Our study demonstrates that TLR3 is localised in the cytoplasm of non-infected neuronal cells, where it is associated mainly with the endosomal compartment. TLR3 subcellular localisation is notably altered by RABV infection, with TLR3 found within viral-induced inclusions identified as RABV-induced Negri Bodies (NBs). The presence of TLR3 in the core of viral NBs, surrounded by a ring of viral N and P proteins, appears to be crucial for NB formation. This finding demonstrates a novel function of TLR3, a molecule mostly known for its immune functions.
In the absence of infection, reactivity with anti-CD63 antibody and weak signal from EEIA-expressing early endosomal structures showed neuronal TLR3 to be confined to the endosomal compartment, associated mainly to the late endosomes and multivesicular bodies. TLR3 is rarely detectable within the Golgi apparatus or at the endoplasmic reticulum. Indeed, we observed the canonical distribution pattern of TLR3 in human non-infected neuronal cells, as previously described for human DCs or fibroblasts 
The subcellular distribution of TLR3 is markedly altered in RABV-infected cells, being localised in perinuclear inclusion bodies, the NBs. These NBs contain viral NC but do not include the envelope viral proteins 
. Relocation of TLR3 during viral infection has only previously been described for respiratory syncitial virus (RSV). In epithelial cells, RSV infection increases the TLR3 levels and targets TLR3 to the membrane 
. In neuronal cells, RABV infection did not increase TLR3 protein levels, or target TLR3 to the membrane. Given that RABV infection did not increase the total pool of TLR3 protein, it is possible that the TLR3 is recruited from pre-existing cytoplasmic-, ER- or endosome-associated TLR3 by NBs. However, NBs were not stained by endosomal or ER markers, suggesting that endosome- or ER-associated TLR3 is not involved in NB formation.
NBs observed by electron microscopy did not reveal the presence of any membrane surrounding the NBs. Thus, it is unlikely NBs are autophagolysomes or lysosomes. Confocal microscopy and 3-D imaging revealed that NBs have a highly organised structure, with a TLR3-containing core surrounded by a halo of viral N and P proteins. The central position of TLR3 suggests that NB formation may be initiated by the aggregation of TLR3 molecules. A primordial role of TLR3 in NB formation is consistent with the absence of NBs from cells in which TLR3
had been silenced. TLR3 is a horseshoe-shaped solenoid with 23 leucine-rich repeats (LRRs) located in the ectodomain of the molecule. Similarly to other solenoid proteins including polyQ, TLR3 may have an intrinsic capacity to form aggregates 
. NBs formed even in the absence of the cytoplasmic domain of TLR3, suggesting a major role for the TLR3 ectodomain in NB formation (Figure S4
). PolyQ proteins such as mutant huntingtin protein form inclusions consisting of an inner dense core of aggregated huntingtin surrounded by a ring structure composed of sequestered cellular proteins 
. Experiments involving the sequential expression of polyQ followed by expression of these cellular proteins led to the conclusion that the ring structure results from the subsequent recruitment of cellular proteins at the exterior surface of an initial polyQ core aggregate. The organisation of NBs may follow a similar pattern. In this case, NBs would result from the initial association of TLR3 molecules followed, in a second step, by the accumulation of viral proteins, forming a ring into which cellular proteins could also be inserted. This would be consistent with the concentric organisation of NB structure and observation of the cellular protein eNos is recruited in NBs in RABV-infected brain. However, the NB structure is distinct from polyQ-initiated aggregates. In particular, TLR3 in NBs is still accessible to Ab. This has not been observed for huntingtin. These observations suggest that the viral ring of protein in NBs is more porous than the ring surrounding the ordered huntingtin aggregate structure. Consequently, NB and polyQ aggregates may have distinct properties. The recruitment and sequestration of cellular proteins in polyQ aggregates may lead to the functional depletion of cellular functional proteins, possibly underlying toxic properties of these aggregates in the cell. This is not necessarily the case for NBs.
The nature of the interaction between TLR3 and viral proteins in NBs remains unclear, since attempts to coimmunoprecipitate TLR3 and NC have failed.
NBs mimic aggresomes but do not exhibit the full set of characteristics of these stress-induced cellular processes. NBs are associated to the chaperone Hsp70 and to the microtubule network, reminiscent of the two principle features of aggresomes 
. However, the marked redistribution of vimentin fibres observed for example in GFP-250-induced aggresomes, and in Theiler's or reovirus-induced aggresome-like inclusions 
was not observed in RABV-infected neuronal cells. In contrast to most aggresomes, such as those formed in HSV2 infection 
, NBs were not associated with the microtubule organising centre (MTOC).
The effect of inhibition of microtubule depolymerisation, by addition of colcemid, on the distribution and size range of NB may indicate that microtubules are required for NB outcome. The finding that NBs do not display all the characteristics of aggresomes has also been observed for inclusions formed in African swine fever virus (ASFV), which have been described to function as “viral factories”. Viral factories are areas of cytoplasm where viral components and cellular components supporting viral replication are concentrated 
. The possibility that NBs function as dynamic structures involved in viral multiplication, as proposed by Lahaye et al
. (submitted), is supported by electron microscopy pictures showing that NBs are found closely associated with newly synthesised viral particles. TLR3 would thus be an essential component for virus multiplication. Consistent with this notion, we showed that NBs were absent from cells with TLR3
silencing and that TLR3−/−
mice were less severely infected by RABV than WT mice. Further studies are now required to show whether RABV production is promoted by TLR3.
A major function of TLR3 is to sense and respond to viral infection. The presence of TLR3 in the core of NBs can be seen as an attempt to inactivate TLR3 function. Sequestration of TLR3 into NBs could reduce the cellular innate immune response. The absence of TRIF –adaptor of TLR3- in NB could support the hypothesis that TLR3 molecules in NB are inactive-. The presence of viral RNA in NBs may reflect an interaction of TLR3 with viral components at an early stage of infection, when the virus enters the cell through the TLR3 decorated endosomal compartment. TLR3 could thus be involved in recruitment of viral NC from the endosomal compartment, before aggregation in NBs. However, the absence of endosomal markers within the NBs remains unexplained. In addition, the presence of dsRNAs in NBs may suggest that NBs can also sequester dsRNAs and thus may play a role in the innate immune response to RABV infection.
Alternatively, given that TLR3 can trigger neuronal apoptosis 
, the sequestration of TLR3 could be seen as an attempt of RABV-infected neuronal cells to escape TLR3-induced apoptosis. Tanaka et al
, 2004 showed an uncoupling of α-synuclein-/synphilin-1-positive aggregate formation and apoptosis 
. They showed that aggregation of α-synuclein-/synphilin-1 seemed to promote cell survival rather than cell death. In our model, sequestration of TLR3 could be a way for the virus to prevent a TLR3-mediated pro-apoptotic response to infection, as neuronal integrity is required for RABV propagation though the CNS. Promotion of neuronal survival and virus replication are not exclusive and could be complementary strategies. Preliminary results indicate that RABV strains promoting either neuronal survival or death display different effects on NB size. Further experiments will be required to understand the role of NBs in the control of neuronal death.
In conclusion, these findings describe a novel role for TLR3 and describe how viruses — in our case, RABV — hijack normal functions of neuronal proteins and exploit cell compartmentalisation to favour the progression of their life cycle.