Expression of an IKKγ Splice Variant Determines IRF3 and Canonical NF-κB Pathway Utilization in ssRNA Virus Infection

Single stranded RNA (ssRNA) virus infection activates the retinoic acid inducible gene I (RIG-I)- mitochondrial antiviral signaling (MAVS) complex, a complex that coordinates the host innate immune response via the NF-κB and IRF3 pathways. Recent work has shown that the IκB kinase (IKK)γ scaffolding protein is the final common adapter protein required by RIG-I·MAVS to activate divergent rate-limiting kinases downstream controlling the NF-κB and IRF3 pathways. Previously we discovered a ubiquitous IKKγ splice-variant, IKKγΔ, that exhibits distinct signaling properties. Methodology/Principal Findings We examined the regulation and function of IKKγ splice forms in response to ssRNA virus infection, a condition that preferentially induces full length IKKγ-WT mRNA expression. In IKKγΔ-expressing cells, we found increased viral translation and cytopathic effect compared to those expressing full length IKKγ-WT. IKKγΔ fails to support viral-induced IRF3 activation in response to ssRNA infections; consequently type I IFN production and the induction of anti-viral interferon stimulated genes (ISGs) are significantly attenuated. By contrast, ectopic RIG-I·MAVS or TNFα-induced canonical NF-κB activation is preserved in IKKγΔ expressing cells. Increasing relative levels of IKKγ-WT to IKKγΔ (while keeping total IKKγ constant) results in increased type I IFN expression. Conversely, overexpressing IKKγΔ (in a background of constant IKKγ-WT expression) shows IKKγΔ functions as a dominant-negative IRF3 signaling inhibitor. IKKγΔ binds both IKK-α and β, but not TANK and IKKε, indicating that exon 5 encodes an essential TANK binding domain. Finally, IKKγΔ displaces IKKγWT from MAVS explaining its domainant negative effect. Conclusions/Significance Relative endogenous IKKγΔ expression affects cellular selection of inflammatory/anti-viral pathway responses to ssRNA viral infection.


Introduction
Activation of the mucosal innate immune response in sentinel epithelial cells is vital to the resolution of mucosal viral infection.
Here, viral replication intermediates are sensed by cytoplasmic pattern recognition receptors, an event that activates two important signaling pathways, one mediated by the NF-kB transcription factor controlling inflammatory cytokine expression, and the second mediated by IRF3 controlling anti-viral type I IFN-a and -b expression. The coordinated expression of these two pathways is responsible for limiting viral replication and activating the adaptive immune response. Significant advances have been made in identifying the structure of these two pathways and their mechanism of control.
Cytoplasmic RNA virus infections, including Sendai (SeV)-, influenza-, Japanese encephalitis-, respiratory syncytial (RSV)-and others, produce 5'triphosphate modified-or ds-RNA products during their replication cycle. These ''non-self'' RNA species are bound by RIG-I, a cytoplasmic DExD/H box RNA helicase [1][2][3]. RNA-bound RIG-I is rapidly polyubiquitylated by E3 ligases (TRIM25 and Riplet/RNF13) that catalyze addition of Lys 63linked ubiquitin polymers into the RIG-I?NH2 terminus [4,5]. Lys 63-ubiquitinated RIG-I, in turn, associates with the mitochondrial antiviral signaling (MAVS) protein via its NH2 terminal caspase recruitment domain (CARD), producing an activated dimeric complex [6]. The assembled RIG-I?MAVS complex, in turn, recruits the TNF Receptor-associated factors (TRAFs)-2, -3 and -6 to multiple TRAF-interaction motifs located in the MAVS proline rich domain [7]. This complex, serves as a scaffold for recruitment of signaling adapters mediating activation of the divergent NF-kB and IRF3 pathways. Downstream activation of the IRF3 pathway results in dramatic upregulation of RIG-I expression and signal amplification.
RIG-I?MAVS activates two distinct pathways controlling NF-kB, termed the canonical and cross-talk pathways. The canonical pathway is mediated by activating the IKK complex, a signaling complex containing the two closely related kinase subunits, IKKa and IKKb, and a third regulatory subunit, IKKc [8]. In the process of IKK activation, IKKc is required for recruiting the catalytic IKK-a and b subunits to activated RIG-I?MAVS, where they are serine phosphorylated in their activation loops. IKK activation effects the phosphorylation and inducible degradation of the IkB inhibitor, resulting in nuclear translocation of the NF-kB/RelA transcriptional activator [9,10]. Here, activated nuclear NF-kB induces expression of inflammatory cytokines such as Grob, IL-6, IL-8 and others [11,12]. By contrast, the cross-talk pathway is mediated by RIG-I?MAVS direct interacting with the IKKa-NF-kB inducing kinase (NIK) complex, in an IKKc -independent manner [13]. This pathway, time-delayed relative to the canonical pathway, results in RelA and RelB release from cytoplasmicsequestered p100. In this way RIG-I?MAVS induces two effector arms converging on NF-kB, producing mucosal inflammation.
RIG-I?MAVS also induces the IRF3 pathway, a pathway controlled by a complex of two IKK-related kinases, TANKbinding kinase 1 (TBK1) and an inducible subunit, IKKe [14]. Here, the TRAF-associated NF-kB activator (TANK) links TBK1and IKKe with upstream TRAF molecules [15,16]. Importantly, IRF3 activation also requires the IKKc signaling adapter; in IKKc-deficient cells, IRF3 activation is also abolished in response to different RNA viruses [15]. As a result of IRF3 activation, the expression of type I IFNs results in a potent upregulation of RIG-I and its ubiquitin ligases, thereby potentiating coordinate signaling by the NF-kB and IRF3 innate signaling responses [17]. In this way, IKKc serves as the final adaptor molecule in RIG-I?MAVS signaling that is shared between the canonical NF-kB and the IRF3 pathways.
In previous work, we identified an alternatively spliced IKKc isoform, termed IKKcD. IKKcD is missing a crucial region in the NH2 terminal coiled coil domain whose functional effect is to couple IKK to distinct upstream signals. Interestingly, IKKcD efficiently mediates cytokine-induced canonical NF-kB activation by associating with the IKKa/b kinases, and mediates TAK/TAB and NIK inducible NF-kB activation, but is resistant to HTLV Tax [18]. Here we investigate its signaling role in response to ssRNA infection. In response to RSV infection, we find that IKKc WT isoform is potently upregulated relative to the IKKcD splice form. In cells only expressing IKKcD, enhanced viral replication and cytopathic effect were seen due to deficient IRF3 signaling and type I IFN production. IKKcD functions as a dominantnegative inhibitor of IRF3 signaling being unable to couple to the TANK-IKKe complex and displaces IKKc from activated RIG-I?MAVS. These data suggest that endogenous expression of IKKcD is involved in balancing inflammatory and anti-viral signaling response to ssRNA infection.

