DDX3 is a DEAD-box RNA helicase that we and others have previously implicated in antiviral immune signalling pathways leading to type I interferon (IFN) induction. We previously demonstrated that it directly interacts with the kinase IKKε (IκB kinase ε), enhances it activation, and then facilitates phosphorylation of the transcription factor IRF3 by IKKε. However, the TLR7/9 (Toll-like receptor 7/9)-mediated pathway, one of the most physiologically relevant IFN induction pathways, proceeds independently of IKKε or the related kinase TBK1 (TANK-binding kinase 1). This pathway induces type I IFN production via the kinases NIK (NF-κB-inducing kinase) and IKKα and is activated when plasmacytoid dendritic cells sense viral nucleic acids. In the present study, we demonstrate that DDX3 also directly interacts with IKKα and enhances its autophosphorylation and -activation. Modulation of DDX3 expression consequently affected NIK/IKKα-mediated IRF7 phosphorylation and induction of type I interferons. In addition, alternative NF-κB (nuclear factor-κB) activation, another pathway regulated by NIK and IKKα, was also down-regulated in DDX3 knockdown cells. This substantially broadens the effects of DDX3 in innate immune signalling to pathways beyond TBK1/IKKε and IFN induction. Dysregulation of these pathways is involved in disease states, and thus, our research might implicate DDX3 as a potential target for their therapeutic manipulation.
Several DExD/H-box RNA helicases, including DDX3, have recently been implicated to participate in innate antiviral responses mediated by pattern recognition receptors (PRRs), either by directly acting as receptors for viral nucleic acids or by functioning as downstream signalling intermediates . Our previous work has shown that DDX3 can act as a multifunctional adaptor molecule in the RIG-I (retinoic acid-inducible gene I) signalling pathway [2–4]: it directly interacts with the kinase IKKε (IκB kinase ε) and enhances its autophosphorylation and -activation. IKKε also directly phosphorylates DDX3 at several serine residues in its N-terminal tail region, with phosphorylation of serine 102 being required for the subsequent recruitment of its substrate IRF3 (interferon regulatory factor 3) into the complex and subsequent induction of type I interferons (IFNs) . Our data therefore suggested that DDX3 acts as a signalling adaptor molecule that facilitates the co-ordinated activation of IKKε and its substrate IRF3 . Other groups have confirmed the role of DDX3 in IFN induction, but have placed it at different levels in the signalling pathways [5–9]: it was suggested that DDX3 can directly bind to the ifnb promoter and act as a transcriptional regulator , and also that DDX3 acts as a receptor for HCV RNA and is placed upstream of MAVS (mitochondrial antiviral signalling) , the crucial mitochondrial adaptor molecule in the RIG-I pathway.
While most innate immune signalling pathways leading to type I interferon induction, including those mediated by Toll-like receptor (TLR) 3, TLR4, the RIG-like helicases (RLHs) and cytosolic DNA receptors, converge on IKKε/TBK1 (TANK-binding kinase 1) for phosphorylation of IRF3 and IRF7, it was shown that TLR7 and TLR9 signalling to IFN induction is independent of these two kinases [10,11]. TLR7 and TLR9 detect viral nucleic acids (single-stranded RNA and unmethylated CpG DNA, respectively) in the endosomal lumen and trigger the induction of large amounts of type I interferons, in particular IFN-α . This pathway is mainly active in plasmacytoid dendritic cells due to their expression of TLR7 and TLR9 and high constitutive levels of the transcription factor IRF7. IRF7 activation in this pathway is mediated by NIK (NF-κB-inducing kinase) and IKKα . NIK facilitates IKKα activation downstream of Myd88 and TRAF6 (TNF receptor-associated factor 6), conferring on it the ability to phosphorylate IRF7 and activate type I interferon promoters [13,14]. TLR7/9-mediated IFN production from plasmacytoid dendritic cells (pDCs) is thought to be highly relevant during viral infections , and thus, we investigated here whether DDX3 also contributes to this physiologically important IKKε/TBK1-independent IFN induction pathway.
In addition, NIK and IKKα also co-operate to activate the alternative NF-κB (nuclear factor-κB) pathway, which leads to processing of p100 to the active p52 NF-κB subunit . This pathway is mainly activated downstream of CD40 and related receptors, and has been shown to play a role in the development of secondary lymphoid organs, bone metabolism, survival and maturation of B-cells and thymic deletion of autoimmune T-cells. Hyperactivation of the p52 pathway has been linked to cellular transformation and the emergence of various haematopoietic and solid tumours . It has also been shown that the alternative NF-κB pathway is activated downstream of RIG-I, where it contributes to the induction of chemokines and antiviral immune responses [18,19].
