Respiratory Syncytial Virus Induces RelA Release from Cytoplasmic 100-kDa NF-κB2 Complexes via a Novel Retinoic Acid-inducible Gene-I·NF-κB-inducing Kinase Signaling Pathway*

Respiratory syncytial virus (RSV) is a primary cause of severe lower respiratory tract infection in children worldwide. RSV infects airway epithelial cells, where it activates inflammatory genes via the NF-κB pathway. NF-κB is controlled by two pathways, a canonical pathway that releases sequestered RelA complexes from the IκBα inhibitor, and a second, the noncanonical pathway, that releases RelB from the 100-kDa NF-κB2 complex. Recently we found that the retinoic acid-inducible gene I (RIG-I) is a major intracellular RSV sensor upstream of the canonical pathway. In this study, we surprisingly found that RIG-I silencing also inhibited p100 processing to 52-kDa NF-κB2 (“p52”), suggesting that RIG-I was functionally upstream of the noncanonical regulatory kinase complex composed of NIK·IKKα subunits. Co-immunoprecipitation experiments not only demonstrated that NIK associated with RIG-I and its downstream adaptor, mitochondrial antiviral signaling (MAVS), but also showed the association between IKKα and MAVS. To further understand the role of the NIK·IKKα pathway, we compared RSV-induced NF-κB activation using wild type, Ikkγ-/-, Nik-/-, and Ikkα-/--deficient MEF cells. Interestingly, we found that in canonical pathway-defective Ikkγ-/- cells, RSV induced RelA by liberation from p100 complexes. RSV was still able to activate IP10, Rantes, and Groβ gene expression in Ikkγ-/- cells, and this induction was inhibited by small interfering RNA-mediated RelA knockdown but not RelB silencing. These data suggest that part of the RelA activation in response to RSV infection was induced by a “cross-talk” pathway involving the noncanonical NIK·IKKα complex downstream of RIG-I·MAVS. This pathway may be a potential target for RSV treatment.

In the United States, respiratory syncytial virus (RSV) 2 causes about 86,000 hospitalizations for severe lower respiratory tract infections (1). Here, pathologic lesions such as bronchiolar epithelial necrosis, bronchiolar occlusion, parenchymal inflammation, and alveolar exudation are found (2). These features, along with the finding that inhibition of mucosal NF-B signaling in a mouse model of RSV disease blocks mononuclear infiltration and disease manifestations (3), suggest that the inflammatory response mediates a lot of clinical disease manifestations.
Previous studies have shown that RSV induces expression of 16 cytokines and chemokines (4), many of which are NF-B dependent (5,6). Five members of the NF-B family have been reported, including three subunits with transactivating function, RelA, RelB, c-Rel, and two DNA binding subunits, NF-B1 (p50) and NF-B2 (p52) (7). The NF-B molecules are sequestered in the cytoplasm by interacting with a group of inhibitory proteins including IB␣, IB␤, IB⑀, p100, and p105 (8). Currently, it is understood that NF-B activation can be controlled by two separate pathways, the canonical and noncanonical pathways, activated by distinct stimuli and under control of distinct IB kinase (IKK) complexes. The canonical NF-B pathway is induced by the monokines TNF and interleukin-1, stimulating the IKK complex composed of the catalytic kinases, IKK-␣ and-␤ and the regulatory subunit, IKK␥ (9). Activated IKK, in turn, induces the phosphorylation-coupled degradation of IB␣, liberating sequestered cytoplasmic RelA allowing RelA to translocate into the nucleus and initiates gene transcription (10). The non-canonical NF-B pathway is induced by lymphotoxin ␤ (LT␤), the TNF superfamily member, LIGHT, or other lymphokines (11)(12)(13)(14). This pathway stimulates a complex of NIK and IKK␣, resulting in the phosphorylation-coupled proteolysis of 100-kDa NF-B2 ("p100") to form the activated 52-kDa NF-B2 ("p52")-RelB complex. Although we found that RSV was able to activate NIK kinase activity and p52 formation (15), the mechanism how RSV activates the NIK⅐IKK␣ complex is unknown.
Recently, we found that retinoic acid-inducible gene I (RIG-I), a DEXD/H box RNA helicase, was the initial intracellular sensor for RSV infection and upstream of the canonical NF-B pathway (16). RIG-I is a central regulator of dsRNA-induced signaling for most of the single-stranded RNA viruses (17,18). RIG-I is upstream of the mitochondrial antiviral signaling (MAVS) protein, also known as interferon-␤ promoter stimulator 1, caspase recruitment domain (CARD) adaptor inducing IFN-␤ (Cardif) or virus-induced signaling adaptor (19 -23). Activated RIG-I binds to MAVS through a homologous CARD motif; interestingly this interaction is absolutely required for downstream RIG-I signaling and occurs on the outer mitochondrial membrane (19).
