MDA5 Localizes to Stress Granules, but This Localization Is Not Required for the Induction of Type I Interferon

Abstract

Virus infection can initiate a type I interferon (IFN-α/β) response via activation of the cytosolic RNA sensors retinoic acid-inducible gene-I (RIG-I) and melanoma differentiation-associated gene 5 (MDA5). Furthermore, it can activate kinases that phosphorylate eukaryotic translation initiation factor 2α (eIF2α), which leads to inhibition of (viral) protein translation and formation of stress granules (SG). Most viruses have evolved mechanisms to suppress these cellular responses. Here, we show that a mutant mengovirus expressing an inactive leader (L) protein, which we have previously shown to be unable to suppress IFN-α/β, triggered SG formation in a protein kinase R (PKR)-dependent manner. Furthermore, we show that infection of cells that are defective in SG formation yielded higher viral RNA levels, suggesting that SG formation acts as an antiviral defense mechanism. Since the induction of both IFN-α/β and SG is suppressed by mengovirus L, we set out to investigate a potential link between these pathways. We observed that MDA5, the intracellular RNA sensor that recognizes picornaviruses, localized to SG. However, activation of the MDA5 signaling pathway did not trigger and was not required for SG formation. Moreover, cells that were unable to form SG—by protein kinase R (PKR) depletion, using cells expressing a nonphosphorylatable eIF2α protein, or by drug treatment that inhibits SG formation—displayed a normal IFN-α/β response. Thus, although MDA5 localizes to SG, this localization seems to be dispensable for induction of the IFN-α/β pathway.

INTRODUCTION

Every nucleated cell in our bodies is equipped with a number of complex systems to guard against invading pathogens. The initial step of this protection is the recognition of the invaders by specialized sensors, the so-called pattern recognition receptors (PRRs). These specialized sensors detect certain pathogen-associated molecular patterns (PAMPs) that are “non-self” to the cell. Recognition of viral PAMPs by PRRs activates downstream signaling pathways and the production of effector proteins to combat viral infection.

The RIG-I-like receptors (RLRs) are a group of cytoplasmic PRRs that belong to the DExD/H-box RNA helicase family and recognize non-self RNA motifs. This RLR family encompasses retinoic acid-inducible gene-I (RIG-I), melanoma differentiation-associated gene 5 (MDA5), and laboratory of genetics and physiology 2 (LGP2). RIG-I recognizes RNA containing 5′-triphosphate (1) as well as relatively small (<2.0-kb) double-stranded RNA (dsRNA) or base-paired RNA molecules (2, 3). MDA5 recognizes long (>2.0-kb) dsRNA by a mechanism that is still poorly understood (4, 5). Recognition of these PAMPs by RIG-I or MDA5 leads to ubiquitin-induced oligomerization (6) and the interaction with and subsequent aggregation of mitochondrial antiviral signaling protein (MAVS) on mitochondria (7). MAVS acts as a signaling hub that results in activation of the IκB kinase epsilon (IKK-ε) and TANK-binding kinase 1 (TBK1) complex as well as the IκB kinase beta (IKK-β) complex. These kinase complexes phosphorylate transcription factors IRF3 and NF-κB, respectively, resulting in the transcription of type 1 interferon (IFN-α/β) genes and other proinflammatory cytokines (8). The production and secretion of IFN-α/β play a key role in the implementation of an antiviral state that restricts virus replication in infected cells as well as in neighboring cells.

Another cellular defense mechanism that limits virus replication is the stress response pathway (for two excellent reviews, see references 9 and 10). Cells react to several types of stress by phosphorylating eukaryotic translation initiation factor 2α (eIF2α) at serine 51, thereby rendering eIF2α inactive and halting cap-dependent translation (11). The stalled translation preinitiation mRNA complexes—together with aggregated prion-like T-cell-restricted intracellular antigen 1 (TIA1), TIA1-related protein (TIAR), Ras-GAP SH3 domain binding protein (G3BP), and several other proteins—form the cytoplasmic stress granules (SG) (12). Four kinases are known to phosphorylate eIF2α upon encountering different forms of cellular stress. Heme-regulated eIF2α kinase (HRI) is predominantly expressed in erythroid cells and is activated when heme concentrations decline (13). General control nonrepressed 2 (GCN2) is a ubiquitously expressed kinase that halts protein translation in amino acid-starved cells (14). Cytosolic protein kinase R (PKR) and PKR-like endoplasmic reticulum (ER)-localized eIF2α kinase (PERK) phosphorylate eIF2α upon recognition of non-self RNA (15, 16) and under conditions of ER stress (17), respectively. The latter two kinases are frequently activated during virus infection. Vaccinia virus, orthoreovirus, respiratory syncytial virus, rotavirus, murine cytomegalovirus, and reovirus all activate a cellular stress response via PKR, while several coronaviruses, vesicular stomatitis virus, Epstein-Barr virus, and human cytomegalovirus activate PERK (9, 10). In cells infected with Sindbis virus, SG are formed in a GCN2-dependent manner (18).

For some viruses, it has been reported that SG induction is associated with increased virus replication (19, 20). Recently, the group of Bartenschlager showed that hepatitis C virus induces a dynamic assembly/disassembly of SG, which correlated with the PKR-mediated phosphorylation and protein phosphatase 1-mediated dephosphorylation of eIF2α (21). This oscillation prevents cell death caused by prolonged translational shutoff and thereby allows chronic infection of cells. In most cases, however, the formation of SG has a negative effect on virus fitness (10). Several mechanisms have been proposed to explain how SG formation limits virus replication. Induction of the stress pathway results in the inhibition of cap-dependent translation and thereby also viral protein synthesis. Additionally, viral mRNA transcripts that are translated in a cap-independent manner can be constrained in these granules and as a result can exclude them from translation (22). Apart from viral RNA, cellular factors essential for viral RNA translation and replication can also be trapped in SG (23, 24). Therefore, many viruses have evolved mechanisms to counteract SG formation.

