Review

Signal Integration via PKR

See allHide authors and affiliations

Science's STKE  03 Jul 2001:
Vol. 2001, Issue 89, pp. re2
DOI: 10.1126/stke.2001.89.re2

Abstract

The vital role of interferons (IFNs) as mediators of innate immunity is well established. It has recently become apparent that one of the pivotal proteins in mediating the antiviral activity of IFNs, the double-stranded RNA (dsRNA)-activated protein kinase (PKR), also functions as a signal transducer in the proinflammatory response to different agents. PKR is a member of a small family of kinases that are activated by extracellular stresses and that phosphorylate the α subunit of protein synthesis initiation factor eIF-2, thereby inhibiting protein synthesis. The activation of PKR during infection by viral dsRNA intermediates results in the inhibition of viral replication. PKR also mediates the activation of signal transduction pathways by proinflammatory stimuli, including bacterial lipopolysaccharide (LPS), tumor necrosis factor α (TNF-α), and interleukin 1 (IL-1). PKR is a component of the inhibitor of κB (IκB) kinase complex and plays either a catalytic or structural role in the activation of IκB kinase, depending on the stimulus. The activities of the stress-activated protein kinases p38 and c-Jun NH2-terminal kinase (JNK) are also regulated by PKR in a pathway that leads to the production of proinflammatory cytokines. This review will focus on the role of PKR in nuclear factor κ B (NF-κB) and mitogen-activated protein kinase (MAPK) pathways, because these have been the subjects of a series of publications over the past year that have reported conflicting findings. Although the conflicts may not be resolved in this review, suggestions are made for experiments that could lead to a clearer understanding of the mechanisms involved.

Introduction

Protein kinase R (PKR) is a ubiquitously expressed serine-threonine kinase that has been implicated as a signal integrator in translational and transcriptional control pathways (Fig. 1). A pair of double-stranded RNA (dsRNA) binding motifs located in the NH2-terminal domain of PKR act as a sensor to bind dsRNA or protein activators (which themselves contain dsRNA binding domains). Following binding, there is a conformational change in PKR, which results in relief of autoinhibition, exposure of the kinase active site, and autophosphorylation [reviewed in (1, 2)]. The known cellular targets of PKR phosphorylation include the α subunit of translation initiation factor eIF-2 and the regulatory subunit of protein phosphatase 2A (PP2A), B56α. The phosphorylation of these substrates by PKR results in inhibition of protein synthesis (Fig. 1).

Fig. 1.

Signaling by PKR. PKR can be activated directly by dsRNA or by viral dsRNA replicative intermediates, and indirectly by cytokines or bacterial products such as LPS or lipoprotein. The binding of dsRNA to PKR causes a conformational change that leads to autophosphorylation. Cytokines and bacterial products activate PKR through upstream activators that could be kinases or other types of protein activators. Once activated, PKR phosphorylates downstream substrates, including the B56α subunit of protein phosphatase 2A (PP2A) and the α subunit of initiation factor eIF-2, inhibiting translation. Transcriptional control is mediated by PKR interactions with signaling modules of either the stress-activated protein kinase family (p38, JNK) or IKK. The requirement for PKR catalytic activity is stimulus dependent (see text for details). PKR also mediates apoptosis by regulating FADD, leading to caspase activation.

Transcription factors, including nuclear factor-κB (NF-κB), which are activated by a variety of stimuli, can also be regulated by PKR-mediated signaling. However, the mechanisms involved in PKR-dependent regulation, and the downstream targets of PKR, remain to be elucidated. For example, there is debate as to whether the kinase activity of PKR is required or whether it plays a structural role as a component of a signaling complex. PKR signals in the stress-activated protein kinase (p38) and c-Jun NH2-terminal kinase (JNK) pathways in response to extracellular signals acting through cell surface receptors (Fig. 1). Whether the kinase activity of PKR is required for p38 activation depends on the stimulus. The activity of the tumor suppressor p53 in cellular stress responses may also be subject to regulation by PKR, and the activation of the transcription factor IRF-1 is PKR dependent (Fig. 2). Neither the signaling pathway to p53 or IRF-1, nor the relevant PKR substrates in those pathways, have been described. The transcription factors signal transducer and activator of transcription 1 (STAT1) and STAT3 can depend on PKR for DNA binding activity or for enhancing transcriptional activity by serine phosphorylation (Fig. 2). PKR has also been implicated in the control of splicing of the tumor necrosis factor-α (TNF-α) mRNA in a kinase-dependent pathway.

Fig. 2.

PKR-dependent and independent signaling by dsRNA. Cells can respond to dsRNA by PKR-dependent or independent pathways (right side). In PKR-null cells (indicated by the dotted lines), dsRNA signaling to NF-κB can be restored by pretreatment with IFN through an unidentified protein kinase X (PKX). A mutant cell line has been isolated in which dsRNA activation of PKR is intact, but induction of IFN-stimulated genes is deficient. The activation and phosphorylation of IRF-3 and IRF-7 occurs by a PKR-independent pathway. JAK; Janus kinase.

