HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Free via Open Access: OA OA Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via CrossRef Citing Articles via Google Scholar Google Scholar Articles by Randall, R. E. Articles by Goodbourn, S. Search for Related Content PubMed PubMed Citation Articles by Randall, R. E. Articles by Goodbourn, S. Agricola Articles by Randall, R. E. Articles by Goodbourn, S.

Interferons and viruses: an interplay between induction, signalling, antiviral responses and virus countermeasures

¹ School of Biology, University of St Andrews, The North Haugh, St Andrews KY16 9ST, UK
² Division of Basic Medical Sciences, St George's, University of London, London SW17 0RE, UK

Correspondence
Richard E. Randall
rer{at}st-andrews.ac.uk
Stephen Goodbourn
s.goodbourn{at}sgul.ac.uk

	ABSTRACT

TOP ABSTRACT Biology of the interferon... Viral evasion strategies Vaccines, antiviral drugs and... Concluding remarks REFERENCES

 
The interferon (IFN) system is an extremely powerful antiviral response that is capable of controlling most, if not all, virus infections in the absence of adaptive immunity. However, viruses can still replicate and cause disease in vivo, because they have some strategy for at least partially circumventing the IFN response. We reviewed this topic in 2000 [Goodbourn, S., Didcock, L. & Randall, R. E. (2000). J Gen Virol 81, 2341–2364] but, since then, a great deal has been discovered about the molecular mechanisms of the IFN response and how different viruses circumvent it. This information is of fundamental interest, but may also have practical application in the design and manufacture of attenuated virus vaccines and the development of novel antiviral drugs. In the first part of this review, we describe how viruses activate the IFN system, how IFNs induce transcription of their target genes and the mechanism of action of IFN-induced proteins with antiviral action. In the second part, we describe how viruses circumvent the IFN response. Here, we reflect upon possible consequences for both the virus and host of the different strategies that viruses have evolved and discuss whether certain viruses have exploited the IFN response to modulate their life cycle (e.g. to establish and maintain persistent/latent infections), whether perturbation of the IFN response by persistent infections can lead to chronic disease, and the importance of the IFN system as a species barrier to virus infections. Lastly, we briefly describe applied aspects that arise from an increase in our knowledge in this area, including vaccine design and manufacture, the development of novel antiviral drugs and the use of IFN-sensitive oncolytic viruses in the treatment of cancer.

Published online ahead of print on 8 November 2007 as DOI 10.1099/vir.0.83391-0.

	Biology of the interferon system

TOP ABSTRACT Biology of the interferon... Viral evasion strategies Vaccines, antiviral drugs and... Concluding remarks REFERENCES

 
The interferons (IFNs) are a group of secreted cytokines that elicit distinct antiviral effects. They are grouped into three classes called type I, II and III IFNs, according to their amino acid sequence. Type I IFNs (discovered in 1957; Isaacs & Lindenmann, 1957) comprise a large group of molecules; mammals have multiple distinct IFN- genes (13 in man), one to three IFN-β genes (one in man) and other genes, such as IFN-, -, -, - and -. The IFN- and -β genes are induced directly in response to viral infection, whereas IFN-, -, - and - play less well-defined roles, such as regulators of maternal recognition in pregnancy. Thus, rather than use the term ‘type I IFN’, we will use IFN-/β when referring to the virally induced cytokines. Although the multigenic nature of IFN- has been known for over 20 years, the significance of this is still debated – i.e. whether these genes are expressed differentially in distinct cell types, whether they are inducible by different types of viruses or whether they are functionally specialized (Brideau-Andersen et al., 2007). For the rest of this review, we will not distinguish between IFN- subtypes. Type III IFNs have been described more recently and comprise IFN-1, -2 and -3, also referred to as IL-29, IL-28A and IL-28B, respectively (reviewed by Ank et al., 2006; Uze & Monneron, 2007). These cytokines are also induced in direct response to viral infection and appear to use the same pathway as the IFN-/β genes to sense viral infection (Onoguchi et al., 2007). Type II IFN has a single member, also called IFN- or ‘immune IFN’, and is secreted by mitogenically activated T cells and natural killer (NK) cells, rather than in direct response to viral infection, and will not be considered in depth in this review.

IFN-/β acts through a common heterodimeric receptor (see below), which appears to be expressed ubiquitously, to activate a signal-transduction pathway that triggers the transcription of a diverse set of genes that, in total, establish an antiviral response in target cells. These genes are referred to as IFN-inducible genes or IFN-stimulated genes (ISGs). A subset of ISGs can also be induced directly (i.e. in an IFN-independent manner) by viral infection, perhaps offering a degree of protection in the primary infected cells, although the dramatic viral sensitivity of IFN-/β receptor-knockout mice suggests that this is much less effective than the IFN response itself. In addition to the cell-autonomous activities of IFN-/β, these cytokines modulate the immune system by activating effector-cell function and promoting the development of the acquired immune response. The IFN-/β system is summarized in Fig. 1.

View larger version (62K): [in this window] [in a new window]   Fig. 1. Overview of the IFN-/β system. Cells that secrete IFN-/β have pattern-recognition receptors (PRRs) to detect molecules associated with infection. These molecules include viral nucleic acids such as dsRNA. These PRRs, once stimulated by their appropriate ligands, activate intracellular signalling cascades leading to transcription of IFN-/β genes. Once secreted, IFN-/β binds to the IFN-/β receptor on neighbouring uninfected cells (as well as on the initial infected cell) and activates an intracellular signalling cascade leading to upregulation of several hundred IFN-/β-responsive genes, many of which have direct or indirect antiviral action. Viruses released from the primary infected cell replicate inefficiently in cells that are in the antiviral state. The image shows a monolayer of cells infected at 0.01 p.f.u. per cell with PIV5 and, 24 h later, the cells were stained with antibody to the viral nucleocapsid protein (virus antigen) and DAPI (4,6-diamidino-2-phenylindole) to stain the nuclei.