In this study, we discover that siRNA-mediated RIG-I silencing inhibits basal and RSV-inducible p52 formation. Further investigation revealed that NIK associates with the RIG-I⅐MAVS complex via the RIG-I NH 2 -terminal CARD domain and the COOH terminus of NIK. Surprisingly, we found RSV induced RelA activation in Ikk␥ Ϫ/Ϫ mouse embryonic fibroblasts (MEFs), cells lacking a functional canonical NF-B pathway. In nondenaturing co-immunoprecipitation experiments, we demonstrated that the NIK⅐IKK␣ complex induced RelA release from cytoplasmic p100 complexes. Together, these findings indicate RIG-I controls RelA activation by two distinct downstream signaling modules, one mediated by canonical pathway activation and the second involving a novel cross-talk pathway dependent on complex formation with NIK⅐IKK␣ whose activation liberates RelA from p100 sequestration. Targeted disruption of this pathway may have significant effects in modulating the inflammatory response to RSV without affecting the viral replication.
Virus Preparation and Infection-The human RSV A2 strain was grown in Hep2 cells and prepared as described (27). The viral titer of purified RSV pools was varied from 8 to 9 log plaque forming units/ml, determined by a methylcellulose plaque assay. Viral pools were aliquoted, quick-frozen on dry ice-ethanol, and stored at Ϫ70°C until they were used. For viral adsorption, cells were transferred into culture medium containing 2% (v/v) fetal bovine serum and RSV infected at a multiplicity of infection of 1 for the times as described in the text.
Plasmid Construction-Expression vectors encoding fulllength and a series of deletion mutants of NIK were produced by PCR and cloned as BamHI/XbaI sequences into the pEGFP-Myc plasmid (Invitrogen). The sequences of the primers are shown in Table 1. A FLAG-tagged full-length NIK expression vector (FLAG_EGPF_NIK) was constructed as a COOH-terminal fusion in pEGFP plasmid. Expression vectors encoding FLAG epitope-tagged RIG-I and its deletion mutants were under the control of a tetracycline response element in a modified pT1S plasmid as described (16). Expression vectors encoding Myc epitope-tagged MAVS and different deletion mutants were produced by PCR and cloned into HindIII/XbaI of the modified pcDNA3_strawberry plasmid; the primers used are listed in Table 2. PEF6_Flag_IKK␣ was cloned into PEF6_V5 vector (Invitrogen) using upstream primer: 5Ј-CTTGGTTAT-GGACTACAAGGACGACGATGACAAGGGGTCGACCA-TGGAGCGGCCCCCGGGGCTGCGGCCG-3Ј, and downstream primer, 5Ј-TCATTCTGTTAACCAACTCCAATC-AAGATTC-3Ј.
siRNA-mediated Gene Silencing-siRNA against human RIG-I (M-012511-00), mouse Relb (M-040784-01), mouse Rela (M-040776-00), and control siRNA (D-001206-13) were commercially obtained from Dharmacon Research, Inc. (Lafayette, CO). The siRNA targeting RIG-I and control siRNA were transfected at 100 nM into A549 cells using a TransIT-siQuest transfection kit (Mirus Bio Corp., Madison, WI) according to the manufacturer's instructions. The control siRNA and the siRNA targeting RelA and RelB were transfected at 50 nM into MEF cells using reverse transfection according to the manufacturer's protocol. Forty-eight hours after transfection, cells were RSV infected at the indicated times. The silencing efficiency of siRNA was evaluated using reverse transcriptase-PCR (RT-PCR) for RIG-I as well as Western immunoblot for RelB and RelA.
RT-PCR and Quantitative Real-time PCR (qRT-PCR)-Total RNA was extracted using acid guanidium phenol extraction (TriReagent; Sigma). One microgram of RNA was reversibly transcribed using Superscript III (Invitrogen) in a 20-l reaction mixture. One l of cDNA product was diluted 1:2, and 2 l was amplified in a 25-l reaction mixture containing 12.5 l of SYBR Green supermix (Bio-Rad) and 0.4 M each of forward and reverse gene-specific primers (Table 3), aliquoted into 96-well, 0.2-mm thin-wall PCR plates, and covered with optical quality sealing tape. The plates were denatured for 90 s at 95°C and then subjected to 40 cycles of 15 s at 94°C, 60 s at 60°C, and 1 min at 72°C in iCycler (Bio-Rad). After PCR was performed, PCR products were run on 2% agarose gels to assure a single amplification product. Static analysis of gene expression was described before (16). Electrophoretic Mobility Shift Assay-A total of 35 g of whole cell extracts were incubated in DNA-binding buffer containing 5% glycerol, 12 mM HEPES, 80 mM NaCl, 5 mM dithiothreitol, 5 mM MgCl 2 , 0.5 mM EDTA, 1 g of poly(dA-dT), and 100,000 cpm of 32 P-labeled double-stranded oligonucleotide containing NF-B binding sites in a total volume of 25 l as described (28). Gels were dried and exposed to Bio-Max film (Kodak) for autoradiography.
