The Interferon Regulatory Factor Family in Host Defense: Mechanism of Action*

Transcription factors of the interferon regulatory factor (IRF) family commands the entire type I interferon (IFN) system from induction of IFNs to diverse IFN responses, thereby providing a principal basis for host resistance against pathogens. However, the family has various additional roles. Regulating the development of the immune system, IRFs shape the establishment and execution of innate and adaptive immunity. IRFs also regulate growth and differentiation of many cell types, thus playing a role in leukemia and other cancers. In addition, evidence indicates that IRFs confer antiviral mechanisms not directly ascribed to the IFN system. This review deals with the diverse roles of IRFs in host defense and discusses the molecular mechanisms by which they regulate target gene transcription.

Transcription factors of the interferon regulatory factor (IRF) family commands the entire type I interferon (IFN) system from induction of IFNs to diverse IFN responses, thereby providing a principal basis for host resistance against pathogens. However, the family has various additional roles. Regulating the development of the immune system, IRFs shape the establishment and execution of innate and adaptive immunity. IRFs also regulate growth and differentiation of many cell types, thus playing a role in leukemia and other cancers. In addition, evidence indicates that IRFs confer antiviral mechanisms not directly ascribed to the IFN system. This review deals with the diverse roles of IRFs in host defense and discusses the molecular mechanisms by which they regulate target gene transcription.
In the past a few years much advancement has been made on the function of IRFs, 2 signaling pathways that lead to innate immunity, their roles in growth and development, and their structures. Progress in these areas is summarized in the first section of this review. The following sections examine the biochemical modifications and molecular mechanisms of IRF action that are less well understood. Table 1 is a compilation of these features.

Role of the IRF Family in the IFN System
Among nine mammalian IRFs, IRF-3 and IRF-7 are early IRFs activated by Toll-like receptor (TLR) 2 and other types of signaling that play a pivotal role in the initial induction of type I IFNs. Upon activation, IRF-3 and IRF-7 are phosphorylated through the IB kinase (IKK) family of kinases, dimerized, and translocated into the nucleus to stimulate IFN␤ and IFN␣ transcription. Two excellent reviews summarize a body of recent publications regarding signaling processes that result in IRF-3/7 activation (1,2). In addition to IFN induction, IRFs direct the entire type I IFN responsive-ness. Following IFN-receptor interaction, a latent cytoplasmic IRF-9 complexes with Stat1 and Stat2 through the activation of the JAK/STAT pathway, binds to the IFN-stimulated response element (ISRE), and stimulates transcription of a large set of IFNstimulated genes (see Ref. 3 for review). Underscoring the tight link between IFN induction and action, type I IFN genes are themselves inducible by IFNs (2,4). In addition to the above noted IRFs, other family members modulate the extent and modes of IFN induction as well as IFN responses in subsequent steps. For example, IFNs are produced in dendritic cells (DCs) at much higher levels than in other cells. This is in part achieved by the activity of IRF-8 (5, 6).