Induction of IFN by viruses
In recent years, our understanding of the nature of the signalling processes in response to viral infection has evolved significantly. Studies on synthetic inducers of IFN throughout the late 1960s and 1970s demonstrated that double-stranded RNAs (dsRNAs), especially poly(rI).poly(rC), were extremely efficient inducers (reviewed by De Clercq, 2006). Seminal studies by Marcus and colleagues in the 1970s and 1980s generated a paradigm in which both RNA and DNA viruses induced IFN through the production of viral dsRNA (Marcus & Sekellick, 1977; Marcus, 1983). Thus, negative-stranded RNA viruses were proposed to generate a dsRNA molecule dependent upon transcription, positive-stranded RNA viruses to generate a dsRNA molecule via replication, and even DNA viruses were proposed to generate dsRNA as a result of convergent transcription.

However, whilst dsRNA, either viral or synthetic, is an efficient inducer of IFN-/β, it is not the only inducer. There are several ways in which a cell can recognize the presence of an invading micro-organism, and multiple distinct routes by which hosts can recognize viruses and signal the induction of IFN through the recognition of pathogen-associated molecular patterns (PAMPs) by pattern-recognition receptors. The importance of any individual route of induction depends upon the specific virus, the cell being infected or the stage of infection, and requires a lot more investigation. The known routes of activation are discussed in detail below.

Before considering the specifics of each route of induction, it is worth reviewing our knowledge of the downstream components, namely the transcription factors, as these may be common to many of the induction pathways. The best-studied model for IFN-/β induction is the production of IFN-β in fibroblastoid cells in response to either the synthetic dsRNA poly(rI).poly(rC) or to simple RNA viruses, such as Sendai virus (SeV; see Table 1 for a list of virus abbreviations) or Newcastle disease virus (NDV). In this case, the induction is at the level of transcription and requires no new cellular protein synthesis, indicating that cells contain all of the required signalling components and transcription factors. In its simplest form, induction of IFN-β requires the activation of nuclear factor kappa B (NF-B) and IFN regulatory factor-3 (IRF-3) although, as discussed below, certain other members of the ten-member IRF family (reviewed by Honda & Taniguchi, 2006; Paun & Pitha, 2007) may play a role under some circumstances.

Optimal induction of the IFN-β gene also requires binding of a c-jun/ATF-2 heterodimer to the promoter. The IRF-3, NF-B and c-jun/ATF-2 complexes assemble on the promoter in a co-operative manner to form the so-called enhanceasome, and formation of this complex is aided by the high-mobility group (HMG) chromatin-associated protein HMGI(Y) (also known as HMGA) (reviewed by Merika & Thanos, 2001). The assembled enhanceasome components aid the recruitment of CREB-binding protein (CBP)/p300 that, in turn, promote the assembly of the basal transcriptional machinery and RNA polymerase II. Although the proposed model for the IFN-β enhanceasome features HMGI(Y) as a structural component, NF-B, IRF-3, and c-jun/ATF-2 can each form a stable structure with the promoter without HMGI(Y) (Berkowitz et al., 2002; Panne et al., 2004). Indeed, although HMGI(Y) could aid crystal formation, it could not be positioned within the resolved structure of NF-B bound to the IFN-β promoter. A revision of the enhanceasome model thus suggests that any role for HMGI(Y) is limited to enhancing complex assembly through modulation of the DNA structure.

A feature of the enhanceasome model is that each of the transcription factors binds the IFN-β promoter with limited affinity and that cooperativity between factors is required for optimal induction. However, promoter-mapping studies have shown that the individual binding sites from the IFN-β promoter can respond independently to inducers and that a subset of binding sites will confer some degree of response to inducers. The consensus view is that binding of IRF-3 and/or IRF-7 (see below) is indispensable for induction, but that activation of both NF-B and c-jun/ATF-2 may not be essential; indeed, IFN induction has been reported under conditions where NF-B or c-jun/ATF-2 are not activated or their binding sites are not required (Goodbourn et al., 1985; Ellis & Goodbourn, 1994; King & Goodbourn, 1994; Peters et al., 2002; Poole et al., 2002). Presumably, the relative contributions that the individual transcription factors make to IFN-β induction will vary depending on cell type and inducer, and there might be alternative enhanceasome structures.

An analysis of IFN-β induction is complicated by the fact that pre-treatment of cells with IFN can sometimes enhance the IFN yield (an effect called priming) and the distinct possibility of positive autoregulation, where IFN produced during the induction cycle can enhance ongoing production by inducing the synthesis of alternative signalling or transcriptional components. The production of IFN during viral infection leads to the induction of at least three transcription factors (IRF-1, IRF-7 and IRF-9) that, under some circumstances, play a role in the induction process (Matsuyama et al., 1993; Kawakami et al., 1995; Sato et al., 1998, 2000; Honda et al., 2005b). The precise role of each of these factors is unclear. IRF-7 expression levels are undetectable or very low in tissues other than plasmacytoid dendritic cells (pDCs – see below) or cells exposed to IFN. Thus, positive-feedback models have been proposed in which direct induction of the ‘primary’ IFN genes (IFN-β and murine IFN-4) (Erlandsson et al., 1998; Marie et al., 1998) takes place, utilizing IRF-3 as the IRF family member; the IFN thus produced feeds back onto cells and induces the synthesis of IRF-7, which, in the presence of a continued infection, then enhances the transcription of the primary IFN genes and allows transcription of the ‘secondary’ IFN genes (the remaining IFN- genes) (Marie et al., 1998; Sato et al., 1998; Prakash et al., 2005). In support of this model, IRF-7 can bind to the IFN-β promoter (Wathelet et al., 1998; Panne et al., 2007b) and can enhance transcription dramatically (Yang et al., 2004). However, in addition to this feedback role, it remains possible that, even at very low levels, IRF-7 may function as a primary transcription factor in IFN-β induction in fibroblasts. Mouse embryonic fibroblasts (MEFs) from both IRF-3^0/0 and IRF-7^0/0 mice show impaired induction of IFN-β (Sato et al., 2000; Honda et al., 2005b), indicating that, although neither factor is essential, both appear important in induction in these cells, although it remains uncertain whether the IRF-7 effect is direct or indirect.

