Multiple Functions of the IKK-Related Kinase IKKε in Interferon-Mediated Antiviral Immunity

See allHide authors and affiliations

Science  02 Mar 2007:
Vol. 315, Issue 5816, pp. 1274-1278
DOI: 10.1126/science.1136567


IKKϵ is an IKK (inhibitor of nuclear factor κBkinase)–related kinase implicated in virus induction of interferon-β (IFNβ). We report that, although mice lacking IKKϵ produce normal amounts of IFNβ, they are hypersusceptible to viral infection because of a defect in the IFN signaling pathway. Specifically, a subset of type I IFN-stimulated genes are not activated in the absence of IKKϵ because the interferon-stimulated gene factor 3 complex (ISGF3) does not bind to promoter elements of the affected genes. We demonstrate that IKKϵ is activated by IFNβ and that IKKϵ directly phosphorylates signal transducer and activator of transcription 1 (STAT1), a component of ISGF3. We conclude that IKKϵ plays a critical role in the IFN-inducible antiviral transcriptional response.

Activation of innate immunity by virus infection begins with type I IFNα and IFNβ gene expression followed by induction of the Janus kinase–signal transducer and activator of transcription (JAK-STAT) pathway, leading to the expression of a large family of IFN-stimulated genes (ISGs) (1). The initial response is triggered by pattern recognition receptors that bind to virus-specific molecular signatures and activate latent kinase complexes such as the stress-activated protein kinases (c-Jun N-terminal kinase and p38), the IKK complex (IKKα/IKKβ/IKKγ), and the IKK-related kinases TANK-binding kinase 1 (TBK1) and IKKϵ [also called IKKi (2)] (3). These kinases coordinate the assembly of the IFNβ enhanceosome, a multisubunit complex composed of the transcription factors ATF2 (activating transcription factor 2)/cJun, NFκB, and interferon regulatory factors 3 and 7 (IRF3 and IRF7) (3). IFNβ induces dimerization of the type I IFN receptor and activates the associated kinases tyrosine kinase 2 (TYK2) and JAK1 (1), leading to the tyrosine phosphorylation of STAT1 and STAT2 and STAT association with IRF9 to form the ISGF3 complex. This complex binds to ISREs (interferon-stimulated response elements) in the promoters of a large group of ISGs, resulting in their transcriptional activation (4).

Initial characterization of the kinases responsible for IFNβ transcription suggested that the virus-inducible expression of IKKϵ functions in a redundant role to its ubiquitous counterpart, TBK1, in the activation of IRF3 and IRF7 ex vivo (58). To investigate the function of IKKϵ in vivo, we generated mice deficient in IKKϵ (fig. S1A). Disruption of Ikbke–/– (the gene encoding IKKϵ) resulted in a complete loss of the kinase in both mice and embryonic fibroblasts (EFs) (fig. S1B). To determine whether IKKϵ is required to protect against virus infection in vivo, we infected wild-type (WT) and Ikbke–/– mice with increasing doses of influenza A/WSN/33 virus (WSN). Although inoculations of more than 7 × 103 plaque-forming units (pfu) resulted in 100% mortality of both cohorts, Ikbke–/– mice displayed an increased susceptibility to WSN at lower titers (Fig. 1A). After inoculation with 700 pfu, both WT and Ikbke–/– mice displayed greater than 10% body weight loss, although unlike WT controls Ikbke–/– mice continued to decline below 80% of their original body mass before succumbing to the infection (Fig. 1B).

Fig. 1.

IKKϵ knockout mice are hypersusceptible to influenza virus infection. (A) Mice were administered increasing amounts of WSN intranasally and observed over a 7-day period (n = 5 mice per cohort per dose). (B) Cohorts of mice were administered 7 × 102 pfu intranasally and weighed daily (n = 20 mice per cohort per dose). Values represent average scores of overall weight loss compared with initial body mass.

