Recent studies have identified the domain structure of RLRs and the various adaptor proteins regulating RLR-mediated antiviral signaling cascades 
, but how RLRs encounter viral RNA in infected cells remain unclear. In this study, we found that all RLRs are recruited into cytoplasmic granules, termed avSGs, upon viral infections. avSGs contain many SG markers, G3BP, TIAR, and eIF3, but unlike canonical SGs, also contained viral RNA and viral NP. We demonstrated that avSGs are critical to virus-induced IFN gene activation. Since RLRs must efficiently find their ligands to act as vital sensors for viral RNA, avSGs may facilitate a proper encounter between viral RNA and RLRs. In addition, OAS and RNase L are recruited to avSGs, supporting the model that RNase L amplifies IFN-inducing signaling by unearthing cryptic ligands for RIG-I and MDA5 
. Furthermore, the specific recruitment of antiviral proteins in avSGs suggests a critical role in the blocking of viral replication without an effect on host normal translation. We also observed that other viruses including SINV, EMCV Adenovirus, Hepatitis C virus and Newcastle disease virus induced avSG (Figure S5C
and data not shown), suggesting that avSGs may function as a general platform for detection of many viruses to initiate antiviral signaling. Because PKR alone cannot trigger IFN gene activation (), PKR contributes at upstream of IAV-induced RIG-I activation. It was reported that SINV activates GCN2 to phosphorylate eIF2á 
, suggesting that the several viruses may activate different eIF2á kinases to form avSG.
Our model for the function of avSGs is summarized in . We do not strictly rule out a possibility that PKR may directly phosphorylate target molecules to participate IFN gene activation, however because activation of PKR alone is not sufficient to trigger IFN production, this effect may be incremental.
avSGs and innate antiviral responses.
Recognition of viral RNA by RIG-I and MDA5 induces ATP-dependent conformational change of the molecules and allows them to interact with mitochondrial IPS-1 via CARD-CARD interaction. Several reports indicated that activated RIG-I forms dimer or oligomer, which is required for efficient signal activation 
. Furthermore, it was demonstrated that IPS-1 is redistributed on mitochondria in response to viral infection and IPS-1 forms prion-like aggregates 
. These observations suggest a possibility that local enrichment of both RLRs and IPS-1 is required for signaling. Although our attempt to detect biochemical interaction between RLR-containing avSG and IPS-1 aggregates in vitro
has been unsuccessful so far because of insoluble property of them, our immunohistochemical analysis strongly indicates that IPS-1-enriched mitochondria are physically attached with avSG in response to IAVÄNS1 infection (4), suggesting critical role of avSG as a platform for RLR-IPS-1 interaction.
The phosphorylation of eIF2á at Ser51 is known to trigger the formation of canonical SGs, however the precise mechanism by which the phosphorylated eIF2á recruits other SG components is not well understood because of difficulty of biochemical analysis 
. We demonstrated that PKR was activated, recruited to avSGs, and essential for avSGs to form after IAVÄNS1 infection. Viral RNA is primarily responsible for triggering the PKR activation because transfection of IAV genomic RNA or poly I:C induced avSG formation and IFN gene activation in a PKR-dependent manner. Consistent with our data, some studies revealed that PKR deficiency impair production of IFN in response to polyI:C and viral infections 
. Schulz et al. reported that PKR is not required for production of IFN-á/â proteins in response to IAV in bone marrow-derived dendritic cells (BM-DCs) 
. This is apparently inconsistent with our data obtained with MEFs. Although we were unable to directly compare between MEF and BM-DC, our explanation for this discrepancy is cell type difference, because IFN protein expression as determined by ELISA () is consistent with mRNA level () and PKR knockdown with HeLa cells dramatically diminished IFN production (Figure S6B
). BM-DC may utilize eIF2á kinase other than PKR to form avSG.
The NS1 of IAV is a multifunctional protein that inhibits various host factors, including PKR 
, RIG-I 
, and tripartite motif-containing protein 25 (TRIM25), known to regulate RIG-I activation 
, and potentially sequesters viral dsRNA through its dsRNA-binding domain. Here, we demonstrated that NS1 of IAV markedly inhibited avSGs and the IFN gene’s activation. Consistently, a recent report demonstrated that formation of IAV-induced SG was inhibited by wild type NS1, but not by mutant NS1, in which Arg38 and Arg41 are substituted to Ala 
. Several classes of viruses are known to inhibit the formation of SGs during infection. The West Nile and Sendai viruses encode RNA that interacts with TIAR and inhibits SG assembly 
. The poliovirus 3C protease cleaves G3BP at Gln 326 during the infection process 
. Moreover, Simpon-Holley et al. recently demonstrated that the E3L protein of Vaccinia virus preventing PKR activation and phosphorylation eIF2á and the Vaccinia virus ÄE3L, lacking E3L genes, generated a granular-like structure distinguished from a granule termed antiviral granule (AVG) 
. They also reported that MEFs lacking the AVG component TIA-1 exhibited increased Vaccinia viral replication, suggesting that AVG is critical for antiviral host responses. These results strongly suggest that viruses acquire means to inhibit the formation of avSGs and subsequent activation of the IFN gene.
The canonical SG has been proposed as a storage compartment for translation-stalled host mRNA in response to various stresses and possible partner with another RNP complex, processing body, which is responsible for mRNA degradation 
. Therefore, SG has been considered as a compartment for dynamic translational regulation upon environmental stress. Our findings discover a new role for newly identified avSG as a platform for interaction between viral RNA and host antiviral molecules to trigger a cascade of events leading to eradication of the virus. Although the difference between SG and avSG is not fully understood at this point, future research will delineate mechanism of their assembly and biological functions in stresses and immune responses.