Co-immunoprecipitation and Western Immunoblot-Whole cell extracts were prepared using modified radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.25% sodium deoxycholate, 1% IGEPAL CA-630, 1 mM phenylmethylsulfonyl fluoride, 1 mM sodium fluoride, 1 mM Na 3 VO 4 , and 1 g/ml each of aprotinin, leupeptin, and pepstatin). Whole cell extracts were pre-cleared with protein A-Sepharose 4B (Sigma) for 10 min at 4°C and immunoprecipitation was conducted for 2 h at 4°C with primary Ab. Immune complexes were then precipitated by adding 50 l of protein A-Sepharose beads (50% slurry) and incubated for 1 h at 4°C. Beads were washed three times with cold TB buffer (150 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, pH 7.4, 0.05% IGEPAL CA-630), and immune complexes were fractionated by 10% SDS-polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane by electroblotting. Membranes were blocked in 5% nonfat dry milk in Tris-buffered saline, 0.1% Tween and probed with the primary Ab indicated in the figure legends. Membranes were washed and incubated with IRDye 700-conjugated anti-mouse Ab or IRDye 800-conjugated anti-rabbit Ab (Rockland, Inc.). Finally, the membranes were washed three times with TBS-T and scanned by an Odyssey infrared scanner. Sources of primary Ab were: anti-FLAG M2 mAb (Stratagene), anti-Myc mAb (Santa Cruz), rabbit anti-RelA C20 polyclonal Ab (Santa Cruz), anti-p52 polyclonal Ab (Upstate, Charlottesville, VA), and goat anti-RSV antibody (Biodesign). For the experiments requiring the second primary antibody, the membrane was not stripped to remove the first primary antibody, and the second primary antibody was incubated with membrane directly.
Electroporation-Two million freshly isolated MEFs were suspended in 100 l of MEF2 nucleofactor solution (Amaxa), and transfected (program A023) with 5 g of plasmid DNA. After transfection, cells were immediately transferred to Dul-becco's modified Eagle's medium and cultured for at least 48 h before treatment.

RSV-induced p52 Formation Is RIG-I-, NIK-, and
IKK␣-dependent-Recently, we have reported that RSV infection activated the noncanonical pathway. To illustrate, we stimulated A549 cells with the LT␤ agonist, LIGHT (29), a noncanonical pathway activator, and compared p52 formation with that induced by RSV infection (Fig. 1A). LIGHT rapidly and strongly induced p52 formation within 30 min after treatment, which persisted for more than 6 h. In addition, we noted p100 expression was up-regulated, a consequence of p100 being downstream of the NF-B pathway (30). Similarly, RSV infection induced p52 formation detectable 12 h after RSV adsorption and persisted for 24 h (Fig. 1A). In contrast to the response to LIGHT, the p100 precursor was transiently induced at the 12-h point and did not persist at 24 h. This is probably due to the transient activation of the noncanonical pathway in RSV infection (15).
Previous work from our laboratory has shown that RSV-induced NF-B activation is dependent on active viral replication, with either UV-inactivated RSV or RSV-conditioned medium being unable to induce NF-B binding upon exposure to naïve cells (5). To determine whether RSV replication is necessary for p100 processing, A549 cells were exposed to RSV, UV-inactivated RSV, or conditioned medium from RSV-infected cells. Although replication competent RSV induced p52 formation, neither UV-inactivated RSV nor CM from RSV-infected cells produced detectable p52 formation (Fig. 1B). Interestingly, p52 formation has previously been shown to be mediated via a cotranslational mechanism, where newly synthesized p100 was the preferential target for processing to p52 (31). In this study, it was shown that cycloheximide (CHX) inhibits the formation of both p100 and p52. To determine whether RSV induced p52 formation also via a co-translational mechanism, RSV infection was repeated in the presence of CHX. We noted that CHX inhibited the formation of the p100 precursor and inducible p52 formation was inhibited (Fig. 1B, right panel). These data indicated that RSV replication was required for p52 formation, and that newly synthesized p100 was the primary source of p52 processing.
RIG-I is the major cytoplasmic sensor of RSV infection upstream of the canonical pathway, but the pathway mediating   Gro␤  CACTCTCAAGGGCGGTCAA  TGGTTCTTCCGTTGAGGGAC  IP10  CGATGACGGGCCAGTGA  CGCAGGGATGATTTCAAGCT  Rantes  TCCAATCTTGCAGTCGTGTTTG  TCTGGGTTGGCACACACTT  GAPDH CATGGCCTTCCGTGTTCCTA GCGGCACGTCAGATCCA noncanonical activation is unknown (16). We next examined whether RIG-I mediated RSV-induced p52 formation. For this purpose, we silenced RIG-I expression using siRNA transfection. In comparison with control siRNA-transfected cells, where RIG-I mRNA was not detectable in uninfected cells and its expression was strongly up-regulated 24 h after RSV infection, accumulation of RIG-I mRNA was significantly reduced in cells transfected with RIG-I-specific siRNA (Fig. 1C, top). Importantly, both the basal and RSV-induced p52 accumulation was significantly decreased in RIG-I-silenced cells (Fig. 1C, bottom).
