INTRODUCTION

PKR¹ is a cAMP-independent serine-threonine kinase that is induced by interferon treatment (1, 2). On activation by double stranded RNA, PKR undergoes autophosphorylation and catalyzes the phosphorylation of the subunit of eukaryotic protein synthesis initiation factor 2 (eIF-2), resulting in inactivation of the latter (3, 4). These events lead to dramatic decreases in protein synthetic rates inside the cell. This situation is not favorable to viruses, which must make their proteins to replicate. Much of the focus on PKR has related to the role of the kinase in mediating the inhibitory interferon response to viral infection (5, 6, 7, 8). However, there is accumulating evidence implicating PKR, which is constitutively expressed in eukaryotic cells in the absence of interferon (9), in the regulation of normal cellular processes. These include signal transduction (10, 11), cellular differentiation (12), growth and proliferation (13, 14, 15), and gene expression at the transcriptional level (16). Finally, the importance of PKR in regulating cell growth and gene expression is underscored by evidence suggesting PKR to be a tumor suppressor gene (17, 18, 19, 20).

PKR, which is efficiently activated by both cellular (21, 22) and viral (6, 23, 24, 25) RNAs, is, in turn, subjected to stringent regulation by both viral and cellular gene products. For example, adenovirus encodes a polymerase III gene product, VAI RNA, which binds to and inactivates PKR (26, 27). In addition, vaccinia virus encodes the K3L protein, which also binds to PKR and inhibits its activity (28, 29). Influenza virus has evolved an unusual mechanism to down-regulate PKR and thus ensure that viral protein synthesis is not compromised during infection. Influenza virus recruits or activates a cellular protein termed P58^IPK, based on its M_r of 58,000 (30, 31). P58^IPK is in a complex with its own inhibitor, I-P58^IPK, and is thus normally inactive in uninfected cells. On influenza virus infection, P58^IPK is dissociated from I-P58^IPK and available to complex with PKR, thereby preventing phosphorylation of eIF-2 (30, 31, 32). Sequence analysis of P58^IPK cloned from bovine, human, and mouse cells showed P58^IPK to be a novel, highly conserved 504-amino acid protein (33, 34). Spanning the P58^IPK molecule are nine tandemly arranged motifs referred to as tetratricopeptide repeats (TPRs). TPRs are 34-amino acid motifs that have been demonstrated to form amphipathic helices that can direct protein-protein interactions (35). The middle region of P58^IPK contains limited homology to the amino-terminal region of eIF-2 , the natural substrate of PKR. Finally, the carboxyl terminus of P58^IPK shows extensive homology to the conserved J domain of the bacterial DnaJ protein (36). Like PKR, P58^IPK likely regulates cellular gene expression in the absence of virus infection. It was demonstrated that overexpression of P58^IPK in NIH 3T3 cells led to the malignant transformation of these cells, suggesting P58^IPK to be involved in cellular growth control (37).

Using in vitro assays with purified components, we previously showed that native or recombinant P58^IPK inhibited both the autophosphorylation of PKR and the phosphorylation of eIF-2 by an already activated kinase (31, 34). We have not, however, directly demonstrated that P58^IPK functions in vivo, inside the cell, to block PKR activity and stimulate protein synthetic rates. We therefore developed an in vivo cotransfection assay using cDNAs encoding P58^IPK and an alkaline phosphatase reporter gene. This allowed us both to analyze P58^IPK inhibitory activity and to precisely map P58^IPK functional domains. We can now report that PKR inhibition and the resultant translational stimulatory activity appear to require two distinct regions of the P58^IPK molecule: the TPR6 domain located in the central region (amino acids 222-255) and the DnaJ similarity region at the COOH terminus (amino acids 391-504). Based on these data, a model of PKR regulation by P58^IPK will be presented.