Thus, IKKα is of major importance for these two immunorelevant signalling pathways, and we therefore aimed to investigate whether DDX3 regulates IKKα activation in a similar manner to IKKε . The data we present here demonstrate that DDX3 can indeed directly enhance IKKα activation and that it is a positive regulator of NIK–IKKα-mediated IRF7 activation and the alternative NF-κB pathway. This further broadens the involvement of DDX3 in innate immune signalling pathways.
Plasmids, recombinant proteins and antibodies
DDX3 and K7 expression constructs (in pCMV-myc, pCMV-Ha or pHisParallel2) have been described in our previous studies [3,4]. IKKα-Flag and NIK-Flag were provided by Prof. Luke O'Neill (Trinity College Dublin), and Flag-IRF7 was provided by Prof. Kate Fitzgerald (University of Massachusetts). GST (glutathione-S-transferase)-IRF7 (aa 468–503) peptides and constructs were kindly provided by Prof. John Hiscott and Dr Qiang Sun (McGill University, Montreal). Purified recombinant protein kinase GST-IKKα was purchased from Millipore and recombinant GST-NIK was purchased from Proqinase (Freiburg, Germany). The antibodies used were anti-Flag-M2 mAb (Sigma–Aldrich), anti-Myc mAb clone 9E10 (Sigma–Aldrich), anti-HA (hemagglutinin) mAb (Covance), anti-DDX3 (Santa Cruz or Bethyl Laboratories), anti-p52/p100 (Millipore), anti-IRF7 (Santa Cruz), anti-IKKα, anti-phospho-Ser176/180-IKKα, anti-phospho-Ser 866/870-p100, and anti-phospho-Ser471/472-IRF7 (all Cell Signaling Technology), anti-His (Sigma–Aldrich), and anti-GST (Promega). Secondary AlexaFluor488- and AlexaFluor594-coupled antibodies were purchased from Invivogen.
Cell culture and transfection
HEK293T and HEK293_TLR7 cells (Invivogen) were maintained in DMEM, and THP1 cells were maintained in RPMI. Media were supplemented with Glutamax™, 10% foetal calf serum, and gentamycin. Transient DNA transfections for immunoprecipitations and pulldowns were performed using the calcium phosphate method. The total amount of DNA was kept constant by the addition of empty vector. Transfection for reporter gene assays was performed in 96-well plate format using Genejuice™ with a total of 230 ng of DNA.
Short hairpin RNA knockdown of DDX3
Lentiviral particles were generated by transfecting HEK293T cells with either pTRIPZ shDDX3 (V2THS_228965, Thermo Scientific) or the corresponding non-silencing control (NSC), and the packaging vectors pSDAX2 and pMD2 in the presence of chloroquine sulfate. Lentivirus-containing supernatants were harvested and concentrated using PEG precipitation. HEK293, HEK293_TLR7, or THP1 cells were transduced with shDDX3 or NSC lentivirus at an MOI of 10. Transduced cells were selected in medium containing 3 µg/ml puromycin. For use in experiments, cells were cultured in the absence of puromycin and DDX3 knockdown was induced by the addition of 0.5 µg/ml doxycycline for 24–48 h. Successful knockdown of DDX3 was confirmed by real-time PCR and Western blot (WB) analysis.
His-tagged DDX3 was purified from Escherichia coli BL21(DE3) cultures transformed with relevant expression constructs using Nickel-Agarose beads. For pulldowns, equal amounts of the different His-tagged DDX3 truncations were used, as estimated by SDS–PAGE and Coomassie staining prior to use. Cell lysates containing Flag-tagged protein expressed in HEK293T cells, or 0.1 µg of GST-IKKα (Millipore) or GST-NIK (Proqinase), were then incubated with His-tagged DDX3 immobilised on Nickel-Agarose. Protein complexes were precipitated and washed thoroughly, before being subjected to SDS–PAGE and WB analysis.
