Enhanced Antiviral and Antiproliferative Properties of a STAT1 Mutant Unable to Interact with the Protein Kinase PKR*

We have previously reported a physical association between STAT1 and the protein kinase double-stranded RNA-activated protein kinase (PKR). PKR inhibited STAT1 function in a manner independent of PKR kinase activity. In this report, we have further characterized the properties of both molecules by mapping the sites of their interaction. A STAT1 mutant unable to interact with PKR displays enhanced interferon γ (IFN-γ)-induced transactivation capacity compared with STAT1. This effect appears to be mediated by the higher capacity of STAT1 mutant to heterodimerize with STAT3. Furthermore, expression of STAT1 mutant in STAT1−/−cells enhances both the antiviral and antiproliferative effects of IFNs as opposed to STAT1. We also provide evidence that STAT1 functions as an inhibitor of PKR in vitro and in vivo. That is, phosphorylation of eIF-2α is enhanced in STAT1−/−than STAT1+/+ cells in vivo, and this correlates with higher activation capacity of PKR in STAT1−/− cells. Genetic experiments in yeast demonstrate the inhibition of PKR activation and eIF-2α phosphorylation by STAT1 but not by STAT1 mutant. These data substantiate our previous findings on the inhibitory effects of PKR on STAT1 and implicate STAT1 in translational control through the modulation of PKR activation and eIF-2α phosphorylation.

Cytokines and growth factors exert a diverse range of biological activities, from host defense, growth regulation, to immunomodulation. Upon ligand binding to cell-surface receptors, JAK 1 kinases are activated and proceed to phosphorylate the receptor on tyrosine residues, which then function as docking sites for cytoplasmic transcription factors of the STAT family (1,2). STATs are subsequently activated by tyrosine phosphorylation, dimerize by phosphotyrosyl⅐SH2 interactions, and translocate to the nucleus to induce transcription of cytokineresponsive genes (3). A single tyrosine phosphorylation site in the carboxyl-terminal activation domain is absolutely essential for STAT dimerization and DNA binding (3), whereas phosphorylation of a serine residue in this region is important for transactivational activity (4).
One of the major STATs intimately involved in both the innate and acquired immune responses is STAT1. Upon virus infection or exposure to interferons (IFNs), STAT1 is found in protein complexes that bind specific DNA sequences upstream to genes responsible for host resistance. For instance, IFN-␣/␤ induces formation of the heterodimeric ISGF3, whereas IFN-␥ induces binding of homodimeric STAT1 (2,3). Moreover, dsRNA, an intermediate produced during virus replication, can also activate STAT1 DNA binding (5,6). The non-redundant role of STAT1 in the antiviral response is further appreciated by findings that stat1 null mice (STAT1 Ϫ/Ϫ ) are highly susceptible to microbial infection (7,8). IFN signaling leads to the expression of a number of genes, one of which encodes for the dsRNA-dependent protein kinase, PKR (9,10). PKR is a serine/ threonine protein kinase that displays two distinct activities: (i) autophosphorylation upon dsRNA binding (9,10) and (ii) phosphorylation of the eukaryotic translation initiation factor eIF-2␣ (9, 10), a modification resulting in inhibition of protein synthesis (11). Several studies with cultured cells provide evidence for antiviral (12,13), antiproliferative, and tumor suppressor functions of PKR (9,10). However, pkr null (PKR Ϫ/Ϫ ) mice exhibit a modest susceptibility to viral infection (14 -17) and show no signs of tumor formation (14,17), suggesting that the lack of PKR can be compensated by other PKR-like molecules (9,10). This hypothesis is supported by the recent identification of the PKR-related genes, PERK/PEK (18) and the mouse homolog of the yeast eIF-2␣ kinase, GCN2 (19).
We previously described an association between PKR and STAT1 (6). This interaction takes place in unstimulated cells and diminishes upon treatment with IFNs or dsRNA. Increased levels of PKR⅐STAT1 complex have a negative effect on STAT1 DNA-binding and transactivation capacities. In this report we have mapped the interaction sites between the two proteins and have identified a novel function of STAT1. Specifically, we demonstrate that STAT1 functions as an inhibitor of PKR activation and eIF-2␣ phosphorylation in vitro and in vivo. A mutant of STAT1, which was unable to interact with PKR, could not inhibit PKR function in yeast and was better able to mediate the transcriptional, antiviral, and antiproliferative responses of IFNs compared with STAT1. Taken together, these findings not only support our previous observations, but they also provide strong evidence for tight regulation of PKR and STAT1 functions by virtue of their interaction. In addition, our data suggest that STAT1 has a dual role in regulation of gene expression by functioning as a transcriptional factor and possibly as translational regulator through PKR activation and eIF-2␣ phosphorylation.