MATERIALS AND METHODS

A 1.6-kb EcoRV-XbaI fragment containing 115 nucleotides of the bovine P58^IPK 5-untranslated region plus 1.5 kb of coding sequence was cloned into pcDNAI/Neo, as described earlier (34). To construct the Ser-241 mutant, a 1.7-kb BglII-BamHI fragment was subcloned from P58-Ser-241 (34) into pcDNAI/Neo (Invitrogen). To construct P58^IPK-8-3, a 1.3-kb BstEII-BamHI fragment containing the P58^IPK gene with a premature stop codon engineered at amino acid 390 was excised from P58^IPK-8-3 (34) and cloned into a BstEII-BamHI-digested WT plasmid. Plasmids N1, TPR5, TPR6, and TPR7 were created using the MUTA-GENE M13 in vitro mutagenesis kit (Bio-Rad) and the following synthesized oligonucleotides: N1-GTTTGTGTTTGGGATCATGAATATACAATT, TPR5-GTTTGTGTTTGGGATACTGAGGCATTTTAT, TPR6-TTGAAAAACGATAATAAAAGGTGTTTTGCA, and TPR7-CTTGACCAGGATCATCATGAATATACAATT. The N1 oligonucleotide was designed to delete the central region of P58^IPK from amino acids 187-301. This region encompasses both TPRs 5-7 and the eIF-2 homology region. The TPR5 variant lacked amino acids 188-221; The TPR6 variant lacked amino acids 222-255, and the TPR7 variant lacked TPR7 plus the 12 amino acids preceding it. After mutagenesis, a 1.4- or 1.6-kb XbaI-BamHI fragment containing the mutated P58^IPK gene was isolated and cloned into pET15b bacterial expression vector (Novagen). Subsequently, a smaller, 1.0- or 1.2-kb internal BstEII-BamHI fragment containing the respective deletion was excised from the pET15b constructs and cloned into pcDNAI/Neo. The precise identity of all mutant constructs was verified by a combination of restriction enzyme digest mapping and sequence analysis using the dideoxy chain termination technique. See Fig. 1 for a schematic of the wild type and mutant P58^IPK constructs used.

The glutathione S-transferase (GST) fusion constructs for the in vitro studies were made as follows. Full-length, wild type P58^IPK was cloned into pGEX2T plasmid (Pharmacia Biotech Inc.) as described previously (34). The GST-TPR6 construct was made by digesting the TPR6 plasmid (described above) with HindIII and EcoRI restriction enzymes and subcloning this 0.9-kb internal fragment into the GST-P58^IPK backbone.

COS-1 cells were transfected using the DEAE-dextran/chloroquine method (38). Monolayers of COS-1 cells were washed once with serum-free Dulbecco's modified Eagle's medium. DNA was added to the cells in serum-free Dulbecco's modified Eagle's medium in the presence of 250 µg/ml DEAE-dextran. After incubation for 2 h, chloroquine was added to a final concentration of 80 µM. After another 2-h incubation, the transfection mixture was removed, and the cells were ``shocked'' with a solution of 20% glycerol in HEPES-buffered saline for 2 min. The cells were then washed twice with Hank's balanced salt solution and then incubated in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. To measure SEAP activity, culture medium was removed from the cells at 40 h after transfection. The cells were then washed and incubated in prewarmed fresh media for 30 min. The media were then collected from the transfected cells, heated to 65 °C for 5 min, and then assayed for SEAP activity as previously published (39).

Transfected cells were labeled for 30 min with [³⁵S]methionine (605 µCi/ml) in methionine-free Dulbecco's modified Eagle's medium. The labeled cells were washed twice with ice-cold Hank's balanced salt solution and lysed in Triton lysis buffer (10 mM Tris-HCl, pH 7.5, 50 mM KCl, 1 mM dithiothreitol, 2 mM MgCl₂, 100 µg/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, and 1% Triton X-100). The clarified extract was then subjected to immunoprecipitation with protein A-agarose that had been prereacted with a polyclonal antibody against placental alkaline phosphatase (Dako). The immunoprecipitates were washed four times with high salt buffer I (20 mM Tris, pH 7.5, 50 mM KCl, 0.4 M NaCl, 1% Triton X-100, 1 mM EDTA, 10 µg/ml aprotinin, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, and 20% glycerol) and three times with high salt buffer II (10 mM Tris, pH 7.5, 100 mM KCl, 0.1 mM EDTA, 10 µg/ml aprotinin, and 20% glycerol). Bound, radiolabeled proteins were separated on a SDS-9% polyacrylamide gel and visualized using autoradiography. The SEAP protein signal was quantified using PhosphorImager analysis (Molecular Dynamics).

