RIG-I and dsRNA-Induced IFNβ Activation

Except for viruses that initiate RNA synthesis with a protein primer (e.g., picornaviruses), most RNA viruses initiate RNA synthesis with an NTP, and at least some of their viral pppRNAs remain unblocked during the infection. Consistent with this, most viruses require RIG-I to mount an innate immune response, whereas picornaviruses require mda-5. We have examined a SeV infection whose ability to induce interferon depends on the generation of capped dsRNA (without free 5′ tri-phosphate ends), and found that this infection as well requires RIG-I and not mda-5. We also provide evidence that RIG-I interacts with poly-I/C in vivo, and that heteropolymeric dsRNA and poly-I/C interact directly with RIG-I in vitro, but in different ways; i.e., poly-I/C has the unique ability to stimulate the helicase ATPase of RIG-I variants which lack the C-terminal regulatory domain.


Introduction
Virus infection elicits potent cellular responses that contain virus spread before the adaptive immune system can intervene, and the production of type I interferons (IFNa/b) is central to this process [1,2]. The sensors involved in coupling recognition of virus infection with the induction of IFNa/b have recently been discovered. These sensors, or pattern recognition receptors (PRRs) that recognize pathogen associated molecular patterns (PAMPs), include RIG-I and mda-5, two cytoplasmic, RNA-binding DExD/ H box helicases (for recent reviews, see [3][4][5]. Both proteins contain N-terminal CARD domains, followed by a DECH box helicase. Both proteins also contain a C-terminal domain, which in the case of RIG-I acts as an internal repressor or regulatory domain (RD) that prevents the CARDs from interacting with their downstream signaling adaptor, IPS-1 [6]. The binding of 59 triphosphorylated RNA ( ppp RNA, which acts as a viral PAMP [7,8]) to the RD of RIG-I leads to its dimerization, which is thought to stimulate the helicase ATPase and release the CARDs for homotypic interaction with IPS-1 [9], the mitochondrial adaptor of both RIG-I and mda-5. IPS-1 activation then leads to the recruitment of a series of kinases which in turn leads to the activation of IRF-3/7 and NF-kB and the expression of the early IFN genes, such as IFNb.
When RIG-I was first described in the seminal paper of Yoneyama et al [10], poly-I/C was proposed as its ligand based on RIG-I over-expression. RIG-I deficient mice, however, were then found not to be defective in their type I IFN response to poly-I/C [11], whereas they were unable to mount an innate immune response to most RNA viruses other than picornaviruses like EMCV (e.g., Influenza A virus, VSV, JEV and Sendai virus (SeV) [12]. Mda-5 deficient mice, in contrast, were found to be entirely unable to mount a type I IFN response to poly-I/C and to EMCV infection [12,13]. The role of these two helicases in the innate immune response to virus infection was thus found to be remarkably specific. Using cell lines derived from these mice, mda-5 2/2 MEFs were found to activate IFN genes in response to various transfected dsRNAs made from complementary ppp RNAs transcribed in vitro, whereas these MEFs did not respond to poly-I/ C. In contrast, RIG-I 2/2 MEFs activated IFN genes in response to transfected poly-I/C, but these MEFs did not respond to dsRNAs made from in vitro transcripts [12,13]. Subsequently, ssRNA transcribed in vitro was also found to be a ligand for RIG-I, and its ability to induce IFN upon transfection depended on the 59 triphosphate moiety of the ssRNA [7,8]. Thus, RIG-I was thought to act as a PRR exclusively for ppp RNA (independent of its singleor double-strandedness), and mda-5 for poly-I/C, or more realistically for the RNA elements of picornavirus infection that are mimicked by poly-I/C. RIG-I and mda-5 are thus thought to recognize different RNA ligands ( ppp RNA and poly-I/C or dsRNA, respectively) that act as PAMPs, which presumably accounts for the virus-specific response of these helicases. This is consistent with our view of RNA virus replication. Except for picornaviruses (and caliciviruses) that initiate all RNA synthesis with a protein primer; the other RNA viruses initiate all RNA synthesis with an NTP, and at least some of the viral ppp RNAs remain unblocked during the infection (e.g., the minus-strands of plus-strand and dsRNA viruses) [14]. Thus, except for picornaviruses (and possibly caliciviruses), cells require RIG-I (and not mda-5) to activate IFNb in response to other RNA virus infections. In order to test this contention, we have designed a SeV infection that generates dsRNA with capped 59 ends [15] to examine whether this SeV infection requires mda-5 rather than RIG-I to activate IFNb. Remarkably, this dsRNA-generating SeV co-infection also requires RIG-I [and not mda-5] to activate IFNb. This study also provides evidence that RIG-I binds dsRNA devoid of free 59 tri-phosphate ends, and that poly-I/C is not a simple analog of dsRNA; i.e., poly-I/C has the unique ability to stimulate the helicase ATPase of RIG-I variants which lack the C-terminal regulatory domain.
