CX3CR1-CX3CL1-dependent cell-to-cell Japanese encephalitis virus transmission by human microglial cells

The neurotropic Japanese encephalitis virus (JEV) is responsible for Japanese encephalitis, an uncontrolled inflammatory disease of the central nervous system. Microglia cells are the unique innate immune cell type populating the brain that cross-communicate with neurons via the CX3CR1-CX3CL1 axis. However, microglia may serve as a viral reservoir for JEV. Human microglia are able to transmit JEV infectivity to neighbouring cells in a cell-to-cell contact-dependent manner. Using JEV-treated human blood monocyte-derived microglia, the present study investigates molecular mechanisms behind cell-to-cell virus transmission by human microglia. For that purpose, JEV-associated microglia were co-cultured with JEV susceptible baby hamster kidney cells under various conditions. Here, we show that microglia hosting JEV for up to 10 days were able to transmit the virus to susceptible cells. Interestingly, neutralizing anti-JEV antibodies did not completely abrogate cell-to-cell virus transmission. Hence, intracellular viral RNA could be a contributing source of infectious virus material upon intercellular interactions. Importantly, the CX3CL1-CX3CR1 axis was a key regulator of cell-to-cell virus transmission from JEV-hosting human microglia. Our findings suggest that human microglia may be a source of infection for neuronal populations and sustain JEV brain pathogenesis in long-term infection. Moreover, the present work emphasizes on the critical role of the CX3CR1-CX3CL1 axis in JEV pathogenesis mediating transmission of infectious genomic JEV RNA.


JEV recovery from JEV-pre-treated human microglia occurs at late periods after virus exposure.
At early periods of exposure e.g. 24 hours, JEV does not alter morphology nor is it cytopathic to human microglia 16 . At later exposure periods of 6 and 8 days, observations under the light microscope demonstrated that both mock-and JEV-treated human microglia were confluent with variable morphologies of flat and elongated or round cells and exerted equivalent granularity ( Fig. 1a and b). Therefore, microglia survived long-term exposure to JEV.
Neutralizing anti-JEV antibodies do not completely abrogate cell-to-cell virus transmission and recovery from human microglia. In order to evaluate a possible contact of JEV with the extracellular space during the process of cell-to-cell virus transmission between JEV-infected human microglia and susceptible cells, neutralizing immune serum of pig vaccinated against JEV was employed 22 . Importantly, the potency of antibody-dependent enhancement of infection by JEV 22 is circumvented by applying antibodies generated in pigs on cells of other species, e.g. human and hamster. Neutralizing titres of immune serum was 1:320 to cell-free JEV 22 and was used at concentrations of 1:100 and 1:1000. Control (Ctrl) and immune neutralizing (NT) porcine sera were added at time of co-culture of BHK-21 cells and JEV-infected human microglia for 6 or 8 days. Then, levels of intracellular JEV envelop protein (JEV E) were measured in BHK-21 cells and virus titres were determined in supernatants.
In order to evaluate cell-to-cell virus transmission, intracellular JEV E protein was analysed in BHK-21 cells by flow cytometry. BHK-21 cells were gated as big cells based on their forward/scatter (FSC/SSC) profile (Fig. 2a, upper left panel). Intracellular JEV E protein expression was measured on BHK-21 cells, as exemplified from co-cultures of 6d JEV-pre-treated microglia with BHK-21 cells (Fig. 2a). Upon co-culture with 6d JEV-pre-treated human microglia, significant frequencies of JEV E-bearing BHK-21 cells were 11.19% (±8.1) and 7.32% (±6.00) in the presence of 1:100 and 1:1000 Ctrl serum, respectively, as compared to mock. As expected, 1:100 and 1:1000 of NT serum significantly reduced the frequencies of JEV E-bearing BHK-21 cells to 3.35% (±1.48) and 2.63% (±1.49) respectively, as compared to corresponding Ctrl serum conditions. However, the frequencies of JEV E-bearing BHK-21 cells in the presence of 1:100 NT serum remained significant compared to mock (Fig. 2b). Although none of the Ctrl and NT sera conditions showed significance, similar tendencies were pictured upon co-culture of 8d JEV-pre-treated human microglia with BHK-21 cells (Fig. 2c). Overall, neutralizing antibodies did not completely block cell-to-cell virus transmission.