Developmental Role and Host Defense
Besides regulating the IFN system, the IRF family participates in directing the development of innate and adaptive immunity, a role best illustrated by the activity of IRF-4 and IRF-8. The two factors are expressed only in immune cells. They not only bind the ISRE, the binding site for all IRF factors, but also bind various Ets/IRF composite elements through their interaction with PU.1, an Ets family transcription factor (6 -8). This protein-protein interaction, occurring through the C-terminal IRF association domain (IAD), extends the range of target genes beyond those carrying ISREs. IRF-4 KO mice are impaired in the maturation of T and B cells, defects most strikingly manifest after antigen stimulation (9). IRF-4 orchestrates B cell development by regulating coordinated proliferation and differentiation programs, e.g. IRF-4 is required for the plasma cell differentiation program and controls immunoglobulin isotype switching (10). IRF-4 is constitutively activated in cells of multiple myeloma as a result of a chromosomal translocation and acts as an oncogene (11). IRF-4 is expressed in other lymphocyte malignancies associated with viral infection (12). IRF-8, also expressed during different stages of B cell development, controls expression of B cell-specific genes in a manner distinct from that of IRF-4 (13). More importantly, IRF-8 coordinates myeloid cell growth and differentiation (14). Reflecting this activity, IRF-8 KO mice develop a chronic myelogenous leukemia (CML)-like disease. IRF-8 promotes macrophage differentiation while inhibiting granulocyte development. It directly stimulates the expression of genes active in macrophages including those for lysosomal and endosomal proteases (6). In myeloid cells IRF-8 negatively regulates cell growth as it inhibits c-myc and stimulates expression of tumor suppressor/ inhibitor of cyclin-dependent kinase (INK4). This growth regulation is relevant to CML in humans, where IRF-8 expression is down-regulated by the Bcr/Abl oncoprotein resulting in abnormal growth promotion (see Ref. 14 for review). Regulation of cell growth and macrophage differentiation requires both the DNA-binding domain (DBD) and the IAD. Revealing a critical role of the IAD, a spontaneous point mutation present in the IAD of IRF-8 causes a CML phenotype very similar to that in IRF-8 KO mice (15). In addition, IRF-4 and IRF-8, among other IRFs, control the development of DC subsets as revealed by gene targeting studies (16). Whereas IRF-4 is important for the development of CD4ϩ and double negative DC subsets, IRF-8 plays a predominant role in the development of plasmacytoid DCs and CD8␣ ϩ DC subsets (16). Rescue experiments revealed that IRF-8, but not IRF-4, is required for the expression of proinflammatory cytokines including type I IFN and IL-12p40, indicating that IRF-8 is critically required for cytokine gene expression (16,17). IRF-4, on the other hand, inhibits expression of proinflammatory cytokine genes (18).
The idea that some IRFs play a role in controlling development is reinforced by recent gene targeting studies of murine IRF-6, which is expressed at high levels in skin (19,20). Results show that IRF-6 is required for the generation of skin and limb as well as craniofacial development. Disruption/mutation of the IRF-6 gene strongly affects keratinocyte proliferation and differentiation, causing deregulation of genes important for epidermis differentiation. Consistent with gene targeting results, mutations in the IRF-6 gene are associated with the Van der Woude syndrome, a congenital abnormality in humans. Thus, as observed for IRF-4 and IRF-8, IRF-6 regulation of skin cell differentiation is coupled with cell growth regulation. Thus far, however, a role for IRF-6 in immunity has not been reported. In line with the pivotal role of IRFs in cell growth control, the activity of IRF-1 and IRF-2 is also linked to cell growth (21).

Viral Regulation of IRFs
As part of anti-IFN mechanisms, some viruses have evolved viral proteins that interact with IRF-3 and IRF-7 to inhibit IFN induction. Among many examples reported, the rotavirus regulatory protein NSP1 binds to and degrades IRF-3 to disable IFN induction (22) Table  S1). Several lines of evidence indicate that other viruses utilize IRFs to ensure their replication, impeding host defense mechanisms not directly related to the IFN system. For example, HIV-1 carries an ISRE-like element at position ϩ200 to ϩ217.

(See other examples in supplemental
HIV-1 infection has been shown to induce expression of IRF-1, which then stimulates HIV-1 transcription through this element both in the presence and absence of Tat (23). Because ectopic IRF-8 expression in T cells represses HIV-1 replication, it has been proposed that IRF-1 and IRF-8 play a role in the replicative and latent phases of viral infection, respectively (23). HTLV-1 is a causal agent for adult T cell leukemia (ATL). IRF-4 expression is up-regulated in HTLV-1-infected cells and in ATL-derived lymphoid cells, and gene expression profile analysis indicates that increased IRF-4 expression is associated with growth promotion and the leukemia phenotype (12). Similarly, increased IRF-4 expression in lymphoma cells with Kaposi's sarcoma-associated herpesvirus infection correlates with the down-regulation of B cell-specific gene expression, suggesting that the viral infection hinders B cell immune responses and promotes tumor progression (24).