Transfected cells were washed twice with ice-cold Hank's balanced salt solution and lysed in Triton lysis buffer. After SDS-polyacrylamide gel electrophoresis, polypeptides were transferred to nitrocellulose (40) and detected with P58^IPK monoclonal antibodies (37) using the ECL chemiluminescence system (Amersham Corp.).

Poly(A)⁺ RNA was isolated from the transfected cells by the acid guanidinium thiocyanate-phenol-chloroform extraction method (41). The poly(A)⁺ RNA was denatured, electrophoresed in a 1% agarose gel containing 0.2 M formaldehyde, and transferred to a nylon membrane (Hybond N, Amersham). The blot was hybridized to a ³²P-labeled 1.6-kb HindIII-XhoI fragment containing the SEAP gene. Afterward, the same blot was stripped and reprobed with a ³²P-labeled 1.7-kb BamHI fragment containing the human actin gene. Both the radiolabeled SEAP and actin RNA signals were quantified using PhosphorImager analysis.

Both WT P58^IPK and the TPR6 variant were expressed in Escherichia coli and purified as GST fusion proteins as described previously (34). To test for kinase inhibitory activity, purified GST fusion proteins were preincubated with purified PKR (4) for 10 min at 30 °C in 10 mM Tris-HCl, pH 7.5, 50 mM KCl, 2 mM MgCl₂, 2 mM MnCl₂, 2 µM ATP, 5 µg/ml aprotinin, 1 mM dithiothreitol, and 500 µg/ml bovine serum albumin. Subsequently, an activator (0.1 µg/ml poly(rI):poly(rC)) was added in the presence of 5 µCi [-³²P]ATP and incubated for an additional 10 min. Finally, exogenous substrate (10 µg of calf thymus histone IIA) was added with an additional 10 µCi [-³²P]ATP, and the mixture was allowed to incubate for another 20 min. The reaction was stopped by the addition of 2 × disruption buffer (180 mM Tris-HCl, pH 6.8, 4.5% SDS, 23% (v/v) glycerol, 17 mM EDTA, 20 µg RNase A, and 3 M 2-mercaptoethanol) and boiling. The samples were then separated by 14% SDS-polyacrylamide gel electrophoresis and visualized by autoradiography.

RESULTS

A cotransfection assay was developed to measure the ability of P58^IPK to inhibit endogenous PKR function, thereby stimulating mRNA translation inside the cell. A similar assay was frequently and successfully used with other PKR inhibitors, including adenovirus VAI RNA and the reovirus 3 protein (42, 43, 44, 45, 46). This in vivo assay involved cotransfecting into COS-1 cells a constant amount of SEAP reporter gene cDNA together with increasing amounts of P58^IPK cDNA. SEAP enzymatic activity was then measured as described under ``Materials and Methods.'' Although this assay does not directly measure PKR enzymatic activity, this protocol provides a quantitative measure of mRNA translational stimulation, which occurs as a result of PKR inhibition. To test the efficacy of our assay, we first measured the SEAP activity of cells cotransfected with plasmids encoding the adenovirus VAI RNA and SEAP and compared it with the activity obtained when cells were transfected with SEAP cDNA alone. VAI RNA stimulated SEAP activity approximately 4-fold (Fig. 2). Similarly, after cotransfection with a cDNA encoding P58^IPK, SEAP enzymatic levels increased approximately 3.5-fold (Fig. 2). This level of reporter gene stimulation was comparable to that previously observed for reovirus 3, another protein inhibitor of PKR activity (46). As a negative control, we cotransfected human actin cDNA along with SEAP cDNA into COS-1 cells. In contrast to P58^IPK or VAI RNA, the cotransfection of actin DNA did not result in stimulation of SEAP reporter activity. Instead, SEAP activity actually declined relative to the SEAP-alone control, probably as a result of enhanced PKR activation that occurs in transfected cells (47). To further demonstrate that the observed stimulatory activity was specific for P58^IPK, a titration of P58^IPK cDNA transfected into the cells was attempted. Both SEAP activity measurements and Western analyses showed there was an increase in SEAP reporter activity, which correlated with increased P58^IPK expression (data not shown). This was not the case for the actin control, in which increasing amounts of actin expression did not result in an enhancement of SEAP enzymatic activity (data not shown).

