The current SARS-CoV-2 pandemic has spurred new interest in interferon signaling in response to viral pathogens. Much of what we know about the signaling molecules and associated signal transduction induced during the host cellular response to viral pathogens has been gained from research conducted from the 1990's to the present day, but certain intricacies of the mechanisms involved, still remain unclear. In a recent study by Vaughn et al. the authors examine one of the main mechanisms regulating interferon induction following viral infection, the RIG-I/MAVS/IRF3 pathway, and find that similar to PKR both DICER interacting proteins, PACT and TRBP, regulate RIG-I signaling in an opposing manner. More specifically, the reported findings demonstrate, like others, that PACT stimulates RIG-I-mediated signaling in a manner independent of PACT dsRNA-binding ability or phosphorylation at sites known to be important for PACT-dependent PKR activation. In contrast, they show for the first time that TRBP inhibits RIG-I-mediated signaling. RIG-I inhibition by TRBP did not require phosphorylation of sites shown to be important for inhibiting PKR, nor did it involve PACT or PKR, but it did require the dsRNA-binding ability of TRBP. These findings open the door to a complex co-regulation of RIG-I, PKR, MDA5, miRNA processing, and interferon induction.
A renewed interest in interferon (IFN) activation and signaling has recently been sparked by findings that severe pathological disease following SARS-CoV-2 infection is directly related to delayed synthesis, activation and signaling of these circulating innate immune factors . The interferons are broken down into Type I (IFNα/β), Type II (IFNγ or immune interferon), and the so called Type III. Type I IFNs, which were once the subject of intense study as cancer therapeutics due to their ability to interfere or slow cellular growth, are still used today as part of certain anticancer protocols . The origins of these cytokines though stem from the field of virology. Isaacs and Lindenmann first reported a soluble substance synthesized by chick chorio-allantoic membrane fragments of fertilized eggs, following exposure to heat-sacrificed influenza virus, that when used to treat naive cells prior to challenge with live virus could inhibit viral infection/production . This substance was later referred to as interferon.
A major antiviral effect of Type I IFNs is the inhibition of translation, more specifically phosphorylation of the eukaryotic initiation factor (eIF)-2 subunit α on serine 51 by the double-stranded (ds) RNA-dependent kinase PKR, encoded by the EIF2AK2 gene . PKR is induced in response to IFN along with a multitude of interferon-stimulated genes (ISGs). Among these are, retinoic acid-induced gene (RIG)-I (also known as DDX58), melanoma differentiation-associated protein (MDA)-5, adenosine deaminase acting on dsRNA (ADAR)-1, and oligoadenylate synthetase (OAS) . Mutations altering the activity or expression of many ISGs have severe pathological consequences. Interestingly, products of several of these ISGs also have the ability to stimulate interferon production. Recent findings by Vaughn et al.  have begun to consolidate what we known about the regulation of the antiviral response and interferon induction in response to dsRNA.
RIG-I, PKR and MDA5 are considered cytoplasmic pattern recognition receptors (PRR) (although PKR is also present in the nucleus) that recognize, bind to, and are activated by dsRNA and 5′-ppp-dsRNA, short dsRNA, or long dsRNA (mainly positive-strand RNA viral genomes), respectively, and mediate intercellular signaling pathways that among other things result in the induction of interferon synthesis. For many years, the mode of activating these pattern recognition receptors was thought to be simple dimerization/oligomerization of the respective PRR in the presence of its preferred RNA activator, as demonstrated in vitro (Figure 1) [7–9]. Upon activation, PKR promotes Type I IFN synthesis in a kinase-dependent manner by stimulating the degradation of inhibitor κB (IκB) and the phosphorylation and nuclear translocation of nuclear factor (NF)-κB p65 (RELA). Similarly, RIG-I and MDA5 activation involves binding of their respectively preferred RNA species by the carboxy-terminal domain (CTD), and oligomerization of its caspase activation and recruitment domain (CARDs) with those of the mitochondrial antiviral signaling protein (MAVS); thereby promoting the phosphorylation and degradation of IκB, the subsequent nuclear translocation and activation of the p50/p65 NF-κB complex; and the phosphorylation, dimerization, and nuclear translocation of interferon response factor (IRF)-3 and -7. The IRF3, IRF7 and p50/p65 NF-κB complexes then bind to the promoters of Type I IFN genes and stimulate their synthesis .