Virus Preparation and Infection
The human RSV A2 strain was propogated in Hep2 cells and purified on sucrose cushion gradient [1]. Cells were infected at an MOI of 1.0 for indicated times. Sendai virus was purchased from Charles River Laboratory. Cells were infected with 100 hemagglutinin units/ml [20].

Transfection
Two million freshly isolated cells were transfected in suspension with indicated plasmids according to the manufacturers recommendation (Amaxa). After transfection, cells were immediately transferred to DMEM and cultured for 24 h before treatment. Luciferase reporter assays were performed as previously described [23]. Data represents mean6SD of triplicate plates of normalized luciferase reporter activity.

Quantitative Real-Time PCR (QRT-PCR)
Total RNA was extracted using acid guanidium phenol extraction (Tri Reagent; Sigma). For IFN and ISG analyses, 1 mgm of RNA was reversely transcribed using Super Script III in a 20 ml reaction mixture. One ml of cDNA product was diluted 1:2, and 2 mL was amplified in a 20 mL reaction mixture containing 12.5 mL of SYBR Green Supermix (Bio-Rad) and 0.4 mM each of forward and reverse gene-specific primers (Supplementary data, Table S1), 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 95uC and then subjected to 40 cycles of 15 s at 94uC, 60 s at 60uC, and 1 min at 72uC in iCycler (BioRAD). The IKKc isoform specific QRT-PCR assays were performed in triplicate in an ABI Prism 7000 Sequence Detection System using the SYBR Green PCR Master Mix (ABI #4364344) as specified by the manufacturer. The final primer concentration was 900 nM (Supplementary data, Table S1). The PCR assays were denatured for 10 min at 95uC, followed by 40 cycles of 15 s at 95uC and 60 s at 60uC. After PCR was performed, PCR products were subjected to melting curve analysis to assure a single amplification product was produced. Quantification of changes in gene expression was using the DDCt method using uninfected cells as a calibrator [1].

Electrophoretic Mobility Shift Assay (EMSA)
A total of 35 mg whole cell extracts (WCEs) were incubated in DNA-binding buffer containing 5% glycerol, 12 mM HEPES, 80 mM NaCl, 5 mM DTT, 5 mM MgCl2, 0.5 mM EDTA, 1 mg of poly (dA-dT), and 100,000 cpm of 32 P-labeled double-stranded oligonucleotide containing NF-kB binding sites [24] and IRF3 binding site [1] in a total volume of 25 mL. After fractionation in TBE acrylamide, gels were dried and exposed to BioMax film (Kodak) for autoradiography.
Native PAGE for IRF-3 Dimer Formation 50 mg protein was fractionated by 7% native acrylamide gel in running buffer containing 1% sodium deoxycholate (Sigma Aldrich) as described [25]. After electrophoresis, proteins were transferred and analyzed by Western Immunoblot.

Selective IKKc Expression in Response to ssRNA Infection
Previously we found that full length IKKc-WT and the alternatively spliced IKKcD transcripts were expressed at 2:1 ratios in uninfected A549 cells [18]. To determine whether IKKc expression is affected by ssRNA virus infection, selective QRT-PCR assays were designed to measure total IKKc isoform expression. Total IKKc was quantified using primer pairs that selectively amplified the region corresponding to Exons 2-3; IKKc-WT was quantified using primer pairs that selectively amplified the Exon 4-5 boundary; and IKKcD was quantified using primers that selectively amplified the Exon 4-6 boundary (Supplementary Table SI online). A549 cells were infected with sucrose cushion purified RSV, and IKKc transcripts quantitated. Strikingly, relative to uninfected cells, total IKKc transcripts were markedly induced 55-fold 6 h after RSV infection and returned to baseline 16 h later ( Figure 1A, top panel). Similarly, full length IKKc-WT was transiently induced 80-fold 6 h after RSV infection ( Figure 1A, middle panel). By contrast, the IKKcD isoforms was only weakly (7-fold) induced by RSV infection at the same time point ( Figure 1A, bottom panel). Together these data indicate that the differential expression of IKKc splice forms is regulated by ssRNA infection.