DNA Binding and Transactivation Assays-Electrophoretic mobility shift analysis was performed using the dsDNA c-Fos c-sis-inducible element (SIE, 5Ј-GATCGTGCATTTCCCGTAAATCTTGTCTACAATT-C-3Ј) according to protocols previously described (6,28). The Dual Luciferase system (Promega) was used to assess the transactivation potential of STAT1. Briefly, STAT1 Ϫ/Ϫ cells or cells expressing STAT1 WT or TM were transfected with Renilla luciferase (pRL-TK) and pGL-2XIFP53 GAS luciferase. Twenty-four hours after transfection, cells were replated and treated with IFN-␥ for 18 h before harvesting. The results presented represent quadruplicate experiments where GAS luciferase activity was normalized to Renilla luciferase activity.
Isoelectric Focusing and PKR in Vitro Kinase Assays-Isoelectric focusing and immunoblot analysis of yeast eIF-2␣ were performed as previously described (29). PKR in vitro kinase assays were carried as previously described (6).
GST Pull-down Assays-Protein production and extraction were performed according to previously described protocols (25,30). Normalized GST fusion proteins were co-incubated with HeLa whole cell lysates or [ 35 S]methionine in vitro translated proteins, washed, subjected to SDS-PAGE, and visualized by fluorography (25).
Yeast Plasmids, Transformations, Growth Protocols, and Protein Extractions-Wild-type and mutants of HA-STAT1 1-413 were subcloned by restriction digest of BamHI-NotI sites into a modified form of the yeast expression vector, pEMBL/yex4 (29), containing a NotI site in the multiple cloning site. Transformation of yeast strains H2544 and J110 and growth analyses were performed as previously described (31).

RESULTS
The catalytic domain of PKR specifically associates with the DNA-binding domain of STAT1. To map the PKR⅐STAT1 interaction we performed a series of binding assays using full-length GST-PKRK296R mutant (GST-PKR WT could not be overexpressed in bacteria (32)) or truncations of PKR bearing either the dsRNA-binding (GST-PKR N, amino acids 1-262) or catalytic (GST-PKR C, amino acids 263-551) domain (Fig. 1A). HeLa S3 extracts expressing human HA-STAT1␣ were incubated with GST-PKR fusion proteins, subjected to SDS-PAGE, and immunoblotted with an anti-HA antibody. As shown in Fig. 1B (top panel), both full-length GST-PKR (lane 5) and the carboxyl terminus of PKR (lane 4) interacted with STAT1 whereas binding to the amino terminus was not detectable (lane 3). Furthermore, this interaction is specific for STAT1, because we could not detect binding of GST-PKR proteins with other STATs (Fig. 1B).
We previously reported that binding of PKR to STAT1 is independent of RNA but requires an intact RNA-binding domain, because the dsRNA-binding-defective mutant PKRLS4 (Arg 58 -Ser 59 -Lys 60 to Gly 58 -Ala 59 -Leu 60 ) (33) does not interact with mouse STAT1 in NIH3T3 cell extracts stably expressing this PKR mutant (6). Based on this, we proposed that the integrity of the dsRNA-binding domain of PKR plays a role in PKR⅐STAT1 interaction in vivo (6). The observation, however, that the carboxyl terminus of PKR is required for binding to STAT1 in vitro (Fig. 1B) prompted us to examine the interaction of STAT1 with GST-PKR bearing either the LS4 or the LS9 mutation (Ala 66 -Ala 68 to Gly 66 -Pro 68 ), which also abolishes RNA binding (33) (Fig. 1C). The RNA-binding-defective mutants were expressed and purified in the GST-PKRK296R background, because expression of LS4 and LS9 in the GST-PKR WT background could not be achieved (data not shown). HeLa S3 extracts containing HA-STAT1␣ were incubated with GST-PKRK296R, GST-PKRK296RLS4, or GST-PKRK296RLS9, and GST-PKR bound proteins were subjected to immunoblotting with anti-HA antibody. All GST-PKR mutants interacted with STAT1 equally well (Fig. 1C, top panel), suggesting that mutations within the dsRNA-binding domain do not interfere with PKR binding to STAT1 in vitro.
