Dominant Negative Function by an Alternatively Spliced Form of the Interferon-inducible Protein Kinase PKR*

The double-stranded RNA (dsRNA)-activated protein kinase PKR (protein kinase dsRNA-dependent) plays an important role in the regulation of protein synthesis by phosphorylating the α-subunit of eukaryotic initiation factor 2. Through this activity, PKR is thought to mediate the antiviral and antiproliferative actions of interferon. Here, we show that the human T cell leukemia Jurkat cells express an alternatively spliced form of PKR with a deletion of exon 7 (PKRΔE7), resulting in a truncated protein that retains the two dsRNA-binding motifs. PKRΔE7 exhibits a dominant negative function by inhibiting both PKR autophosphorylation and eukaryotic initiation factor 2 α-subunit phosphorylation in vitro and in vivo. Reverse transcriptase-polymerase chain reaction assays showed that PKRΔE7 is expressed in a broad range of human tissues at variable levels. Interestingly, expression of PKRΔE7 is higher in Jurkat cells than in normal peripheral blood mononuclear cells, raising the possibility of a role in cell proliferation and/or transformation. Thus, expression of alternatively spliced forms of PKR may represent a novel mechanism of PKR autoregulation with important implications in the control of cell proliferation.

In eukaryotes, RNA translation plays an important role in cell proliferation induced by various extracellular stimuli (reviewed in Ref. 1). Among many regulatory proteins involved in this process, the interferon (IFN) 1 -inducible double-stranded RNA (dsRNA)-activated protein kinase PKR (protein kinase dsRNA-dependent) is an important regulator of translation initiation through its capacity to phosphorylate the ␣-subunit of eukaryotic initiation factor 2 (eIF-2␣) (reviewed in Ref. 2).
Binding of PKR to dsRNA results in its activation by autophosphorylation and subsequently in the phosphorylation of eIF-2␣ (2). Phosphorylation of eIF-2␣ by PKR on serine 51 leads to an increased affinity of the initiation factor for eukaryotic translation initiation factor 2B, also known as guanine exchange factor, and thus increases the proportion of the latter that is trapped as an inactive complex with eIF-2 and GDP (reviewed in Ref. 3). The reduction in free eukaryotic translation initiation factor 2B results in a fall of the overall rate of guanine nucleotide exchange on the remaining unphosphorylated eIF-2, eventually leading to an inhibition of translation initiation (3). In addition to translational control, PKR has been implicated in signaling pathways leading to transcriptional activation by dsRNA, virus infection, various cytokines, or genotoxic stress (reviewed in Ref. 4).
Several reports have assigned to PKR a tumor suppressor function in vitro (2). Specifically, expression of wild type (WT) human PKR in yeast (5) or in mouse cells (6) results in cell growth inhibition and in some cases in the induction of cell death by apoptosis (7). On the other hand, expression of PKR mutants in NIH3T3 cells that are catalytically inactive or dsRNA binding-defective causes malignant transformation and induction of tumorigenesis after injection of the transformed cells in nude mice (6, 8 -10). Contrary to these in vitro functions, deletion of the pkr gene by homologous recombination is not tumorigenic (11,12). In addition, PKR knockout (PKR Ϫ/Ϫ ) mice are not susceptible to virus infection (11,12) with the exemption of encephalomyelocarditis virus after priming with IFN-␥ (12) or vesicular stomatitis virus after intranasal infection (13,14). Therefore, it has been suggested that the lack of PKR may be compensated by the expression of other PKR-like molecules whose function is possibly blocked by the expression of the PKR mutants in vitro (2,11). This is supported by the cloning and characterization of PKR-related genes, such as PKR-like endoplasmic reticulum kinase/pancreatic eIF-2␣ kinase, which functions as an eIF-2␣ kinase (15), and the mouse homologue of the yeast eIF-2␣ kinase GCN2 (16). Thus, PKR may be the prototype of a family of kinases with overlapping biochemical and biological functions (17).
Work in many laboratories using in vitro mutagenesis of human and mouse PKR has led to an extensive characterization of the structure-function relationship of the molecule (2). Briefly, in both species the amino-terminal half of PKR contains two RNA-binding motifs (dsRBMs) (2), which are conserved among most of the RNA-binding proteins (18). On the other hand, the carboxyl-terminal half of PKR is divided into 11 subdomains, which are required for catalytic activity (2) and are conserved among many serine/threonine protein kinases (19). At the genomic level, the human PKR gene contains 17 exons that vary in size from 18 nucleotides in exon 1 to 840 nucleotides in exon 17 (20), whereas the mouse gene contains * This work was supported by a grant and a postdoctoral fellowship from the Cancer Research Society Inc. (to A. E. K. and S. L., respectively). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Despite the tumor suppressor function in vitro, naturally occurring mutants of PKR in tumor cells have not as yet been identified with the exemption of a mutant form of mouse PKR in a pro-B leukemia cell line (22). Here, we report the cloning of a point mutant of human PKR (Y176H) from Jurkat leukemia cells encoding for a protein that retains the RNA-binding and catalytic properties of wild type PKR in vitro. We also describe the cloning and characterization of an alternatively spliced form of PKR (PKR⌬E7) from Jurkat cells. PKR⌬E7 is a splicing product of exon 7 of the human PKR gene, which contains the two dsRBMs and exhibits a dominant negative function in vitro an in vivo. PKR⌬E7 is differentially expressed in various types of normal human tissue, and we provide evidence that its expression may play a role in induction of cell proliferation as a result of PKR inactivation.

