Innate Discrimination between Viral and Self Nucleic Acids
The innate recognition pathways described above are exquisitely sensitive to the presence of viral nucleic acids but are generally silent in the absence of infection. However, the reasons for this robust virus versus self discrimination are still not fully understood. Clearly, some viral nucleic acids have features of PAMPs, which allow for the distinction of potential pathogens from the host (Janeway, 1989).
For example, the 5' triphosphate on many ssRNAs of viral origin is absent from mRNA, which bears a 5' methylguanosine cap (Alberts et al., 2002) that does not allow RIG-I activation (Hornung et al., 2006). A 5' triphosphate is also absent from transfer RNA (tRNA), which, although uncapped, is processed to yield a 5' monophosphate (Alberts et al., 2002).
However, a 5' triphosphate is found on 5S ribosomal RNA (rRNA), an abundant species of cellular RNA. Why this does not lead to RIG-I activation is unclear, but it might have to do with the fact that the 50 end of the rRNA is obscured by ribosomal proteins (Ramakrishnan, 2002).
A similar argument applies to the 7S RNA component of the signal recognition particle, which also contains a 50 triphosphate (Wild et al., 2002). In addition, rRNA, tRNA, and other structural RNAs are heavily modified by conversion of uridine to pseudouridine, 2-thiouridine, and 2-O-methyluridine, all of which decrease RIG-I stimulatory activity when incorporated into 50 triphosphate RNA oligonucleotides (Hornung et al., 2006). Finally, as discussed above, RIG-I binding and activation can also be favored by structural elements that are more prevalent in viral RNAs, such as the panhandle structures at the 50 ends of influenza genomic RNA segments (Lamb and Krug, 2001) or the untranslated portions of the HCV genome (Saito et al., 2007).
These considerations highlight the fact that innate recognition of viruses is not as simple as the recognition of bacterial PAMPs such as LPS. Strictly speaking, there are no viral PAMPs because, to some extent, they are all shared by the host. In some cases, this could actually be used to amplify antiviral immunity: A recent report suggests that RNase L activation in virus-infected cells generates small self RNAs that potentiate the activation of RIG-I and MDA5 (Malathi et al., 2007) (Figure 1). Therefore, there is no absolute sense in which the innate immune system can be said to discriminate virus from self.
The robustness of innate responses to viral infection appears to come from coupling sensing of quantitative differences in prevalence of certain patterns, e.g., high abundance of unmodified uridine in viral genomes, to another important parameter, that of abnormal localization. For example, the cytosolic pathway effectively relies on the principle that certain species of nucleic acids are only found in the nucleus. The cytosolic presence of DNA or 5' triphosphate-bearing RNA must therefore reflect an abnormal situation, such as virus infection, that requires response. Nowhere is the importance of localization better exemplified than in TLR-mediated virus recognition.
As for the cytosolic pathway, there is a quantitative component to endosomal TLR activation in that the dearth of CpG motifs in self DNA or the presence of modified uridine in self RNA raise the threshold for TLR9 or 7 activation, respectively (Kariko´ et al., 2005; Latz et al., 2007). However, this is not sufficient for absolute discrimination, as indicated by the fact that the increased delivery of self DNA or self mRNA to endosomes can lead to TLR9 or 7 activation (Diebold et al., 2006; Lau et al., 2005; Leadbetter et al., 2002).
Most notably, the mislocalization of TLR9 to the plasma membrane allows it to respond to self DNA present in the extracellular space (Barton et al., 2006). In other words, the TLRs involved in viral nucleic acid recognition are restricted to endosomes because self DNA and self RNA have limited access to those compartments (Marshak-Rothstein and Rifkin, 2007). In this scenario, the biggest challenge to self versus virus discrimination is the clearance of apoptotic cells. It is perhaps no coincidence that phagocytes taking up apoptotic cells, including macrophages and, in the mouse, the CD8a+ subset of cDC, do not express TLR7 (Reis e Sousa, 2004).
It is also perhaps so that innate responses can be diminished that DNA degradation is initiated in cells undergoing apoptosis and is completed in the phagolysosomes of scavenger cells (Nagata, 2005). Notably, a failure of the latter process leads to IFN-a and -b production via the cytoplasmic pathway, resulting in autoimmunity (Kawane et al., 2006; Okabe et al., 2005; Yoshida et al., 2005). In addition, defects in apoptotic cell clearance are often associated with systemic lupus erythematosus (SLE) or other autoimmune diseases (Gaipl et al., 2005). This suggests that the system is carefully titrated to deal with the normal load of apoptotic cells and that phagocytic scavengers rapidly dispose of cell corpses before secondary necrosis allows the release of internal nucleic acids. It further suggests that although phagocytes can be alerted to the presence of viruses within dying cells (Schulz et al., 2005), they have efficient pathways to avoid responding to nucleic acids within uninfected apopotic bodies, through lack of sensors and/or the presence of active nucleases. Changes in this balance, by the increase of apoptotic cell load, the decrease of clearance or digestion by scavengers, or the increase of innate detection [e.g., through TLR7 duplication (Pisitkun et al., 2006)] can allow antiviral detection mechanisms to begin to react to self, often with devastating consequences.
