The kinase insert domain of interferon-induced protein kinase PKR is required for activity but not for interaction with the pseudosubstrate K3L.

Interferon-induced protein kinase (PKR) is a member of a family of kinases that regulate translation initiation through phosphorylation of eukaryotic initiation factor 2α. In addition to the conserved catalytic subdomains that are present in all serine/threonine kinases, the eukaryotic initiation factor 2α kinases possess an insert region between catalytic subdomains IV and V that has been termed the kinase insert domain. To investigate the importance of the kinase insert domain of PKR, several deletions and point mutations were introduced within this domain and analyzed for kinase activity both in vitro and in vivo. Here we show that deletion of the kinase insert sequence or mutation of serine 355, which lies within this region, abrogates kinase activity. In addition, the kinase insert domain of PKR and adjacent amino acids (LFIQME) in catalytic subdomain V are not required for binding of the pseudosubstrate inhibitor K3L from vaccinia virus. A portion of the catalytic domain of PKR between amino acids 366 and 415 confers K3L binding in vivo, suggesting a possible role for this region of PKR in substrate interaction.


Interferon-induced protein kinase (PKR) is a member of a family of kinases that regulate translation initiation through phosphorylation of eukaryotic initiation factor
2␣. In addition to the conserved catalytic subdomains that are present in all serine/threonine kinases, the eukaryotic initiation factor 2␣ kinases possess an insert region between catalytic subdomains IV and V that has been termed the kinase insert domain. To investigate the importance of the kinase insert domain of PKR, several deletions and point mutations were introduced within this domain and analyzed for kinase activity both in vitro and in vivo. Here we show that deletion of the kinase insert sequence or mutation of serine 355, which lies within this region, abrogates kinase activity. In addition, the kinase insert domain of PKR and adjacent amino acids (LFIQME) in catalytic subdomain V are not required for binding of the pseudosubstrate inhibitor K3L from vaccinia virus. A portion of the catalytic domain of PKR between amino acids 366 and 415 confers K3L binding in vivo, suggesting a possible role for this region of PKR in substrate interaction.
The double-stranded RNA (dsRNA) 1 -activated protein kinase, PKR (also called p68, DAI, and P1 kinase) is an interferon-inducible protein that plays a key role in the antiviral and antiproliferative interferon response (for a review see Ref. 1). PKR is a serine/threonine kinase that undergoes autophosphorylation upon binding to dsRNA activators, such as those generated upon viral infection or stem-loop structures in certain RNAs (e.g.. Refs. 2 and 3). Upon autophosphorylation, PKR phosphorylates the ␣ subunit of eukaryotic initiation factor 2 (eIF2) (2,4). eIF2 is composed of three subunits (␣, ␤, and ␥) and forms a ternary complex (eIF2⅐Met-tRNA i ⅐GTP) that is required for binding of Met-tRNA i to the 40 S ribosomal subunit. Prior to the association of the 60 S ribosomal subunit, GTP hydrolysis results in formation of an inactive eIF2⅐GDP complex. Upon phosphorylation of eIF2␣, the exchange of GDP for GTP is blocked due to inactivation of the guanine nucleotide exchange factor, eIF2B, resulting in the shut off of protein synthesis (5).
PKR plays a critical role in the establishment of the interferon-induced antiviral state (6). Many viruses have evolved strategies to counteract the inhibition of translation mediated by PKR (for review see Ref. 7). For instance, vaccinia virus encodes a protein called K3L, which shares extensive homology with the N terminus of eIF2␣ (8) and inhibits phosphorylation of eIF2␣ by PKR in vitro (9) and in vivo (10,11). Taken together with the finding that K3L binds directly to PKR (9,11,12), these data suggest that K3L acts as a pseudosubstrate inhibitor of PKR.
