Translational Mini-Review Series on Toll-like Receptors:
Recent advances in understanding the role of Toll-like receptors in anti-viral immunity
Andrew G. Bowie, School of Biochemistry and Immunology, Trinity College Dublin, Dublin 2, Ireland.
Toll-like receptors (TLRs) respond to pathogens to initiate the innate immune response and direct adaptive immunity, and evidence to date suggests that they have a role in the detection of viruses. Many viral macromolecules have been shown to activate anti-viral signalling pathways via TLRs, leading to the induction of cytokines and interferons, while viruses also have means of not only evading detection by TLRs, but also of subverting these receptors for their own purposes. This review discusses the role of TLRs in the context of other known viral detection systems, and examines some of the often surprising results from studies using mice deficient in TLRs and their adaptors, in an attempt to unravel the particular contribution of TLRs to anti-viral immunity.
In order for the immune system to respond appropriately to the presence of a certain pathogen, host pattern recognition receptors (PRRs) must detect pathogen-associated molecular patterns (PAMPs). Until 5 years ago, how mammalian cells detect viruses to initially trigger innate anti-viral signalling pathways was almost a complete mystery, although important downstream anti-viral transcription factors such as nuclear factor kappa B (NFκB) and interferon regulatory factor 3 (IRF3) were known to be activated by viral infection of cells, and to be important in initiating the anti-viral response through the induction of type I interferons (IFNs). At that stage, only one viral PRR was known, namely dsRNA-dependent kinase (PKR). With the discovery that TLRs were viral PRRs, a key missing link was identified connecting viral infection to IFN induction . Further, in Drosophila melanogaster, where proteins involved in the immune response display a striking homology to those with a role in mammalian innate immunity, the Toll pathway, which was known previously to respond to Gram-positive bacteria and fungal infections, has now been shown to be required for insect anti-viral immunity . However, in the past 2 years the identification of further viral PRRs in the form of the RNA helicases retinoic acid-inducible gene I (RIG-I) and Mda5, together with unexpected results from TLR–/– mice infected with viruses, have led to the role of TLRs in the anti-viral response being questioned and refined.
In this review the particular role of TLRs, compared to other viral PRRs, is discussed in the light of recent discoveries. Recent progress in understanding virus–TLR interactions is presented. These interactions are complex, as although viral PAMPs can activate TLRs leading to gene induction, other viral proteins work to block and evade TLR function. Furthermore, some viruses actually subvert the TLR system for their own purposes, for example to facilitate entry into the cell or replication, or to manipulate the extracellular cytokine environment to the benefit of the virus.
TLRs mediate cellular responses to viral proteins and nucleic acids
Table 1 shows PRRs that have been implicated in responding to viruses and, where this is known, the particular macromolecular structure recognized. Viral dsRNA is a prototypic PAMP, which is thought to be produced during many viral infections. The first viral PRR to be proposed was PKR, an intracellular detector of dsRNA and polyinosinic–polycytdylic acid (poly(I:C) (a synthetic dsRNA widely used to mimic viral infection). Subsequent knock-out studies confirmed an important role for PKR in anti-viral innate immunity . A key function of PKR is to inhibit translation of host cell mRNA under conditions of viral invasion by phosphorylating the translation initiation factor eIF2α. However, PKR was unlikely to be capable of mediating all the known downstream effects of viral infection, and it was the observation that mice defective in PKR can still respond to poly(I:C) that led to the discovery of TLR3 as a second PRR for dsRNA . Induction of proinflammatory cytokines and type I IFNs in response to poly(I:C) in macrophages and splenocytes from TLR3–/– mice were impaired . Furthermore, dsRNA from a reovirus failed to induce CD69 expression in splenocytes from TLR3–/–animals, in comparison to wild-type cells . TLR3 is now thought to act predominantly in endosomes but, at least for some cell types, also at the cell membrane . More recently, the human TLR3 ectodomain structure has been solved [6,7], revealing a large horseshoe-shaped solenoid comprising 23 leucine-rich repeats. Although the extracellular domain was covered largely by carbohydrates one face was glycosylation-free, and this contained two patches of positively charged residues, suggesting the probable site of binding of negatively charged phosphate-containing dsRNA. Although there is still no direct evidence of viral dsRNA binding to TLR3, both Choe et al.  and Bell et al.  showed that synthetic dsRNA [including poly(I:C)] could indeed bind to the ectodomain. Bell et al. went on to locate the poly(I:C) binding site on TLR3 by mutational analysis , which confirmed that the glycan-free surface was indeed the site of ligand binding. They also identified two amino acids critical for function (H529 and N541), verifying their hypothesis that two bound sulphate ions (which have a very similar atomic arrangement to phosphate groups) marked the ligand binding site. This represents the first molecular description of how TLRs bind their ligands. However, although TLR3 can no doubt mediate cellular responses to dsRNA and poly(I:C), its particular role in anti-viral innate immunity is still being defined (see below).
