IRF
The nine human interferon regulatory factors, IRF1 through IRF9, bind DNA with an N-terminal winged helix-turn-helix domain studded with five conserved tryptophans. Each domain reads only a short AANTGAAA half-site, so functional elements are tandem repeats of it: the interferon-stimulated response element (ISRE), written RNGAAANNGAAACT or, on the other strand, AGTTTCNNTTTC. The same half-site also appears in two named composites - the EICE with PU.1 or SPIB, and the AICE with BATF-JUN - which is how IRF4 and IRF8, weak DNA binders on their own, reach their targets. IRFs drive the type I and type II interferon responses, myeloid and lymphoid development, and plasma cell differentiation; IRF6 is the non-immune outlier, acting in epithelium, where mutations cause orofacial clefting.
The factors
Nine IRF genes exist in humans, IRF1 to IRF9 18303999 (an IRF10 is present in birds and fish but not in mammals). Each encodes an N-terminal DNA-binding domain of roughly 115 residues characterized by a cluster of five conserved tryptophans - the "tryptophan cluster" that defines the family - folded as a variant helix-turn-helix 9422515. Most members also carry a C-terminal IRF-association domain that mediates homodimers, IRF-IRF heterodimers, and complexes with non-IRF partners: STAT2 for IRF9, BATF-JUN for IRF4 and IRF8, PU.1 and SPIB for IRF4 and IRF8 12453417,22983707,29317535. Functionally the family splits along lines that sequence does not see. IRF1 and IRF2 are broadly expressed and act as an interferon-inducible activator/repressor pair on the same sites. IRF3 is constitutive and cytoplasmic until phosphorylated on virus sensing, whereupon it dimerizes and drives type I interferon genes; IRF7 amplifies that response and is highly expressed in plasmacytoid dendritic cells 18303999. IRF9 is the constitutive DNA-binding subunit of ISGF3 29317535. IRF4 and IRF8 are lymphoid and myeloid, with low intrinsic DNA affinity; IRF5 acts in myeloid cells and B cells. IRF6 is the outlier - epithelial, expressed in skin and oral periderm, with no established role in interferon signaling 12219090.
The motif
The unit is a short half-site, AANTGAAA, of which only about four bases (the GAAA) are read strongly 9422515,10487755. The crystal structure of the IRF1 domain on the PRD I element of the interferon-beta promoter showed why: the recognition helix lies obliquely in the major groove, contacts a short GAAA core through three of the five conserved tryptophans, and bends the DNA toward the protein 9422515. Because so little sequence is read per domain, affinity and specificity come from repetition, and the IRF2 structure showed cooperative binding to a tandem GAAA repeat driven by DNA distortion 10487755. The canonical element is therefore the ISRE, written RNGAAANNGAAACT 29317535; the data consensus for this family, CGAAACCGAAACT, is the same element read on the opposite strand, whose reverse complement AGTTTCGGTTTCG is the familiar AGTTTCNNTTTC form. Two composites reuse the same half-site with a different partner. The EICE, GGAAGTGAAAC, is a PU.1 or SPIB site fused to an IRF half-site 7797077,12453417; the AICE joins an IRF half-site to an AP-1 site, either as AICE1 (TTTCNNNNTGASTCA, four bases of spacer) or AICE2 (GAAATGASTCA, directly abutting) 22983707,28346410. Paralogue resolution from sequence is close to zero: the in vitro models for all nine IRFs are near-identical GAAA repeats, and what separates them in vivo is which partner they cooperate with, not which bases they read.
Where it acts
ISREs sit predominantly in promoters of interferon-stimulated genes, usually within a few hundred base pairs of the transcription start site, and in virus-inducible enhancers such as the interferon-beta enhanceosome, where PRD I and PRD III are IRF elements 9422515. In lymphoid and myeloid cells, IRF4 and IRF8 occupancy is overwhelmingly composite: EICEs at immunoglobulin light-chain enhancers 7797077, where DNA-bound and Ser148-phosphorylated PU.1 raises IRF4 affinity 20- to 40-fold, and AICEs with BATF-JUN in T cells and dendritic cells 22983707. AICE affinity grades the response to signal strength: high-affinity AICEs remain occupied under weak T-cell receptor stimulation while low-affinity ones only under strong stimulation 28346410. ISGF3, the STAT1-STAT2-IRF9 trimer, accounts for most ISRE occupancy in a type I interferon response, with the STAT1-STAT2 pair on the 5' GGAA and IRF9 on the second repeat 29317535. IRF6 works instead in epithelial enhancers of the periderm and keratinocyte programs.
