|Home | About | Journals | Submit | Contact Us | Français|
Induction of type I interferon is a central event of innate immunity, essential for host defense. Here we report that the transcription factor ELF4 is induced by type I interferon and upregulates interferon expression in a feed-forward loop. ELF4 deficiency leads to reduced interferon production, resulting in enhanced susceptibility to West Nile virus encephalitis in mice. After viral infection, ELF4 is recruited by STING, interacts with and is activated by the MAVS-TBK1 complex, and translocates into the nucleus to bind interferon promoters. Cooperative binding with ELF4 increases the binding affinity of interferon regulatory factors IRF3 and IRF7, which is mediated by EICE elements. Thus, in addition to identifying a regulator of innate immune signaling, we uncovered a role for EICE elements in interferon transactivation.
The innate immune system serves as the first-line sentinel of host defense against invading pathogens by recognizing various conserved molecular motifs called pathogen-associated molecular patterns (PAMPs) and initiating cellular host defense countermeasures. PAMPs are detected by several classes of host pattern recognition receptors (PRRs), including Toll-like receptors (TLRs), RIG-like receptors (RLRs), NOD-like receptors (NLRs), C-type lectins and double-stranded DNA (dsDNA) receptors. The PRR responses trigger the activation of downstream signaling events, leading to the transcription of NF-κB–dependent and IRF3-IRF7–dependent genes, including type I interferon1.
TLRs recognize a range of products derived from a variety of pathogens, including bacteria, fungi, viruses and parasites2. All TLRs characterized so far signal through the adaptor proteins MyD88 or TRIF, the latter downstream of TLR3 and TLR4. RLRs recognize cytoplasmic viral RNA and recruit the adaptor MAVS (also called Cardif, IPS-1 or VISA)3–6. Currently known dsDNA receptors include AIM2, which forms an inflammasome with ASC and caspase-1 (ref. 7), RNA polymerase III, which converts DNA into 5′-triphosphorylated RNA that activates RIG-I (ref. 8) and the DNA cytoplasmic sensors IFT16 (ref. 9), DDX41 and cGAS10,11. Signals from IFI16, DDX41 and cGAS are transduced by the endoplasmic reticulum membrane protein STING (also termed ERIS, MITA, MPYS or TMEM173), which has a key role in the cytoplasmic nucleic acid sensing pathway12–15. Most innate immune signaling pathways result in the induction of type I interferon and other cytokines. Type I interferon is induced within hours after infection, often in high amounts, to initiate an antiviral state in cells and is essential for survival in acute viral infection and modulation of the immune response.
IRF3 and IRF7 are critical transcriptional factors that regulate production of type I interferon (IFN-α and IFN-β) by binding to the inter-feron stimulated response element (ISRE). Irf3−/−, Irf7−/− or Irf3−/− × Irf7–/– mice are much more prone to viral infection than wild-type mice, and do not induce type I interferon efficiently. However, Irf3−/− × Irf7–/– macrophages and dendritic cells have minimally diminished induction of IFN-β after viral infection, indicating that other transcription factors may be necessary for interferon expression16. ELF4 (refs. 17,18) belongs to the ETS transcription factor family, which has at least 27 mammalian members involved in various biological processes19. ELF4 controls the quiescence of endothelial cells20, promotes cellular transformation18 and is associated with ovarian cancer and leukemias21,22. These studies suggest that ELF4 is an important transcription factor, though its physiological functions are largely unknown. A previous study has shown that the ETS transcription factor PU.1 cooperates with IRF4 to regulate the expression of immunoglobulin23, suggesting a correlation between the ETS binding motif and the ISRE.
Here we demonstrate that ELF4 regulates interferon induction and is critical for host defense. ELF4 was activated in response to innate immune signals and initiated transcription of type I inter-feron genes to control diverse pathogens. Elf4−/− mice had increased susceptibility to viral infection. In the absence of ELF4, IRF3, IRF7 and NF-κB bind inefficiently to interferon gene promoters, resulting in impaired type I interferon responses. ELF4 has a global role in TLR, RLR and STING-mediated DNA sensing signaling pathways, indicating its fundamental importance in the innate immune system.
We investigated the role of STING, an adaptor protein that has a pivotal role in antiviral immune signaling, in infection with West Nile virus (WNV), a single-stranded RNA virus that can cause lethal meningoencephalitis in humans. STING loss-of-function Tmem173gt/gt mice (also known as Goldenticket and here referred to as Stinggt/gt)24 were more susceptible to WNV than wild-type animals (Fig. 1a). To more clearly understand STING-related antiviral immunity and to identify interacting factors of this antiviral pathway, we used an infection interaction screen to detect virus-induced genes that associate with STING. We infected HeLa cells that expressed Flag epitope–tagged STING with WNV, and purified Flag-STING by immunoprecipitation. Analysis of STING-associated proteins resolved by SDS-PAGE and analyzed by mass spectrometry identified ELF4 as a protein recruited to STING after infection with WNV (Fig. 1b and Supplementary Tables 1 and 2). We confirmed the interaction between ELF4 and STING by co-immunoprecipitation in HeLa cells (Fig. 1c).