RNA Isolation and RT-PCRs-
The primers used in RT-PCRs are summarized in Table I. For sequencing of the protein-coding sequence of PKR, 1 g of RNA isolated by the guanidium thiocyanate method (23) was reverse transcribed using the P303 primer. The single-stranded PKR cDNA was then amplified by PCR using the P504/P304, P501/ P301, P502/P302, and P503/P303 sets of primers with denaturing at 94°C for 1 min, annealing at 60°C for 1 min, and extension at 72°C for 1 min for a total of 30 cycles. The PCR products, which spanned the entire protein-coding sequence of PKR, were subcloned into pCR™II (Invitrogen) and sequenced with T7 DNA Polymerase (U.S. Biochemical Corp.) or subjected to direct sequencing using the Deaza T7 Sequencing™ Kit according to the supplier's instructions (Amersham Pharmacia Biotech). The human WT PKR cDNA was subcloned into HindIII/BamHI sites of either pcDNA3.0-neo (Invitrogen) or pFLAG-CMV-2 vector (Eastman Kodak Co.). For transient PKR⌬E7 expression, the 900-bp NcoI/AccI fragment of the WT PKR in pcDNA3.0-neo or pFLAG-CMV-2 vector was replaced with the corresponding fragment from pCR™II vector carrying the ⌬E7 deletion to produce PKR⌬E7 or FLAG-PKR⌬E7 under the control of the human cytomegalovirus promoter. For expression in yeast, the HindIII/BamHI fragments of PKR⌬E7 and FLAG-PKR⌬E7 were subcloned into the corresponding sites of pYES2 vector (Invitrogen).
For PKR⌬E7 RNA expression, 1 g of RNA was subjected to RT using the P305 primer and PCR amplification using the P505/P305 set of primers. For expression of PKR⌬E7 in human tissues, the human multiple tissue cDNA (K1420-1) and the human immune system panel (K1426-1) from CLONTECH were used. The normalized cDNAs were amplified by PCR-denaturation at 94°C for 1 min, annealing at 55°C for 1 min, and extension at 72°C for 1 min for a total of 30 cycles using the P505/P305 set of primers. The PCRs were subjected to agarose or polyacrylamide gel electrophoresis and visualized by ethidium bromide or silver staining (24), respectively.
Protein Expression with the Vaccinia/T7 Virus System-One day before transfection, 0.8 ϫ 10 6 HeLa S3 or PKR Ϫ/Ϫ cells were seeded in 6-cm plates. One h before transfection, the cells were infected with recombinant vaccinia virus containing the bacterial T7 RNA polymerase gene (25). Transfection was performed using the ratio of 1 g DNA to 2.5 g of LipofectAMINE (Life Technologies) per plate, and cells were incubated in serum-free medium at 37°C for 5 h. Subsequently, complete medium was added, and cells were grown for an additional 16 h before analysis.
Protein Extraction and PKR Autophosphorylation-Cells were washed twice with ice-cold 1ϫ phosphate buffer saline, and proteins were extracted with a lysis buffer containing 10 mM Tris-HCl, pH 7.5, 50 mM KCl, 2 mM MgCl 2 , 1% Triton X-100, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 3 g/ml aprotinin, 1 g/ml leupeptin, and 1 g/ml pepstatin. After incubation on ice for 20 min, the cell lysate was centrifuged at 10,000 ϫ g for 10 min. The cytoplasmic supernatant (S 10 fraction) was transferred to a fresh tube, the protein concentration was measured by Bradford assay (Bio-Rad), and stored at Ϫ85°C.
For PKR autophosphorylation in vitro, 50 -500 g of protein extracts were subjected to immunoprecipitation with mouse monoclonal antihuman PKR antibodies (clone F9 or E8; Ref. 26) and anti-mouse IgGagarose beads. PKR immunoprecipitates were equilibrated in 1 ϫ PKR kinase buffer consisting of 10 mM Tris-HCl, pH 7.7, 50 mM KCl, 2 mM MgCl 2 , 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 3 g/ml aprotinin, 1 g/ml leupeptin, and 1 g/ml pepstatin. PKR autophosphorylation was performed in the presence of 0.1 g/ml activator reovirus dsRNA and 1 Ci of [␥-32 P]ATP (ICN). After incubation at 30°C for 20 min, the reactions were subjected to SDS-PAGE, and radioactive bands were visualized by autoradiography. Alternatively, the in vitro kinase assay of PKR was performed with S 10 protein extracts in 1ϫ PKR kinase buffer in the conditions described above. The autophosphorylated PKR was then immunoprecipitated with mouse monoclonal anti-human PKR antibodies, subjected to SDS-PAGE, and visualized by autoradiography.