Original homodimerization model for dsRNA-mediated activation of PPRs PKR, RIG-I and MDA5.
Not surprisingly, as activation of these PRRs represents a major host antiviral response, a number of viruses encode inhibitors of these innate immune proteins; but it was not until identification of the TAR RNA-binding protein (TRBP), a 39 kDa dsRNA-binding protein comprised of two dsRNA-binding domains (dsRBDs) and a C-terminal pseudo-dsRBD, that a host encoded inhibitor of PKR was first identified . The PKR inhibiting effects of TRBP are believed to require dsRNA-binding ability (but this has never been adequately tested) as well as the dsRBDs and pseudo-dsRBD. In addition, phosphorylation of TRBP on Ser142, Ser152, Ser283 and Ser286 destabilizes TRBP homodimers and favors TRBP:PKR interaction and inhibition of PKR . While dsRNA (poly I:C) and virus associated RNA species remained the main focus of PKR activation, a number of observations demonstrating the role of non-pathogenic activation of PKR for normal homeostasis and general stress signaling, suggested the existence of a cellular PKR activator, as well . Four years after TRBP was reported to be an endogenous inhibitor of PKR, two labs simultaneously identified the first endogenous activator of PKR, PACT (and its mouse ortholog RAX), a 34 kDa dsRNA-binding protein 46% identical with TRBP at the protein level composed of two dsRBDs and a C-terminal pseudo-dsRBD [13,14]. Both groups reported that PACT-mediated activation of PKR required the dsRBDs of PACT, but activation could occur in a dsRNA-independent manner. In addition, general stresses (oxidative, growth factor removal, etc.) were reported by both groups to result in PACT phosphorylation, increased PACT:PKR association and enhanced PKR activation. Interestingly, the sites of phosphorylation reported by these groups were different; the main differing factor being the stress used. In HT1080 cells stressed by low dose actinomycin D, Ser246 and Ser287 in the C-terminal pseudo-dsRBD were required for PACT phosphorylation . In contrast, serum withdrawal in COS7, REH lymphocytic leukemia or IL3-withdrawal in BaF/3 and H7 hematopoietic progenitor cells resulted in phosphorylation of Ser18 . In both cases, phosphorylation of the identified site(s) was required for PKR activation in response to the respective stress, suggesting that differing stresses could promote PACT-dependent activation of PKR through differential phosphorylation of PACT on stress-specific sites, although the model of PKR activation in response to viral infection remained that of dsRNA-dependent homodimerization and autophosphorylation. It wasn't until Bennett et al. demonstrated that knockdown of RAX, using siRNAs, inhibited PKR activation and eIF2α S51 phosphorylation in mouse embryo fibroblasts following vesicular stomatitis virus (VSV) infection that PACT was shown to also be required for the host antiviral response .