The family Picornaviridae consists of a large number of small RNA viruses. They possess a single-stranded genome of positive polarity with a length of 7.5 to 8.5 kb. The viral genome has a single open reading frame that codes for a large polyprotein, which is processed by viral proteases into the structural and nonstructural proteins. During viral RNA replication, a fully complementary dsRNA product is synthesized that has recently been identified as the ligand that activates MDA5 (25). Members of the genera Enterovirus, represented by important human pathogens such as poliovirus, coxsackieviruses (CV), and rhinoviruses, circumvent this induction of IFN-α/β by degrading MDA5 via the proteasome degradation pathway and by cleaving downstream signaling proteins by viral proteases 2A^pro and 3C^pro (26). Additionally, poliovirus actively reverses the formation of SG by cleaving the essential component G3BP (27).

Viruses that belong to the Cardiovirus genus, such as Theiler's murine encephalomyelitis virus (TMEV), encephalomyocarditis virus (EMCV), and the recently identified human-tropic Saffold virus (SAFV), also efficiently suppress IFN-α/β induction (28–31). The cardiovirus leader (L) protein plays a major role in antagonizing the IFN-α/β induction. In addition, L modulates nucleocytoplasmic trafficking and suppresses apoptosis (32). Mutations in L equally affect these different activities, suggesting that they are linked. However, the mechanism of action of L remains to be established.

Interestingly, a recent report showed that TMEV L also represses the formation of SG (23). Strikingly, nonstructural protein 1 (NS1) of influenza A virus, a well-known IFN-α/β antagonist (33), was also recently shown to inhibit SG formation (34). The observation that two independent viral IFN-α/β antagonists also block formation of SG suggests that a link between these two antiviral pathways may exist.

Here, we present a comprehensive analysis of SG induction upon infection of cells with mengovirus, a strain of EMCV. We show that a mutant mengovirus with a compromised L, but not wild-type (wt) virus, induces SG formation in a PKR-dependent manner. Moreover, we demonstrate that MDA5 is recruited to SG in cells infected with mutant mengovirus as well as by other stress stimuli. The importance of MDA5 localization to these SG for induction of IFN-α/β is investigated in detail.

MATERIALS AND METHODS

Chemical inhibitors and RNA ligands.

Emetine, cycloheximide, and puromycin were purchased at Sigma-Aldrich and used at a final concentration of 10 μg/ml (emetine and cycloheximide) and 20 μg/ml (puromycin). The PKR inhibitor (PKRi) was purchased at Merck-Millipore and used at a final concentration of 10 μM. Poly(I·C) was purchased from GE Healthcare. Triphosphate-containing RNA (pppRNA) was produced by runoff RNA transcription using a T7 RiboMAX kit (Promega) from a 250-bp PCR product template encoding the 5′ end of the coxsackievirus B3 (CVB3) genome containing a T7 promoter sequence.

Cells and viruses.

RIG-I wild-type (RIG-I^wt), RIG-I knockout (RIG-I^KO), MDA5^wt, and MDA5^KO mouse embryonic fibroblasts (MEFs) were provided by S. Akira (2). The MAVS^wt and MAVS^KO MEFs were provided by Z. J. Chen (35). PKR^wt and PKR^KO MEFs were provided by J. Bell (36) through T. Michiels. The PERK^wt, PERK^KO, GCN2^wt, and GCN2^KO MEFs were provided by D. Ron (37, 38), and eIF2α S⁵¹S and eIF2α S⁵¹A MEFs were provided by R. J. Kaufman (39) through C. A. de Haan. MEFs and HeLa, BGM, and BHK-21 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS) and ciproxin (1 mg/ml). Mengovirus and a mengovirus Zn-finger domain mutant (mengo-Zn) (28) were propagated on BHK-21 cells. Coxsackivirus B3 strain Nancy (40) was propagated on BGM cells.

Plasmids.

The expression plasmid encoding the RIG-I caspase recruitment domain (CARD) was described previously (41). Plasmid encoding green fluorescent protein (GFP)-tagged MAVS was constructed by PCR amplification of the GFP gene using primers flanked by NheI (Fw; 5′-GCTAGCGCCGCCACCATGGTGAGCAAGG-3′) and NotI (Rv; 5′-GCGGCCGCCCTTGTACAGCTCGTCCATG-3′) restriction sites (underlined). The PCR product was cloned into NheI-NotI-digested pcDNA4/V5-His-A vector, resulting in the pcDNA-GFP vector. The gene encoding the human MAVS was PCR amplified using primers flanked by NotI (Fw 5′-GCGGCCGCATGCCGTTTGCTGAAGACAAG-3′) and XhoI (Rv 5′-CTCGAGCTAGTGCAGACGCCGCCGGTAC-3′) restriction sites and cloned into the NotI-XhoI-digested pcDNA-GFP vector, yielding the pcDNA-GFP-MAVS expression plasmid.

Single-step growth curve.

Confluent monolayers of HeLa cells were infected with the different picornaviruses (multiplicity of infection [MOI] = 10), and RNA was harvested at 2-h intervals until 10 h postinfection (h.p.i.). Total RNA was isolated using a GenElute mammalian total RNA miniprep kit (Sigma-Aldrich) according to the manufacturer's instructions. Isolated RNA was used to determine IFN-β and viral RNA levels using real-time quantitative reverse transcription-PCR (RT-qPCR). Cells grown on glass coverslips were infected simultaneously, fixed at the indicated time points, and used for an immunofluorescence assay.

RT-qPCR.

Total cellular RNA from confluent monolayers of a 24-well cluster (∼2 × 10⁵ cells) was isolated using a GenElute mammalian total RNA miniprep kit (Sigma-Aldrich) according to the manufacturer's instructions. Isolated RNA was DNase I treated (Invitrogen) prior to reverse transcription using a TaqMan reverse transcription reagent kit (Applied Biosystems) with random hexamer primers according to the manufacturer's instructions. Quantitative analysis of mRNA levels was performed using a LightCycler 480 system (Roche).