Biology and disease
IRFs are the transcriptional output of innate immune sensing and are also lineage regulators. Human IRF8 mutations cause dendritic cell immunodeficiency, with biallelic loss abolishing monocytes and dendritic cells 21524210. IRF4 is required for plasma cell differentiation, and multiple myeloma lines die when IRF4 is depleted whether or not IRF4 itself is mutated 18568025. A common IRF5 haplotype that alters splicing and expression is a reproducible risk factor for systemic lupus erythematosus 16642019. IRF6 mutations cause Van der Woude syndrome (cleft lip or palate with lower lip pits) and popliteal pterygium syndrome 12219090, the clearest case of an IRF acting outside immunity.
Reading a hit
A single AANTGAAA half-site is not evidence of anything: with roughly six specified bases it is expected on the order of a million times per strand in a 3.1-Gb genome. Require the tandem repeat, or a composite. The most common misreading is confusion with STAT. A STAT homodimer site (GAS) is an inverted repeat, TTCNNNGAA, with the two half-sites facing each other; an ISRE is a direct repeat, GAAANNGAAAC, with both half-sites in the same orientation - and its reverse-complement spelling AGTTTCNNTTTC superficially resembles a GAS. The two families genuinely overlap in biology as well, because ISGF3 contains STAT1 and STAT2 29317535, so ISRE-shaped matrices appear under both labels; one footprint archetype grouped in this family is named AC0672:IRF/STAT precisely because the clustering could not separate them. Evidence for this family is comparatively good - most HOCOMOCO models here carry in vitro HT-SELEX or methyl-HT-SELEX support, and most JASPAR profiles here are SELEX-derived - but two ChIP-derived entries have absorbed a partner's site. HOCOMOCO's ChIP-based IRF8 model (IRF8.H14CORE.0.P.B, consensus AGDGGAAGTGAAAGT) contains GGAAGTGAAA, an EICE whose 5' half is a PU.1 site, not an IRF site 7797077,12453417; and one CIS-BP model here (M07978, TGHGGAAATGARTCA) is AICE2-shaped, an IRF half-site abutting an AP-1 heptamer 22983707, and its best match across the databases lies outside the IRF family. In myeloid or B-lineage data, check for an adjacent GGAA or TGASTCA before concluding that an IRF acted alone.
How this was worked out
IRF-1 is cloned as the nuclear factor that binds the upstream regulatory region of the interferon-beta gene and mediates its virus-induced transcription; the IRF1 gene is itself virus-inducible.
Still open then: Which bases IRF-1 read, and how the same factor could act on both interferon-beta and interferon-stimulated genes, was unknown.
Pip (IRF4) is cloned as the lymphoid partner that binds the immunoglobulin light-chain enhancer composite element only together with DNA-bound, Ser148-phosphorylated PU.1, defining the ETS-IRF composite element.
Still open then: How IRF4 targets genes in cells that lack PU.1 or SPIB, such as most T cells, was unexplained.
The IRF-1 DNA-binding domain on PRD I of the interferon-beta promoter is a new helix-turn-helix that grips DNA through three of the five conserved tryptophans and reads a short GAAA core through an obliquely angled recognition helix while bending the DNA toward the protein.
Still open then: The structure suggested, but did not show, why IRF elements are built from GAAA repeats.
The IRF-2 DNA-binding domain structure gives the recognition sequence AANNGAAA and shows cooperative binding to a tandem GAAA repeat driven by DNA distortion, explaining binding to both single and repeated sites.
The PU.1/IRF4/DNA ternary structure shows the composite element contorted into an S shape that juxtaposes the two DNA-binding domains across the central minor groove, explaining cooperativity with IRF4 and IRF8 and anti-cooperativity with IRF1 and IRF2.
In T helper 17 cells IRF4 binds AP-1-IRF composite elements (AICEs) cooperatively with BATF-containing AP-1 dimers, resolving how IRF4 and IRF8 find targets without an ETS partner; the same elements operate in TH2, B and dendritic cells.
Bases flanking the AICE set the affinity of the BATF-IRF4 complex, and enhancers occupied only at strong T-cell receptor signal are low-affinity AICEs, linking motif affinity to signal strength; a high-affinity AICE2 variant at CTLA4 is associated with resistance to autoimmunity.
Crystal structures of the IRF9 association domain alone and bound to the STAT2 coiled-coil domain explain the selective STAT2-IRF9 pairing in ISGF3, and a model of ISGF3 on an ISRE places the STAT pair on the 5' GGAA repeat and IRF9 on the second GAAA repeat.
Still open then: The full ISGF3 complex on DNA has not been solved; its placement on the ISRE is a model.
Still unresolved
- Why in vitro and ChIP-derived IRF models disagree on the base before each GAAA is unresolved: most HT-SELEX and methyl-HT-SELEX models read CGAAACCGAAAC, placing a CpG in each half-site, while ChIP-derived models read TGAAACTGAAA or GAAAGTGAAA, and whether genomic CpG depletion and methylation suppress a genuine in vitro preference or the SELEX enrichment overstates it has not been tested in the sources cited here.