Promyelocytic Leukemia Protein (PML)-mediated Antiviral Activity and Role for IRFs
PML is a major constituent of nuclear bodies, a spherical nuclear substructure with many reported functions (see Ref. 25 for review). PML and Sp100, another nuclear body component, are induced after IFN stimulation leading to an increase in the number and size of nuclear bodies. Studies on some DNA viruses indicate that PML/nuclear bodies contribute to elicitation of antiviral states in the cells. Infection of herpes simplex virus-1 (HSV-1) disrupts nuclear body structure through the viral protein ICP0. Cells with reduced PML expression show increased HSV-1 viral gene expression and viral yield (26). ICP0 interacts with PML and affects SUMO (small ubiquitin-related modifier) modification of PML, thereby counteracting PML inhibition of viral replication. Similarly, human cytomegalovirus causes disruption of nuclear bodies. Reduced PML expres-

TABLE 1 Structural and functional comparison of IRF family members
The size (amino acid lengths), chromosomal localization (in italics), and percent identity in the DBD and IAD are for human IRF members. Percent identity of the DBD and IAD is compared with IRF-1 and IRF-3, respectively.
sion is, likewise, associated with increased expression of human cytomegalovirus proteins and viral replication (27). Additionally, infection with rabies virus, an RNA virus, also alters PML distribution through the interaction of the viral P protein with PML (28). In keeping with the protective role suggested for PML, rabies virus infection in PML Ϫ/Ϫ cells results in a much greater viral yield than PML ϩ/ϩ cells. Given that PML and Sp100 are induced by IFN through ISREs, involvement of IRFs in nuclear body activity has been expected. Indeed, recent studies show that IRF-8 is required for the expression of PML and Sp100 in macrophages and DCs, where nuclear bodies are larger and more intense than in other cell types (6,29). In light of the predominant role of macrophages and DCs in the control of viral infection, IRF-8 and other IRFs may play a role in PMLmediated host antiviral responses.

Structure and Mechanism of IRF Action
The crystal structure of the IRF DNA-binding domain shows that it is composed of three ␣-helices, a four-stranded ␤-sheet, and large loops to bind to an ISRE with the consensus motif GAAA (30), presumably as a homodimer as shown for IRF-3 (31). On the other hand, IRF-4 and IRF-8 bind to the Ets/IRF composite site as a heterodimer with PU.1 (7). Structural analyses of the C-terminal regulatory region demonstrate that the IAD consists of the ␤-sandwich core flanked by helices and loops (32,33). This IAD fold resembles the MH2 domain fold of the Smad protein family. IAD crystal structures also reveal an autoinhibitory mechanism for IRF-3 and IRF-7, in that dimerization and DNA binding are prevented by an interaction of the H1 and H5 helices prior to stimulation. Phosphorylation disrupts this autoinhibition leading to the liberation of the DNA binding and dimerization activities. Interestingly, the overall IAD fold is conserved in all IRFs from IRF-3 to IRF-9 (not IRF-1/2), although autoinhibition is predicted only for IRF-3 and -7 by this analysis (33).