Similar to PKR, PACT was also demonstrated by Kok et al. to associate with RIG-I via the CTD domain, promote oligomerization of RIG-I CARD domains, ATPase activity; and the subsequent activation of MAVS, phosphorylation of IRF3 and synthesis of IFNβ. PACT-dependent activation of IFNβ synthesis through RIG-I was induced by poly I:C, but was not dependent on PKR, DICER or phosphorylation of PACT on Ser246 . Like PKR, RIG-I signaling could be activated by PACT in the absence of poly I:C. Moreover, similar to PKR, siRNA-mediated knockdown of PACT also inhibited the ability of Sendai virus infection to stimulate RIG-I-mediated IFN production. Although altogether suggestive of a common mechanism of activation for PKR and RIG-I, a major piece of the puzzle was still missing. Now Vaughn et al.  demonstrate that similar to PKR, PACT and TRBP co-regulate RIG-I-mediated signaling in an opposing manner. In contrast with PACT, TRBP inhibited RIG-I-mediated IFN production in response to both poly I:C, and 5′-ppp-dsRNA. Both PACT and TRBP demonstrated the ability to regulate RIG-I signaling in a dose-dependent manner, in the absence of a RNA activator, although further analysis is needed to clarify the mode in which TRBP inhibits RIG-I-mediated IRF3 activation and nuclear localization in the presence of RIG-I ATPase activity. Phosphorylation sites in PACT and TRBP, which were shown to be important for the regulation of PKR in response to poly I:C, were not found to be important for the regulation of RIG-I signaling by these proteins . Although this does not rule-out whether PACT- or TRBP-dependent regulation of RIG-I is dependent on phosphorylation or some other post-translational modification, it does demonstrate that phosphorylation of sites in the C-terminal pseudo-dsRBD of both proteins, important for regulating PKR, are not necessary for RIG-I regulation and may represent a point of discrimination for PACT and TRBP in regulating PKR or RIG-I. Studies have also suggested that PACT regulates the activation of RIG-I and MDA5 through its interaction with LGP2 . In the presence of short dsRNA (short poly I:C), PACT does not associate with LGP2 and associates with RIG-I leading to its activation. In contrast, in the presence of long dsRNA (long poly I:C or positive-sense RNA viral infection) PACT associates with LGP2 and MDA5 facilitating MDA5 oligomerization and activation [19,20]. While TRBP may also be involved in MDA5 regulation, no clear role for TRBP has been elucidate as of yet. Komuro et al. reported that the LGP2–TRBP association enhanced LGP2–MDA5 viral RNA sensing in response to picornavirus. Additionally, Sanchez David et al. reported difficulty finding TRBP associated with LGP2, but suggested that the LGP2–TRBP complex present had no effect on RIG-I signaling. The potential for TRBP to regulate MDA5 directly in the absence of LGP2 remains a possibility.
It is not surprising that several viruses have found a way to target PACT:TRBP regulation in order to subvert the antiviral response [21–23]. In the case of HIV, viral infection was shown to actually convert PACT into a PKR inhibitor via a mechanism involving the formation of a viral-induced complex consisting of PACT, PKR, TRBP and ADAR1, a known inhibitor of PKR in stress bodies. Other than suppress the antiviral response this complex appeared to also temporally regulate viral gene and protein expression during the lytic phase . Infection with certain viruses may also demonstrate a similar effect in regards to RIG-I and MDA5. Although RIG-I and MDA5 are not known to interact directly with ADAR proteins, both PACT and TRBP can, and ADAR1-dependent editing is known to suppress RIG-I and MDA5 activation. Thus, under similar circumstances to that reported in HIV infection, PACT has the potential to be transformed into an inhibitor, by proxy.
Beyond innate immunity, both TRBP and PACT have been demonstrated to be important for normal development. Targeted disruption of TRBP resulted in runted mice that died shortly after weaning. These mice where shown to have suppressed translation of mRNA among other abnormalities . As regards PACT, Rowe et al. reported that targeted disruption of PACT exon 8, which encodes the C-terminal pseudo-dsRBD required for PKR activation, resulted in mice that for most intensive purposes were normal with the exceptions of small size, minor cranial-facial abnormalities, and abnormal inner and outer ear development resulting in impaired hearing. In contrast with MEFs following siRNA-mediated knockdown of PACT, PKR could be activated following VSV infection in MEFs established from these PACT−/− mice . On the contrary, Bennett et al. reported that targeted knockout of the entire PACT coding sequence resulted in healthy heterozygous (RAX+/−) mice, but proved to be early embryonic lethal in homozygous (RAX−/−) offspring . Other than stress and antiviral signaling, some of the findings in the knockout mouse models might also be explained by the fact that both TRBP and PACT regulate miRNA production through association with DICER, TRBP actually being part of the RISC complex . Thus loss of either TRBP or PACT would be expected to have systemic consequences on miRNA metabolism as well as the stress/innate immune response. Not surprisingly, Drosophila harboring a disruption in dRAX and humans bearing certain mutations in PACT manifest severe defects in nervous system coordination and neuromuscular function [27,29,30]. Recent publications by the Patel group have demonstrated that mutations in PACT, which are now considered the causative genetic alterations in dystonia type 16 (DYS16) patients, promote PACT:PACT and PACT:PKR interactions but reduce PACT:TRBP interactions [29–31]. The resulting consequences of this effect are enhanced PKR activation and a pro-inflammatory/stress phenotype. Whether RIG-I- or MDA5-mediated signaling is also induced has not been reported, although aberrant activation of both are associated with diverse autoimmune diseases [32,33]. Such results have begun to fine tune the way in which TRBP and PACT are thought to regulate antiviral and innate immune signaling (Figure 2).