In addition to its antiviral role, PKR plays an important role in the control of cell growth, inasmuch as expression of inactive mutants of PKR (PKR⌬6 (13), PKR-K296R (14), and PKR-M6, M7 (15)) causes the malignant transformation of NIH 3T3 cells, whereas expression of wild-type PKR in yeast causes a slow growth phenotype (16). Also, overexpression of a PKR inhibitor (p58) transforms NIH 3T3 cells (17). PKR mutants cause the malignant transformation of cells possibly through formation of inactive heterodimers with endogenous PKR (18), thus diminishing the levels of eIF2␣ phosphorylation. In support of this hypothesis, a recent study has shown that expression of the eIF2␣ subunit, in which the phosphorylation site residue (serine 51) was mutated to alanine, also resulted in malignant transformation of NIH 3T3 cells (19).
Several eIF2␣ kinases have been cloned: PKR (20, 21), a heme-sensitive kinase (named heme-regulated inhibitor) ( 22,23), and a yeast kinase GCN2 (24). Although there is sequence similarity among the catalytic domains of the eIF2␣ kinases and other serine/threonine kinases, a unique feature that distinguishes eIF2␣ kinases from other kinases is the large spacer (28 amino acids) between kinase subdomains IV and V, called the kinase insert domain (21).
In this study we examined the effects of mutations in the kinase insert domain on PKR activity in vitro and in vivo. Also, we tested mutants of PKR for interaction with the pseudosubstrate K3L in vitro and in vivo. Our findings indicate that the kinase insert domain is dispensible for K3L interaction, but it is required for kinase activity, suggesting that the kinase insert domain is not involved in binding of K3L. Mapping studies of K3L binding to PKR implicate a predicted ␣-helical sequence between amino acids 366 and 415 of the catalytic domain to be sufficient for the binding of K3L to PKR.

MATERIALS AND METHODS
PKR Mutagenesis and Plasmid Construction-All mutagenesis was carried out by a two-step polymerase chain reaction approach (25). The polymerase chain reaction products were digested with MscI and AflII and subcloned into a similarly digested KS-PKR clone, which contained the entire cDNA from pcDNA1neo-p68 (26). PKR⌬KI and PKR⌬330 -336 mutants were generated to delete amino acid codons 330 -355 and 330 -336, respectively. Wild type PKR and mutants were subcloned into the mammalian expression vector pRc/ CMV (Invitrogen) at unique HindIII/NotI sites. PKR⌬6 was subcloned into the yeast two-hybrid system vectors pGBT9 and pGAD.GH (gifts from P. Bartel, SUNY Stony Brook, and G. Hannon, Cold Spring Harbor, NY) as described previously (12). An EcoRI/BglII fragment encoding amino acids 366 -415 of PKR was subcloned into pGBT9 digested with EcoRI and BamHI. The PKR constructs were also subcloned into the galactose-inducible yeast expression vector pDAD1 (27) by digestion of KS-PKR plasmids with HindIII/SmaI and ligation to pDAD1 digested with HindIII/PvuII.
For expression of an internal portion of PKR, the pET-HMK vector (a gift from M. Blanar, University of California, San Francisco, CA) was used to provide the initiation codon and 15 additional amino acids (31). The FlagPKR281-415 mutant was generated by digestion of KS-PKR with MscI/AflII and pET-HMK with EcoRI, followed by blunt ending and ligation. This generates a fusion protein containing the "Flag" peptide sequence (31) followed by amino acids 281-415 of PKR. The stop codon is generated from vector sequences. A glutathione S-transferase-K3L fusion protein was generated by ligation of a blunt NdeI-BamHI fragment containing vaccinia virus K3L (9) with pGEX-HMK (a gift from M. Blanar, UCSF), which was digested with EcoRI and blunted.
Transient Transfections-COS-1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Life Technologies, Inc.) and were transfected with 15 g of pRc/CMV-PKR plasmid in 100-mm culture dishes using the DEAE-dextran method (28). For the luciferase assay and preparation of RNA for "RNase protection," 1.2 g of Rous sarcoma virus-luciferase DNA (50) was co-transfected with 12 g of pRc/CMV or pRc/CMV-PKR DNA in 100-mm plates. COS-1 cells were harvested 48 h after transfection.