Quantitative polymerase chain reaction (Q-PCR) for virus nucleocapsid mRNA and virus plaque assays from lung tissue indicated that Ikbke–/– mice had an elevated viral load compared with that of their control littermates (Fig. 2A and fig. S2A). Histopathology of lungs at seven days post infection (dpi) revealed that infected Ikbke–/– mice exhibited an inflammatory infiltrate consisting of lymphocytes, macrophages, and neutrophils (fig. S2B). However, reverse transcription PCR (RT-PCR) and enzyme-linked immunosorbent assay (ELISA) analyses of lung and serum, respectively, revealed no obvious deficiency in virus-induced cytokines, such as IFN α, β, and γ, RANTES, and IL2, or in the amounts of virus-specific antibodies (Fig. 2B and figs. S3 and S4A). By contrast, array (Affymetrix, Santa Clara, CA) analysis of RNA from pooled primary lung samples revealed a subset of ISGs, including Ifit3, Ifi203, and the double-stranded RNA (dsRNA)–activated adenosine deaminase gene (Adar1), that were poorly induced in the absence of IKKϵ as corroborated by RT-PCR analyses (Fig. 2B and fig. S4B). To determine whether the resulting loss of ADAR1 has functional consequences, we amplified and sequenced a stem loop structure present in WSN matrix (M) mRNA from infected lung tissue. ADAR1 binds double-stranded RNA (dsRNA) and deaminates adenosine residues, thus converting adenine to guanine and thereby mutagenizing viral RNA (9). Analysis of over 95 individual clones from each cohort revealed that greater than 30% of the viral M mRNA bears at least one or more A-to-G transitions in control mice, compared with less than 5% in Ikbke–/– mice (Fig. 2C). We conclude that IKKϵ is required for the induction of a subset of ISGs such as ADAR1, which displays antiviral activity in vivo.

Fig. 2.

IKKϵ is required for the normal host response to infection. (A) Viral expression from infected mice (7 × 102 pfu) was determined by Q-PCR from pooled lung samples. Values determined as ratios of viral nucleocapsid protein (NP) relative to Hprt. (B) Lung transcriptional profiles of IFNβ (Ifnb), IFNα (Ifna), Adar1, tumor necrosis factor α (Tnfa), NP, Ikbke, and Hprt as determined by RT-PCR. (C) ADAR1 activity as measured by influenza matrix (M) RNA sequencing. M was reverse transcribed from pooled lung samples (8 dpi) and sequenced for A to G substitutions (n = 96 transcripts per cohort). Underlined adenosines represent positions of base transitions. (D) Viral titers of primary EFs infected with WSN (multiplicity of infection = 0.01) as determined by plaque assay. Values represent the mean average of triplicate experiments. Error bars indicate standard error of the mean. (E) EF transcriptional profiles of Ifnb, Adar1, IFN-activated gene 203 (Ifi203), NP, Ikbke, and Hprt by RT-PCR.

To determine whether the in vivo phenotype of Ikbke–/– mice could be recapitulated ex vivo, we compared the rates of influenza virus replication in primary EFs from WT and Ikbke–/– mice. Comparable replication rates were observed at 2 dpi, although the titer continued to increase in Ikbke–/– cells relative to WT cells thereafter (Fig. 2D). RT-PCR analyses from the WT and Ikbke–/– samples showed comparable induction of IFNβ but revealed a decrease in the expression of a subset of ISGs (Fig. 2E). A low amount of virus-inducible ADAR1 mRNA was observed in Ikbke–/– EFs; however, the induction of other ISGs such as IFI203 was completely blocked. Thus, it appears that IKKϵ controls the optimal expression of a subset of ISGs required to control viral load both in vivo and ex vivo.