Co-immunoprecipitation assays were performed from cell lysates of transiently transfected HEK293T or HEK293_TLR7 cells, which were harvested 24 h after transfection. Cells were lysed in IP (immunoprecipitation) lysis buffer (50 mM Hepes pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% NP-40, 10% glycerol, 10 mM NaF, 1% Triton X-100, 10 mM DTT, and protease inhibitors). Cell lysates were incubated with 20 µl of Flag-M2 Agarose (Sigma–Aldrich) or Protein G agarose that had been pre-coupled with relevant antibodies at 4°C overnight and blocked with 5% bovine serum albumin. The immunoprecipitated protein complexes were washed thoroughly and then eluted by boiling in Laemmli sample buffer, before being subjected to SDS–PAGE and WB analysis. Endogenous co-immunoprecipitation between DDX3and IKKα was carried out from lysates of CLO75-stimulated HEK293_TLR7, and a matched isotype control antibody was used as a control for non-specific pulldown.
Immunofluorescent staining and confocal microscopy
HEK293 cells were seeded on sterile glass cover slips in six-well plates and transfected with expression constructs as indicated using the calcium phosphate method. After 24 h, they were washed, fixed with 4% paraformaldehyde, and permeabilised with 0.5% Triton X-100. After blocking with BSA, cells were stained with the appropriate primary antibodies, followed by Alexa-Fluor488- or AlexaFluor594-labelled anti-species secondary antibody (Invitrogen). After staining, cells were mounted in SlowFade Gold Antifade Reagent containing DAPI (Invitrogen) on microscope slides and sealed using clear nail varnish. Cells were viewed using an Olympus Fluoview Confocal Microscope, and analysed using the Olympus Fluoview FV10-ASW software. Cells in multiple different frames were examined and recorded in each of two repeat experiments. The determination of Pearson's correlation coefficients (PCCs) was carried out using the Fluoview software on transfected cells.
In vitro kinase assays
In vitro kinase assays were carried out with 55 ng of recombinant GST-IKKα (Proqinase). This was incubated with purified recombinant substrates in kinase assay buffer (50 mM Hepes, pH 7.5, 5 mM glycerophosphate, 2 mM dithiothreitol, 0.1 mM Na3VO4, and 10 mM MgCl2) in the presence of 20 µM unlabelled ATP for 30 min at 30°C. Samples were analysed by SDS–PAGE and WB analysis with phospho-specific antibodies. Alternatively, kinase assays were carried out in the presence of 10 µM unlabelled ATP and 5 µCi of [γ32P]-ATP. Samples were then analysed by SDS–PAGE and autoradiography.
Reporter gene assays
Ifnb or Ifnα4 reporter assays were carried out in HEK293 or HEK293_TLR7 cells in 96-well plate format. Cells were transfected with reporter constructs and expression vectors using GeneJuice (Merck/Millipore). About 60 ng of ifn promoter Firefly luciferase reporter gene construct was used in conjunction with 20 ng of a pGL3-Renilla luciferase construct (ifnb reporter and pGL3-Renilla provided by Andrew Bowie, Trinity College Dublin, ifna4 reporter provided by Pamela Osterlund, National Public Health Institute Finland). The total amount of DNA transfected was kept constant at 230 ng/well by the addition of matching empty vector DNA. Firefly and Renilla Luciferase activity was measured 24 h after transfection. Renilla Luciferase readings were used to normalise for transfection efficiency.
Real-time PCR analysis
RNA was isolated from stimulated and unstimulated THP1 NSC and shDDX3 cells using a Macherey Nagel RNA isolation kit according to manufacturer's instructions. cDNA was synthesised using iScript Reverse Transcriptase (Bio-Rad), and cDNA concentration was then adjusted to 200 ng/µl. Real-time PCR analysis was conducted using Evagreen® Mastermix (Metabion) according to manufacturer's instructions, using an annealing temperature of 57°C for IFNβ and 60°C for BAFF, HPRT, and DDX3 primers. Primer sequences are available upon request. Relative expression levels were calculated using the ΔΔCt method with HPRT as the housekeeping gene and unstimulated NSC cells as the reference. DDX3 knockdown of between 60 and 90% was achieved in individual experiments.
HEK-Blue™ IFN assay
Supernatants were collected from CL075-stimulated THP1 cells and added to HEK-Blue IFNα/β sensor cells (Invivogen) plated in 96-wps. SEAP production from the exposed reporter cells was then measured using QUANTI-Blue™ reagent. SEAP production correlates to IFNα/β levels in the supernatant.