The Replication of RNA Viruses Is Increased in IKKcD Reconstituted MEFs
To selectively compare the function of IKKc and IKKcD in response to RNA viruses, IKKc 2/2 mouse embryonic fibroblasts (MEFs) were transfected with full length IKKc WT or IKKcD expression vectors and isoform expression quantified by Western immunoblot ( Figure 1B). At equivalent amounts of expression vector, IKKcD expressed at a 2-fold higher level than did IKKc-WT, and was not affected by RSV infection. To determine if IKKc isoform expression affects RSV replication, the expression of RSV proteins were detected in Western immunoblot using a pan-anti-RSV Ab. As expected from their inability to produce type I IFN, RSV replicated to high levels in the IKKc 2/2 cells. In cells expressing IKKc-WT, the level of RSV replication was significantly reduced ( Figure 1C), consistent with the robust IFN production in RSV infected cells and the actions of type I IFN to restrict RSV replication [26]. By contrast, despite the findings that ectopic IKKcD had a slightly higher expression level than that of IKKc-WT, there was a significant increase of RSV proteins produced 16 h after RSV infection ( Figure 1C, compare G, N and M protein expression). Quantification of the RSV N protein by near-infrared scanning (LiCOR Odyssey) showed that the normalized abundance of N was 53 arbitrary units (AU) in IKKc 2/2 MEFs, 24 AU in IKKc-WT expressing cells and 46 AU in IKKcD expressing cells. Similar findings were seen for RSV G and M proteins.
To confirm this result, multinucleated cell (MNC) formation, a consequence of RSV Fusion protein expression, was measured in IKKc 2/2 MEFs expressing either EGFP, EGFP-IKKc-WT or EGFP-IKKcD [18]. MNCs were quantified by scoring 100 EGFPpositive cells in 5 randomly selected images by an observer blinded to the experimental condition. Twenty MNCs were observed in RSV-infected empty vector transfectants, whereas 9 were observed in EGFP-IKKc-WT and 18 in EGFP-IKKcD transfectants. The reduction in MNC formation in IKKc-WT transfectants is highly significant compared to empty vector transfectants (p,0.01, x2 statistic ), whereas the number of MNCs in IKKcD was not different from empty vector. These data indicate that IKKcD is more permissive for RSV replication than IKKc-WT. Similar findings were produced in IKKc 2/2 cells stably expressing IKKc-WT, IKKcD or empty vector (Supplementary Figure S1 online).
We also investigated Sendai virus (SeV) replication in IKKc-WT and IKKcD reconstituted IKKc 2/2 MEFs. Both at 12-and 24 h after infection, a significant increase of Sendai viral protein was observed in IKKc 2/2 and IKKcD reconstituted MEFs, as compared to those reconstituted with IKKc-WT ( Figure 1E). Quantification of the normalized protein abundance is shown in the adjacent graph ( Figure 1E, right). Together these data suggested that IKKcD reconstituted MEFs were defective in restricting viral expression relative to those expressing the IKKc-WT isoform.

IKKcD Transfectants Are Defective in Type I IFN Production
To understand the mechanism for the enhanced viral replication rate in IKKcD-reconstituted MEFs, viral induced expression of type I IFN (IFN-b, -a1 and -a4), were quantified by QRT-PCR. In IKKc 2/2 MEFs, RSV infection did not induce a detectable change in expression of any type I IFN. Conversely, in cells reconstituted with IKKc-WT, RSV induced a 150-fold increase in IFN-b, a 10-fold increase in IFN-a1 and a 350-fold increase in IFN-a4 ( Figure 2A). Strikingly, in IKKcD-reconstituted cells, the expression of all three type I IFNs was significantly less, and for IFNa1, indistinguishable from that of IKKc 2/2 MEFs (Figure 2A).
A major mechanism for IFN induced antiviral activity involves the expression of downstream IFN-stimulated genes (ISGs). To confirm that the attenuated type I IFN production in IKKcD expressing cells was biologically relevant, we next measured ISG expression. In RSV infected cells reconstituted with IKKc-WT, robust 1,200-fold induction of the IFN response factor-7 (IRF7), a 550-fold induction of IFN inducible gene -10 (IP10), and a 350-fold induction of RANTES were observed ( Figure 2B). Conversely, in both IKKc 2/2 MEFs and those reconstituted with IKKcD, ISG expression was significantly reduced ( Figure 2B).
To exclude the possibility that IKKc reconstitution by transient transfection affects cellular signaling in response to ssRNA virus infection, the response of IKKc 2/2 -deficient cells stably expressing IKKc-WT and IKKcD, were investigated [18]. These cells have been previously shown to have intact NF-kB signaling in response to TNFa stimulation and IKKa/b expression [18]. In response to RSV infection, we found that these cells also had defective type I IFN expression (Supplementary Figure S2A online). Similar findings were produced in stable transfectants in response to SeV infection, both in terms of defective type I IFN production as well as impaired ISG expression (Supplementary Figures S2B, S2C  online).
A biological action of epithelial type I IFN production is to induce paracrine activation of the Jak-STAT pathway in neighboring cells to produce a mucosal anti-viral state [27]. In this process, STAT1 is tyrosine phosphorylated and its expression upregulated via a positive feedback loop [28]. We therefore analyzed RSV-induced inducible STAT1 tyrosine phosphorylation and expression in IKKc 2/2 MEFs transfected with either empty-, IKKc-WT and IKKcD expression vectors. The induction of phospho-Tyr 701 STAT1 and upregulation of STAT1 protein were only observed in IKKc-WT reconstituted cells ( Figure 2C). Similar results were observed in response to SeV infection ( Figure 2D). Collectively, these data suggest that, in contrast to IKKc-WT, IKKcD does not effectively differently couple to type I IFN induction resulting in a deficient ISG response after ssRNA virus infection.