The Link Between dsRNA Signaling and PKR

In 1988, Phillip Marcus and Margaret Sekellick proposed that PKR was required for the induction of the expression of the gene for the cytokine interferon (IFN) (3). This suggestion was based on experiments in which the nucleoside analog 2-aminopurine (2-AP), a PKR inhibitor, blocked IFN gene expression in chick embryo fibroblasts or mouse L cells in response to Newcastle Disease virus or the dsRNA analog polyrI:polyrC (pIC), respectively. In earlier studies, 2-AP selectively blocked the induced expression of various genes in response to dsRNA or IFN (4, 5). Although the specificity of 2-AP as a PKR inhibitor has not been established, the use of this drug to identify possible PKR-dependent pathways has been productive. For example, in accord with the results above, 2-AP blocked the increased transcription of the human IFN-β gene in cells exposed to virus or pIC (6). Specificity was shown by the lack of effect of 2-AP on the expression of either the heat shock gene hsp70 stimulated by high temperature, or the metallothionein gene stimulated by cadmium or dexamethasone. Interestingly, however, 2-AP also blocked the increased expression of the c-fos and c-myc protooncogenes caused by serum growth factors or viral infection, suggesting that PKR regulates these genes in addition to the IFN genes and IFN-regulated genes. PKR contributes to the increased expression of c-fos and other immediate early genes in cells exposed to platelet-derived growth factor (PDGF) through the regulation of transcription factor STAT3 (7). Although the increased expression of c-myc caused by serum has not been studied further, the regulation of c-myc expression by IFN-γ or PDGF in fibroblasts is dependent on PKR-mediated phosphorylation of STAT1 on S727 (8).

PKR Mediates dsRNA Activation of NF-κB

The NF-κB link to PKR was first suggested by a series of experiments showing that dsRNA could activate NF-κB in various cell lines (9). Although the 2-AP experiments described above suggested that PKR might be involved in activating NF-κB, the cloning and expression of PKR allowed a more direct test of this hypothesis (10). Human PKR expressed as a glutathione-S-transferase (GST) fusion protein activated NF-κB DNA binding activity in mouse SVT2 cell extracts (11). This activation could be inhibited by subsequent incubation of the extracts with inhibitor of κB (IκB). PKR phosphorylated IκB in vitro, and this phosphorylated form did not inhibit NF-κB activation in vitro. However, subsequent work indicated that direct phosphorylation of IκB by PKR is unlikely to be a major mechanism of NF-κB activation. Catalytically inactive mutant (K296R) PKR inhibited dsRNA-NF-κB dependent reporter gene activity in transient transfection assays in mouse macrophage cells, further implicating PKR in dsRNA signaling of NF-κB (11). Results from targeted mRNA degradation experiments in HeLa cells also support a role for PKR as a mediator of NF-κB signaling by dsRNA in human cells (12). When PKR RNA was targeted for degradation by 2′-5′ oligoadenylate (2-5A)-linked antisense oligonucleotide chimeras to PKR, activation of NF-κB by dsRNA was blunted. The 2-5A-RNaseL pathway is a component of the IFN-regulated antiviral response pathway. IFN increases the expression of 2-5A synthetases, which require dsRNA for activation. This dsRNA is produced during viral infection. Once activated, the 2-5A synthetases synthesize 2-5A from adenosine triphosphate (ATP). The 2-5A activates the latent endoribonuclease RNaseL, which proceeds to degrade viral and host RNA, stopping the infection. In the 2-5A antisense chimera experiment, the usually nonspecific ribonuclease RNaseL, which is activated by 2-5A, is rendered specific by attaching antisense DNA to the 2′-5′ oligoadenylate activator, thereby directing RNaseL to targets specified by the antisense moiety. Although TNF-α-induced activation of NF-κB appeared normal, close examination of the results [Fig. 3 (12)] revealed a slight inhibition of the effect of TNF-α. Recently, specific degradation of PKR mRNA using the same 2-5A antisense approach has implicated PKR in IFN-γ activation of NF-κB (13).

Fig. 3.