The previous analysis measured only the enzymatic activity of SEAP secreted into the culture media during a 30-min pulse. To determine whether the increase in SEAP enzymatic activity was due directly to enhanced SEAP mRNA translation, cotransfected cells were pulse-labeled with [³⁵S]methionine. Cytoplasmic extracts were prepared and subjected to immunoprecipitation using a SEAP-specific polyclonal antibody as described under ``Materials and Methods.'' When increasing amounts of P58^IPK cDNA were cotransfected into the cells with a constant amount of SEAP cDNA, there was a corresponding increase in SEAP protein synthesis (Fig. 3A, lanes 3-5). This increase in the rate of SEAP protein synthesis (3-6-fold as determined by PhosphorImager analysis) paralleled the up to 3-4-fold increase in SEAP enzymatic activity (Fig. 3A, bottom). To determine whether the increased rate of SEAP protein synthesis was due to increased SEAP mRNA levels, Northern blot analysis of poly(A)⁺ RNA from the P58^IPK- and SEAP-cotransfected cells was performed (Fig. 3B). SEAP RNA levels were quantitated using PhosphorImager analysis and normalized according to control actin mRNA levels. Cotransfection of WT P58^IPK cDNA with SEAP cDNA resulted not in increased but, rather, decreased levels of SEAP mRNA (Fig. 3B, compare lanes 1 and 2 or 4 and 5). These data demonstrate that cotransfection with P58^IPK cDNA enhanced the translation of SEAP mRNA and caused minor decreases in SEAP mRNA levels.

Previous data from our laboratory, using in vitro assays to map the functional domains of P58^IPK, implicated P58^IPK amino acids 168-277 as critical for function (34). To directly test the importance of the central region (which contains the eIF-2 homology region) for PKR inhibitory activity and mRNA translational stimulation in vivo, we tested the N1 mutant (deleted for amino acids 188-301; see Fig. 1) in the cotransfection assay. The N1 variant lacks not only the central domain of P58^IPK, but also TPR motifs 5-7. The N1 mutant was first tested in our in vitro assay and, as expected, was demonstrated to be approximately 30-fold less active than the WT P58^IPK protein (data not shown). Similarly, the N1 variant, over a range of concentrations, was unable to stimulate SEAP protein synthetic rates (Fig. 3A, compare lanes 6-8 with lanes 3-5) or enzymatic activity (Figs. 3A, bottom of lanes 6-8, and 4A). Western blot analysis confirmed that the N1 protein was expressed at approximately the same level as WT P58^IPK (Fig. 4B, compare lanes 2 and 3). Consistent with the pulse-labeling experiments, Northern blot studies revealed only minor decreases in SEAP mRNA levels between cells transfected with the N1 mutant and wild type P58^IPK (Fig. 3B, compare lanes 5 and 7). We conclude that the P58^IPK central domain, spanning amino acids 188-301, is important for inhibitory function both in vitro and in vivo.

To more precisely map P58^IPK functional domains, we generated three smaller TPR deletion mutants: TPR5, TPR6, and TPR7 (see Fig. 1). Similar to N1, the TPR6 mutant protein was unable to stimulate SEAP enzymatic activity compared with WT P58^IPK in cotransfected cells. Indeed, SEAP levels were reduced in TPR6- and SEAP-cotransfected cells compared with cells transfected by SEAP cDNA alone (Fig. 4A). Western blot analysis confirmed efficient synthesis of the TPR6 variant (Fig. 4B, compare lanes 5 and 6). In addition, Northern blot analysis confirmed that the inactivity of this mutant was at the level of translation, since only minor decreases in SEAP mRNA levels were observed when compared with the WT (52 versus 77%, respectively, of mRNA compared with levels in cells transfected by SEAP cDNA alone; Fig. 3B, compare lanes 5 and 6). In contrast to TPR6, the TPR5 and TPR7 mutants induced enhanced translational stimulatory activity compared with WT P58^IPK (Fig. 5A). This enhanced activity was not due to enhanced variant expression, as Western blot analysis revealed that TPR5 and TPR7 were expressed at lower levels than the WT P58^IPK (Fig. 5B). It is tempting to speculate that this stimulation by the two mutants is due to lack of binding of negative regulatory factors, which normally interact with these TPR domains. In support of this notion, we recently identified a novel gene product that was found to interact with the TPR7 domain using a yeast two-hybrid screen.²