SeV-GFP(+), which expresses green fluorescent protein (GFP) from a transgene between the M and F genes, and SeV-GFP(2), which expresses antisense GFP mRNA from a similarly located transgene, were prepared as previously described [16]. DI-H4 stocks were described previously [17].
Primary antibodies used included anti-Flag MAb (F1804; Sigma), rabbit anti-mda-5 and mouse anti-Rig-I (J. Tschopp, Lausanne, Switzerland). Rabbit anti-RIG-I which reacts with both the human and murine helicases was provided by S. Akira (Osaka, Japan).

Plasmids, transient transfections, infections, inductions, and luciferase assay
Flag-tagged RIG-I, and mda-5 were obtained from Klaus Conzelmann (Munich) and Jurg Tshopp (Lausanne). Mda5-DCARD was obtained from S Goodbourn (London). N-terminal deletion mutants of RIG-I (residues 242-925) were constructed by PCR amplification with mutagenic sense primers that introduced a Kpn I site and a met codon in lieu of phe241. C-terminal deletion mutants were constructed with antisense primers that introduced a stop codon and a Kpn site in lieu of Pro 797. The PCR products were digested with kpn and then inserted into pEF-BOS (kindly provided by J. Tschopp). The inserts of the resulting pEF-BOS Rig-I plasmids were confirmed by sequencing.
pb-IFN-fl-lucter, which contains the firefly luciferase gene under the control of the human IFN-b promoter, was described previously [18]. pTK-rl-lucter, used as a transfection standard, contains the herpes simplex virus TK promoter region upstream of the Renilla luciferase gene (Promega).
Transfections. 100,000 cells were plated into six-well plates 20 h before transfection with 1.5 mg of pb-IFN-fl-lucter; 0.5 mg of pTK-rl-lucter; 1 mg of plasmids expressing RIG-I and MDA-5; 1.5 mg of plasmids expressing RIG-DCARD, Mda-DCARD (as indicated); and TransIT-LT1 transfection reagent (Mirus). At 24 h posttransfection, the cells were (or were not) infected with various SeV stocks or transfected with 5 mg of poly(I-C) using TransIT-LT1 transfection reagent. Twenty hours later, cells were harvested and assayed for firefly and Renilla luciferase activity (dual-luciferase reporter assay system; Promega). Relative expression levels were calculated by dividing the firefly luciferase values by those of Renilla luciferase.
Immunoblotting. Cytoplasmic extracts were prepared using 0.5% NP-40 buffer. Equal amounts of total proteins were separated by SDS-PAGE and transferred onto Immobilon-P membranes by semi-dry transfer. The secondary antibodies used were alkaline phosphatase-conjugated goat anti-rabbit (or mouse) immunoglobulin G (Bio-Rad). The immobilized proteins were detected by light-enhanced chemiluminescence (Pierce) and analyzed in a Bio-Rad light detector using Quantity One software.
Recombinant RIG-I cloning and expression. The open reading frame of human RIG-I was amplified by PCR using primers designed to introduce a HindIII site upstream of the start codon and a XhoI site downstream. The PCR products were digested and then inserted between the same sites of pET28-His 10 Smt3, to fuse the RIG-I proteins in-frame with an aminoterminal His 10 Smt3 domain. RIG-I (1-796) was constructed with an antisense primer that introduced a stop codon in place of pro797 and a XhoI site. RIG-I (242-796) was constructed with sense primer that introduced a HindIIII site upstream of lys241and the antisense primer used above. The K270A mutation was introduced by a PCR-based two-stage overlap extension method. The plasmid inserts were sequenced completely to ensure that unwanted mutations during amplification and cloning had not occurred.