Effect of TNF-α on NF-κB activation when PKR is inhibited by 2-5A PKR-directed antisense [for experimental details, see (12)]. (A) Activation of NF-κB in response to dsRNA is blocked after the depletion of PKLR by 2-5A antisense. (B) Lack of effect of the chimeric oligonucleotides on NF-κB activation in response to TNF-α. Nuclear extracts from HeLa cells treated (as indicated) were used for electrophoretic mobility shift assays. In lanes 6 through 8, 17 through 19, 0.5 ng, 1 ng, and 2.5 ng of cold PRDII DNA oligonucleotide competitor (as compared with 0.1 ng of radiolabeled PRDII probe; ~30,000 cpm) were added. In lanes 9 and 20, antibody (Ab) to the p50 subunit of NF-κB was added to nuclear extracts of pI:pC-treated cells or TNF-α-treated cells. In lane 10, the antibody to p50 was added to a nuclear extract of control cells. Arrowheads indicate the migration of the NF-κB complex. Notice that when comparing lanes 12 and 16 with lane 14, there is a slight inhibition of the effect of TNF-α. [Copyright AAAS 1994, reprinted with permission.]

Examination of the phenotype of mouse embryo fibroblasts (MEFs) with a targeted deletion in the PKR gene (PKR-/-) confirmed its role in dsRNA-stimulated activation of NF-κB (14). NF-κB reporter gene activation could be restored by coexpression of wild-type, but not mutant, forms of PKR. The defect in NF-κB activation in PKR-/- MEFs was accompanied by a defect in the production of IFN in response to pIC. However, this defect in NF-κB activation was not apparent in the mice themselves. A possible explanation for this was provided by the rescue of the defect in NF-κB activation by dsRNA by pretreatment of the PKR-/- MEFs with IFN (14) (Fig. 2). All mice housed under normal conditions have circulating IFN (15, 16), which likely rescues the PKR defect at the level of the organism. This is discussed further below (in "PKR-Independent Signaling of NF-κB") in light of recent work calling into question the role of PKR in NF-κB activation by dsRNA (17). Independent confirmation of a role for PKR in mediating signaling to NF-κB was provided by experiments using inducible expression of PKR in vaccinia virus (VV) recombinants (18, 19). Infection of HeLa cells with PKR-expressing VV recombinants resulted in activation of NF-κB DNA binding activity concomitant with translocation to the nucleus.

PKR Is Part of the IKK Complex

Further mechanistic analyses had to await the description of what has been termed the IκB kinase (IKK) signalsome (20-24). IKK is the immediate upstream effector kinase that phosphorylates critical serine residues in the IκB family of inhibitors. PKR physically interacts with the upstream effector kinase IKK (25). IKK is a large multicomponent enzyme complex consisting minimally of three components: two closely related kinase subunits, IKKα and IKKβ, which have identical structural domains and interact as a heterodimer; and a third regulatory subunit termed IKKγ (or NEMO). Although other proteins are probably components of the IKK complex, they have not been unambiguously confirmed and it seems likely that the complexes may differ in various cell types and depend on the specific stimulus. In the case of proinflammatory stimuli, including those that signal through PKR, activation of the IKK complex is dependent on IKKβ phosphorylation (23, 26). PKR appears to associate with IKK through its catalytic domain, because mutants lacking the dsRNA binding domain still form a complex with IKK (27). Careful examination of the kinetics of NF-κB activation by dsRNA (by the addition of pIC) compared to activation by TNF-α, in HEK293 cells and human T98G cells, revealed that pIC elicits a slower response than does TNF-α, but both stimulate sustained increases in NF-κB activity (25). IKK is activated in response to pIC in T98G cells with kinetics that correlate with NF-κB activation, but the activity is more prolonged than that seen with TNF-α stimulation (25). The dsRNA signal appears to be mediated through IKKβ, because cells lacking this subunit no longer respond efficiently to dsRNA or vesicular stomatitis virus (VSV) (28). In PKR-/- MEFs, IKK is not measurably activated by pIC at early time points, although some activity can be measured after four hours (25, 28). Coexpression of IKKβ with wild-type PKR in PKR-/- MEFs restores IKK activation by dsRNA (or VSV), in accord with the defect in dsRNA signaling observed in IKKβ null cells (27, 28). Also in line with these results, IκB is transiently degraded in PKR wild-type but not in PKR null cells (25).

Clearly, PKR is required for the efficient activation of NF-κB through the activation of IKK and degradation of IκB. But this begs the question of how PKR is mediating this response. The picture so far is incomplete and some of the answers remain elusive. There is good evidence that PKR can be detected as part of the IKK signalsome. However, there is some confusion as to whether the catalytic activity of PKR is required for IKK activation by dsRNA, and which subunit of the IKK complex is responsible for the interaction. A catalytically inactive mutant of PKR ectopically expressed in human 293T cells can be detected in association with the IKK complex. PKR (K296R) and IKKα mutants coexpressed from VV recombinants associate (19, 25). PKR can also associate with IKKβ in GST pull-down assays (28, 29). Wild-type PKR transiently transfected into mouse NIH3T3 fibroblasts efficiently stimulates IKK, and although catalytically inactive PKR is a poor activator of IKK at low levels of expression, higher levels can activate IKK (28). Moreover, either purified wild-type or mutant (K296R) PKR can activate recombinant IKKβ, although the dsRNA dependency for wild-type PKR activity was not tested (28). These results lead to the suggestion that PKR activates IKK by way of protein-protein interactions stimulating the autophosphorylation of IKKβ and not by direct phosphorylation.