Post-translational Modifications
IRF activities are regulated through a variety of post-translational modifications. Phosphorylation is a major mechanism governing type I IFN induction and action. TLR-and retinoic acid inducible gene-I (RIG-I)-mediated phosphorylation of IRF-3 and IRF-7 by IKKi/⑀ and Tank-binding kinase-1 (TBK-1) is a prerequisite of IFN induction (1,2). IFN responses depend on kinases of the JAK and MAPK (mitogen-activated protein kinase) families that phosphorylate Stat proteins (see Ref. 34 for review). However, phosphorylation of IRFs by these kinases in IFN signaling has not been fully explored. IRFs are likely to be modulated by other kinases. For example, IRF-8 is phosphorylated at the tyrosine residues in the DBD, which is inhibited by the protein tyrosine phosphatases SHP1 and SHP2 (35). Tyrosine dephosphorylation coincides with the inhibition of IRF-8dependent transcription. It is suggested that repressed IRF-8 function in some forms of leukemia, including CML, is in part due to elevated SHP1/2 activity (35).
IRFs, like other transcription factors, are acetylated in vivo and in vitro, causing an alteration of DNA binding and transcriptional activities. IRF-1, -2, and -7 are acetylated by histone acetylases GCN/PCAF and CBP/p300 (36,37). IRF-2 and IRF-7 are acetylated at the lysine residue in the DBD that flanks tryptophan. Both residues are conserved throughout the IRF family and are critical for ISRE binding. However, not all IRF members are acetylated under similar conditions, suggesting selectivity of IRF acetylation. IRF-2 is acetylated in a cell cycle-dependent manner, correlating with its ability to regulate cell cycle-dependent H4 transcription (36). Acetylation of IRF-7 results in reduced DNA binding activity (37). This may be part of the post-activation events to promptly lower IFN gene transcription after activation.
IRF proteins are subject to ubiquitination and proteasomedependent degradation. As noted above, some viral proteins degrade IRF-3 and -7 to weaken IFN induction. However, destabilization appears to be part of a normal process for activated IRF-3. Saitoh et al. (38) showed that the peptidylprolyl isomerase Pin1 interacts with IRF-3 and changes the isomerization status in a phosphorylation-dependent manner, leading to rapid destabilization of IRF-3. In accordance, reduced Pin1 expression or gene disruption increases IFN-␤ induction. In some cases ubiquitination of transcription factors enhances transcription, rather than causing degradation (see Ref. 39 for review). Ubiquitination that enhances transcription appears to involve mono-, multi-, or polyubiquitination through lysine 63 rather than lysine 48 of ubiquitin, the latter being a common mark for degradation. Kawai et al. (40) show that IRF-7 is ubiquitinated upon TLR stimulation through an E3 ubiquitin ligase, TRAF6 (40), leading to increased IFN gene transcription. In a separate study, TRAF6 was shown to associate with IRF-8 in TLR-stimulated macrophages, consistent with the involvement of IRF-8 in ubiquitination-coupled transcription (41). Given that there are numerous E3 ligases, some of which are induced in response to IFN, other IRF members may also be subject to ubiquitination-dependent changes in activity.
Some IRFs have extensive variations in length, the extent of which is regulated by viral infection in some cases. IRF-2 is shown to be truncated following viral stimulation, resulting in the removal of much of the C-terminal regulatory region (42). IRF-3 is differentially spliced upon papillomavirus infection, leading to the truncation of the DBD (43). In addition, IRF-5 is reported to have as many as nine distinct isoforms derived from differential promoter usage and splicing (44). These isoforms are differentially expressed in monocytes and lymphocytes and are thought to assume different functions. IRF-7 in human leukocytes also exhibits several splice variants (45). Given cell type specificity and signal-regulated changes, these variants are likely to modulate the functional outcomes of IRFs, although details of their activities are not available at present.