The pET28-His 10 Smt3-Tgs1 plasmids were transformed into E. coli BL21. Cultures (500 ml) derived from single transformants were grown at 37uC in LB medium containing 50 mg/ml kanamycin until the A 600 reached 0.6. The cultures were adjusted to 0.2 mM IPTG and 2% ethanol and incubation was continued for 20 h at 17u. Cells were harvested by centrifugation and stored at 280uC. All subsequent procedures were performed at 4u. Thawed bacteria were resuspended in 25 ml of buffer A (50 mM Tris-HCl, pH 8.0, 200 mM NaCl, 10% glycerol) and one tablet of protease inhibitor cocktail (Roche) and lysozyme (100 mg/ml) were added. After incubation for 30 min, imidazole was added to a final concentration of 5 mM and the lysate was sonicated to reduce viscosity. Insoluble material was removed by centrifugation. The soluble extracts were mixed for 30 min with 1.6 ml of Ni 2+ -NTAagarose (Qiagen) that had been equilibrated with buffer A containing 5 mM imidazole. The resins were recovered by centrifugation, resuspended in buffer A containing 5 mM imidazole, and poured into columns. The columns were washed with 8 ml of 10 and 20 mM imidazole in buffer A and then eluted step-wise with 2.5 ml aliquots of buffer A containing 50, 100, 250, and 500 mM imidazole. The elution profiles were monitored by SDS-PAGE. The 250 mM imidazole eluates containing the recombinant RIG-I polypeptides were dialyzed against 50 mM Tris-HCl, pH 8.0, 200 mM NaCl, 2 mM DTT, 1 mM EDTA, 10% glycerol and then stored at 280uC. The protein concentration was determined using the Bio-Rad dye binding method with BSA as the standard.
qRT/PCR of endogenous IFNb mRNA RNA was extracted from cell lysates with TRIzol reagent (Invitrogen) and reverse transcribed with random hexamers (Roche) and Reverse Transcriptase Superscript II (Invitrogen) according to manufacturer's instructions. The cDNA was then amplified using a TaqManH 7500 (Applied Biosystems) thermocycler. Results were analysed using the SDS 1.4 programme (Applied Biosystems). 18S rRNA primers and probe sequences used were as described previously [15]. Murine beta IFN primers and probes used were as described in [19].
In vitro synthesis of RNA, purification, and transfection. DNA for T7 RNA polymerase synthesis of model RNA1 was prepared by PCR using the following partially complementary primers: 59-TAATACGACTCACTATAgggAC-ACACCACAACCAACCCACAAC-39 (forward) (start sites are in lowercase type) and 59-GAAAGAAAGGTGTGGTGTTGG-TGTGGTTGTTGTGGGTTGGTTGTGG-39 (reverse). Transcription was performed on 100 pmol of purified PCR product using T7 MEGAshortcript from Ambion, according to the manufacturer's instructions. Biotinylated RNA1 was synthesized using equal amounts of 59 biotin-UTP and UTP. The T7 transcripts were purified on NucAway Spin columns from Ambion (to remove unincorporated nucleotides).
The various homo-polymers were from Sigma. 59 OH chemically synthesized Tr41 (the first 41 nt of the SeV trailer RNA) and its complement, as well Tr41 with a 39 C 12 extension (539mer) were purchased from MicroSynth. TransIT-LT1 transfection reagent was from Mirus.