IKKcD Is Deficient in Viral Induced IRF3 Activation
Previous studies have demonstrated that IKKc is an essential adapter for IRF3 activation downstream of RIG-I?MAVS [15]. Because IRF3 is a major regulator of type I IFN production, we therefore tested whether IKKc and IKKcD differentially affected viral induced IRF3 or NF-kB transcription. Myc epitope-tagged IKKc and IKKcD were co-transfected in the absence or presence of MAVS along with NF-kB-selective (IFNb PRDII domain) or IRF3selective (PRDIII) luciferase reporter genes into IKKc 2/2 MEFs. MAVS expression was determined by anti-Flag Ab in Western Immunoblot ( Figure 3A, left, top panel) and that of IKKc and IKKcD by anti-Myc Ab in in Western immunoblot ( Figure 3A, left, middle panel). We noted that IKKc isoform expression did not affect MAVS expression.
As expected, MAVS was unable to activate NF-kB-driven luciferase reporter activity in IKKc 2/2 MEFs, and mediated a 4-fold increase in IKKc-WT transfectants ( Figure 3B). Importantly, MAVS activated NF-kB-driven luciferase reporter activity to a slightly greater degree (5-fold) in cells expressing IKKc2D ( Figure 3B, left, top panel). Conversely, although MAVS induced 3.5-fold increase in IRF3-driven luciferase reporter activity in cells expressing IKKc-WT, no detectable induction of IRF3-driven luciferase activity was seen in cells expressing IKKcD ( Figure 3B, right, top panel). A similar experiment was performed with activated form of RIG-I, encoding the NH2 terminal CARD domain (RIG-N). Expression of RIG-N was determined by Western immunoblot ( Figure 3A, top right), and that of co-transfected IKKc isoforms ( Figure 3A, middle right). Consistently with the findings of MAVS, RIG-N was unable to activate NF-kB-dependent reporter activity in IKKc 2/2 MEFs ( Figure 3B, left, bottom panel) but did so in cells expressing either IKKc-WT or IKKcD isoforms, where a 3-fold induction of PRDII was observed. Strikingly, and in a manner consistent with the expression of MAVS, RIG-N activated IRF3dependent transcription only in the presence of IKKc-WT, but was unable to activate IRF3 in IKKcD expressing cells ( Figure 3B, right, bottom panel). Together these data indicated that IKKc-WT mediates both NF-kB and IRF3 pathways, whereas IKKcD is unable to support IRF3 signaling.
To further define this mechanism, we examined whether IRF3 was induced to translocate into the nucleus in SeV-infected IKKc 2/2 deficient MEFs stably expressing IKKc-WT or IKKcD. Sucrose cushion-purified nuclear extracts, free of cytoplasmic markers (tubulin [29]), were assayed by Western immunoblot using an anti-IRF3 Ab. In cells expressing IKKc-WT, nuclear IRF3 was undetectable at 0-and 6 h, but appeared in the nuclear compartment after 12-and 24 h of SeV infection ( Figure 3C). By contrast, in IKKcD expressing cells, no IRF3 was detected in the nucleus ( Figure 3C), despite effective viral replication ( Figure 1E). Next, nuclear extracts from SeV-infected MEFs were assayed for IRF3 DNA binding activity in EMSA using a radiolabeled ISRE site (taken from the ISG15 promoter). SeV induced a specific DNA binding activity 12-and 24 h after infection only in IKKc-WT expressing cells; no DNA binding activity was seen in IKKcD expressing cells ( Figure 3D; this band was previously shown to be DNA sequence specific and contain IRF3 [1]). To further confirm defective IRF3 activation, IRF3 dimer formation was quantified by Western immunoblot of native gel-fractionated whole cell extracts prepared from IKKc 2/2 , IKKc-WT-reconstituted and IKKcDreconstituted MEFs infected for various times by either SeV or RSV. IRF3 dimer formation was detected only in IKKc-WTexpressing cells in response to either type of viral infection ( Figures 3E,F). We conclude that, in contrast to IKKc-WT, IKKcD is unable to mediate IRF3 nuclear translocation, DNA binding, dimerization or transcriptional activation.
We next investigated whether IKKcD was coupled to the NF-kB pathway. First, confocal immunofluorescence experiments were performed for RelA nuclear translocation in IKKc 2/2 MEFs complemented with either EGFP-IKKc or EGFP-IKKcD. Transfectants were then either treated with TNF (30 ng/ml, 1 h) or infected with RSV (MOI = 1, 24 h). Cells were fixed, RelA stained using anti-RelA Ab and transfectants imaged using confocal microscopy. Nuclei were counterstained with DAPI, and the presence of RelA examined in EGFP-expressing cells. In untreated controls, RelA was cytoplasmic in empty vector, IKKc-WT or IKKcD expressing cells ( Figure 4A). In response to TNF stimulation, a strong nuclear concentration of RelA was observed in either IKKc-WT or IKKcD expressing cells but not those transfected with empty vector ( Figure 4B). Similarly, RSV induced RelA nuclear translocation only in either IKKc-WT and IKKcD expressing cells ( Figure 4C).
A hallmark of the activated canonical NF-kB pathway involves cytoplasmic IkBa proteolysis via a ubiquitin proteasome-independent pathway [10], a phenomenon that is IKKc-dependent [13]. To confirm that RSV-induced RelA nuclear translocation was mediated by canonical NF-kB pathway activation, IkBa proteolysis was measured in cytoplasmic extracts using Western immunoblot. In IKKc-WT expressing cells, cytoplasmic IkBa proteolysis is clearly evident 6 h after RSV infection ( Figure 4D), and is resynthesized 24 h after viral exposure via the RelA-IkBa positive feedback loop [30]. In IKKcD expressing cells, cytoplasmic IkBa proteolysis is also observed, although with slower kinetics. Conversely, in nuclear extracts, NF-kB DNA binding increases in IKKc-WT-complemented cells 6 h after RSV infection, at times when cytoplasmic IkBa is degraded, and declines as IkBa is resynthesized, trapping NF-kB back in its cytoplasmic location ( Figure 4D; supershifting experiments in RSV infected IKKc 2/2 MEFs have previously demonstrated this complex to be composed of RelA?p50 complexes [13]). Consistent with the qualitative differences in kinetics of IkBa proteolysis, NF-kB DNA binding increases in IKKcD-expressing cells, peaking at later times and persisting 24 h after infection ( Figure 4D).
We sought to further understand the mechanism for qualitative difference in NF-kB activation in the IKKcD background. Previously we showed that RIG-I is strongly induced in response to RSV infection, and its expression is required for RSV inducible NF-kB activation [1]. Because RIG-I is type I IFN dependent, we examined whether RSV-induced RIG-I upregulation was attenuated in IKKcD-expressing cells by QRT-PCR. Although RIG-I mRNA is induced by over 100-fold in IKKc-WT expressing cells, in both IKKc 2/2 and IKKcD expressing cells, viral inducible RIG-I expression was significantly reduced, accounting, in part, for the reduced NF-kB activation ( Figure 4E). Together, these data indicate that the primary defect in IKKcD signaling is the IRF3 pathway, and the attenuated NF-kB activation is because of secondarily reduced RIG-I expression ( note ectopic RIG-I or MAVS expression can activate the canonical NF-kB pathway, seen in Figures 3A,B).