The carboxyl terminus of P58^IPK shows considerable homology to the conserved J domain of the E. coli DnaJ heat shock family of proteins (36, 48). Previous in vitro assays determined that the DnaJ homology region at the COOH terminus was dispensable for P58^IPK inhibitory activity in vitro (34). However, because eukaryotic DnaJ homologs have been suggested to participate in protein-protein interactions (36), we next determined whether the DnaJ similarity region contributed to the PKR inhibitory function inside the cell. The P58^IPK-8-3 mutant (which lacks the DnaJ region and amino acids 391-504; Fig. 1) failed to stimulate SEAP protein synthetic rates in a pulse-labeling analysis (Fig. 3A, lanes 9-11) or in a SEAP enzymatic activity assay (Figs. 3A, bottom of lanes 9-11, and 5A). Western blots revealed efficient synthesis of the 8-3 variant (Fig. 5B, lane 7), and Northern blot analysis failed to show significant differences in SEAP mRNA levels between cells cotransfected with SEAP cDNA and wild type P58^IPK cDNA or 8-3 mutant DNA (Fig. 3B, compare lanes 2 and 3). These data demonstrate that, like the TPR6 domain, the DnaJ similarity region is required for PKR inhibitory activity and the resultant translational stimulation in vivo.

To provide a direct link between our in vivo cotransfection assay and the ability of P58^IPK to regulate PKR, we examined the PKR inhibitory activity of the key TPR6 mutant using an in vitro kinase inhibition assay. Purified GST-P58^IPK or GST-TPR6 fusion proteins were preincubated with purified PKR for 10 min at 30 °C, after which the poly(rI):poly(rC) activator and histone substrate was added, as described under ``Materials and Methods.'' As shown in Fig. 6, neither the vector control (GST, lanes 2 and 3) nor the GST-TPR6 fusion construct (lanes 6 and 7) was able to reduce the amount of PKR-mediated histone phosphorylation. In contrast, GST-P58^IPK (Fig. 6, lanes 4 and 5) efficiently blocked PKR phosphorylation activity. At the higher concentration of GST-P58^IPK, the percentage of inhibition of PKR activity was 71% as determined by PhosphorImager analysis. These data confirmed our cotransfection analyses demonstrating the inability of the TPR6 P58^IPK mutant to inhibit PKR.

DISCUSSION

In the present study, we developed a cotransfection assay to analyze P58^IPK function inside the cell. As was the case in similar assays with viral-encoded PKR inhibitors, such as the reovirus 3 protein (44, 45, 46) and the adenovirus VAI RNA (42, 43, 49), P58^IPK down-regulated endogenous PKR activity, leading to the stimulation of reporter gene mRNA translation. Although cotransfection with wild type P58^IPK and variants caused minor decreases in SEAP mRNA levels (possibly due to competition for transcription factors; Ref. 44), the predominant effect of P58^IPK action was at the level of mRNA translation. This was perhaps most dramatically demonstrated by the cotransfections with WT P58^IPK and SEAP, in which the levels of SEAP mRNA were reduced approximately 2-fold, whereas the levels of SEAP mRNA translation were concurrently stimulated up to 3-6-fold (Fig. 3). Unexpectedly, we found that both the central region, containing the TPR6 motif, and the COOH terminus, encompassing the DnaJ similarity region, were required for P58^IPK activity in vivo (summarized in Fig. 1). In accordance with this functional data, we found that P58^IPK lacking the TPR6 motif failed to interact with PKR using the yeast two-hybrid system to analyze in vivo binding (50). In addition, the TPR6 variant failed to inhibit PKR phosphorylation activity in vitro (Fig. 6). In contrast, the COOH region of P58^IPK is not required for P58^IPK-PKR interactions (32, 50), nor is the DnaJ region required for P58^IPK function in vitro (34). This region shows homology as high as 55% identity and 80% similarity to select DnaJ proteins that play an essential role in the chaperone function of the hsp70-like DnaK protein of E. coli (36, 48, 51).