To map the region on STAT1 that facilitates its interaction with PKR, we used truncated STAT1 proteins corresponding to its amino-terminal, DNA-binding, linker, SH2, or transactiva-tion domains ( Fig. 2A). The pull-down assays were performed with GST-PKR C, because this protein binds to STAT1 (Fig.  1A) and is more stable than GST-PKRK296R (data not shown). We observed that GST-PKR C specifically interacted with the DNA-binding domain of STAT1 ( Mutations of STAT1 That Affect Binding to PKR-To identify amino acids in STAT1 that form contacts with PKR, we next constructed mutations within amino acids 343-348 of HA-STAT1 1-413 that abolished binding to PKR. Alanine-scan mutagenesis of each of the six amino acids did not yield a point mutant of HA-STAT1 1-413 that disrupted interaction with PKR (data not shown). However, a three-amino acid substitution (TM; Arg 346 -Leu 347 -Leu 348 to Ala 346 -Asp 347 -Asp 348 ) within this region disrupted the ability of STAT1 to interact with GST-PKR C (Fig. 2E, lane 6). Interestingly, HA-STAT1 1-413 TM possessed faster mobility on SDS-PAGE gels compared with WT most likely through changes in the overall charge of the molecule.
To further characterize the interaction of full-length HA-STAT1␣ TM with PKR, we utilized the human fibrosarcoma, U3A cell line, which lacks endogenous STAT1 (35). HA-STAT1␣ WT or TM were co-transfected with PKRK296R into U3A cells, after which, the protein extracts were immunoprecipitated against PKR and immunoblotted with HA antibodies.
As seen in Fig. 2F, STAT1␣ WT associated with both transfected and endogenous PKR (upper panel, lanes 2 and 5). Conversely, STAT1 TM binding with endogenous PKR was completely abolished (lane 3) and displayed very marginal binding to transfected PKR, which was detectable only after long exposures (lane 6). In contrast to HA-STAT1 1-413 TM, full-length STAT1␣ TM did not display any difference in its migration pattern compared with STAT1␣ WT. These in vitro findings were also verified in vivo by the yeast two-hybrid assay (data not shown). Taken together, it appears that the DNA-binding domain of STAT1 interacts with PKR in vitro and in vivo.
DNA Binding and Transcriptional Properties of STAT1 TM-To test the ability of STAT1 TM to respond to IFN-␥ treatment, we performed transient transfection assays in STAT1 Ϫ/Ϫ cells using STAT1 WT or STAT1 TM and a luciferase reporter construct driven by two copies of the GAS element from the IFP53 gene (28). As shown in Fig. 3A, we observed that luciferase expression in cells transfected with STAT1 WT was induced by IFN-␥ treatment. However, in cells transfected with STAT1 TM, we observed, after normalization to Renilla luciferase, a much higher basal luciferase activity (ϳ5-fold) that could be slightly induced by IFN-␥ stimulation.
To better characterize STAT1 TM, we infected STAT1 Ϫ/Ϫ fibroblasts with retroviruses harboring the puromycin-resistant gene and HA-STAT1␣ WT or HA-STAT1␣ TM. As a control, retroviruses containing only the puromycin-resistant gene were used. After puromycin selection, polyclonal populations of STAT1 WT-expressing cells showed ϳ5-fold greater expression over STAT1 TM pools (Fig. 3B, compare lanes 2 and 3). Transactivation assays using the 2XIFP53-GAS luciferase reporter correlated with our findings in transient transfection experiments that STAT1 TM confers higher basal reporter activity, which can be induced by IFN-␥ treatment (Fig. 3B). We next tested whether STAT1 TM could bind DNA after IFN treatment. Although we could not detect ISGF3 formation in STAT1 TM cells in response to IFN-␣/␤ (data not shown), IFN-␥ stimulation resulted in the formation of DNA-binding complexes consisting of either STAT1⅐STAT3 heterodimers or STAT3 homodimers, but not that of STAT1 homodimers (Fig. 3C, compare lanes 7-16). This finding is consistent with previous re-ports that STAT3 is activated following IFN treatment (3). Moreover, the intensity of the STAT3 homodimer appears to be higher compared with control or STAT1 WT cells (compare lanes 2, 4, and 6).