293T cells transfected with plasmids encoding either human or mouse ELF4 activated a luciferase reporter driven by the Ifnb1 promoter (IFN-β–Luc), which resulted in up to 200-fold induction of the reporter, comparable to levels induced by overexpressed MAVS (Fig. 1d and Supplementary Fig. 1b)25. Real-time PCR analysis showed that ELF4 induced Ifnb1 mRNA in 293T cells (Fig. 1e and Supplementary Fig. 1a). The ELF4-mediated activation of Ifnb1 was specific, as related ELF family members that also have ETS DNA-binding domains did not induce Ifnb1 (Fig. 1f). To assess whether ELF4 led to the production and secretion of interferon proteins, we measured amounts of type I interferon in the supernatants after overexpression of ELF4 via a bio-assay using 2fTGH cells with a luciferase reporter driven by the ISRE (2fTGH-ISRE-Luc cells), which generate luciferase in response to type I interferons. The release of type I interferon was markedly increased after ELF4 overexpression (Fig. 1g). To study the function of ELF4 in immune responses, we examined induction of cytokines related to interferon, including various forms of IFN-α and IFN-λ3. Exogenous expression of ELF4 modestly upregulated INF-α2, IFN-α8 and IFN-λ3, and did not upregulate IFN-α4 or IL-22 (Supplementary Fig. 1d–h).
To determine the breadth of the antiviral signals transduced by ELF4, we quantified secretion of interferon in ELF4-expressing cells triggered by the following: an RNA virus, a DNA virus, HIV, poly(I:C) or poly(dA:dT) (Fig. 1h and Supplementary Fig. 1c). ELF4 enhanced induction of interferon triggered by all viral stimuli and agonists tested. To identify the critical domains of ELF4 required for the induction of type I interferon, we determined the ability of several ELF4 variants (Supplementary Fig. 2a) to induce type I interferon. The ETS domain, which mediates binding between ELF4 and cis-acting elements, was essential for production of IFN-β induced by ELF4, as assessed by reporter assay. In addition, we observed marked abrogation of IFN-β induction when we deleted the nuclear localization sequence region (NLS, as delineated on the Eukaryotic Linear Motif resource website; http://elm.eu.org/), TRAF2 binding motif (T2B; http://elm.eu.org/) or amino acids 1–87 (amino acids 1–52 comprise a putative transactivation domain)26. Proteins lacking the C terminus of ELF4, which is responsible for the interaction with STING (Supplementary Fig. 2b), induced the IFN-β reporter poorly, but both the putative transmembrane helices and the sumoylation sites were dispensable (Fig. 1i). Collectively, these results demonstrate that ELF4 is a potent stimulator of type I interferon.
As ELF4 stimulates production of interferon, which is critical for anti-viral immunity, we examined the response of ELF4 to viral infection. Elf4 mRNA was highly induced when we infected cells with either Sendai virus (SeV) or vesicular stomatitis virus (VSV) (Fig. 2a), suggesting the involvement of this protein in antiviral responses. Primary monocyte–derived macrophages from human donors showed significant induction of Elf4 and Ifnb1 mRNA after infection with WNV (Fig. 2b). We also observed that expression of ELF4 was induced by interferon, indicating that Elf4 is an interferon-stimulated gene (Fig. 2c). In addition, the replication of SeV, VSV and WNV was dramatically attenuated by ELF4 overexpression. Inhibition of viral replication was dose-dependent as measured by immunoblot and a plaque assay (Supplementary Fig. 3a–c).
To further delineate the function of ELF4 in host defense, we infected KEK 293T cells after knockdown of endogenous Elf4 by small interfering (si)RNA. Transfection of an ELF4-specific siRNA reduced ELF4 expression and resulted in increased viral replication, when compared to transfection with control (scrambled) siRNA (Fig. 2d and Supplementary Fig. 3d,e). Consistent with this, VSV and HSV-1 replication were markedly enhanced in Elf4−/− mouse embryonic fibroblasts (MEFs) and mouse macrophages compared with wild-type cells (Fig. 2e–h and Supplementary Fig. 3f–h). Elf4−/− MEFs reconstituted with full-length ELF4, but not the variant with the ETS domain deleted, reduced VSV or HSV-1 infectivity significantly (Fig. 2e–g), indicating that the ETS domain is essential for the anti-viral activity of ELF4. IFN-β production induced by viral infection was impaired in Elf4−/− MEFs, and could be restored by full-length ELF4 but not the variant with ETS deleted (Fig. 2i). ELF4 therefore has a role in host defense against viral infection.
To evaluate the importance of ELF4 in viral infection in vivo, we infected wild-type and Elf4−/− mice with WNV. In keeping with our in vitro data, and consistent with the results using Stinggt/gt and Mavs−/− mice27, Elf4−/− mice were significantly more susceptible to lethal WNV infection compared to wild-type mice (Fig. 3a). Viremia was significantly higher in Elf4−/− mice than in wild-type mice at day 3 after infection, as determined by real-time PCR of transcripts encoding WNV envelope proteins (here referred to as WNVE transcripts; Fig. 3b). Because meningoencephalitis contributes to WNV lethality28, we quantified viral burdens in brains of wild-type and Elf4−/− mice. Consistent with the survival data, the viral burden (as assessed by WNVE mRNA and plaque assays) was higher in the central nervous system of Elf4−/− mice than wild-type mice at day 8 after infection (Fig. 3c,d). We observed similar differences in the spinal cord, kidney and splenic tissues (Fig. 3e–g). Consistently, we observed more immune cell infiltrate in the central nervous system (Fig. 3h,i) and a more pronounced antiviral serologic response (Supplementary Fig. 3i,j) in the Elf4−/− mice. In addition, we detected less circulating IFN-β and IFN-α2 in the blood or plasma of Elf4−/− mice on days 1 and 3 after WNV infection (Fig. 3j–l) and less interferon-induced protein ISG15 (Fig. 3m). Induction of Ifng, Il6 and the IFN-γ–inducible gene Cxcl10, however, was similar in Elf4−/− and wild-type mice (Fig. 3n and Supplementary Fig. 3k,l). Consistent with the in vitro data, the Elf4 mRNA in blood was upregulated at day 1 and 3 after WNV infection compared to day 0 (Fig. 3o). Furthermore, induction of type I interferon and antiviral activity were impaired in macrophages isolated from Elf4−/− mice after WNV infection (Fig. 3p,q). Thus, ELF4 is involved in the induction of type I inter-feron and protection against infection with WNV.