Innate Sensing Pathways in Resistance to Viral Infection
With multiple pathways for virus sensing, an important question is to what extent they are redundant for protection from viral infection. The availability of mice in which single TLRs, RLRs, or downstream signaling components have been genetically ablated offers an opportunity to address this question. Tlr3/ mice do not show increased susceptibility to infection with VSV, LCMV, or reovirus (Edelmann et al., 2004; Johansson et al., 2007) and show either unchanged or only slightly lower resistance to MCMV (Edelmann et al., 2004; Tabeta et al., 2004). Nevertheless, a role for TLR3 has been found in the regulation of immunopathology associated with infection. For instance, Tlr3/ mice are paradoxically more resistant to infection with lethal doses of West Nile virus because TLR3 contributes to the inflammatory response to the virus, resulting in leakiness of the blood-brain barrier and allowing the infection of the brain (Wang et al., 2004). Similarly, in a mouse model of acute pneumonia induced by influenza virus, the lack of TLR3 promotes increased survival despite higher virus titers in the lungs (Le Goffic et al., 2006). Finally, Tlr3/ mice are also more resistant to infection with Punta Toro virus, perhaps because TLR3 deficiency restricts production of proinflammatory mediators, such as IL-6, which contribute to immunopathology (Gowen et al., 2006). The role of TLR9 in responses to DNA viruses is also unsettled. TLR9-, MyD88-, or Unc93b-deficient mice show greater mortality and increased virus titers in response to MCMV infection, in part because of a role of the TLR9 and MyD88 pathway in promoting cytokine production, leading to the activation of NK cells (Krug et al., 2004a; Tabeta et al., 2004; Tabeta et al., 2006). TLR9 is also important for protection from intravaginal HSV-2 infection, mediating early production of IFN-a and -b by locally recruited pDCs and thereby limiting virus spread (Lund et al., 2006). However, TLR9 or MyD88 are not required for protection from HSV-1 infection in a footpad injection or corneal scarification model (Krug et al., 2004b). The susceptibility of Tlr7/ mice to infection has not been analyzed although MyD88-deficient mice are susceptible to intranasal infection with VSV (Zhou et al., 2007), a virus that is recognized in part via TLR7. The lack of a consistent role for individual TLRs in protection from viral infection should not come as a surprise given the more restricted distribution of these receptors compared to the RLR family, the possibility of redundancy among distinct TLRs, and the overlap between the endosomal and cytosolic pathways (Figure 2). However, the latter might be a lot less redundant: Preliminary evidence indicates that RIG-I-deficient mice are highly susceptible to infection with JEV, whereas mice lacking MDA5 succumb to EMCV as early as those lacking IFNAR (Kato et al., 2006). Similarly, LGP2-deficient mice show enhanced susceptibility to EMCV (Venkataraman et al., 2007). The experimental infection of mice is a highly contrived situation that does not mimic exposure to natural inoculating routes and virus doses and is markedly affected by genetic background, including the absence of certain antiviral pathways in most inbred mouse strains (Pichlmair et al., 2004). In contrast, analysis of primary immunodeficiencies in patients is more informative in determining the relative importance of innate recognition pathways in human populations naturally exposed to the repertoire of human viruses. So far, patients lacking IRAK4 have been found to possess normal resistance to viruses even though they cannot couple TLR7 and 9 recognition to IFN-a and -b or IFN-l induction (Yang et al., 2005). Similarly, patients with an autosomal recessive defect in UNC93B are resistant to most viral infections, although they suffer from sporadic HSV encephalitis (Casrouge et al., 2006). Thus, the TLR pathway might be dispensable for responses to many viruses in human populations. Patients with defects in the cytosolic pathway have not been reported.