Based on our in vitro and in vivo functional and binding data, we present what has become an increasingly complex model of PKR regulation by P58^IPK (Fig. 7). As we have previously shown, P58^IPK likely exists in an inactive complex with its own inhibitor, I-P58^IPK, in uninfected cells. After influenza virus infection, P58^IPK dissociates from I-P58^IPK, allowing P58^IPK to interact with PKR. Although we found that P58^IPK can interact with PKR in the absence of other protein or RNA factors in vitro (32, 34), the present study suggests that additional factors must interact with P58^IPK and/or PKR to allow for P58^IPK-PKR interactions inside the cell. We hypothesize that such a factor (Fig. 7, X) binds to the P58^IPK DnaJ region, a region already known to promote protein-protein interactions (36). Two alternative pathways can further explain how factor X facilitates interactions between P58^IPK and PKR. In one scenario, we propose that X binds to P58^IPK and targets P58^IPK to the correct cellular compartment and then to the protein kinase. This model would likely require that X interact with both P58^IPK and PKR. Such a model could explain why P58^IPK variants, lacking the DnaJ region, can still interact with and inhibit PKR in vitro. Alternatively, factor X, which might itself be a molecular chaperone (52, 53), may bind to the DnaJ region of P58^IPK and alter the conformation or folding of P58^IPK, such that it can now interact with PKR and inhibit the protein kinase. Our earlier studies mapped the P58^IPK-interactive site of PKR to amino acids 244-296, spanning the regulatory and catalytic borders of PKR but including the ATP binding region of the protein kinase catalytic domain (50). In any case, the end result of either of these two model pathways would be the P58^IPK induced inhibition of PKR activity and resultant stimulation of mRNA translation. The interplay between P58^IPK, PKR, and protein X is reminiscent of other systems in which a TPR protein either interacts with other TPR proteins or else another regulatory protein, such as 2, to control its activity (54, 55). Moreover, there is even precedence in the literature for the interaction of another TPR protein, Hip, with the molecular chaperone Hsc70 (56).

Before closing, it is worth commenting on the requirement of the TPR6 motif for both P58^IPK binding and function. Curiously, the central TPR motifs, often including TPR6, were found to be critical for the function of other TPR proteins, including Nuc2, CDC-23, CDC-27, and BimA (57, 58, 59, 60). Furthermore, single amino acid substitutions in TPR domains 5-7 have been shown to abolish the biological function of the yeast CDC23 gene (58, 59). These data suggest a critical function for this region of these TPR proteins despite their diverse regulatory roles. The P58^IPK TPR6 domain contains sequences with limited similarity to a region of eIF-2, the natural PKR substrate. This homologous region, however, is not extensive, with a maximum 31% identity with five gaps. Nonetheless, this region of P58^IPK does contain the PKR consensus eIF-2 phosphorylation motif ELS (P58^IPK amino acids 239-241 and eIF-2 amino acids 49-51). Since the eIF-2 serine 51 is phosphorylated by PKR (61), these data suggested that P58^IPK may function as a PKR substrate (62). This would require the presence of serine 241 to functionally mimic the eIF-2 phosphorylation site. However, we found that a serine to alanine mutation at amino acids 241 (see Fig. 1) did not inactivate P58^IPK function either when P58^IPK was assayed in vitro (34) or in vivo (data not shown), suggesting that P58^IPK probably does not function as a PKR pseudosubstrate. In accordance with these results, we found that P58^IPK likely binds at or near the ATP binding domain on PKR, a region clearly distinct from the substrate binding site. In contrast, the vaccinia virus-encoded PKR inhibitor K3L likely does function as a PKR substrate (28, 29, 63). This argument is supported by our recent evidence that, unlike P58^IPK, K3L binds to a region of PKR spanning amino acids 367-551, which localizes to the large lobe of the PKR protein kinase, thought to be largely responsible for substrate recognition and binding (50, 64).