SeV Infections of RIG-I 2/2 and mda-5 2/2 MEFs
Unnatural SeV infection is commonly used to induce IFN, as wt SeV infection induces IFN very poorly [17]. This is because the wt SeV genome expresses C and V proteins which counteract the innate immune response in several ways, most notably by their ability to inhibit IFN activation by transfected ppp RNA and poly-I/C [15]. The extent to which IFNb is activated during SeV infection presumably depends both on the level of the RNA PAMPs produced and that of the viral products that counteract the innate immune response. For mononegaviruses whose genome and antigenome RNAs are tightly covered with N protein during their synthesis (and thus are unlikely to act as PAMPs), small promoter-proximal (leader and trailer) ppp RNAs are made independent of assembly with N [15,20,21]. These promoterproximal ppp RNAs are essential for the control of genome replication. Viral dsRNA could also be generated by the annealing of trailer ppp RNA and L mRNA read-thru transcripts. During wt SeV infection of non-immune cells, the activity of both PAMPs is presumably neutralized by the viral C proteins [15]. Akira and coworkers have previously used SeV-C minus infection to induce IFN [12]. The SeV which is more commonly used to induce IFN (''Cantell strain'') is in fact a mixed virus stock composed mostly of copyback defective-interfering (DI) genomes, and this DI infection both under-produces the C and V proteins and overproduces trailer ppp RNA. In addition, because some of the copyback DI genomes are exceedingly small (546 nt vs 15,264 for wt), some of these genomes may also be made without being assembled with N [17]. If so, these unassembled DI genomes would self-anneal to form ppp dsRNA panhandle structures because their ends are selfcomplementary [22]. Thus, both ppp ssRNA and ppp dsRNA (i.e., in which one strand contains a 59 tri-phosphate end) are thought to induce IFNb activation during SeV-C minus and DI infections.
More recently, we have also used quasi-wt SeV co-infections that express GFP mRNA and anti-GFP mRNA (from separate genomes) to activate IFNb [15]. The ability of this co-infection to induce IFN depended on the presence of both complementary GFP mRNAs in the cytoplasm, and RFP mRNA expression could not substitute for one of the GFP mRNAs. IFNb activation induced by the GFP+/2 co-infection thus presumably results from the generation of GFP dsRNA in which both 59 ends are capped.
Moreover, this IFNb activation also appeared to depend on RIG-I, as a dominant-negative form of RIG-I (RIG-I-DCARD) inhibited this response. However, the precise manner in which RIG-I-DCARD acts as a dominant-negative inhibitor of RIG-I is not known, and this mutant helicase may be acting non-specifically when over-expressed.
To further investigate the helicase requirement for the GFP+/2 infection, RIG-I 2/2 and mda-5 2/2 MEFs were transfected with plasmids expressing luciferase under the control of the IFNb promoter (and control plasmids), and then infected with SeV DI-H4 or GFP+/2 stocks. Cells extracts were prepared at 20 hpi and their luciferase activities were determined. As shown in fig 1A, neither infection of RIG-I 2/2 MEFs activated the IFNb promoter above background levels. When RIG-I was re-expressed in these cells by transfection (along with the reporter plasmids), this restored the ability of both infections to activate IFNb. In contrast, the DI-H4 and GFP+/2 infections of mda-5 2/2 MEFs activated the IFNb promoter in both cases. The re-expression of mda-5 in these cells roughly doubled their response to the GFP+/2 infection, but did not stimulate their response to the DI-H4 infection ( fig 1C). Thus, the presence of RIG-I is essential for IFNb activation during GFP+/2 infection that generate capped-dsRNA, as well as during DI-H4 infections. In contrast, the presence of mda-5 is not essential, but can stimulate IFNb activation in response to capped dsRNA in mda-52/2 cells.
To determine the levels of RIG-I generated by transfection relative to those of the endogenous helicase in MEFs, wt MEFs were also transfected with (human) p-Flag-RIG-I (and pGFP as a transfection indicator) under identical conditions as in panels A and C. Cell extracts were then Western blotted with an antibody that reacts with both human and murine RIG-I. This estimation of RIG-I levels is also more informative in wt MEFs because they are efficiently transfected (ca 50%), in contrast to the helicase-deficient MEFs that are poorly transfected (,5%). As shown in panel B, the level of flag-RIG-I expressed by transfection (the upper band of the doublet) is estimated to be 50-100% as strong as the endogenous band. Given that 48% of these wt MEFs were transfected (as indicated by GFP expression), Flag-RIG-I is expressed under these condition at levels that are 1 to 2 times those of the endogenous helicase (assuming that our anti-RIG-I reacts equally with both forms of the helicase).