Induction of the IFN- genes is less well understood (reviewed by Civas et al., 2002). In fibroblastoid cells, these genes do not display a primary induction profile, with the exception of murine IFN-4. Unlike the IFN-β promoter, the IFN- genes lack NF-B sites in their promoters, but contain several binding sites for members of the IRF family, including some sites that fit poorly to the consensus. The identity of the IRF family member that stimulates IFN- gene transcription is uncertain, but it is notable that IRF-7 has a more relaxed DNA-binding specificity for IRF sites than other members of the IRF family (Morin et al., 2002), and there is considerable evidence that IRF-7 stimulates IFN- gene transcription preferentially (Au et al., 1998; Lin et al., 2000), probably in association with IRF-3 (Morin et al., 2002). The apparent requirement for IRF-7 is consistent with the lack of primary IFN- gene induction in fibroblasts, where induction would be dependent upon the feedback induction of the IRF-7 gene (Yeow et al., 2000).

Whilst the general applicability of this model remains to be seen, there are clearly many situations where different profiles are observed. In lymphocytes, viral infection leads to the production of IFN- without the need to pre-produce IFN-β, indicating that these cells are somehow primed constitutively (Hata et al., 2001). This may correlate with the pre-existence of IRF-7 in these cells, although the evidence for this is equivocal. However, in the case of pDCs, there is constitutive IRF-7 expression, and these cells respond to IFN inducers by making massive amounts of IFN-/β (hence their alternative name of IFN-producing cells or IPCs). Recent work has also suggested a role for IRF-1 in the induction of IFN-β in myeloid-derived dendritic cells (mDCs) in response to the Toll-like receptor 9 (TLR9) ligand CpG (Negishi et al., 2006; Schmitz et al., 2007).

We will next consider the known individual routes of activation in more detail.

(i) Extracellular dsRNA or dsRNA delivered through endosomes.
It has been known for over 30 years that many cell lines and cell types respond to the addition of synthetic dsRNA to the culture medium (referred to here as extracellular dsRNA). Given the size of this molecule, it seemed likely that cells either have a surface-bound dsRNA receptor or contain such a receptor in endosomes, which is activated by internalizing dsRNA by endocytosis. As the kinetics of response to extracellular dsRNA can vary widely between cells, it is probable that a dsRNA receptor can be localized in both compartments. The nature of this receptor remained elusive until 2001, when TLR3 was identified as a molecule that permitted signal responses to added dsRNA (Alexopoulou et al., 2001). TLR3 shows a relatively wide tissue distribution, but is expressed at a high level in mDCs, especially the CD4^–CD8⁺ subset. TLR3 relocalizes from the endoplasmic reticulum to endosomes in mDCs (Johnsen et al., 2006), to lysosomes in bone marrow-derived macrophages (de Bouteiller et al., 2005; Lee et al., 2006) and to both endosomes and the cell surface in fibroblasts (Matsumoto et al., 2002). The location of TLR3 (and, indeed of TLR7 and TLR9 – see below) enables the detection of viral nucleic acid present in the extracellular environment or produced by uncoating or degradation of entering viral particles; these features permit the development of an IFN response without the need for viral replication. Furthermore, under some circumstances, pDCs can respond through TLR3 to the dsRNA presented in the phagocytosed apoptotic cells derived from viral infection (Schulz et al., 2005).

The importance of TLR3 in antiviral defence has been probed by using knockout mice; TLR3-deficient mice succumb to infection by murine cytomegalovirus (MCMV) due to reduced IFN production (Tabeta et al., 2004), but remain resistant to some other viruses, such as lymphocytic choriomeningitis virus (LCMV), vesicular stomatatis virus (VSV) and reovirus (Edelmann et al., 2004). Engagement of TLR3 by dsRNA triggers a complex signal-transduction pathway (summarized in Fig. 2), starting with the dimerization of TLR3 and its tyrosine phosphorylation (Sarkar et al., 2004) and recruitment of an adaptor called Toll–interleukin (IL)-1-resistance (TIR) domain-containing adaptor inducing IFN-β (TRIF) (Hoebe et al., 2003; Yamamoto et al., 2003a), as well as phosphatidylinositol 3 (PI3) kinase (Sarkar et al., 2004). Engagement of TRIF signals the activation of both the NF-B and the IRF-3 ‘arms’ of the IFN-induction pathway (Jiang et al., 2004). In the NF-B ‘arm’, TRIF activation leads to the recruitment of tumour necrosis factor (TNF) receptor-associated factor 6 (TRAF6) (Sato et al., 2003; Jiang et al., 2004) and receptor-interacting protein 1 (RIP1) (Meylan et al., 2004; Cusson-Hermance et al., 2005). TRAF6 recruitment and oligomerization activate its lysine 63-linked ubiquitin E3 ligase activity, leading to polyubiquitination of itself and RIP1 (reviewed by Chen, 2005). The polyubiquitin chains are recognized by TAK1-binding proteins 2 and 3 (TAB2 and TAB3) (Kanayama et al., 2004), which chaperone transforming growth factor β-activated kinase 1 (TAK1) to the complex (Deng et al., 2000; Wang et al., 2001). Polyubiquitinated RIP1 is recognized by the NF-B essential modifier (NEMO) [also known as the subunit of the IB kinase (IKK) complex; Ea et al., 2006; Li et al., 2006a; Wu et al., 2006] and thus the IKK complex is recruited to the TRIF–RIP1–TRAF6–TAB–TAK1 complex; as a result of this juxtaposition, the IKKβ subunit of the IKK complex is phosphorylated directly by TAK1 (Wang et al., 2001), leading to the downstream phosphorylation of IB, its subsequent ubiquitination and degradation and the eventual nuclear uptake of NF-B. To emphasize the importance of the ubiquitination profile of this complex in innate immune responses, the de-ubiquitination enzyme A20 is required for the termination of TLR-dependent NF-B activation, in a manner that requires the removal of ubiquitin from TRAF6 (Boone et al., 2004).