IRFs on Chromatinized Targets and Transcription
IRFs interact and cooperate with other transcription factors to express target genes. IRF-interacting transcription factors include proteins of the NFB, NFAT, Ets, and Stat families, which are also broadly involved in immune responses (Fig. 1). The cooperative activity of IRFs with these factors is reinforced by the fact that promoters that possess ISREs also tend to carry regulatory elements for the interacting partners. A major question, the answer to which has remained elusive, is how IRFs assemble additional regulatory factors, such as co-activators/ repressors, a number of histone modifiers, and chromatin remodeling factors, to finally activate RNA polymerase II. Although our understanding of this subject is not extensive at present, IRFs are likely to interact directly with histone modifiers and control the chromatin environment of target genes, as demonstrated for the IFN␤ gene (46). By so doing, IRFs may affect not only an early transcriptional initiation event but also influence subsequent steps of transcription such as elongation and termination. In this context several IRFs are shown to interact with histone acetylases (47) as well as with components of the Brg-1-or Brm-associated factor (BAF) complex that remodel chromatin structure in an energy-dependent manner (48,49).
An intriguing aspect of IFN-related transcription is that unlike many other genes, histone deacetylase activity is required for active transcription of IFN and IFN-inducible genes. Thus, histone deacetylase inhibitors such as trichostatin potently inhibit induction of a large number of IFN-stimulated genes, despite the fact that transcription of many other genes is stimulated by these inhibitors (50,51). These agents are believed to inhibit an event that follows binding of IRF-3 to DNA and to cause complete disruption of transcriptional preinitiation complex assembly, because RNA polymerase II is no longer recruited to the promoters in the presence of trichostatin. Given that activation of IFN and IFN-inducible genes coincides with an increase in histone acetylation (46), it is thought that histone deacetylase activity is required not because it causes global chromatin deacetylation in and around the promoter but because it acts on a hitherto uncharacterized event essential for transcription activation. These reports suggest a somewhat unusual mechanism by which IRF proteins may control IFN-mediated transcription.
Another significant aspect of IRF-mediated gene expression is that transcriptional activation is, for the most part, short-lived. Following activation, transcription is often quickly reversed, a process reminiscent of other proinflammatory cytokine genes. This may be because of rapid IRF degradation occurring in the post-activation period as noted for IRF-3 and -7, facilitating timely cessation of transcription (38). In addition, it is possible that this process is coupled with the recruitment of corepressor complexes that reverts the chromatin environment to preactivation states (52). It is also possible that IRFs known to have negative transcriptional activity, such as IRF-2 and IRF-4, may be involved in the process of limiting the length of activation.

Dynamic Movement of IRFs in Living Cells
Real time analysis of protein movement has revealed that transcription factors are constantly moving within the nucleus, binding to chromatinized DNA in a very transient manner (see Ref. 53 for review). A similar rapid movement has been noted for many chromatin-binding proteins, although nucleosomal core histones are largely immobile. The dynamic movement of nuclear proteins indicates that interactions between transcription factors are also highly dynamic and occur only transiently.
Our photobleaching experiments showed that IRF-8 and other factors move quite fast in living macrophages. A large majority of IRF-8 bound to chromatin for only Ͻ0.1s at a time, although a small fraction of IRF-8 showed more stable binding (54). Likewise, Bosisio et al. (55) showed that NFB complexes move rapidly in the living nucleus, and their binding is turned over in Ͻ30 s even at specific NFB binding sites. Thus, it appears that transcription factors are in dynamic equilibrium in living cells, binding to chromatin targets for a very short period of time and moving away soon thereafter. Similarly, transcription factors are assembled into protein complexes for a short time, which then rapidly dissolves. These dynamic events indicate that transcriptional processes are constantly influx, adjusting to shifting needs in the cells (Fig. 1).

Conclusions and Future Perspectives
IRF family proteins play diverse roles in innate and adaptive immunity. They regulate the development of immune cells and control the entire IFN system, thereby providing essential mechanisms for host defense against pathogens. Growing evidence indicates that IRFs undergo various forms of post-translational modification to activate and attenuate target gene transcription. IRFs interact with specific and general transcription factors and recruit chromatin modifiers, thereby affecting the extant chromatin environment and driving transcription. Recent studies of in vivo IRF movement have begun to show that IRFs interact with nuclear factors and bind to chromatin targets in a surprisingly transient manner, acting in rapidly shifting equilibria. Future studies will elucidate the dynamic, genome-wide behavior of individual IRFs before, during, and after transcription.