Western Blot Analysis-S10 extracts were prepared as described previously (13). Total cell protein (75 g) was subjected to SDS-10% PAGE followed by transfer to 0.45-m nitrocellulose membranes. PKR wild type and mutant proteins were detected using a mouse monoclonal antibody against PKR (13B8-F9; a kind gift from G. Barber In Vitro Kinase Assay-The activity of the PKR proteins expressed in COS-1 cells was tested by immunoprecipitation with monoclonal antibody 71/10 (30), followed by a kinase assay in the presence of purified eIF2 (a generous gift of W. Merrick) and 1 g/ml poly(I)⅐poly(C), as described previously (26).
In Vitro Translation and Co-precipitations-PKR plasmids were transcribed in vitro, and mRNAs were translated in rabbit reticulocyte lysate in the presence of [ 35 S]methionine (ICN) as suggested by the manufacturer (Promega). PKR1-415 and PKR1-280 were generated by linearization of pRc/CMV-PKRwt DNA with AflII and MscI, respectively.
Glutathione S-transferase (GST) and the GST-K3L fusion protein were expressed in Escherichia coli and purified using glutathione-Sepharose beads (Pharmacia Biotech, Inc.). Translation products were incubated with GST or GST-K3L proteins (1 g) for 60 min at 4°C, and Glutathione-Sepharose beads were added for an additional 30 min. Following extensive washing with co-immunoprecipitation buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% Nonidet P-40), Laemmli buffer was added, and samples were subjected to SDS-10% PAGE. 35 S-Labeled translation products were detected by autoradiography. The FlagPKR281-415 mutant was expressed in BL21(DE3) bacteria and subjected to co-precipitation with GST and GST-K3L proteins as described above. The precipitated products were subjected to SDS-PAGE, transferred to nitrocellulose membrane, and Western blotted using an anti-Flag monoclonal antibody (M2; Eastman Kodak Co.). Proteins were detected by ECL (Amersham Corp.).
Luciferase Assay-Luciferase assays were performed according to the manufacturer's instructions (Promega) and analyzed using a BIOORBIT bioluminometer. To determine levels of luciferase mRNA, total RNA was prepared from duplicate transfections and harvested by the guanidinium/CsCl method as described (32). RNA was digested with RNase-free DNase I (Boehringer Mannheim) and analyzed using the RNase-One protection assay (as described in Ref. 19), according to the recommendations of the manufacturer (Promega).
Yeast Transformations-The yeast strain S150 -2B (leu2-3,112, his3⌬1, trp1-289, ura3-52) was transformed with pDAD1 constructs using the lithium acetate method (33) and plated on SD-agar plates supplemented with leucine, histidine, and tryptophan. Single colonies were picked and grown overnight in liquid SD medium containing the same amino acids. 20 l of these cultures were streaked onto parallel SD-agar media containing either glucose or galactose as the source of carbon. Slow growth phenotype was monitored after 3 days.
For the two-hybrid assay, the yeast strain Y526 was transformed with various combinations of pGBT9 and pGAD.GH constructs using the lithium acetate method (33) as described (12).

Mutagenesis of the Kinase Insert Domain of PKR-
To understand the function of the kinase insert domain of PKR with respect to catalytic activity and possible substrate interactions, we designed a series of mutants (Fig. 1). The mutations included a large and small deletion (PKR⌬KI and PKR⌬330 -336, respectively) and point mutations in three residues (PKR-Y323F, PKR-S337A, and PKR-S355A) in the kinase insert domain, which are conserved in all eIF2␣ kinases (Fig. 1). The PKR⌬KI mutant contains a deletion of amino acids 330 -355 that removes the entire kinase insert domain. The PKR⌬330 -336 mutant contains a deletion of a seven-amino acid sequence that is duplicated in the human PKR kinase insert sequence (21), whereas there is only one copy in the murine PKR sequence ( Fig. 1). In addition, we used the dominant negative mutant, PKR⌬6 (13), which harbors a deletion of amino acids 361-366. This sequence was proposed as a possible substrate interaction site because of the high sequence conservation among the eIF2␣ kinases (17). Also shown in Fig. 1 is the enzymatically inactive PKR-K296R mutant (14,26), which harbors a mutation of the conserved lysine in catalytic subdomain II that is involved in the phosphotransfer reaction.