ELF4 has been shown to regulate two important antiviral mechanisms: development of NK and NK T cells29, and proliferation and homing of CD8+ T cells by activating Klf4 expression17. NK and NKT cell numbers are profoundly reduced in Elf4−/− mice compared to wild-type mice29. However, NK and NKT cells appear to be dispensable for the host anti-WNV immune response30 likely because WNV and other flaviviruses partially suppress NK cell activity in mice31. Consistently, reconstitution of NK and NKT cells in Elf4−/− mice by tail-vein injection did not affect the susceptibility of these mice to WNV (Supplementary Fig. 3m). In addition, Elf4−/− mice had a similar number of NK cells as wild-type mice after WNV infection (Fig. 3r). CD8+ T cells in particular are required for WNV clearance from the central nervous system32. However, differences in proliferation of CD8+ T cells between Elf4−/− and wild-type mice are only notable in mice older than 6 months17. Consistent with this, we observed a similar proportion of splenic CD8+ T cells in Elf4−/− and wild-type mice before and after viral infection as well as comparable proportions of CD4+ T cells (Supplementary Fig. 3n,o). Elf4−/− and wild-type mice in our experiments were 6 weeks old and had similar extent of CD8+ T cell proliferation17. These data suggest that ELF4 controls WNV replication primarily via regulation of type I inter-feron response and not through modulation of NK cell, NKT cell or CD8+ T cell numbers during viral infection. Nevertheless, ELF4 may participate in the regulation of CD8+ T cell function indirectly via induction of type I interferons33 and also by influencing NK and NKT cell survival and proliferation via type I interferon–induced IL-15 (ref. 34).
To study the functions of ELF4, we next evaluated the role of ELF4 in innate immune signaling pathways, such as TLR-, RLR- and STING-dependent dsDNA-sensing pathway. Activation of the Ifnb1 promoter by poly(I:C), SeV, poly(dA:dT) or HSV-1 was attenuated when we silenced Elf4 by a specific siRNA (Fig. 4a and Supplementary Fig. 4a), indicating that ELF4 was required for RLR- and STING-dependent activation of Ifnb1 transcription. Bone marrow–derived macrophages (BMDMs) from Elf4−/− animals showed significantly impaired production of IFN-β compared to wild-type BMDMs when triggered by poly(I:C), lipopolysaccharide (LPS) or CpG DNA as well as after infection with SeV, EMCV or HSV-1 compared to wild-type BMDMs (Fig. 4b). Therefore, ELF4 is also required for IFN-β production mediated by TLR3, TLR4, TLR7 and TLR9. We observed similar results when we measured the activation of IFN-β, IFN-α2 and ISG15 by TLR or RLR agonists in BMDMs, peritoneal macrophages and bone marrow–derived dendritic cells (BMDCs), indicating a global role of ELF4 in innate antiviral response (Fig. 4c–f). Consistent with this, production of IFN-β induced by LPS, R848 or poly(I:C) decreased markedly after siRNA-mediated knockdown of ELF4 in 293T cells that stably express TLR4, TLR7 or TLR3 (Fig. 4g). However, ELF4 was dispensable for induction of IL-1β and tumor necrosis factor (TNF) in macrophages, suggesting a specific role in type I interferon–related gene regulation (Fig. 4h and Supplementary Fig. 4b). In contrast to the in vivo results, induction of Cxcl10 was impaired in Elf4−/− MEFs (Supplementary Fig. 4c). As Cxcl10 is induced in response to IFN-α signaling and also IFN-γ signaling, the deficiency of Cxcl10 induction resulting from impaired IFN-α may be compensated by IFN-γ in Elf4−/− mice.
To characterize the role of ELF4 in signaling pathways downstream of pathogen-recognition receptors, we quantified the induction of an IFN-β reporter in a luciferase assay in untransfected or in ELF4 plasmid–transfected MEFs derived from animals deficient in the innate adaptors MyD88, TRIF, MAVS or STING. ELF4-induced activation of IFN-β was similar in the adaptor-deficient cells and in wild-type cells (Fig. 4i). We obtained similar results when we subjected 293T cells to siRNA-mediated knockdown of TRIF, MAVS, MyD88 or STING (Supplementary Fig. 4d), which indicated that ELF4 functions downstream of these adaptors. In ELF4-deficient MEF cells transfected with plasmids encoding the adaptors TRIF, MAVS, MyD88 or STING and luciferase reporters for IFN-β or ISRE, activity induced by adaptor overexpression was decreased compared to wild-type cells (Fig. 4j). Consistent with this, siRNA-mediated knockdown of ELF4 in 293T cells reduced the production of IFN-β induced by TRIF, MAVS or STING overexpression, as well as the activation of ISRE induced by MyD88 (Supplementary Fig. 4e–h). Thus, ELF4 appears to function downstream of TRIF, MyD88, MAVS and STING.