In order to confirm virus recovery, infectious virus particles from culture supernatants were titrated. After 6d cultured JEV-pre-treated microglia, significant virus titres were measured as compared to mock, i.e. 9.97 × 10 4 TCID 50 /mL (±2.35 × 10 5 ) and 5.96 × 10 4 TCID 50 /mL (±1.17 × 10 5 ) in the presence of 1:100 and 1:1000 of Ctrl serum, respectively. As expected, the presence of 1:100 and 1:1000 NT serum decreased virus titres to 2.17 × 10 1 Figure 1. JEV recovery from JEV-pre-treated human microglia after late periods of exposure. (a,b) Human microglia were treated with Mock and JEV (Nakayama isolate used at a multiplicity of infection (MOI) of 10 TCID 50 /cell) at 37 °C for indicated time-periods. Representative micrographs showing cell monolayer and morphology at magnification of (a) 10x and (b) 40×. Scale bars are of (a) 200 μm and (b) 50 μm. In (c), human microglia were pre-treated with Mock and JEV (Nakayama isolate used at a MOI of 10 TCID 50 /cell) at 37 °C for indicated time-periods. The latter cell were intensively washed with cold PBS and subsequently co-cultured with BHK-21 cells for 2 additional days in the presence of indicated concentrations of control porcine serum (Ctrl serum). (c) Curve lines representing virus titres in supernatants. The symbol represents the mean value; the error bars the standard deviation. Micrographs and data are of 3 independent experiments with each condition performed in duplicate cultures. Significant differences between Mock and JEV isolate is indicated by the letters a and b for conditions in the presence of 1:100 and 1:1000 PS, respectively. Statistics are calculated with the Mann-Whitney test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). (  www.nature.com/scientificreports www.nature.com/scientificreports/ TCID 50 /mL (±5.09 × 10 1 ) and 8.75 × 10 1 TCID 50 /mL (±2.40 × 10 2 ) respectively. This diminution in virus titres was significant compared to corresponding concentration of Ctrl serum (Fig. 2d). Similarly, significant virus titres were measured in 8d cultured JEV-pre-treated microglia compared to mock, i.e. 8.50 × 10 4 TCID 50 /mL (±2.21 × 10 5 ) and 1.03 × 10 4 TCID 50 /mL (±2.35 × 10 4 ) in the presence of 1:100 and 1:1000 of Ctrl serum, respectively. Compared to corresponding concentration of Ctrl serum, virus titres significantly decreased to 8.57 × 10 2 TCID 50 /mL (±2.41 × 10 3 ) and 1.87 × 10 2 TCID 50 /mL (±3.15 × 10 2 ) in the presence of 1:100 and 1:1000 NT serum, respectively. However, only the condition of co-cultures of 8d JEV-pre-treated human microglia and BHK-21 cells in the presence of 1:100 NT serum showed significance compared to mock (Fig. 2e). Therefore, infectious virus could be recovered from co-cultures in the presence of NT serum, demonstrating that de novo virus production could overcome neutralizing activities. Viral RNA may be a contributing source of infectious viral material upon cell-to-cell virus transmission. Considering the fact that human microglia do not produce infectious virus particles, but sustain the replication of viral RNA 16 and that neutralizing antibodies do not completely block virus transmission and recovery, further experiments aimed to identify the potential source of transmission of infectious viral material upon intercellular interactions. In confocal microscopy experiments of co-cultures, the Iba-1 marker was used to label microglia cells (Iba-1 pos ) and to discriminate from BHK-21 cells (Iba-1 neg ). Immune fluorescence labelling of intracellular dsRNA allowed detection of replicating viral RNA 23,24 and labelling of intracellular JEV E protein showed viral envelop protein. Mock were free of both dsRNA and JEV E (Fig. 3). Short co-culture periods of 6 hours showed that Iba-1 pos microglia contained dsRNA, but no JEV E protein upon pre-treatment with JEV ( Fig. 3a and b). After 24 hours of co-cultures, both Iba-1 pos microglia and Iba-1 neg BHK-21 cells presented dsRNA (Fig. 3c). In parallel, JEV E was mostly found in Iba-1 neg BHK-21 cells, but was also observed in Iba-1 pos microglia cells (Fig. 3d). This indicates a possible re-infection of the latter cells by de novo JEV particles released from the JEV-producing BHK-21 cells.
