The Essential, Nonredundant Roles of RIG-I and MDA5 in Detecting and Controlling West Nile Virus Infection

Mouse studies.

RIG-I^−/− mice and their wild-type littermate controls have been described previously (44, 46) and were obtained as a generous gift from S. Akira (Osaka University). MDA5^−/− mice were kindly provided by M. Colonna (Washington University, St. Louis, MO). MAVS^−/− mice were previously described (47). For production of double knockout (DKO) mice lacking both RIG-I and MDA5, RIG-I^−/− and MDA5^−/− mice were intercrossed, and the resulting DKO offspring were backcrossed into a C57BL/6 background through the F3 generation. The resulting DKO line used in this study has an approximately 94% C57BL/6 genetic background, as determined by microsatellite DNA analysis. All mice were genotyped and bred under pathogen-free conditions in the animal facility at the University of Washington. Experiments were performed with approval from the University of Washington Institutional Animal Care and Use Committee. Age-matched 6- to 12-week-old mice were inoculated subcutaneously (s.c.) in the left rear footpad with 100 PFU of WNV isolate TX 2002-HC (WNV-TX) in a 10-μl inoculum diluted in Hanks balanced salt solution (HBSS) supplemented with 1% heat-inactivated fetal bovine serum (FBS). Mice were monitored daily for morbidity and mortality.

Cells and viruses.

Working stocks of WNV-TX were generated by propagation on Vero E6 cells, and titers were determined by standard plaque assay on BHK-21 cells as previously described (14). This virus preparation and methods differ slightly from those used for WNV-NY in the study described in the accompanying paper by Lazear et al. (48). WNV infections were performed by incubating virus at the indicated multiplicity of infection (MOI) with cells in serum-free medium for 1 h, followed by removal of the virus and addition of Dulbecco modified Eagle medium (DMEM) supplemented with 10% FBS, 1 mM sodium pyruvate, 2 mM l-glutamine, 1× HEPES (pH 7.3), antibiotic-antimycotic solution, and 1× nonessential amino acids (complete DMEM). Primary mouse embryonic fibroblasts (MEFs) were generated from RIG-I^−/−, MDA5^−/−, MAVS^−/−, RIG-I^−/− × MDA5^−/− DKO, and WT control mice as previously described (17) and grown in complete DMEM. To facilitate direct comparison of signaling in our PAMP characterization studies, we also generated primary MEFs from RIG-I^−/− embryos of the F3 generation backcrossed onto the C57BL/6 background and used these cells to compare directly to WT, MDA5^−/−, and DKO MEFs. Primary bone marrow-derived macrophages (Mϕ) and dendritic cells (DCs) were generated as described previously (17). Sendai virus, Cantell strain, was purchased from Charles River. Cells were infected with 100 HA units per ml of Sendai virus and harvested at 24 h postinfection. Encephalomyocarditis virus (ECMV) was a gift of R. Silverman (Cleveland Clinic), and cells were infected at an MOI of 5 and harvested at 24 h postinfection.

IFN-β ELISA.

IFN-β in cell culture supernatants was measured by using the mouse-specific IFN-β enzyme-linked immunosorbent assay (ELISA) kit according to the manufacturer's protocol (PBL Biomedical Laboratories).

RNA extraction and analysis.

Total cellular RNA from cultured cells was collected for quantitative reverse transcription-PCR (qRT-PCR) using the RNeasy kit (Qiagen) and reverse transcribed using the iScript select cDNA synthesis kit using both oligo(dT) and random primers (Bio-Rad). Cellular mRNA and viral RNA expression levels were determined by SYBR green qRT-PCR using gene- or virus-specfic primers. Specific primer sets are as follows: mGAPDH forward, CAACTACATGGTCTACATGTTC; mGAPDH reverse, CTCGCTCCTGGAAGATG; mIFNβ forward, GGAGATGACGGAGAAGATGC; mIFNβ reverse, CCCAGTGCTGGAGAAATTGT; mIL6 forward, GTTCTCTGGGAAATCGTGGA; mIL6 reverse, TGTACTCCAGGTAGCTATGG; WNV forward, CGCCTGTGTGAGCTGACAAAC; WNV reverse, CATAGCCCTCTTCAGTCC; mIFN-a2a, purchased from SA Biosciences; mIRF-7 forward, CCCATCTTCGACTTCAGCAC; mIRF-7 reverse, TGTAGTGTGGTGACCCTTGC; mIFIT2 forward, CTGGGGAAACTATGCTTGGGT; and mIFIT2 reverse, ACTCTCTCGTTTTGGTTCTTGG.