Defective IRF3 Activation in Cells Expressing Increased IKKcD:IKKc-WT Ratios
Previous work from our lab demonstrated that IKKcD was universally expressed in various ratios in different tissue-and cell types with IKKc-WT [18]. To illustrate, the expression of IKKc and IKKcD was surveyed in 7 different cell types by Western immunoblot, where the 43 kDa IKKcD and 50 kDa IKKc-WT isoforms could be resolved. Both bands are specific as demonstrated by peptide competition experiments (Supplementary Figure S3 online). Both IKKc and IKKcD isoforms are expressed from ,2:1 IKKc:IKKcD ratios in HepG2 or HEK293 cells, to 1:2 IKKc:IKKcD ratios in WT MEF cells, to 1:4 ratios in HeLa S3 cells ( Figure 5A). We postulated that cells natively expressing high amounts of IKKcD could be defective in IRF3 activation. We therefore assayed IRF3 dimer formation in Hela S3 cells by Western immunoblot after native gel fractionation. Despite the finding that RSV replicated in Hela S3 better than that of Hela CCL2 cells (Supplementary Figure S4 online), HeLa S3 cells showed no evidence of IRF3 dimer formation. By contrast, efficient IRF3 dimer formation was observed in HeLa CCL2 cells ( Figure 5B). To confirm HeLa cells could efficiently couple to the canonical NF-kB pathway, the response to TNF was measured. We observed that Hela S3 cells have an intact canonical NF-kB activation pathway indicated by rapid cytoplasmic IkBa proteolysis and appearance of nuclear NF-kB DNA binding activity in EMSA ( Figure 5C these complexes have been extensively characterized by competition and supershift as containing RelA?p50 heterodimers [11,31]).
To confirm that HeLa S3 cells had an otherwise intact RIG?MAVS-IRF3 pathway, HeLa S3 were complemented with IKKc-WT. For this experiment, HeLa S3 cells were co-transfected with empty(pcDNA), IKKc-WT, or IKKcD expression vectors and the IRF3driven IFNb PRDIII luciferase reporter gene. Cells were then RSV infected, and luciferase activity measured 24 h later. RSV was unable to activate PRDIII luciferase reporter activity in the cells transfected with empty vector or in those reconstituted with IKKcD ( Figure 5D). However, RSV induced a 4-fold increase of IRF3-dependent reporter gene activity in the cells transfected with IKKc-WT; by contrast empty vector-and IKKcD transfected cells did not induce IRF3 dependent transcription ( Figure 5D). These data suggested IKKc-WT complemented the IRF3 signaling defect in Hela S3 cells, cells selectively defective in IRF3 signaling but having an otherwise have an intact canonical NF-kB pathway.