Generation of an Anti-phosphoserine 51-specific eIF-2␣ Antibody-Rabbit antiserum was produced against a chemically synthesized phosphopeptide ILLSELpSRRRIRS (where pS represents phosphoserine) that contains serine 51 of human eIF-2␣. The antibody was purified from rabbit serum by sequence-specific chromatography and was negatively preadsorbed using a nonphosphopeptide corresponding to the site of phosphorylation to remove antibody that is reactive with nonphosphorylated eIF-2␣ protein. The final product was generated by affinity chromatography using an eIF-2␣-derived peptide phosphorylated at serine 51.
Western Blotting-Protein extracts or PKR immunoprecipitates were subjected to SDS-PAGE and proteins were transferred onto nylon Immobilon TM P membrane (Millipore Corp.). Immunoblots were performed with mouse monoclonal anti-human PKR, anti-FLAG (Kodak; catalog no. IB13025), anti-human eIF-2␣ antibodies, rabbit polyclonal anti-phosphoserine 51 eIF-2␣ antibodies (homemade or from BIO-SOURCE, catalog no. , and rabbit antiserum to TrpE-yeast eIF-2␣ fusion protein (CM-217) at a concentration of 1 g/ml using the standard protocol (27). After incubation with horseradish peroxidaseconjugated anti-mouse or anti-rabbit IgG antibodies (1:1000 dilution; Amersham Pharmacia Biotech), proteins were visualized with the enhanced chemiluminescence (ECL) detection system according to the manufacturer's instructions (Amersham Pharmacia Biotech). Quantification of the bands in the linear range of exposure was performed by densitometry using the NIH Image 1.54 software.
Yeast Strains and Growth Analysis-The yeast strains used in this study are summarized in Table II. PKR, PKR⌬E7, and PKRLS9⌬E7 in pYES2 vector were transformed into these strains and selected in SD-Trp medium as described (28). Single colonies were picked up and grew in SD-Trp liquid medium at 30°C to A 600 ϭ 1.5, and the liquid cultures were streaked on SGLU and SGAL minimal plates containing 10% glucose and 20% galactose, respectively, and incubated at 30°C for up to 72 h. The yeast growth was measured by diluting the liquid cultures to A 600 ϭ 0.01 in synthetic medium containing 10% galactose, a According to gb/M35663.
2% raffinose, and the required amino acid supplement (SGAL). The cultures were incubated at 30°C for various time periods, and a 0.5-ml liquid culture from each point was used to measure A 600 .

Cloning of a Point Mutant and an Alternatively Spliced Form of PKR from the Human Leukemia Jurkat T Cells-We have
reported a diminished PKR activation in various human leukemia cell lines including Jurkat T lymphocytes (26). We speculated that PKR inactivation in these cells might be caused by mutations in the PKR gene. To test this possibility, we amplified and sequenced the PKR cDNA from Jurkat cells and normal PBMCs after RT-PCR as described under "Materials and Methods." Direct sequencing of the PCR products verified the presence of a T to C point mutation in Jurkat PKR cDNA at nucleotide 526 downstream from the initiator ATG ( Fig. 1A), which results in a single substitution of tyrosine 176 to histidine. When the PCR products were subcloned into pCRII vector and sequenced (see "Materials and Methods"), nine out of 10 clones contained the T to C mutation. One clone, however, harbored a 77-bp deletion, which corresponds to the entire exon 7 of the human PKR gene (Fig. 1B). Deletion of exon 7 leads to the conjunction of exons 6 and 8 with a frameshift that introduces a stop codon (TGA) within exon 8 and produces an RNA encoding for a 174 amino acid protein. This truncated protein contains the two dsRBMs of PKR and herein is named PKR⌬E7 (Fig. 2).
We confirmed the expression of the PKR⌬E7 RNA in Jurkat cells by an RT-PCR assay (see "Materials and Methods"). The PKR⌬E7 PCR product contains a 77-bp deletion and therefore migrates faster than the WT PKR PCR product on agarose gel electrophoresis (Fig. 3A, compare lanes 3 and 4). The ratio of the band intensities of the two PCR products is proportional to the amount of each PKR transcript within the cells. To examine whether PKR⌬E7 expression is unique for Jurkat cells, we performed an RT-PCR assay with RNA from normal PBMCs. We found that PBMCs contain very low levels of PKR⌬E7 RNA (Ͻ1% of full-length PKR RNA; see also Fig. 8), whereas PKR⌬E7 RNA levels in Jurkat cells is about ϳ10% of WT PKR transcript (Fig. 3A). We also verified the PKR⌬E7 protein expression by immunoblot analysis (Fig. 3B). To facilitate the detection of PKR⌬E7, Jurkat cells were treated with IFN-␣/␤ to induce PKR RNA expression. The protein extracts before (lane 2) and after IFN treatment (lane 3) were subjected to immunoprecipitation and immunoblotting with antibodies specific to the N terminus domain of PKR to detect both full-length PKR (top panel) and PKR⌬E7 (bottom panel). These experiments showed that PKR⌬E7 protein is expressed in Jurkat cells at low levels (lane 2), and its expression is induced after IFN treatment (Fig. 3B, lane 3).
