**4. Discussion**

By use of a novel immunofluorescence-compatible technique for 3D immunofluorescence imaging of solvent-cleared brain tissue slices [38,39] adapted to the imaging of RABV infections in brain tissue [40], we investigated the cell tropism of six different RABV isolates and lab-adapted strains. Both neurons and astrocytes display heterogeneous morphologies with pronounced three-dimensional projections. Compared to conventional thin sections or even in silico 3D reconstructions of serial thin sections, optical slicing of 1 mm-thick tissue samples during image acquisition allowed fast and seamless 3D reconstruction of immunostained tissues and high-resolution dissection of the detected antigens [40]. For the first time, this allowed systematic analysis and comparison of the tropism of the different viruses in large 3D volumes of infected mouse brains. One example for the superiority of the employed technique is the unambiguous detection of rCVS-11-infected astrocytes after i.c. inoculation (Figures 6 and 7), whereas conventional thin layer immunohistochemistry analyses led to the assumption that infection of glial cells by CVS does not occur, even after i.c. inoculation [55].

Importantly, more than 10<sup>4</sup> neurons and 10<sup>3</sup> astrocytes (Supplementary Table S1 and S2) were quantitatively investigated for RABV infection in their authentic environment per virus and inoculation route. This provided reliable datasets about the frequencies of RABV detection in the two cellular subpopulations in vivo and statistically significant differences in astrocyte infections, which depended on the degree of virulence and the inoculation route of the tested RABVs (Figures 5 and 6).

Because of the clonal and the defined nucleotide sequences, the recombinant viruses rRABV Fox and rRABV Dog [42], rRABV Rac (this work), SAD L16 [43], and rCVS-11 (this work) were selected for the analyses (full genome nucleotide and G protein amino acid sequence comparisons are provided in Supplementary Tables S3 and S4, and Supplementary Figure S6). Since SAD L16 was directly derived from the live vaccine SAD B19 vaccine strain [56], which is equally apathogenic after intramuscular inoculation, the non-recombinant SAD B19 progenitor vaccine virus strain ERA was included to allow for the investigation of attenuated live vaccine virus astrocyte tropism after peripheral i.m. inoculation. Compared to SAD B19, the ERA strain has a less intense cell culture passage history [45] and is still pathogenic in mice after peripheral inoculation [57]. Indeed, whereas SAD L16 was not able to cause disease after i.m. inoculation, ERA was even more pathogenic than rCVS-11 at the high i.m. inoculation dose of 10<sup>5</sup> TCID50 with 100% and 16.7% of mice developing clinical signs, respectively (Supplementary Figure S3).

Both the progenitor field virus isolates and the respective recombinant virus clones rRABV Fox and rRABV Dog have been shown to share comparable pathogenicity with efficient disease development after i.m. infection of mice [42]. For the two aforementioned field virus isolates as well as for the raccoon RABV isolate used here for the generation of rRABV Rac, virulence in raccoons has been demonstrated [46]. Notably, although the raccoon virus isolate was highly virulent in its natural reservoir host, the recombinant clone rRABV Rac generated here was less virulent in mice, showing a pathogenicity of 50% only at the high i.m. inoculation dose (Supplementary Figure S3).

In striking contrast to rRABV Fox and rRABV Dog, the lab virus clone rCVS-11 was less efficient via the i.m. route but still able to cause disease in one out of six mice at a high virus dose of 10<sup>5</sup> infectious units per mouse (Supplementary Figure S3). This perfectly matched previous studies, where the non-recombinant progenitor CVS-11 strain caused disease in only 16.7% of the mice at high i.m. dose inoculation, and, as observed here (Supplementary Figure S3), was apathogenic at low dose i.m. infection [51].

In contrast to the in vivo results, in vitro infection of mixed rat hippocampal neuron/astrocyte cultures with rRABV Fox, rRABV Dog, SAD L16, and rCVS-11 led to comparable levels of astrocyte infections, as demonstrated by GFAP-positive, RABV-infected cells (Figure 1) in a range from 4.9% to 6.7% for all viruses (Table 1). These data were in accordance with the general susceptibility of cultivated astrocytes for RABV infection repeatedly shown for field and lab-adapted viruses [29,34], with preferential replication in neurons [32]. Whereas Tsiang et al. reported 90–99% of the primary astrocytes free of RABV antigen, our results indicate that there is less variation within the tested viruses, with virus antigen detection in about 5% of the cultivated astrocytes independent of whether they were of field virus or lab strain origin (Table 1). Overall, comparison of the in vitro and the in vivo data showed that a general susceptibility of a primary CNS cell subpopulation in vitro (Figure 1) does not necessarily reflect the situation in vivo (Figures 2, 4 and 6).

