specificity toward the 40 S subunit.
It is interesting to note that in the presence of Mg 2ϩ and cycloheximide, PKR was heavily enriched in the 40, 60, and 80 S fractions, suggesting that the kinase may be preferentially associated with ribosomal subunits involved in the process of initiation (Fig. 4). When PKR-⌬K was similarly analyzed, about 85% of the mutant kinase was found in the 80 S and polysome fractions of the sucrose gradient, indicating that the kinase domain does not directly participate in ribosomal targeting (Fig. 5). However, the fact that PKR-⌬K mutant is enriched with polysomes, whereas wild-type kinase is primarily located in the 40, 60, and 80 S fractions, suggests that kinase activity may affect the association of PKR for certain ribosomal populations. Building on the model that wild-type PKR interacts primarily with initiating ribosomes, the large portion of PKR-⌬K complexed with polysomes may indicate that kinase activity facilitates the release of PKR from ribosomes that are beginning the elongation phase of translation. This proposal explains why almost no PKR-⌬K protein was found in the free peak and an increased portion of kinase mutant complexed with elongating ribosomes.
What is the nature of the interaction between PKR and ribosomes? Both DRBD sequences are required for kinase association with the translational machinery. This conclusion was most specifically supported by the observation that PKR-A68P and PKR-A158P were blocked in their association with ribosomes (Fig. 6). Both mutants have been reported to be severely reduced in their ability to bind ds-RNA (19). These results suggest that motif 1 and motif 2 interaction with rRNA is central to PKR association with ribosomes. The fact that PKR was bound almost exclusively with 40 S subunits indicates a specificity for rRNA sequences, although protein-protein interactions may also participate in ribosomal targeting.
DRBD Sequences May Carry Out Multiple Functions in the Regulation of PKR Activity-The DRBD sequences appear to perform a dual function important for regulating PKR. To control kinase catalytic activity, motifs 1 and 2 would bind activating ligand, triggering autophosphorylation of the kinase. As discussed above, stimulation of kinase activity may involve ligand-induced relief of the negatively acting DRBD regions. Phosphorylation of PKR is thought to lock the kinase into a catalytically induced conformation that no longer requires ds-RNA binding for eIF-2␣ kinase activity (6,7,15). The high level of PKR activity would be retained until the kinase is dephosphorylated by protein phosphatases (66,67). Motif 1 and motif 2 sequences would now be available to target activated PKR to ribosomes, specifically to 40 S subunits. Localization of the kinase to ribosomes may be essential for PKR to come into contact with its eIF-2 substrate. The eIF-2 couples with 40 S ribosomal subunits during translation initiation. Ramirez et al. (59) have utilized sucrose gradient methods similar to those described in this report to show directly that yeast eIF-2␣ migrates in the ribosome fractions. PKR recognition and phosphorylation of eIF-2 would reduce the exchange of GDP-GTP exchange catalyzed by eukaryotic initiation factor-2B, leading to a reduction in the rate of protein synthesis.
As discussed in the Introduction, expression of certain kinase-inactive PKR mutants in NIH3T3 cells leads to a malignant transformation phenotype (42)(43)(44). Two different molecular models have been proposed to explain this trans-dominant inhibition of PKR (8,24,28,29,34,42). One model suggests that overexpressed mutant PKR binds all of the available ds-RNA in cells and, thus, prevents ligand activation of the endogenous wild-type kinase. The second model proposes that the kinase-defective mutant forms a heterodimer with wild-type PKR and impedes intermolecular autophosphorylation required for activation of the kinase. Both mechanisms have drawn experimental support. The basis for the apparent contradictions between these studies may be that both mechanisms can contribute to trans-dominant inhibition depending on the level of activating ligand found in the particular cell system.
The important role of ribosome targeting in in vivo phosphorylation of eIF-2␣ supports the idea that overexpressed mutant kinase could also reduce the function of endogenous wild-type PKR by competing for ribosomal binding sites. As noted earlier, the PKR-⌬K mutant binds efficiently to the ribosomes and may alter the release of the kinase from the translation machinery. By preventing the endogenous wild-type PKR from targeting to ribosomes, the mutant kinase would block phosphorylation of eIF-2␣. Experiments are ongoing to determine whether coexpression of mutant forms of PKR in yeast alters ribosome association of wild-type PKR.
In closing, the DRBD sequences of PKR appear to be important for proper localization of this kinase to the ribosomes. Given the large number of different proteins containing DRBD sequences, it is interesting to speculate that many of these sequences carry out a similar ribosomal targeting function. Another likely candidate is the yeast protein encoded by YML3. The YML3 protein contains a single DRBD sequence in its carboxyl terminus, from amino acid residue 315 to 375, and has been shown to be associated with the large ribosomal subunit in mitochondria (68). While initial reports suggested that YML3 encodes the L3 protein of the mitochondrial ribosome (68), more recent work indicates that this is not the case (69). The physiological role of the YML3 protein and the possible contribution of DRBD-mediated ribosomal targeting to this function remain to be determined.