RIG-I and MAVS Associate with NIK and IKK␣-To further investigate the novel functional interaction between RIG-I and NIK, and determine the domains involved, co-immunoprecipitation experiments were conducted. First, full-length Myc_NIK and different FLAG_RIG-I constructs were co-transfected into HEK293 cells. The RIG-I constructs included full-length RIG-I, the RIG-I NH 2 terminus containing the two CARD domains (RIG_N), and the RIG-I COOH-terminal part containing the helicase domain (RIG_C). NIK was precipitated using the anti-Myc Ab, and RIG-I association was detected by Western immunoblot using anti-FLAG Ab ( Fig. 2A, first panel from top). These results indicated that NIK bound to full-length RIG-I and RIG_N, but not RIG_C, suggesting that the two CARD domains of RIG-I are required for RIG-I⅐NIK complex formation.
To confirm this finding, the reverse experiment was performed using the anti-FLAG Ab to precipitate RIG-I and NIK association determined using anti-Myc Ab in the Western immunoblot. The same result was observed in this experiment ( Fig. 2A, third panel from top).
We next investigated whether the downstream RIG-I adapter, MAVS, complexed with NIK. Co-immunoprecipitation experiments were performed using full-length FLAG-and GFP-tagged NIK and different deletion mutations of MAVS. For this experiment, 4 forms of Myc and Strawberry-tagged MAVS plasmids were used including wild type MAVS (MAVS), MAVS deleted in its NH 2 -terminal CARD domain A, A549 cells were treated with LIGHT at 0, 0.5, 1, 3, and 6 h, as well as infected with RSV (multiplicity of infection 1) at 12 and 24 h. Whole cell extracts were collected and Western immunoblot was conducted to detect the expression of p100 and its proteolytic product p52. ␤-Actin staining was used as a loading control. B, A549 cells were treated with RSV, UV-inactivated RSV, or RSV conditioned medium (CM) for the indicated times (in h, bottom). Right panel, cells were RSV infected and treated with cycloheximide (CHX, 25 g/ml) for the indicated times. The abundance of p100 and p52 were determined by Western immunoblot. ␤-Actin staining was used as a loading control. C, A549 cells transfected with control (Con) siRNA and RIG-I siRNA for 48 h were then RSV infected for 0 and 24 h. Total RNA was extracted and assayed by RT-PCR to measure the expression of RIG-I (upper panel). ␤-Actin is a control. Shown is an ethidium bromide-stained agarose gel. Whole cell lysates were prepared from the same cell treatment and p100/p52 by Western immunoblot using NH 2 -terminal anti-NF-B2 Ab (lower panel). The blot was probed with ␤-actin as a loading control. D, WT, Nik Ϫ/Ϫ , Ikk␣ Ϫ/Ϫ , and Ikk␥ Ϫ/Ϫ MEFs were infected by RSV (multiplicity of infection 1) for 0 and 24 h. Whole cell lysates were assayed by Western immunoblot to detect p100 expression and p52 formation (indicated by asterisk). ␤-Actin was a loading control. p52 was detected only in RSV-infected WT and Ikk␥ Ϫ/Ϫ cells. (MAVS_dN), MAVS deleted in its COOH-terminal transmembrane domain (MAVS_dC), and MAVS without both its CARD and transmembrane domains (MAVS_Dou). First, full-length NIK was precipitated using anti-FLAG Ab and MAVS association detected by anti-Myc Ab in Western immunoblot (Fig. 2B,  top panel). The converse experiment was conducted to precipitate MAVS using anti-Myc Ab, and NIK detected using anti-FLAG in Western blot (Fig. 2B, third panel from top). Both experiments produced the same result, only the full-length MAVS associated with NIK.
To confirm that this observation was not the result of artifactual overexpression, we sought to co-immunoprecipitate the endogenous proteins. For this experiment, RSV-infected cells were immunoprecipitated with IgG, or with anti-MAVS Abs. The presence of NIK was then assayed in the washed immune complexes by Western immunoblot. A specific NIK band was observed only in the MAVS immunoprecipitate (Fig. 2C). These data confirmed that the endogenous proteins formed a complex in cellulo.
To identify which NIK domains were required for MAVS complex formation, co-immunoprecipitation experiments were conducted using different NIK deletion mutants. A series of expression vectors encoding Myc epitope-tagged NH 2 and COOH-terminal domain deletions were tested (Fig. 3A). These constructs were co-transfected with full-length FLAGtagged MAVS, and MAVS-associated complexes precipitated using anti-FLAG Ab. Our results suggested that the N-terminal deletion mutants of NIK (C1, C2, and C3) associate with MAVS, but the COOH-terminal deletions (N1, N2, and N3) did not (Fig. 3B). This result suggested that the COOH terminus of NIK spanning amino acids 660 -947 was required for complexing with MAVS.