RNA preparation.

Infected cell RNA (icRNA) or mock-infected cell RNA (mcRNA) was extracted using TRIzol LS reagent according to the manufacturer's protocol (Invitrogen). WNV virus particle RNA (virion RNA) was purified by layering precleared infected cell supernatants over a 20% sucrose cushion, followed by centrifugation for 4 h at 70,000 × g in a Beckman Coulter SW 40 Ti rotor. RNA was extracted from sedimented virions using TRIzol LS. In vitro-transcribed 5′ and 3′ nontranslated regions (NTR) of WNV were generated as previously described (40, 49). RNA concentrations were determined by absorbance using a NanoDrop spectrophotometer. icRNA was treated with Antarctic phosphatase (AP) to remove phosphate moieties or RNase III to digest dsRNA, followed by ethanol and sodium acetate precipitation (New England BioLabs). Transfections of RNA were performed with a TransIT-mRNA transfection kit according to the manufacturer's protocol (Mirus Bio) into cells pretreated for 30 min with 20 μg/ml of cycloheximide (CHX) in complete DMEM. Total cellular RNA from transfected cells was collected for qRT-PCR 8 to 10 h posttransfection using an RNeasy kit.

Immunoblot analysis.

Protein extracts were prepared as previously described (50), and 20 μg of protein lysate was analyzed by SDS-polyacrylamide gel electrophoresis followed by immunoblotting. The following primary antibodies were used: anti-murine IFIT2 (gift of G. Sen, Cleveland Clinic), goat anti-WNV NS3 (R&D Systems), mouse antitubulin (Sigma), and mouse anti-glyceraldehyde-3-phosphate dehydrogenase (anti-GAPDH; Santa Cruz). All secondary antibodies were purchased from Jackson ImmunoResearch. Immunoreactive bands were detected with Amersham ECL Plus reagent (GE Healthcare).

Statistical analysis.

In vivo Kaplan-Meier curves were analyzed by log rank test. All in vitro statistics were obtained by an unpaired, two-tailed Student t test. A P value of <0.05 was considered significant. All data were analyzed using Prism software (GraphPad Prism).

RIG-I and MDA5 are essential for protection against WNV infection in vivo.

To determine the individual roles of RIG-I and MDA5 in WNV infection and immunity, we assessed WNV pathogenesis in WT mice and in RIG-I^−/− or MDA5^−/− mice. An in-depth, direct analysis of the respective roles of the individual RLRs has been hampered by the embryonic lethality of the RIG-I^−/− genotype in a complete C57BL/6 background, while MDA5^−/− mice are fully viable on several genetic backgrounds (44). To circumvent embryonic lethality, the RIG-I^+/+and RIG-I^−/− mice were generated on a mixed ICR × 129Sv × C57BL/6 genetic background (44). MDA5^−/− mice, in comparison, were generated originally on a 129Sv background without any noted developmental defects (51) and were backcrossed subsequently to a 99% pure C57BL/6 background to facilitate WNV pathogenesis studies. Each RLR single-knockout mouse line was evaluated in comparison to its own individual wild-type (WT) control. We infected mice subcutaneously with 10² PFU of WNV-TX and monitored morbidity and mortality. RIG-I^−/− mice were more susceptible to WNV infection than were their WT controls, with enhanced lethality (50% versus 17%, P < 0.05) observed over the 20-day monitoring period of infection (Fig. 1a). Similar results were obtained when we compared MDA5^−/− mice with their respective WT controls, with significantly more (70% versus 30%; P < 0.02) MDA5^−/− mice dying from WNV infection (Fig. 1b). We also intercrossed RIG-I^−/− × MDA5^−/− double-knockout (DKO) mice lacking expression of both RIG-I and MDA5, and this line was backcrossed into a C57BL/6 background through the F3 generation to yield a strain of ∼94% C57BL/6, as defined by microsatellite DNA analysis (data not shown). DKO mice exhibited markedly increased susceptibility to WNV infection compared to that of WT C57BL/6 control mice or RIG-I^−/− or MDA5^−/− mice, and they had a more rapid mean time to death, ∼8 days (100% versus 50%; P < 0.0001) (Fig. 1c). The susceptibility of DKO mice to WNV infection was remarkably similar to that of MAVS^−/− mice, which were generated on a pure C57BL/6 background (Fig. 1c) (14). These results demonstrate that RLR signaling from both RIG-I and MDA5 is required for protection against WNV infection in vivo.