The ability of STAT1 to be phosphorylated upon IFN stimulation was also examined (Fig. 3D). To compare STAT1 phosphorylation per equal amounts of STAT1 protein, we used a 5-fold higher amount of STAT1 TM extracts versus STAT1 WT before and after IFN stimulation. These reactions were also normalized to total protein concentration by the addition of treated or untreated STAT1 Ϫ/Ϫ control extracts. Although STAT1 WT was tyrosine-phosphorylated following IFN-␣/␤ or IFN-␥ treatment (top panel, lanes 5 and 6), we failed to detect STAT1 TM tyrosine phosphorylation (lanes 8 and 9). In contrast, phosphorylation of serine 727 did not significantly differ between STAT1 WT and STAT1 TM after IFN treatment (middle panel, lanes 5-6 and 8 -9). Reprobing of the membrane with antibodies to HA revealed the hypo-and hyperphosphorylated forms of STAT1 usually observed after IFN treatment (lower panel). Because STAT1 is also known to form heterodimers with STAT3 following IFN stimulation (3), we tested whether STAT1 TM could associate with STAT3. A much higher amount of STAT3 co-precipitated with STAT1 TM before and after IFN treatment (Fig. 3E, top panel, lanes 7-9), although STAT1 protein levels were approximately equal (bottom panel). We also analyzed expression and activation of STAT3 in the same protein extracts used for the STAT1/STAT3 co-immunoprecipitation. STAT3 phosphorylation was slightly elevated (ϳ50%) in cells expressing STAT1 TM before or after treatment with either IFN-␣/␤ or IFN-␥ (Fig. 3F, lanes 7-9). This increase in STAT3 activity may account for increased STAT3 DNA binding in STAT1 TM cells.
STAT1 TM Enhances the Antiviral and Antiproliferative Effects of IFNs-The biological effects of STAT1 TM activation were examined by cell cycle analysis after treatment with IFN-␣/␤ or IFN-␥ (Fig. 4A). A greater proportion of STAT1 TMexpressing cells (IFN-␣/␤, 6 -8%; IFN-␥, 10 -11%) were arrested in G 0 /G 1 phase after treatment with either type I or type II IFNs (right panel) compared with control (left panel) or STAT1 WT-expressing (middle panel) cells. In addition, the ability of STAT1 TM cells to resist virus infection was also investigated. Control, STAT1 WT, and STAT1 TM cells were primed with IFNs and subsequently infected with serially diluted VSV. The amount of virus needed to induce CPE was qualitatively measured. As shown in Fig. 4B (upper panel), STAT1 TM cells that were treated with IFN-␥ were ϳ50-fold more resistant to VSV infection compared with STAT1 WT cells, and ϳ10 4 -fold more resistant versus control STAT1 Ϫ/Ϫ cells. In contrast, IFN-␣/␤-treated STAT1 TM cells were 10-fold more susceptible to VSV CPE compared with control, STAT1 WT, and STAT1 ϩ/ϩ cells. Interestingly, even untreated STAT1 TM cells provided a greater degree of protection compared with STAT1 WT and STAT1 ϩ/ϩ cells. This enhanced ability of STAT1 TM cells to resist virus infection was also observed after encephalomyelocarditis virus infection (data not shown). Western blotting against STAT1␣ revealed that STAT1 TM is expressed at much lower levels than STAT1 WT and endogenous STAT1 from STAT1 ϩ/ϩ cells (bottom panel). Taken together, these data suggest that STAT1 TM enhances the antiproliferative and antiviral effects of IFNs on a per molecule basis.