In order to assess the role of viral RNA in cell-to-cell virus transmission, RNAse A was applied to degrade RNA content in co-cultures such as viral RNA. Using flow cytometry, BHK-21 cells were gated as big cells and microglia as small cells based on their FSC/SSC profile. Then, intracellular dsRNA and JEV E protein were detected in BHK-21 cells and microglia (Fig. 4a). Treatment of co-cultures with RNAse A decreased the frequencies of dsRNA + -BHK-21 cells in a dose dependent manner and a dose of 60 μg/mL of RNAse A was used for further experiments ( Supplementary Fig. 1a). In the absence of RNAse A, frequencies of dsRNA + -BHK-21 cells were of 1.88% (±0.54) whereas 60 μg/mL of RNAse A significantly reduced the frequencies of dsRNA + -BHK-21 cells to 0.56% (±0.43). Only conditions in the absence of RNAse A showed significance compared to mock with frequencies of 0.45% (±0.16) (Fig. 4b, left panel). In parallel, the absence of RNAse A revealed frequencies of dsRNA + -microglia of 0.98% (±0.39) whereas 60 μg/mL of RNAse A showed frequencies of dsRNA + -microglia www.nature.com/scientificreports www.nature.com/scientificreports/ of 0.50% (±0.56). Again, only conditions in the absence of RNAse A showed significance compared to mock with frequencies of 0.37% (±0.24) (Fig. 4b, right panel).
Deeper analysis of JEV E expression revealed that in the absence of RNAse A, frequencies of JEV E + -BHK-21 cells were of 12.39% (±6.93) and 60 μg/mL of RNAse A resulted in frequencies of JEV E + -BHK-21 cells of 4.65% (±4.05). Both conditions were significant compared to mock with frequencies of 0.34% (±0.16) (Fig. 4c, left  panel). In parallel, the absence of RNAse A showed frequencies of dsRNA + -microglia of 2.71% (±2.3) and 60 μg/ mL of RNAse A lead to frequencies of dsRNA + -microglia of 0.78% (±0.42). Again, both conditions were significant compared to mock with frequencies 0.35% (±0.18) (Fig. 4c, right panel). www.nature.com/scientificreports www.nature.com/scientificreports/ However, treatment with RNAse A did not abrogate nor diminished the generation of infectious JEV particles, which titres ranged at 10 5 -10 6 TCID 50 /mL ( Fig. 4d and Supplementary Fig. 1b). Overall, RNAse A effectively reduced intracellular dsRNA in cells but not JEV E. Moreover, the production of de novo virus particles overcome RNAse A activity.

The CX 3 CR1-CX 3 CL1 axis contributes to intercellular interactions for cell-to-cell virus transmission and recovery.
Upon short period of exposure to JEV (e.g. 2 days), human microglia increase their cell surface expression of CX 3 CR1 16 . After 6 days, ~5% of human microglia expressed CX 3 CR1 on their cell surface upon JEV exposure and was comparable to mock (Fig. 5a). In counterpart, ~2% of BHK-21 cells expressed CX 3 CL1 on their surface, the ligand for CX 3 CR1 (Fig. 5b). Therefore, the role of the CX 3 CR1-CX 3 CL1 axis upon intercellular interactions during virus transmission was investigated using a non-competitive, potent and highly specific antagonist for CX 3 CR1 in co-cultures 25,26 . Importantly, doses of 10 μM of CX 3 CR1 antagonist did not affect the capacity of virus propagation by BHK-21 cells (Supplementary Fig. 2a). Overall, treatment with CX 3 CR1 antagonist decreased the frequencies of JEV E-expressing cells and virus titres in a dose-dependent manner. Since 10 μM of CX 3 CR1 antagonist efficiently inhibited virus transmission and recovery, this dose was employed for further analysis (Supplementary Fig. 2b and c).
Cell-to-cell virus transmission was investigated by the detection of intracellular dsRNA and JEV E protein in BHK-21 cells and microglia by flow cytometry after 2 days of co-cultures since earlier time-periods of co-cultures did not allow quantification of dsRNA and JEV E protein (Supplementary Fig. 3a and b). In the absence of CX 3 CR1 antagonist, frequencies of dsRNA + -BHK-21 cells were of 2.75% (±1.57) and treatment with 10 μM of CX 3 CR1 antagonist lead to reduced frequencies of dsRNA + -BHK-21 cells to 1.54% (±0.87). Both conditions showed significance compared to mock with frequencies of 0.49% (±0.34) (Fig. 5c, left panel). In parallel, frequencies of dsRNA + -microglia were of 0.35% (±0.31) and 0.51% (±0.45) in the absence and the presence of 10 μM of CX 3 CR1 antagonist, respectively. No significance was found compared to mock with frequencies of 0.23% (±0.29) (Fig. 5c, right panel).
