The Atg5–Atg12 conjugate associates with innate antiviral immune responses

Involvement of the Atg5-Mediated Autophagic Process in Vesicular Stomatitis Virus (VSV) Replication.

To examine the roles of the autophagic process in RNA virus replication, mouse embryonic fibroblasts (MEFs) derived from WT and Atg5-deficient mice [Atg5 knockout (KO)] were infected with VSV. VSV induction of autophagy was determined by visualizing LC3-GFP expressed exogenously (25). Eight hours after VSV infection, LC3-positive vacuole-like signals were prominent in WT, but not in Atg5 KO MEFs (Fig. 1A). The conversion of LC3 was also characterized after VSV infection. The amount of the modified form (LC3-II), a hallmark of the autophagic process, was increased, whereas that of the free form (LC3-I) was decreased upon VSV infection of WT MEFs. Such LC3 conversion was completely abrogated in Atg5 KO MEFs (Fig. 1B). These results confirmed that VSV infection causes autophagy through a process mediated by Atg5.

We next examined the efficiency of VSV replication in WT and Atg5 KO MEFs. As shown in Fig. 1C, the levels of viral titer in the culture supernatants were significantly lower in Atg5 KO MEFs compared with those in WT MEFs, suggesting that the autophagic machinery, at least Atg5, regulates the VSV replication. The cytopathic effect of VSV was also examined by neutral red staining, and the result showed that the number of plaque formation was lower in Atg5 KO MEFs compared with WT MEFs [supporting information (SI) Fig. 6]. As a control, the efficiency of human herpes simplex virus (HSV) replication in Atg5 KO MEFs was comparable with that in WT MEFs (Fig. 1C). Interestingly, although the type I IFN and IFN-stimulated gene (ISG) production was elicited by VSV infection, the levels of IFN-β and IP-10 mRNA were significantly higher in Atg5 KO MEFs relative to WT MEFs (Fig. 1D). In contrast, the induction of these mRNA in response to HSV infection was impaired in Atg5 KO MEFs (Fig. 1D). We also examined the activation of the master transcription regulator of the type I IFN gene, IRF-3, after viral infection. Although the nuclear transition of phosphorylated IRF-3 was induced by VSV infection, the levels of such transition were higher in Atg5 KO MEFs compared with WT MEFs (Fig. 1E). Neutralization of IFN-α and IFN-β reversed the VSV replication efficiency and the difference in viral replication between WT and Atg5 KO MEFs was abolished (Fig. 1F). These results suggest that Atg5-mediated autophagic machinery regulates VSV replication by repressing type I IFN production.

Involvement of Atg5 in isRNA-Mediated Type I IFN Production.

The yield of most viral replication is largely dependent on the efficiency of the replication machinery within the host cell and the activity of antiviral responses mediated by type I IFNs. Such type I IFN production is induced mainly by isRNA, a byproduct of RNA virus replication within the host's cell (26). To examine the involvement of Atg5 in innate antiviral immune responses, the levels of type I IFN production were compared between WT and Atg5 KO MEFs after dsRNA [poly(I:C)] stimulation. As shown in Fig. 2 A and B, Atg5 KO MEFs produced significantly higher amounts of type I IFN mRNA than the WT isogenic counterparts in response to dsRNA (> 25-fold in IFN-α and >10-fold in IFN-β at both 4 and 8 h after stimulation), which was similar to that of VSV infection. The results were consistent when culture supernatants were examined for type I IFN protein levels (SI Fig. 7). Similarly, dsRNA stimulated the higher levels of IL-6 and IP-10 production in Atg5 KO MEFs compared with WT MEFs (Fig. 2 C and D). The specificity of this hyperresponse to dsRNA stimulation was further confirmed by using Atg5 KO MEFs complemented with the atg5 gene under the control of the doxycycline-regulated expression system (m5-7 cells) (27) (SI Fig. 8). The expression levels of Atg5 in m5-7 cells were decreased by increasing concentrations of doxycycline in the culture media (SI Fig. 8A). Of note, the levels of type I IFN production from dsRNA-stimulated m5-7 cells were increased with decreasing cellular expression of Atg5 (SI Fig. 8 B and C). Taken together, these results suggest that the autophagic process mediated by Atg5 is involved in the induction of type I IFN production upon VSV infection, which seems to play a pivotal role in the mechanism underlying suppression of viral replication.

The Atg5–Atg12 Conjugate Negatively Regulates IPS-1-Mediated NF-κB and Type I IFN Promoters.