Because of the central role of IKK␣ in mediating the noncanonical pathway, and its functional requirement for RSV-induced p52 formation (shown in Fig. 1D), the association of IKK␣ and MAVS was also investigated. Full-length FLAG-tagged IKK␣ was co-expressed with full-length or deletion mutants of MAVS. The complex was precipitated using anti-Myc Ab and IKK␣ association detected using anti-FLAG Ab in the Western immunoblot. Our result suggested that only full-length MAVS binds IKK␣ (Fig. 3C). To confirm this finding, GFP-tagged IKK␣ and YFP-tagged MAVS were cotransfected into A549 cells. 36 h after transfection, colocalization of IKK␣ and MAVS were studied by confocal microscopy. Three images are shown in Fig. 3D. The first image shows two cells expressing both GFP_IKK␣ and YFP_MAVS. The colocalization of these two molecules is demonstrated by the yellow color in the merged figure (at right). In the second image, most cells express only GFP_IKK␣; here the IKK␣ distribution is evenly dispersed through the cytoplasm. In the third image, one cell is expressing GFP_IKK␣ only and the fluorescence distribution is even throughout the cytoplasm. The other cell expressing both YFP_MAVS and GFP_IKK␣, the redistribution of green fluorescence to a mitochondrial pattern is again demonstrated. Together these data indicate that IKK␣ is recruited to colocalize with MAVS on the mitochondrion.
RSV Activates RelA Translocation in Ikk␥ Ϫ/Ϫ MEFs-To further understand the function of the NIK⅐IKK␣ complex in RSVinduced signaling, we infected WT, Ikk␥ Ϫ/Ϫ , Nik Ϫ/Ϫ , and Ikk␣ Ϫ/Ϫ MEFs with RSV and assayed nuclear extracts (NE) for canonical NF-B DNA binding activity using a RelA/NF-B1selective probe (28). In WT cells, RSV induces the presence of a RelA-containing DNA binding complex 12 and 24 h after viral adsorption. Surprisingly, in Ikk␥ Ϫ/Ϫ cells, although RelA DNA binding activity was not detected 12 h after RSV infection, RelA binding was strongly induced 24 h after infection (Fig. 4A, second panel from left). Similar complexes were observed in Nik Ϫ/Ϫ and Ikk␣ Ϫ/Ϫ MEFs. To confirm which member of the NF-B family forms this DNA protein complex, NE from RSVinfected Ikk␥ Ϫ/Ϫ cells were incubated with anti-RelA, anti-RelB, or anti-p52 Abs and a supershift assay was performed. Only the sample incubated with RelA produced a retarded band, leading us to conclude that RelA is contained in the DNA binding complex in RSV-infected Ikk␥ Ϫ/Ϫ cells.
To further examine the complexes in Ikk␥ Ϫ/Ϫ , Nik Ϫ/Ϫ , and Ikk␣ Ϫ/Ϫ MEFs, RSV-infected NEs were subjected to supershift assay for RelA, RelB, NF-B1/p50, and NF-B2/p52 complexes (Fig. 4B). With all three cell types, the predominant members of the inducible DNA binding complex were RelA and NF-B1/ p50 as indicated by the ability of these Abs to reduce the RSV inducible complex and form a supershifted band (* in Fig. 4B).
To more convincingly demonstrate RelA translocation, sucrose cushion purified NE from RSV-infected WT, Ikk␥ Ϫ/Ϫ , Nik Ϫ/Ϫ , and Ikk␣ Ϫ/Ϫ MEFs were assayed for changes in RelA abundance by Western immunoblot. Here, RSV infection strongly induced the nuclear RelA signal in all cell types. Importantly, RSV infection increased the nuclear accumulation of RelA in Ikk␥ Ϫ/Ϫ MEFs (Fig. 4C), leading us to conclude that RSV induces RelA translocation independently of the canonical pathway.
To determine whether that RelA was transcriptionally competent, a NF-B luciferase reporter gene (the IFN␤ PRD II domain (32)) was transfected into Ikk␥ Ϫ/Ϫ MEFs and exposed to the absence or presence of RSV. A 3-fold increase in normalized luciferase activity was observed in the RSV-infected cells (Fig. 4D).
We further tested the requirement of RelA in RSV-induced gene expression in Ikk␥ Ϫ/Ϫ cells. To isolate its function, we separately silenced RelA or RelB by siRNA transfection. The efficiency of siRNA knockdown was evaluated using Western immunoblot to detect RelA and RelB expression. Here, a Ͼ50% inhibition of RelA or RelB expression was produced by the cognate siRNA transfection (Fig. 5). In the Control, RelA-, or RelB-silenced IKK␥ Ϫ/Ϫ cells, the RSV-induced expression of IP10, Rantes, and Gro␤ were measured by qRT-PCR. In cells transfected with control siRNA, Ip10, Rantes, and Gro␤ increased about 8-, 15-, and 7-fold, respectively. A similar level of induction was observed in the cells transfected with siRNA targeting RelB, whereas the expression of all the three genes decreased significantly after RelA was silenced in IKK␥ Ϫ/Ϫ cells (Fig. 5). These data suggested the existence of an RSV-inducible IKK␥-independent signaling pathway that activates RelA translocation whose presence is required for transcriptional activity of endogenous target genes.