Biochemical Characterization of PKR⌬E7-Characterization of Y176H mutation showed that PKRY176H retains both the dsRNA binding and catalytic activities of WT PKR in vitro (data not shown). As a result of it, we concentrated our efforts on characterizing the function of PKR⌬E7. The N terminus domain of PKR is involved in dsRNA-binding (29). PKR⌬E7 is similar to an artificially made N terminus-truncated form of PKR, known as p20, which contains the two contiguous dsRBMs. p20 can bind to dsRNA (30 -33) and heterodimerize with WT PKR in yeast two-hybrid assays (32). Based on this, we wished to examine the ability of PKR⌬E7 to self-associate and associate with WT PKR in the presence and absence of dsRNA. To do so, we constructed a fusion protein of PKR⌬E7 bearing the FLAG epitope in the N terminus end. When FLAG-PKR⌬E7 and PKR⌬E7 were transiently co-expressed into HeLa cells, an equal amount of the two proteins was co-imunoprecipitated with anti-FLAG antibodies (Fig. 4A, lane 8), confirming their ability to self-associate. In similar assays, an equal amount of endogenous WT PKR (Fig. 4B,  These data suggested that self-association of PKR⌬E7 may take place in the absence of dsRNA, whereas its association with full-length PKR is dsRNA-dependent. To further investigate this possibility, we used a FLAG-PKR⌬E7 construct bearing the LS9 mutation (substitutions of alanine 66 and alanine 68 to glycine 66 and proline 68), which completely abolishes dsRNA binding (34). Co-expression of FLAG-PKRLS9⌬E7 and PKR⌬E7 in HeLa cells and immunoprecipitation with anti-FLAG antibody revealed the lack of association of FLAG-PKR⌬E7 with either PKR⌬E7 (lanes 3 and 4, bottom row) or the endogenous PKR (lanes 3 and 4, top row). Since the LS9 mutation may affect the conformation of the dsRNA-binding domain of PKR (34), these data suggested that the integrity of dsRNA-binding domain is essential for PKR⌬E7 self-association and association with full-length PKR. In these experiments, we noticed that a higher amount of FLAG-PKR⌬E7 was immunoprecipitated with anti-FLAG antibodies after MN treatment. One plausible explanation is that binding of RNA to FLAG-PKR⌬E7 impedes the accessibility of the antibody to FLAG epitope, and this inhibition may be alleviated by MN treatment.
PKR⌬E7 Exhibits a Dominant Negative Function-The ability of PKR⌬E7 to associate with WT PKR prompted us to examine for a possible dominant negative function in PKR activation. To this end, first we assessed the ability of FLAG-PKR⌬E7 to inhibit the autophosphorylation of endogenous PKR when transiently expressed in HeLa cells using the vaccinia/T7 virus system (25). Because the vaccinia/T7 virus system is a two-step procedure utilizing transfection with Lipo-fectAMINE and infection with recombinant virus (see "Materials and Methods"), we measured PKR activation in cells treated with LipofectAMINE and vector DNA (Fig 5A,  . The activation of endogenous PKR was measured first by autophosphorylation in the protein extracts in vitro followed by immunoprecipitation with an anti-human PKR antibody (Fig. 5A). We observed that PKR autophosphorylation, after normalization to protein levels, was more highly induced by the virus (Fig. 5A, compare lane 2 with lane 1) presumably by the production of activator dsRNA during infection. On the other   , lanes 2 and 6). Note the low (undetectable) levels of endogenous eIF-2␣ phosphorylation with this antibody (lane 1) and the lack of its cross-reactivity with the serine 51 to alanine mutant of eIF-2␣ (lane 3). The phosphorylation levels of endogenous HeLa eIF-2␣ were detected, however, when a commercially available phosphoserine 51-specific antibody was used (Fig. 5D) 1 and 2), which was diminished by PKR⌬E7 expression (lane 3) through the inhibition of endogenous PKR. Taken together, the above data demonstrate the dominant negative function of PKR⌬E7 in both PKR activation and eIF-2␣ phosphorylation.