STAT1 Functions as a PKR Inhibitor in Vitro-To gain better insight into the PKR⅐STAT1 interaction, we next mapped Transformants were streaked on control 10% glucose (upper plate) or 10% galactose agar plates and monitored for slow growth phenotype. B, transformants were grown in galactose medium, and relative growth rates were monitored at the indicated times by trypan blue cell counting. The upper graph represents the growth curves of control J110 transformants whereas the bottom graph shows the growth curves of H2544 transformants. C, J110 and H2544 were grown in galactose medium to induce HA-STAT1 1-413 WT and TM expression. 2 mg of protein extracts were immunoprecipitated with HA antibodies and immunoblotted with HA antibodies. D, J110 and H2544 were grown in galactose medium to induce PKR expression. Total protein extracts were subjected to immunoblot analysis with rabbit antisera to human PKR. The upper band, which is present in J110 extracts lacking PKR, is nonspecific (NS). E, extracts from J110 control and H2544 strains transformed with vector alone, K3L, HA-STAT1 1-413 WT, or HA-STAT1 1-413 TM were subjected to isoelectric focusing after induction with galactose and probed with antibodies against yeast eIF-2␣. the STAT1-binding site on PKR (Fig. 5A). In agreement with earlier results, the full-length kinase domain (lane 4), but not the dsRNA-binding domain (lane 3), bound PKR. Truncation of the kinase domain from either the amino or carboxyl terminus (Fig. 5A) defined a region critical for STAT1 binding: amino acids 367-415 (lanes 5-9). Interestingly, this corresponds to the same region in the large lobe of the PKR kinase domain, to which the vaccinia virus K3L protein binds to block PKR activation (36,37). In view of this fact, we tested whether the two proteins could compete for binding to PKR. Bacterially expressed FLAG-K3L was co-incubated with GST-PKR C and extracts from HeLa cells expressing increasing amounts of HA-STAT1␣. Immunoblotting was performed to detect either binding of HA-STAT1␣ or FLAG-K3L to GST-PKR C. As seen in Fig. 5B, STAT1 displaced K3L from PKR in a dose-dependent manner (bottom panel, lanes 10 -13), indicating that K3L and STAT1 bind to the same region on PKR.
The PKR pseudosubstrate, K3L, has been shown to bind PKR and block access of eIF-2␣ to the catalytic pocket of PKR (29,37). To address the importance of STAT1 binding to PKR, we investigated the ability of STAT1 to act as a cellular inhibitor of PKR. A series of in vitro assays were performed to assess the effect of STAT1 on PKR activation. Human PKR was activated by reovirus dsRNA in vitro from HeLa S3 cells in the presence of increasing amounts of recombinant full-length STAT1 (Fig. 5C, middle panel). We observed that PKR autophosphorylation diminished in a dose-dependent manner by the addition of STAT1 (top panel), although PKR protein levels were equal (bottom panel).

STAT1 Inhibits the Antiproliferative Properties of PKR in
Yeast-To date, the best approach to test for the translational function of PKR is in Saccharomyces cerevisiae. It has been shown that high levels of PKR expression in yeast are toxic due to inhibition of general translation (38). However, at lower levels of expression PKR can substitute the function of GCN2 (39), the only eIF-2␣ kinase known to exist in S. cerevisiae (40), by phosphorylating eIF-2␣ on serine 51 to inhibit protein synthesis. Through this approach, a number of PKR inhibitors have been identified and characterized (31). Given that PKR expression in yeast results in inhibition of cell growth (38,39), we wanted to analyze whether STAT1 could block this effect when co-expressed with PKR. Yeast strain H2544, which lacks the yeast eIF-2␣ kinase GCN2, contains a stable integration of human PKR WT cDNA downstream to a galactose-inducible promoter, whereas the isogenic strain J110 is identical except that the PKR cDNA was not inserted. Previous studies have shown that induction of PKR expression in H2544 results in inhibition of cell growth through phosphorylation of eIF-2␣ (39). Conversely, co-expression of a PKR inhibitor, such as K3L, leads to rescue of PKR-mediated growth inhibition (29). To test the inhibitory activity of STAT1 on PKR, strains J110 and H2544 were transformed with vector alone, K3L, HA-STAT1 1-413 WT, or HA-STAT1 1-413 TM. Transformants were streaked onto minimal media 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, upper plate, and Fig. 6B, upper graph). H2544 transformants containing only empty plasmid DNA demonstrated a slow-growth phenotype after PKR induction (Fig. 6A, lower plate, labeled C). However, expression of K3L reversed this growth inhibitory phenotype (lower plate, labeled K3L). Likewise, expression of STAT1 1-413 WT also rescued yeast growth (lower plate, labeled WT), although in contrast, the interaction mutant of STAT1 was unable to counteract the growth inhibitory effects of PKR (lower plate, labeled TM). Growth curves to assess the degree of rescue showed that the ability of HA-STAT1 1-413 WT transformants to rescue growth was half as potent relative to K3L (Fig. 6B, lower graph). Because it was not possible to quantify the relative levels of K3L and STAT1 1-413 WT expression, the degree of rescue may be dependent on their different levels of expression. Efforts to rescue growth by full-length HA-STAT1␣ were unsuccessful, because we could not detect HA-STAT1␣ expression in yeast (data not shown). However, the truncated HA-STAT1 1-413 proteins were readily detectable in yeast protein extracts (Fig. 6C) as was the expression of human PKR in H2544 transformants (Fig. 6D, lanes 5-8). In correlation with the growth curves, the extent of eIF-2␣ phosphorylation, as assessed by isoelectric focusing experiments (29), was diminished in H2544 transformants expressing either K3L or HA-STAT1 1-413 WT (Fig. 6E, lanes 3 and 4). The induction of PKR levels in Fig. 6D, lane 6, was probably translational in nature as a result of inhibition of eIF-2␣ phosphorylation and up-regulation of PKR protein synthesis by K3L (Fig. 6E, lane  3). This PKR up-regulation was not evident in lane 7 most likely due to the weaker inhibitory effect of STAT1 1-413 WT on eIF-2␣ phosphorylation compared with K3L (Fig. 6E, compare lanes 3 and 4).

Increased PKR Activation and eIF-2␣ Phosphorylation in STAT1
Ϫ/Ϫ Cells-Next we investigated whether the loss of STAT1 would augment PKR activity. To do so, protein extracts from untreated or IFN-treated STAT1 ϩ/ϩ and STAT1 Ϫ/Ϫ cells were used to assess PKR activity in vitro. We noticed that the basal activity of PKR was ϳ5-fold higher in STAT1 Ϫ/Ϫ cells compared with STAT1 ϩ/ϩ cells (Fig. 7A, upper panel, compare  lanes 1 and 3), although PKR protein levels were equivalent, as assessed by in vivo [ 35 S]methionine labeling (lower panel, compare lanes 1 and 2). The increase in PKR activity after IFN treatment in STAT1 ϩ/ϩ cells (top panel) reflects increased PKR expression, whereas no such increase in PKR activity/protein was observed in STAT1 Ϫ/Ϫ cells. This is consistent with the notion that transcriptional up-regulation of PKR after IFN treatment is dependent on the JAK/STAT pathway (upper panel, compare lanes 2 and 4). To further substantiate our findings that PKR activity is elevated in STAT1 Ϫ/Ϫ cells, we compared the levels of eIF-2␣ phosphorylation in STAT1 ϩ/ϩ and STAT1 Ϫ/Ϫ cells in vivo. The effect of STAT1 on eIF-2␣ phosphorylation in STAT1 ϩ/ϩ and STAT1 Ϫ/Ϫ cells was examined by treatment with dsRNA and immunoblotting with a phosphospecific antibody to phosphoserine 51 of eIF-2␣ (Fig.  7B, top panel). These experiments showed that a higher amount of eIF-2␣ was phosphorylated in STAT1 Ϫ/Ϫ cells compared with STAT1 ϩ/ϩ cells (top panel, compare lanes 1 and 3) and that this phosphorylation was more highly induced after dsRNA treatment (compare lanes 2 and 4). Taken together, the inhibition of PKR activity by STAT1 in mammalian and yeast cells supports our findings that STAT1 can inhibit PKR activity in vitro and in vivo. DISCUSSION We have mapped the sites of interaction between PKR and STAT1 in vitro using GST pull-down assays and also in vivo. We found that STAT1 interacts with PKR on amino acids 367-415, an area located within the large lobe of the kinase domain of PKR that is also bound by the vaccinia virus encoded PKR inhibitor, K3L (36,37). Sequence comparison with other eIF-2␣ family members, GCN2, HRI, and the newly discovered PERK/PEK revealed that this part of the kinase domain is highly conserved and suggests that these members might be capable of interacting with STAT1. Our previous observations that the dsRNA-binding mutant PKRLS4 was unable to associate with STAT1 in vivo (6), at first, appear to be challenged by our in vitro studies herein, which show that the carboxyl terminus of PKR is required for binding to STAT1 and that GST-PKRK296RLS4 can interact with STAT1 ( Fig. 1). One possible explanation for this difference may be inferred, based on structural data of the amino terminus of PKR (41) and also on previous mutational studies (33), that the LS4 and LS9 mutations could introduce conformational changes that would affect its interaction with STAT1 in vivo. Such conformational changes may not take place when GST-PKRK296RLS4 and GST-PKRK296RLS9 are purified from bacteria. Another conceivable explanation is that the association of PKR with STAT1 in vivo is facilitated by the presence of other protein(s) whose action is modulated by the amino terminus of PKR and may become limited when the interaction is tested in vitro.