Because ELF4 associates with STING in virus-infected cells, we investigated the possibility that ELF4 regulated the activation of NF-κB or IRF3. As a positive control, ectopic expression of MAVS activated IRF3 and NF-κB in a dose dependent manner (Fig. 5a) as revealed by the phosphorylation of IRF3 and NF-κB. However, ectopic expression of ELF4 did not activate IRF3 and NF-κB (Fig. 5a). Furthermore, SeV-induced IRF3 dimerization was blocked in MAVS-deficient MEFs or in 293T cells in which Mavs was knocked down but not in ELF4-deficient MEFs or in 293T cells in which Elf4 was knocked down (Fig. 5b and Supplementary Fig. 5a). These results indicate that ELF4 is not involved in the activation of IRF3 and NF-κB. We next examined the possibility that association with STING may recruit ELF4 to the MAVS-TBK1 signaling complex and activate ELF4. ELF4 coimmunoprecipitated with MAVS or TBK1 in the presence of STING (Supplementary Fig. 5b). Further, viral infection promoted ELF4 interaction with STING as well as with TBK1 (Fig. 5c). Overexpression of ELF4 did not induce type I interferon in Tbk1−/− MEFs (Fig. 5d). Reconstitution of Tbk1−/− MEFs with wild-type TBK1, but not with a TBK1 variant lacking kinase activity, restored the ability of ELF4 to induce IFN-β (Fig. 5d). ELF4 is phosphorylated after viral infection (Supplementary Fig. 5c), which prompted us to examine whether TBK1 could phosphorylate ELF4. In vitro kinase assays showed that TBK1 mediated phosphorylation of ELF4 (Fig. 5e). We examined four main phosphorylation regions in ELF4 (using the NetPhos 2.0 Sever) to identify amino acids phosphorylated by TBK1. The third ELF4 phosphorylation region is necessary for induction of IFN-β expression (Supplementary Fig. 5d), in which Ser331 is indispensable (Fig. 5f and Supplementary Fig. 5e). To confirm that TBK1 is needed for activation of ELF4, we examined type I interferon induction in plasmacytoid dendritic cells (pDCs), which are known to not use TBK1 for regulation of interferons35. CpG DNA–induced production of interferons was normal in Elf4−/− pDCs (Supplementary Fig. 5f,g) indicating that TBK1 and ELF4 are not involved in signaling of type I interferon in this cell type.
In addition, the dimerization of endogenous ELF4 was induced by viral infection as well as by stimulation with poly(I:C), LPS, R848 and CpG DNA in macrophages (Fig. 5g). The ETS domain was required for ELF4 dimerization (Supplementary Fig. 5h,i). We observed nuclear translocation of ELF4 after SeV infection and ELF4 overexpression in HeLa cells (Fig. 5h and Supplementary Fig. 5j). Furthermore, MAVS, STING, TBK1 and phosphorylation of ELF4 were essential for the nuclear translocation of ELF4 in response to viral infection, as assessed by western blotting and immunofluorescence (Fig. 5i and Supplementary Fig. 5k,l). In contrast with endogenous ELF4, ectopically expressed ELF4 resided in the nucleus (Fig. 5j). ELF4 contains two putative nuclear-localization sequences (NLSs): at amino acids 196–202 (NLS1)26 and 173–183 (NLS2; http://elm.eu.org/). We examined subcellular localization of ELF4 variants with deleted NLS1 (ELF4-ΔNLS1), NLS2 (ELF4-ΔNLS2) or both NLS1 and NLS2 (ELF4-ΔNLS1-2). ELF4-ΔNLS1-2 and ELF4-ΔNLS2 were retained in the cytoplasm, whereas ELF4-ΔNLS1 localized both in the nucleus and cytoplasm (Fig. 5j). ELF4-ΔNLS1-2 and ELF4-ΔNLS2 did not activate the Ifnb1 promoter, whereas ELF4-ΔNLS1 induced weak promoter activity (Supplementary Fig. 5m). Taken together, these results suggest that in response to viral infection, ELF4 was phosphorylated by TBK1 and translocated to the nucleus in a MAVS- and STING-dependent manner.
The ETS domain mediates sequence-specific DNA binding36 with two highly conserved arginines in this domain critical for DNA binding34. Overexpression of an ELF4 variant in which both of these conserved arginines were replaced by alanines (ELF4-RRAA) in 293T cells abrogated activation of IFN-β. Induction of IFN-β was restored when we replaced the alanine with arginine in the mutants (Fig. 6a), suggesting that ELF4-dependent induction of interferon depends on DNA binding by ELF4. Chromatin immunoprecipitation and quantitative PCR (ChIP-qPCR) indicated that ELF4 and IRF3 were recruited to the Ifnb1 and Ifna2 promoters (Fig. 6b,c) in macrophages infected with SeV. Electrophoretic mobility shift assays (EMSAs) also showed that ELF4 bound to the Ifnb1 promoter after infection with SeV (Fig. 6d,e).
