The new crystal structure of the IRF-3 transactivation domain reveals an evolutionarily conserved protein domain whose activity is modulated by phosphorylation controlling oligomerization and protein-protein interactions.
New protein structures often bring excitement to molecular and cell biologists. This discovery process has widespread appeal because it provides the means to derive evolutionary postulates and make inferences about molecular function based on the understanding of structural organization at atomic resolution. Two papers in this issue of Nature Structural Biology, reported by Takahashi et al.1 and by Qin et al.2, bring about such fresh ideas and deeper understanding of critical signal transduction processes involved in antiviral and antibacterial host responses.
The player at stake is a member of the interferon regulatory factor (IRF) family, IRF-3 (ref. 3). IRF-3 is a multidomain transcription factor, with a conserved N-terminal DNA-binding domain that has novel DNA-recognition motifs4,
5,
6. The two papers discussed here now report the structure of the C-terminal transactivation, oligomerization and regulatory domain of IRF-3. These studies reveal that this domain shares structural similarity with those in other proteins, including the Smads7, signaling effectors of the transforming growth factor- (TGF-) pathway, and forkhead-associated (FHA) domain proteins8, many of which are involved in the DNA damage response. The structures therefore verify the relationship between IRFs and Smads, which was first predicted by Eroshkin and Mushegian9.
Immediate cellular response to viral or bacterial invasion constitutes a critical first step of the complex immune response of multi-cellular organisms. This primary response to pathogenic challenges is called innate immunity, and one of its core characteristics is the activation and secretion of interferon family of cytokines, which subsequently propagate more global immune response to the pathogenic agent10. Of central importance for the induction of interferons in response to viral RNA or bacterial surface lipopolysaccharides is the activation of cellular signaling pathways that involve Toll-like receptors (TLRs) in the plasma membrane, followed by the recently connected cascades of kinase complexes that include (i) TAK-1 (TGF--activated kinase-1) in complex with regulatory subunits and multifunctional adaptors such as TRAF-6 (TNF receptor−associated factor-6), acting upstream of (ii) IKK (inhibitor of NF- (I) kinase ) coupled to TANK-1 (TRAF family member−associated NF-B activator-1) and TBK-1 (TANK-binding kinase-1)11,
12 (Fig. 1). The ultimate targets of this phosphorylation cascade are members of the IRF family, a prominent member of which is IRF-3. Upon phosphorylation, IRF-3 dimerizes and translocates from the cytosol into the nucleus where it binds to the enhancers of various interferon genes in order to induce rapid expression3,
13.
(a) A schematic diagram of IRF-3 signaling. Upon recognition of viral double-stranded RNA or bacterial lipopolysaccharides, Toll-like receptors (TLRs) on host cells activate a kinase complex containing TAK-1, its regulators TAB-1 and TAB-2, the multifunctional adaptor-E3 ligase TRAF-6 and additional components (not shown). Activated TAK-1 leads to phosphorylation and activation of the TANK-1/TBK-1/IKK multi-kinase complex, which eventually phosphorylates monomeric IRF-3 at its C-terminal domain (C). Phosphorylation potentially changes the conformation of both C- and N-terminal (N) domains, leading to dimerization of IRF-3 and rapid translocation to the nucleus where IRF-3 is a component in the enhanceosomes of the interferon genes. (b) A schematic diagram of Smad signaling. Dimeric TGF- is recognized by cell surface receptor serine/threonine kinases (RSTKs) of which the type II receptor phosphorylates and activates the type I receptor kinase. Type I receptor kinase then phosphorylates the C terminus (C) of a receptor-activated Smad (R-Smad) proteins, leading to conformational changes. Phosphorylated Smads hetero-oligomerize with the nonphosphorylated Smad4, translocate to the nucleus and engage in transcriptional complexes together with cofactors that mediate induction or repression of target gene expression. Smad dimers are shown for simplicity. A double arrow indicates potential interaction and crosstalk between IRF-3 and Smad signaling complexes. Curved arrows indicate phosphorylation events. Small straight arrows indicate the flow of signal transduction. The plasma membrane lipid bilayer and the nuclear envelope double bilayer with embedded nuclear pores indicate cellular compartmentalization.