View larger version (31K): [in this window] [in a new window]   Fig. 2. TLR3-dependent signalling in response to dsRNA. dsRNA, presented to the outside of the cell or presented to endosomes by endocytosis of extracellular dsRNA, uncoating of endocytosed viral particles or degradation of engulfed apoptotic cells, binds to TLR3. Activated TLR3 recruits the adaptor TRIF that, in turn, acts as a scaffold to recruit signalling components that feed into either the IRF-3 or the NF-B pathways. NF-B activation requires TRAF6 and RIP1 recruitment to TRIF and their co-operation in recruiting the IKK complex and TAK1. TAK1 phosphorylates the IKKβ subunit of the IKK complex, leading to its activation and phosphorylation of IB. Phosphorylated IB is ubiquitinated and subsequently degraded by proteasomes, releasing NF-B for migration to the nucleus (green arrow) and assembly on the IFN-β promoter. IRF-3 activation requires recruitment of TRAF3 to TRIF. TRAF3 binds to TANK, which then binds to TBK-1 and/or IKK, which are activated in an uncharacterized manner and can phosphorylate IRF-3 directly. The related proteins NAP1 and SINTBAD may function in a non-redundant manner at the same level as TANK (indicated as TANK etc.). IRF-7, where present due to the feedback action of IFN, is activated by TBK-1 and IKK in a similar manner (NB this is distinct from the TLR7- and TLR9-dependent pathway described in Fig. 3). The activated IRFs also migrate to the nucleus (green arrows) and assemble on the IFN-β promoter with NF-B and ATF-2/c-jun, leading to the recruitment of co-factors such as CBP/p300 and RNA polymerase II and, ultimately, stimulation of transcription. See text for more details and references.

(ii) ssRNA delivered through endosomes.
During some viral infections, pDCs can make up to half of the circulating IFN (reviewed by Cao & Liu, 2007). One key feature of pDCs is their TLR-expression profile. They are one of the few cell types that express TLR7, and this is expressed exclusively in endosomes. The ligands for TLR7 include immunomodulatory compounds such as imiquimod and R848 (Hemmi et al., 2002; Jurk et al., 2002; Heil et al., 2003) and single-stranded RNA (ssRNA) molecules, either synthetic (Diebold et al., 2004) or derived from viruses such as human immunodeficiency virus (HIV), influenza and VSV (Diebold et al., 2004; Heil et al., 2004; Lund et al., 2004). TLR7 shows no sequence specificity for the ssRNA, requiring only the presence of several uridines in close proximity (Diebold et al., 2006). Interestingly, RNA modifications such as those found in mammalian tRNA and rRNA suppress the TLR7-agonist properties, suggesting that these modifications may have evolved to enable discrimination between host and pathogen RNA (Kariko et al., 2005).

Thus, pDCs can mount robust IFN responses when exposed to ssRNA viruses. Uniquely, these cells are dependent on TLRs for IFN induction by NDV, with the RNA helicases melanoma differentiation-associated gene-5 (mda-5) and retinoic acid-inducible gene-I (RIG-I) (discussed below) playing no part (Kato et al., 2005). There is a functional diversity within RNA viruses, in that some (such as influenza A virus; Diebold et al., 2004) can activate TLR7 signalling without the need for replication, and presumably do so as a consequence of uncoating and RNA release within the endosome, whilst others (such as VSV and SeV) require replication to activate TLR7 signalling. This observation was puzzling, because the replication would be cytoplasmic and therefore in a different subcellular compartment from TLR7. A recent observation has offered a resolution of this issue by showing that cytoplasmic material (including replicated viruses) can be engulfed by autophagy and the resultant vesicles can fuse back to endosomes to present nucleic acids to TLR7 (Iwasaki, 2007; Lee et al., 2007). It remains to be seen how this system prevents the sensing and response to cellular RNAs.

The mechanism of induction of IFN follows a different profile from dsRNA-induced activation of TLR3, in that ssRNA-activated TLR7 recruits a distinct adaptor called myeloid differentiation factor 88 (MyD88) that, in turn, recruits a complex containing the kinases interleukin-1 receptor-associated kinase 4 (IRAK-4), IRAK-1 and TRAF6 (Fig. 3); there is also an involvement of TRAF3 in this process (Hacker et al., 2006). TRAF6 recruitment can activate NF-B through TAK1–TAB2–TAB3 and the canonical IKK complex, and this route is followed for NF-B activation in response to TLR7 engagement. However, in contrast to the activation of IRF-3 seen in response to TLR3 engagement (discussed above), the unique presence of significant constitutive levels of IRF-7 in pDCs offers an alternative route for induction of IFN. In these cells, the MyD88–IRAK-1–IRAK-4–TRAF6 complex binds directly to IRF-7 (Honda et al., 2004; Kawai et al., 2004; Uematsu et al., 2005) and TRAF6 uses its ubiquitin E3 ligase function to polyubiquitinate IRF-7 (Kawai et al., 2004). This process also requires RIP1, and IRF-7 interacts preferentially with polyubiquitinated RIP1 (Huye et al., 2007). Recruited IRF-7 is phosphorylated by IRAK-1 in a TBK/IKK-independent manner and translocates into the nucleus (still in association with MyD88, TRAF6 and IRAK-1), where it can bind to DNA and stimulate transcription. Recent results also suggest a role for IKK (Hoshino et al., 2006) and osteopontin (Shinohara et al., 2006) in IRF-7 activation.

View larger version (27K): [in this window] [in a new window]   Fig. 3. TLR7- and TLR9-dependent signalling. In pDCs, ssRNA or CpG DNA is presented to TLR7 or TLR9, respectively, in endosomes by endocytosis of extracellular nucleic acids or uncoating of endocytosed viral particles, or by degradation of engulfed apoptotic cells. TLR7 is also stimulated by viral PAMPs taken into the endosomes from the cytoplasm by autophagy. Activated TLRs recruit the adaptor MyD88 that recruits IRAK-4 and IRAK-1. This complex acts as a scaffold to recruit signalling components that feed into either the IRF-7 or NF-B pathways. NF-B activation follows a route similar to that described in Fig. 2, although the role for RIP1 remains to be clarified. IRF-7 recruitment to the MyD88 adaptor complex requires polyubiquitination by TRAF6 in a RIP1-dependent manner. IRF-7 is phosphorylated by IRAK-1 and a complex containing IRF-7, MyD88, TRAF6, IRAK-1 and possibly IRAK-4 is released and migrates to the nucleus (green arrows). Here, it assembles on the IFN-β promoter with NF-B and other factors, leading to the stimulation of transcription. IRF-7 can also stimulate IFN- promoters strongly. See text for more details and references.