Functional Characterization of the Dominant Negative Function of PKR⌬E7 in Yeast-It has been shown that high level of PKR expression in Saccharomyces cerevisae is toxic due to inhibition of general translation (5). However, at a lower level of expression, PKR can substitute the function of GCN2, the only eIF-2␣ kinase known to exist in yeast (37), by phosphorylating eIF-2␣ on serine 51 and stimulating GCN4 translation, a transcription factor involved in amino acid biosynthesis (38). To verify the dominant negative function of PKR⌬E7 in vivo, we used a yeast strain that lacks endogenous GCN2 (J110) (39) and two strains that contain one (H2544) and two (H2543) alleles of WT human PKR, respectively, under the control of the galactose-inducible promoter (40). Induction of PKR expression in H2544 strain partially inhibits growth, whereas PKR induction in H2543 completely abolishes growth. Strains J110, H2544, and H2543 were transformed with vector alone, FLAG-PKR⌬E7, or FLAG-PKR⌬E7LS9. As positive control, we used the PKR inhibitor vaccinia virus K3L (40). The transformants were streaked onto minimal medium plates containing either glucose or galactose as a carbon source, and the effect of each of these proteins on PKR-mediated growth inhibition was monitored. All transformants of the isogenic J110 strain grew well in either glucose or galactose, indicating that expression of these exogenous proteins did not perturb normal yeast growth characteristics (Fig. 6A). In agreement with previous studies (40), H2544 transformants containing vector DNA without an insert demonstrated a slow growth phenotype after PKR induction (Fig. 6A, bottom plate). However, expression of K3L reversed this growth-inhibitory phenotype (bottom plate). Likewise, expression of FLAG-PKR⌬E7 also rescued yeast growth consistent with previous findings that the N terminus domain of PKR from amino acid 1 to 262 rescues yeast growth inhibition by WT PKR (5). In contrast to this, FLAG-PKRLS9⌬E7 was unable to counteract the growth-inhibitory effects of PKR (bottom plate). Growth curves showed that the ability of FLAG-PKR⌬E7 transformants to rescue growth was equally potent to K3L (Fig. 6B, middle and bottom graphs).
The expression of PKR, FLAG-PKR⌬E7, and FLAG-PKRLS9⌬E7 in yeast was then examined by Western blotting (Fig. 6C). PKR expression was detected in H2544 (top panel, lanes 3 and 4) and H2543 (top panels, lanes 5 and 6) but not in the control J110 strain (lanes 1 and 2). However, expression of WT PKR in H2544 and H2543 strains was more highly induced in the presence of FLAG-PKR⌬E7 than FLAG-PKRLS9⌬E7  lanes 1 and 2). However, expression of FLAG-PKRLS9⌬E7 in strains H2544 and H2543 was very low under growth conditions in which PKR expression was induced (bottom panel, lanes 4 and 6). Since FLAG-PKRLS9⌬E7 does not exhibit a dominant negative function, we speculated that its low expression was due to translation inhibition by PKR. To further investigate this possibility, WT human PKR and FLAG-PKR⌬E7 or FLAG-PKRLS9⌬E7 were co-expressed into yeast strains J80 and J82, which lack GCN2 but contain wild type eIF-2␣ and the serine 51 to alanine mutant eIF-2␣, respectively (38). As shown in Fig. 6D, FLAG-PKR⌬E7 was equally expressed in both strains in the absence (lanes 2 and 5) or presence of WT PKR (lanes 3 and 6). On the other hand, expression of FLAG-PKRLS9⌬E7 was significantly reduced in both strains when WT PKR was induced (lanes 9 and 12). These data indicated that the inhibition of FLAG-PKRLS9⌬E7 expression by WT PKR may not be translational in nature, since PKR-mediated inhibition of protein synthesis cannot take place in the eIF-2␣ mutant-containing strain (38). The mecha- nism of down-regulation of FLAG-PKRLS9⌬E7 by PKR is not presently known.
