Innate Sensing of Viruses by Toll-Like Receptors

Recognition of dsRNA.

It is now well appreciated from the work of numerous laboratories that dsRNA, along with its synthetic analog, polyinosine-poly(C) (poly I:C), is a potent inducer of IFN responses. It was recently demonstrated that TLR3 is a cellular receptor that recognizes dsRNA and initiates these responses (4). dsRNA is a replication intermediate for RNA viruses, and lysis of virus-infected cells is hypothesized to release dsRNA that can be detected by TLR3 molecules expressed on neighboring cells. Interestingly, cells such as fibroblasts express TLR3 on the cell surface; however, in certain dendritic cells, TLR3 expression is restricted to intracellular compartments (56, 57). Activation of TLR3 and the subsequent secretion of IFN-α/β help to establish a localized antiviral state that limits viral replication at the site of infection.

TLR3-mediated responses are unique among TLR family members because activation of TLR3 elicits lower levels of inflammatory cytokines than other TLRs. However, TLR3 activation induces very robust secretion of IFN-β as well as expression of costimulatory molecules (4). Each of these activities is retained in MyD88^−/− mice, suggesting that TLR3 does not require the MyD88-dependent pathway (37). Recent studies have revealed that TLR3 signals through a unique adaptor molecule, TIR domain-containing adaptor-inducing IFN-β (TRIF), also known as TIR-containing adaptor molecule-1 (TICAM-1), in place of MyD88 (Fig. 1) (61, 77). Through TRIF/TICAM-1, TLR3 activates NF-κB and IFN regulatory factor 3 (IRF3), a key regulatory transcription factor for cellular antiviral responses, including IFN-β secretion. The molecular basis for the TRIF/TICAM-1 preference exhibited by TLR3 is likely due to the presence of an alanine residue within a key region of the cytoplasmic domain of TLR3. All other TLRs contain a proline residue at this position and utilize MyD88.

Recognition of CpG DNA.

TLR9 was originally identified as the receptor for unmethylated bacterial CpG DNA (8, 35). Recently, herpes simplex virus type 1 (HSV-1), HSV-2, and murine cytomegalovirus (MCMV), all of which contain genomes rich in CpG DNA motifs, were shown to activate inflammatory cytokine and IFN-α secretion via TLR9 (48, 54, 73). An important aspect of TLR9 biology is that its activation is inhibited by the addition of chloroquine, which prevents acidification of endosomes (2, 29). The current model of TLR9 activation is that in which virus particles are taken up by cells and subsequently degraded within endocytic vesicles, exposing the viral genome which can activate TLR9 (Fig. 1). Intriguingly, TLR9-mediated IFN-α secretion in response to HSV-1, HSV-2, and MCMV occurs in a MyD88-dependent manner (48, 54, 73), which is in direct contrast to IFN responses mediated by other TLRs, which induce IFN-α/β secretion through MyD88-independent pathways. Furthermore, the TLR9-mediated IFN-α response to HSV-1 and HSV-2 may be limited to plasmacytoid dendritic cells (pDCs), a dendritic cell subpopulation characterized by their ability to secrete high levels of IFN in response to viral infection (7, 15, 20, 21, 24, 27, 63, 68, 70). Murine dendritic cell subsets distinct from pDCs do not secrete IFN-α in response to HSV-1 or HSV-2, indicating that this cell type may be specifically designed to combat viruses (48, 54).

Recognition of ssRNA.