Previous studies have reported that p100 forms a heterodimer with RelA and acted as an inhibitor to sequester RelA in the cytoplasm (14,33). However, the pathways controlling RelA release from p100 are not known. Because our findings showed that p52 formation was IKK␥-independent (Fig. 1D), we investigated whether RSV induced RelA release from p100associated complexes in WT and Ikk␥ Ϫ/Ϫ MEFs. To quantitate the p100⅐RelA complex, nondenaturing co-immunoprecipitation was performed. Here, p100 was precipitated in cytoplasmic extracts from RSV-infected WT and Ikk␥ Ϫ/Ϫ MEFs and Western immunoblot was performed using anti-NF-B2 and anti-RelA Abs. In uninfected WT MEFs, p100 was strongly associated with RelA; 24 h after RSV adsorption, p100-associated RelA was significantly decreased. Importantly, the same phenomenon occurred in the Ikk␥ Ϫ/Ϫ MEFs (Fig. 6B). These data suggested that RSV-induced RelA translocation is, in part, mediated by an IKK␥-independent proteolysis of p100.
These data indicate that RSV-induced activation of the noncanonical pathway results in RelA translocation. To determine whether a similar phenomenon occurred in response to a prototypical activator of the noncanonical pathway, we examined whether LIGHT induced RelA translocation. In this experiment, A549 cells were stimulated with LIGHT or TNF␣ for 0, 1, or 3 h. NE extracts were then assayed for RelA and RelB translocation by Western immunoblot. We observed that LIGHT induced both RelA and RelB translocation within 1 h of stimu-lation, whereas TNF␣ only induced RelA translocation (Fig. 6C).
RSV Replication Was Increased in Ikk␥ Ϫ/Ϫ , but Not in Nik Ϫ/Ϫ and Ikk␣ Ϫ/Ϫ MEFs-To determine the role of the canonical and noncanonical/cross-talk pathways in antiviral response, WT, Nik Ϫ/Ϫ , Ikk␣ Ϫ/Ϫ , and Ikk␥ Ϫ/Ϫ MEFs were RSV infected. Expression of viral proteins was then determined using Western immunoblot using anti-RSV P, N, and M protein Ab. Although the level of viral protein expression was similar in WT, Nik Ϫ/Ϫ , and Ikk␣ Ϫ/Ϫ MEFS, a significant increase in viral protein expression was observed in the Ikk␥ Ϫ/Ϫ MEFs (Fig. 7A). In addition, in contrast to the modest cytopathic effect and cell fusion observed in WT, Nik Ϫ/Ϫ , and Ikk␣ Ϫ/Ϫ MEFs, a significant increase in cytopathic effect and cell fusion was observed in the Ikk␥ Ϫ/Ϫ MEFs (Fig. 7B). These data suggested that IKK␥ mediates anti-viral signaling but the noncanonical/cross-talk pathways do not. By extension, this finding suggests that inhibition of the NIK-IKK␣ signaling pathway may reduce inflammatory chemokine expression, but will not increase viral replication.

DISCUSSION
RIG-I is a major initial intracellular sensor that detects RSV infection and activates the downstream NF-B and IRF3 pathways by complexing with the MAVS adapter (16). In this study, we focused on the mechanistic details for how RSV induces NF-B pathways. Currently it is thought that NF-B is regulated by two separate mechanisms, termed the canonical and noncanonical pathways. The canonical NF-B pathway is IKK␥dependent and liberates RelA from cytoplasmic IB␣ complexes. By contrast, the noncanonical pathway is both NIK-and IKK␣-dependent and results in RelB release from cytoplasmic p100 complexes. Although most stimuli activate either the canonical pathway (TNF/IL-1) or the noncanonical pathway (LIGHT/LT␤), RSV efficiently activates both (5,15,34). The mechanism by which RIG-I couples to the canonical pathway is largely understood, but the mechanism how RSV activates the noncanonical pathways via the NIK⅐IKK␣ kinase complex is unclear. Here we make the surprising findings that RIG-I activates RelA translocation from p100 complexes by the noncanonical NIK⅐IKK␣ subunits in Ikk␥ Ϫ/Ϫ cells. The findings that noncanonical pathway activation is dependent on active RSV replication, an event that results in production of vRNAs, and that this pathway is inhibited upon RIG-I knockdown indicate that RIG-I activation is required and upstream of the noncanonical pathway. These data indicate the existence of an addi- tional RSV-inducible "cross-talk" pathway that mediates translocation of the potent transcriptional RelA transactivator.