Reasons for this could be altered antiviral response profiles in the more disordered cell culture conditions, di fferent replication and spreading kinetics of the viruses in vivo, or di fferences in the non-synaptic release of infected neurons to allow infection of non-synaptically connected CNS cells. Whereas no information is available about the latter thus far, it is known that astrocytes are abortively infected by a chimeric SAD L16 virus expressing CVS-11 glycoprotein in mouse brains [33]. This supports the idea that replication of rCVS-11, ERA, and SAD L16 was similarly blocked in astrocytes by potent innate immune responses after infection via the i.m. or the i.c. inoculation routes, respectively. However, since neither viral genome copies nor virus mRNA levels were measured here, it remains to be clarified whether replication kinetics of the lab strains indeed di ffer from those of the field strains in the infected brains.

Nevertheless, it is conceivable that a more e fficient replication of rRABV Fox, rRABV Dog, and rRABV Rac, and therefore the accumulation of the major interferon antagonist phosphoprotein P [8,9] in astrocytes, led to inhibition of antiviral responses. Indeed, lab-attenuated RABV has been shown to di ffer from wild-type RABV by the induction of an increased type I interferon (IFN) production and expression of inflammatory cytokines via the mitochondrial antiviral-signaling protein (MAVS) pathway [34]. Accordingly, the immunofluorescence detection of RABV P for the identification of infected cells (Figures 2, 4 and 6) confirmed the presence of abundant levels of the major interferon antagonist in the field virus-infected astrocytes. It is highly unlikely that the lack of P detection in the astrocytes of lab RABV-infected animals was due to a major di fference in P gene expression of field viruses, since P was readily detectable in lab RABV-infected neurons.

Besides supporting virus replication in astrocytes by decreasing release of inflammatory cytokines, robust field virus replication in these cells may result in a less pronounced or delayed general antiviral response to field viruses. This would be in accordance with the observed immune escape of field viruses in infected animals [35] and may represent a major immunological di fference to lab-adapted strains. However, further experimentation will be needed to test this hypothesis.

Notably, and in contrast to the ERA strain not being detected in astrocytes after i.c. inoculation, the astrocyte infection by rCVS-11 was at a frequency of 13.4% after i.c. inoculation (Figure 6 and Supplementary Table S2) and also at higher levels than observed for the field viruses rRABV Fox and rRABV Dog after both i.c. and i.m. infection. This revealed that the ability for rCVS-11 to establish an infection in these cell types depended on the infection route. Since infection from the periphery relies on trans-synaptic spread of the viruses and the virus may not become visible for immune and non-neuronal target cells, astrocyte infection could represent a late phase phenotype of brain infection, where abundant infection of neurons and non-synaptic release of virus particles may facilitate astrocyte infection. Consequently, astrocyte-mediated immune reactions would be delayed and virus elimination prior to disease onset not possible. Compared to the field viruses, slower virus replication and/or spreading kinetics of rCVS-11 could lead to lower levels or the absence of detectable astrocyte infection after i.m. inoculation (Figure 5), although the virus is principally able to infect these cell types (Figure 7b). Indeed, the e fficacy of trans-synaptic retrograde spread can di ffer between a highly neurotropic CVS-24 virus variant and an SAD L16-like vaccine virus [24]. On a higher level, this may also distinguish highly virulent field viruses such as rRABV Fox and rRABV Dog from rCVS-11. Furthermore, it is conceivable that i.c. inoculations represent a shortcut to brain infection with simultaneous and multiple infection of neurons and astrocytes in a non-transsynaptic manner. Higher numbers of infected cells at the beginning of CNS infection compared to trans-synaptic invasion after i.m. infection may lead to faster virus spread and infection of multiple regions of the brain with abundant late phase virus release and astrocyte infection, similar to that speculated above for i.m. infections with highly virulent field viruses.

Indeed, astrocyte activation by other viruses have been shown to occur earlier after i.c. than after peripheral infections [58]. Both virus and cell response kinetics may differ between the two infection routes. Increased induction of neuronal cell death after i.c. RABV infection compared to no detectable apoptosis in i.m.-infected animals [59] further indicates qualitative differences in host reaction to the virus. The special role of astrocytes in the CNS as a main source of IFN-β expression and virus control through TLR (Toll-like receptor) and RLR (RIG-I-like receptor) activation pathways [33,60,61] in combination with the infection route-dependent differences described here, as observed for rCVS-11, may contribute to such differences in apoptosis and other host reaction patterns. However, further studies must clarify whether different infectious route-dependent and -independent virus kinetics can determine astrocyte tropism in vivo and how this affects downstream host reactions. Since differences in innate immune induction through dsRNA between field and attenuated viruses were demonstrated [34], the innate immune induction potential of rRABV Fox, rRABV Dog, rRABV Rac, rCVS-11, ERA, and SAD L16 in astrocytes has to be investigated in order to assess whether the route dependency of rCVS-11 is also affected by different levels of innate immune induction.