Interaction of RIG-I with poly-I/C in vivo
We would of course like to confirm that RIG-I also senses dsRNA without 59 tri-phosphate ends, by examining whether capped GFP dsRNA (from SeV infected cells) activates IFNb upon transfection. SeV infected cells, however, will also contain viral leader and trailer ppp ssRNAs, and ppp dsRNAs (made from trailer RNA and L mRNA read-through transcripts), and determining the level of purity of the GFP dsRNA is problematic. We have also tried to prepare such capped dsRNA by capping GFP transcripts (made in vitro) with the vaccinia virus guanylyl transferase, but we were unable to cap more than 70% of each strand. Natural dsRNA that can be obtained in pure form, like reovirus RNA, contains a free 59 tri-phosphate (minus-strand) end. We therefore turned to poly-I/C that contains 59 di-phosphate ends, as model RNAs containing these 59 ends were found not to activate RIG-I [8].
We examined the possible interaction of poly-I/C and RIG-I in MEFs using helicases that lack their N-terminal CARDs, which appear to act as dominant-negative inhibitors of the helicases. For example, over-expression of the tandem CARDs of RIG-I alone induce IFN independently of the presence of viral RNA, suggesting that the CARDs mediate IPS-1 activation and downstream signaling [10]. Mutation of the RIG-I ATP binding site abolishes RIG-I activity, suggesting that ATP and RNAdependent conformational changes are essential for sensing viral RNA [10]. Finally, over-expression of the RIG-I RD alone inhibits RIG-I signaling in response to SeV DI infection, by apparently interfering with the oligomerization of wt RIG-I [6]. RIG-I-DCARD could then act as a dominant-negative inhibitor of RIG-I because of its ability to bind viral RNAs and oligomerize with wt RIG-I, but this mixed oligomer would not activate IPS-1. Less is known about the manner in which mda-5 signals to IFNb.
When IFNb activation in MEFs in response to transfected ppp ssRNA or poly-I/C is compared, this activation is largely inhibited by the co-expression of RIG-I-DCARD in both cases (fig 2). In contrast, the co-expression of mda-5-DCARD has no effect on the activation induced by ppp ssRNA, and a mimimal effect on that induced by poly-I/C. This does not appear to be because mda-5-DCARD is inactive, or because RIG-I-DCARD is acting non-specifically. When the CARD-less helicases are expressed in RIG-I 2/2 MEFs in which mda-5 is (or is not) expressed by transfection, mda-5-DCARD clearly inhibits the poly-I/C induced activation due to the (over-)expressed mda-5, whereas RIG-I-DCARD does not at all inhibit this activation that is exclusively due to the presence of mda-5 ( fig 3A). In contrast, when the CARD-less helicases are expressed in RIG-I 2/2 MEFs in which RIG-I is (or is not) expressed, mda-5-DCARD now does not inhibit (but rather stimulates) the poly-I/C induced activation due to the (over-)expressed RIG-I, whereas RIG-I-DCARD clearly inhibits this activation in the presence of both helicases ( fig 3B). Thus, mda-5-DCARD does indeed act as an inhibitor of poly-I/C induced IFNb activation, but only when this activation is due exclusively to mda-5. When both helicases are present, it is RIG-I-DCARD (and not mda-5-DCARD) that inhibits poly-I/C induced IFNb activation. Although the helicase-deficient MEFs transfect poorly with plasmid DNA, they appear to be more efficiently transfected with either relatively small poly-I/C (400 bp on average) or ppp RNA (55 nt) (fig 4). When RIG-I 2/2 MEFs are transfected with poly-I/ C or ppp RNA, the level of endogenous IFNb mRNA increases only in response to poly-I/C as expected (fig 4A), as mda-5 does not respond to ppp RNA [12]. The increased IFNb mRNA apparently leads to the secretion of IFN, as pretreatment of wt MEFs with the supernatants from the above experiment efficiently prevented the growth of VSV-GFP in these cells only when poly-I/C had been transfected ( fig 4C). In contrast, when mda-5 2/2 MEFs are transfected with poly-I/C or ppp RNA, the level of endogenous IFNb mRNA increases in response to poly-I/C as well as to ppp RNA (fig 4B), and the supernatants from both these transfections have the capacity to inhibit VSV-GFP replication when used to pretreat other MEFs (fig 4D). These results further indicate that RIG-I responds to poly-I/C as well as to ppp RNA. We also examined whether the level of IFNb activation was proportional to that of RIG-I expression in RIG-I 2/2 MEFs. Increasing amounts of RIG-I were expressed in these cells, which were then subsequently transfected with either ppp ssRNA or poly-I/C. Expression of increasing amounts of RIG-I had little or no effect on IFNb activation in the absence of transfected RNA (none, fig 5). In the presence of transfected RNA, the level of IFNb activation was indeed proportional to that of RIG-I expression for both poly-I/C and ppp ssRNA, and poly-I/C was, remarkably, half as efficient as ppp ssRNA. Although it is possible that the combined effect of poly-I/C and increasing RIG-I levels act indirectly to increase IFNb activation (e.g., by increasing mda-5 levels), the fact that this increase in IFNb activation depends on the presence of both poly-I/C and increased RIG-I suggests that RIG-I can interacts with poly-I/C in vivo.

