IRF7 and ISGF3 engage overlapping interferon-stimulated response elements and drive largely overlapping antiviral transcriptomes, so that a substantial interferon-like gene program is still induced when type I and type III interferon signaling are both absent.
Schmid S, Mordstein M, Kochs G, García-Sastre A, tenOever BR. Transcription Factor Redundancy Ensures Induction of the Antiviral State. Journal of Biological Chemistry. 2010. Volume 285, issue 53, pages 42013-42022. DOI 10.1074/jbc.m110.165936. PMID 20943654. PMCID PMC3009927.
IRF7 and ISGF3 engage overlapping interferon-stimulated response elements and drive largely overlapping antiviral transcriptomes, so that a substantial interferon-like gene program is still induced when type I and type III interferon signaling are both absent.
The prevailing account of the antiviral transcriptional response placed interferon secretion and subsequent ISGF3 activation at its center, with interferon regulatory factors acting mainly upstream by inducing interferon itself. Infection of mice lacking both the type I and the type III interferon receptors with an NS1-deficient influenza A virus showed that a large block of genes normally classed as interferon-stimulated genes was still induced, which raised the question of which transcription factor was responsible and how its DNA binding preferences relate to those of ISGF3.
The study combined biochemical dissection of a single well-characterized element, the ISG15 interferon-stimulated response element, with transcriptional profiling in cells and animals unable to signal through interferon. Systematic mutagenesis of the core motif and its flanking sequence, read out by electrophoretic mobility shift assay against reconstituted IRF7 and reconstituted ISGF3, separated positions that confer IRF7 specificity, ISGF3 specificity, or binding by both. Profiling of STAT1-deficient U3A cells in which IRF7 was activated by IKK epsilon defined an IRF7-driven transcriptome in the absence of interferon signaling, and about 80 percent of the genes induced in that setting were also induced in the infected knockout mouse lung.
The work reframes the interferon-stimulated response element as a family of motifs with distinguishable factor preferences and supports a model in which IRF7 provides redundancy that secures the antiviral state when interferon signaling is unavailable.
Type I and type III interferons signal through distinct receptors that converge on the assembly of ISGF3, a complex of STAT1, STAT2 and IRF9, which binds interferon-stimulated response elements upstream of more than one hundred interferon-stimulated genes. The IRF family binds a partially overlapping element, and IRF3 and IRF7 had been assigned the upstream role of inducing interferon itself. IRF3 is ubiquitously expressed, while basal IRF7 is largely restricted to hematopoietic cells and is induced elsewhere by interferon or tumor necrosis factor alpha. Earlier profiling of IRF3 and of IRF7 activity existed, but no distinction was drawn between an element that binds IRFs, one that binds ISGF3, and one that binds both. The paper states the gap in those terms. A prior report by Mordstein and colleagues on mice lacking both interferon receptors provided the animal model, and the observation of interferon-stimulated gene induction in those animals despite the absence of ISGF3 activity is what motivated the study.
Do IRF7 and ISGF3 act on the same DNA elements, and can IRF7 alone reconstruct an interferon-like transcriptome in the absence of type I and type III interferon signaling, thereby providing redundancy in establishing the antiviral state?
The design pairs an in vivo demonstration that the phenomenon exists with an in vitro dissection of why it exists. Mice carrying disruptions in IFNAR1 and IL28R alpha were infected with wild-type or NS1-deficient influenza A virus, which removes the viral antagonist of IRF activation, so that virus-driven, interferon-independent transcription could be measured by microarray and quantitative PCR. For biochemistry, IRF7 or the three ISGF3 components were expressed exogenously in fibroblasts and activated with IKK epsilon or interferon beta, giving matched extracts whose binding to a panel of synthetic probes could be compared directly. The ISG15 element was chosen because it carries both a complete ISGF3 consensus and a complete IRF element, so that core positions and flanking positions could be mutated one at a time. STAT1-deficient U3A cells provided a human cellular system in which IRF7 activity can be read transcriptionally with no possibility of interferon feedback, and primary fibroblasts from Irf3, Irf7 and double knockout mice, treated with cycloheximide to block autocrine signaling, provided a genetic test using endogenous factors.