Data representation and statistical analysis
All data shown are representative of at least two independent repeat experiments, most are representative of three to five repeat experiments. Luciferase reporter gene assays are representative of at least three independent repeat experiments. Statistical analysis was carried out using a paired Student's t-test unless indicated otherwise. For clear representation, WBs have been cropped. Where a black box has been placed around WB panels, this indicates that they were taken from the same original autorad exposure and can thus be compared for intensity. Any adjustments for brightness and contrast that have been applied have been applied evenly across the whole (boxed) panel.
DDX3 interacts directly with IKKα
We first tested whether DDX3 and IKKα physically interact. Initially, we overexpressed Flag-tagged IKKα and myc-tagged DDX3 in HEK293T cells and immunoprecipitated IKKα with a Flag antibody. DDX3 clearly co-immunoprecipitated with Flag-IKKα, but was not pulled down by the Flag antibody in the absence of Flag-IKKα (Figure 1a). In this experiment, we also observed that overexpression of IKKα induced a higher-molecular-weight form of DDX3 (Figure 1a, middle panel) that likely represents a phosphorylated form of DDX3. We next carried out a similar co-immunoprecipitation experiment with overexpressed Flag-tagged NIK and myc-tagged DDX3 to determine whether DDX3 also interacts with NIK, the kinase that can activate IKKα in the alternative NF-κB pathway and downstream of TLR7, 8, and 9. DDX3 clearly co-immunoprecipitated with Flag-NIK, but was not pulled down by the Flag antibody in the absence of Flag-NIK (Figure 1b). Endogenous IKKα also strongly co-immunoprecipitated with NIK (Figure 1b, bottom panel), meaning that the observed interaction between DDX3 and NIK could be indirect and bridged by IKKα. Next, we therefore conducted pulldown experiments with purified recombinant proteins. Recombinant His-tagged DDX3 purified from E. coli was used in pulldowns with recombinant GST-tagged IKKα. In these pulldowns, we observed direct binding of GST-IKKα to His-tagged DDX3 coupled to Ni-Agarose (Ni-Ag) beads (Figure 1c). IKKα did not bind to empty Ni-Ag beads that were used as a control. This suggests that DDX3 and IKKα can directly interact with each other. Similar pulldown experiments with purified GST-NIK were inconclusive, and thus, we are unable to conclude whether DDX3 and NIK also directly interact or whether the interaction is bridged by other proteins, such as IKKα. In addition to these co-immunoprecipitation and pulldown experiments, we also visualised the cellular localisation of co-expressed myc-DDX3 and Flag-IKKα/NΙK using immunofluorescent staining and confocal microscopy. DDX3 and IKKα co-localised in the cytoplasm of transfected cells with an average PCC of 0.63. When Flag-NIK was also co-expressed, the staining pattern or PCC did not change significantly (Supplementary Figure S1).
DDX3 interacts directly with IKKα following stimulation of the TLR7 pathway.
We next mapped the interaction between DDX3 and IKKα, using a range of His-tagged truncation mutants of DDX3. IKKα interacted with full-length DDX3, DDX3 (aa 1–408) and DDX3 (aa 139–408), but not with DDX3 (aa 409–662) or DDX3 (aa 172–408) (Figure 1d). This suggests that IKKα interacts with a region between aa 139 and 172 of DDX3. Interestingly, this means that the main interaction site for IKKα in DDX3 appears to be different from that for IKKε, which we previously determined to lie between aa 100 and 110 of DDX3 .