Varying Ratios of IKKcD (Keeping IKKc Constant) Affects Viral Induced Type I IFN Production
Our findings in HeLa S3 cells suggested that the endogenous ratio of IKKc-WT:IKKcD is one determinant of type I IFN production in response to ssRNA virus infection. To more fully explore this hypothesis, we conducted an experiment varying the ratios of IKKc-WT:IKKcD in the IKKc 2/2 MEF background. In this experiment, we fixed the total amount of IKKc to a constant level while only changing the IKKcWT:IKKcD ratio ( Figure 6A). Type I IFN production was then quantified in response to RSV infection using QRT-PCR. In cells expressing only IKKc-WT, a 3,000-fold induction of IFNb transcripts were observed in response to RSV infection, a response that was reduced in a dose-dependent manner upon the expression of IKKcD ( Figure 6B). Under expression conditions where IKKcD was the predominant isoform, type I IFN production was significantly blunted ( Figure 6B). Similar findings were observed for the ,90-fold induction of IFNa1, 350-fold induction of IRF7 and 500-fold induction of IFNa4 mRNAs ( Figure 6B). As expected from the blunted RIG-I induction in cells expressing IKKcD (see Figure 4E), the RSV inducible expression of NF-kB dependent Grob and A20 genes were also blunted. Together we conclude that the relative IKKc-WT:IKKcD ratio, independent of changes in total IKKc abundance, controls cellular IRF3-IFN responsiveness to ssRNA virus infection.

IKKcD Overexpression Is a Dominant Negative Inhibitor of IKKc-WT Mediated IFN Production
Earlier we showed that IKKcD is a dominant negative inhibitor of HLTV-I Tax induced NF-kB activation, despite the ability of both isoforms to bind HTLV-I Tax protein [18]. To explore whether IKKcD functioned as a dominant negative inhibitor of type I IFN production, we performed an alternative experiment where IKKcD was expressed in increasing amounts in the presence of a constant amount of IKKc-WT in IKKc 2/2 MEFs (Figure 7). Type I IFN production was quantified in RSV using QRT-PCR. Co-transfection of IKKcD at 2.5 mg reduced type I IFN expression by ,50% that was further reduced in a dosedependent manner up to 7.5 mg expression plasmid, where the response was almost completely abolished.
Together we conclude that the relative IKKc-WT:IKKcD ratio primarily mediates IRF3-IFN responsiveness in response to RNA virus infection through its dominant negative effect.

IKKcD Is Defective in Recruiting the TBK1 Adapter, TANK
Previous work has demonstrated that an interaction between the TBK1 adapter, TANK, and IKKc is required for coupling RIG-I?MAVS complex to IRF3 activation [15]. Because IKKcD fails to activate IRF3, we first tested whether IKKcD associates with MAVS. For this purpose, Myc epitope-tagged IKKc-WT or IKKcD expression vectors were co-transfected with Flag-tagged MAVS and subjected to nondenaturing coimmunoprecipitation using anti-Myc Ab. MAVS association was detected by immunoblot using anti-Flag Ab. We observed that both IKKc and IKKcD effectively bound to MAVS ( Figure 8A). We next asked whether IKKcD was able to bind TANK. In this experiment, Myc epitope tagged IKKc-WT or IKKcD expression vectors were co-transfected with V5-epitope tagged TANK. A nondenaturing co-immunoprecipitation assay experiment was then performed using anti-Myc Ab as the primary immunoprecipitating Ab, followed by Western immunoblot of the immunoprecipitates using anti-V5 Ab. V5-TANK was only observed in immunoprecipitates in cells expressing IKKc-WT, but not IKKcD ( Figure 8B, top panel). Equivalent amounts of IKKc-WT and IKKcD were seen in the immunoprecipitates ( Figure 8B, bottom  panel).
TANK is an external adaptor that mediates the recruitment of the atypical IKK, IKKe, to IKKc. Because IKKcD is unable to bind TANK, we investigated whether IKKcD was defective in IKKe recruitment. For this purpose, either Flag epitope tagged IKK-a, -b or -e was co-transfected with Myc tagged IKKc-WT or IKKcD and subjected to nondenaturing coimmunoprecipitation. To demonstrate the essential role of TANK for recruiting IKKe, V5-labeled TANK was also co-transfected with IKKe. After IKKc isoforms were precipitated using anti-Myc Ab, the association of respective IKK was detected by anti-Flag Ab. Consistent with our previous work, IKKc-WT associates with IKK-a and IKK-b [18]. IKKe did not bind to IKKc-WT in the absence of TANK, but in cells cotransfected with TANK, IKKe could bind ( Figure 8C, left  panel). Conversely, although IKKcD bound IKKa and IKKb, it did not recruit IKKe, even in the presence of TANK ( Figure 8C, right panel). Based on these data, we conclude that IKKcD is

The Dominant Negative Effect of IKKcD Is Mediated by Displacement of IKKc-WT from MAVS
To determine the mechanism for the inhibitory effect of IKKcD (Figure 7), we sought to determine whether IKKcD could displace the IRF3 signaling-competent IKKc-WT isoform from the activated MAVS complex. For this experiment, Myc epitope tagged IKKc-WT was co-transfected with FLAG-tagged MAVS in the absence or presence of IKKcD. Lysates were then subjected to nondenaturing coimmunoprecipitation using anti-FLAG, and IKKc isoform association was determined by Western blot using anti-Myc Ab. In the absence of IKKcD, IKKc-WT was associated with MAVS, however, the expression of IKKcD completely displaced IKKc-WT from the MAVS complex ( Figure 9).