Expression and Characterization of PKR Mutants-To characterize the kinase insert domain mutants, COS-1 cells were transiently transfected with wild type and mutant PKR constructs, and PKR activity was assayed in vitro. Western blot analysis was performed on extracts from transfected cells with a monoclonal antibody that is specific for human PKR (13B8-F9; Ref. 29). In agreement with earlier studies (34,35), wild type PKR was expressed to a much lower level than the catalytically inactive PKR-K296R mutant ( Fig. 2A, compare lanes 2  and 3). The amount of PKR⌬6 protein was higher than wild type PKR, but not as high as the PKR-K296R mutant (compare lane 4 with lanes 2 and 3). The relatively poor expression of wild type PKR is likely due to the activation of PKR by its own mRNA, resulting in localized translational inhibition (35,36). The amounts of PKR⌬KI and PKR⌬6 proteins were similar (lanes 4 and 5), whereas the PKR⌬330 -336 mutant was present at 10-fold higher amounts than wild type PKR but lower than the other deletion mutants (lane 6). Of the three point mutants, PKR-Y323F and PKR-S355A were more highly expressed than the PKR-S337A mutant (lanes 7-9).
To assess the kinase activity of the PKR mutants, an in vitro kinase assay was performed in which immunoprecipitated PKR was assayed for autophosphorylation and phosphorylation of exogenously added eIF2. Using extracts of interferon-treated HeLa cells as a positive control, strong autophosphorylation (position marked by an open arrow to the left and by dots) and eIF2␣ phosphorylation (marked by a closed arrow to the left and by dots) were detected upon addition of dsRNA (Fig. 2B,  lanes 1 and 2). There is a background band at 35 kDa that migrates slightly faster than eIF2␣. It is present in all lanes, and it is not affected by dsRNA addition. It has been previously shown that the best indicator of PKR kinase activity is its ability to phosphorylate eIF2, since the level of autophosphorylation does not always correlate with that of kinase activity (37,38). Cells transfected with the empty vector showed no eIF2␣ phosphorylation (lanes 3 and 4). Extracts of wild type PKR-transfected COS cells displayed enhanced eIF2␣ phosphorylation in the presence of dsRNA (lanes 5 and 6). The autophosphorylation signal for wild type PKR was very weak due to low levels of protein expression ( Fig. 2A), and the lack of induction by dsRNA has been reported previously to be due to the phosphorylation state of PKR prior to harvesting of the cells (35). The substrate phosphorylation signal is stronger than PKR autophosphorylation, since a molar excess of eIF2 substrate is added, compared with PKR. Extracts from cells transfected with PKR-K296R (lanes 7 and 8) and PKR⌬6 (lanes 9 and 10) showed no kinase activity. The deletion of the entire kinase insert domain (PKR⌬KI) abolished both autophosphorylation and eIF2␣ phosphorylation (lanes 11 and 12), suggesting that this domain is important for kinase activity. PKR-Y323F and PKR-S337A mutants phosphorylated eIF2␣ almost as well as wild-type PKR (lanes [13][14][15][16]. Since the amount of PKR-Y323F protein was 4-fold greater than that of the PKR-S337A mutant ( Fig. 2A, lanes 7 and 8) and the amount of eIF2␣ phosphorylation was similar, this suggests that the kinase activity of the Y323F mutant is 25% of the S337A mutant. The slightly faster mobility of the autophosphorylated Y323F mutant (compare lane 14 to lanes 2 and 15) may reflect a reduced phosphorylation state compared with that of wild type PKR and the S337A mutant. The PKR-S355A mutant showed no detectable autophosphorylation or eIF2␣ phosphorylation upon addition of dsRNA activator (lanes 17 and 18), suggesting that this residue is critical for kinase activity. The loss of kinase activity observed for PKR⌬KI and PKR-S355A mutants was not due to diminished amounts of protein, since both mutant proteins were detected by immunoprecipitation followed by Western blotting (data not shown).