Interaction of poly-I/C and RIG-I in vitro
DExH/D box helicases share 7 conserved sequence motifs that mediate ATP and nucleic acid binding [23]. Nucleic acid binding stimulates the helicase ATPase and results in a conformational power stroke [24]. ATP binding is essential for RIG-I signaling, but the mechanistic role of the ATPase activity in RIG-I signaling to IFNb is unclear. RIG-I is required for non-immune cells to mount an IFN response to SeV-C minus and DI-H4 infections, which presumably express different levels of ppp ssRNA and ppp dsRNA. RIG-I is also required for non-immune cells to mount an IFN response to capped dsRNA (i.e., without any free 59 triphosphate ends). Consistent with these results, the RIG-I ATPase is stimulated not only by ppp ssRNA, but also by several dsRNAs, including those made by annealing chemically synthesized complementary RNAs (dsRNA-tr41, containing two 59-OH ends) (fig 6C), bluetongue (reo)virus dsRNA (BTV RNA; 10 segments ranging from ca. 400 to 4000 bp, which contain one capped and one 59-ppp end) (fig 6D), and poly-I/C, poly-G/C and poly-A/U (all originally containing 59-pp ends)(fig 6E and data not shown). In contrast, the RIG-I ATPase is not stimulated by chemically synthesized OH ssRNA (data not shown) [9]. Cui et al have reported that ppp ssRNA is much more efficient in stimulating wt RIG-I (1-925) ATPase than OH dsRNA. This difference is less pronounced, but clear in fig 6B and C. Two groups have recently reported that the RD of RIG-I alone binds ppp ssRNA, in a shallow, positively-charged groove [9,25]. These two studies were largely in agreement, but differed in that the RD in one case also bound relatively short OH dsRNA (25 bp) [25], whereas that of Cui et al did not interact with OH dsRNA (50 bp) even though the ATPase of a RIG-I variant lacking the CARDs was clearly stimulated by OH dsRNA [9]. We have used streptavidin beads containing biotinylated ppp ssRNA to study RNA/RIG-I interaction. As shown in fig 7, RIG-I binds to ppp ssRNA beads (lanes ''none'' (no competitor) vs bald beads). As expected, this binding can be efficiently out-competed with an excess of unmodified ppp ssRNA (559mer, top panel), but the same amount of OH ssRNA (419mer, top and middle panels), poly-I or poly-C (middle panel) had no effect. More importantly, RIG-I binding to ppp ssRNA can be efficiently out-competed with relatively short OH dsRNA (41 bp, bottom panel) under conditions where an equal amount of OH ssRNA (419mer or 539mer (bottom panel)) has no effect. Poly-I/C also competed with the binding of RIG-I to ppp ssRNA (middle panel). Thus, the binding of short OH dsRNA or poly-I/C and ppp ssRNA to RIG-I appear to be mutually exclusive, either because their binding sites on RD overlap [25], or because OH dsRNA binds to the helicase core in an RD-dependent manner. This latter possibility is suggested by the finding that the ATPase of a RIG-I variant lacking RD (residues 1-793) could not be stimulated by either ppp ssRNA or OH dsRNA [9].