Most likely, lack of astrocyte infection by SAD L16 and ERA was the outcome of strong virus inhibition and elimination, as abortive astrocyte infection by a comparable virus has previously been suggested [33]. Whether less antiviral response induction may allow the more virulent rCVS-11 or the highly virulent field viruses to overcome a threshold of virus replication and antagonist expression—and thus may support further replication—will be addressed in future studies. Even though the underlying mechanisms cannot be clarified here, comparable results for SAD L16 and the still pathogenic ERA strain strongly sugges<sup>t</sup> that a general block of detectable astrocyte replication of these vaccine strains represents a major difference to the other tested viruses.

Only by using novel 3D immunofluorescence techniques, we were able to provide comprehensive analyses of the cell tropism of highly virulent field RABVs and less virulent or attenuated lab strains. In particular, high-resolution 3D images and quantitative downstream analysis provided novel insights in the infection processes at the clinical phase of this deadly disease. Although independent of detectable astrocyte infection, symptoms and lethal progression of the disease occur once the virus efficiently spreads in the brain after infection with all six tested viruses (Figure 7 and Supplementary Figure S3). However, different virus kinetics and astrocyte-related innate immune reactions may affect the progression kinetics, immune pathogenicity, and further spread of the virus to peripheral salivary glands. The latter may represent a key issue in terms of field virus transmission and maintenance in host populations.

Whereas these aspects must be addressed in separate trials, this study provides a novel and quantitative basis for a new, dynamic view on RABV host interactions in vivo on the cellular level. Also, this approach showcases the potential of immunostaining-compatible tissue clearing/3D imaging techniques to specifically investigate virus–host interactions at high-resolution on cellular and subcellular levels. While this study focused on standard cellular markers for neurons and astrocytes, future approaches including immune and host pathway markers will pave the way for direct high-resolution imaging-based analysis of infection processes in complex and morphologically preserved tissues.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/9/2/412/s1, Captions\_SupplementaryFiles: Captions and legends to supplementary figures, tables, and videos. Figure S1: Immunofluorescence of non-infected astrocytes and neurons in vitro. Figure S2: Details of rRABV Fox-infected, NeuN-positive neurons. Figure S3: Kaplan-Meyer survival plots for rRABV Rac, rCVS-11, ERA, and SAD L16. Figure S4: Details of field virus (rRABV Fox, rRABV Dog, and rRABV Rac) and lab RABV (rCVS-11 and ERA) infections after i.m. inoculation. Figure S5: Details of field virus (rRABV Fox, rRABV Dog, and rRABV Rac) and lab RABV (rCVS-11, ERA, and SAD L16) infections after i.c. inoculation. Figure S6. Amino acid alignment of glycoprotein G of rRABV Rac, rCVS-11, SAD L16, rRABV Fox, and rRABV Dog. Table S1: Quantification of RABV-infected neurons and astrocytes after i.m. infection. Table S2: Quantification of RABV-infected neurons

and astrocytes after i.c. infection. Table S3. Nucleotide sequence alignment of full genome sequences of rRABV Rac, rCVS-11, SAD L16, rRABV Dog, and rRABV Fox. Table S4. Alignment of G protein amino acid sequence of rRABV Rac, rCVS-11, SAD L16, rRABV Dog, and rRABV Fox. Video S1: 3D projections of rRABV Fox-infected astrocytes and neurons in two di fferent areas of a mouse brain. Video S2: 3D projection of an rRABV Fox-infected astrocyte in a mouse brain. Video S3: 3D projection of the quantification of RABV-infected neurons and astrocytes. Video S4: 3D projections of field and lab RABV-infected brains after i.m. infection with rRABV Fox, rRABV Dog, rRABV Rac, rCVS-11, and ERA, and SAD L16 after i.c. infection.

**Author Contributions:** Conceptualization, S.F., L.Z., T.M. and C.M.F.; methodology, L.Z., V.t.K., T.N. and M.P.; investigation, M.P., L.Z., M.C. and A.K.; writing—original draft preparation, M.P., L.Z. and S.F.; writing—review and editing, M.P., L.Z., M.C., V.t.K., A.K., T.N., C.M.F., T.M. and S.F.; visualization, M.P. and L.Z.; supervision, S.F., T.M. and C.M.F.; project administration, S.F.; funding acquisition, S.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by an intramural collaborative research gran<sup>t</sup> on Lyssaviruses at the Friedrich-Loe ffler-Institute (Ri-0372) to S.F. and C.M.F. L.Z. was supported by the Federal Excellence Initiative of Mecklenburg Western Pomerania and European Social Fund (ESF) Grant KoInfekt (ESF/14-BM-A55-0002/16).

**Acknowledgments:** We thank Dietlind Kretzschmar and Angela Hillner for technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.