ETS transcription factors bind to related, but distinct, sites containing a core purine-rich motif, GGAA. We identified three GGAA motifs in the Ifnb1 promoter, four GGAA motifs in the Ifna2 promoter and one GGAA motif in the Ifna4 promoter (Supplementary Fig. 5n). Activation of the Ifnb1 promoter was impaired when the first (position −80 relative to the transcription start site) or second (position −50) GGAA was replaced by GATC (Fig. 6f). In EMSA, in vitro–expressed ELF4 protein bound to the wild-type Ifnb1 promoter probe, did not bind a probe that contained mutations in the first GGAA motif (IFN-mut1) and bound weakly to probes that contained mutations in the second (IFN-mut2) or third (IFN-mut3) GGAA motifs (Fig. 6g). These data suggest that ELF4 binds to the Ifnb1 promoter directly. Furthermore, activation of Ifna2 was significantly reduced after replacing the second GGAA with GATC (Fig. 6h), indicating that ELF4 also binds the Ifna2 promoter. Taken together, our results demonstrate that ELF4 binds and activates interferon gene promoters.
Combinatorial interactions between distinct classes of sequence-specific transcription factors are important in regulating eukaryotic gene expression in response to diverse environmental signals37. The Ifnb1 promoter contains DNA-binding sites for several transcription factors: NF-κB, IRF3-IRF7 and activator protein 1 (AP1; a heterodimeric protein composed of ATF-2 and c-Jun)38 (Supplementary Fig. 5n). We observed that ELF4 could activate a luciferase reporter of the interferon regulatory factor (IRF)-binding element (ISRE-Luc) and NF-κB–Luc but not AP1-Luc (Supplementary Fig. 6a). We first examined the role of IRF3 and p65 on ELF4-mediated activation of Ifnb1 promoter in 293T cells, which do not express IRF7. Knockout of Irf3 in MEFs or knockdown of Rela in 293T cells reduced the ability of ELF4 to activate the Ifnb1 promoter (Fig. 7a,b and Supplementary Fig. 6b). Overexpression of ELF4 strongly synergized with IRF3 to activate the Ifnb1 promoter in 293T cells (Supplementary Fig. 6c). Flag-tagged ELF4 co-immunoprecipitated with hemagglutinin (HA)-tagged p65 but did not associate with HA-tagged p50 or IRF3 in 293T cells (Supplementary Fig. 6d). Endogenous ELF4 associated with p65 after infection with SeV in 2fTGH cells (Fig. 7c). SeV-mediated recruitment of ELF4 to the Ifnb1 promoter, but not to the Ifna2 or Ifna4 promoters, was attenuated by siRNA-mediated silencing of p65 in MEFs (Fig. 7d and Supplementary Fig. 6e). Conversely, p65 did not bind efficiently to the Ifnb1 promoter in Elf4−/− MEFs (Fig. 7e). However, ELF4 had no effect on p65-mediated activation of NF-κB–Luc reporter activity (Supplementary Fig. 6f), indicating that both NF-κB binding and ELF4 binding are required for the cooperative activation of interferon promoters.
The ETS domain protein PU.1 can recruit IRF family transcription factors IRF2, IRF4 and IRF8 to specific promoters carrying composite ETS-IRF elements (EICEs)23. Assembly of the PU.1-EICE complex requires phosphorylation of PU.1 at Ser148. Because EICE are present in interferon promoters, we tested whether ELF4 enhances the binding of IRF3 or IRF7 to the interferon promoters and synergizes with IRFs for interferon expression. Overexpressed ELF4 synergized with IRF3 to activate the Ifnb1 promoter in 293T cells (Supplementary Fig. 6c). We found a single GGAA motif in the Ifna4 promoter that is mainly activated by IRF7. We next tested the effect of ELF4 on IRF7-mediated activation of the Ifna4 reporter. ELF4 overexpression alone did not activate the Ifna4 reporter (Supplementary Fig. 1e), whereas overexpression of IRF7 alone resulted in a 20-fold induction of IFN-α4–Luc reporter activity (Fig. 7f). Co-expression of ELF4 strongly enhanced IRF7 transactivation activity in a dose-dependent manner (up to 150-fold; Fig. 7f and Supplementary Fig. 6g). In contrast, ELF4 has minimal effect on the IRF7-mediated activation of a ISRE-Luc reporter (Supplementary Fig. 6h,i).
To determine the physiological function of ELF4 in regulating IRF3-IRF7 and NF-κB activity, we performed ChIP-qPCR in untreated or VSV-infected wild-type or ELF4-deficient peritoneal macrophages (Fig. 7e,g). VSV-induced recruitment of IRF3 and p65 to the Ifnb1 promoter was significantly reduced in Elf4−/− macrophages (Fig. 7e). Similarly, the binding of IRF7 to the Ifna4 promoter was decreased in Elf4−/− macrophages (Fig. 7g). Furthermore, we observed cooperative binding of ELF4 and IRF–NF-κB to IFN-β in EMSA (Supplementary Fig. 6j). The EICE element of the Ifnb1 promoter was indispensable for the cooperative transactivation mediated by ELF4, IRF3 and p65, indicating an EICE-dependent formation of ELF4-IRF complex (Fig. 7h). Comparative kinetic analyses showed parallel virus-induced activation of ELF4, NF-κB and IRF-3 MEFs (Fig. 7i). Similar to the expression of IRF7, ELF4 expression was low in most tissues (Supplementary Fig. 7a). We noted comparable reduction in the production Ifnb1 and Isg15 in Elf4−/− and Irf3−/− MEFs (Supplementary Fig. 7b,c). Taken together, our data indicate that ELF4 is important for the recruitment of IRF3, IRF7 and p65 to interferon promoters and for the induction of interferon gene expression.