Finally, we wanted to ascertain that endogenous DDX3 and IKKα interact following activation of a relevant immune signalling pathway. Thus, we immunoprecipitated endogenous DDX3 from CL075-stimulated HEK293_TLR7 cells, using a matching isotype antibody as a control for specificity. We observed co-immunoprecipitation of endogenous IKKα 45 min after CL075 stimulation (Figure 1e), while the isotype control antibody did not pull down any IKKα. In addition, we also carried out a similar co-immunoprecipitation in stable HEK293_TLR7 cell lines that we engineered to express an inducible DDX3 shRNA (short hairpin RNA) construct or an NSC shRNA construct. DDX3 knockdown was induced by treating cells with doxycycline for 48 h, before stimulating cells with CL075 for different time periods, and immunoprecipitating endogenous DDX3 from resulting cell lysates. In this experiment, we again saw endogenous IKKα co-immunoprecipitating with DDX3 45 min after CL075 stimulation in control (NSC) cells, while no IKKα was pulled down with the DDX3 antibody in cells depleted for DDX3 (DDX3 shRNA cells), indicating that the observed DDX3–IKKα interaction bands in NSC cells are specific (Figure 1f). The interaction between DDX3 and IKKα was still detected 90 min after stimulation, but had disappeared at the 4 h time point (Figure 1f). These data confirm that endogenous DDX3 and IKKα interact after physiological activation of the TLR7 pathway. Finally, we also immunoprecipitated DDX3 from CL075-stimulated THP1 cells that either expressed an NSC shRNA or the DDX3 shRNA. As THP1s are a human monocyte/macrophage cell line, they likely respond to the TLR7/8 ligand CL075 through TLR8. Again, we observed co-immunoprecipitation of IKKα with DDX3 in CL075-stimulated THP1 cells, but not in unstimulated cells (Figure 1g), indicating that this function is shared between TLR7 and TLR8. There was no pulldown of IKKα in DDX3 knockdown cells, confirming that the interaction is specific.
DDX3 enhances IKKα activation
Seeing as we previously demonstrated that DDX3 enhances IKKε autophosphorylation , we next investigated whether DDX3 has a similar effect on IKKα activation. In radioactive kinase assays with commercially obtained GST-IKKα, we detected an IKKα autophosphorylation band that was reproducibly enhanced in the presence of recombinant purified DDX3. The N-terminal domain (aa 1–408) of DDX3 was sufficient to mediate this effect (Figure 2a). Phosphorylation of IKKα at serines 176 and 180 in the activation loop of the kinase is required for and indicative of its activation. To confirm the effect of DDX3 on IKKα phosphorylation and clarify whether it enhances IKKα phosphorylation at these relevant serine residues, we conducted kinase assays with unlabelled ATP followed by detection of phospho-IKKα in WBs with an antibody specific for IKKα phosphorylated at serines 176/180. IKKα phosphorylation at these residues was indeed enhanced in the presence of recombinant DDX3 (aa 1–408) (Figure 2b). Total recombinant IKKα was also detected in this experiment with an anti-IKKα antibody. We suspect that the total IKKα band appears thicker due to the appearance of phosphorylated higher-migrating forms of IKKα rather than uneven loading, as we repeated this experiment many times with reproducible results (Figure 2b). Next, we co-expressed Flag-IKKα and either DDX3 or the vaccinia virus (VACV) protein K7, which antagonises DDX3 , in HEK293T cells, lysed the cells after 24 h, and probed for phosphorylated IKKα (Ser176/180) in WBs. As shown in Figure 2c, ectopic expression of DDX3 (aa 1–408) strongly enhanced IKKα phosphorylation, while expression of its VACV antagonist K7 reduced levels of phosphorylated IKKα. This demonstrated that the effect of DDX3 on IKKα autophosphorylation also occurs in cells. Finally, we wanted to confirm that endogenous DDX3 is involved in mediating IKKα phosphorylation following activation of a relevant immune signalling pathway. To this end, we used the HEK293_TLR7 cell lines expressing DDX3 shRNA or a NSC. Knockdown of DDX3 was induced for 48 h before stimulation of the cells with the TLR7 agonist CL075. While there was some background activation present in unstimulated cells, CL075 enhanced IKKα/β phosphorylation from 15 min to 4 h after stimulation in NSC cells, with a peak at 1 h post stimulation. In cells with shRNA-mediated DDX3 knockdown, IKKα/β phosphorylation was clearly reduced, with some residual activation detectable at the 2 h and 4 h time points (Figure 2d). It should be noted that the antibody for phospho-IKKα is unable to differentiate between phospho-IKKα and phospho-IKKβ due to conservation of their activation loops, and thus it is possible that phosphorylation of both kinases is detected and affected by DDX3 knockdown in this experiment. However, the fact that we see such a clear reduction with DDX3 knockdown despite a possible contribution of IKKβ phosphorylation to the overall signal rather supports our conclusion that DDX3 has a marked role in this pathway.
DDX3 enhances autophosphorylation of IKKα in the activation loop.