Imidazoquinolines and specific guanine nucleoside analogs are potent antiviral compounds that activate TLR7 and TLR8 (32, 34, 42, 51). Recently, the PAMP for TLR7 and TLR8 was identified as ssRNA. Like TLR9, TLR7 and TLR8 require endosome acidification for proper activity and induce IFN-α production in a MyD88-dependent manner. Thus, these three TLRs utilize similar mechanisms of action to recognize and respond to distinct nucleic acid targets (22, 32). Indeed, TLR7, TLR8, and TLR9 are believed to have evolved from one another, and these similarities support the hypothesis that they comprise a distinct TLR subfamily (18, 23). Influenza A virus, vesicular stomatitis virus (VSV), and human immunodeficiency virus type 1 (HIV) have recently been implicated as subjects of TLR7 surveillance (22, 33, 55). Purified influenza virus genomic RNA elicits IFN-α from wild-type but not from TLR7^−/− mouse dendritic cells (22). GU-rich synthetic oligonucleotides derived from the U5 region of HIV-1 RNA stimulate costimulatory molecule expression as well as inflammatory cytokine and IFN-α secretion in a TLR7-dependent manner (33). Furthermore, poly(U) but not poly(A), poly(C), poly(G), or poly(I) compounds activate TLR7. Together, these data suggest that a simple RNA motif is recognized by TLR7 and also raise the question of how cells distinguish between viral and self RNAs (22). These groups suggest that self RNAs are subject to degradation by extracellular RNases. Thus, RNAs taken up by phagocytic cells within the confines of a virus particle are able to activate TLR7, whereas self RNAs rarely reach the endocytic compartment where TLR7 resides. The studies described above utilized mice lacking TLR7; however, genetic complementation experiments with human TLR7 and TLR8 suggest that TLR8, not TLR7, is responsible for the recognition of ssRNA in humans (33). These authors point out that murine TLR7 expression does not confer ssRNA responsiveness in their system and may represent a limitation of their experimental approach, leaving open the possibility that TLR7 may function in humans in vivo. However, these data do suggest that TLR7 and TLR8 mediate a species-specific recognition of ssRNA.

Recognition of envelope glycoproteins.

Viral envelope glycoproteins are another broad class of molecules that are emerging as the subject of TLR detection. One common theme among the viral glycoproteins implicated to date is that they play critical roles in the binding and/or entry of their respective viruses. Thus, viral pathogens can be detected at the earliest stages of infection simply via contact between glycoproteins on the viral envelope and TLRs on the cell surface, rather than requiring viral gene expression and/or replication.

Respiratory syncytial virus (RSV) was the first virus demonstrated to activate a member of the TLR family, eliciting NF-κB activation and inflammatory cytokine secretion in a TLR4- and CD14-dependent manner (50). (Although CD14 was initially identified as an LPS receptor, it is now generally viewed as an enhancer of TLR responses; however, the precise role that CD14 plays in promoting TLR responses is not well understood [50].) The fusion (F) protein from RSV was identified as a viral component that activates TLR4. This observation is consistent with the notion that viral entry, gene expression, and replication may not be required for induction of innate immunity in response to RSV. Furthermore, the importance of TLR4 activation during RSV infection in vivo was demonstrated with mice lacking TLR4 (30). TLR4-deficient mice exhibited lower levels of infiltrating natural killer and CD14⁺ cells and reduced levels of IL-12 in serum than their wild-type counterparts. The physiological consequence of these shortfalls was a reduction in the rate of viral clearance. Taken together, these studies established that viruses were subject to the same form of innate sensing as bacterial and fungal pathogens and that TLR activation by viruses in vivo is a critical aspect of the overall immune response to viral pathogens.

Mouse mammary tumor virus (MMTV) was also shown to activate TLR4 (66). As was the case for RSV, the viral envelope protein (Env) was shown to be the viral component responsible for TLR4 activation (66). J. C. Rassa et al. showed that MMTV Env coimmunoprecipitates with TLR4, providing the first evidence for a physical interaction between a TLR and a viral glycoprotein. Furthermore, the physical interaction between TLR4 and MMTV Env appears to occur via a protein-protein interaction, since both can be coimmunoprecipitated in the presence of tunicamycin (66). These authors speculate that TLR4 activation by Env may have a positive effect on viral replication by directly activating B cells, allowing an initial round of viral replication or feedback of TLR4-induced cytokines, which may promote provirus transcription. A recent study has revealed that MMTV activation of TLR4 on dendritic cells enhances expression of its entry receptor, CD71 (14), which suggests that activation of TLR4 by MMTV facilitates viral entry in addition to the replication-enhancing effects described above. This report also showed that TLR2 is activated by MMTV and that TLR2 responses may contribute to these processes (14).