Immunofluorescence assay.

Cells on glass coverslips were washed once with phosphate-buffered saline (PBS) and fixed with paraformaldehyde (4%)–PBS for 15 min. Cells were permeabilized with PBS–0.2% Triton X-100, washed trice with washing buffer (PBS–0.1% Tween 20), and incubated with blocking buffer (PBS–0.1% Tween 20–2% bovine serum albumin) for 1 h. Cell monolayers were incubated for 1 h with primary antibody rabbit-α-MDA5 (Barral et al. [42]) (1:200), goat-α-MDA5 (Imgenex) (1:25), mouse-α-G3BP (BD) (1:1,000), rabbit-α-TIA1 (Santa-Cruz) (1:50), mouse-α-dsRNA (J2) (English & Scientific Consulting) (1:1,000), goat-α-eIF3 (Santa-Cruz) (1:100), rabbit-α-Sam68 (Santa-Cruz) (1:100), mouse-α-PKR (BD) (1:100), or rabbit-α-Flag (Sigma) (1:200) and then for 30 min with goat-α-rabbit-Alexa 488 (Invitrogen) (1:100), goat-α-mouse-Alexa 594 (Invitrogen) (1:100), or donkey-α-goat-Alexa 594 (Invitrogen) (1:100) and Hoechst-33258 (1:2,000) diluted in blocking buffer. Between and after the incubations, the cell monolayers were washed, thrice each time, with washing buffer. Finally, the cells were washed once with distilled water and coverslips were mounted on glass slides in Mowiol (Polysciences). Cells were examined by standard fluorescence microscopy (Leica DMR) or confocal microscopy (Leica SPE-II).

Stress induction.

HeLa and MEF cells were grown on coverslips in 24-well clusters, and confluent monolayers were treated with 0.5 mM arsenic acid for 30 min (oxidative stress), exposed to heat shock by incubation for 30 min in a water bath of 46°C (ER stress), treated with 0.1 mM MG132 for 4 h at 37°C (amino acid deprivation), or treated with 2 μM thapsigargin for 1 h at 37°C (ER stress). Formation of stress granules was determined using an immunofluorescence assay.

siRNA knockdown.

PKR mRNA knockdown in HeLa cells was performed by reverse transfection of 20 pmol small interfering RNA (siRNA) duplex per well (5′-GCAGGGAGUAGUACUUAAAUA[dT][dT]-3′ and 5′-UAUUUAAGUACUACUCCCUGC[dT][dT]-3′, where the last two nucleotides in each sequence are deoxyribonucleotides [dT]) using Lipofectamine RNAiMAX and Opti-MEM (Invitrogen) in a 24-well cluster. Briefly, PKR siRNA duplex or scrambled siRNA (Qiagen) was diluted in 100 μl Opti-MEM, 1 μl Lipofectamine RNAiMAX was added, and the reaction mixture was incubated 25 min in a 24-well cluster. HeLa cells were diluted in FCS-supplemented medium (without antibiotics), 5 × 10⁴ cells in 0.5 ml medium was added per well, and the reaction mixture was incubated 48 h at 37°C. Following the siRNA transfection, cells were used for RNA transfection or mengo-Zn infection. Knockdown efficiency was determined by RT-qPCR, immunofluorescence assay, and Western blot analysis.

Western blot analysis.

For Western blot analysis of PKR, poly(ADP-ribose) polymerase (PARP), and LC3 expression, HeLa cells transfected with either scrambled siRNAs or siRNAs directed against PKR were suspended in ice-cold cell lysis buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 0.5% Triton X-100, 1 mM EDTA) and incubated 30 min on ice. Cell debris was pelleted for 15 min at 15,000 × g, and the protein concentration of the supernatant was determined by a Bradford protein assay (Bio-Rad) according to the manufacturer's instructions. Proteins (50 μg cell lysate) were separated using reducing sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes by semidry electrophoretic transfer. Membranes were washed once with washing buffer (PBS–0.1% Tween 20) and incubated 1 h in blocking buffer (PBS–0.1% Tween 20–5% nonfat milk). Membranes were successively incubated for 1 h with primary antibody mouse-α-PKR (BD) (1:1,000), rabbit-α-PARP (Roche) (1:5,000), rabbit-α-LC3 (Novus Biologicals) (1:3,000), or mouse-α-actin (Sigma-Aldrich) (1:20,000) and then for 30 min with goat-α-mouse-IRDye680 (Li-COR) (1:15,000) or goat-α-rabbit-IRDye800 (Li-COR) (1:15,000) diluted in blocking buffer. Between and after the incubations, the membranes were washed, thrice each time, with washing buffer. Finally, membranes were washed once with PBS and scanned using an Odyssey Imager (Li-COR).

TCID₅₀ assay.

Infected HeLa and MEF cells were freeze-thawed three times. Cells were pelleted using high-speed centrifugation, and supernatants were used for endpoint dilution. HeLa cells in 96-well clusters were infected with 3-fold serial dilutions of the cleared supernatants, and 50% tissue culture infective dose (TCID₅₀) values were calculated 2 days after infection.

PKRi treatment of HeLa and MEF cells.

HeLa and MEF cells were grown in 24-well clusters and infected with mengo-Zn (MOI = 10) for 1 h. Following infection, medium was replaced by 0.5 ml fresh medium or medium supplemented with 10 μM PKR inhibitor (PKRi). Total RNA was isolated 6 h postinfection and used to determine intracellular viral RNA levels by RT-qPCR. Cells grown on glass coverslips were simultaneously infected and PKRi treated and used for an immunofluorescence assay.

Recombinant IFN treatment.

HeLa and MEF cells were grown on coverslips and either mock treated or incubated with recombinant human IFN-α2 (Roferon-A; Roche) (500 U/ml) and mouse IFN-β (Sigma) (500 U/ml), respectively, for 24 h. Cells were paraformaldehyde fixed and used for an immunofluorescence assay. Simultaneously, RNA was isolated and used to determine PKR and MDA5 levels by RT-qPCR.

