The innate immune response is the first line of defence against infectious agents. It involves the release of type 1 interferons (IFN-α/β), inflammatory cytokines and chemokines, the activation of complement and the attraction of macrophages, neutrophils and NK cells into infected tissues. This innate immune response is triggered in the first hours following the entry of pathogens and is not pathogen specific. It contrasts with the adaptive immune response that is tailored to a specific pathogen and requires several days to develop. The innate immune system can sense the presence of micro-organisms through “pattern recognition receptors” (PRRs) that recognize danger signals and “pathogen associated molecular patterns” (PAMPs) expressed by microbes. Toll like receptors (TLRs) or retinoic acid inducible gene (RIG) like receptors (RLRs) are important PRRs for the recognition of viral double-stranded RNAs and single-stranded RNAs (Liu et al. 2008; Hornung et al. 2006). Resulting signal transduction cascades involve TRIF, Myd88 or IPS-1 as adaptors of TLR3, or TLRs other than TLR3 and of RLRs, respectively.

Once RABV is inoculated in the skin or in muscle by bites or scratches, the presence of the virus is rapidly detected by host defence mechanisms. The IFN response triggered at the site of entry has an antiviral effect. This was demonstrated by comparing the viral load in the thigh muscle of two groups of mice, parental mice and mice lacking the type I IFN receptor (IFNAR) after injection in the hindlimb with the encephalitic RABV strain, CVS. The viral load measured by the accumulation of viral RNA was increased in mice lacking IFNAR compared to parental strain of mice (Chopy et al. 2011a). This observation suggests that some viral particles might be readily eliminated at this early step of infection.

Like most tissues in the organism, the NS expresses different types of PPRs capable of sensing dangerous and pathogen signals (Boivin et al. 2002; Nguyen et al. 2002; Bottcher et al. 2003; Koedel et al. 2004; McKimmie et al. 2005). Central neurons express TLR1-4 as well as TLR 7 and 8 (Prehaud et al. 2005; Ma et al. 2006; Kim et al. 2007; Ma et al. 2007; Tang et al. 2007; Barajon et al. 2009). These express the RLRs (RIG-I and Mda-5) (Lafon et al. 2006; Menager et al. 2009; Peltier et al. 2010; Chopy et al. 2011b), but not the RLR LGP2, which seems to be actively degraded in neurons (Chopy et al. 2011b). Peripheral nerve plexuses and nerves (dorsal root ganglion sensory neurons and fibers of sciatic nerves) express TLRs (TLR3, 4 and 7) with prominent expression of TLR3 (Cameron et al. 2007; Barajon et al. 2009; Goethals et al. 2010). Neurons take an active part in the innate immune response in the brain being both responders to IFN and IFN producers, secreting type I IFN (predominantly IFN-β in the brain, no IFN-α and no Type III IFN λ) (Prehaud et al. 2005; Delhaye et al. 2006; Sommereyns et al. 2008).

RABV is known to trigger a RIG-I-mediated innate immune response in neurons (Hornung et al. 2006) by detecting the 5′tri-phosphate base pairing of the viral genome (Pichlmair et al. 2006). After infection, human neurons can mount a classical primary IFN response (activation of IRF3 and NF-kappa B), as well as a secondary IFN response, (activation of STATs and IRF7), leading to the production of cytokines (IL-6, TNF-alpha) and chemokines (CXCL10 and CCL5) (Prehaud et al. 2005; Chopy et al. 2011a, b).

After the injection of the CVS strain of rabies in the hind limb of mice, the size of the draining popliteal lymph nodes and those of spleen increase. Draining lymph nodes are populated with activated T cells expressing the marker of activation CD69. Activation can also be observed among peripheral blood lymphocytes (Vuaillat et al. 2008).

A strong B cell response is mounted in the spleen. When mice were injected with a less pathogenic virus (the PV strain), similar activation of T cells was observed in lymph nodes and blood suggesting that adaptive immune response is independent of the virulence of the RABV strain. Indeed, when mice were injected with an encephalitic RABV bat strain (silver-haired bat rabies virus, SHBRV) or with a less pathogenic virus (CVS-F3, mutant of CVS encoding a mutation in the G protein), the resulting adaptive immune responses (neutralizing antibodies, CD4⁺, CD8⁺ T cells response) were not different (Roy and Hooper 2007). This suggests that adaptive immune response triggered by RABV strains in the periphery, an event that occurs late after the virus has entered the NS, is unrelated to RABV pathogenicity. This may explain why patients die of rabies despite having mounted an immune response in the periphery attested by the presence of neutralizing antibodies in the blood (Hunter et al. 2010). Nevertheless, in an experimental model of rabies in skunks, a cyclophosphamide-induced immunosuppression was found to reduce the infection of the salivary glands (Charlton et al. 1984), suggesting that the adaptive immune response may control to some extent the final stage of RABV infection. This control may also be exerted on viral particles at the site of entry if entry into the NS has been delayed.