It has been established that after binding to isRNA, RIG-I or MDA5 interacts with IPS-1 and transmits intracellular signaling to activate type I IFN genes (23). To examine the functional involvement of Atg5 in the IPS-1-mediated cellular signaling, a reporter gene assay was performed. Overexpression of a constitutively active mutant of RIG-I (a COOH terminus-truncated mutant, RIG-I ΔC) resulted in activation of the NF-κB-dependent promoter and type I IFN promoters in human epithelial kidney (HEK)293 cells. This activation was suppressed by coexpression of Atg5, but not by control LacZ (data not shown), suggesting that Atg5 negatively regulates IPS-1-mediated cellular signaling (Fig. 2E). Because Atg5 molecules are conjugated with Atg12 under physiological conditions (27), Atg12 and Atg5-K130R, a mutant that is incapable of conjugation with Atg12 (28), for their effects on IPS-1-mediated signaling. Similar suppression was observed with Atg12 alone, but not with Atg5-K130R (Fig. 2E). Although the expression levels of RIG-I ΔC were comparable among samples cotransfected with different doses of the Atg expression plasmid, exogenously expressed Atg5 or Atg12, but not Atg5-K130R, was coupled with its endogenous counterpart to form the Atg5–Atg12 conjugate (Fig. 2E). Thus, these results suggest that the Atg5–Atg12 conjugate, rather than monomeric Atg5 or Atg12, is functionally involved in this suppression mechanism. A series of overexpression experiments was also performed using MDA5 ΔC, and similar suppression was observed with Atg5 or Atg12, but not with Atg5-K130R (SI Fig. 9). Interestingly, MDA5 ΔC-mediated IFN-α4 promoter activation was not suppressed by the Atg5–Atg12 conjugate, suggesting that the pathway for IFN-α4 promoter activation is distinct from that for IPS-1-dependent signaling (SI Fig. 9).

To further confirm the involvement of the Atg5–Atg12 conjugate in the suppression of IPS-1-dependent signaling, we examined the responses of Atg7 KO MEFs to isRNA stimulation. Atg7 has an E1-like activating activity in the Atg12 conjugation system; deficiency of Atg7 results in a loss of the Atg5–Atg12 conjugate but does not otherwise affect Atg5 or Atg12 molecules (29). Consistent with the results obtained from overexpression experiments described above, Atg7 KO MEFs produced higher amounts of type I IFNs in response to dsRNA when compared with the WT isogenic counterpart (SI Fig. 10). Collectively, these data suggest that the Atg5–Atg12 conjugate, but not monomeric Atg5 or Atg12, acts as a suppressor for IPS-1-mediated signaling, resulting in the down-regulation of innate antiviral immune responses.

Physical Association Between the Atg5–Atg12 Conjugate and CARD-Containing Signaling Molecules.

To further examine physical association between the Atg5–Atg12 conjugate and isRNA-mediated signaling molecules, an immunoprecipitation assay was performed. When expressed in HEK293 cells, the Atg5–Atg12 conjugate coprecipitated with RIG-I, MDA5, and IPS-1, but not with GFP control (Fig. 3A). RIG-I ΔC bound better to the Atg5–Atg12 conjugate relative to full-length RIG-I, suggesting that one or more elements present in the COOH terminus weakened RIG-I association with the Atg5–Atg12 conjugate. Monomeric Atg5 and Atg12 each also coprecipitated with RIG-I, MDA5, and IPS-1, but not with GFP, suggesting that Atg5 or Atg12 monomer is sufficient to interact with these signaling molecules (Fig. 3A). Monomeric Atg5-K130R was also coprecipitated with RIG-I and IPS-1 (SI Fig. 11). To confirm the interaction between endogenous Atg5–Atg12 conjugate and IPS-1, cell lysates prepared from HEK293 cells infected with or without VSV were subjected to immunoprecipitation analysis (Fig. 3B). Endogenous IPS-1 coprecipitated with the Atg5–Atg12 conjugate under physiological conditions, suggesting that viral infection is not necessary for interaction between the Atg5–Atg12 conjugate and IPS-1. However, Atg5–Atg12 conjugate–IPS-1 interaction is further enhanced upon VSV infection (Fig. 3B).

To examine the intracellular localization of these molecules, endogenous Atg5 and IPS-1 in HeLa cells were stained separately and analyzed by confocal microscopy. As shown in Fig. 3C, Atg5 was present diffusely within the cytoplasm, whereas most IPS-1 localized to mitochondria, which is consistent with a previous report (19) (yellow signal in Fig. 3C“Merge”). A portion of Atg5 and IPS-1 colocalized with a signal specific for mitochondria (white signal in Fig. 3C“Merge”). The intracellular association between Atg5 and IPS-1 was further confirmed by FRET assay. As a control, a FRET signal was observed when cyan fluorescent protein (CFP)–IPS-1 (donor) and yellow fluorescent protein (YFP)–Mit (encoding mitochondria localization signal) (acceptor), but not YFP-KDEL (encoding an ER localization signal), were expressed in HeLa cells, suggesting that IPS-1 localizes to mitochondria, but not to the ER (Fig. 3D). When YFP-Atg5 was coexpressed, the FRET signal was also observed from IPS-1 WT, but not from the CARD-null mutant IPS-1 T54A, suggesting that the Atg5–Atg12 conjugate physically associates with IPS-1 within a cell by targeting the CARD (Fig. 3D).