Table 1. Pattern recognition receptors (PRRs) and the proposed virally derived macromolecules to which they respond. See Table 2 for explanation of abbreviations. PRRs are listed in the order in which they were identified as having a role in detecting viruses.
|TLR2||CMV||Glycoproteins gB and gH||[11,12]|
|TLR7||IAV, HIV, VSV,||ssRNA containing uridines and ribose||[23,80]|
|RIG-I||EBV||EBV-encoded small RNAs|||
|SeV, VSV, HCV||dsRNA|||
|RV||5′ phosphorylated genomic ssRNA|||
|IAV||5′ phosphorylated genomic ssRNA|||
The first indication that viral proteins could activate TLRs was the case of the fusion (F) protein of respiratory syncitial virus (RSV) (see Table 2 for virus nomenclature) and TLR4 . RSV commonly causes severe lower respiratory tract infections in children. F protein stimulated secretion of interleukin (IL)-6 from wild-type cells, but not those isolated from C3H/HeJ mice (which have a mutation in the TLR4 Toll/IL-1R (TIR) domain) or C57BL10/ScCr mice (in which the gene encoding TLR4 is deleted). However, another study found no significant role for TLR4 in murine RSV infection , and therefore the importance of F protein-induced cytokine production via TLR4 remains unclear. Similarly, TLR2 has been shown to mediate cellular responses to viral glycoproteins. For cytomegalovirus (CMV), TLR2 was shown to be responsible for the activation of NFκB and the induction of IL-6 and IL-8 by ultraviolet light-inactivated virions, demonstrating a replication-independent effect . More recently, two CMV entry-mediating envelope glycoproteins, gB and gH, have been shown to mediate these TLR2-dependent effects, and to co-immunoprecipitate with TLR2 and TLR1 .
Table 2. Virus nomenclature; (+), positive-stranded RNA; (–), negative-stranded RNA; (RT), reverse-transcribed RNA.
|EMCV||Encephalomyocarditis virus||Picornaviridae||ssRNA (+)|
|HCV||Hepatitis C virus||Flaviviridae||ssRNA (+)|
|HIV||Human immunodeficiency virus||Retroviridae||ssRNA (RT)|
|HPEV||Human parechovirus||Picornaviridae||ssRNA (+)|
|HSV||Herpes simplex virus||Herpesviridae||dsDNA|
|IAV||Influenza A virus||Orthomyxoviridae||ssRNA (–)|
|JEV||Japanese encephalitis virus||Flaviviridae||ssRNA (+)|
|LCMV||Lymphocytic choriomeningitis||Arenaviridae||ssRNA (–)|
|MMTV||Mouse mammary tumour virus||Retroviridae||ssRNA (RT)|
|NDV||Newcastle disease virus||Paramyxoviridae||ssRNA (–)|
|RSV||Respiratory syncytial virus||Paramyxoviridae||ssRNA (–)|
|RV||Rabies virus||Rhabdoviridae||ssRNA (–)|
|SeV||Sendai virus||Paramyxoviridae||ssRNA (–)|
|VSV||Vesicular stomatitis virus||Rhabdoviridae||ssRNA (–)|
|WNV||West Nile virus||Flaviviridae||ssRNA (+)|
TLRs have also been implicated in sensing other types of viral nucleic acid, apart from dsRNA. TLR9 was shown originally to be activated by bacterial DNA sequences containing unmethylated CpG dinucleotides . These motifs are also found in abundance in some dsDNA viral genomes. Herpes simplex virus (HSV), which has such a genome, has been shown to induce type I IFN in plasmacytoid dendritic cells (pDCs) through TLR9 [14,15]. In TLR9–/– mice injected with HSV no IFN-α was detected, while pDCs treated with purified HSV DNA released IFN-α. Recognition of HSV by pDCs did not require virus replication and was achieved through an endocytic pathway that was inhibited by chloroquine or bafilomycin A. This is consistent with the fact that TLR9 is located in and signals from intracellular endosomal compartments [16,17]. Therefore, as in the case of TLR3, recognition of viral nucleic acid by this TLR is largely endosomal. Thus TLR9 can probably sample endocytosed material and trigger an immune response whenever an endocytosed dsDNA virus is present.