Infiltrating T cells control most of the infections of the NS. This is, for example, observed during the course of West Nile virus brain infection, where CD8⁺ T cells attracted by the chemokines produced by inflammatory cells in the infected NS are a critical factor for controlling the infection (Klein et al. 2005; Zhang et al. 2008). In rabies, sterilisation of the infection by T cells is inefficient, and their activities are specifically inactivated by the virus (Lafon 2008). Immunohistochemical studies performed on autopsy materials obtained from deceased rabies-infected patients revealed that the cells undergoing death were leukocytes and not neurons (Hemachudha et al. 2005; Tobiume et al. 2009). This observation was reproduced in mice infected with the encephalitic RABV strain CVS. Immunocytochemistry of brain and spinal cord sections revealed that despite a heavy load of viral antigens, infected neurons do not undergo apoptosis. In contrast, the migrating T cells (CD3+) were apoptotic (Baloul and Lafon 2003; Baloul et al. 2004; Lafon 2005a; Kojima et al. 2009; Rossiter et al. 2009). Moreover, pathogenicity of the CVS strain was similar in immunocompetent mice Balb/c mice and in Nu/Nu Balb/c mice, indicating that T cells do not control the outcome of encephalitic rabies (Lafon 2005a). In striking contrast, deprivation of T cells transformed an abortive infection into a encephalitic rabies similar to that caused by the encephalitic strain CVS infection, showing that T cells is a critical factor in the restriction of the NS infection caused by an abortive RABV strain. Indeed, when apoptosis was analyzed in the spinal cord of immunocompetent mice infected with the abortive RABV strain PV, killing of T cells was not observed; instead, infected neurons died (Galelli et al. 2000). Altogether, these observations indicate that T cells have a protective potential to control RABV infection in the NS, nevertheless their capacity to control RABV infection is impeded with the encephalitic RABV strain. The mechanisms by which the encephalitic RABV strain evades the host T cell response was further studied as described below.

Tumors evade immune surveillance by multiple mechanisms, including the inhibition of tumour-specific T cell immunity. In order to escape attack from protective T cells, tumor cells up-regulate expression of certain surface molecules such as B7-H1, Fas-L, and HLA-G, which triggers death signalling in activated T cells expressing the corresponding ligands PD-1 for B7-H1, Fas for FasL and CD8—among others—for HLA-G (Gratas et al. 1998; Dong et al. 2002; Rouas-Freiss et al. 2003). Studies evaluating whether RABV-infected neurons up-regulate immunosubversive molecules to kill activated T cells following an evasive strategy similar to that selected by tumors cells have been undertaken both in vivo and in vitro. In vitro, RABV infection was found to up-regulate the expression of HLA-G at the surface of human neurons (Lafon et al. 2005; Megret et al. 2007). In vivo, comparison of experimental rabies in mice caused by CVS, which kills T cells, or by PV, which does not kill T cells, led to the finding that the CVS-infected NS, but not the PV-infected NS, up-regulates the expression of FasL. In mice lacking a functional FasL, there was less T cell apoptosis in the NS than in control mice. Remarkably, RABV morbidity and mortality were reduced in these mice. Destruction of T cells through the Fas/FasL pathway can be enhanced by indoleamine 2, 3 dioxygenase (IDO), the expression of which in the infected neurons and the brain is up-regulated by RABV (Prehaud et al. 2005; Zhao et al. 2011). The enzyme IDO converts extracellular tryptophan into kynurenine, thereby reducing its concentration in the microenvironment that in turn markedly enhances the sensitivity of any nearby T cell for Fas-ligand induced apoptosis (Kwidzinski et al. 2003).

In addition, RABV-infected brain up-regulates the expression of another immunosubversive molecule, B7-H1 (Lafon et al. 2008). Whereas non-infected NS was almost devoid of B7-H1 expression, RABV infection triggers neural B7-H1 expression that increases as the infection progresses. Infected neurons and also other non-infected neuroectodermal cells, including astrocyte-like cells, were found positive for B7-H1. RABV infection of B7-H1 deficient mice resulted in a drastic reduction in clinical signs and mortality. Reduction of RABV virulence in B7H1^−/− mice was concomitant of a reduction of CD8⁺ T cell apoptosis among the migratory T cells.

Thus, despite the triggering of a classical adaptive immune response in the periphery and the infiltration of the lymphocytes into the infected NS, the protection, which could have been conferred in the NS by this immune response, is drastically impeded by RABV infection.