Indirect support for a structural role for PKR comes from transfection assays in which NF-κB activation was measured by NF-κB luciferase reporter gene expression (29). Transfection of either wild-type or catalytically inactive mutant PKR can activate the reporter genes, although careful examination of the data reveals that wild-type PKR is much more active at lower levels of expression. These data are complicated to interpret because wild-type PKR will inhibit translation through eIF-2 phosphorylation. This effect is seen quite clearly in a comparison of the levels of luciferase activity between cells transfected with wild-type PKR or catalytically inactive PKR. The expression of dominant negative mutant PKR inhibits eIF-2 phosphoryation by endogenous PKR and boosts translation almost tenfold (29). The inhibitory effects of PKR on protein translation in transient reporter assays have often been used to indirectly measure the activity of PKR and PKR inhibitors (30). In the experiments reported (29), the effects of dsRNA were not measured.

A kinase-independent role of PKR in activating NF-κB has been also been claimed from experiments in which catalytic or dsRNA binding defective mutants of PKR were expressed stably in NIH3T3 cells (31). The mutant-expressing cell lines exhibited enhanced IKK and NF-κB activity. Endogenous PKR activity or dsRNA activation of NF-κB were not measured, but TNF activation of NF-κB appeared normal. In these experiments, it is difficult to rule out that the activation of NF-κB might result from transformation of NIH3T3 cells by overexpression of the different PKR mutants, rather than from a direct effect on IKK. However, the authors report that NIH3T3 cells transformed by activated Ras or by mutant eIF-2α (S51A) did not show enhanced IKK activity.

The transfection of wild-type PKR, but not equivalent amounts of mutant (K296R) PKR, rescued a defect in dsRNA signaling to the expression of a reporter construct dependent on transcription factors interferon regulatory factor 1 (IRF-1) and NF-κB (32). Interestingly, IRF-1 itself can activate NF-κB in a PKR-dependent manner (33). These experiments established not only that dsRNA signaling in fibroblasts is dependent on PKR, but also that IFN activation of specific transcription factors, such as IRF-1, requires a PKR-dependent signal. It seems reasonable to conclude that under normal conditions PKR catalytic activity, as a component of the IKK complex, is required for dsRNA signaling to activate NF-κB (and other transcription factors). Expression of excess amounts of dominant negative mutants of PKR may disturb this complex through heterodimeric interactions with wild-type PKR, resulting in perturbation of the complex and autophosphorylation of IKKβ. This interpretation is supported by experiments in which IKK and NF-κB activation were measured following infection of PKR-/- MEFs with inducible wild-type or mutant PKR VV recombinants (27). Only the wild-type PKR VV recombinants activated IKK and stimulated NF-κB activity, although both mutant and wild-type PKR could be found associated with the IKK complex. In these experiments, it is assumed that PKR is activated by viral dsRNA. It would be interesting to know whether expression of mutant PKR VV recombinants in PKR+/+ cells activates IKK. Activation would be expected if these mutants disturbed the complex by heterodimerization with wild-type PKR.

PKR Contributes to TNF-α Signaling

A model in which PKR contributes to the stability of the IKK signalsome is supported by experiments measuring TNF-α activation of NF-κB by different inducers in different cells types. There is a slight but reproducible diminution of NF-κB activation by TNF-α in HeLa cells in which PKR has been ablated using 2-5A antisense (12). An investigation of the cooperative interaction between TNF-α and IFN-γ resulting in the synergistic enhancement of gene expression has also revealed a role for PKR in neuronal cells (34). The synergism can be blocked by 2-AP or by overexpression of catalytically inactive PKR. There appears to be cell specificity to this response, because synergism between TNF-α and IFN-γ in endothelial cells is not affected by 2-AP. In PKR-/- MEFs, although the response to pIC is severely blunted (but not ablated), the TNF-α response remains intact at early time points (25). However, when the kinetics of TNF-α activation of NF-κB are examined in PKR-/- MEFs, a defect in sustained signaling compared to wild-type cells is apparent. This is revealed by measuring NF-κB activation or by following IκBβ degradation. TNF-α and IFN-γ cooperativity in activating NF-κB can also be seen in PKR+/+ MEFs but is lost in PKR-/- MEFs. IFN-γ on its own does not activate detectable NF-κB in neuronal cells, endothelial cells, or MEFs; however, in fibrosarcoma cells, IFN-γ alone can activate NF-κB (13). This effect can be blocked by pretreatment with 2′-5′ oligonucleotide-linked antisense chimeras against PKR mRNA, and is independent of activation by STAT1 but dependent on Janus kinase (JAK) activity. How a signal is transmitted from the tyrosine kinase JAK-dependent IFN-γ receptor to PKR and NF-κB remains to be determined.