Viral Evasion of Innate Immunity
Viruses have to multiply extensively in infected hosts to ensure successful transmission, a challenging task in the face of immune attack. Therefore, it is not surprising that viruses that infect vertebrates have developed a myriad of immune evasions strategies, a significant number of which target the IFN system. Viruses can inhibit IFNa and -b synthesis, inactivate secreted IFN molecules, interfere with IFNAR signaling, and/or block the activation of antiviral effector proteins upregulated by IFNs (Weber et al., 2004). Here, we focus exclusively on the strategies used by viruses to block the PRR pathways discussed above. A number of viruses might inhibit IFN-a and -b induction by sequestering viral agonists from PRRs. For example, escape from cytosolic recognition could be accomplished by replication within membrane-bound compartments known as virus factories (Novoa et al., 2005). Other viruses encode dsRNA-binding proteins, such as the vaccinia virus E3L protein, which are believed to prevent IFN-a and -b induction by sequestering dsRNA (Xiang et al., 2002). Notably, E3L also possesses a DNA-binding site (Kim et al., 2003), raising the possibility that it might additionally sequester viral DNA from the DAI recognition pathway. The nonstructural protein 1 (NS1) of some strains of influenza virus also possesses a dsRNA-binding site and was thought to inhibit IFN-a and -b production through dsRNA sequestration (Garcia-Sastre, 2001). However, this site also allows binding to ssRNA and the formation of a complex with RIG-I in the presence of 50 triphosphate RNA (Pichlmair et al., 2006). This suggests that the NS1 protein might in fact block IFN-a and -b induction by targeting RIG-I (Guo et al., 2007; Mibayashi et al., 2007; Opitz et al., 2007; Pichlmair et al., 2006), although it might also use additional mechanisms, some of which are virusstrain specific (Kochs et al., 2007). Like influenza virus, other viruses may inhibit IFN-a and -b induction through targeting PRRs or downstream signaling molecules. The V proteins of paramyxoviruses can block MDA5 activation (Andrejeva et al., 2004), perhaps explaining why Sendai virus is recognized primarily via the RIG-I pathway (Kato et al., 2006). The NS3-4A protein of HCV can cleave both IPS-1 and TRIF, thereby antagonizing both the RLR and TLR3 pathway (Foy et al., 2005; Li et al., 2005; Meylan et al., 2005). Similarly, vaccinia virus A52R protein interferes with TLR signaling (Harte et al., 2003). Finally, several viruses belonging to different classes have been reported to interfere with the activation of IRF3, suggesting either direct interaction with IRFs or inhibition of upstream kinases (Weber et al., 2004). In sum, viruses target the IFN system at multiple points of the innate response, including blocking IFN-a and -b gene induction. Viruses that have lost their ability to block the IFN system have reduced pathogenicity and are often used as vaccine strains in experimental model systems. However, it should be noted that suppression of the IFN system by viruses is never as severe as, for example, that seen upon experimental ablation of IFNAR or the STAT proteins. This might reflect the coevolution of virus and host, leading to an equilibrium that preserves both organisms.
Conclusion
In 1957, Isaacs and Lindenmann reported the discovery of IFNs (Isaacs and Lindenmann, 1957). In the 50 years since, much has been learned about IFN receptor signaling and IFN-dependent antiviral immunity. However, the molecular mechanisms coupling detection of viruses to induction of the IFN-a and -b genes have only recently begun to be understood. The vertebrate strategy for the innate sensing of viruses is similar to that used for detection of bacteria and fungi and relies on detection of PAMP-like nucleic acids prevalent in viruses. However, because such nucleic acids are, to a greater or lesser extent, also found in the host, sensing virus presence requires an additional strategy. Hence, a fundamental aspect of innate recognition of viruses is to monitor the presence of nucleic acids in locations where the self equivalents are not normally found. The compartmentalization of innate recognition of viruses has interesting parallels in the strategy used by the adaptive immune system. Both systems employ specialized molecules to survey the two cellular compartments where potential pathogens might be found. TLRs and MHC class II molecules sample the endosomal compartment, whereas RLRs, DAI, and MHC class I perform an analogous function in the cytosol. RLRs and MHC class I are ubiquitously expressed by all nucleated cells, whereas MHC class II and endosomal TLRs have a more restricted distribution, primarily in immune cells. Like MHC molecules, virus PRRs can be upregulated by exposure to IFN-a and -b and/or IFN-g. Finally, the two systems are separate but in constant communication. The delivery of endosomal contents to the cytoplasm allows both MHC class I crosspresentation and RLR or DAI activation. Conversely, autophagy provides cytoplasmic proteins for MHC class II presentation (Deretic, 2006) and viral nucleic acids for TLR recognition. The notion that innate pathogen recognition and MHC presentation are intimately related is further suggested by the observation that Unc93b controls both endosomal TLR delivery and crosspresentation (Tabeta et al., 2006) and that TLR signaling can regulate MHC class II presentation (Blander and Medzhitov, 2006). The membrane biology aspects of immune recognition and of viral infection are therefore fertile ground for future discoveries. The next fifty years of IFN research might be marked by the intimate merging of innate immunity and cell biology research.