Further evidence for a relationship between TLR4 and MMTV is demonstrated by the observation that wild-type MMTV is retained in mice expressing functional TLR4 (C3H/HeN), but in mice expressing a nonfunctional TLR4 (C3H/HeJ), wild-type MMTV is replaced with a recombinant of MMTV and the endogenous provirus (Mtv) (41). Secretion of the immunosuppressive cytokine IL-10 by B cells is responsible for inhibiting the antiviral response to MMTV and preserves wild-type MMTV in TLR4-expressing mice. However, B-cell production of IL-10 is not a direct consequence of TLR4 activation; rather, an unidentified TLR4-dependent signal from dendritic cells and macrophages prompts B cells to secrete IL-10.

Measles virus (MV), through its hemagglutinin protein, induces NF-κB and inflammatory cytokine secretion via TLR2 (9). In contrast to the wild-type virus, tissue culture-adapted vaccine strains were not able to stimulate TLR2. This observation is intriguing because one would predict that an efficacious vaccine would be more likely to trigger innate immune responses than virulent wild-type strains. The resolution to this apparent contradiction may lie in the observation that the receptor for wild-type MV, CD150, is up-regulated in human monocytes upon TLR2 activation. Vaccine MV strains, in contrast, utilize CD46 as an entry receptor (9).

Recently, our laboratory demonstrated that human CMV is able to activate and signal through TLR2 (19). NF-κB is activated and inflammatory cytokine secretion is induced in a TLR2-dependent manner in response to CMV. Furthermore, similar to most TLR2-mediated responses, the response to CMV is enhanced by the presence of CD14. The observation that dense bodies (fusion-competent defective particles lacking capsid and viral DNA but rich in envelope and tegument proteins) trigger responses identical to those of live and UV-inactivated virions initially suggested that CMV is in the group of viruses that trigger TLR activation during virus-cell contact and/or entry. In contrast to paramyxoviruses and retroviruses, CMV is extremely structurally complex, displaying at least 12 envelope glycoproteins. Soluble forms of envelope glycoprotein B (gB) activate NF-κB (80) and IRF3, induce IFN-β secretion, and render an antiviral state in cells (11, 12, 16, 71). More recently, we have found that soluble gB also induces inflammatory cytokine secretion in a TLR2-dependent manner. Furthermore, a direct interaction between gB and at least one other CMV envelope protein with TLR2 was found, suggesting that multiple glycoproteins from a single virus can display recognition motifs (M. Guerrero and T. Compton, unpublished data). At this point, it is unclear whether induction of antiviral responses to CMV is a TLR2-mediated process, and to date no link has been established between TLR2 and activation of the IFN pathway.

HSV-1 is another herpesvirus that may activate TLRs via one of its envelope glycoproteins. Like CMV, HSV-1 elicits inflammatory cytokine secretion through TLR2; however, the viral trigger has yet to be defined. HSV-1 has an intimate relationship with the host immune system, and immunomodulation by TLR2 activation appears to play a prominent role in HSV-1-associated immunopathology (49). Challenge of adult TLR2^−/− mice with lethal doses of HSV-1 (KOS strain) resulted in lower levels of IL-6 in serum and decreased levels of monocyte chemotactic protein 1 in the brain compared to those of the control mice. Interestingly, virus dissemination was similar in wild-type and TLR2-null mice, but TLR2^−/− mice were significantly more resistant to lethal HSV-1 challenge than wild-type mice. This trend was also observed in neonatal mice. Consistent with the decreases in cytokine levels, infiltration of mononuclear cells was lower in TLR2^−/− mice. These results suggest that TLR2-mediated cytokine responses to HSV-1 are responsible for a significant portion of the morbidity and mortality associated with HSV-1 infection. As described above, HSV-1 also activates TLR9 (49). Activation of multiple TLRs by a single pathogen may be advantageous to the host, as it would allow the immune system to more precisely identify an invading pathogen and then mount an appropriate response. Experiments comparing double- and single-knockout mice may provide exciting new insights into how activation of multiple TLRs by a single pathogen is coordinated.

The ability of glycoproteins to stimulate TLRs is intriguing because they are synthesized within host cells and therefore do not intuitively harbor modifications or moieties that differ significantly from those of the host. It is possible that these proteins possess unique structural conformations and/or clusters of glycans that host proteins do not assume. It seems highly likely that some underlying common pattern is displayed that is particular to the glycoproteins of viruses. Future work will undoubtedly be aimed at defining the molecular basis of PAMPs on viral envelope glycoproteins.