In mice lacking both interferon receptors, NS1-deficient influenza A virus induced interferon beta and interferon lambda 2 transcripts and a set of genes annotated as interferon-stimulated genes, at comparable viral nucleoprotein levels between viral cohorts. Some interferon-stimulated genes, notably Mx1, were not induced (Figure 1). The observation is direct. The inference drawn by the authors, that this residual program reflects IRF3 and IRF7 activity, is interpretation at this stage of the paper.
Exogenously expressed IRF7 activated by IKK epsilon, and reconstituted ISGF3 activated by interferon beta or IKK epsilon, both drove an ISG15 element reporter and both bound an ISG15 element probe. Supershift assigned the IRF7 complex to an IRF7 homodimer rather than an IRF3 and IRF7 heterodimer, and confirmed STAT1, STAT2 and IRF9 in the ISGF3 complex. Monomeric IRF9 binding was detectable and was inversely related to ISGF3 assembly (Figure 2).
Systematic probe mutagenesis separated the requirements of the two complexes. Minimizing the element to a single core raised ISGF3 and IRF9 binding, randomizing both flanks abolished ISGF3 and IRF9 binding while leaving IRF7 binding intact, and individual core substitutions produced positions that affect only ISGF3, only IRF7, or both (Figure 3). The authors derive a minimal IRF7 motif of AAWNCGAAA, with WWNNGAAANNGAAA also compatible, and an ISGF3 consensus of WBVGGAAANNGAAACT. These consensus statements are the authors' generalization from the probe panel.
Favorable bases at the minus 2 and minus 3 positions compensated for an unfavorable base at core position 7 that otherwise abolishes IRF7 binding (Figure 3N). The authors read this as IRF7 making minor groove contacts at those upstream positions.
In STAT1-deficient U3A cells, IRF7 with IKK epsilon induced interferon alpha 1 without inducing MxA, and microarray profiling showed induction not only of known IRF3-regulated genes such as IFIT1, IFIT2, IFIT3 and RSAD2 but also of IFIT5, IFIH1, GBP1, OAS2, OASL and IRF9 (Figure 4). About 80 percent of the genes upregulated in this setting overlapped the set induced in the infected knockout mouse lung. The overlap figure is an observation. The further statement that the redundancy is evolutionarily conserved among vertebrates is an extrapolation offered by the authors.
In U3A cells stably reconstituted with GFP, IRF7 or STAT1 and challenged with interferon beta or an NS1-mutant influenza A virus, MxA behaved as an ISGF3-dependent gene, OAS1 was induced by both ISGF3 and IRF7, and CXCL10, MAP3K8 and interferon alpha 1 were induced by IRF7 but not by interferon beta treatment (Figure 5). Knockdown of endogenous IRF3 did not abolish these IRF7-dependent inductions.
In primary mouse embryonic fibroblasts treated with polyinosinic-polycytidylic acid under cycloheximide, CXCL10 and MAP3K8 were induced in Irf3 knockout cells but not in Irf7 knockout or double knockout cells, placing the requirement on endogenous IRF7 and excluding an autocrine interferon contribution (Figure 5C).
Elements taken from the promoters of IRF7-regulated genes sorted into the two predicted motif classes, and direct binding tests assigned the MxA element to ISGF3 only, the OAS1 element to IRF7 and ISGF3, the ISG15 element to IRF3, IRF7 and ISGF3, and the CXCL10 element to IRF7 (Table 1 and Figure 6). In extracts from interferon-treated, polyinosinic-polycytidylic acid stimulated cells, the single inducible CXCL10 element complex was supershifted only by an IRF7 antibody, which ties the ectopic expression model to endogenous activity.
The mechanism supported here is a DNA-level one and it is established at the level of binding and transcript abundance rather than by structural work. Interferon-stimulated response elements are not a single class. Core positions plus the immediate flanking sequence determine whether an element is read by IRF7, by ISGF3, or by both, with ISGF3 requiring contacts that extend beyond the consensus core, presumably through STAT1 and STAT2, and IRF7 tolerating more variation because favorable upstream minor groove contacts can offset a poor core contact. Because a large fraction of interferon-stimulated gene promoters carry elements of the permissive class, activated IRF7 alone reproduces much of the interferon-stimulated transcriptome without any interferon receptor engagement, and the cellular level of IRF7 therefore sets how much of that program a cell can mount independently of interferon.
The study does not establish the structural basis of the differential contacts. The major and minor groove assignments are inferred from mutational effects and from published structures of related complexes rather than demonstrated here. Stoichiometry of ISGF3 on DNA is explicitly left open by the authors. The relative contribution of IRF7 homodimers and any residual IRF3 and IRF7 heterodimers in the infected animal is constrained but not settled, and the authors say that heterodimer formation cannot be ruled out.