DDX3 enhances NIK/IKKα-mediated IRF7 activation
We next hypothesised that increased IKKα activation should result in increased phosphorylation of its substrates. As mentioned in the introduction, we were particularly interested in determining whether DDX3 can regulate IFN induction pathways that are independent of IKKε/TBK1, such as TLR7-induced production of IFNα. We therefore investigated whether DDX3 affects NIK/IKKα-mediated phosphorylation of IRF7, the main transcription factor required for activation of IFNα promoters. IRF7 is not constitutively expressed in cells other than pDCs and was undetectable by Western blotting in HEK293 cells. In all experiments involving IRF7, we therefore ectopically expressed a small amount of Flag-IRF7. We first overexpressed NIK, IKKα, IRF7 and DDX3 in HEK293T cells, followed by detection of phospho-IRF7 by Western blotting (using an antibody-specific IRF7 phosphorylated at serines 471/472, which is indicative of its activation). As expected, co-expression of NIK and IKKα led to activatory phosphorylation of IRF7 (Figure 3a, lane 2), which was further enhanced by DDX3 expression (Figure 3a, lane 3). To confirm that endogenous DDX3 is required for mediating NIK/IKKα-induced IRF7 activation, we next used stable HEK293 cell lines containing the inducible DDX3 shRNA construct or an NSC construct, as described above. We induced knockdown of DDX3 for 48 h, followed by transfection of cells with NIK, IKKα, and IRF7 expression constructs. Phosphorylation of IRF7 was significantly reduced in cells with DDX3 knockdown compared with control (NSC) cells (Figure 3b), confirming that DDX3 is an upstream regulator of IRF7 activation by NIK and IKKα.
DDX3 regulates IKKα-mediated IRF7 phosphorylation and interferon induction.
DDX3 is part of the TLR7/8 signalling pathway to IFN induction
IRF7 activation leads to induction of type I interferon promoters. We therefore investigated next whether DDX3 overexpression or knockdown modulates IFN induction downstream of TLR7/8 stimulation and NIK/IKKα. We first used an IFNα4 promoter luciferase construct, which is predominantly activated by IRF7 (rather than IRF3) . Stimulation of HEK293_TLR7 cells with CL075 activated the IFNα4 promoter, and this activation was further enhanced by ectopic expression of increasing amounts of DDX3 (Figure 3c). While this did not reach statistical significance (P = 0.085, n = 3, paired t-test), there was nonetheless a trend towards enhanced IFNα4 promoter activation in DDX3-overexpressing cells. We next expressed NIK and IKKα together in HEK293 cells, which - as expected - also strongly activated the IFNα4 promoter. Co-expression of DDX3 significantly enhanced IFNα4 promoter activation induced by NIK/IKKα (P = 0.034, n = 3, paired t-test), suggesting that it might act downstream of or at the level of these kinases (Figure 3d). We confirmed this further by conducting NIK/IKKα-induced IFNα4 reporter gene assays after knocking down expression of endogenous DDX3 in our stable inducible DDX3 shRNA HEK293 cell line. When endogenous DDX3 was knocked down in shDDX3 cells, IFNα4 promoter activation induced by NIK and IKKα was significantly reduced compared with control (NSC) cells (P = 0.00014, n = 9, paired t-test), suggesting that DDX3 acts downstream of the NIK/IKKα kinase complex in IFNα4 promoter activation (Figure 3e).
We next used the THP1 cell line expressing either DDX3 shRNA or the corresponding NSC construct, in order to analyse effects of DDX3 knockdown in cells that endogenously express TLR8. Following induction of DDX3 knockdown, THP1 cells were differentiated with PMA and then stimulated with CL075 for 24 h. Ifnb mRNA levels were up-regulated by CL075, as detected by real-time PCR, and were significantly reduced in DDX3 knockdown (sh) cells compared with control (NSC) cells (Figure 3f) (P = 0.005, n = 3, paired t-test). We also measured type I IFN and RANTES protein levels in the supernatant using the HEK-Blue™ IFNα/β reporter cell line and RANTES ELISAs, respectively. IFN protein levels were significantly reduced in DDX3 knockdown (sh) cells compared with control (NSC) cells (Figure 3g) (P = 0.005, n = 3, paired t-test). RANTES levels were also significantly reduced in CL075-stimulated DDX3 knockdown (shDDX3) THP1 cells compared with control (NSC) cells. These results confirm that endogenous DDX3 is involved in the signalling pathway leading to IFN induction downstream of TLR7/8.