TLR7 and TLR8, which are related more closely to TLR9 than to other TLRs in terms of sequence similarity, also act in endosomes where they have been shown to trigger gene induction in response to viral ssRNA. The first indication of an anti-viral role for TLR7 was the observation that the imidazoquinoline resiquimod (or R-848), which was known to have potent anti-viral properties, was shown to activate murine macrophages in a TLR7-dependent manner . It was shown subsequently that human TLR7 or TLR8, but not murine TLR8, could also confer cells with responsiveness to R-848 . Subsequently, guanosine- and uridine-rich ssRNA oligonucleotides derived from HIV-1 were shown to stimulate dendritic cells (DCs) and macrophages to secrete IFN-α and proinflammatory cytokines via human TLR8 and murine TLR7 . In addition, Diebold et al.  showed that the production of large amounts of IFN-α by pDCs in response to wild-type influenza virus required endosomal recognition of influenza genomic RNA and signalling via murine TLR7. In corroboration of these studies, Lund et al.  demonstrated that as well as recognizing influenza, TLR7 was required for pDC and B cell responses to another ssRNA virus, vesicular stomatitis virus (VSV). More recently, Diebold et al.  have clarified the molecular basis for the recognition of ssRNA by TLR7: they found that uridine and ribose, the two defining features of RNA, are both necessary and sufficient for TLR7 stimulation, that short ssRNAs act as TLR7 agonists in a sequence-independent manner as long as they contain several uridines in close proximity and that such RNA species derived from either viral or self RNA trigger TLR7 activation equally efficiently if delivered to endosomes . Thus they concluded that TLR7 seems to recognize uracil repeats in RNA and that it discriminates between viral and self-ligands on the basis of endosomal accessibility rather than sequence.
Activation of anti-viral signalling by TLRs
The signalling pathways triggered by TLR engagement of viral PAMPs leading to the induction of both cytokines and type I IFNs have been well described recently in many reviews [24–27], and therefore in this review only a few of the key aspects of these pathways will be highlighted. The most studied anti-viral transcription factors activated by TLR signalling pathways are NFκB and IRF family members, particularly IRF3, 5 and 7. TLR-activated transcription factors co-operate to induce a huge number of effector genes, many of which have proven roles in anti-viral innate immunity. For example, IFN-β induction by TLRs can involve co-operation between NFκB, IRF3 and IRF7, IFN-α requires mainly IRF7 while TNF-α and IL-6 are NFκB- and IRF5-dependent .