The kinase activity of the PKR⌬330 -336 mutant was assessed in a similar assay (Fig. 2C). In this experiment, extracts from IFN-induced HeLa cells showed strong autophosphorylation (marked by an open arrow) and eIF2␣ phosphorylation (marked by a closed arrow; Fig. 2C, lanes 1 and 2). No catalytic activity was obtained in COS cells transfected with the vector alone (lanes 3 and 4). The addition of dsRNA to immunoprecipitates of wild type PKR caused extensive eIF2␣ phosphorylation (lanes 5 and 6). Only slight eIF2␣ phosphorylation was detected for the PKR⌬330 -336 mutant (lanes 7 and 8). The amount of PKR⌬330 -336 protein was 10-fold higher than wild type PKR, yet eIF2␣ phosphorylation was only 10% of that for wild type PKR (compare lanes 6 and 8), indicating that this mutant retains only approximately 1% of the catalytic activity of wild type PKR. This suggests that this seven-amino acid sequence (amino acids 330 -336), although duplicated in human PKR (Fig. 1), is very important for kinase activity.
Effects of PKR Mutants on Yeast Growth-The expression of wild type PKR in yeast causes growth suppression, which results from elevated eIF2␣ phosphorylation (17,18,39). To test for the effect of mutations in PKR on yeast growth, PKR mutants were expressed in yeast under a galactose-inducible promoter. As expected, all yeast transformants grew when plated on glucose-containing media (Fig. 3); the slower growth phenotype of the PKR-S337A transformant reflects a poor streak and is not observed in all experiments. Upon transfer to galactosecontaining media, wild type PKR-expressing yeast showed the characteristic slow growth phenotype (17), while yeast transformed with the empty vector or the PKR-K296R mutant displayed normal growth. In accordance with the in vitro results, expression of PKR⌬KI and PKR-S355A did not affect yeast growth, consistent with a lack of kinase activity. The PKR⌬330 -336 mutant did not cause inhibition of yeast growth, suggesting that the very low level of kinase activity retained by this mutant was not sufficient for growth inhibition. Expres- sion of PKR-Y323F and PKR-S337A mutants caused growth suppression similar to that observed for wild type PKR (Fig. 3). The lack of an effect on yeast growth for the inactive PKR mutants is not due to lack of expression, since the proteins were detected by Western blotting (data not shown). The in vivo results, together with the in vitro kinase assay data (Fig.  2B), indicate that deletion of the entire kinase insert domain or deletion of a seven-amino acid motif within this domain (PKR⌬330 -336) results in inactivation of PKR. Mutation of serine 355 to alanine also caused a loss of kinase activity, raising the possibility that phosphorylation of this residue might be important for activation of PKR.
Inactive PKR Mutants Stimulate Translation in Vivo-To assess the effects of the PKR mutants on translation, the luciferase reporter gene was co-transfected with the PKR constructs into COS-1 cells. Luciferase activities were corrected for the amount of luciferase mRNA as measured by RNase protection (data not shown; less than 30% variability between PKR transfectants). Translation of luciferase mRNA was stimulated (between 2-and 6-fold) by the catalytically inactive PKR-K296R, PKR⌬6, and PKR⌬KI mutants (Fig. 4). The PKR-K296R mutant had the largest effect (about 6-fold stimulation), which could be explained by the better expression of this mutant ( Fig. 2A, lane 3). The kinase insert deletion mutant and the PKR⌬6 mutant in three different experiments caused a 3-fold stimulation of translation of the reporter mRNA (Fig. 4). The enhanced translation could be explained by dominant negative inhibition, or by sequestration of dsRNA activators from the endogenous kinase, which becomes activated upon transfection with plasmid DNA (28). A study of PKR function in yeast suggests that inactive PKR mutants generated by deletions are dominant alleles, while those containing point mutations are recessive (18). However, in this assay the largest effects were observed for the PKR-K296R and not the deletion mutants (PKR⌬6, PKR⌬KI). This could be due to the much higher expression levels of the PKR-K296R protein, compared with PKR⌬6 and PKR⌬KI proteins ( Fig. 2A). Wild type PKRexpressing cells did not show a significant reduction in luciferase mRNA translation. This was also observed in previous studies using another reporter gene (35). This likely reflects the low level expression of wild type PKR in this expression system (see Fig. 2A).