Activation of PKR and RIG-I by dsRNA is regulated in an opposing manner by PACT and TRBP.
Still some questions remain. It is rather intriguing that similar to PKR, PACT-dependent activation of RIG-I requires the dsRBDs but not the ability to bind dsRNA. In contrast, although not tested with PKR, TRBP-dependent inhibition of RIG-I requires the dsRNA-binding ability of TRBP [6,18]. Oddly enough, PKR requires its dsRNA-binding ability for activation, as the PKR dsRNA-binding mutants demonstrate reduced activation in vivo. Likewise, mutations in the CTD of LGP2 inhibit its association with PACT and disrupt the oligomerization and activation of MDA5. Although mutations in the CTD domain affecting RIG-I (or MDA) RNA binding and/or interaction with PACT would also be expected to inhibit activation, no such studies have been reported. But if PACT activation of PKR and RIG-I requires the association of the dsRBDs but not dsRNA, why are dsRNA-binding mutants of PKR unable to be activated in the presence of PACT; one might assume a similar scenario for RIG-I. Is dsRNA binding simply a way to induce a conformational change in PKR, RIG-I and LGP2–MDA5 that is favorable for PACT association? Do dsRNA-binding mutants of these PRRs recruit PACT less efficiently? As with PKR, TRBP seems to regulate RIG-I by sequestering dsRNA. This has long been viewed as a sequestration from PKR, and in this most recent publication RIG-I; but what if TRBP sequestration of dsRNA serves the purpose of enhancing the PACT:TRBP interaction and reducing the level of free PACT available to activate PKR and RIG-I. Moreover, as PACT prefers forming homodimers over binding dsRNA while TRBP prefers binding dsRNA over homodimers, is it possible that either a stoichiometrically elevated level of RNA activator needs to be reached, as during virus infection, to compete PACT away from TRBP or stress-mediated phosphorylation of PACT is actually needed to reduce its ability to interact with TRBP as indicated by enhanced homodimerization and/or association with PKR and RIG-I, while phosphorylation of TRBP by more mitogenic factors, in contrast, results in enhanced dsRNA binding and/or association with the respective PRR and/or association with PACT? This model would fit several observations in the literature: (i) PACT-dependent activation of PKR and RIG-I is not dependent on dsRNA but is facilitated by it, while TRBP-dependent inhibition of RIG-I (and presumably PKR) requires the dsRNA-binding ability of TRBP [6,13,14]; (ii) although not observed using PACT−/− MEFs, siRNA-mediated knockdown of PACT by independent groups has been reported to inhibit PKR, RIG-I and MDA5 activation and subsequent IFN synthesis in response to virus infection [17,18,22]; (iii) mutations in PACT that influence its interaction with TRBP, enhance its association with PKR, promoting a pro-inflammatory state [30,31]. The findings reported by Vaughn et al. demonstrate the presence of a common regulatory mechanism shared by PKR and RIG-I. The next steps will be to determine if this same mechanism of regulation extends to MDA5 (some recent findings suggest that TRBP may not be involved in regulating MDA5 [19,34]) and to better define the role of dsRNA species in these complex interactions, as the simple dimerization/oligomerization model for PKR, RIG-I, and MDA5 activation no longer adequately fits. Accumulating evidence suggest that PACT and TRBP may actually regulate which PRR gets activated when the cell is presented with a specific RNA species. Thus, the potential is present for PACT and TRBP to regulate the activation of PKR, RIG-I, MDA5 and DICER and discriminate which protein(s) are active under a given condition dictated by the relative abundance of various RNA species. Results from these studies may actually begin to shed light on the link between pathogen recognition, the resulting stress response, and regulation of miRNA metabolism.
The author declares that there are no competing interests associated with this manuscript.