RIG-I and MDA5 are both required for innate immune gene induction and control of WNV replication.

To determine how RIG-I and MDA5 individually regulate innate immune gene expression and control of WNV replication, we performed a detailed time course analysis of gene expression and virus replication within low-passage-number primary MEFs from RIG-I^+/+ WT, RIG-I^−/−, C57BL/6 WT, MDA5^−/−, and DKO mice. For each analysis, we measured viral RNA and compared innate immune gene expression in mock-infected cells with that in WNV-infected cells at a high multiplicity of infection (MOI = 5), determining the fold change in RNA expression levels (Fig. 2). Whereas WNV RNA replicated to higher levels (4-fold difference [P < 0.003] at 48 h postinfection [hpi]) throughout the 48-h time course in RIG-I^−/− than in cognate WT MEFs (Fig. 2a), no appreciable differences in viral infection were observed between WT and MDA5^−/− cells (P > 0.05 at 48 hpi) (Fig. 2b). The IFN-β gene is an acute innate immune response gene that is induced early after virus infection (10). The increase in WNV replication in RIG-I^−/− MEFs corresponded with an early deficit (14- and 13-fold differences [P < 0.03] at 8 and 10 hpi) of IFN-β mRNA induction, with expression reaching WT levels subsequently (Fig. 2c). In comparison, IFN-β expression was induced equivalently in WNV-infected WT and MDA5^−/− cells throughout the time course (P > 0.05 at 10 and 48 hpi) (Fig. 2d). We also examined the WNV-induced expression of the gene for IFN-α2a, a comparably late-expressed innate immune response cytokine that amplifies and diversifies innate immune gene expression. IFN-α2a is induced after IFN-β signaling, as it requires IRF-7 induction and activation (17, 52). We failed to detect a major difference in IFN-α2a mRNA expression over a course of 22 to 48 hpi between WT and RIG-I^−/− cells (Fig. 2e). Remarkably, MDA5^−/− cells showed a significant deficit in IFN-α2a expression (2- and 3.8-fold differences [P < 0.003] at 34 and 42 hpi, respectively) compared to WT controls at late time points of infection (Fig. 2f). These results verify that RIG-I is essential for early innate immune gene induction and virus control (12) and reveal a specific role for MDA5 in the later, amplification phase of the innate immune response after WNV infection. We note that the early deficit in IFN-β production resulted in increased viral replication in infected cells during time points before WNV can efficiently antagonize IFN signaling, whereas the late defect in IFN-α2a production did not result in increased WNV replication in MEFs under single-step growth kinetic conditions, possibly because at these points accumulated viral nonstructural proteins can antagonize IFN signaling (53, 54).

We next assessed the combined roles of RIG-I and MDA5 in controlling WNV replication and promoting innate immune gene induction in MEFs. DKO cells were mock infected or infected with WNV over a 72-h time course. As expected, WNV RNA accumulated to much higher levels (>60-fold difference at 72 hpi) in DKO than in WT cells (Fig. 2g). We also performed qRT-PCR analysis to examine the expression of a subset of RLR-responsive innate immune genes previously identified from transcriptional profiling studies of WNV-infected cells (12). Remarkably, while these genes were induced highly in WT cells, none were significantly induced in DKO cells after WNV infection (Table 1). These observations are consistent with recent results demonstrating that in target cells of WNV infection, MAVS-dependent signaling is the predominant pathway through which viral RNA is sensed for host defense gene induction (29). Our results demonstrate that RIG-I and MDA5 are the two essential PRRs that sense WNV infection and induce the antiviral response through MAVS-dependent signaling.

MDA5 is required to control virus replication in myeloid cells.