In the case of STAT1, amino acids 343-348 provide a major site of interaction with PKR. This region of STAT1 corresponds to ␤-sheet 3, a structure located on the back side of the DNAbinding domain of STAT1 (33,34). Mutation of three amino acids within this region (STAT1 TM; Arg 346 -Leu 347 -Leu 348 to Ala 346 -Asp 347 -Asp 348 ) abolished binding to PKR in vitro and in vivo. Paradoxically, sequence alignment of this three-residue stretch of wild-type STAT1 with other STAT members showed almost complete conservation. Similarly, residues in ␤-sheet 3 of the DNA-binding domain of STAT3 responsible for specific interaction with c-Jun are also nearly identical between STAT members; however, c-Jun specifically interacts with STAT3 (42). Thus, our observation that PKR specifically interacts with STAT1 can probably only be explained in the context of the presentation of ␤-sheet 3 in STAT1 compared with other STAT molecules. In the end, structural analysis might be necessary to determine the exact points of contact between PKR and STAT1.
In light of our previous observations that the ability of PKR to interact with STAT1 can block STAT1 DNA binding (6), we examined whether release of STAT1 from this inhibitory mechanism might augment its transcriptional activity. In contrast to STAT1 WT, we were unable to observe tyrosine phosphorylation of STAT1 TM following IFN stimulation, and as a result, ISGF3 and STAT1 homodimer formation could not be detected in DNA binding experiments. We believe that the inability of STAT1 TM to be phosphorylated cannot be the result of a misfolded protein, because this mutant is phosphorylated on serine 727 after IFN treatment. Rather, it appears that some undefined negative regulatory mechanism is responsible for this phenomenon. The protein kinase that directly phosphorylates serine 727 of STAT1 in vivo is not as yet known (4). It is unlikely to be PKR because STAT1 is not phosphorylated by PKR in vitro (6). Interestingly, a recent report shows a defective serine 727 phosphorylation of STAT1 in PKR Ϫ/Ϫ fibroblasts after IFN stimulation providing evidence for an indirect role of the kinase in this process (43). Although unable to bind DNA as a homodimer, STAT1 TM was competent and more capable in forming DNA-binding complexes with STAT3 relative to STAT1 WT cells. Interestingly, tyrosine phosphorylation of STAT3 was elevated by 50% before and after IFN treatment in polyclonal cell populations expressing STAT1 TM. The mechanism behind this finding is not clear at this time, but it might be possible that recruitment of STAT3 to the JAK⅐receptor complex is more efficient in STAT1 TM cells compared with STAT1 WT cells. Nevertheless, the net effect of increased STAT1⅐STAT3 heterodimer and STAT3 homodimer formation probably contributes to up-regulation of STAT1⅐STAT3 DNAbinding and GAS-dependent transactivation. Because STAT3 homodimers appear to play a minimal role in the transactivation of IFN-dependent genes by IFN-␥ (44), it is likely that the transcriptional activity observed in our reporter assays is contributed by the STAT1⅐STAT3 heterodimer. This is the second instance where a transcriptional role of STAT1 has been shown to require serine 727 but not tyrosine 701 phosphorylation. An earlier study demonstrated that mutation of serine 727 to alanine, but not tyrosine 701 to phenylalanine, on STAT1 significantly ablated the TNF-␣-dependent induction of caspase genes (45). It was speculated that STAT1 could potentially participate in signaling pathways independent of its tyrosine phosphorylation state (46); such a hypothesis is supported by our findings in here.