Many ETS domain–containing proteins are targets of signal transduction pathways and are activated in response to a wide array of extra-cellular stimuli. We demonstrated that ELF4 was induced after viral infection and interferon treatment, and upregulated production of type I interferon in a feed-forward loop. Genetic deletion of ELF4 rendered mice highly susceptible to WNV infection. Analysis of Elf4−/− macrophages, DCs and MEFs revealed that ELF4 was essential for efficient production of type I interferons. ELF4 was downstream the adaptor proteins TRIF, MyD88, STING and MAVS and was phosphorylated by TBK1. As is the case with p65 (ref. 39), ectopic expression of ELF4 was constitutively active and resulted in formation of dimers. We speculate that overexpressed ELF4 may recruit and be activated by TBK1. Overexpressed ELF4 bypassed upstream signals and induced type I interferon in the absence of these adaptor proteins. In Tbk1−/− cells, overexpressed ELF4 was inactive, as phosphorylation did not occur. ELF4 cooperated with IRF3, IRF7 or p65 to bind to ELF4-IRF or ELF4–NF-κB composite motifs in type I interferon promoters and to activate interferon gene expression.
Rapid induction of type I interferon expression is a central event in establishing the innate antiviral response. Molecular regulation of interferon gene expression is tightly regulated by extracellular and intracellular signals generated during primary infection and culminates in the activation of NF-κB and IRFs. These in turn trigger an immediate early interferon response characterized by the release of IFN-β and IFN-α1. IRF3 and IRF7 have essential roles in virus-induced production of type I interferon40. IRF3 is constitutively expressed and is crucial for the initial induction of immediate early genes such as those encoding IFN-β, IFN-α1 and CCL5 (ref. 14). In contrast, IRF7 is synthesized de novo upon interferon stimulation and contributes to the expression of delayed-type genes including other interferon subtypes36. Similarly to the expression of IRF7, ELF4 expression is low in most tissues and is induced by type I interferons, suggesting a major role in the subsequent feed-forward amplification of interferons. Like IRF3 and IRF7, ELF4 is phosphorylated after viral infection. Thus, in many respects, ELF4 has a mode of expression and activation similar to that of IRF7. Moreover, we noted comparable reduction in the production type I interferon in Elf4−/− and Irf3−/− MEFs. Thus, ELF4 is a newly recognized and distinct IRF that has functional similarity with IRF3 and IRF7. In addition to IRF3 and IRF7, recent studies have shown that IRF5 contributes to activation of type I interferon in myeloid DCs41. Induciton of IFN-β was impaired in Elf4−/− BMDCs in response to TLR and RLR agonists, suggesting that cooperative binding of IRF5 and ELF4 to the Ifnb1 promoter may exist in myeloid DCs.
In common with many other transcription factors, ETS proteins exhibit low selectivity in binding-site preference, which suggests that the specificity of promoter targeting by ETS proteins relies on cooperation with transcription factors. Indeed, combinatorial control is a characteristic property of the ETS family, which involves interaction between ETS members and other key transcription factors such as SRF, Pax, AP1, IRF and NF-κB family members42–44. In response to parathyroid hormone, ELF4 is phosphorylated by JNK. JNK activates the oncoprotein c-Jun, thereby forming the activator protein-1 (AP-1) transcription factor as a homodimer or heterodimer45.
ETS proteins are a family of multifunctional transcription factors that direct gene expression by binding to target genes carrying specific consensus GGAA DNA-binding recognition sites. The IRFs regulate expression of target genes by binding to an ISRE that usually contains a potential ETS-binding site (GGAA)23. A cooperative correlation between ETS proteins and IRFs has been described23. PU.1 can interact with IRF2, IRF4 and IRF8 to form a heterodimeric complex in its target gene promoters and thereby regulate gene transcription. We here identified and characterized an ETS protein ELF4 as an interferon transcription factor that enhances the DNA-binding affinity of IRF3 and IRF7. We did not observe a direct protein interaction between ELF4 and IRF3 or IRF7 (data not shown) suggesting that protein-protein interaction is not essential for cooperative binding to the promoters46. Instead, cooperation arises mainly through co-binding of transcription factors dependent on nucleotide sequence (ETS-binding motif and ISRE). We found a new enhancer box (E-box), EICE, in interferon promoters. Thus ELF4 synergizes with IRF3, IRF7 and NF-κB to induce type I interferons by cooperative binding of these transcription factors via the ETS binding motif. As both ETS proteins and IRFs are widely expressed and participate in many important biological processes, the characterization of EICE may contribute to understanding gene regulation in diverse fields, including innate immunity.
In conclusion, ELF4 is a new type I IRF that activates the innate immune response by promoting the production of type I interferons, which are critical for immune surveillance of different classes of viral pathogens. Exploiting or enhancing ELF4-mediated signaling may lead to new strategies for combatting infectious diseases; conversely, inhibiting ELF4 may result in immunosuppression.
