We conducted functional studies to directly compare the roles of RIG-I, MDA5, and IPS-1 in innate immunity against members of the paramyxovirus, flavivirus, reovirus, and orthomyxovirus families. Studies from knockout mice have shown that RIG-I is essential for the innate immune response to the nonhuman paramyxoviruses SenV and NDV (17
). We therefore conducted initial experiments to compare innate immune signaling and response in cells infected with SenV, NDV, or RSV, a human pathogen of public health importance (5
). Paramyxoviruses encode a monopartite genome of negative-sense single-stranded RNA (20
). Infection of primary MEFs by SenV triggered accumulation of the active, nuclear isoforms of IRF-3 in wt and MDA5−/−
cells, but infection failed to induce IRF-3 nuclear translocation in RIG-I−/−
cells (Fig. ). IRF-3 distribution was also predominantly cytoplasmic in uninfected cells. In side-by-side immunoblot analyses, SenV, NDV, and RSV infection induced the expression of IRF-3-dependent genes including ISG15, ISG54, and ISG56 in wt and MDA5−/−
cells but not in RIG-I−/−
cells (Fig. ). In the case of RSV, RIG-I was essential for signaling innate immune defenses against both a clinical strain and a recombinant virus that expresses GFP (Fig. ). This is consistent with a recent study showing RIG-I-dependent cytokine and Toll-like receptor 3 (TLR3) expression in RSV-infected airway epithelial cells (24
). Thus, RIG-I but not MDA5 is essential for triggering innate immune defenses against RSV and other paramyxoviruses. Cells lacking RIG-I were overall more permissive to RSV and NDV infection than were wt or MDA5−/−
cells (Fig. ), indicating that RIG-I actions restrict initial infection. We found that while MDA5 is not essential for innate immune activation by paramyxoviruses, virus-induced gene expression in MDA5−/−
cells was relatively attenuated or delayed compared to that in wt cells (Fig. ), suggesting that MDA5 may play an auxiliary role in amplifying innate immune signaling initiated by RIG-I during paramyxovirus infection. It should be noted that SenV, NDV, and RSV infections also led to the accumulation of RIG-I and MDA5 proteins in wt cells. MDA5 expression was not readily detectable above background levels in the absence of RIG-I during paramyxovirus infection.
FIG. 1. RIG-I-dependent signaling of the innate immune response during paramyxovirus infection. (A) wt, RIG-I−/−, or MDA5−/− mouse fibroblasts were either mock infected (right panels) or infected with SenV at 100 HA units/ml of (more ...)
We further examined the roles of RIG-I and MDA5 in the innate immune response to infection by Dengue virus type 2 (DEN2), a flavivirus encoding a monopartite positive-sense single-stranded RNA genome, and prototypic reoviruses, encoding a multipartite double-stranded RNA genome, of two distinct serotypes: T1L and T3D (20
). Whereas reovirus T3D exhibits severe neurovirulence and high fatality when injected into infant mice, infection with reovirus T1L is rarely fatal and exhibits considerably milder symptoms (1
). Infection by DEN2 triggered the expression of IRF-3-responsive genes in wt, RIG-I−/−
, and MDA5−/−
cells and concomitant accumulation of IRF-3 in the nuclei of infected cells (Fig. ; also data not shown). Reovirus infection of MEFs similarly triggered ISG expression, albeit to different degrees, and this response occurred in the absence of either RIG-I or MDA5 (Fig. ). However, neither RIG-I−/−
cells attained wt levels of virus-induced gene expression during DEN2 or reovirus infection. Furthermore, although RIG-I and MDA5 expression increased in each case following virus infection, only wt cells attained the highest levels of expression. These results indicate that RIG-I and MDA5 are independently dispensable for innate immune signaling during DEN or reovirus infection. Alternatively, our results may indicate the involvement of a yet to be defined cytoplasmic receptor other than RIG-I or MDA5 in innate immune signaling during DEN2 or reovirus infection. To examine this possibility, MDA5 expression in RIG-I−/−
MEFs was silenced by shRNA to generate double-knockout cells lacking both RIG-I and MDA5. DEN2 and reovirus T3D both triggered robust ISG expression in RIG-I−/−
cells that had been treated with a nontargeting shRNA control (Fig. , Neg). In contrast, DEN2 or reovirus T3D infection triggered little or no ISG expression (which directly correlated with MDA5 expression levels) in the double-knockout cells, despite robust viral replication. These results indicate that DEN2 and reoviruses trigger both RIG-I- and MDA5-dependent innate immune responses in mouse fibroblasts.