In addition, recovery of infectious virus particles was evaluated by virus titration in supernatants. Importantly, recovery of infectious virus particles was confirmed in the absence of CX 3 CR1 antagonist by significant titres of 5.21 × 10 7 TCID 50 /mL (±1.09 × 10 8 ) when compared to mock. In line with previous observations by flow cytometry, treatment with 10 μM of CX 3 CR1 antagonist significantly reduced virus titres to 2.71 × 10 4 TCID 50 /mL (±5.04 × 10 4 ) but these titres remained significant compared to mock (Fig. 5e). Upon co-culture of 8d JEV-pre-treated human microglia with BHK-21 cells, similar picture were observed although no significant changes were obtained regarding the frequencies of JEV E-expressing cells from the various conditions. Otherwise, virus titres in the absence of the CX 3 CR1 antagonist were significant in comparison to mock. Importantly, treatment with 10 μM of CX 3 CR1 antagonist significantly reduced the content of infectious virus particles in BHK-21 cells (Supplementary Fig. 4a and b). Overall, antagonistic treatment of CX 3 CR1 inhibited cell-to-cell virus transmission and recovery.

The CX 3 CR1-CX 3 CL1 axis may influence the type of intercellular interactions. JEV transmission
and recovery requires cell-cell contact 16 . Already at short co-culture periods, two types of cell-to-cell interactions were observed between Iba-1 pos microglia-containing dsRNA and Iba-1 neg BHK-21 cells. In the majority of the observations, the two cell types were close to each other with a large portion of the surface cell membranes in contact (Fig. 6a, upper panel), indicating virological synapses 27 . To a lesser extent, the two cell types were more distant from each other, but forming thin membrane protrusions 27 containing dsRNA in the Iba-1 pos microglia tubular extensions, which interacted with BHK-21 cells (Fig. 6a, lower panel).
Co-cultures in the presence or the absence of the CX 3 CR1 antagonist were also analysed by confocal microscopy using a similar approach as described before. In both conditions, Iba-1 pos microglial cells contained dsRNA and interacted with Iba-1 neg BHK-21 cells. Intercellular interactions were abundant and of a virological synapses appearance 27 in the absence of the antagonist. Although the frequency had substantially decreased, membrane protrusions 27 were occasionally observed in the presence of CX 3 CR1 antagonist (Fig. 6b).

Discussion
Microglia are the first line of defence against CNS insults 14 . In the context of JE, microglial cells are the major producers of antiviral inflammatory mediators within the CNS 28 . Indeed, human microglia produce various inflammatory mediators upon exposure to JEV 16 . However, microglia may also serve as reservoir for JEV, since they sustain viral replication for long period after JEV exposure 16,17,29 . For cell-to-cell JEV transmission, the present study highlights the potential involvement of human microglia in JEV propagation and spreading in the brain. Cell-to-cell contact-dependent virus transmission may substantially contribute to the spread of the virus infection 30 . Moreover, virus recovery after late periods may contribute to persistent JEV infection. Persistency of JEV in the human CNS has been suggested, since infectious virus can be found in the cerebrospinal fluid of patients beyond the third week of illness 31 . Although it is unknown whether JEV can be reactivated, a biphasic pattern of JE has been described with a relapse of the disease 14-32 days after illness recovery 32 . Thus, JEV transmission by microglia may initiate another cycle of JEV activation leading to JE relapse in persistent infection.
www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ Cell-to-cell contact-dependent virus transmission is a mechanism used for viral escape to immunity. In our study, antibodies efficiently neutralized cell-free JEV 22 , but were inefficient in blocking cell-to-cell transmission of microglia-associated JEV. First, resistance to neutralizing antibodies may result from the inefficacy of a certain subset of neutralizing antibodies to inhibit cell-to-cell virus transmission 27 . Otherwise, resistance to neutralizing antibodies may be due to the inaccessibility of antibodies within intercellular interaction structures. Cytosol-to-cytosol connectivity between cells, as well as cytoplasmic connections-containing vesicles prevent exposure of viral material to the extracellular space 33 . Finally, the present study used antibodies derived from pigs treated with a lentivirus-based vaccine against JEV G3 prM and E proteins 34 . Therefore, other potential sources of viral material such as genomic material and capsid protein are not recognized by antibodies-derived vaccine.