DDX3 regulates p100 processing and affects expression of p52-target genes
Another immune-relevant signalling pathway that relies on the interplay between NIK and IKKα is the alternative NF-κB pathway, which leads to processing of p100 to p52 . We therefore investigated whether DDX3 also regulates this separate IKKα-dependent signalling pathway. First, we expressed and stained Ha-p100 in HEK293T cells in order to examine its subcellular localisation (both p100 and the processed p52 carry the Ha-tag when expressed from our expression construct and will therefore be visualised by our immunofluorescent staining approach). Co-expression of NIK/IKKα with Ha-p100 increased the nuclear localisation of p100/p52 as expected (due to processing of p100 to p52, which can then translocate into the nucleus). Additional co-expression of DDX3 led to a further increase in nuclear p52, suggesting that DDX3 can promote p100 processing to p52 (Figure 4a). Next, we used WB analysis to confirm that DDX3 affects p100 processing. We induced knockdown of DDX3 in our stably transduced HEK293 DDX3 shRNA (or NSC) cells by treating them with doxycycline for 48 h, and then ectopically expressed NIK and IKKα, before harvesting cells for WB analysis. NIK and IKKα co-expression clearly induced p100 phosphorylation and subsequent degradation to p52 in control (NSC) cells (Figure 4b, lane 3), while p100 phosphorylation was greatly reduced and p52 levels were lower in DDX3 knockdown (shDDX3) cells (Figure 4b, compare lane 3 and lane 6 and see quantification graphs). We then proceeded to show that DDX3 knockdown affects downstream gene expression mediated by p52 and chose the well-established p52-target gene BAFF (B-cell activation factor) as our read-out . We again used the THP1 cell lines stably transduced with either the DDX3 shRNA plasmid or an NSC plasmid, and induced knockdown by treating cells for 48 h with doxycycline. We then differentiated THP1 cells with PMA for 24 h before stimulation with CL075. We observed up-regulation of BAFF mRNA expression (measured by real-time PCR) at 8 or 24 h in NSC cells, which was reduced in DDX3 knockdown (shDDX3) cells (Figure 4c), reaching statistical significance for the 24 h time point (P = 0.015, n = 3, paired t-test). This data demonstrate that DDX3 is also involved in mediating alternative NF-κB activation downstream of NIK/IKKα in the TLR7/8 pathway and thus broadens the downstream effects of DDX3 modulation beyond IFN-inducing pathways.
DDX3 regulates p100 processing and expression of a p52-target gene.
Our study demonstrates that DDX3 directly interacts with IKKα and enhances its kinase activity, thereby regulating the activity of IKKα-dependent downstream signalling pathways. We had previously shown that DDX3 interacts with IKKε and facilitates its autophosphorylation and auto-activation . According to our new data, DDX3 exerts a similar effect on the related kinase IKKα and can thereby contribute to additional IKKε/TBK1-independent innate immune signalling pathways. A very interesting study published while this project was ongoing also demonstrated the DDX3–IKKα interaction, in that case occurring during HCV infection or transfection of poly(I:C) . The study showed that DDX3 co-localised with the HCV 3′UTR RNA and IKKα. Interestingly, the DDX3–IKKα interaction in this paper was suggested to result from direct recognition of HCV RNA by DDX3 , thus indicating that DDX3 acts as a viral sensor in this context. This is in line with a study published by Oshiumi et al. , which suggested that DDX3 recognises HCV RNA and then signals via RIG-I/MDA5/MAVS to IFN induction. However, our previous studies have characterised DDX3 as a downstream signalling adaptor linking IKKε and the transcription factor IRF3 [3,4], which is also supported by several other studies [5,6,8,9]. In this present study, we observed the DDX3–IKKα interaction in the absence of viral RNA, namely following stimulation of the TLR7/8 pathway with the synthetic ligand CL075, a thiazoloquinolone derivative. Knockdown of DDX3 also affected signals downstream of NIK/IKKα, suggesting that DDX3 is able to exert an equivalent role as a signalling adaptor also in NIK/IKKα-dependent pathways. Li et al.'s paper  demonstrated that IKKα was required for SREBP-mediated lipogenesis and lipid droplet formation in HCV-infected Hu7.5.1 cells. As lipid droplets are crucial for efficient HCV replication and assembly, IKKα knockdown strongly suppressed HCV replication in these cells. DDX3 is also required for HCV replication [22,23] and it is therefore conceivable that enhancement of IKKα signalling contributes to DDX3's proviral role in HCV replication. Our study complements this work by demonstrating that the interaction between DDX3 and IKKα is direct and promotes IKKα autophosphorylation and -activation, resulting in enhanced phosphorylation of IKKα's downstream targets. While Li et al.  focussed on SREBP expression mediated by CBP, we demonstrate that IRF7 activation and p100 phosphorylation and processing are also affected by DDX3's regulation of IKKα. Our work therefore suggests that the DDX3–IKKα link might be a double-edged sword for the virus. On the one hand, it promotes lipid droplet formation, which favours the virus , but on the other hand it promotes production of interferons and pro-inflammatory cytokines (this study). In this context, it is very interesting that, in a follow-up study to their original work, the Liang laboratory showed that the DDX3–IKKα interaction appears to be dynamically regulated during HCV infection, with it only occurring early during infection. When viral protein production was high, DDX3 clustered with the HCV core protein on lipid droplets instead .