Discussion
IKKc was identified as an essential regulatory subunit of the canonical IKK complex because IKKc deficient cells were unable to activate NF-kB in response to most known stimuli [19,32]. IKKc plays multiple adapter roles in IKK activation through its ability to organize the assembly of IKKs into the activated high molecular weight complex [33,34], bind ubiquitylated upstream signaling adapters [32,35,[35][36][37], and recruit the IkBa inhibitor into the activated IKK complex [33]. Through these activities, IKKc forms a molecular bridge between IKK, its upstream activators, and its substrate. In the innate immune response pathway, IKKc recruits IKK2a and -b catalytic complexes to RIG-I?MAVS, resulting in IkBa proteolysis and canonical NF-kB activation. Similarly IKKc is a binding target for TANK, an adapter that links TBK1 and IKKe, two key kinases controlling IRF3 activation [16,38]. In this manner, IKKc is the final common shared signaling adapter upstream of the divergent IRF3 and the canonical NF-kB pathways. Both IRF3 and NF-kB signaling play important, yet distinct, roles in anti-viral and inflammatory signaling in response to ssRNA viral infection. For example, IRF3 is a major mediator of type I IFN production, important in mucosal anti-viral response; in IKKc 2/2 cells, the replication of RNA viruses is significantly increased due to the inability to produce type I IFNs [15]. Similarly, NF-kB signaling is important in initiating mucosal inflammation and the adaptive immune response. The coordination and timing of these two arms of innate immune signaling response may affect the resolution of viral infection, yet the mechanisms for selection of these two pathways are not yet fully elucidated.
In this study, we have extended our previous work describing the signaling properties of a ubiquitously expressed IKKc alternative splice product. Previously we reported that IKKcD is defective in mediating HTLV-Tax recruitment, but more efficiently mediates NF-kB activation by IKKa/b and MAP3Ks, NIK and TAK/TAB [18]. These earlier studies showed that IKKcD efficiently binds to IKKa/b isoforms in coimmunoprecipitation experiments, and induces IKK kinase activity to a greater degree than does IKKc-WT. Here we find that IKKcD is primarily defective in IRF3 signaling, reducing type I IFN production and ISG signaling by displacing IKKc-WT from MAVS complex with an isoform deficient in recruiting TANK-IKKe. Moreover, in cells naturally expressing high levels of endogenous IKKcD to IKKc-WT ratios are defective in viral inducible IRF3 activation but respond via cytokine induced NF-kB activators. Strikingly, IKKc-WT mRNA expression is highly inducible by ssRNA infection relative to IKKcD, suggesting that the cellular ratios of the two isoforms are dynamic. The mechanisms for this induction, and consequences in signaling will require further exploration. Together these findings indicate that relative endogenous expression of IKKcD isoforms may produce IKK complexes that differentially couple to distinct upstream signals, affording heterogeneity in cellular responses to otherwise similar activating stimuli ( Figure 10).
IKKc is encoded by a 10-exon-containing gene located at chromosome Xq28 [39]. Mutations in the IKKc gene including different truncations of the IKKc protein have been linked to the human syndromes of incontinentia pigmenti and anhidrotic ectodermal dysplasia associated with immunodeficiency [40]. Although human IKKc transcripts containing an alternatively spliced first (noncoding) exon have been deposited in GenBank (AI24572, AF091453), these alternatively spliced transcripts encode wild type IKKc [39]. IKKcD is the only alternative splice form known that affects the IKKc coding region, and is caused by occlusion of exon 5. Using both 2D gel electrophoresis and a reverse transcription-PCR assay that distinguished the two isoforms, we found that IKKcD is widely expressed in normal human tissues in various relative ratios with fully spliced IKKc-WT [18]. For example in normal breast and cervical tissue, IKKcD is the predominant isoform detected, whereas in normal liver and lung IKKc-WT is predominant (HeLa S3, derived from a human cervical tumor maintains this IKKcD predominance). The findings that IKKcD expression reduces IRF3 signaling and type I IFN response in response to ssRNA virus infection may yield new insight for tissue-selective differences in anti-viral responses and tissue tropism in RNA virus infections [41].
We note that viral replication in IKKcD-expressing MEFs is enhanced relative to those cells expressing IKKc-WT, but not to the degree seen in IKKc 2/2 MEFs ( Figure 1C). In these experiments, viral inducible expression of the major type I IFNs, IFN-b, -a4 and -a1, is nearly absent and indistinguishable from that in IKKc 2/2 cells. That this degree of inhibition is biologically significant is demonstrated by the lack of detectable phospho-STAT formation or STAT autoregulation (Figure 2). It is surprising, then, that virus does not replicate in IKKcD2expressing cells to a similar degree as that seen in IKKc 2/2 cells. One interpretation of these findings is that IKKcD expressing cells, residual NF-kB activity may play a  role in anti-viral response. In this regard, we note that others have suggested that NF-kB signaling mediates anti-viral activity [42]. Inhibition of NF-kB signaling in response to hPIV or RSV infection resulted in enhanced viral replication in an IFN-independent manner. The selective deficiency in IRF3 signaling in IKKcD expressing cells may allow the isolated study or identification of this potential anti-viral pathway.