TLRs engage downstream signalling pathways to activate these transcription factors via five TIR adaptors, in signalling processes involving IL-1R-associated kinases (IRAKs) and TNF receptor-associated factor 6 (TRAF6). These signalling processes lead ultimately to activation of kinases such as IκB kinases and TBK1, which phosphorylate NFκB and IRFs, respectively, to variously direct their dimerization, nuclear translocation and transactivating potential . The five TIR adaptors, MyD88, Mal, TRIF, TRAM and SARM, are differentially utiilized by TLRs leading to different signalling specificities (see Table 3). TRAM is uniquely utilized by TLR4 for both NFκB and IRF3 activation, while Mal has been shown to have a role in NFκB activation for both TLR4 and TLR2. TRIF is important in mediating IRF3 and NFκB activation for TLR4. Importantly, TRIF is also the sole adaptor used by TLR3, leading to activation of IRF3, IRF7 and NFκB . MyD88 has the most ubiquitous role and has been implicated in all TLR signalling pathways, with the exception of TLR3 . It was shown originally to be essential for signalling to TLR-induced NFκB activation, but was also shown subsequently to be required for TLR-induced IRF5 activation , and for TLR7- and TLR9-induced IRF7 activation . SARM, the fifth adaptor, has been shown to have a negative regulatory role on the TLR3 and TLR4 TRIF-dependent pathways .
Table 3. Adaptors used by Toll-like receptors (TLRs) and the transcription factors they activate. Examples of genes induced by each pathway and cell types in which the pathways operate are also shown.
|TLR2||Mal and MyD88||NFκB||IL-6, TNF-α||Macrophage|
|TLR3||TRIF||NFκB, IRF3, 5, 7||IL-6, TNF-α, IFN-β||mDC|
|TLR4||Mal and MyD88||NFκB, IRF5||IL-6, TNF-α, IL-12||Macrophage|
|TRAM and TRIF||NFκB, IRF3||IFN-β, RANTES|| |
|TLR7||MyD88||NFκB, IRF5, 7||IL-6, TNF-α, IFN-α||pDC|
|TLR9||MyD88||NFκB, IRF5, 7||IL-6, TNF-α, IFN-α||pDC|
|MyD88||IRF1||IFN-β, iNOS, IL-12||mDC|
A detailed knowledge of which TLRs activate particular transcription factors, together with a growing awareness of which cell types express distinct TLRs, is helping to define the particular role of TLRs in the anti-viral response and to provide important immunological context. For example, the TLR7/9–MyD88–IRF7 axis is especially important for pDC-induced IFN-α production . Most recently it has been shown that in myeloid DCs, TLR9-induced IFN-β, IL-12 and inducible macrophage type nitric oxide synthase (iNOS) is dependent on MyD88-mediated IRF1 activation . Because IRF1 expression is strongly up-regulated by IFN-γ, this provides a mechanism whereby IFN-γ enhances TLR signalling.
Distinguishing the TLRs from other viral PRRs
Although TLRs can respond to numerous viral PAMPs leading to activation of anti-viral signalling pathways, for numerous reasons TLRs cannot account fully for the anti-viral response . In contrast to the fact that TLRs that can induce type I IFN are expressed on restricted cell types, such as macrophages and DCs, virtually all cells in the body are capable of producing type I IFN in response to viral infection. Further, the viral nucleic-acid sensing TLRs (TLR3, 7, 8, 9) are located in endosomes of these restricted cells, and so would not sense intracellular cytosolic infection. Importantly, mice lacking MyD88 (utilized by all TLRs except TLR3) or TLR3 have been shown to be resistant to infection with several viruses [5,30].
Subsequent to the discovery that TLRs recognize viral PAMPs a further class of viral PRRs was discovered, the cytosolic RNA helicases RIG-I and Mda5 [33,34], which seem to account for many of these discrepancies. These helicases were thought initially to detect viral dsRNA, although recently RIG-I has been shown to recognize uncapped 5′-triphosphate viral ssRNA [35,36]. RIG-I and Mda5 each contain two caspase-recruitment domains (CARDs) and signal through the same signalling adaptor molecule, IPS-1 [also called mitochondrial anti-viral signalling protein (MAVS)], virus-induced signalling adapter (VISA)  and CARD adapter inducing IFN-β (CARDIF) ), leading to NFκB, IRF3 and IRF7 activation. Importantly, studies of mice lacking either RIG-I or Mda5 demonstrated distinct roles for the helicases in vivo[41,42]. RIG-I was required for type I IFN production in response to paramyxoviruses (Newcastle disease virus (NDV) and Sendai virus (SeV)), VSV, influenza A virus (IAV) and Japanese encephalitis virus (JEV), while Mda5 mediated IFN induction by the picornavirus encephalomyocarditis virus (EMCV) . Further, RIG-I–/– and Mda5–/– mice were very susceptible to infection by these respective viruses compared to control mice. Additionally, Mda5 was shown to contribute to poly(I:C) responses in vivo, unlike RIG-I, which senses in vitro-transcribed synthetic dsRNA .