PKR-Independent Signaling of NF-κB by dsRNA

In PKR-/- MEFs, there is residual activation of NF-κB by dsRNA, which becomes more apparent after prolonged exposure to dsRNA (Fig. 2). In cells derived from PKR-/ mice, the activation of NF-κB by dsRNA can be restored by pretreatment, or priming, of the cells with either IFN type I or IFN type II (32). The rescue of the PKR null phenotype by priming suggests that there must be a PKR-independent pathway involved in NF-κB activation by dsRNA (32, 35). Although the mechanisms involved have not been further examined at a molecular level, genetic support for a PKR-independent mechanism was provided by the identification of a cell line with a defect in dsRNA-induced gene expression that nevertheless had intact dsRNA-induced PKR activity and NF-κB activation (36). One such alternative pathway may be JNK activation, which may take place independent of PKR (37). However, in low-passage primary MEFs derived from PKR-/- mice, JNK activation by dsRNA is defective (38). This is not the case in late-passage PKR-/- MEFs or in derivative NIH3T3-like cell lines (17, 28). An explanation for these results is that the cells produce an autocrine factor (possibly IFN-β), which is able to rescue the JNK defect more efficiently than the NF-κB defect in dsRNA signaling. Although there is a defect in IFN-β production in PKR-/- MEFs, it is not absolute. An experiment to test this possibility has not yet been reported, but could be conducted by measuring dsRNA activation of NF-κB in cells derived from PKR/IFN type I receptor double knockouts when they become available. A recent report describes activation of NF-κB by dsRNA in the absence of PKR (37); however, the observations reported can be ascribed to the rescue phenomenon described above. Another cellular response in which PKR-dependent and independent pathways have been recently implicated is in macrophage activation by dsRNA (39).

PKR-Dependent Signaling of Stress-Activated Pathways

IFNs upregulate the expression of MyD88, an adaptor protein necessary for signal transduction by the Toll-like receptor (TLR) superfamily (40). An increase in MyD88 could sensitize cells to exposure to any agent that signals through pattern-recognition receptors. PKR can mediate JNK activation by dsRNA in primary low-passage MEFs (38). In fact, PKR can participate as a signal integrator for ligand-activated stress-activated protein kinase (SAPK) pathways, stimulating the activity of both JNK and p38 (Fig. 1). Different proinflammatory stimuli, including dsRNA, lipopolysaccharide (LPS), IFN-γ, interleukin 1-β (IL1-β), and TNF-α, require PKR for efficient activation of both p38 and JNK in primary MEFs (38). Pretreatment of cells with IFN type I to raise the levels of PKR amplifies the magnitude of p38 activation (38). Whereas the requirement for PKR is maintained in immortalized cell lines derived from MEFs, JNK activation no longer requires PKR regardless of the stress stimuli, suggesting that the pathway from mitogen-activated protein kinase kinase 4 (MKK4) to JNK becomes uncoupled from PKR during the immortalization process and passage through crisis (38). The requirement for PKR does not extend to stress stimuli that are not coupled to cell surface receptors, such as anisomycin, ultraviolet light, osmotic shock, arsenite, hydrogen peroxide, or heat shock.

Because p38 activation by ligand-activated stress pathways is preserved in immortalized fibroblasts, it has been possible to determine whether the catalytic activity of PKR was required for activation of p38. Restoration of p38 activation by different stimuli could be achieved using inducible expression of PKR, but the requirement for catalytic activity depended on the stimulus (38). For example, the catalytic activity of PKR was not necessary for the restoration of TNF-α signaling, but was required for LPS and dsRNA signaling. This suggests that PKR plays a structural role for TNF-α signaling to p38, similar to its proposed function in TNF-α signaling to NF-κB. The requirement for PKR catalytic activity for LPS and dsRNA signaling raises the possibility that dsRNA signals through a member of the Toll family of receptors (41). The requirement for PKR catalytic activity also raises the question of the substrate for PKR in this context. The residues on IKKβ phosphorylated as a result of dsRNA or TNF-α treatment or PKR overexpression are the same (28). However, the clear identity of an upstream IKK kinase(s) remains inconclusive, despite the fact that there is no shortage of candidates. PKR does not phosphorylate IKKβ directly but may, once activated by dsRNA, phosphorylate and activate an IKK kinase. The similarity in behavior of PKR in the activation of NF-κB, and dsRNA in the activation of SAPK, suggests that the component (or components) that is phosphorylated by PKR, yet to be identified in these signaling cascades, may be a MAPK kinase. The activation of both MKK3/6 and MKK4/7 in response to proinflammatory stimuli is defective in primary PKR-/- MEFs, placing PKR upstream of these MAPKKs. This observation suggests that PKR should be able to phosphorylate MAPKKs in response to dsRNA, but this has not been reported. However, it seems likely that PKR can be added to a long list of MAPKKKs that can also activate IKK.