The antiproliferative and antiviral effects of IFNs were also investigated, because a number of cell cycle regulatory proteins are regulated at the transcriptional level through the JAK/ STAT signaling pathway (44). IFN-dependent cell cycle arrest, as measured by FACS analysis, was significantly higher in cells expressing STAT1␣ TM versus control cells or cells expressing STAT1␣ WT. In addition, the ability of STAT1 TM cells to resist VSV infection was ϳ10-fold greater than control or STAT1 WT cells following IFN-␥ stimulation. These differences become more significant when the variation in the expression levels between STAT1 WT and TM is considered (i.e. 5-fold higher expression of WT than STAT1 TM; Fig. 3B, lanes  2 and 3; Fig. 4B, bottom panel). Interestingly, even untreated STAT1 TM cells provided greater protection relative to control and STAT1 WT cells, suggesting that the anti-viral responses available within STAT1 TM cells are already enhanced prior to IFN pretreatment.
Given the fact that STAT1 and K3L, a pseudosubstrate in-hibitor of PKR, can compete for binding with PKR, we next analyzed whether STAT1 could inhibit PKR activation. We observed that recombinant STAT1 inhibited PKR activation in vitro in a dose-dependent manner. However, it is unlikely that STAT1 is a substrate or pseudosubstrate for PKR, because sequence alignment of STAT1 with K3L and eIF-2␣ did not reveal any significant similarities, findings that coincide with our earlier observations that PKR does not phosphorylate STAT1 in vitro or in vivo (6). Instead, the interaction of the two proteins might prevent opening of the cleft that separates the two lobes of PKR's kinase domain, thus inhibiting both PKR activation and eIF-2␣ phosphorylation. The capacity of STAT1 to neutralize PKR activity is further appreciated from in vivo studies. PKR exhibits the same substrate specificity as the yeast GCN2 kinase that regulates protein synthesis by phosphorylating eIF-2␣ to inhibit cell growth and replacement of GCN2 with PKR leads to inhibition of translation and yeast growth (31). As such, this system has been consistently used as a means to probe for PKR activation, as well as for characterizing inhibitors of PKR, like K3L (29). The observation that HA-STAT1 1-413 WT, but not the interaction mutant, TM, was able to block PKR activity and subsequently relieve the growth inhibitory effects of PKR, supports our model whereby PKR activity is tightly regulated by its interaction with STAT1. These results from yeast were further substantiated in STAT1 Ϫ/Ϫ cells where we observed augmentation of both PKR activity and in vivo eIF-2␣ phosphorylation in comparison to wild-type cells.
We rationalize that the interaction of PKR and STAT1 modulates cellular proliferation in normally growing cells and in cells challenged by viruses. In unstimulated cells, the balance of PKR⅐STAT1 heterodimer formation versus their free monomers may dictate cellular proliferative capacity. For example, perturbing this equilibrium by the expression of catalytic mutants of PKR would sequester a larger fraction of STAT1 (6) and could contribute to increased virus susceptibility and to the transformed phenotype of cells expressing such mutants of PKR (9,10). Conversely, loss of PKR's repressive activity in PKR Ϫ/Ϫ mice would favor the antiviral and antiproliferative functions of STAT1. As such, PKR Ϫ/Ϫ mice are still capable of providing host resistance to many viruses (14,17). It would thus appear that STAT1 provides a first line of defense against virus infection, whereas the role of PKR in viral resistance appears to be secondary. This is illustrated by the observation that STAT1 Ϫ/Ϫ mice are exquisitely sensitive to a variety of pathogens (7,8), whereas PKR Ϫ/Ϫ mice exhibit a modest sensitivity (14,17). Increased eIF-2␣ phosphorylation in STAT1 Ϫ/Ϫ cells could provide a compensatory mechanism for the loss of STAT1 and may play a role in differences in the susceptibility of STAT1 Ϫ/Ϫ mice to virus infection (8,24). Therefore, eIF-2␣ phosphorylation may play a role in the differential sensitivity of various viruses to interferon treatment, an effect that could be mediated through the interaction of STAT1 with PKR and other eIF-2␣ kinases, because they share high sequence homology in the STAT1 binding region. Generation of mice lacking multiple members of this family of kinases 2 may be useful in determining such virus-specific responses.