Apart from unique upstream detection systems, a comparison of the different anti-viral transcription factors and genes activated by TLRs versus the helicases provides little information so far on distinct functions for each system, as similar signalling pathways are activated downstream of their key adaptors (the TIR adaptors and IPS-1, respectively). One notable exception may be IRF5 as so far only the TLRs, and not the helicases, have been shown to activate it. Nevertheless, the key difference between the two detection systems seems to centre on their location, both in terms of which cells they are expressed by and where in the cell they function.
It has been shown that in non-immune cells such as fibroblasts, as well as in non-plasmacytoid DCs, RIG-I rather than TLRs mediates IFN and cytokine induction in response to RNA viruses such as SeV, NDV and VSV [43,44]. In contrast, studies in mice lacking IPS-1 and RIG-I confirmed that pDCs use the TLR system (TLR7 and TLR9) to produce large amounts of IFN-α[44–46]. For example, IPS-1–/– mice produced normal levels of type I IFNs in response to VSV, although they were still highly susceptible to VSV infection and had greatly increased viral loads [45,46]. These apparently conflicting results are probably explained by the fact that although pDCs produce high amounts of circulating IFN systemically, the reduced amount of local IFN production at the site of infection (due to absence of IPS-1 and hence lack of RIG-I signalling) is critical to control viral load . One important point to note is that these data show that some viruses at least are clearly detected by both the TLR and helicase systems. Further, a study with RSV has shown a distinct temporal role for RIG-I and TLR3 in mediating RSV-induced innate responses in that siRNA-induced knockdown of either RIG-I or TLR3 in airway epithelial cells showed that RIG-I was essential for early induction of IFN-β and other IRF-dependent genes, whereas TLR3 mediated induction of the same genes at late times of infection . In fact, TLR3 induction following RSV infection was regulated by RIG-I-dependent secretion of IFN-β.
Subcellular localization is also important for understanding the relative role of TLRs to the other viral PRR systems. This becomes apparent upon considering the viral life cycle, and how viruses enter cells and replicate. Viral entry usually occurs by either receptor-mediated endocytosis followed by endosomal fusion, or by direct viral membrane fusion with the cell membrane. In the latter case, uncoating (which exposes the viral nucleic acid so that viral gene expression and replication occur) will happen in the cytosol. The TLRs probably have a role in sensing viruses in the initial virus–cell interaction, whereby TLR2 and TLR4 can respond to some viral surface glycoproteins, as discussed above for CMV and TLR2. For viruses utilizing endocytosis and then endosomal fusion, TLR3, 7, 8 or 9 could then encounter viral nucleic acid, depending on the particular cells infected. In contrast, viruses that enter cells by direct fusion of the viral and cell membrane and then uncoat will encounter RIG-I and Mda5 in the cytosol. The fact that RIG-I and Mda5 have been linked with so many RNA viruses is consistent with the fact that all RNA viruses except influenza and retroviruses replicate in the cytosol. Because viral replication products will accumulate in the cytosol during an infection, RIG-I induced responses may be more sustained and hence prominent than TLR-induced responses in endosomes. However, endosomal TLRs in phagocytic cells can also detect viral products by phagocytosis of infected cells or of debris from lysed cells.
A further cytosolic PRR has been proposed but not identified (Receptor X in Table 1) , which can induce IFN in response to transfected DNA, but not via IPS-1 [45,46]. This PRR might respond to viral genomic dsDNA in the cytosol after uncoating, especially for poxviruses such as vaccinia virus (VACV) which, in contrast to herpesviruses (such as CMV and HSV), replicate in the cytoplasm.