Our results suggest that RIG-I mediates early/initial PAMP recognition and signaling, while MDA5 mediates late signaling to amplify innate immune actions during WNV infection. While previous studies support a role for RIG-I in initial triggering of innate immune defenses against WNV infection, the role of MDA5 in this process has not been evaluated (12, 28, 32). To assess the role of MDA5 in innate immune signaling in bona fide in vivo target cells of WNV infection, we performed time course analysis to evaluate the levels of infectious WNV production and gene induction in primary bone marrow-derived macrophages (Mϕ) and dendritic cells (DCs) prepared from WT and MDA5^−/− mice. Mϕ and DCs from MDA5^−/− mice supported increased virus growth compared to WT cells (4-fold difference [P < 0.04] at 36 h and 2- and 3-fold differences [P < 0.008 and P < 0.02], respectively) (Fig. 3a and b). Immunoblot analysis demonstrated that WNV proteins accumulated to higher levels in MDA5^−/− Mϕ and DCs than in WT cells (especially at 36 h), whereas DCs exhibited a concomitant reduction in the virus-induced expression of IFIT2, an IFN-stimulated gene (ISG) that is downstream of RLR signaling and restricts WNV infection (Fig. 3c and d) (50, 55). We also found that MDA5^−/− Mϕ and DCs produced lower levels of IFN-β postinfection than WT cells (Fig. 3e and f), despite the increased viral burden. As controls, we also assessed IFN-β production in response to Sendai virus (a RIG-I-specific stimulus [44]) and encephalomyocarditis virus (ECMV; an MDA5-specific stimulus [51]), respectively, in WT and MDA5^−/− Mϕ and DCs. Although Sendai virus induced robust production of IFN-β, ECMV infection failed to induce IFN-β production in MDA5^−/− DCs. These results demonstrate that MDA5 is essential for optimal control of WNV replication and induction of antiviral host defense genes in cells that are targets of infection in vivo.

To assess the combined roles of RIG-I and MDA5 in detecting and controlling WNV infection in myeloid cells, we generated DKO DCs and compared their response with those of WT and MAVS^−/− cells. WNV replicated to increased levels (18-fold higher) in DKO cells compared to WT cells, similar to observations in parallel cultures of MAVS^−/− DCs (17-fold higher) (Fig. 3g). Moreover, and in contrast to MDA5^−/− cells, there was no detectable induction of IFN-β production or innate immune gene expression (compare Fig. 3a to d with Fig. 3h and i) by DKO or MAVS^−/− DCs after WNV infection. Thus, while MDA5 is required for optimal innate immune gene induction and control of WNV infection, in myeloid cells the combination of RIG-I and MDA5 and subsequent signaling through MAVS is essential for WNV detection and innate immune gene induction.

WNV PAMP characteristics for RLR detection.

To determine the properties of the PAMPs that trigger RLR signaling during WNV infection, we examined the signaling actions of RNA recovered from mock-infected control cells (mcRNA), cells infected with WNV for 24 h (icRNA), specific WNV RNA secondary-structure motifs, and native virion RNA when transfected into primary MEFs. We first purified RNA from mock- and WNV-infected cells to generate control mcRNA and icRNA. mcRNA transfection induces only very low IFN-β mRNA levels, and we thus normalized all data sets against the response to mcRNA (Fig. 4a). RNA also was obtained from WNV virions (vRNA) by extraction and purification after ultracentrifugation of infected cell supernatants. As the WNV genome 5′ and 3′ nontranslated regions (NTR) contain dsRNA loop structures that might confer RLR recognition, we also prepared in vitro-transcribed RNA fragments of the 5′ and 3′ NTR of the viral genome RNA. Each RNA preparation was transfected into WT MEFs in the presence of cycloheximide to prevent translation and de novo viral transcription, thus allowing us to assess the ability of the input RNA to stimulate innate immune signaling of IFN-β mRNA expression. Notably, transfection of icRNA significantly induced IFN-β mRNA expression in recipient cells compared to control cells transfected with mcRNA (1,050-fold [P < 1 × 10⁻⁵]) (Fig. 4b). We treated icRNA with Antarctic phosphatase (AP) to remove 3′ and 5′ phosphate moieties or with RNase III to digest dsRNA. In parallel, we recovered icRNA from WNV-infected RNaseL^−/− cells, allowing us to assess innate immune signaling induced by possible RNA products of RNase L cleavage (56). icRNA stimulation of innate immune signaling was reduced by ∼50% following phosphatase treatment, and it was completely ablated following RNase III treatment (1,050-fold versus 550-fold [P < 0.001]) (Fig. 4b). icRNA recovered from RNaseL^−/− MEFs contained ∼8 times more WNV RNA (data not shown); consequently, we adjusted the amount of input icRNA into recipient cells to equalize input RNA levels based on WNV genome equivalents. icRNA from RNaseL^−/− cells induced levels of innate immune signaling comparable to those induced by icRNA recovered from WT cells. These results suggest that the WNV stimulatory PAMPs are generated independently of RNase L cleavage products. Neither vRNA nor the in vitro-transcribed viral NTR RNA induced appreciable levels of IFN-β compared to icRNA. This outcome was despite the presence of a 5′-ppp on the viral NTR RNA and the fact that larger copy number of viral RNA was transfected into cells from the NTR RNA compared to icRNA because equivalent mass quantities of RNA were transfected into cells for each condition. Despite an absence of innate immune signaling induction, there were in fact >470-fold more WNV genomes transfected from the vRNA than from icRNA, as determined by qRT-PCR (Fig. 4b). Thus, icRNA but not WNV virion RNA or viral 5′ppp-NTR RNA motifs induce innate immune signaling. These results imply that RLR PAMPs are not carried within the incoming WNV genome RNA of the virion but instead are produced within infected cells, and that PAMP recognition of icRNA by the RLRs includes RNA components with phosphate moieties and dsRNA motifs.