Translational Versus Transcriptional Signaling by PKR

The translation factor eIF-2 is exquisitely sensitive to regulation by phosphorylation of its α subunit. When as little as 20% of eIF-2α is phosphorylated on serine 51, protein translation initiation may be inhibited. PKR is member of a small family of evolutionarily conserved eIF-2α kinases distinguished by the presence of a signature amino acid sequence that constitutes part of the eIF-2α binding site (42). Members of the eIF-2α kinase family, in addition to PKR, include the heme-regulated inhibitor (HRI), the endoplasmic reticulum resident kinase (PERK/PEK), and GCN2 family members (43). Members of the GCN2 family are highly conserved throughout evolution and have been identified in yeast, mammals, invertebrates, and plants. Yeast GCN2 regulates the translation of the transcriptional activator GCN4. GCN2-mediated phosphorylation of eIF-2α in response to amino acid starvation results in the translation of GCN4. Activation of GCN2 occurs through an autophosphorylation reaction resulting from binding of uncharged tRNA binding to the histidyl tRNA synthetase-like COOH-terminal domain of GCN2. HRI is a hemin binding kinase that is activated under hemin deficiency conditions by autophosphorylation. In the presence of hemin, ATP binding by HRI is decreased, inhibiting autophosphorylation and activation.

In PKR-/- cells, the regulation of protein translation by eIF-2α phosphorylation is intact (presumably through the action of one of these other eIF-2α kinase family members), but is no longer responsive to the presence of dsRNA. It has been proposed that dsRNA can activate an MKK4-JNK pathway through a translational inhibitory mechanism that is largely independent of PKR but is instead mediated by the activation of the 2-5A-dependent RNaseL-mediated cleavage of 28S rRNA (17). These conclusions were based on experiments performed on cell lines derived from RNaseL/PKR double knockout mice. As mentioned above, immortalized cell lines and primary MEFs give conflicting results with dsRNA signaling, probably because of autocrine IFN production. Moreover, whereas fibroblast lines from single RNaseL knockouts were studied, lines from PKR null mice were not included. Curiously, although JNK activation was deficient in RNaseL/PKR null cells, inhibition of protein synthesis or transcription rescued the response, suggesting the presence of a labile negative regulator of signaling to JNK. These experiments bear repeating with careful kinetic measurements of eIF-2α phosphorylation, NF-κB and activating transcription factor 2 (ATF2) activation, and 2-5A-dependent rRNA cleavage, to better establish the link to translational regulation by these pathways. Inhibition of the PKR and RNaseL pathways using alternative means, such as expression of dominant negatives or antisense inhibition, could be used in different cells to establish the universality of the phenomenon. Phenotypic rescue of the null phenotypes with wild-type and mutant constructs, as has been done with PKR-/- fibroblasts in the case of p38 signaling, would be necessary to make a link to translational control by JNK through RNaseL and PKR. Interestingly, when PERK and GCN2 are activated by the unfolded protein response and amino acid starvation, respectively, they repress translation of most mRNAs by eIF-2α phosphorylation, but selectively increase translation of ATF4, resulting in the increased expression of the downstream gene CHOP (also known as GADD153) (44). It remains possible that PKR could influence transcription by similar means, that is, by inhibiting translation of some proteins but promoting the expression of others.

Other ways by which PKR could influence mRNA accumulation include regulating the splicing of selective transcripts. For example, the human TNF-α mRNA 3′ untranslated region (UTR) harbors a cis-acting element, designated 2-APRE, that renders splicing of precursor transcripts dependent on activation of PKR (45). Whether other transcripts are subject to regulation by similar means needs to be explored, and the existence of a database of genes with similar elements in their 3′ UTRs makes this feasible. It is not known whether this form of control requires external stimuli, because the stable, 17-bp stem-loop structure in the TNF-α transcript is sufficient to activate PKR.