Clarifying the role of TLR3
Although TLR3 is probably the most ubiquitously expressed anti-viral TLR, and is clearly a receptor for dsRNA, its particular role in innate immunity is still perplexing . An initial study in TLR3–/– mice showed that the immune response to four different viruses, lymphocytic choriomeningitis virus (LCMV), VSV, a reovirus and CMV, was unaffected compared to control mice . A further study on CMV, however, did find a role for TLR3 (see below). None the less, it was surprising that TLR3 seemed to have no role in controlling infection for RNA viruses, especially for reovirus, given its dsRNA genome, and the fact that reovirus RNA had been shown previously to stimulate TLR3 .
TLR3 was shown subsequently to play an important, but not protective, role in infection by the ssRNA virus West Nile virus (WNV). Wang et al.  demonstrated that a peripheral inflammatory response (in the form of IL-6 and TNF-α production) was initiated through TLR3, leading to disruption of the blood–brain barrier, which allowed for virus entry into the brain (see Table 4). Therefore, TLR3–/– mice were more resistant to lethal WNV infection, so that in this case the virus appeared to benefit from its interaction with TLR3.
Table 4. Representative results from mouse Toll-like receptor 3 (TLR3) knock-out studies with viruses.
|↑Viral load in blood|
|↓IL-6 and TNF-α|
|↓Viral load in brain|
|↑Blood–brain barrier integrity|
|EMCV||↑Viral load in heart|||
|↓Proinflammatory cytokines and chemokines|
|Normal IFN-β expression|
|↑Viral load in lungs|
|↓Proinflammatory cytokines (IL-6, IL-12)|
|↓CD8+ T cells in lung|
|CMV||↑Viral load in spleens|||
|↓Mortality (but not significant)|
|↓Type I IFN in serum|
|↓IL-12 and IFN-γ in serum|
|↓NK and NK T cell activation|
Similar to WNV, other RNA viruses also appear to induce inflammation via TLR3. Although inflammation may be seen as a protective anti-viral response, it seems that in the absence of TLR3 there is often decreased virally induced pathology and increased survival, as shown in Table 4. EMCV-induced myocardial injury can lead to severe heart failure, and mice deficient in TLR3 were found to be more susceptible to EMCV infection and have a higher viral load in the heart . TLR3 was found to have no role in IFN induction by the virus but to be responsible for increased inflammation in the heart, and for virus-induced proinflammatory cytokine and chemokine production . Similarly, in TLR3–/– mice infected with IAV, a virus which causes a highly contagious acute respiratory disease, there was a reduced production of inflammatory mediators leading to increased survival . Here the authors concluded that the TLR3–IAV interaction therefore leads to a detrimental host inflammatory response, thus contributing to IAV-induced pneumonia. Virally mediated inflammation and pathology via TLR3 has also been seen for other RNA viruses , and this seems to be an emerging theme. Gowen et al.  have proposed recently that in naturally acquired viral infections, where exposure to smaller infectious doses would occur, TLR3 may play a role in managing infection and co-ordinating the balance of inflammatory mediators. This would be in contrast to the situation experimentally, where the higher viral doses used may lead to over-production of proinflammatory mediators.
In contrast to the ambiguous role of TLR3 described above, a study by Reis e Sousa and colleagues has provided a clear specific role for TLR3 in the anti-viral response, and also solved an immunological puzzle related to the regulation of CD8+ T cells by cross-presentation of antigens not expressed by antigen-presenting cells . Although DCs were known to be the principal cells involved in cross-presentation, it was unclear which signals determine whether cross-presentation resulted in cross-priming, leading to a cytotoxic T cell (CTL) response, or in cross-tolerance, resulting in CD8+ T cell inactivation. In this study, murine CD8α+ DCs (known to express TLR3 strongly in endosomes and to have a central role in cross-presentation in mice) were activated to mature and to produce cytokines by phagocytosis of apoptotic bodies from virally infected cells containing dsRNA in a TLR3-dependent manner . Furthermore, immunization with virally infected cells or with cells loaded with poly(I:C) led to an impressive increase in TLR3-dependent CTL cross-priming against cell-associated antigens. The authors concluded that TLR3 may have evolved to permit cross-priming of CTLs against viruses that do not directly infect DCs.