Elf4−/− mice on a C57BL/6J background have been described previously29. Stinggt/gt mice, which lack a functional STING protein, were a gift from R. Vance (University of California, Berkeley). Irf3−/− mice were a gift from R. Medzhitov (Yale University), Myd88−/− mice were a gift from R. Flavell (Yale University). 293T, HeLa, L929 (mouse fibroblast cells), 2fTGH-ISRE (human fibrosarcoma cells expressing an ISRE), L929-ISRE and Vero cells were cultured in DMEM. Isolation of MEFs, BMDMs and macrophages was performed as described12. BMDCs were isolated from tibia and femur. Cells were cultured in 10-cm Petri-dish at 37 °C for 5 d. The medium for BMDCs was RPMI-1640 medium containing GM-CSF at 100 U/ml and TNF at 50 U/ml. 293-hTLR4A-MD2-CD14 and 293XL-hTLR7 cells were from InvivoGen. Heparinized human blood was obtained with written informed consent from healthy donors, and primary macrophages were derived from peripheral blood monocytes and infected with WNV as described47. VSVGFP (Indiana strain) was a gift from J. Rose (Yale University), and HSV-1 was from A. Iwasaki (Yale University). SeV (Cantell strain; VR-907) and EMCV (VR-129B) were purchased from American Type Culture Collection (ATCC). Single-cycle HIV virus was prepared by cotransfecting 293 cells with VSV-G (0.07 μg) or YU2 (1.33 μg), pOPRIgagpol (1.5 μg), Ps-ReV(0.33 μg) and HIVec2GFP (0.66 μg). Poly(I:C) (catalog number tlr-pic), poly(I: C)-LMW (tlr-picw), flagellin (tlrl-bsfla), lipoteichoic acid (LTA; tlrl-lta), R848 (tlrl-r848), Pam3CSK4 (tlrl-pms) and CpG oligonucleotide (tlr-1585) were from Invivogen. LPS (L4391) was from Sigma. Poly(dA-dT) was synthesized as 5′-d(AT)30-3′ and was annealed by heating to 95 °C and cooling to 25 °C. Lipofectamine 2000 (Invitrogen) and TransIT-LT1 (MIR2300, Mirus) was used for lipid transfection. The following antibodies were used: anti-ELF4 (ELF4) (AV38028), anti-Flag (F3165), anti-GAPDH (G8795), anti-HA (H3663) and anti-VSV (A5977; all from Sigma); anti-phospho-IRF3 (Ser396; Cell Signaling) anti-IRF3 (D83B9) and; anti-MAVS (sc-68881 (Santa Cruz) and 3993 (Cell Signaling)); and anti–histone H3 (31560; Upstate). Anti-Sendai virus polyclonal antibody (PD029) was from MBL. The antibodies were diluted 1,000 times for immunoblots, 200 times in immunofluorescence and ChIP assay. Human interferon beta (C11420-1) was from PBL interferon. The IFN-β ELISA kits were obtained from PBL.
The work described in this paper involves studies in experimental animals. All relevant, institutional approvals were obtained (protocol number: 2010-10404, Yale University's Institutional Animal Care and Use Committee). Heparinized blood from healthy volunteers was obtained with written informed consent under an approved protocol of the Human Investigations Committee of Yale University School of Medicine. Informed consent was obtained from all subjects. The sample size was chosen based on power analysis. We did not exclude any samples. No randomization or blinding was done. We tested and authenticated the cell lines for myco-plasma contamination.
293T cells seeded on 24-well plates were transiently transfected with 50 ng of the luciferase reporter plasmid together with a total of 600 ng of various expression plasmids or empty control plasmids. As an internal control, 10 ng of pRL-TK was transfected simultaneously. Dual luciferase activity in the total cell lysates was quantified 24–36 h after transfection.
These experiments were performed as described12 with cell imaging using an Olympus BX51 microscope or a Leica TCS SP2 confocal system under a 25× objective.
The activity of type I interferon and TNF was measured as described25, with reference to recombinant human or mouse IFN-β (R&D Systems) as a standard with 2fTGH cells stably transfected with an interferon-sensitive (ISRE) luciferase construct.
The VSV plaque assay was performed as described48. For the HSV-1 and WNV plaque assays ~1 × 105 cells were infected with virus, and 24 h later supernatants were collected and used to infect confluent Vero cells. After a 1-h incubation, supernatants were removed, and cells were washed with PBS, culture medium containing 2% (wt/vol) methylcellulose was overlaid for 60 h, cells were fixed for 30 min with 0.5% (vol/vol) glutaraldehyde and then stained with 1% (wt/vol) crystal violet dissolved in 70% ethanol. Plaques were counted, and average counts were multiplied by the dilution factor to determine the viral titer as plaque-forming units per milliliter.
293 cells were transfected with plasmids encoding ELF4 or MAVS, or infected with SeV and harvested after 20 h. Nuclear-cytoplasmic fractionation was conducted using the NE-PER Nuclear and Cytoplasmic Extraction Reagents kit (Thermo Fisher Scientific) according to the manufacturer's protocol.
293T cells, HeLa cells, MEFs and macrophages were infected with VSV-GFP and harvested at the indicated time. Cells were fixed by intracellular fixation buffer (00-8222-49, eBioscience) and analyzed by FACS Calibur (BD Biosciences).
Total RNA was isolated from cells using the RNeasy RNA extraction kit (Qiagen), and cDNA synthesis was performed using 1 μg of total RNA (iScriptcDNA Synthesis kit). qPCR was done with gene-specific primers and 6FAM-TAMRA (6-carboxyfluorescein–N,N,N′,N′-tetramethyl-6-carboxyrhodamine) probes or inventoried gene expression kits from Applied Biosystems (6FAM-MGB (6-carboxyfluorescein minor-groove binder) probes).
ChIP was conducted using Magna Chip A/G (Millipore). Ifnb1and Ifna2 promoter primers are listed in Supplementary Table 3. ChIP samples were analyzed by real-time qPCR with the SYBR-Green Master Mix system (Bio-Rad).