Effects of Mutations in the PKR Kinase Domain on Binding of the Pseudosubstrate K3L-The substrate binding site of PKR has not been characterized, although some regions have been proposed, including the kinase insert domain (40,41). We attempted to determine whether the kinase insert domain is the binding site of eIF2 by co-immunoprecipitation of in vitro translated PKR products with purified eIF2. However, an interaction between PKR and eIF2 was not detected (data not shown). This is consistent with previous studies in which the binding of eIF2 to PKR was suggested to be much weaker than the binding of PKR to the vaccinia virus K3L protein (11). The K3L protein has extensive homology to the N-terminal portion of eIF2␣ (8) and is thought to act as a pseudosubstrate inhibitor of PKR, since expression of K3L can inhibit phosphorylation of eIF2␣ by PKR in vitro (9) and in vivo (10,11).
Since the K3L protein is a pseudosubstrate inhibitor of PKR, delineation of the K3L binding site may be helpful in identifying the substrate binding site within PKR. In particular, we wished to determine if the kinase insert domain constitutes the K3L binding site on PKR. PKR proteins were synthesized in vitro and incubated in the presence of either recombinant GST or a recombinant GST-K3L fusion protein. Glutathione-Sepharose beads were added to precipitate GST and GST-K3L proteins, and bound 35 S-labeled PKR proteins were resolved by SDS-PAGE. Wild type PKR was expressed at lower levels than PKR⌬6 and PKR⌬KI proteins (Fig. 5A, compare lane 1 with  lanes 4 and 7; full-length product indicated by dot). As reported in earlier studies, a number of 35 S-labeled smaller molecular weight products were detected (21,26). The three PKR products were all co-precipitated with GST-K3L with similar efficiencies but not by GST alone (compare lanes 3, 6, and 9 with  lanes 2, 5, and 8). Three to five percent of the 35 S-labeled PKR proteins were recovered with the GST-K3L precipitates. The low levels of recovery could be due to partial masking of K3L by the larger GST moiety. Also, endogenous PKR and heme-regulated inhibitor likely compete with the PKR products for K3L binding (9). The results indicate that whereas the kinase insert region is required for kinase activity (Figs. 2 and 3) it is not required for K3L binding. Also, it is noteworthy that the deletion of amino acids 361-366 (PKR⌬6 mutant; Ref. 13), which were proposed to constitute part of the substrate binding site (17), has no effect on the binding of the pseudosubstrate K3L

in vitro.
To delineate the K3L binding site, C-terminal deletion mutants of PKR were synthesized in vitro and tested for interaction with K3L. The PKR1-415 mutant migrated at its predicted molecular mass (59 kDa, indicated by a dot). However, there were trace amounts of full-length PKR resulting from incomplete digestion of the cDNA (Fig. 5B, lane 1). The PKR1-415 mutant co-precipitated with GST-K3L with efficiency com-parable with that of wild type PKR and did not bind the GST control (lanes 1-3). The PKR1-280 migrated according to its expected size (40 kDa, lane 4) and failed to bind K3L (lane 6). These results suggest that there are critical residues between amino acids 281 and 415 of PKR that are involved in K3L binding. However, these results do not rule out the possibility of a conformational change in another portion of the protein, resulting in a loss of K3L binding. To address this issue, we generated an N-terminal deletion mutant of PKR, which contains a Flag motif (encoding 15 amino acids; Ref. 31) at its N terminus fused to a portion of the PKR kinase domain (FlagPKR281-415). The FlagPKR281-415 mutant contains a portion of PKR spanning catalytic subdomains I-VI (amino acids 281-415; Fig. 5D). It was expressed in E. coli and subjected to the K3L binding assay described above (Fig. 5C). The FlagPKR281-415 mutant that encodes a 20-kDa protein was detected using an anti-Flag monoclonal antibody (Fig. 5C, lane  1). This protein was precipitated by K3L (lane 3; 5% of input) but not by the GST control (lane 2). Taken together with the results for the C-terminal deletion mutants (Fig. 5B), these data suggest that the sequence between amino acids 281 and 415 contains critical binding sites for the pseudosubstrate K3L and that the kinase insert domain and LFIQME sequence are dispensible for K3L binding (Fig. 5D).