Finally, TLR3 seems to be the TLR expressed most strongly in the brain, particularly in astrocytes . Further TLR3 expression is induced on human adult astrocytes in culture upon stimulation with proinflammatory cytokines, or TLR3 or TLR4 agonists and, when activated, TLR3-induced neuroprotective mediators. Thus TLR3 may have a specific role in the brain, but this needs to be investigated further .
TLRs and large DNA viruses
The data so far with TLRs and DNA viruses suggests a more direct protective role for TLRs compared to the situation with RNA viruses. TLR3 has been implicated clearly for at least one DNA virus (CMV), and indirectly for another (VACV). It may seem a contradiction that a dsRNA receptor protects more clearly from DNA rather than RNA viruses, but DNA viruses probably produce dsRNA as a result of bi-directional transcription of their genomes, and this has now been demonstrated for HSV and VACV . Of note, mice with a defective trif gene were more susceptible to CMV, and had much higher viral titres in their spleens than infected control mice . Compellingly, although high levels of type I IFN was detected in the serum of infected control mice, none was seen in the serum of mice with the trif mutation . Because TRIF is used by both TLR3 and TLR4 for signalling (see above) conceivably either TLR could be involved here. However, a later paper by the same group showed clearly, using TLR3–/– mice, a role for this TLR in the anti-CMV response (see Table 4). In the absence of TLR3, mice infected with CMV showed increased viral titres in the spleen, hints of higher mortality, reduced serum IFNs and cytokines and decreased activation of NK and NK T cells .
This study also showed that for CMV, TLR9 also contributes to protection against the virus, due presumably to recognizing the CMV dsDNA genome (Table 1), and that TLR9 makes a stronger contribution than TLR3 , probably via viral detection in endosomes of pDCs (see above). It seems that both TLR3 and TLR9 contribute independently to the anti-viral response to CMV. TLR2 may also contribute to protection, given that CMV virion proteins can interact with it and trigger cytokine induction (see Table 1).
Another DNA virus linked strongly with TLRs is HSV. As already mentioned above, both intact HSV and purified HSV DNA have been shown induce type I IFN in pDCs through TLR9 [14,15]. As well as TLR9, TLR2 may also contribute to the host response to HSV. Kurt-Jones et al.  found that neonatal mice lacking TLR2 were resistant to HSV-induced encephalitis. Similar to the case of TLR3 and WNV, TLR2 seemed to be causing HSV-induced inflammation, which in natural infections may facilitate viral clearance, but in this instance was causing tissue damage. Another study found that although TLR2–/– mice showed no enhanced susceptibility to intranasal infection with HSV, MyD88–/– mice were highly susceptible, showing 100% mortality by day 10 post-infection . For HSV, a full anti-viral response requires detection of viral PAMPs by TLRs from both infected epithelial cells in the mucosal barrier as well as stimulation of TLRs in non-infected DCs . Hence virally infected tissues direct DCs to initiate an appropriate effector T cell response, and both steps require TLR engagement. However, non-TLR pathways are also clearly involved in IFN induction and controlling HSV replication in vivo[15,64,65].
A third DNA virus strongly linked to TLRs is the poxvirus VACV. This was the first virus for which immune evasion strategies employed against TLRs were identified , which suggested that TLRs are probably important in containing VACV. It was shown subsequently that in macrophages isolated from mice lacking a functional trif gene, VACV replicated to a much higher level than in control cells, thus directly linking the TRIF pathway with suppression of VACV replication, presumably via IFN induction . Another study recently showed TLR2- and MyD88- (but not TRIF-) dependent production of IL-1 and IL-6 in response to VACV infection of murine immature bone marrow-derived DCs . In vivo production of IL-6 in response to VACV infection was also TLR2- and MyD88-dependent . It has also now been shown that in human keratinocytes, VACV infection induces the production of an anti-microbial peptide via TLR3 stimulation, which can suppress VACV replication . Thus, TLRs probably play an important role in containing poxviruses in both mice and humans.