Levels of biologically active interferon in serum were measured using an EMCV L929 cytopathic effect bioassay as described49. Results were compared with a standard curve using recombinant mouse IFN-α (PBL Laboratories).
Oligonucleotides for the EMSA were amplified with either a biotin-labeled or unlabeled primer using the wild-type and mutant IFN-β–Luc as template, and purified by Qiagen PCR kits. EMSAs were performed by incubating 0.5 ng of labeled probe with nuclear extract or purified protein (10 μg) and competing oligonucleotides in binding buffer (10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol and 1 μg/μl poly(dI·dC) at room temperature for 30 min. Mixtures were size-fractionated on a nondenaturing 4% polyacrylamide gel followed by transfer to nitrocellulose membranes and detected using streptavidin-HRP chemiluminescence for biotin-labeled probes (Light Shift Chemiluminescent EMSA Kit; Pierce).
PCDNA 3.1PCBP2, ELF4 and STING were expressed in vitro using TnT T7 Quick Coupled Transcription-Translation System (Promega, L1170). The kinase activity was analyzed by TBK1 kinase enzyme system and ADP-Glo assay (Promega, V8291).
NK or NKT cells were isolated from splenocytes of wild-type mice using NK cell isolation kit II (Miltenyi Biotec, 130-096-892) or NK1.1+ iNKT cell isolation kit (Miltenyi Biotec, 130-096-513). Purity of isolated cells was >95% upon analysis by flow cytometry. 1 × 106 cells NK cells and 1 × 106 NKT cells were injected intravenously into recipient mice.
ELF4 siRNA (catalog number: SASI_Hs01_00189458, SASI_Hs01_00189459, SASI_Hs01_00189460), IRF3 siRNA (catalog number: SASI_Hs02_00332144, SASI_Hs02_00332145), TRIF siRNA (catalog number: SASI_Hs01_00226929, SASI_Hs01_00226930, SASI_Hs01_00226931), MyD88 siRNA (catalog number: SASI_Hs01_00111539, SASI_Hs01_00111540, SASI_ Hs01_00111541), STING siRNA (catalog number: SASI_Hs02_00371843, SASI_Hs01_00031029, SASI_Hs01_00031030), MAVS siRNA (catalog number: SASI_Hs01_00128708, SASI_Hs01_00128711) and scrambled siRNA (catalog number: SASI_Hs01_00224995) were purchased from Sigma. RelA siRNA (catalog number: SR304030A, SR304030B, SR304030C) and scrambled siRNA were from Origene. TBK1 siRNA (catalog number: cs-39058) was from Santa Cruz.
Primary MEF cells were transfected using mouse/rat nucleofector solution (Lonza, M-FB-481) and nucleofector II system.
‘Pcmv5-humanELF4’ has been described previously34, and the sequence encoding full-length ELF4 was subcloned to Pcmv7.1 and Pcmv3xHA vectors. Deleted, truncated and point mutants of ELF4 were generated by the QuickChange site-directed mutagenesis kits (Stratagene) and the construct encoding wild-type protein as the template (Supplementary Table 3). Each mutation was confirmed by sequencing. After cDNA encoding human ELF1, ELF2, ELF3, ELF5 or mouse ELF4 was amplified from cDNA made from 293, U937 or L929 cells, it was cloned into plasmid p3xFLAG-CMV-7.1 (Sigma), pCMV-3XHA or pcDNA3.1(+)-3′VSV (Invitrogen) to generate plasmids encoding Flag-tagged, HA-tagged or VSV-tagged protein, respectively. pCMV4-p50(21965) and pCMV4-p65(21966) are from Addgene. The plasmids encoding human STING, MAVS, TBK1, IRF3, IRF7 and STK38 have been described28. For reporter assays, pISRE-Luc (Stratagene) and pNifty-Luc (Invivogen), an NF-κB–dependent E-selectin luciferase reporter plasmid, were used. Other promoters were cloned into pGL3-Basic vector (Promega) with the following promoter regions: −1 to −155 for mouse Ifnb1, and −1 to −476 for mouse Ifna4. All constructs were verified by sequencing the relevant regions. Point mutants of IFN-β reporter were generated by the QuickChange site-directed mutagenesis kits (Stratagene), and the construct encoding wild-type IFN-β–Luc as the template (Supplementary Table 3). Each mutation was confirmed by DNA sequencing.
Student's t-test, log-rank test and nonparametric Mann-Whitney analysis were used to analyze data. Data are presented are means ± s.e.m. unless otherwise noted, and experiments were repeated at least three times. Results were considered significant at P < 0.05.
We thank S. Nimer (University of Miami) for the Elf4−/− mice, Z. Jiang (Peking University) for the reporter plasmids. This work was supported by the US National Institutes of Health (N01-HHSN272201100019C, AI099625 and AI08992). E.F. and A.I. are funded by the Howard Hughes Medical Institute.
F.Y. and E.F. designed the study, analyzed the data and wrote or revised the paper. F.Y. performed most experiments. R.L., L.Y. and P.W. analyzed the data. P.W. provided technical support and contributed to animal work. P.W. and L.Y. contributed equally to the second authorship. F.Q. G.Y. and Y.O.Z. provided expertise and contributed to flow cytometry and immunofluorescence experiments and viral infections. L.Y. and W.W. provided technical help. R.M., R.S. and A.I. provided expertise and contributed to experiment with viral infection.
Note: Any Supplementary Information and Source Data files are available in the online version of the paper.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.