(iii) DNA delivered through endosomes.
Certain peripheral blood mononuclear cells (PBMCs) show responses to engulfed ‘foreign’ DNA and can activate IFN production. In this case, the receptor is TLR9, again localized on the endosomes, and the DNA is recognized as foreign because it is usually unmethylated (CpG), unlike host genomic DNA. The importance of TLR9 in innate antiviral immunity is illustrated by the susceptibility of TLR9-deficient mice to MCMV, as a result of reduced IFN production (Krug et al., 2004; Tabeta et al., 2004; Delale et al., 2005).

pDCs, mDCs and macrophages each mount responses to TLR9 agonists, but these cell types use distinct signalling pathways. In pDCs, IFN induction requires endosomal retention of the CpG ligand (Honda et al., 2005a) and involves a MyD88-dependent signalling pathway that is identical to the TLR7-mediated activation of IRF-7 and NF-B, as discussed above (Fig. 3). However, mDCs are unable to retain CpG ligands in their early endosomes and are unable to use this IRF-7 pathway, although they still activate NF-B; rather, they use IRF-1 (Negishi et al., 2006; Schmitz et al., 2007), which is also activated by TLR9-dependent engagement of the MyD88 adaptor and IRAK kinases.

IRF-5 is also involved in the production of IFN by pDCs in response to TLR7 and TLR9 agonists (Barnes et al., 2004; Schoenemeyer et al., 2005; Takaoka et al., 2005; Yasuda et al., 2007). IRF-5 is not involved in IFN production by fibroblasts, but makes a contribution to serum IFN levels during viral infection (Yanai et al., 2007). TLR7 or TLR9 engagement activates IRF-5 via the MyD88 pathway, as discussed above for IRF-7. Mutations in the IRF-5 promoter or allelic variations in the structure of the IRF-5 protein are associated with a predisposition to systemic lupus erythematosus (reviewed by Graham et al., 2007; Kozyrev et al., 2007; Kyogoku & Tsuchiya, 2007) and arthritis (Sigurdsson et al., 2007).

(iv) Intracellular viral RNA.
Cells possess TLR-independent pathways that respond to viral nucleic acids generated in the cytoplasm by viral replication. The activities of two widely expressed RNA helicase molecules (RIG-I and mda-5) are linked to these responses. Although the nature of the activators of RIG-I and mda-5 remain uncertain (see below), RIG-I was identified initially as an essential regulator for poly(rI).poly(rC)-induced signalling in a functional screen (Yoneyama et al., 2004) and binds to poly(rI).poly(rC) in vitro (Rothenfusser et al., 2005; Yoneyama et al., 2005). Ectopic expression of RIG-I enhances poly(rI).poly(rC) responses, and small interfering RNA (siRNA) knockdowns limit IFN-β induction by poly(rI).poly(rC) (Yoneyama et al., 2004, 2005). The related protein mda-5 was identified initially as a binding partner for the IFN-induction antagonist protein of PIV5 (Andrejeva et al., 2004) and has similar properties to RIG-I (Andrejeva et al., 2004; Yoneyama et al., 2005), although it binds less avidly to poly(rI).poly(rC). When activated, these molecules recruit and activate a mitochondrion-associated adaptor called CARD adaptor inducing IFN-β (Cardif)/virus-induced signalling adaptor (VISA)/mitochondrial antiviral signalling protein (MAVS)/IFN-β promoter stimulator protein 1 (IPS-1) (Kawai et al., 2005; Meylan et al., 2005; Seth et al., 2005; Xu et al., 2005; reviewed by Hiscott et al., 2006a; Johnson & Gale, 2006; summarized in Fig. 4). Experiments with knockout mice indicate that this adaptor is essential for responses to both helicases (Kumar et al., 2006; Sun et al., 2006). Strikingly, although these mice have functionally intact TLR-signalling systems, they are extremely susceptible to RNA viruses, indicating that IFN production by the TLR and helicase pathways is not redundant and that IFN production by the helicase pathway is essential for controlling infection by these viruses.

View larger version (23K): [in this window] [in a new window]   Fig. 4. mda-5- and RIG-I-dependent signalling. Viral RNA, generated in the cytoplasm by uncoating, transcription or replication, activates the RNA helicases mda-5 and RIG-I. mda-5 and RIG-I are both activated by dsRNA, whilst RIG-I can also be activated by RNA molecules with 5' triphosphates. Both helicases have N-terminal CARD domains that recruit the adaptor Cardif/VISA/MAVS/IPS-1. This adaptor, in turn, acts as a scaffold to recruit signalling components that feed into either the IRF-3 or the NF-B pathways. Although the details of these downstream signalling pathways remain incomplete, for Cardif/VISA/MAVS/IPS-1 activation, they seem very similar to those events described in Fig. 2 downstream of TRIF. The assembly of an enhanceasome complex on the IFN-β promoter is also equivalent to that described in Fig. 2. See text for more details and references.

A third member of the helicase family (LGP2) has also been described; this protein lacks the CARD domains present in mda-5 and RIG-I and appears to function as a negative regulator of IFN production (Rothenfusser et al., 2005; Yoneyama et al., 2005; Saito et al., 2007). However, this simple model has been challenged by the phenotypes of LGP2^0/0 mice, which show unpredictable antiviral phenotypes (Venkataraman et al., 2007). Engagement of Cardif/VISA/MAVS/IPS-1 leads to independent signalling to both the NF-B and IRF-3 ‘arms’ of the IFN-induction pathway, as discussed above for the TLR3 adaptor TRIF. Cardif/VISA/MAVS/IPS-1 interacts directly with TRAF6 (Xu et al., 2005) and probably with TRAF3, and the activation pathways are very similar. Specifically, in the context of intracellular RNA signalling, the IKK component NEMO acts as an essential adaptor both for NF-B activation and TBK-1/IRF-3 activation, through its interaction with TANK (Zhao et al., 2007).