RIG-I CARD and GFP-MAVS overexpression.

Activation of the RLR pathway was achieved by overexpression of the Flag-tagged CARD of RIG-I or GFP-tagged MAVS. Confluent monolayers of HeLa cells were transfected with 800 ng plasmid DNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Briefly, plasmid DNA was diluted in 50 μl Opti-MEM, 2 μl Lipofectamine 2000–50 μl Opti-MEM was added, and the reaction mixture was incubated 25 min and added to confluent monolayers grown in 24-well clusters. At 24 h posttransfection, the total RNA was isolated and used to determine IFN-β levels by RT-qPCR. Cells grown on glass coverslips were simultaneously transfected and used for an immunofluorescence assay.

RNA ligand transfection.

For transfection of MDA5 and RIG-I ligands, cells were grown in 24-well clusters in 0.4 ml medium. Confluent monolayers were transfected with either 100 ng pppRNA or poly(I·C) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Briefly, RNA ligand was diluted in 50 μl Opti-MEM, 1 μl Lipofectamine 2000–50 μl Opti-MEM was added, and the reaction mixture was incubated 25 min and added to the cells. Total RNA was isolated and used to determine IFN-β levels by RT-qPCR. Cells grown on glass coverslips were simultaneously transfected and used for an immunofluorescence assay.

Drug treatment.

HeLa cells were grown in 24-well clusters. Prior to RNA ligand transfection, the medium was replaced by 0.4 medium supplemented with 12.5 μg/ml emetine or cycloheximide or 25 μg/ml puromycin. Cells were mock, pppRNA, or poly(I·C) transfected as described before. Total RNA was isolated and used to determine IFN-β levels by RT-qPCR. Cells grown on glass coverslips were simultaneously transfected and used for an immunofluorescence assay.

RESULTS

SG formation in picornavirus-infected cells.

It was recently shown that TMEV L represses the formation of SG and also that the L proteins of mengovirus and SAFV are endowed with this function when introduced in TMEV in place of its L (23). Consistent with this observation, we found that a mutant mengovirus in which the Zn-finger domain of L is disrupted (mengo-Zn) (28) induced clear cytoplasmic aggregates of G3BP, a hallmark of SG (Fig. 1A). In contrast, wild-type mengovirus (mengo-wt) failed to induce SG. This was not due to differences in replication kinetics, as indicated by similar levels of dsRNA (Fig. 1B) and intracellular viral RNA (Fig. 1E). SG formation by mengo-Zn differed from that by coxsackievirus B3 (CVB3), an enterovirus. This enterovirus induced small G3BP-containing granules early in infection which gradually decreased in size later in infection (Fig. 1A). A similar observation has been made in cells infected with poliovirus, another enterovirus (27, 43).

Fig 1.

Induction of the IFN-α/β pathway and SG formation by picornavirus infection. (A and B) Immunofluorescence images of HeLa cells infected with coxsackievirus B3 (CVB3), mengovirus (mengo-wt), and mengovirus with a Zn-finger domain mutation in L (mengo-Zn) (MOI = 10). Cells were fixed at indicated time points and for G3BP (A) or dsRNA (B). Nuclei were stained using Hoechst-33258. (C) Immunofluorescence images of either mock-treated or mengo-Zn-infected (MOI = 10) HeLa cells. Cells were fixed 6 h.p.i. and stained using an antibody against TIA1. Nuclei were stained with Hoechst-33258. (D) Immunofluorescence images of mock-treated or mengo-Zn-infected (MOI = 10) MEF cells. Cells were stained using antibodies against G3BP (red) and SAM68 (green). Nuclei were stained with Hoechst-33258. TIA1 and G3BP form cytoplasmic aggregates upon mengo-Zn infection, while Sam68 maintains its nuclear localization. (E and F) In the same experiment as described for panel A, total RNA from infected cells was isolated and used for RT-qPCR analysis of intracellular viral RNA levels (E) and induction of IFN-β mRNA (F). Data are presented as the means ± standard deviations (SD) of the results of triplicate experiments.

It has been proposed that virus infection leads to the formation of a unique SG—marked by presence of Src-associated protein in mitosis of 68kDa (Sam68) (22, 44)—that is distinct from the SG induced by other stimuli. Nevertheless, although mengo-Zn clearly induced granular localization of G3BP (Fig. 1A) and TIA1 (Fig. 1C), we were unable to detect Sam68 in these virus-induced SG (Fig. 1D). Sam68 was also absent in SG induced by an L mutant TMEV (23), suggesting that this protein is not recruited to SG in cardiovirus-infected cells.

The kinetics of SG induction correlates flawlessly with the transcription of IFN-α/β mRNA (Fig. 1A and F). While CVB3 and mengo-wt efficiently repressed IFN-α/β induction (31, 45), mengo-Zn that lacks L's IFN-α/β antagonizing activity (28) induced high levels of IFN-β mRNA (Fig. 1F).

Mengovirus-induced SG formation is PKR dependent and represses viral RNA replication.

The induction of cellular stress by virus infection is often mediated by the activation of PERK, GCN2, and/or PKR (9, 10, 18). Activation of these kinases leads to phosphorylation of eIF2α that, in turn, induces SG formation. The identity of the kinase that is activated by mengo-Zn, or any other picornavirus, is as yet unknown. To investigate this, we used mouse embryonic fibroblasts (MEFs) with a specific disruption of the PERK, GCN2, or PKR gene (these cells are referred to here as knockout [KO] MEFs). To confirm that the KO MEFs used were behaving properly, they were subjected to stress inducers that are known to specifically activate each of these kinases. As expected, the PERK KO cells, but not the wt cells, failed to form SG upon treatment with thapsigargin (38) (Fig. 2A). Likewise, the GCN2 KO cells showed no SG induction upon MG132 treatment (46) (Fig. 2B). However, both PERK and GCN2 KO cells were still able to form SG upon mengo-Zn infection (Fig. 3A). In contrast, PKR KO cells (36) failed to induce cytoplasmic granules upon mengo-Zn infection (Fig. 3A). The prerequisite of PKR activation for mengo-Zn-induced SG was confirmed in HeLa cells in which PKR expression was depleted by siRNAs (Fig. 3B and C). Of note, the absence of SG in cells lacking PKR was not due to an intrinsic defect in the stress response pathway since they were still able to form SG upon exposure to other stress stimuli (Fig. 2C and D). These data strongly suggest that mengo-Zn infection induces SG in a PKR-dependent manner.