We have repeated and extended these ATPase studies, using RIG-I-DRD (in this case residues 1-796) and RIG-I-DCARD/ DRD (242-796), as well as full-length (wt) RIG-I (1-925). As a negative control, we examined RIG-I containing a mutation in the Walker A box of the helicase ATP binding site (RIG-I-K270A). We find that the RIG-I-DRD ATPase is essentially inactive with not only ppp ssRNA and OH dsRNA, but is inactive as well with poly-G/C, poly-A/U and BTV dsRNA (closed circles, fig 6, and data not shown). Unexpectedly, we found that the RIG-I-DRD ATPase was nevertheless clearly stimulated by poly-I/C (albeit not as efficiently as wt RIG-I; fig 6E). In addition, the RIG-I-DCARD/DRD ATPase was hyper-stimulated by poly-I/C as compared to that of wt RIG-I ( fig 6E). These latter results rule out the possibility that the RIG-I ATPase somehow requires the presence of the RD for its activity. These results are thus more consistent with the binding of ppp ssRNA and OH dsRNA to overlapping sites on RD. It is also possible that the binding of OH dsRNA to RD is helped more by its simultaneous interaction with the helicase core than the binding of ppp ssRNA to RD. In summary, the ability of poly-I/C to stimulate the ATPase of RIG-I is clearly different from that of heteropolymeric dsRNA. It appears that poly-I/C has the unique ability to bind to both the helicase domain and the RD, and can stimulate the helicase ATPase independently of the RD.

Stimulation of the RIG-I ATPase is necessary but insufficient for IFNb activation in vivo
Given that the RIG-I-DRD ATPase is stimulated by poly-I/C, we examined whether poly-I/C could activate IFNb in vivo via this variant of RIG-I. When additional wt RIG-I was expressed in wt MEFs, IFNb activation in response to a fixed amount of poly I/C increased ca. 3-fold (fig 8). However, when additional RIG-I-DRD or RIG-I-DCARD/DRD was expressed, there was little or no effect on poly-I/C induced IFNb activation. When additional RIG-I-DCARD was expressed, this variant acted as a dominantnegative inhibiter of poly-I/C induced IFNb activation, as expected. Similar results were obtained when ppp ssRNA, rather than poly-I/C was used to induce IFNb activation (data not shown). Thus, stimulation of the ATPase is necessary but not sufficient for poly-I/C induced IFNb activation. These results are consistent with the finding that the RD is required for RIG-I selfassociation, and that this self-association is required for downstream signaling [6,9]. The finding that poly-I/C also competed with the binding of RIG-I to ppp RNA (which clearly binds to the RD) (fig 7) suggests that poly-I/C binds to both the helicase core and the RD. These latter results are consistent with the demonstration that the binding of ppp RNA or short dsRNA to RIG-I protects a 17kD fragment from trypsin digestion (essentially the entire RD), whereas the binding of poly-I/C produces a 66 kD fragment of the helicase core and the RD [25].

Discussion
The dsRNAs generated during nondefective SeV infection (due to convergent transcription) or to copyback DI infection (due to intramolecular annealing of the DI genome's complementary ends) would both contain one 59 tri-phosphate end. As RIG-I is thought to be activated by ppp RNA independent of its single-or double-strandedness [7,8], the finding that IFNb activation during SeV C minus and DI-H4 infections require RIG-I was not unexpected. dsRNA devoid of free 59 tri-phosphate ends, however, is normally found only during (e.g., picornavirus) infections that initiate RNA synthesis with protein primers, and these infections activate IFNb via mda-5 rather than RIG-I. The finding that SeV-GFP+/2 infections also requires RIG-I (and not mda-5; figs 1 and 2) was thus not expected. This latter finding, however, is consistent with the recent report that ppp ssRNA and short OH dsRNA bind to overlapping sites of the RIG-I RD, and that short dsRNAs will activate IFNb via RIG-I upon transfection if they contain 59 mono-phosphates that apparently stabilize the dsRNAs intracellularly [25]. We have confirmed that short OH dsRNA competes with ppp ssRNA for its binding to RIG-I, whereas OH ssRNA is inactive (fig 7).