To our knowledge, this data is the first to demonstrate that NIK associates with the RIG-I-MAVS signaling complex. NIK is a serine-threonine kinase of the mitogen-activated MAP kinase family known to associate with the TNF receptor-associated factors (TRAFs)-2, and 3, the TRAF-and NIK-associated factor (TNAP), and the IKK␣ kinase (35)(36)(37)(38)(39). TRAF association allows NIK to couple with activated receptors in the TNF superfamily. Additionally, previous work shows that NIK serves as a scaffolding molecule, binding IKK␣, an event that permits IKK␣ to complex with p100 to phosphorylate and initiate p52 formation (11). These multiprotein interactions have been partially mapped to the NIK molecule. The NH 2 terminus of NIK binds TRAF3 and TNAP (38,39), whereas the COOH terminus is known to interact with TRAF2 (amino acids 624 -947 of NIK) and IKK␣ (amino acids 735-947 of NIK) (35)(36)(37). Our deletion experiments indicate that the NIK COOH terminus (amino acids 660 -947) is also required for MAVS interaction. Because of the number of protein interactions, it is highly likely that a macromolecular complex is being formed between NIK, IKK␣, RIG-I, and MAVS to produce a functional signaling complex. More detailed work will be required to identify which proteins directly interact.
MAVS is an essential signal transducer for mediating activated RIG-I signaling. MAVS does not have known enzymatic activity and apparently serves as a site for signaling complex assembly. In this regard, the RIG-I⅐MAVS complex has been shown to be uniquely localized to the surface of mitochondria via a short COOH-terminal transmembrane domain on the MAVS protein (19). Previous work has shown that in the absence of MAVS, cells are unable to activate the canonical NF-Bor the IRF3 signaling pathways in response to dsRNA or viral infections (19 -23). Interestingly, in the absence of mitochondrial targeting, MAVS is unable to associate with RIG-I or mediate its signaling. Our data indicate that mitochondrial localization is required for NIK interaction, because NIK being unable to bind to the COOH terminally deleted MAVS, which aberrantly targets to the cytoplasm (Fig. 2B and Ref. (19)). A similar finding is made for the IKK␣-MAVS interaction (Fig. 3C). The explanation for a mitochondrial requirement for signaling is currently unknown. Our data is the first to demonstrate that NIK association with the RIG-I⅐MAVS com-FIGURE 6. RelA is released from p100 associated complexes. A, IB␣ degradation is IKK␥-dependent. WT, Nik Ϫ/Ϫ , Ikk␣ Ϫ/Ϫ , and Ikk␥ Ϫ/Ϫ MEFs were mock or RSV infected for 24 h. IB␣ was measured using Western immunoblot (top panel); ␤-actin was used as an internal control (bottom panel). Note that IB␣ is degraded in WT, Nik Ϫ/Ϫ , and Ikk␣ Ϫ/Ϫ MEFs, but not significantly in Ikk␥ Ϫ/Ϫ MEFs. B, RelA is released from p100 complexes. WT and Ikk␥ Ϫ/Ϫ MEFs were infected by RSV for 0 and 24 h. p100 was subjected to nondenaturing co-immunoprecipitation (IP) using anti-p100 Ab; nonimmune rabbit IgG was used as nonspecific binding control. RelA association was detected by Western immunoblot (IB) using anti-RelA Ab (top). Specific RelA staining is indicated by a black arrowhead. To monitor p100 recovery, the membrane was re-probed with anti-NF-B2 Ab (bottom). Note that RelA associates with p100 in uninfected cells, but this binding is lost in response to RSV infection. C, RelA release upon noncanonical pathway activation. A549 cells were stimulated with LIGHT or TNF for the indicated times (in h, bottom). NE were prepared and RelB or RelA translocation determined by Western immunoblot. ␤-Actin staining is used as an internal control. Note that LIGHT induces both RelA and RelB translocation, whereas TNF only induces RelA translocation. plex mediates a third signaling pathway, a cross-talk pathway involved in RelA release from p100 complexes.
RelA is sequestered in the cytoplasm through association with discrete IB-like proteins, including IB␣, IB␤, IB⑀, BCL-3, and p100, which serve to function as reservoirs for NF-B (14, 33, 40 -42). The canonical pathway primarily involves stimulus-induced RelA liberation from IB␣-, IB␤-, and IB⑀-sequestered complexes. In this study, we find that RSV induces the degradation of IB␣ in WT, Nik Ϫ/Ϫ , and Ikk␣ Ϫ/Ϫ MEFs, but not in Ikk␥ Ϫ/Ϫ MEFs. This result suggests that RSV-induced IB␣ proteolysis is IKK␥ dependent. However, surprisingly, RSV is still able to activate RelA translocation and transcriptional activation in Ikk␥ Ϫ/Ϫ MEFs. We note that Lin and co-workers (43) also reported a similar 3-fold increase of NF-B-dependent reporter gene expression in Ikk␥ Ϫ/Ϫ MEFs infected with vesicular stomatitis virus. Although the identity of the transactivator was not investigated, they did note that reporter gene activity was inhibited by dominant negative IB␣, suggesting RelA involvement (43). Our study provides a mechanistic link for the viral inducible activation of RelA via forming a complex with RIG-I-MAVS and the noncanonical NIK-IKK␣ kinases. Our working model of the viral-induced cross-talk pathway is shown in Fig. 8.