- How ISGF3 sits on the ISRE is inferred rather than observed: the STAT2-IRF9 interface has been crystallized, but the placement of the STAT pair on the 5' GGAA and IRF9 on the second repeat comes from a docked model, so which subunit makes which base contacts is not settled.
- The databases cannot separate ISRE-shaped STAT models from IRF models (one archetype is named IRF/STAT), and whether ISRE-shaped STAT1 and STAT2 ChIP motifs reflect contacts by IRF9 within ISGF3 or genuine STAT preference for that element cannot be decided from the matrices alone.
A hit identifies an IRF-type element, not an IRF protein: all nine paralogues select near-identical GAAA repeats in vitro, and their specificity in cells comes from partners (PU.1/SPIB at an EICE, BATF-JUN at an AICE, STAT2 in ISGF3), not from the bases they read. A lone AANTGAAA half-site is far too short to be meaningful; insist on a tandem repeat or a composite. Expect cross-talk with the STAT family in both directions, since ISGF3 contains STAT1 and STAT2 and reads the ISRE - one footprint archetype here is named IRF/STAT. Most models in this family have in vitro support, but the ChIP-derived IRF8 model contains an embedded PU.1 half-site and one CIS-BP model in this family is actually an AICE, half AP-1. Neither should be read as intrinsic IRF specificity.
References
- Escalante CR, Yie J, Thanos D, et al. Structure of IRF-1 with bound DNA reveals determinants of interferon regulation. Nature 1998;391:103-6. PMID 9422515
- Fujii Y, Shimizu T, Kusumoto M, et al. Crystal structure of an IRF-DNA complex reveals novel DNA recognition and cooperative binding to a tandem repeat of core sequences. EMBO J 1999;18:5028-41. PMID 10487755
- Tamura T, Yanai H, Savitsky D, et al. The IRF family transcription factors in immunity and oncogenesis. Annu Rev Immunol 2008;26:535-84. PMID 18303999
- Rengachari S, Groiss S, Devos JM, et al. Structural basis of STAT2 recognition by IRF9 reveals molecular insights into ISGF3 function. Proc Natl Acad Sci U S A 2018;115:E601-E609. PMID 29317535
- Escalante CR, Brass AL, Pongubala JM, et al. Crystal structure of PU.1/IRF-4/DNA ternary complex. Mol Cell 2002;10:1097-105. PMID 12453417
- Eisenbeis CF, Singh H, Storb U. Pip, a novel IRF family member, is a lymphoid-specific, PU.1-dependent transcriptional activator. Genes Dev 1995;9:1377-87. PMID 7797077
- Glasmacher E, Agrawal S, Chang AB, et al. A genomic regulatory element that directs assembly and function of immune-specific AP-1-IRF complexes. Science 2012;338:975-80. PMID 22983707
- Iwata A, Durai V, Tussiwand R, et al. Quality of TCR signaling determined by differential affinities of enhancers for the composite BATF-IRF4 transcription factor complex. Nat Immunol 2017;18:563-572. PMID 28346410
- Kondo S, Schutte BC, Richardson RJ, et al. Mutations in IRF6 cause Van der Woude and popliteal pterygium syndromes. Nat Genet 2002;32:285-9. PMID 12219090
- Hambleton S, Salem S, Bustamante J, et al. IRF8 mutations and human dendritic-cell immunodeficiency. N Engl J Med 2011;365:127-38. PMID 21524210
- Shaffer AL, Emre NC, Lamy L, et al. IRF4 addiction in multiple myeloma. Nature 2008;454:226-31. PMID 18568025
- Graham RR, Kozyrev SV, Baechler EC, et al. A common haplotype of interferon regulatory factor 5 (IRF5) regulates splicing and expression and is associated with increased risk of systemic lupus erythematosus. Nat Genet 2006;38:550-5. PMID 16642019