Next, we examined the dominant negative effect of PKR⌬E7 on PKR-mediated eIF-2␣ phosphorylation in H2544 and H2543 strains (Fig. 6E). To do so, H2544 and H2543 strains were transformed with vector DNA alone (lanes 1 and 5), the vaccinia virus inhibitor K3L (lanes 2 and 6), PKR⌬E7 (lanes 3 and  7), or PKRLS9⌬E7 (lanes 4 and 8) followed by the induction of WT PKR in the presence of galactose. WT PKR expression was detected by immunoblot analysis using an anti-human PKR FIG. 6. Functional characterization of PKR⌬E7 in yeast. A, the yeast strains J110, H2543, and H2544 were transformed with pYES2 vector, pEMBL/yex4 vector containing K3L, and pYES2 vector containing FLAG-PKR⌬E7 or FLAG-PKR LS9⌬E7. The transformants were streaked on SD agar plates containing either glucose or galactose and maintained at 30°C for 72 h. B, the transformants were incubated in SD liquid medium containing 10% galactose and 2% raffinose at 30°C. The growth rate was quantified by measuring the optical density (A 600 ) values at different time points. C, expression levels of PKR, FLAG-PKR⌬E7, and FLAG-PKRLS9⌬E7 in yeast. The transformants of strains J110 ( lanes  1 and 2), H2544 (lanes 3 and 4), and H2543 (lanes 5 and 6) were incubated in SD liquid medium containing 10% galactose and 2% raffinose. 20 g of whole protein extracts was used for Western blotting with a rabbit polyclonal anti-human PKR antibody (top panel) or a mouse monoclonal anti-human FLAG antibody (bottom panel). The band above PKR that is also present in J110 cells, which lack PKR (lanes 1 and 2), is nonspecific (NS). D, protein extracts (20 g) from the J80 (WT eIF-2␣) and J82 (serine 51 to alanine eIF-2␣ mutant) transformants expressing WT PKR ( lanes  3, 6, 9, and 12), FLAG-PKR⌬E7 (lanes 2, 3, 5, and 6), or FLAG-PKRLS9⌬E7 (lanes 8, 9, 11, and 12) were subjected to Western blotting analysis using a mouse monoclonal anti-FLAG antibody. E, protein extracts from H2544 and H2543 cell (20 g) transformed with vector DNA only (lanes 1 and 5), vaccinia virus K3L DNA (lanes 2 and 6), PKR⌬E7 cDNA (lanes 3 and 7), or PKRLS9⌬E7 DNA (lanes 4 and 8) were subjected to immunoblot analysis using a rabbit polyclonal anti-human PKR antibody (top panel), the homemade phosphoserine 51 eIF-2␣-specific antibody (middle panel), or mouse monoclonal anti-eIF-2␣-specific antibody (bottom panel). The intensity of the bands was quantified with the NIH Image 1.54 software, and the ratio of eIF-2␣ serine 51 phosphorylation to the amount of eIF-2␣ protein is indicated. specific antibody (top panel). Phosphorylation of eIF-2␣ was detected by immunoblotting using the homemade phosphospecific antibody (middle panel) and normalized to eIF-2␣ protein levels using a rabbit polyclonal antibody to yeast eIF-2␣ (bot-  6 or 7). These data clearly demonstrate the dominant negative function of PKR⌬E7 in PKR activation and eIF-2␣ phosphorylation in yeast.
Activation of Reporter Gene Expression by PKR⌬E7-The dominant negative function of PKR⌬E7 was further verified in reporter assays in HeLa cells or in mouse fibroblasts derived from two different PKR knockout (PKR Ϫ/Ϫ ) mice (11,12) (Fig.  7). The first PKR Ϫ/Ϫ mouse was generated by the disruption of the N terminus domain of the kinase (deletion of exons 2 and 3; N-PKR Ϫ/Ϫ ; Ref. 12), whereas the second was generated by the disruption of the catalytic domain of the molecule (deletion of exon 12; C-PKR Ϫ/Ϫ ; Ref. 11). Cells were co-transfected with the ␤-galactosidase reporter gene and K3L or PKR⌬E7 cDNA in the absence or presence of WT human PKR cDNA. Expression of K3L or PKR⌬E7 alone induced ␤-galactosidase activity in HeLa cells (Fig. 7A), most likely due to the relief of translational inhibition caused by the activation of the endogenous PKR during transfection (36). As expected, expression of WT PKR in HeLa cells resulted in the inhibition of ␤-galactosidase activity compared with control, which was relieved by the coexpression of either K3L or PKR⌬E7. Note that ␤-galactosidase activity was more highly inhibited when HeLa cells were transfected with a higher amount of WT PKR cDNA (data not shown). On the other hand, in N-PKR Ϫ/Ϫ (Fig. 7B) and C-PKR Ϫ/Ϫ cells (Fig. 7C) K3L but not PKR⌬E7 expression resulted in an induction of ␤-galactosidase activity. This effect of K3L may indicate the presence of other eIF-2␣ kinase(s) that can be activated during transfection. Whether this is PKR-like endoplasmic reticulum kinase (15), GCN2 (16), or another PKR-like kinase is not presently known. Expression of WT human PKR in both knockout cells led to the inhibition of ␤-galactosidase activity (Fig. 7, B and C), which was relieved by the co-expression of either K3L or PKR⌬E7. Taken together, these data demonstrate the dominant negative function of PKR⌬E7 in PKR-mediated inhibition of reporter gene expression. Consistent with our data, Tian and Mathews (42) have recently reported that induction of reporter gene expression by p20 in transient transfection assays in human 293 cells.
Tissue Distribution of PKR⌬E7 RNA-To investigate the physiological relevance of PKR⌬E7 expression, we examined the expression levels of PKR⌬E7 relative to WT PKR RNA in various types of normal human tissue by a RT-PCR assay. As shown in Fig. 8, PKR⌬E7 RNA was expressed in a broad range of human tissues but at variable levels. In most tissues, expression was below 5% of WT PKR RNA, whereas expression in heart, placenta, liver, and skeletal muscle was as high as 5.3, 5.7, 5.3, and 8.4%, respectively. Expression of PKR⌬E7 RNA in spleen was undetectable.