RelA liberated as a result of p100 processing appears to be under a separate stimulus-specific control than the canonical pathway. For example, this cross-talk pathway was recently described as being downstream of LT␤, a TNF superfamily ligand that also induces IB␣ independent RelA release from cytoplasmic p100 complexes (44). We demonstrate here that LIGHT also induces RelA translocation. Because LT␤ and LIGHT signaling are independent of RIG-I-MAVS, we conclude that several signaling cascades can converge on the cytoplasmic p100⅐RelA complexes including those activated by TNF superfamily of receptors and cytoplasmic RIG-I-like helicases. Finally, we emphasize that although the existence of this cross-talk pathway was initially indicated by the ability of RSV to induce RelA activation in Ikk␥ Ϫ/Ϫ MEFs, this pathway is activated in wild type MEFs and A549 epithelial cells. In both cell types, cytoplasmic RelA is associated with p100, and that p100 processing is induced (as exemplified by p52 accumulation) in response to RSV infection.
Our study indicates that the activation of the noncanonical NIK⅐IKK␣ complexes induces the release of p100-associated RelA as well as the noncanonical RelB⅐NF-B2 (p52) complex (Fig. 8). The prototypical NF-B DNA binding complex is composed of a heterodimer of a transactivating subunit (RelA, RelB) with a DNA binding subunit (NF-B1, NF-B2). Currently we understand the canonical RelA⅐NF-B1 heterodimer is the predominant DNA binding complex in A549 cells detected in gel binding studies, confirmed by our supershift studies (Fig. 3, A  and B). In this complex, RelA is responsible for most of the transcriptional activating properties (Fig. 5 and Refs. 5 and 28). Discussed earlier, RSV-induced RelA is coming from two separate cytoplasmic pools, IB␣/IB␤ and p100. Although p100 processing releases both RelA and RelB in Ikk␥ Ϫ/Ϫ MEFs, we find here by side-by-side siRNA knockdown that RelA, and not RelB, is the major mediator of RSV-induced Rantes, Ip10, and Gro␤ expression.
Our findings that p52 formation is sensitive to the protein synthesis inhibitor cytoheximide (CHX) is similar to those of Mordmuller et al. (31), who demonstrated that p52 processing is coupled to p100 translation in response to LPS and LT. We interpret these findings to indicate that RNA virus infection also activates RelA release from sequestered p100 via a co-translational mechanism. The exact role of the noncanonical RelB⅐NF-B2 (p52) complex in epithelial cells is less clear. Previous work using siRNA-mediated knockdown of p52 indicates that p52 complexes only contribute to a minor degree RSV-inducible cytokine expression (15). In lymphocytes, the RelB⅐NF-B2 complex activates a distinct spectrum of genes, including BAFF, SDF, and BLC-3 (11). Because these genes are not expressed by epithelial cells the role of the noncanonical pathway is presently unclear. Presently the only suggestive data is that the RelB⅐NF-B2 complex modifies the rate of canonical pathway activation in RSV infection (15).
Previously our group showed that inhibition of the canonical NF-B pathway using a peptide that disrupted IKK␥-IKK␤ interaction (the NEMO binding peptide) significantly downregulated the inflammatory reaction in RSV-infected mice, despite robust, increased, RSV replication (3). Recently, IKK␥ was also shown to mediate viral-induced activation of the IRF3 pathway downstream of the MAVS complex; inhibition of IKK␥ was permissive for increased viral replication (43). Shown is a schematic model that shows the signaling cascade downstream of RIG-I that diverges to activate the canonical, noncanonical, and cross-talk pathways. The presence of RSV dsRNA is sensed by the cytosolic RIG-I helicase, which associates with the mitochondrial bound MAVS via heterotypic CARD domain interactions. This complex mediates activation of the typical IKK resulting in RelA release from IB␣-sequestered complexes. In parallel, the NIK⅐IKK␣ complex is activated, resulting in activation of the noncanonical and cross-talk pathway by stimulating the 100-kDa NF-B2 proteolysis. Mito, mitochondria; p65, RelA.
Together, these findings indicate that inhibiting IKK␥ function will affect both the NF-B-dependent inflammatory response as well as the IRF-3-mediated anti-viral response. In our study of the cross-talk pathway, we have made the intriguing finding that a portion of RSV-induced inflammatory cytokine production is mediated by the noncanonical NIK⅐IKK␣ complex. Inhibiting NIK and IKK␣ signaling does not result in increased RSV replication. These characteristics of the NIK⅐IKK␣-mediated cross-talk pathway makes NIK and IKK␣ potential targets for therapeutics targeting acute broncholitis caused by RSV infection.