Certain ISGs are inducible by both virus infection and IFN treatment, whereas others are inducible only by IFN directly (10). To determine whether the IKKϵ-dependent ISGs are defective in IFN signaling, we treated WT and Ikbke–/– EFs with recombinant IFNβ (rIFNβ). About 30% of the IFN-inducible ISGs were poorly induced in Ikbke–/– EFs (Fig. 3A). This result was confirmed by RT-PCR analyses (Fig. 3B). Although the amount of IFN induction of several ISG genes, such as Adar1, Ifit3, and Ifi203, decreased in the absence of IKKϵ, others, such as Irf 7, Prkra (RNA-activated protein kinase), and Stat1, were unaffected. Alignment of the DNA sequences of IKKϵ-dependent ISREs identified a consensus sequence similar to previously characterized ISGF3 binding sites (4, 11)(fig. S5A). Electrophoretic mobility shift assays (EMSAs) using these ISREs identified an IFN-inducible DNA-protein complex that failed to bind to IKKϵ-dependent promoters in extracts from IFN-treated Ikbke–/– cells (Fig. 3C). This complex was shown to be ISGF3, because IRF9, STAT1, and STAT2 antibodies all disrupted DNA-protein complex formation (Fig. 3D). Expression of IKKϵ but not a dominant negative mutant of IKKϵ [Lys38→Ala38 (K38A)] rescued IFNβ induction of ISGF3 binding in Ikbke–/– cells (Fig. 3E). Thus, ISGF3 binding to IKKϵ-dependent promoters requires IKKϵ kinase activity. To determine whether IKKϵ can be activated by IFN treatment, we transfected EFs, which express minute quantities of IKKϵ, with WT IKKϵ or the K38A mutant and treated the EFs with rIFNβ. Because Thr501 (T501) of IKKϵ has been implicated in virus-induced activation (12), we used a phosphospecific T501 antibody to observe a rapid induction of IKKϵ phosphorylation in IFN-treated cells (Fig. 4A). TBK1, which is not activated by IFN treatment, does not contain this T501 residue. On the basis of the amino acid sequence of the phosphorylation site in IKKϵ and previous studies that implicate p38 kinase in IFN- and virus-inducible phosphorylation, we speculate that IKKϵ activation may be mediated through p38 kinase signaling (13).

Fig. 3.

The type I IFN response is defective in Ikbke–/– EFs. (A) Gene expression profile in primary EFs treated with rIFNβ (0.1 U/ml) for 6 hours. The heat map depicts the average mean induction of ISGs from triplicate Affymetrix samples. (B) Transcript quantities of Adar1, IFN-induced protein with tetratricopeptide repeats 3 (Ifit3), Ifi203, IRF7 (Irf7), Stat1, Ikbke, and Hprt as determined by RT-PCR from rIFNβ-treated EFs. hpt, hours posttreatment. (C) Cell extracts from rIFNβ-stimulated EFs were analyzed by EMSA with ISRE elements derived from the 2′-5′ oligoadenylate synthetase 1B (Oas1b), myxovirus resistance 1 (Mx1), and Adar1 genes. The ns band denotes nonspecific binding. (D) Protein composition of ISGF3 was analyzed by antibody (Ab.) competition assays with IRF9, STAT1, and STAT2 antibodies. CTRL, control. (E) Ikbke–/– EFs were transfected with vector, IKKϵ, or K38A. Transfections were untreated or treated with rIFNβ and analyzed by EMSA on an Adar1 ISRE.

Fig. 4.

IKKϵ is activated by IFNβ and determines ISGF3 binding specificity. (A) IFNβ-treated whole cell extract from Ikbke–/– EFs, transfected with human IKKϵ WT or K38A, were immunoblotted with IKKϵ phosphospecific T501, IKKϵ, or β-actin. (B) rIFNβ-stimulated EFs were analyzed by EMSA using Adar1 and Irf7 ISREs. (C) rIFNβ-stimulated EFs were cross-linked and assayed by ChIP assay. (D) rIFNβ-stimulated cell extracts from Stat1–/– EFs transfected with STAT1 WT, S708A, Δ744/747, or the double mutant (AΔ) were analyzed by EMSA using Adar1 and Irf7 ISREs. (E) Stat1–/– EFs treated as in (D) were cross-linked and assayed by ChIP assay.