An Alternative Translational Control Pathway Regulated by PKR

A search for PKR substrates using a PKR mutant with reduced kinase activity in a yeast two-hybrid screen (wild-type PKR is growth inhibitory when expressed in yeast) identified a regulatory subunit of PP2A, B56α, as a novel PKR substrate (46). As the major cellular protein phosphatase, PP2A plays key roles in different processes, including the cell cycle, apoptosis, signal transduction, transcription, and translation. The association of different regulatory B subunits with the core dimer (composed of A and C subunits) of the enzyme can determine substrate specificity, catalytic activity, and subcellular localization. B56α can be efficiently phosphorylated by PKR in vitro and in vivo in response to dsRNA. When B56α is phosphorylated by PKR, the activity of PP2A is increased. B56α overexpression in cells upregulates protein synthesis, but this can be prevented by PKR. The target for this regulation appears to be the translational control protein eIF-4E, which is dephosphorylated by PP2A. Phosphorylation of eIF-4E increases its efficiency of binding to capped mRNA, aiding translation initiation. PKR-dependent phosphorylation of B56α is proposed to increase PP2A activity, resulting in decreased eIF-4E activity and reduced translation. Accordingly, PKR can regulate protein synthesis by either targeting eIF-2α or eIF-4E, through the regulation of the activity of PP2A (Fig. 1). This is dramatically illustrated in PKR-/- cells or in cells transfected with dominant negative mutants of PKR (47). However, because PP2A has such a broad range of functions, other effects of PKR regulation of this phosphatase cannot be excluded. Once the site or sites of PKR phosphorylation on B56α are mapped, mutants can be derived and tested in physiological assays. Conservation of the phosphorylation sites can also be determined for other B regulatory subunits, indicating possible targets for PKR regulation.

PKR Signaling in Apoptosis

PKR-/- fibroblasts are variably resistant to apoptosis induced by different stimuli, including dsRNA, LPS, and TNF-α (48). We have noted that there may be strain differences in the MEF response, because fibroblasts derived from isogenic C57BL6 mice do not exhibit the same sensitivity to LPS as those derived from mixed 129/sv/BL6 or isogenic 129/sv mice (49). Resistance to apoptosis triggered by TNF-α was not observed in mice from a different genetic background carrying a targeted mutation in the PKR catalytic domain (50). Nevertheless, there are many independent reports describing a proapoptotic role for PKR. VV recombinants overexpressing PKR trigger apoptosis in NIH3T3 cells; although this was originally attributed to inhibition of translation through eIF-2α phosphorylation, subsequent analysis has revealed other mechanisms, including Fas associated death domain (FADD)-mediated activation of caspase 8 (18, 51). Accordingly, inducible overexpression of PKR in NIH3T3 fibroblasts sensitizes them to apoptosis induced not only by dsRNA or TNF-α, but also by influenza virus, through activation of the FADD-caspase 8 pathway (52, 53). These cells were not more sensitive to apoptosis induction by VSV or Sindbis virus, although these do induce eIF-2α phosphorylation, suggesting that regulation of eIF-2α is not a major mechanism of PKR-dependent apoptosis. Apoptosis triggered by PKR-mediated transcriptional events have also been proposed, including upregulation of Fas and p53 (48, 54). Increased expression of Fas can sensitize MEFs to the apoptosis-inducing ligand activity of the anti-Fas monoclonal antibody Jo2, or in human promonocytic U937 cells overexpressing PKR, can sensitize the cells to TNF-α stimulated apoptosis (54-56). Recall that PKR can also stimulate the antiapoptic protein NF-κB. Thus, conflicting pathways appear to be activated by PKR. Taken together, the question of the role of PKR-dependent induction of antiapoptotic NF-κB in the context of a proapoptotic signal, such as dsRNA, is raised. This issue has recently been addressed using VV recombinants expressing PKR along with a mutant form of IκBα that cannot be phosphorylated by IKK (27). PKR-dependent NF-κB activation and apoptosis can be blocked in this system, but eIF-2α phosphorylation remains intact. Proteosome inhibitors block NF-κB activation and PKR-induced apoptosis (18). Unraveling the role of translational inhibition and NF-κB activation in a physiological signaling context will require careful examination of the dynamic features of the systems being investigated, where the balance and timing of expression of NF-κB-dependent pro- and antiapoptotic genes is crucial to the final outcome.

Modulation of PKR by Pattern Recognition Receptors

Although PKR can clearly act as a direct dsRNA receptor in vitro, whether it functions this way in vivo has not been demonstrated. As described above, dsRNA can signal through both PKR-dependent and PKR-independent pathways. The properties of dsRNA as a negatively charged polysaccharide resemble LPS and suggest that it may signal through a pattern-recognition receptor, such as a member of the TLR family analogous to the signaling pathway for CpG DNA (28) (Fig. 4). Genomic DNA of vertebrates contains a high degree of methylation at the C-5 position of the cytosine in the CpG dinucleotide (5-methylCpG). In contrast, bacterial DNA is usually not methylated on cytosine. Accordingly, mammalian cells have evolved to respond to a bacteria-specific DNA structure or pattern, hence the term "pattern recognition receptor." This was confirmed by the recent discovery that the pattern recognition receptor TLR9, and components of its signaling pathway, mediate cellular responses to CpG DNA (57). Moreover, like dsRNA, CpG DNA appears to use two independent pathways to mediate its downstream effects. The binding of CpG DNA to TLR9 activates a signaling pathway that results in the recruitment of the adaptor protein MyD88. MyD88 interacts through a death domain with the serine-threonine kinase IRAK, resulting in autophosphorylation of IRAK and stimulation of an interaction of IRAK with the adaptor protein TRAF6. This oligmerizes and activates the transforming growth factor-β-activated kinase 1 (TAK1), which in turn can phosphorylate and activate IKKβ, leading to NF-κB activation. TRAF6 can also regulate the activity of mitogen-activated or extracellular signal-regulated protein kinase kinase kinase 1 (MEKK1) through the adaptor ECSIT. This leads to activation of the MAPKs, ERK, p38, and JNK, and activation of the AP1 transcription factor. Mice with a mutation in TLR9 are not susceptible to CpG DNA-induced toxic shock, and cells derived from the mice do not respond to CpG DNA (57). In accord with the signaling pathway described above, MyD88 null mice are also insensitive to CpG DNA.