Our work shows that IKKα is another kinase whose activity can be regulated directly by DDX3, similar to IKKε  and CK1ε (casein kinase 1ε) . These studies suggest that DDX3 employs similar mechanisms to facilitate activation of different kinases and mediate phosphorylation of their downstream targets. Interestingly though, the C-terminal region of DDX3 was required and sufficient for CK1 activation , whereas the N-terminus was required and sufficient for IKKε and IKKα activation. We showed here that, despite the relatedness of the IKKs, IKKα interacts with DDX3 in a slightly different manner compared with IKKε. While IKKε binds to N-terminal and C-terminal regions of DDX3 (aa 1–110 and 409–662), but not to DDX3 (aa 139–408) , IKKα still binds to DDX3 (aa 139–408) but fails to bind to the C-terminus of DDX3 (aa 409–662).
Importantly, the finding that DDX3 can directly regulate IKKα greatly broadens the downstream effects of DDX3 manipulation on innate immune responses: (1) it affects TBK1/IKKε-independent induction of type I interferons downstream of TLR7/8 (and most probably TLR9 given the similarity of the two signalling pathways) and (2. it affects IKKα-mediated signalling pathways that are independent of IRF3/7, such as the alternative NF-κB pathway, and thereby regulates expression of additional sets of genes beyond type I interferons. These findings, in combination with the work published by the Liang laboratory [21,24] that showed an effect on the CBP-SREBP pathway, suggest that DDX3 generally affects downstream signalling pathways emanating from IKKα. It is not unlikely that additional downstream pathways, such as the conventional NF-κB pathway, could also be affected by DDX3's regulation of IKKα .
Its relevance for physiological processes regulated by these pathways is worth investigating in future studies. For example, by affecting the alternative NF-κB pathway, DDX3 could be involved in CD40 signalling, development of secondary lymphoid organs, and B-cell survival and maturation of B-cells . While it remains to be confirmed that our finding of DDX3 regulating TLR7-mediated IFN induction extends to pDCs, the physiologically relevant cell type for this signalling pathway, this could potentially indicate a link between DDX3 and systemic lupus erythematosus (SLE). TLR7 signalling is strongly linked SLE pathogenesis, with TLR7 activation being triggered by anti-RNP immune complexes, resulting in the production of large amounts of type I IFN from pDCs [27–29]. SLE is an autoimmune condition with a much higher prevalence in women compared with men, and interestingly, DDX3(X) escapes X-inactivation  and might indeed be expressed to higher levels in females compared with males [31–33]. Furthermore, there is also some evidence that the TLR7 pathway is more active in females compared with males [34,35]. Based on our data demonstrating the involvement of DDX3 in TLR7 signalling, higher expression of DDX3 in females could conceivably contribute to increased TLR7 pathway activity and thereby to the increased female prevalence of SLE.
B-cell activation factor
human embryonic kidney
interferon regulatory factor
mitochondrial antiviral signalling
Pearson's Correlation Coefficient
plasmacytoid dendritic cell
pattern recognition receptor
retinoic acid-inducible gene
short hairpin RNA
systemic lupus erythematosus
A.F., L.G., and Y.H. designed and carried out experiments and analysed data. M.S. devised the hypothesis and experimental strategy, guided experimental design, and analysed data. A.F. and M.S. co-wrote the paper and all authors checked the manuscript before publication.
We gratefully acknowledge support of this research through funding from Science Foundation Ireland [09/RFP/BIC2188] and the Irish Health Research Board [HRA/2009/171 and HRA_POR/2011/58].
The Authors declare that there are no competing interests associated with the manuscript.