Even though it has been reported that RIG-I by itself does not respond to poly-I/C in mda-5 2/2 MEFs [12,13], especially when poly-I/C of 4-8 kbp in length is used [26], this does not exclude the possibility that RIG-I participates in the cellular response to poly-I/C when both helicases are present, or that RIG-I can directly respond to poly-I/C when RIG-I levels are elevated. When RIG-I levels are gradually increased in MEFs, the level of IFNb activation is proportional to that of RIG-I expression in response to either poly-I/C or ppp ssRNA ( fig 5). There is other indirect evidence that RIG-I participates in the poly-I/C induced IFNb activation when both helicases are present. For example, whereas Huh 7 cells respond well to poly-I/C, a sub-line of these cells containing an incapacitating mutation in a CARD of RIG-I (Huh 7.5 cells) has lost most of its response to this dsRNA [27]. In addition, even though the ability of paramyxovirus V proteins (including that of SeV) to bind mda-5 and prevent poly-I/C induced signaling is well documented [28,29], it is the SeV C protein (that inhibits RIG-I signaling) and not the SeV V protein which acts to inhibit the cellular response to poly-I/C in a SeV infection [15].
We found that both short dsRNAs containing two 59 OH ends (dsRNA-tr41) and BTV (reovirus) dsRNAs stimulate the RIG-I ATPase in vitro almost as efficiently as ppp ssRNA ( fig 6). Unexpectedly, we also found that poly-I/C stimulated the ATPase of RIG-I variants lacking the C-terminal RD, in strong contrast to either ppp ssRNA or heteropolymeric dsRNA ( fig 6). Poly-I/C is an unusual dsRNA, composed of complementary homopolymers, and one strand being composed of the rare base inosine. Poly-I/C has been the transfected dsRNA of choice to study IFN induction for many years, not only because it is commercially available, but because it works so well. It has been known for decades that poly-I/C has a special ability to induce IFN [30]. Using DEAE-dextran to transfect various dsRNAs into chick embryo fibroblasts, the efficiency of IFN induction of poly-I/C was found to be orders of magnitude greater than that of poly-G/C or poly-A/U. Moreover, these differences in activity among the various polynucleotides did not appear to reflect differences in their thermal stability, sensitivity to RNase A, the rate or amount of uptake into the cells or in the rate of intracellular breakdown. Colby and Chamberlin (1969) presciently concluded that the high degree of specificity of the induction process was consistent with the existence of a specific intracellular receptor, most probably a protein. One of the possible reasons for the remarkable ability of poly-I/C to induce IFN may be because, unlike other dsRNAs or ppp ssRNAs, it can directly bind to both the RIG-I helicase domain as well as to mda-5, stimulating their ATPases and inducing the conformational changes that liberate the CARDs for interaction with IPS-1. At present, there is no clear understanding of the RNA elements of poly-I/C that mimic the PAMPs generated during mouse picornavirus infections, or why poly-I/C alone among various dsRNAs stimulates the RIG-I-DRD ATPase. However, it is becoming increasingly clear that poly-I/C is not a simple analog of dsRNA.
While this work was being prepared for publication, Kato et al [26] reported that the length of dsRNA is important for differential recognition by RIG-I and mda.5. Long poly-I/C (4-8 kbp) was found to be a specific ligand for mda-5, whereas relatively short poly-I/C (300 bp) was a specific ligand for RIG-I. We have examined the length of our poly-I/C by agarose gel electrophoresis and found that it was relatively short (200-600 bp relative to DNA markers), and the maximum length of our GFP dsRNA would be 714 bp. Our results are thus consistent with those of Kato et al. Figure 7. Competition of various RNAs for RIG-I binding to pppRNA beads. Streptavidin beads to which 1.5 ug of biotinylated ppp RNA1 had been added (or not; bald beads) were incubated with 1 ug of purified RIG-I (Methods) and 5 ug of various competitor RNAs as indicated. After 2 h at 4u, the beads were washed 3 times with base buffer, SDS protein sample buffer was added, and the beads were Western blotted with anti-RIG-I. The amount of RIG-I remaining on the beads (% bound) relative to that in the absence of competitor is given below each panel. The right-hand lanes ''input'' show 1/8 of the RIG-I used for the RNA binding that were directed Western blotted (i.e., 1/8 of the input). doi:10.1371/journal.pone.0003965.g007