Evasion and subversion of TLRs by viruses
In order to replicate successfully and spread within a host, every virus must have ways of overcoming or suppressing the anti-viral response. VACV has many proteins which can inhibit NFκB and IRF3 activation, but so far A46 and A52 are the only two which block specifically TLR-induced transcription factor activation. They have distinct modes of action and target separate host proteins on the TLR signalling pathways, most probably leading to the disruption of the formation of crucial host signalling complexes. A52 can interact with both TRAF6 and IRAK2 and inhibit TLR-induced NFκB activation, especially for TLR3 , while A46 can interact with human MyD88, Mal, TRIF and TRAM and thus inhibit both TLR-induced NFκB and IRF activation, although it has less of an effect on TLR3-induced NFκB activation than A52 . A46 and A52 are not redundant, as a deletion of either gene in isolation led to an attenuated phenotype in a murine intranasal model of infection [69,70]. Thus, targeting TLR signalling pathways in infected cells confers an advantage on VACV in vivo. A strategy to evade detection by TLRs using TRIF has also been identified in HCV. This virus has a serine protease which can cleave TRIF and thus prevent TLR3 signalling to NFκB and IRF3 . The same protease can also target IPS-1 , and thus HCV can disable both the TLR3 and RIG-I detection pathways with a single protein.
Finally, not only do viruses possess counter-measures for evading detection by TLRs, but it is also becoming apparent that they have strategies to actually hijack TLR pathways for their own purposes. In contrast to the recognition of viral nucleic acid by TLRs, the interaction between viral proteins and TLRs (Table 1) may sometimes represent subversion of the TLR system for the benefit of the virus, rather than detection of the virus by the host. For example, the haemagglutinin (HA) protein of wild-type, but not vaccine, strains of measles virus activates murine and human cells via TLR2, leading to induction of proinflammatory cytokines such as IL-6 in human monocytic cells, and up-regulation of surface expression of CD150, the receptor for measles virus . Hence, activation of TLR2-dependent signalling by wild-type measles virus is likely to contribute to viral spread.
Similarly, mouse mammary tumour virus (MMTV) seems to utilize TLR4 for the benefit of the virus, as it has been shown that maintenance of MMTV infection is dependent on TLR4 signalling, which triggers production of the immunoregulatory cytokine IL-10 . Further, in C3H/HeJ mice which lack functional TLR4, the virus was eliminated by the cytotoxic immune response . Interestingly, VACV A52 can also manipulate IL-10 via the TLR pathway, because as well as inhibiting NFκB activation and the NFκB-dependent gene IL-8, A52 expression also enhances MAP kinase activation via the TRAF6 interaction, leading to induction of the IL-10 promoter and enhancement of TLR-induced IL-10 production . Hence, in VACV-infected cells where TLRs have been activated, A52 may direct the signalling pathway towards IL-10 induction and away from chemokine production. Consistent with these results, and the fact that some other viruses either encode their own IL-10 orthologues or have alternative mechanisms to enhance host IL-10, two landmark papers have now shown that elevated IL-10 production is important for viral persistence for LCMV in vivo[76,77].
This review has illustrated the rapid developments which have taken place in the past few years in understanding how viruses interact with the TLR system. It is clear that a complex interplay between viruses and the host via TLRs exists, in which it is often not immediately apparent whether the virus or the host benefits from a given interaction. Much remains to be discovered, but no doubt an increased understanding of how TLRs and viruses interact, at the molecular, cellular and whole animal levels, will give rise to many therapeutic opportunities in diverse viral diseases.
Work in the author's laboratory is supported by Science Foundation Ireland (Investigator Programme; 02/IN.1/B192), the Irish Research Council for Science, Engineering and Technology and the Irish Health Research Board.