The present study suggests that genomic viral RNA is contributing, but not the exclusive source of infectious viral material, to cell-to-cell virus transmission. In various studies, virus particles have been the source of infectivity in cell-to-cell virus transmission [35][36][37] . Here, the absence of JEV E protein in JEV-infected microglia cells reflects the inability of forming virus particles. Thus the absence of virus particles, but the detection of intracellular dsRNA demonstrates a replicative form of JEV viral RNA, similar to other ssRNA + viruses 24 . Viral RNA in its dsRNA and/or ssRNA forms may be sufficient for cell-to-cell transmission and the recovery of infectious virus including JEV 38,39 . Despite decreasing intracellular dsRNA content, treatment with RNAse A does not abrogate infectious JEV recovery. Because RNAse A degrades ssRNA, this suggests that dsRNA is more likely to be the source of virus material than ssRNA. However, other viral factors may also contribute to cell-to-cell virus transmission.
A cell-to-cell contact-dependent mode of transmission of viral genomic RNA represents an alternative mechanism enabling the generation of de novo viable virus particles 33 . Human microglia are inefficient to generate infectious virus particles and recovery of microglia-associated JEV by target cells depends on cell contact 16 . JEV induces virological synapses in a contact-dependent mode of virus transmission between virus-pulsed dendritic cells and T cells 35 . Here, microglia may use virological synapses and membrane protrusions such as nanotubes and/or filopodia structures for intercellular interactions. Virological synapses require the contribution of adhesion molecules together with microtubules and actin cytoskeleton for stabilization, whereas membrane extension are actin-rich structures 27 .
Importantly, virus transmission from JEV-treated microglia to target cells involves CX 3 CR1-CX 3 CL1 interaction. In the CNS, microglia express CX 3 CR1 whereas neurons express CX 3 CL1 and the CX 3 CR1-CX 3 CL1 axis is a main regulator of chemotaxis and cross-communication between microglia and neurons 18 . In the present study, low levels of CX 3 CR1 expression on microglia allowed cell-to-cell virus transmission indicating an extreme sensitivity of this mechanism. The fact that JEV-treated microglia transiently up-regulates surface expression of CX 3 CR1 16 suggests that cell contact-mediated virus transmission may significantly contribute to infection of neuronal cells at early time of infection. Therefore, the CX 3 CR1-CX 3 CL1 axis could be a potential therapeutic target candidate in JEV-infected patients. However, CX 3 CL1 is crucial in the inhibition of the production and neurotoxicity of JEV-infected microglia-derived inflammatory factors 40,41 . Moreover, polymorphism of CX 3 CR1 42 may influence the specificity of potential therapeutics. Polymorphism of CX 3 CR1 may also explain the differences observed in virus transmission and recovery between cells of blood donors. Nevertheless, the CX 3 CR1 antagonist did not completely abrogate cell-to-cell JEV transmission indicating that additional or alternative factors may be involved in transmission. For example, DC-SIGN, which is expressed in human microglia 43,44 , promotes dendritic cell-to-T cell JEV transmission 35 and mediates cellular modifications such as cytoskeleton remodelling promoting filopodia extension 45 .

Methods
Authorization, ethics approval and consent to participate. The Federal Office for the Environment (FOEN, Bern, Switzerland) provided authorization for the collection of human samples and manipulation of the various cells and viruses (authorization number A130522). Ethics approval 034/13-CER-FR for the characterization and differentiation of human white blood cells has been granted by the Ethics Committee of the Canton of Fribourg according to corresponding laws and regulations, based on the Declaration of Helsinki. No consent form was required as the buffy coats were provided from the Swiss Red Cross Blood Bank. All samples were analysed anonymously.