DISCUSSION
In this paper, we have characterized the function of an alternatively spliced form of PKR produced by a deletion of exon 7 (PKR⌬E7). Although alternative splicing has been previously described for the 5Ј-untranslated region of PKR mRNA (43), to our knowledge this is the first study to describe the expression of an alternatively spliced product of human PKR with a dominant negative function. PKR⌬E7 is composed of the two copies of the dsRBMs of PKR, a sequence motif found in many dsRNAbinding proteins (18).
Analysis of the biochemical characteristics of PKR⌬E7 has shown that it binds to dsRNA and is capable of both selfassociating and associating with full-length WT PKR. Both properties of PKR⌬E7 appear to require the presence of dsRNA, since treatment with MN abolishes its association with WT PKR but partially diminishes its self-association (Fig. 4). These findings are in accord with previous observations by Wu and Kaufman (44) that dimerization of intact PKR with p20 requires the dsRNA binding activity. Consistent with this, recent studies by Tian and Mathews have shown that the efficacy or rate of p20/PKR dimerization through a protein/ protein interaction is considerably less than that of their dsRNA-mediated dimerization (42). Our data show that selfassociation of PKR⌬E7 can also take place independently of dsRNA, and this is in line with previous data showing that dimerization of p20 is independent of RNA (45). In accord with this, the intrinsic ability of RNA-free preparations of p20 to dimerize in the absence of dsRNA has been recently reported (42). Therefore, PKR⌬E7 dimerization in vivo may take place in the absence of dsRNA, whereas the presence of dsRNA may induce conformational changes that facilitate heterodimerization and/or heterodimer stabilization between PKR⌬E7 and WT PKR (42,45).
We have seen that self-association of PKR⌬E7 is stronger than its association with full-length PKR, in agreement with a previous study showing that homodimerization of p20 in yeast two-hybrid assays is better than its heterodimerization with the full-length PKR (32). The requirement of dsRNA for PKR⌬E7 binding to PKR might give the specificity for PKR⌬E7 to selectively associate with PKR that is bound to dsRNA supporting the notion that dsRNA binding is required for the dominant negative activity of the N terminus truncated form of PKR (44). In regard to this, the dominant negative function of PKR⌬E7 could be mediated through its interaction with the WT PKR, resulting in the formation of inactive heterodimers (Fig. 9). This may be limited to those latent PKR molecules that are accessible to dsRNA, and this could relieve possible localized inhibitory effects of PKR on protein synthesis. Alternatively, expression of PKR⌬E7 may lead to the sequestration of cellular activator dsRNA, resulting in the inhibition of WT PKR activation (Fig. 9). Consistent with this notion, Tian and Mathews have recently shown that the dominant negative function of p20 correlates with the ability of the two dsRBMs to bind to dsRNA but not with their ability to dimerize, supporting the view that dsRNA sequestration may underlie the dominant negative effect (42).
PKR localizes in the cytoplasm, strongly in the nucleolus, and diffusely throughout the nucleoplasm (46 -48). Studies by Tian and Mathews (42) recently showed that the dsRBMs are required for the localization of PKR, and this activity correlates with dsRNA binding. The same authors demonstrated that p20 exhibits a localization indistinguishable from that of WT PKR, suggesting a similar function for PKR⌬E7 (42). Interestingly, nuclear localization of PKR was recently shown to be rapidly induced upon treatment with DNA-damaging agents (49), but it is not known whether this process requires dsRNA binding. The two dsRBMs of PKR were reported to be required for its association with ribosomes, and targeting to ribosomes may bring PKR closer to the translation machinery, thus facilitating phosphorylation of eIF-2␣ (50). In the same study, a PKR mutant with deletion of the entire kinase catalytic domain from amino acid residue 271 to 551 was found to associate strongly with ribosomes (50). Based on this finding, it is reasonable to speculate that PKR⌬E7 competes with WT PKR for binding to ribosomes, thus preventing the accessibility of PKR to eIF-2␣. However, expression of PKR⌬E7 compared with PKR is less than 10% in most tissues, suggesting that the competition between the two molecules for binding to ribosomes could be local. In fact, several observations have supported the theory of localized activation of PKR in regulation of translation of specific genes (2).
Expression of PKR⌬E7 RNA was detected in a broad range of human tissues at variable levels (Fig. 8). Heart, brain, pla-centa, liver, and skeletal muscle were among the top five tissues that exhibited expression over 5% compared with WT PKR RNA, with the highest levels of PKR⌬E7 RNA expression in the skeletal muscle (8.4%). These five tissues are all energy-demanding, and this could possibly indicate a role of PKR in energy/intermediate metabolism pathways. The other tissues contained less than 5%, whereas expression of PKR⌬E7 RNA in spleen was undetectable. Interestingly, expression of PKR⌬E7 was higher in Jurkat cells than in normal PBMCs (Fig. 3). Whether or not this difference is a cause or an effect of the transformed phenotype of Jurkat cells is an issue that requires further investigation.