The element rather than the cytokine becomes the unit of analysis. Assigning a gene to the interferon-stimulated class no longer implies that its induction requires interferon signaling, and the paper supplies sequence criteria by which a promoter can be sorted into IRF7-responsive, ISGF3-responsive, or universal categories. The authors suggest that the defined requirements should support computational prediction of a gene's inducibility given the levels of IRF3, IRF7 and ISGF3, which is put forward as a prospect rather than demonstrated. Practically, the STAT1-deficient U3A system with inducible IRF7 activation gives a way to read IRF7 transcriptional output with interferon feedback removed.
The paper sits in the tenOever laboratory's continuing interest in how the cell-intrinsic antiviral transcriptional response is wired and in what the influenza A virus NS1 protein suppresses. The use of NS1-deficient and NS1-mutant influenza A virus as a tool to unmask host transcriptional responses recurs across the corpus. The mapping of which host genes are inducible independently of interferon signaling also informs later work in the corpus on interferon-independent antiviral mechanisms. Category 3 synthesis, visible only when this paper is placed beside Shapiro and colleagues 2014 on Drosha as an interferon-independent antiviral factor, is that the laboratory pursued interferon-independent arms of antiviral defense from two directions, one transcriptional and one post-transcriptional, in the same period.
The transcriptomic conclusions rest on Affymetrix arrays from pooled material, three lungs pooled per cohort at each time point with no biological replication at the array level in the animal experiment, with validation by quantitative PCR on a small number of genes. The IRF7 transcriptome is defined under exogenous overexpression of IRF7 together with IKK epsilon in a single transformed human fibrosarcoma line lacking STAT1, which is a strong and non-physiological activation regime, and the endogenous validation is limited to one element, CXCL10, and to a small panel of genes in primary mouse fibroblasts. The binding analysis is built almost entirely on derivatives of one element, the ISG15 element, so the derived consensus sequences are generalizations from a single sequence context rather than from an unbiased selection. Binding was measured with cell extracts on naked oligonucleotides, so chromatin context, cooperativity with other virus-activated factors, and promoter occupancy in cells are not addressed. The infection work uses one virus, influenza A virus, in one mouse background carrying intact Mx1 alleles, and the interferon-independent program was defined with an NS1-deficient or NS1-mutant virus rather than with wild-type virus. The proposal that restricted IRF7 expression exists to avoid unnecessary toxicity is stated by the authors as speculation and is not tested.
Cells lacking interferon receptors still switch on much of the antiviral gene program, because IRF7 reads many of the same promoter elements that the interferon-activated ISGF3 complex uses.
Antiviral gene induction is usually attributed to interferon signaling through the ISGF3 complex. Mice lacking both type I and type III interferon receptors nonetheless induced many interferon-stimulated genes after influenza A virus infection. Systematic mutagenesis of a model promoter element showed that IRF7 and ISGF3 bind overlapping but distinguishable sequences, and activating IRF7 in interferon-unresponsive human cells reproduced most of that gene set, indicating a redundant route to the antiviral state.
The antiviral transcriptional response has been framed as a consequence of interferon secretion and ISGF3 assembly. Infection of IFNAR1 and IL28R alpha double knockout mice with NS1-deficient influenza A virus induced a substantial set of interferon-stimulated genes despite the absence of ISGF3 activity. Working from the ISG15 interferon-stimulated response element, the authors mutated the core motif and its flanks and compared binding of reconstituted IRF7 and reconstituted ISGF3 by mobility shift, resolving positions that confer specificity for one complex, for the other, or for both, and deriving candidate consensus sequences for each. Activation of IRF7 in STAT1-deficient U3A cells induced a gene set overlapping about 80 percent with the set induced in infected knockout lungs, and promoter elements from individual genes bound the predicted factors, with MxA restricted to ISGF3 and CXCL10 restricted to IRF7. Endogenous IRF7 was required for CXCL10 and MAP3K8 induction in knockout fibroblasts under translational blockade.
lab-led for Schmid 2010 and Schmid 2014, co-led with the Maniatis laboratory for Ng 2011, and training period for tenOever 2007, which was carried out with the Maniatis and García-Sastre laboratories and is not the independent program's work