IRF-3 activation Biochemical analyses of the steps of the functional cycle of IRF-3 have suggested successive molecular alterations (reviewed in refs. 3,13). The studies reported in this issue now define the structural elements and provide insights into the activation mechanism of IRF-3 and possibly other IRF family members, such as IRF-7. IRF-7 contributes to the sustained response of interferon gene expression and interferon-stimulated genes; its expression is induced by the action of interferons via their plasma membrane receptors and downstream effectors of the Janus-associated kinase (JAK) and signal transducer and activator of transcription (STAT) families14. Newly synthesized latent IRF-7 in the cytosol may follow a similar cascade of activation events as for IRF-3, which could lead to an overall amplification of the antiviral response.
The structures of the IRF-3 C-terminal domain reported by Takahashi et al.1 and Qin et al.2 are virtually identical and contain a -sandwich core that resembles the FHA domain8. This core is decorated by flanking -helices at both the N- and C-termini. The N-terminal -helix makes direct contacts with the C-terminal helices, creating a compact, largely hydrophobic structural unit. This helical bundle blocks the surface of the core that could potentially mediate protein-protein interactions. Notably, the overall fold of the entire domain bears a striking similarity with the Mad-homology 2 (MH2)7 domain of Smads.
Even though the two groups report very similar structures, they developed different models for IRF-3 activation, based on different interpretations of the crystal structures. The IRF-3 C-terminal domain constructs used by the two groups differ only very slightly, and they both crystallized as a dimer in the asymmetric unit. Takahashi et al.1 assumed that this dimer interface might resemble that in the phosphorylated, activated IRF-3. Additional mutagenesis and biochemical experiments verified that the residues mediating crystal dimer contacts are indeed important for dimer formation in solution. Thus, Takahashi et al.1 concluded that phosphorylation of the IRF-3 C-terminal domain could potentially promote dimer formation, leading to activation.
The IRF-3 C-terminal domain construct of Qin et al.2 is monomeric in solution, leading the authors to conclude that the dimer interface is a result of crystal packing and not likely to be functionally relevant. They formulated an auto-inhibition model for IRF-3, functionally analogous to that of the structurally similar Smads7,
15. In this model, addition of negatively-charged phosphate groups to serine and threonine residues in the helical bundle is predicted to induce a dramatic conformational change that would disrupt the 'auto-inhibited' state. This 'activated' state could then oligomerize to initiate subsequent signaling events. Mutagenesis and functional analyses accompanying the reported structure support their model. Thus, structural data for the phosphorylated (or pseudo-phosphorylated) IRF-3 C-terminal domain will now be necessary to resolve the models developed by these two groups.
Oligomerization 'epitopes' Following phosphorylation, IRF-3 is known to undergo homo- and possibly hetero-oligomerization with other IRF members, similar to the oligomerization that occurs in the activation of Smads16. The oligomer stoichiometry of signaling proteins can be complex. For example, inactive Smads are monomeric, and form either dimers or trimers upon activation16. The two studies shed some light on the oligomerization of IRF-3. Although Takahashi et al.1 and Qin et al.2 do not agree on the structural model of IRF-3 activation, they identified similar features important for mediating protein-protein interactions, namely, a hydrophobic patch and a basic surface in the C-terminal domain of IRF-3. In particular, Qin et al.2 noted that both Smad MH2 and FHA domains contain similar features, which are involved in recognition of phospho-serine or -threonine on their partner proteins. IRF-3 phosphorylation could potentially expose these 'epitopes' by inducing conformational changes in the C-terminal domain. These epitopes can then interact with the phospho-serines or -threonines of another IRF-3 subunit or possibly of another IRF member, such as IRF-7, during the late interferon response. In Smads, phosphorylation induces conformational alterations in the MH2 domain; this leads to recognition of the phospho-serines of another Smad subunit (homo-oligomerization). Alternatively, the MH2 domain of another non-phosphorylated Smad subunit, the product of the tumour suppressor gene Smad4, recognizes the phospho-serines of the Smad MH2 and thus contributes to formation of hetero-oligomers7,
16 (Fig. 1).