FIG. 2. DEN and reoviruses trigger the innate immune response independently of RIG-I or MDA5. (A) wt, RIG-I−/−, or MDA5−/− mouse fibroblasts were either mock infected or infected with DEN2 at a multiplicity of infection of 1. (Left) (more ...)
RIG-I and MDA5 signal downstream immune actions through the essential adaptor protein, IPS-1 (14
). To determine the requirement for signaling by either RIG-I or MDA5 in the innate immune response to DEN2 and reovirus infection, we examined whether IPS-1 was required for triggering virus-induced gene expression. In control experiments, cells lacking IPS-1 failed to signal IRF-3 nuclear translocation and lacked ISG expression upon SenV infection (Fig. ). Similarly, when infected with either reovirus T3D or DEN2, IPS-1−/−
cells failed to induce IRF-3-dependent gene expression above basal background levels (Fig. ) or to trigger IRF-3 nuclear translocation following DEN2 (Fig. ) or reovirus T3D (data not shown) infection. To verify these results, we measured IFN-β promoter signaling in HeLa cells transfected with siRNA targeted to IPS-1 or RIG-I. Cells were then either mock infected or infected with SenV or DEN2 and assessed for IFN-β promoter activation. As expected, knockdown of RIG-I expression abrogated IFN-β promoter activation by SenV but had only a partial impact on promoter signaling by DEN2 (Fig. ). Importantly, knockdown of IPS-1 completely abrogated signaling to the IFN-β promoter induced by SenV or DEN2. Thus, IPS-1 is essential for triggering innate immune defenses during infection with SenV, DEN, or reovirus. Taken together, our results suggest that reovirus and DEN signal the innate defenses through an IPS-1-regulated pathway that is likely initiated by either RIG-I or MDA5.
FIG. 3. IPS-1 is essential for triggering of the innate immune response during paramyxovirus, reovirus, or DEN infection. (A) wt or IPS-1−/− mouse fibroblasts were either mock infected or infected with SenV at 100 HA units/ml. (Left) Cells were (more ...)
We also directly compared and assessed the roles of RIG-I and MDA5 in triggering innate immune actions during orthomyxovirus infection. Influenza A and B viruses are orthomyxoviruses, each encoding a multipartite genome comprised of single-stranded, negative-sense RNA (20
). We found that the A/PR/8/34 (Fig. ) or A/Udorn (data not shown) strain of influenza virus triggered IRF-3 nuclear accumulation in a low frequency of cells in cultures of wt (Fig. ) or MDA5−/−
(data not shown) MEFs. Uninfected cells show typical cytoplasmic staining of IRF-3 (Fig. ). In parallel experiments, we found that nuclear accumulation of IRF-3 occurred within a larger frequency of cells during infection of cultures with an isogenic mutant virus, A/PR/8delNS1, that expresses a truncated NS1 gene (data not shown). This supports previous findings that the influenza virus A/PR/8/34 NS1 protein can attenuate RIG-I signaling of IRF-3 activation (11
). Importantly, IRF-3 was never observed to accumulate in the nuclei of influenza virus-infected RIG-I−/−
cells (Fig. ). Immunoblot analysis revealed that each strain of influenza A virus (Fig. ) as well as the B/Yamagata strain of influenza B virus (Fig. ) triggered innate immune response gene expression and concomitant increases in RIG-I and MDA5 protein levels in a manner that was completely dependent on RIG-I. We note that influenza B virus triggered a reduced response in MDA5−/−
cells, which suggests that MDA5 may play a supportive role in signaling the innate immune response to influenza B virus infection. Alternatively, the reduced response may be a direct reflection of low levels of viral growth observed in MDA5−/−
FIG. 4. RIG-I- and IPS-1-dependent innate immune signaling by orthomyxoviruses. (A) (Left panels) wt or RIG-I−/− mouse fibroblasts were either mock infected or infected with influenza virus A/PR/8/34 at a multiplicity of infection of 5. At 12 (more ...)