Fig 2.

Integrity of cells lacking PERK, GCN2, or PKR expression. (A to C) Immunofluorescence images of MEFs from mice deficient in PERK (PERK^KO) (A), GCN2 (GCN2^KO) (B), or PKR (PKR^KO) (C) and their litter controls (wt) that were either mock treated or treated with MG132, thapsigargin, heat shock, or arsenic acid. Cells were stained for eIF3, and nuclei were stained with Hoechst-33258. Note that PERK^KO cells failed to form SG upon thapsigargin treatment (PERK-mediated stress induction), while GCN2^KO cells failed to form SG upon MG132 treatment (GCN2-mediated stress induction). (D) Immunofluorescence analysis of HeLa cells transfected with scrambled siRNA (scramble) or PKR siRNA (PKR^KD) and subsequently mock treated or treated with heat shock or arsenic acid. Cells were stained using an antibody against G3BP, and nuclei were stained with Hoechst-33258.

Fig 3.

Mengo-Zn induces SG in a PKR-dependent manner. (A) Immunofluorescence images of knockout (KO) MEFs and cells of their litter control (wt) infected with mengo-Zn (MOI = 10). Cells were fixed at 6 h postinfection (h.p.i.) and stained for eIF3. Nuclei were stained with Hoechst-33258. Cells containing SG were counted, and the results are shown in the graphs. While MEFs deficient in GCN2 and PERK expression showed SG formation, MEFs deficient in PKR expression were devoid of SG formation. (B) Immunofluorescence images of scrambled siRNA (scramble)- or PKR siRNA (PKR^KD)-transfected HeLa cells infected with mengo-Zn (MOI = 10). Cells were fixed at 6 h.p.i. and stained for G3BP or PKR. Nuclei were stained with Hoechst-33258. Cells containing SG were counted, and the results are shown in the graph. (C) Lysates of scrambled siRNA (scr.)- and PKR siRNA (PKR^KD)-transfected HeLa cells were made in parallel and analyzed for expression of PKR by Western blotting using an antibody against PKR. Additionally, detection of β-actin expression was used as a loading control. (D) MEFs from PKR-deficient mice (PKR^KO) and cells of their litter control (wt) and HeLa cells transfected with scrambled siRNA (scramble) or PKR siRNA (PKR^KD) were infected with mengo-Zn (MOI = 10). Total RNA was isolated at 8 h.p.i. and used for RT-qPCR to determine intracellular viral RNA levels (Viral RNA). Additionally, viral particle formation was determined using endpoint dilution and TCID₅₀ values are shown. Data are presented as the means ± SD of the results of triplicate experiments analyzed using an unpaired t test (* and **, significant differences at P < 0.05 and P < 0.01, respectively). (E) HeLa cells, PKR knockout MEFs (MEF PKR^KO), and cells of their litter control (MEF^wt) were incubated in the presence or absence of PKR inhibitor (PKRi) and subsequently infected with mengo-Zn (MOI = 10). Intracellular viral RNA levels were determined by RT-qPCR. Data are presented as the means ± SD of the results of triplicate experiments analyzed using an unpaired t test (ns, no significant difference; **, significant difference at P < 0.01). (F) In the same experiment, HeLa cells were fixed and stained for G3BP. Nuclei were stained with Hoechst-33258. Cells containing SG were counted, and results are shown in the graph.

To investigate the effect of stress pathway activation on viral RNA replication, intracellular viral RNA levels were determined in HeLa and MEF cells that lacked PKR expression and thus SG formation (Fig. 3A and B). In both cell types, mengo-Zn replicated to higher yields (2-to-3-fold increase in intracellular viral RNA level) than their litter controls (Fig. 3D). This difference in intracellular viral RNA levels reflects the difference in infectious particles as determined by endpoint dilutions (Fig. 3D). Additionally, mengo-Zn replicated to 2-fold-higher levels in HeLa and MEF cells treated with a PKR inhibitor (PKRi; Fig. 3E) that reduced SG formation (Fig. 3F). This increase was not due to an off-target effect of the drug, since viral RNA replication was unaffected in PKR KO MEFs (Fig. 3E).

Importantly, mengovirus infection induces also the autophagy and apoptosis pathways (47, 48). We investigated whether virus-induced activation of the stress pathway and the consequent formation of SG influenced activation of the apoptosis or autophagy pathway in virus-infected HeLa cells in which PKR was depleted by siRNAs. Inhibition of the stress pathway had no effect on PARP cleavage (a hallmark of apoptosis) or LC3 lipidation (a hallmark of autophagy) in either wt- or mengo-Zn-infected cells (Fig. 4).

Fig 4.

Inhibition of the stress response had no effect on the apoptosis or autophagy pathways. Lysates of scrambled siRNA (scramble)- and PKR siRNA (PKR^KD)-transfected HeLa cells infected with mengo-wt or mengo-Zn (MOI = 10) were analyzed for expression of PARP, PKR, LC3, and actin by Western blotting. PKR knockdown, which prevents SG formation, had no effect on the induction of apoptosis as detected by the degradation of full-length (FL) PARP into the smaller cleavage product (indicated by the asterisk). Additionally, lack of SG formation had no effect on the activation of the autophagy pathway as determined by the conversion of LC3 I into LC3 II.

Together, our results suggest that the PKR-mediated activation of the stress response pathway acts as an antiviral response that limits virus replication, although the impact on virus titers is relatively small.