The structural similarities between IRF-3 and Smad C-terminal domains raise the question of whether IRFs and Smads can form heterooligomers. Both IRF-3 and Smad proteins are ubiquitously expressed in mammalian cells. Is it possible that, under conditions of viral attack and local TGF- activation, induction of IRF−Smad complexes may mediate unique signaling events in the anti-viral response? Indeed, recent data have identified functional complexes formed between Smads and IRF-7 (R. Derynck, personal communication). This scenario brings about the exciting principle of combinatorial signaling that permeates the modern signal transduction world. The structural work summarized here offers multiple angles on how future studies could address such questions in a rational manner.
Cofactor interactions The IRF-3 -sandwich core also offers interaction surfaces for recruitment of additional transcriptional cofactors, such as the coactivator p300/CBP. Both studies measured the contribution of specific residues of IRF-3, as predicted from the crystal structures, to the transcriptional activation potential of this factor, in the context of the full-length protein. Residues from the basic surface discussed above, as well as from an acidic surface, play critical roles in virus-mediated interferon- enhancer transactivation. The acidic pocket is shown to be directly involved in interacting with p300/CBP. Based on the complex nature of the interferon- enhanceosome (a transcriptional machinery containing IRF-3)17, one predicts that these two IRF-3 structural motifs may have an important role in the interactions and assembly of additional transcriptional cofactors of the complex. Once again, this molecular paradigm finds a faithful replica in the structure-function relationships of the Smad MH2 domain, which interacts not only with p300/CBP, but also with a plethora of transcription factors that participate in various gene-specific promoter-enhancer or silencer complexes that elicit the multifunctional physiology of the TGF- pathways7,
16.
Remaining issues One aspect of this system remains unknown—the structure of the entire IRF-3 molecule. Do the structural changes of the IRF-3 C terminus upon phosphorylation (proposed by the two studies) directly link to the activation of the DNA-binding domain at the N terminus? It seems likely, but of course much depends on the linker segment (amino acids 140-174) connecting the well-folded N- and C-terminal domains. Similarly, do these changes directly link to the mechanism of IRF nuclear translocation, which is also induced by its phosphorylation18—that is, does phosphorylation induce the exposure of nuclear localization signals residing in the N-terminal domain or mask the nuclear export signals of IRF-3 in the short linker? In an identical scenario for Smads, only structures of isolated MH1 or MH2 domains have been obtained, but the less structured linker holds the key to the following questions. How do the linkers of IRFs or Smads transmit information between the separate domains? Does the mechanism involve specific structural adaptation of these flexible regions? Alternatively, can this effect be explained by invoking cofactors that bind to the linkers, the release or recruitment of which mediates in trans the allosteric effect to the other end of the protein? Here, again, Smads may offer useful examples on mechanisms of IRF cytoplasmic retention. For example, SARA (Smad anchor for receptor activation) is an endocytic membrane protein that recognizes monomeric, latent Smads and is thought to serve as a retention/anchoring protein in the cytoplasm7,
16. The structural similarity between IRF-3 and Smads may now facilitate future investigations on IRF cytoplasmic anchoring. All these present open challenges for future structural and biochemical dissection. In summary, the Smads and IRF C-terminal domains represent versatile signaling modules that are activated by phosphorylation, recognize phospho-serine/threonine motifs and become part of large transcription complexes. This central principle identifies the heart of regulation of innate immunity and cellular growth and morphogenesis.
(TGF-
(inhibitor of NF-
-helices at both the N- and C-termini. The N-terminal 