The Atg5–Atg12 Conjugate Directly Associates with IPS-1 and RIG-I Through the CARDs.

Next, we set out to identify domains in IPS-1 that are responsible for interaction with the Atg5–Atg12 conjugate by employing IPS-1 mutants (Fig. 4A). The deletion of the transmembrane domain (TMD) abrogated the interaction (Fig. 4B). IPS-1 CARD alone was sufficient for the association with the Atg5–Atg12 conjugate. Site-directed mutagenesis in the third α-helical region of the CARD by T54A substitution resulted in loss of interaction with the Atg5–Atg12 conjugate. Such an interaction was also abrogated in the presence of proline rich domain (PRD) immediately juxtaposed to the CARD (IPS-1 CARD–PRD), suggesting that the PRD blocks this interaction. Collectively, these results suggest that the IPS-1 CARD is the target of the Atg5–Atg12 conjugate and IPS-1 TMD supports this interaction by interfering with the PRD blockade of the CARD (Fig. 4B).

To confirm the direct interaction between the Atg5–Atg12 conjugate and IPS-1, a pull-down assay was performed using recombinant proteins. Because recombinant Atg5 and Atg12 are not conjugated in vitro system, we tested free Atg5 for interaction with IPS-1. Atg5 alone was sufficient for direct interaction with IPS-1 and RIG-I through the CARD, suggesting that the Atg5–Atg12 conjugate directly interacts with the CARDs of both IPS-1 and RIG-I within the cells (SI Fig. 12).

Because the CARD–CARD homotypic interactions between IPS-1 and RIG-I are essential for subsequent signaling, we examined whether the Atg5–Atg12 conjugate interferes with this interaction or not. As shown in Fig. 4C, the interaction between IPS-1 and RIG-I was enhanced by increasing the amount of the Atg5–Atg12 conjugate but did not change with control molecule expression (ARPC-2). Such evidence suggests that RIG-I CARDs may gain access to both IPS-1 CARD and the Atg5–Atg12 conjugate upon isRNA stimulation, although the Atg5–Atg12 conjugate does not dissociate RIG-I CARDs from IPS-1, but intercalates between RIG-I and IPS-I by binding directly to the CARDs of these molecules.

Cells, Viruses, and Reagents.

HEK293, HeLa, and Vero cells were purchased from American Type Culture Collection (Manassas, VA). Atg5 KO MEFs were kindly provided by N. Mizushima (Tokyo Medical and Dental University, Tokyo, Japan) and have been described in refs. 27 and 33–35. These cells were maintained in DMEM supplemented with 10% FCS and 50 μg/ml penicillin/streptomycin. VSV was provided by National Institute of Animal Health (Tokyo, Japan). Human HSV type I (KOS strain) was kindly provided by T. Suzutani (Fukushima Medical University, Fukushima, Japan). Endotoxin free poly(I:C) was obtained from Invivogen (San Diego, CA). Anti-LC3 and Atg5 antibodies were provided by N. Mizushima (25, 35, 36). Anti-APG5 (Atg5) and anti-Sp1 antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-MAVS (IPS-1) or anti-phospho-IRF-3 (Ser-396) antibody was purchased from Bethyl Laboratories (Montgomery, TX) or Cell Signaling Technology (Danvers, MA), respectively.

Expression Plasmids.