We have shown that PKR⌬E7 exhibits a dominant negative function in PKR activation in mouse, human, and yeast cells, which results in growth inhibition through the induction of eIF-2␣ phosphorylation. Whether the dominant negative function of PKR⌬E7 plays a role in pathways other than inhibition of eIF-2␣ phosphorylation is currently under investigation. For example, we have shown that PKR phosphorylates human p53 on serine 392 in vitro, which may account for some of the translational properties of p53 (51). Therefore, PKR⌬E7 may also exhibit a dominant negative function in p53 phosphorylation through its capacity to physically associate with PKR and p53 (data not shown). Also, PKR⌬E7 contains the two dsRBMs, which have been found in many dsRNA-binding proteins (18). These proteins bind dsRNA in a largely sequence-independent fashion and are involved in a myriad of biological processes such as RNA editing (52), RNA trafficking (53), RNA processing (54), transcriptional regulation (55), and the interferon antiviral response (56). Also, the dsRBMs of PKR and other dsRBM-containing proteins have been shown to possess dsRNA annealing activity and may play a role as chaperones by facilitating the folding of cellular RNAs (42,57). Therefore, the possibility that PKR⌬E7 exhibits functions that are independent of PKR cannot be excluded (Fig. 9).
Given the potential importance of the tight regulation of PKR activity in growth control, it is not surprising that several cellular inhibitors of PKR have been identified and characterized. For example, the human immunodeficiency virus-1 TAR FIG. 8. Tissue distribution of PKR⌬E7 RNA. Human multiple normal tissue cDNA (Stratagene) was used for the detection of PKR⌬E7 RNA expression by PCR amplification followed by 10% polyacrylamide gel electrophoresis and silver staining (upper panel). The proportion of PKR⌬E7 RNA compared with full-length PKR RNA in each tissue was quantified using the NIH Image 1.54 software (lower panel).
FIG. 9. Possible biochemical and biological functions of PKR⌬E7 in vivo. In inactive form, the N terminus dsRNA-binding domain (DRBD) of PKR folds over the C terminus kinase domain (KD) keeping it in a "closed" conformation (61). Binding of dsRNA induces PKR homodimerization and exposes the kinase domain, resulting in activation by autophosphorylation (61). Activated PKR induces the phosphorylation of eIF-2␣ on serine 51, leading to the inhibition of translation initiation, cell growth, and/or virus replication (step 1). PKR⌬E7 may exert the dominant negative function by binding to and blocking PKR homodimerization (step 2) and/or by sequestering dsRNA from binding to and activating PKR (step 3). The low levels of PKR⌬E7 (10% compared with full-length PKR) are compatible with the notion for a local dominant negative function in PKR activation and eIF-2␣ phosphorylation. Such local effects have been thought to be maintained by compartmentalization or anchoring of translational factors to cytoskeletal framework and be critical for the translation of specific mRNAs encoding for proteins that play an important role in cell growth, transformation, or differentiation (1). Also, PKR⌬E7 may exhibit functions that are independent of PKR as an RNA-binding protein (step 4; see "Discussion").
RNA-binding protein (TRBP) is a dsRNA-binding protein that inhibits PKR (58) by binding to dsRNA and forming heterodimers with endogenous PKR (32). Interestingly, TRBP overexpression can transform mouse NIH3T3 cells in culture through the inactivation of endogenous PKR (59). However, unlike TRBP, PKR⌬E7 does not exhibit a dominant negative effect on mouse PKR (data not shown), providing evidence for differences in the specificity between the two dsRNA-binding inhibitors of PKR.
The question arises as to what is the physiological significance of inactivation of PKR by PKR⌬E7 and how the inhibitory function of PKR⌬E7 differs from that of other dsRNAbinding PKR inhibitors (e.g. TRBP). One possibility is that association of each of these dsRNA-binding proteins with PKR requires a specific RNA structure, resulting in a local and specific inhibition of PKR activation and eIF-2␣ phosphorylation. Another possibility is that each of the heterodimers between PKR and dsRNA-binding proteins plays a role in RNAmediated biological processes other than translation (i.e. RNA trafficking, editing, splicing, or transport). In this regard, activation of PKR has been implicated in the splicing of human TNF-␣ mRNA (60).
In conclusion, the data presented here demonstrate the expression and the dominant negative function of a dsRNA-binding alternatively spliced product of human PKR. Further understanding of the basis of regulation and function of alternatively spliced PKR products may yield important insights into biological function of the kinase in regard to RNA metabolism, transcription, translation, and regulation of signaling pathways that affect cell proliferation.