Binding of K3L to PKR in Vivo-To substantiate our results from the in vitro system, the yeast two-hybrid system was used to assay substrate interactions with PKR in vivo (42). In this assay, the proteins of interest are expressed as GAL4 DNAbinding or GAL4 transactivation domain fusion proteins, and their interaction results in the expression of ␤-galactosidase, which is monitored by blue color formation in media containing 5-bromo-4-chloro-3-indoyl ␤-D-galactoside. In agreement with previous results (12), we were able to detect binding of the PKR mutants to K3L in this assay (Table I, bottom). As positive controls, we utilized two yeast proteins, SNF1 and SNF4 (42), which gave a strong signal (Table I,  PKR constructs were co-transfected with Rous sarcoma virus-luciferase (50), luciferase activity was measured using a luminometer, and luciferase mRNA was measured by RNase-One (Promega) protection assay (data not shown). Units of luciferase translation are given as luciferase activity (mV/g of protein) divided by luciferase mRNA levels. The mean of three experiments is shown with the standard deviation. the SNF4 fusion was tested against PKR⌬6, PKR366 -415, and K3L fused to the GAL4 DNA binding domain. No signal was obtained for these combinations or for the SNF1 fusion tested against PKR⌬6 and K3L fused to the GAL4 transactivation domain. The PKR⌬6 mutant showed an interaction with p20, which corresponds to the N-terminal dimerization domain of PKR (12) and with the pseudosubstrate K3L (Table I, bottom). This shows that the PKR⌬6 mutant retains the ability to bind the pseudosubstrate K3L in vivo. This result further indicates that the LFIQME sequence is not a substrate binding site as was originally proposed (17). In addition, we tested the fragment containing amino acids 366 -415 of PKR for its binding to K3L. This combination showed a positive signal, suggesting that the K3L binding site identified in vitro (amino acids 281-415; Fig. 5) can be further demarcated to between amino acids 366 and 415 of PKR (Fig. 6, dashed line). DISCUSSION In this study we report that the kinase insert domain of human PKR is required for kinase activity but does not play a role in binding of PKR to the pseudosubstrate K3L protein from vaccinia virus. We also observed that mutation of a conserved serine residue in the kinase insert domain (serine 355) abolished kinase activity, raising the possibility that phosphorylation of this residue may be critical for activation of PKR. The context of this serine residue is a PKC consensus site (43), but it is not conserved in the murine PKR sequence (Fig. 1), and there is no evidence for phosphorylation of PKR by another kinase.
Since K3L strongly inhibits the phosphorylation of eIF2␣ by PKR through direct interaction with PKR (9), and because of the ability of eIF2␣ to compete out the interaction between PKR and K3L (11), K3L is thought to be a pseudosubstrate inhibitor of PKR. Also, the residues in K3L showing greatest homology to eIF2␣ are essential for inhibition of substrate phosphorylation by PKR. 2 This suggests that K3L inhibits eIF2␣ phosphorylation by PKR through binding and blocking at least a portion of the PKR substrate binding site. Therefore, identification of the K3L binding site within PKR may provide important information as to the location of the substrate binding site within PKR. We showed that residues 361-366, which are deleted in the PKR⌬6 mutant (13), are not required for K3L binding in vitro or in vivo (Fig. 5A, Table I). These findings demonstrate that the PKR⌬6 mutant still contains the pseudosubstrate binding domain. In fact, upon alignment of the catalytic domains of PKR and the cell cycle regulatory kinase Cdc2 (44), the LFIQME sequence, which is deleted in the PKR⌬6 mutant, aligns with the highly homologous sequence LYLIFE, which comprises ␤-sheet 5 in subdomain V of Cdc2 (Fig. 6). Therefore, the loss of kinase activity in the PKR⌬6 mutant (13) results from the deletion of a conserved structural motif (␤-sheet 5) and not from loss of a substrate binding site.