Fig. 4.

Regulation of PKR activity by pattern recognition receptors. A proposed scheme for PKR activation mediated by the TLR family. PKR can be implicated in signaling pathways activated through TLR2 (lipoprotein) and TLR4 (lipopolysaccharide), but not through TLR9 (CpG DNA). Activation of PKR by dsRNA could take place either by means of an unknown pattern recognition receptor (TLRX) or by a more direct mechanism analogous to the pathways proposed for CpG DNA. Whether the adaptor molecules, MyD88 or IRAK, are involved in PKR activation by the TLRs has yet to be shown (indicated by the dotted lines). The striped regions represent the death domains of IRAK and MyD88.

The second pathway activated by CpG DNA involves the DNA damage response protein kinase (DNA-PK) (58). In bone marrow-derived macrophages from mice lacking the catalytic subunit of this heterodimeric protein, CpG DNA no longer stimulates the production of the proinflammatory cytokines IL-6 and IL-12. DNA-PK can also phosphorylate and activate IKKβ when stimulated by CpG DNA. Although the link between TLRs and DNA-PK signaling by CpG DNA has yet to be clarified, the parallels with dsRNA signaling are obvious (Fig. 4). Because LPS can signal through PKR and LPS activates a TLR, a link to TLRs has been established (38, 48). Lipoprotein also uses a PKR-dependent signaling pathway to activate p38 activity in MEFs (59). Because CpG signaling is intact in PKR null mice, TLR9 is unlikely to be involved in dsRNA signaling (60). However, the role of other TLRs needs to be explored, including the TLR2 partner required to recognize lipoproteins.

Protein Activators of PKR

The PKR-associated protein PACT, and its mouse homolog RAX, activate PKR in the absence of dsRNA in vitro and in vivo (61, 62). Whether these molecules are mediators of cellular signaling in response to cytokines or dsRNA has not been demonstrated. Although PACT and RAX are phosphorylated and stimulate PKR activity in response to signals such as growth factor or serum deprivation, or physiochemical stress, there is no evidence that their activation can be connected to ligand-activated pathways (63, 64). The upstream kinase(s) responsible for PACT and RAX stress-induced phosphorylation has not been identified. Moreover, it cannot be excluded that, under circumstances where PACT or RAX is phosphorylated, the activation of PKR is not contributed to by endogenous Alu RNAs, which are transcribed and bind PKR in response to cellular stress (65). Recently, a modular structure has been proposed for PACT, where the PKR activation domain has been identified as separate from the PKR binding domain (66). The PKR binding domain of PACT is strongly conserved among some members of a large family of dsRNA binding proteins, and it remains possible that other members of this family also contribute to PKR regulation or downstream effects. For example, DRBP76, a PKR substrate, also shares the PKR binding domain. However, whereas cotransfection of PACT and PKR into PKR null cells activates an apoptosis-inducing signal in actinomycin D-sensitized cells, cotransfection of PKR with DRBP76 does not induce apoptosis (66). Therefore, although dsRNA binding domains are common features of PKR-interacting proteins, there is likely specificity in the PACT-PKR interactions. However, this needs to be tested under more physiological conditions. Different dsRNA binding proteins could be responsible for mediating cellular responses to other agents, such as the genotoxic stress response stimulated in response to bulky adduct DNA damage (67), or the PKR-dependent regulation of T cell function (68, 69).

Conclusion

PKR has emerged as an important signaling mediator in a wide variety of cellular responses to extracellular stimuli. This kinase is not essential in most of these pathways, but is required for the maintenance and propagation of signals necessary to sustain balanced host responses in proinflammatory and apoptotic situations. More research is needed to clarify how ligand-activated signals from extracellular receptors are relayed through PKR to downstream targets. The down-regulation of PKR catalytic activity has not been explored extensively, and the investigation of the role of protein activators or modulators is just beginning. The role of PKR as a signal transducer was expanded using knockout mice, but more effort is required to place the findings in cell culture models on a firmer physiological basis, perhaps by selective gene targeting. Investigation of the signaling role of PKR in disease processes other than virus infection promises to be a productive venture.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
View Abstract

Navigate This Article