The Ikbke–/– phenotype is similar to that of Stat1–/– mice, which also displays an increased susceptibility to influenza virus infection both in vivo and ex vivo. However, in contrast to Ikbke–/– cells, where 30% of ISGs fail to be induced by IFN, no ISGs are induced in Stat1–/– cells (14, 15). This suggests that STAT1 signaling is not completely compromised in the absence of IKKϵ. This observation is consistent with EMSA and chromatin immunoprecipitation (ChIP) analyses of IKKϵ-independent ISREs showing normal binding to ISGF3 in Ikbke–/– cells treated with IFNβ (Fig. 4, B and C). In addition, bone marrow–derived macrophages from Ikbke–/– mice display normal induction of IFNγ-stimulated genes (fig. S5B).

As critical components of ISGF3, STAT1 and STAT2 are potential targets for IKKϵ. We therefore investigated the status of these two transcription factors in response to IFN in Ikbke–/– mice. Initially, we carried out Western blot experiments with primary lung extracts from mice infected with different viruses. The quantities of both STAT1 and STAT2 increased normally in response to infection; however, discrete break-down products of STAT1 were observed in WT mice but not in Ikbke–/– mice (fig. S6, A and B). In primary EFs, the amount of STAT1 in Ikbke–/– IFNβ-treated cells exceeded that of WT cells (fig. S6C). In considering these results together, we suggest that STAT1 appears to be subject to IKKϵ-dependent processing and/or degradation, an event that has been ascribed to STAT1 previously (16). This explanation also suggests the possibility that STAT1 may be a target of IKKϵ.

To determine whether IKKϵ acts directly on STAT1, we carried out in vitro phosphorylation studies with recombinant IKKϵ. Mass spectrometry analyses revealed that IKKϵ phosphorylates Ser708 (S708), S744, and S747 (fig. S7A). To determine the function of these serine residues, we created STAT1 mutants and transfected them into Stat1–/– EFs. Transfected cells were treated with IFN and assayed for ISGF3 binding (Fig. 4, D and E). The substitution of S708 with alanine (S708A) dramatically decreased ISGF3 binding to the IKKϵ-dependent Adar1 ISRE but not to the IKKϵ-independent Irf7 ISRE, as determined by both ChIP and EMSA experiments. In contrast, the binding of ISGF3 to both ISREs was unaffected by a C-terminal deletion of STAT1 that removed serine residues 744 and 747 (ΔS744/747).

Examination of the three-dimensional structure of STAT1 predicts that phosphorylation of S708 would favor the formation of STAT1-STAT2 heterodimers, rather than STAT1 homodimers (17). Heterodimer formation is necessary for the assembly of ISGF3 because STAT2 is thought to tether STAT1 and IRF9 to the ISRE (18). ISGF3 binds to adjacent GAAA repeats recognized by STAT1 and IRF9, respectively, on complementary major and minor grooves. STAT2 in turn, contacts both proteins while making additional contacts with sequences upstream from the GAAA repeats (fig. S7B) (4, 11, 17, 19). Computational analyses of the DNA sequences of IKKϵ-dependent and -independent ISREs identify a purine-rich region upstream of the IKKϵ-independent ISREs. This purine tract could serve as an additional STAT2 binding site, permitting specific contacts with DNA. In the absence of this purine-rich sequence, ISGF3 binding may require a more stable STAT1-STAT2-IRF9 interaction, which in turn may require the phosphorylation of STAT1 S708 by IKKϵ. Although the structural consequences of this phosphorylation remain to be determined, the function of IKKϵ in type I IFN signaling is to guide the transcriptional machinery to a subset of ISGs required for a direct antiviral response. By contrast, the IKKϵ-independent genes may function primarily in regulating the IFN signaling machinery, which is required for the integration of innate and adaptive immune systems. These results emphasize the importance of the interplay between local and systemwide antiviral mechanisms. Even when the systemwide antiviral response is intact, defects in the local response lead to an increase in viral load, ultimately overwhelming the immune defenses.

Supporting Online Material

Materials and Methods

Figs. S1 to S7

References and Notes

View Abstract

Navigate This Article