We have determined that the kinase insert domain is not required for K3L binding (Fig. 5A). We also defined a portion of 2 T. Dever, personal communication.
FIG. 5. Co-precipitation of PKR mutants and K3L. A, wild type PKR, PKR⌬6, and PKR⌬KI were translated in vitro and incubated with either purified GST or GST-K3L before the addition of glutathione-Sepharose beads. Following extensive washing, the bound material was removed by boiling in Laemmli buffer and subjected to SDS-10% PAGE. One-fifth of the amount used for co-precipitation was loaded directly onto the gel (indicated as Input). B, C-terminal deletion mutants (PKR1-415 and PKR1-280) were subjected to the same assay described above. C, a fusion protein encoding amino acids 281-415 of PKR (FlagPKR281-415) was expressed in E. coli and co-precipitated with GST or GST-K3L and glutathione-Sepharose beads. Following SDS-PAGE, proteins were transferred to a nitrocellulose membrane and detected by Western blotting using an anti-Flag monoclonal antibody (M2, Kodak). D, summary of PKR/K3L interaction results. For description of PKR domain structure, see the legend to Fig. 1.

TABLE I Yeast two-hybrid assay for interaction of PKR mutants with K3L
Indicated combinations of GAL4-DNA binding domain (GAL4-DB) or GAL4 transactivation domain (GAL4-TA) were used to transform yeast strain Y526 and assessed for the ability to activate lacZ transcription as described (42). SNF1 and SNF4 were transformed together as a positive control and with the PKR mutants as negative controls.
PKR between amino acids 281 and 415 that confers K3L binding ability in vitro (Fig. 5C). The binding site was further delineated using the two-hybrid system, in which amino acids 366 -415 of PKR were sufficient for interaction with K3L (Table I). In another recent study, it was found that K3L binding is conferred by amino acids 367-551 (45), which is in agreement with our results.
The structural motifs present within the pseudosubstrate binding domain can be predicted from the deduced threedimensional structures of other serine/threonine kinases. The kinase domain possesses a small N-terminal lobe rich in ␤-sheets and a C-terminal large lobe that is predominantly ␣-helical. The cleft between the small and large lobes contains the catalytic site (46,47). Amino acids 367-415 of PKR, containing the minimal K3L binding site, form an ␣-helical structure, which links the small and large lobes of the kinase (between subdomains V and VI). This region corresponds to the ␣-helices present between ␤-sheets 5 and 6 and is not highly conserved between kinases (46). Studies with protein kinase A and its inhibitor, protein kinase inhibitor (48), have shown that protein kinase inhibitor contacts protein kinase A within the ␣-helical region that corresponds to the K3L binding site to PKR. Alignment of this sequence of the different eIF2␣ kinase sequences does not yield a clear sequence motif that is exclusive to this kinase family. 3 Further studies will be necessary to determine the nature of the K3L-PKR interaction interface.
In summary, we have identified the binding site within PKR of the pseudosubstrate K3L. It is possible that in addition to the K3L binding site (Fig. 6) other portions of PKR, including the kinase insert domain, may interact with the larger eIF2 complex. Further characterization of the substrate binding site of PKR is of importance to the understanding of the role of substrate interactions in growth regulation by PKR. FIG. 6. An alignment of a portion of the catalytic domains of human PKR and CDC2 kinase. A portion of the catalytic domains of PKR (21) and CDC2 (44) were aligned using the program Gene Jockey. Amino acid identity is indicated by dots, whereas conserved substitutions are marked by vertical dashed lines. The positions of the catalytic subdomains (49), the kinase insert domain, and the PKR⌬6 mutation are underlined. The amino acid positions corresponding to the PKR deletion mutants are shown by closed arrows. The dashed line indicates the portion of PKR that is sufficient for K3L binding.