Differential kinetics of RIG-I and MDA5 PAMP production in WNV-infected cells.

To determine how RIG-I and MDA5 individually contribute to the recognition of icRNA PAMPs, we examined icRNA signaling of innate immune genes in WT, RIG-I^−/−, and MDA5^−/− MEFs. We isolated mcRNA from WT mock-infected cells and icRNA from WNV-infected cells at 6, 10, 12, and 34 h after infection. Equal mass amounts of the recovered icRNA or mcRNA were transfected into WT, RIG-I^−/−, and MDA5^−/− cells in the presence of cycloheximide, and IFN-β mRNA levels were measured by qRT-PCR. icRNA recovered after 6 h postinfection stimulated IFN-β mRNA induction in WT cells, but this response was reduced in both RIG-I^−/− and MDA5^−/− cells (Fig. 5a). We observed a difference in early signaling between RIG-I^−/− and MDA5^−/− cells in response to icRNA harvested at 10 and 12 h after infection but not at 34 h after infection. As shown in Fig. 5a, the early signaling of IFN-β mRNA induction was impaired in RIG-I^−/− cells in which signaling was mediated exclusively by MDA5. Signaling was comparable in both RIG-I^−/− and MDA5^−/− cells upon transfection of icRNA recovered from later times (34 h) postinfection. Thus, icRNA from early time points of WNV infection contains PAMPs that were sensed preferentially by RIG-I, whereas icRNA from later time points of infection contains PAMPs that were detected by both RIG-I and MDA5.

As phosphatase treatment of 24 h-derived icRNA reduced PAMP stimulation by ∼50% (see Fig. 4b), we next assessed the requirement for RNA phosphate moieties to affect PAMP sensing by either RIG-I or MDA5. icRNA was recovered from cells 24 h after WNV infection and treated with phosphatase prior to transfection of WT, RIG-I^−/−, or MDA5^−/− MEFs. Phosphatase treatment of icRNA had a minimal effect on reducing IFN-β signaling in RIG-I^−/− MEFs, suggesting that the remaining MDA5 sensing of the WNV PAMP was not affected by loss of exposed phosphates on PAMP RNA. In contrast, phosphatase treatment of icRNA caused an almost complete loss of innate immune signaling in MDA5^−/− MEFs such that less than 10% signaling remained compared to that observed for untreated icRNA transfected into MDA5^−/− MEFs (Fig. 5b). As a control, these RNAs also were transfected into DKO and MAVS^−/− cells; we failed to see significant signaling in either cell type (Fig. 5b). These results reveal a temporal distribution of PAMP detection by RIG-I and MDA5 during WNV infection. Recognition of PAMP RNA within WNV-infected cells at early times occurs in a RIG-I-predominant manner that depends on exposed phosphate moieties, whereas at later times PAMP recognition is performed cooperatively by RIG-I and MDA5. Additionally, our results suggest that WNV generates RIG-I and MDA5-specific PAMPs with differential kinetics over the course of viral replication, and that both PAMPs feature a component of dsRNA that imparts RLR recognition. Results from these RNA transfection experiments agree with infection data (Fig. 2) showing that the loss of RIG-I and MDA5 or MAVS abolishes innate immune signaling in response to cytosolic RNA.