The relative roles of mda-5 and RIG-I have been studied intensively to determine whether the helicases are redundant parallel sensors or whether they sense different virus-derived signals. In cell-culture systems, both mda-5 and RIG-I signal responses to transfected poly(rI).poly(rC) (Andrejeva et al., 2004; Yoneyama et al., 2004, 2005; Yamashita et al., 2005; Cardenas et al., 2006; Taima et al., 2006). Furthermore, helicase domain-swap experiments have revealed that there are no detectable differences between mda-5, RIG-I or either chimaera in their ability to respond to poly(rI).poly(rC) (Childs et al., 2007). However, studies on knockout mice indicate that mda-5 plays a much more important role than RIG-I in the regulation of total serum IFN-/β levels in response to injected poly(rI).poly(rC) (Kato et al., 2006), in IFN-β production in poly(rI).poly(rC)-treated dendritic cells or macrophages (Gitlin et al., 2006; Kato et al., 2006) and in poly(rI).poly(rC)-transfected embryo-derived fibroblasts (Kato et al., 2006). Curiously, RIG-I plays a limited role in poly(rI).poly(rC) responses in these systems (Kato et al., 2006).

Despite the ability of both helicases to mediate responses to poly(rI).poly(rC), the two molecules differ in their abilities to recognize other types of RNA. The observation that RIG-I was much more efficient than mda-5 at responding to dsRNA generated by in vitro transcription raised the possibility that the two helicases recognized different types of structures within dsRNA. However, RIG-I can bind and respond to ssRNA molecules that contain a 5' triphosphate (Hornung et al., 2006; Pichlmair et al., 2006). Such a triphosphate moiety is present on the RIG-I-activating in vitro-transcribed dsRNA and on many RNA molecules generated during viral replication, but absent from most cellular cytosolic RNAs. For example, cellular mRNAs have 5' cap structures, and RNA polymerase I and most RNA polymerase III transcripts have 5' monophosphates. Specifically, mda-5 was not activated by 5' triphosphate-containing RNAs. The well-documented response of RIG-I to poly(rI).poly(rC) is not explained by these findings, because the bulk of poly(rI).poly(rC) contains little or no 5' triphosphate. Probably, RIG-I can be activated by both 5' triphosphate RNAs and dsRNAs or, alternatively, the 5' triphosphate RNAs examined may contain some secondary structure.

Intriguingly, the magnitude of IFN induction is enhanced by the activation of RNase L (Malathi et al., 2007), an endonuclease that degrades both cellular and viral RNAs to generate short fragments with 3' monophosphates. The induction of IFN by RNase L utilized both mda-5 and RIG-I, and the activation of both was limited by removal of the 3' monophosphates. The authors speculate that the degradation products contain both ssRNA and dsRNA components, but these would both lack 5' triphosphates. Overall, knowledge of the ligands for both helicases remains incomplete, and a full understanding may require the structures of the helicases to be solved.

The above properties offer some explanation for the relative importance of mda-5 and RIG-I in inducing IFNs in response to different RNA viral infections. Experiments on knockout mice show that mda-5 is essential for IFN production in response to picornaviruses (Gitlin et al., 2006; Kato et al., 2006), but is less important in responses to other types of RNA virus, such as influenza A virus, where RIG-I is critical (Kato et al., 2006). These results are consistent with the properties of mda-5 and RIG-I discussed above and the reported status of viral RNA structure generated during the viral infection. Thus, the picornaviruses poliovirus (PV) and encephalomyocarditis virus (EMCV) depend on mda-5 for IFN-β induction and generate detectable levels of dsRNA during infection, whilst influenza A virus and paramyxoviruses do not generate dsRNA and depend on RIG-I (Pichlmair et al., 2006). However, results from cell culture suggest a less clear-cut situation. Ectopic overexpression or gene knockout achieved in cell culture by using siRNA or dominant interfering forms of factors indicate a role for both RIG-I and mda-5 in IFN induction by NDV and SeV (Andrejeva et al., 2004; Yoneyama et al., 2004, 2005; Melchjorsen et al., 2005), whilst mda-5, but not RIG-I, is required for IFN-β induction by measles virus (MeV) (Berghall et al., 2006) and picornaviruses (Kato et al., 2006). In all of these experiments, it should be remembered that the viruses will often encode inhibitors of IFN induction. For example, most paramyxoviruses (including NDV, SeV and MeV) encode an inhibitor of mda-5 (the V protein) that would help to explain why mda-5 gives the impression of making only a limited contribution to IFN induction. It is also important to stress that different preparations of some viruses might produce distinct types of IFN-inducing signal and, if so, that this might influence the apparent contribution of RIG-I and mda-5 to IFN induction. For example, the optimal induction of IFN by SeV was achieved with a mixture of live virus and defective interfering (DI) particles (Johnston, 1981), due to the presence of copyback DI genomes that could self-anneal and form dsRNA (Strahle et al., 2006). It is probable that, during a viral infection, cells may be exposed to and need to block signals emanating from more than one type of PAMP.

(v) Cytoplasmic DNA.
Some mammalian cells, most notably macrophages and DCs, respond to ‘foreign’ DNA present in the cytoplasm (Ishii et al., 2006; Stetson & Medzhitov, 2006a). Such DNA might be distinguished from self DNA by its lack of methylation, higher A+T content (perhaps on the basis of minor-groove topology) or cytoplasmic localization. There seems a considerable degree of disagreement in the literature about the nature of the foreign DNA-recognition system, for example over the involvement of Cardif/VISA/MAVS/IPS-1 (Stetson & Medzhitov, 2006a; Cheng et al., 2007; Takaoka et al., 2007) or whether NF-B is activated (Ishii et al., 2006; Stetson & Medzhitov, 2006a; Cheng et al., 2007). It is possible that there are several distinct novel components, perhaps expressed in a tissue-specific manner. Studies utilizing transfected DNA generally do not investigate whether the DNA remains cytoplasmic, or indeed whether it is transcribed into RNA.