To assess the mechanisms of RIG-I signaling during influenza virus infection, we conducted IFN-β promoter signaling experiments in HeLa cells and in Huh7 cell clones. HeLa cells were first transfected with siRNA to knock down RIG-I or IPS-1 expression. As shown in Fig. , RIG-I or IPS-1 knockdown abrogated signaling of IFN-β promoter activation by the influenza A virus NS1 truncation mutant. In further experiments we transfected Huh7 cells or Huh7.5 cells with a synthetic double-stranded RNA polymer consisting of annealing strands of inosine and cytosine [poly(I:C)] or with vRNA isolated from influenza virus A/PR/8/34. Huh7 cells are competent for signaling through the RIG-I pathway and trigger innate immune activation in response to poly(I:C), whereas Huh7.5 cells encode defective RIG-I and cannot mediate RIG-I-dependent signaling (37
). Influenza virus vRNA and poly(I:C) control RNA both activated the IFN-β promoter reporter in Huh7 cells (Fig. ). However, neither the vRNA nor the poly(I:C) control was able to trigger IFN-β promoter signaling in the absence of functional RIG-I in Huh7.5 cells. Furthermore, when infected with influenza virus A/PR/8/34 or A/Udorn/1972, cells lacking IPS-1 failed to induce expression of ISG above basal background levels (Fig. ). Taken together, our results demonstrate that RIG-I is essential for triggering the innate immune response to influenza viruses A and B, and they indicate that the orthomyxoviruses signal innate defenses through the IPS-1 adaptor protein. Our observations validate influenza A virus vRNA as a trigger of RIG-I signaling (31
) and indicate that vRNA contains signatures or specific motifs that present a pathogen-associated molecular pattern that is recognized by RIG-I and sufficient to trigger the innate immune response to infection.
To further define the role of RIG-I in signaling host gene expression, we conducted functional genomics analyses to define the cellular genes whose expression is controlled by RIG-I during influenza A virus infection. We designed our bioinformatics analysis to identify RIG-I-responsive genes in cells by comparing wt and RIG-I−/− cells that had been either mock infected or infected with influenza A/PR/8/34 virus for 8 or 26 h. Overall, the number of differentially expressed genes in RIG-I−/− cells far exceeded that in wt cells after influenza virus A/PR/8/34 infection at both time points (Fig. ). Our data show that infection with influenza virus A/PR/8/34 triggered the expression of a number of immune system-related genes that are not expressed, or whose expression is altered, in the absence of RIG-I (Fig. ). In particular, RIG-I−/− cells show a profound lack of expression of a bioset of genes whose products are involved in innate defenses during influenza A virus infection. In agreement with the immunoblot results, we found that IFN-α/β gene expression is significantly attenuated in the absence of RIG-I (expression of IFN-α6 and IFN-β [Fig. ]). There was also a lack of expression of genes whose products are involved in IFN-α/β production and signaling (such as IRF3, IRF7, Stat1, and Stat2) and of ISGs with direct antiviral actions, including PKR, OAS, Mx-1, ISG54, ISG56, and Viperin, in RIG-I−/− cells (Fig. and data not shown). Moreover, TLR3 and MDA5 (Ifih1) expression was attenuated overall in RIG-I−/− cells.
Expression profiling also showed a lack of expression of several genes involved in antigen presentation and the secretion of proinflammatory cytokines in RIG-I−/−
cells (Fig. ), suggesting that RIG-I signaling may further impact those immune processes during influenza A virus infection. The absence of RIG-I also resulted in reduced expression of a subset of genes following influenza virus A/PR/8/34 infection (Fig. ), suggesting that RIG-I may further regulate the maintenance of basal-level expression of these genes. In contrast, the expression of genes involved in TLR signaling of the proinflammatory response, apoptosis, and interleukin-1 signaling of chemokine secretion was elevated in the absence of RIG-I (Fig. ); the primary microarray data for these genes are available at http://expression.viromics.washington.edu
. Viral gene expression was further verified by qPCR, and the analyses showed that at all times tested, viral HA and NP genes were transcribed to higher levels in RIG-I−/−
cells than in wt cells (Fig. ). Taken together, our results indicate that influenza viruses can engage RIG-I during infection, resulting in innate immune signaling and induction of genes broadly involved in immunity and inflammation.