The data in this study and that of our previous work have indicated that IKKcD is competent to mediate signals through the canonical NF-kB pathway. This conclusion is supported by multiple lines of evidence: 1. the ability of ectopic RIG-I or MAVS to activate NF-kB dependent gene expression in IKKcD expressing MEFs ( Figure 3); 2. the ability of RSV infection to induce IkBa proteolysis in IKKcD expressing MEFs ( Figure 3); 3. the ability of TNF to induce IkBa proteolysis in Hela S3 cells predominately expressing the IKKcD isoform ( Figure 4); and 4. the ability of catalytic IKKa/b isoforms to bind IKKcD in nondenaturing coimmunoprecipitation assays (Figure 8). Previous protein interaction mapping studies that show the IKKa/b interaction motif lies in a 119 aa region in the far NH2 terminus of IKKc [37,43,44], upstream of residues encoded by exon 5. Despite the ability to mediate productive IKKa/b binding and interaction, we note the qualitative difference in the kinetics of IkBa proteolysis in IKKcD expressing MEFs induced by RSV infection (Figure 4). Our data suggest that the slower kinetic response of NF-kB activation in IKKcD expressing cells is due to reduction in IRF3-IFN-RIG-I cross talk pathway.
Viral activation of RIG-I?MAVS signaling is thought to occur in two sequential phases. The first phase is mediated by low ambient concentrations of RIG-I early in the course of viral infection, where viral RNA is in low abundance [45]. Here, initial activation of IRF3 is mediated by the TRIM25 or Riplet/RNF135 ubiquitin ligases, inducing RIG-I ubiquitylation vis Lys 63-linked ubiquitin polymers, a modification that promotes its association with MAVS, initiating downstream IFN production [46]. The subsequent potent upregulation of RIG-I expression induced by this first wave of IFN [46] produces amplification of the signaling pathway. We observed that RSV-induced RIG-I upregulation is attenuated in IKKcD expressing cells, nearly to that seen in IKKc 2/2 MEFS (Figure 3). Taken in context with our earlier work showing that IFN-induced RIG-I upregulation is required for RSV-induced NF-kB activation [47], these data may suggest a mechanism for cross-talk between the IRF3 and NF-kB pathway in viral induced inflammation. In the absence of IRF3-IFN signaling, seen in IKKcD expressing cells, reduced RIG-I upregulation may result in delayed and or attenuated NF-kB activation.
Secondary structure predictions of IKKcD reveals that the occlusion of exon 5 affects the COOH terminus of an extended NH2 terminal coiled-coil motif at aa 174-224, a motif important in protein-protein interaction [18]. This alternative splice variant exhibits differential signaling by binding distinct signaling molecules. In this regard, IKKc is essential for TNF signaling via its ability to recruit IKK to activated TNF receptors, mediate interaction with upstream MAP3Ks, and directly and/or recruit IkBa into the activated IKK for stimulus-induced phosphorylation. Recent work has shown that this scaffolding function of IKKc may be due to its inducible post-translational modification via a unique chemistry of head-to-tail ubiquitin polymers catalyzed by the LUBAC ubiquitin ligase complex. These inducible ubiquitin polymers enhance the binding of upstream MAP3Ks, that phosphorylate associated catalytic IKKa/b subunits, resulting in their activation [36]. Interestingly, the LUBAC-mediated ubiquitin polymerization is involved in TNF inducible NF-kB signaling, but not by the related cytokine, IL-1. The role of posttranslational modifications of IKKc in response to ssRNA viral infection is not known. We note from co-immunoprecipitation experiments, that IKKcD associates with MAVS in the absence of ssRNA infection ( Figure 8A), suggesting to us, that IKKc-MAVS complex may be a stable, pre-formed complex whose activity is initiated by binding activated RIG-I.
Our findings that IKKcD displaces IKKc-WT from the MAVS complex explains the dominant negative effect of IKKcD to reduce IRF3 activation. IKKcD is defective in TANK binding and consequently recruiting downstream TBK1?IKKi, a kinase complex essential in IRF3 phosphorylation. Like IKKc, TANK is itself a scaffolding protein responsible for recruiting IKKe and TBK1 into an activated complex [38]. Structure-function studies of TANK have revealed a C2H2-type zinc finger in the TANK COOH terminus essential for IKKc association [48]. Conversely, sequential mutagenesis and mapping studies on IKKc have identified aa 150-250 aa as the TANK binding domain [15,38]. Our study is consistent with these findings where IKKcD, lacking exon5-encoded 174-224 aa, is unable to bind TANK. In the absence of TANK binding, in IKKcD expressing cells, RIG?I MAVS is unable to signal to the IRF3 pathway, induce type I IFN expression or activate ISG signaling.
In summary, our findings reveal ubiquitously expressed IKKc splice variant differentially couples IKK signaling to the IRF3 pathway and the induction of type I IFNs. Our studies further indicate that the relative level of expression of IKKc splice forms affects IFN-mediated antiviral signaling in the host cell and may affect viral tissue tropism. Manipulation of the expression of these two isoforms may provide a mechanism to modulate the two arms of the innate immune pathway where preferential expression of NF-kB inflammatory signaling or IRF3 induced anti-viral signaling would be desired.  IKKc 2/2 MEFs stably reconstituted with IKKc-WT or IKKcD were RSV infected for 16 h (MOI = 1). Total RNA was extracted and QRT-PCR was conducted using probes for IFN-b, -a1, -a4. (b) WT MEFs, or IKKc2/2 MEFs stably reconstituted with IKKc-WT or IKKcD were Sendai virus infected for 16 h. Total RNA was extracted and QRT-PCR was conducted using probes for IFN-b, and -a4. (c) Same experiment as in (b) where QRT-PCR was performed with probes for IRF7, IP10 and RANTES. Found at: doi:10.1371/journal.pone.0008079.s003 (5.61 MB TIF) Figure S3 Antibody specificity. The staining specificity of anti-IKKc Ab was evaluated using peptide preadsorption. Anti-IKKc Ab was preadsorbed with nothing or 10-fold molar excess of recombinant purified GST-IKKcD-WT, and used as primary Ab in Western immunoblot. Both bands are reduced by 50%.