MDA5 is recruited to SG under conditions of cellular stress.

In our studies aimed at elucidation of the potential link between the stress pathway and IFN-α/β pathway, we focused on MDA5, as this receptor plays a key role in activating the IFN-α/β response upon picornavirus infection (25, 49). MDA5 has a cytosolic localization in naive cells (50). However, the distribution of MDA5 upon virus infection is largely unknown. Therefore, we investigated the localization of MDA5 in picornavirus-infected cells using immunofluorescence microscopy. In both CVB3- and mengo-wt-infected cells, MDA5 maintained its cytosolic localization (Fig. 5A). Surprisingly, infection with mengo-Zn resulted in granular localization of MDA5 (Fig. 5A), which was confirmed by another assay using MDA5-specific antibody (data not shown). The MDA5 granules were distinct from the smaller dsRNA-containing puncta that most likely represent the replication complexes (Fig. 5A and B). Hence, we considered that MDA5 might localize to SG. Indeed, MDA5 was demonstrated to colocalize with G3BP in both mengo-Zn-infected HeLa and MEF cells (Fig. 5C) and also with TIA1 in mengo-Zn-infected HeLa cells (data not shown). Taken together, these data show that MDA5 migrates to SG upon mengo-Zn infection.

Fig 5.

MDA5 is recruited to SG upon stress induction. (A) Immunofluorescence images of mock-treated and picornavirus-infected HeLa cells (MOI = 10). Cells were fixed at 6 h.p.i. and stained with an antibody developed by Barral et al. (42) against MDA5 (and was used to stain for MDA5 unless stated otherwise) (green) and dsRNA (red). Nuclei were stained with Hoechst-33258. (B) Confocal microscopy analysis of mengo-Zn-infected HeLa cells (MOI = 10) stained for MDA5 and dsRNA as described for panel A. Single-cell magnified images clearly showed that dsRNA does not colocalize with the MDA5 granules. (C) Immunofluorescence images of mock-treated and mengo-Zn-infected HeLa and MEF cells (MOI = 10). Cells were fixed at 6 h.p.i. and stained for G3BP (red) and MDA5 (green). Nuclei were stained with Hoechst-33258. (D) Immunofluorescence images of heat shock- or arsenic acid-treated HeLa and MEF cells. Cells were fixed after treatment and stained as described for panel C.

To investigate whether MDA5 is recruited to SG only during virus-induced stress or whether it is also recruited under non-viral stress conditions, we activated the cellular stress response by the use of different stimuli. In Fig. 5D, it is shown that MDA5 also localizes to SG under conditions of ER stress induced by heat shock and oxidative stress induced by arsenic acid treatment. It is noteworthy that MDA5 was detected in all SG induced by any kind of stress, which suggests that MDA5 is a general component of SG. Importantly, neither treatment resulted in the activation of MDA5 and the downstream IFN-α/β pathway (data not shown). Therefore, we conclude that MDA5 localizes to SG upon the induction of cellular stress and that the localization is independent of its activation.

Activation of the RLR pathway does not induce and is not needed for SG formation.

The observation that SG are formed only in cells infected with a virus deficient in suppressing IFN-α/β (Fig. 1) suggested a possible link between the stress pathway and the IFN-α/β pathway. To investigate whether secreted IFN-α/β might induce a cellular stress response, HeLa and MEF cells were treated with recombinant human IFN-α2 and mouse IFN-β, respectively. Although the transcription of interferon-stimulated genes (MDA5 and PKR) was induced (data not shown), neither of the cell types showed formation of SG upon IFN treatment (Fig. 6A).

Fig 6.

Activation of the RLR pathway does not lead to SG formation. (A) Immunofluorescence images of mock- and recombinant IFN-treated HeLa and MEF cells. Cells were fixed 24 h posttreatment and stained for G3BP. Nuclei were stained with Hoechst-33258. (B) Plasmids encoding the Flag-CARD of RIG-I or GFP-MAVS were transfected in HeLa cells. Cells were fixed 24 h posttransfection and stained for Flag (green) and G3BP (red). Nuclei were stained with Hoechst-33258. (C) In the same experiment, total RNA was isolated from plasmid-transfected cells and used for RT-qPCR analysis of IFN-β mRNA levels. Data are presented as the means ± SD of the results of triplicate experiments. (D) Immunofluorescence images of MEFs from mice deficient in RIG-I (RIG-I^KO), MDA5 (MDA5^KO), or MAVS (MAVS^KO) expression and wild-type control cells (wt) were mock transfected, transfected with pppRNA ligand, or infected with mengo-Zn (MOI = 10). Cells were fixed 6 h postinfection or -transfection and stained for G3BP. Nuclei were stained with Hoechst-33258. In the same experiment, total RNA was isolated from infected and transfected MEFs and used for RT-qPCR analysis of IFN-β mRNA levels (right panels). Data are presented as the means ± SD of the results of triplicate experiments. Note that cells that lacked IFN-β mRNA induction still formed SG via activation of PKR.

Alternatively, the activation of the RLR pathway rather than downstream responses of IFN-α/β could one way or another be involved in SG formation. To test this possibility, we overexpressed the caspase recruitment domain (CARD) of RIG-I or its downstream interacting partner MAVS, both of which have been reported to trigger phosphorylation of IRF3 and transcription of IFN-β mRNA upon ectopic expression (41). Cells overexpressing RIG-I CARD or MAVS were devoid of SG (Fig. 6B), although clear induction of IFN-β mRNA was observed (Fig. 6C). Thus, activation of the RLR pathway and subsequent induction of IFN-α/β gene transcription also do not induce SG formation.

To rule out the possibility that activation of the RLR pathway is required for SG formation, we activated PKR by mengo-Zn infection and transfection of a triphosphate-containing RNA ligand (pppRNA) in MEFs lacking MDA5, RIG-I, or MAVS. SG formation was observed in all cells (Fig. 6D). This observation shows that mengovirus RNA and pppRNA efficiently induce SG in the absence of RLR pathway activation.