The receptor for cytoplasmic DNA is probably distinct from the TLRs and RIG-I/mda-5, and a candidate receptor molecule called DAI/DLM-1/ZBP1 was identified recently (Takaoka et al., 2007). Consistent with a role in signalling responses to foreign DNA, depletion of DAI/DLM-1/ZBP1 led to a decrease in IFN-β mRNA production during herpes simplex type 1 (HSV-1), but not NDV, infection (Takaoka et al., 2007). The induction of IFN-β by DAI/DLM-1/ZBP1 in response to cytoplasmic DNA involves the activation of TBK1 and IRF-3, but the potential involvement of NF-B remains to be resolved.

(vi) Induction of IFN by viral proteins.
In addition to viral nucleic acids, several viral envelopes or particles can induce IFN-/β, although this is not a general feature of viruses. For example, HIV gp120 is a potent inducer of IFN-/β in PBMCs (Capobianchi et al., 1992), but the best-studied systems in this regard are herpesviruses. In response to exposure to either human cytomegalovirus (HCMV) or HSV-1, human fibroblasts activate IRF-3 complex formation without new protein synthesis (Navarro et al., 1998; Mossman et al., 2001; Preston et al., 2001). Moreover, for herpesviruses, the activation of ISGs by IRF-3 seems sufficient to establish an antiviral state without NF-B activation or IFN production (Mossman et al., 2001). Importantly, although both HCMV and HSV-1 activate IRF-3 rapidly, these viruses encode powerful antagonists of IFN production to counteract activation. Whether viral recognition at the cell surface or viral entry is required to initiate IRF-3 activation is uncertain. However, HSV-1 virions deficient in glycoprotein D (gD), gH, gB or gI were unable to activate IRF-3, suggesting that viral entry is required (Mossman et al., 2001; Preston et al., 2001). Similarly, HCMV-mediated ISG induction was blocked by fusion inhibitors (Netterwald et al., 2004). On the other hand, soluble HSV-1 gD (Ankel et al., 1998) and soluble HCMV gB (Boehme & Compton, 2004) induced IFN-β or IRF-3 activation in some cell types. In contrast to HCMV or HSV-1, Kaposi's sarcoma-associated herpesvirus (KSHV)/human herpesvirus 8 (HHV-8) virions do not activate IRF-3 or IFN-β, despite soluble envelope glycoprotein gpK8.1 doing so (Perry & Compton, 2006), suggesting that HHV-8 has evolved a mechanism to avoid this effect.

In addition to IRF-3 activation, efficient induction of IFN-β usually requires the activation of NF-B. At low m.o.i., HSV-1 and HCMV particles do not induce NF-B in the absence of viral replication (Paladino et al., 2006), implying the existence of a signalling mechanism that activates IRF-3, but not NF-B. Interestingly, at higher m.o.i. with HCMV, NF-B activation and IFN-/β production were observed (Paladino et al., 2006).

Expression of MeV nucleocapsid can activate IRF-3 and cause IFN induction (tenOever et al., 2002), but the generality of this phenomenon remains uncertain. Moreover, the nucleocapsid is an avid RNA-binding protein and could function simply to present RNA to the signalling system.

Macrophages and conventional DCs respond to bacterial pathogens and lipopolysaccharide (LPS) by making IFN. This activation is TLR4-dependent and, as for TLR3, TLR7/8 and TLR9 (discussed above), TLR4 activates IRF-3 in addition to NF-B. However, TLR4 is unique among these TLRs in responding entirely to an extracellular signal. Whilst the biological significance of induction of IFN in response to LPS remains unclear, there is evidence that respiratory syncytial virus (RSV) can induce IFN synthesis through TLR4 via recognition of its F protein (Kurt-Jones et al., 2000; Haynes et al., 2001). Furthermore, VSV gG can activate IFN-/β production through a TLR4-dependent pathway (Georgel et al., 2007).

Although TLR4 utilizes MyD88 to activate NF-B, it uses the same adaptor as TLR3 (namely TRIF) to activate IRF-3 (Yamamoto et al., 2002; Oshiumi et al., 2003). Both MyD88 and TRIF are recruited indirectly to TLR4, via the TIRAP/MyD88 adaptor-like (Mal) (Fitzgerald et al., 2001; Horng et al., 2001) and TRIF-related adaptor molecule (TRAM) (Yamamoto et al., 2003b) proteins, respectively.

Signalling responses to IFN
The basic signalling pathway activated in response to IFN-/β has been characterized in detail and reviewed comprehensively (Platanias, 2005; Fig. 5). Key players in this signalling pathway are the Janus/just another kinase (JAK) family (cytoplasmic tyrosine kinases that are recruited to a range of receptors in response to ligand binding) and the signal transducers and activators of transcription (STATs) (members of a family of transcription factors that are phosphorylated by JAKs and bind to DNA; reviewed by Murray, 2007).

View larger version (23K): [in this window] [in a new window]   Fig. 5. Signalling pathway activated by IFN-/β. The biological activities of IFN-/β are initiated by binding to the type I IFN receptor. This leads to the activation of the receptor-associated tyrosine kinases JAK1 and Tyk2, which phosphorylate STAT1 on tyrosine 701 and STAT2 on tyrosine 690. Phosphorylated STAT1 and STAT2 interact strongly with each other by recognizing SH2 domains, and the stable STAT1–STAT2 heterodimer is translocated into the nucleus, where it interacts with the DNA-binding protein IRF-9. The IRF-9–STAT1–STAT2 heterotrimer is called ISGF3 and it binds to a sequence motif (the IFN-stimulated response element or ISRE) in target promoters and brings about transcriptional activation. In addition to the phosphorylation of tyrosine, STAT1 also requires phosphorylation on serine 727 for function. See text for more details and references.