The LC3-GFP expression plasmid was provided by N. Mizushima (35). The YFP-Mit and YFP-KDEL expression plasmids were described in refs. 25 and 37. The expression plasmid for biotinylated actin-related protein complex p34 (pcDNA3.2/capTEV-NT/V5-GW/ARPC2) was purchased from Invitrogen (Carlsbad, CA). The IPS-1 expression plasmid was described in ref. 17. The truncated form of IPS-1 cDNA was generated by ligating the amino acid 1-100 and the amino acid 170-540 of hIPS-1 ORF amplified by PCR (IPS-1ΔPRD); amplifying the amino acid 1-514 by PCR (IPS-1ΔTMD); amplifying the amino acid 1-170 by PCR (IPS-1 PRD); amplifying the amino acid 1-100 by PCR (hIPS-1 CARD); ligating the amino acid 1-100 of hIPS-1 ORF and EGFP cDNA fragment amplified by PCR (IPS-1CARD-GFP); ligating the CARD and TMD (amino acid 514-540) cDNA amplified by PCR (IPS-1 CARD-TMD); or ligating the CARD, TMD, and EGFP cDNA amplified by PCR (IPS-1CARD-TMD-GFP). Atg5 and Atg12 cDNAs were kind gift from N. Mizushima. RIG-I, MDA5, RIG-I ΔC (amino acid 1-604), and MDA5 ΔC (amino acid 1-575) cDNAs were amplified by PCR using a human spleen cDNA library (TAKARA, Tokyo, Japan). The hIPS-1 CARD mutant (IPS-1 CARD-TMD T54A) and Atg5-K130R expression plasmids were generated by site-directed mutagenesis. These fragments were introduced in-frame into pFLAG-CMV4 (Sigma, St. Louis, MO), pCIneo-HA (38), pCAGGS-Flag-m1SECFP (39), or pCAG-His Venus vector (39). Sequences of the PCR products were confirmed using an ABI PRISM Genetic Analyzer (Applied Biosystems, Foster City, CA)

Fluorescence and Confocal Microscopy Analysis.

Two million WT or Atg5 KO MEFs were transfected with the LC3-GFP expression plasmid by using MEF Nucleofector Kit 1 (Amaxa, Gaithersburg, MD) according to the manufacture's protocol and 4 × 10⁵ cells were plated on a slide glass. Twenty-four hours after transfection, the cells were infected with or without VSV at multiplicity of infection (moi) = 1.0. Eight hours after infection, the cells were fixed and analyzed by a fluorescent deconvolution microscope, BIOZERO (Keyence, Tokyo, Japan). HeLa cells were treated with Mitotracker (Invitrogen) 45 min before fixation with 3% paraformaldehyde and incubated with goat anti-APG5 (Atg5) and rabbit anti-MAVS (IPS-1) antibodies followed by treatment with Alexa 488-conjugated anti-goat IgG and Alexa 405-conjugated anti-rabbit IgG antibodies (Invitrogen). Confocal microscopy analysis was performed using FV500 (Olympus, Tokyo Japan).

Virus Titration.

WT or Atg5 KO MEFs (5 × 10⁵ cells per ml) were infected with VSV at moi = 0.1 or 1.0 in the presence or absence of 20 μg/ml rabbit normal IgG (Cedarlane, Ontario, Canada) or a mixture of 1,000 units/ml of anti-mouse IFN-α and IFN-β antibodies (PBL Biomedical Laboratories, Piscataway, NJ). Twenty-four hours after infection, the cells and culture supernatants were recovered and kept at −80°C. Vero or HeLa cells (5 × 10⁵ cells per ml) were used in the plaque assay for measuring virus titer. After the addition of the supernatants, the cells were overlaid with MEM containing 1% agarose and 5% FCS and were incubated for 24 h. After staining with neutral red, the number of plaques was enumerated.

Real-Time PCR.

Real-time PCR was performed as described in ref. 40.

Luciferase Assay.

Luciferase assay was conducted as described in ref. 40.

Immunoprecipitation and Immunoblot Analysis.

Immunoprecipitation and immunoblot analysis was performed as described in ref. 38 by using anti-FLAG M2 (Sigma), anti-HA (Covance, Berkeley, CA), anti-HA-Peroxidase (3F10) (Roche Diagnostics, Indianapolis, IN), anti-Atg5, anti-MAVS (IPS-1), or anti-phospho-IRF3 (Ser-396) antibody.

FRET.

HeLa cells were transfected with various combinations of plasmid expressing CFP- or YFP-fusion proteins. Cells were imaged on an inverted microscope (Olympus IX70) with a standard 75-W xenon lamp and a 40× objective lens (Uapo/340, N.A. 1.35). Interference filters (excitation and emission filters) contained in wheels were automated using Lambda 10–2 hardware (Sutter Instruments, Novato, CA). Interference filters used in this study were as follows: donor channel excitation, 440DF20; donor channel emission, 480DF30; acceptor channel excitation, 490DF20; acceptor channel emission, 535DF25; FRET channel excitation, 440DF20; FRET channel emission, 535DF25. Emitted light was captured by a cooled CCD camera (CoolSNAP fx; Roper Scientific, Tucson, AZ). The whole system was controlled using the MetaFluor 6.0 software (Universal Imaging, Media, PA). The FRET signal is shown after following the calculation by using Image J software (National Institutes of Health, Bethesda, MD). FRET = 1 − (donor emission before bleaching)/(donor emission after bleaching).

Statistical Analysis.

Student's t test was used for statistical analysis.