Collectively, these data indicate that activation of the stress pathway and formation of SG do not rely on the integrity or activation of the RLR pathway.

Localization of MDA5 to SG is not required for IFN-α/β induction.

Finally, we considered the possibility that localization of MDA5 to SG is important for ligand recognition and IFN-α/β induction. To investigate this, we measured IFN-α/β responses in cells that are unable to form SG by three different approaches.

First, we transfected pppRNA and poly(I·C)—a synthetic dsRNA ligand that activates MDA5 (49)—in HeLa and MEF cells that were deficient in PKR expression. In contrast to wild-type cells, PKR-deficient cells failed to show SG formation upon RNA ligand transfection (Fig. 7A). Yet transcription of IFN-β mRNA was efficiently induced by RNA ligand transfection in cells lacking SG formation (Fig. 7B). In HeLa cells where PKR expression was reduced by siRNA knockdown, a small reduction in IFN-β mRNA induction was observed upon mengo-Zn infection; however, this was not observed in MEFs lacking PKR expression (Fig. 7B).

Fig 7.

SG formation is not required for ligand recognition and IFN-ß mRNA induction. (A) HeLa cells transfected with scrambled siRNA (scramble) or PKR siRNA (PKR^KD) and MEFs from PKR-deficient mice (PKR^KO) and cells of their litter control (wt) were transfected with pppRNA and poly(I·C) ligands. Cells were fixed at 6 h.p.i. and analyzed by immunofluorescence. HeLa cells were stained for G3BP and MEF cells for eIF3. Nuclei were stained with Hoechst-33258. (B) In the same experiment, total RNA from transfected HeLa cells (top) and MEF cells (bottom) was isolated and used for RT-qPCR analysis of IFN-β mRNA levels. Additionally, RT-qPCR analysis of IFN-β mRNA levels from mengo-Zn-infected HeLa and MEF cells is presented. Data are presented as the means ± SD of the results of triplicate experiments analyzed using an unpaired t test (ns, no significant difference; *, significant difference at P < 0.05). (C) Immunofluorescence images of mock- and RNA ligand-transfected HeLa cells in the presence or absence of puromycin, cycloheximide, or emetine. Cells were fixed 6 h posttransfection and stained for G3BP, and nuclei were stained with Hoechst-33258. (D) In the same experiment, total RNA from pppRNA (top)- and poly(I·C) (bottom)-transfected HeLa cells was isolated and used for RT-qPCR analysis of IFN-β mRNA levels. Data are presented as the means ± SD of the results of triplicate experiments analyzed using an unpaired t test (ns, no significant difference; *, significant difference at P < 0.05). (E) Immunofluorescence images of MEFs from mice expressing a nonphosphorylatable eIF2α protein (MEF eIF2α S⁵¹A) and cells of their litter control (MEF eIF2α S⁵¹S) that were mock treated, transfected with RNA ligands, or infected with mengo-Zn (MOI = 10). Cells were fixed at 6 h postinfection or -transfection and stained for eIF3. Nuclei were stained with Hoechst-33258. (F) In the same experiment, total RNA was isolated and used for RT-qPCR analysis of IFN-β mRNA levels. Data are presented as the means ± SD of the results of triplicate experiments analyzed using an unpaired t test (ns, no significant difference).

To investigate the involvement of SG formation in cells expressing normal levels of PKR, we next treated HeLa cells with drugs that stall protein translation and in some cases also repress the assembly of SG. Puromycin is a known inhibitor of protein translation that causes disassembly of the ribosome complex, thereby making the mRNA available for the incorporation into SG (51). In contrast, emetine and cycloheximide are compounds that fix complete ribosomes on mRNA transcripts, which results also in a halt in protein translation but prevents the formation of SG (52). Accordingly, transfection of pppRNA and poly(I·C) into cells treated with puromycin resulted in SG formation, while cells treated with emetine and cycloheximide were devoid of SG (Fig. 7C). Still, transfection of RNA ligands in cells that were unable to form SG displayed induction of the IFN-α/β pathway similar to that seen with mock-treated cells (Fig. 7D).

Last, we also used MEFs expressing the nonphosphorylatable eIF2α S⁵¹A protein (39). These cells are deficient in the formation of SG upon arsenic acid and heat shock treatment (data not shown) and also during mengo-Zn infection or RNA ligand transfection (Fig. 7E). Although these cells were unable to form SG, both mengo-Zn infection and transfection of RNA ligands potently induced IFN-β mRNA transcription (Fig. 7F).

In conclusion, three lines of evidence show that SG formation, and thus MDA5 localization to these granular structures, is dispensable for triggering IFN-α/β responses.

DISCUSSION

Virus infection triggers several antiviral responses that limit virus replication. To counteract these responses, viruses express dedicated proteins that antagonize these antiviral pathways. Previously, we showed that a mutant mengovirus with a disabled L protein was unable to repress the induction of IFN-α/β (28). Here, we show that the same mutant mengovirus activates the stress response pathway via PKR, resulting in SG formation. Furthermore, we show that infection of cells that are unable to form SG resulted in an increase, albeit modest, in intracellular viral RNA levels. Previously, it was shown that SG formation slightly repressed viral RNA replication of poliovirus, another picornavirus (43). Thus, our observation lends support to the idea that SG formation acts as an intrinsic antiviral mechanism against (picorna)virus infection by repressing viral RNA replication.

Our data demonstrate that a single viral protein, L, antagonizes the induction of the IFN-α/β pathway as well as the antiviral stress pathway. Interestingly, the NS1 protein of influenza A virus, a well-known IFN antagonist, was recently also recognized to suppress the stress response pathway (34). Although these evolutionarily conserved systems are believed to act strictly autonomously, accumulating evidence suggests that some antiviral mechanisms have coevolved and are intertwined. For instance, components of the autophagy pathway seem also to be involved in dampening the induction of the IFN-α/β pathway (53, 54). Moreover, ligand binding by some Toll-like receptors (TLR) stimulates the activation of the autophagy pathway (55, 56).