Signalling in response to type III IFN follows a very similar pattern to that in response to type I IFN (Zhou et al., 2007), and signalling in response to type II IFN also shares some features of these pathways. The IFN- receptor is also a heterodimeric glycoprotein, comprising two species of major subunit, IFNGR1 and IFNGR2, that pre-associate weakly in unstimulated cells. The cytoplasmic domains of the IFNGR1 and IFNGR2 receptor subunits are associated with a different set of JAKs, JAK1 and JAK2, respectively. When dimeric IFN- binds to the receptor subunits, the signalling pathway is initiated by triggering receptor dimerization. This brings JAK1 and JAK2 into close proximity, resulting in the activation of JAK2 which, in turn, trans-phosphorylates JAK1, thereby activating it. Subsequently, the activated JAKs phosphorylate a tyrosine-containing region in the C terminus of IFNGR1 (between tyrosine 440 and tyrosine 444), creating a pair of binding sites for STAT1. Two STAT1 molecules then interact via their SH2 domains with IFNGR1 and are phosphorylated at tyrosine 701, resulting in their activation and dissociation from the receptor. The activated STAT1 molecules dimerize through SH2 domain–tyrosine phosphate recognition, forming a STAT1–STAT1 homodimer. This homodimer translocates to the nucleus and binds to the unique element of ISGs, the gamma-activation sequence (GAS), and stimulates transcription. Importantly, this sequence is distinct from the ISRE, and the binding of the STAT1–STAT1 homodimer does not require IRF-9.

Although not a major component in IFN-/β signalling, STAT1 homodimers can also be activated by IFN-/β (the so-called AAF complex) and trigger the transcription of some genes (Decker et al., 1991). Furthermore, the STAT1–STAT2 heterodimers formed by IFN-/β stimulation bind to some promoters and stimulate transcription (Li et al., 1996), but the importance of this in the overall antiviral response is unknown.

Additional signalling pathways that can influence the outcome of IFN-induced transcription are activated by IFN (reviewed by Platanias, 2005). Transcriptional activation by STAT1, whether activated by IFN-/β or IFN-, also requires phosphorylation on serine 727 for full activity and for mounting a full antiviral response. Several kinases catalyse this phosphorylation, including the protein kinase C isoform PKC-, which interacts directly with STAT1 (Uddin et al., 2002; Deb et al., 2003). PKC- is activated by the PI3 kinase pathway that is also activated by IFN-/β (reviewed by Kaur et al., 2005). Serine 727 phosphorylation facilitates the interaction of STAT1 with the basal transcription machinery and with other adaptor proteins. STAT2 is not serine-phosphorylated in response to IFN, but it also binds CBP/p300 (see above) and facilitates interaction with the basal transcriptional machinery (Bhattacharya et al., 1996). Treatment of cells with IFN-/β also activates the p38 MAP kinase (mitogen-activated protein kinase) pathway, which can have important effects on the antiviral programme, although the molecular basis of this is unknown (reviewed by Katsoulidis et al., 2005).

Whilst the activation of an IRF-9–STAT1–STAT2 heterotrimer is the canonical mode of ISRE activation, there are alternative transcription-factor arrangements that activate ISGs through ISREs, and which may show promoter-specific effects or distinct kinetic profiles (reviewed by van Boxel-Dezaire et al., 2006). Activation of a subset of ISRE-dependent genes by IFN-/β depends on the phosphorylation of STAT1 on tyrosine 708 by IKK (which is activated in infected cells) and this phosphorylation may stabilize the association of ISGF3 with these ISREs (tenOever et al., 2007). A heterotrimer containing IRF-9 and two molecules of STAT1 (as opposed to a STAT1–STAT2 pair) is activated by IFN- (Bluyssen et al., 1995) and activates several ISRE-containing promoters, including the CXCL10 promoter (Majumder et al., 1998). This type of complex, activated differentially by IFN-, may mediate the observed subtle differences in transcript profile induced by IFN- and IFN-/β. STAT2–STAT2 homodimers can form a complex with IRF-9 (Bluyssen & Levy, 1997), although the biological role of this is unclear.

Importantly, many ISGs show either sustained induction in response to IFN or biphasic induction kinetics. These responses are dependent on new protein synthesis and, in many cases, depend on IRF-1, which binds to most, if not all, ISREs. In several cases, IRF-1 interacts with STAT1 for optimal transcriptional activation (e.g. Chatterjee-Kishore et al., 2000a, b). Although IRF-1 is induced by IFN-/β and IFN-, it is more responsive to IFN- and this is probably a major reason for the difference in gene expression between the two types of IFN. Notably, all of the components of ISGF3 are themselves IFN-inducible, and the increase in subunit concentration may permit the formation of an ISGF3-like complex that does not require STAT tyrosine phosphorylation. In this context, IRF-9 induction by IFN-β requires IRF-9 and STAT1 without a requirement for STAT1 phosphorylation (Rani & Ransohoff, 2005).

Nature of the IFN-induced antiviral state
Treatment of cells with IFN-/β upregulates the expression of several hundred genes and, in combination, these specify the antiviral state (Fig. 6). No single gene is pivotal and, for any given virus, a subset of genes is probably required to limit viral replication. Several of the upregulated genes encode enzymes that have been studied intensively and reviewed comprehensively, e.g. dsRNA-dependent protein kinase R (PKR) (Garcia et al., 2006), 2'5'-oligoadenylate synthetase (OAS) (Clemens, 2005; Silverman, 2007) and Mx (Martens & Howard, 2006; Haller et al., 2007).

View larger version (51K): [in this window] [in a new window]   Fig. 6. Biological properties of IFN-/β. IFN-/β binds to its receptor and initiates the signalling programme outlined in Fig. 5. The IFN-induced transcripts encode proteins that mediate the antiviral response. Some of these proteins (e.g. PKR and OAS) are enzymes whose activities are dependent upon viral co-factors (e.g. dsRNA) and, when such co-factors are provided, the enzymes can bring about dramatic changes in cellular function (such as translational arrest). Other IFN-inducible factors trigger cell-cycle arrest (e.g. the G1/S phase-specific cyclin-depende