- Miyamoto M, Fujita T, Kimura Y, et al. Regulated expression of a gene encoding a nuclear factor, IRF-1, that specifically binds to IFN-beta gene regulatory elements. Cell 1988;54:903-13. PMID 3409321
40 motifs in this family — JASPAR 9 · HOCOMOCO 17 · CIS-BP 9 · Vierstra 5
| motif | database | consensus | bits |
|---|---|---|---|
| IRF9 (MA0653.1) | JASPAR | AaCGAAACCGAAACT | 24.8 |
| M05531_2.00 | CIS-BP | CaGTTTCgncGAAACCGAAAc | 24.1 |
| IRF3 (MA1418.2) | JASPAR | gaAAhGGAAACCGAAAC | 23.2 |
| IRF4 (MA1419.2) | JASPAR | CGAAACCGAAACt | 23 |
| AC0666:IRF:IRF | Vierstra | CdGTTTCChgGAAACcGAAA | 22.8 |
| IRF8 (MA0652.2) | JASPAR | CGAAACCGAAACT | 22.4 |
| IRF3.H14CORE.1.SM.BSMB | HOCOMOCO | CaGTTTCChGGAAACtGaaa | 21.9 |
| M05527_2.00 | CIS-BP | aGTtTCGntnndncGAtACT | 20.7 |
| IRF2 (MA0051.2) | JASPAR | gGAAAGcGAAAccaaa | 20.6 |
| M03335_2.00 | CIS-BP | CGAAACCGAAACt | 19.3 |
| IRF6.H14CORE.1.S.CSC | HOCOMOCO | CGAaACCGAAACt | 18.8 |
| IRF2.H14CORE.0.PSM.APSMA | HOCOMOCO | CGAAAGCGAAAg | 18.4 |
| IRF8.H14CORE.1.SM.BSMB | HOCOMOCO | tGAAACTGAAACt | 18.2 |
| M08142_2.00 | CIS-BP | AAnnGAAAgTGAAAgt | 18 |
| M05549_2.00 | CIS-BP | CGAAACcGAAAC | 17.7 |
| M03337_2.00 | CIS-BP | CGAAAgCGAAAvT | 17.6 |
| IRF7 (MA0772.2) | JASPAR | CGAAAgCGAAAvT | 17.4 |
| IRF1.H14CORE.0.P.BPB | HOCOMOCO | aAnnGAAAgTGAAAg | 17.4 |
| AC0673:IRF:IRF | Vierstra | AAnCGAAAaTCGnTT | 17.4 |
| IRF9.H14CORE.0.PSM.APSMA | HOCOMOCO | GAAACTGAAACt | 17.3 |
| IRF8.H14CORE.0.P.BPB | HOCOMOCO | AgdGgAAgTGAAAgt | 15.8 |
| IRF4.H14CORE.2.SM.BSMB | HOCOMOCO | AcCGAAACCGaaA | 15.7 |
| IRF5 (MA1420.1) | JASPAR | CCGAAACCGAAACt | 15.3 |
| M03332_2.00 | CIS-BP | CCGAAACCGAAACt | 15.3 |
| IRF5.H14CORE.0.SM.BSMB | HOCOMOCO | GAAACCGAAACt | 15.1 |
| IRF7.H14CORE.1.SM.BSMB | HOCOMOCO | AAnCGAAAvcGAAA | 14.7 |
| IRF3.H14CORE.0.PS.APSA | HOCOMOCO | GAAAvgGAAACbGAAA | 14.6 |
| IRF6 (MA1509.1) | JASPAR | ACCGAAACT | 14.1 |
| IRF7.H14CORE.0.S.BSB | HOCOMOCO | cAAAhCGAAAaT | 14 |
| IRF8.H14CORE.2.SM.BSMB | HOCOMOCO | aAAcCGAAACT | 13.4 |
| AC0672:IRF/STAT:IRF | Vierstra | GAAAccGAAAC | 13.3 |
| Irf1 (MA0050.4) | JASPAR | tGAAACTGAAA | 13.2 |
| AC0670:IRF/ZBED:IRF | Vierstra | GTTTCGntnnnncGAtAC | 13.2 |
| IRF4.H14CORE.0.P.BPB | HOCOMOCO | GgAAcTGAAAc | 12.8 |
| IRF6.H14CORE.0.SM.BSMB | HOCOMOCO | aCCGAAACt | 12.7 |
| IRF5.H14CORE.1.S.BSB | HOCOMOCO | CGAAAC | 11.2 |
| M07978_2.00 | CIS-BP | tghgGAAaTGArtca | 10.7 |
| IRF4.H14CORE.1.S.BSB | HOCOMOCO | ACCGAAAC | 10.5 |
| M05545_2.00 | CIS-BP | GAAAgcGAAA | 10.3 |
| AC0671:IRF:IRF | Vierstra | CCGAAACc | 10 |
HOCOMOCO tags read as assay + quality grade: P ChIP-seq, S HT-SELEX, M methyl-HT-SELEX, G GHT-SELEX, I SMiLE-seq, B PBM; grade A means two or more assay types agreed. A P-only matrix is a record of where the protein was crosslinked, which need not be a sequence it binds itself.
The factors, elsewhere
These entries are written from the primary literature and every citation is checked against PubMed, but this one can still be out of date, too confident, or simply wrong — especially where a motif’s published site and its database matrices disagree. Corrections and additions are welcome, and they are the fastest way to make this page better for everyone.