The finding that single viral proteins of two unrelated viruses can repress the induction of both IFN-α/β and the stress pathways suggested a possible link between these two antiviral systems. In our pursuit for this connection, we found that MDA5, the sensor for picornavirus RNA, displayed a granular distribution in cells infected with mengo-Zn but not in cells infected with CVB3 or mengo-wt. The MDA5-containing granules colocalized with G3BP and TIA-1, suggesting that MDA5 localizes to SG in mengo-Zn-infected cells. SG are storage places for preinitiated mRNA complexes, which are formed under cellular stress conditions (12). One possible explanation of why MDA5 would relocate to SG upon viral infection is that these RNA-rich granules also serve as a ligand recognition platform for MDA5. Recently, it was found that MDA5 specifically recognizes the dsRNA replication intermediate in picornavirus-infected cells (25). Therefore, we hypothesized that the dsRNA ligand could be trapped in SG and thereby could be attracting MDA5 to these structures. However, using confocal microscopy analysis we showed that the SG were devoid of dsRNA. Previously, the single-stranded genomic RNAs of TMEV and poliovirus were also found to be absent from SG (23, 44). These data argue that it is unlikely that SG act as sites for viral RNA recognition in picornavirus-infected cells.

To better understand the physiological relevance of the MDA5 localization in SG, we determined IFN-α/β induction in cells that were deficient in SG formation. We showed that infection with mengo-Zn and transfection of non-self RNA ligands such as poly(I·C) and pppRNA (i.e., a specific RIG-I ligand) induced high levels of IFN-β mRNA in PKR KO MEFs and PKR knockdown HeLa cells. Additionally, cycloheximide and emetine treatment of HeLa cells, which represses SG formation, did not affect the IFN-β mRNA induction by pppRNA and poly(I·C) transfection. Furthermore, MEFs that are incapable of forming SG by an eIF2a S⁵¹A mutation showed levels of induction of the IFN-α/β pathway upon mengo-Zn infection and transfection of pppRNA and poly(I·C) similar to those seen with wild-type MEFs. These data strongly suggest that SG formation is dispensable for the induction of the IFN-α/β pathway. This is in agreement with data from studies by Sen et al. and Clavarino et al., who showed efficient induction of IFN-β mRNA transcription by poly(I·C) transfection in PKR KO MEFs as well as in MEFs that are incapable of forming SG due to the eIF2a S⁵¹A mutation (57, 58). Moreover, a recent report by Schulz et al. showed that EMCV infection resulted in comparable levels of IFN-β mRNA in wild-type and PKR KO MEFs (59). Collectively, these data strongly suggest that SG formation is not needed for efficient induction of the IFN-α/β pathway via RLR activation.

While this paper was in preparation, a study by Onomoto et al. reported the localization of RLRs to SG in cells infected with influenza A virus lacking the NS1 gene (IAV-ΔNS1) (60). In that paper, the authors suggested that SG fulfill an essential role in activation of the IFN-α/β pathway by functioning as sites of RNA recognition. The authors showed that IAV-ΔNS1 induced SG in a PKR-dependent manner and that the viral ssRNA colocalized with RIG-I, the known sensor of IAV, and SG markers TIAR, eIF3, and G3BP. They also demonstrated that siRNA knockdown of G3BP reduced the amount of IAV-ΔNS1-induced SG by 3-fold and reported a 5-fold reduction in IFN-β mRNA transcription. Additionally, an almost 10-fold reduction of IFN-β mRNA transcription was observed in IAV-ΔNS1-infected PKR KO MEFs compared to control cells whereas a complete lack of IFN-α/β pathway activation was observed in PKR KO MEFs upon transfection of IAV ssRNA and poly(I·C).

The differences between our data and those of Onomoto et al. might be explained by comparing the sources of the PKR KO MEFs. Onomoto et al. used MEFs derived from mice with a disruption in exons 2 and 3, including the initiating methionine, of the PKR gene (61). This results in the expression of a truncated PKR lacking the N-terminal dsRNA-binding domain region (62). In contrast, we used MEFs derived from mice with a disruption of the PKR kinase domain in exon 12 (36). This disruption results in the expression of a PKR protein lacking a part of the kinase domain and thus kinase activity (62). Importantly, apart from inducing the stress pathway, PKR is also involved in activating IκB kinase beta (IKK-β), which occurs independently of PKR's kinase activity (63, 64). Activation of IKK-β is essential for the induction of the NF-κB pathway and, thereby, transcription of IFN-α/β mRNA (65). Thus, expression of a N-terminally truncated PKR may negatively influence the induction of the IFN-α/β pathway as a consequence of the inability to activate IKK-β. Indeed, in a comparative study by Iordanov et al., a reduction in IFN-α/β mRNA transcription was observed in MEFs expressing N-terminally truncated PKR but not in MEFs expressing PKR lacking kinase activity due to disruption of the kinase domain (66). Therefore, results obtained with MEFs expressing the N-terminally truncated PKR should be evaluated with caution. Importantly, we used not only MEFs expressing PKR lacking kinase activity but also several other cell systems and approaches to prevent SG formation, all of which indicated that SG are dispensable for efficient induction of the IFN-α/β pathway.

The issue remains why MDA5 localizes to SG upon mengo-Zn infection. Our observation that MDA5 also localized to SG under nonviral stress conditions, such as heat shock and arsenic acid treatment, strongly suggests that MDA5 is recruited to SG independently of its function as a PRR. Similar to the observations concerning MDA5, other members of the DExD/H-box RNA helicase family such as RHAU, DDX1, DDX3, and DDX6 were also shown to reside in SG under stress conditions (reviewed in reference 12). Therefore, it might be that the RNA screening function of MDA5, which includes also binding to nonviral RNA (4, 5), results in its SG association under stress conditions, but this requires further investigation.