Antibodies and stains. The viral envelop protein was detected using the pan-immune anti-flavivirus antibody (mouse IgG1/IgG2a, clone ATCC-HB-112 D1-4G2-4-15 hybridoma, ATCC, Wesel, Germany) and double stranded RNA (dsRNA) was detected with the anti-dsRNA J2 antibody (mouse IgG2a, Scicons, Budapest, Hungary). Cell markers were detected using anti-human antibodies against CX 3 CL1 (rabbit polyclonal IgG, clone PA5-23062, Thermofischer Scientific, Waltham, MA), Iba-1 (rabbit polyclonal IgG, Wako, Richmond, VA) and fluorescent-labelled CX 3 CR1-R-Phycoerythrin (rat IgG2b, clone 2A9-1, Miltenyi Biotec GmbH, Bergisch-Gladbach, Germany). represent replicates where each colour is an individual donor. The black line represents the median value, the red line the mean value and the error bars the standard deviation. (e) Scatter dot plots representing virus titres in supernatants. The symbols represent replicates where each colour is an individual donor and the black line represents the mean value. Data are of 2 or 4 independent experiments with each condition performed in triplicate or duplicate or triplicate cultures. Asterisks on top of a condition show significant differences compared to mock; asterisks on black line show significant differences between the indicated conditions. Statistics are calculated with (c,d) the t-test or (e) the Mann-Whitney test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). ( Virus preparation and titration. JEV Nakayama isolate (National collection of pathogenic viruses, NCPV, Salibury, UK) was propagated in BHK-21 cells as previously described 16 . Briefly, 80% confluent BHK-21 monolayer cell culture was infected with JEV suspended in RPMI-1640 GlutaMAX TM -I medium (Thermofischer Scientific) supplemented with 2% FBS and cultured until cytopathogenic effects (approx. 36-48 hours). Virus stock suspension was obtained after disruption of remaining cells by freezing and centrifugation of cell soup at 3000 g at 4 °C for 30 minutes to eliminate cell debris. In parallel, mock was prepared accordingly from uninfected BHK-21 cells and served as negative control of virus exposure in experiments. www.nature.com/scientificreports www.nature.com/scientificreports/ Virus titres of virus stocks and experimental supernatants were determined by end-point titration. Briefly, 10-fold serial dilutions of suspension in GMEM supplemented with 10% FBS and Tryptose Phosphate Broth solution were applied on BHK-21 cells at 37 °C, 5% CO 2 . After 36-48 hours, intracellular viral particles were detected with the pan-immune anti-flavivirus antibody followed by peroxidase enzymatic reaction.

Generation of human microglial cells.
Human blood monocyte-derived microglia were generated from buffy coats of anonymous healthy donors obtained from Blutspendedienst (Bern, Switzerland) using a protocol adapted from 46 . Human peripheral blood mononuclear cells (PBMC) were isolated from buffy coat after Ficoll-Paque density gradient centrifugation (1.077 g/L, Amersham Pharmacia Biotech AG, Dubendorf, Switzerland). Monocytes were enriched using positive selection of CD14 + cells with selection columns and magnetic sorting system (Miltenyi Biotech GmbH). CD14 + monocytes were cultured at a concentration of 0.5 × 10 6 cells/mL in RPMI-1640 GlutaMAX TM -I medium. Medium was supplemented with antibiotic/antimycotic and bioactive human recombinant granulocyte macrophage colony-stimulating factor (GM-CSF) (10 ng/mL), macrophage colony-stimulating factor (M-CSF) (10 ng/mL), nerve growth factor (NGF)-β (10 ng/mL) and CC chemokine ligand 2 (CCL2) (50 ng/mL) (all purchased from Miltenyi Biotech GmbH), at 37 °C and 5% CO 2 for 7 days. Half of the medium was renewed after the 3 days of culture.
Treatment of human microglial cells with JEV and co-culture with susceptible target cells. Human microglia were treated with Mock or JEV (at a multiplicity of infection (MOI) of 10 TCID 50 /cell) in RPMI-1640 GlutaMAX TM -I medium at 37 °C and 5% CO 2 during various time-periods. Supernatants were collected and pre-treated human microglia were washed 5 times with cold PBS. Both supernatants and last washing was verified negative for infectious JEV by end-point titration. Subsequently, susceptible target BHK-21 cells were co-cultured with pre-treated human microglia in RPMI-40 GlutaMAX TM -I medium, at 37 °C in 5% CO 2 . Co-cultures were performed in commercial porcine serum (Thermofischer Scientific). After various time-periods of co-culture, supernatants and cells were collected for further analysis.
In some experiments, immune porcine serum of vaccinated pigs with neutralizing activity against Nakayama isolate at a titre of 1:320 22 and previously described commercial porcine serum was used as control. In other experiments, CX 3 CR1 antagonist (AZD8797, Axon Biochemicals, Groningen, The Netherlands) suspended in DMSO (Thermofischer Scientific) and/or RNAse A (Macherey Nagel, Düren, Germany) suspended in PBS were used. DMSO and PBS were respectively used as controls. These reactive were put on top of treated microglial cells before the addition of BHK-21 cells for co-cultures.
Flow cytometry. Cells were analysed using multi-colour flow cytometry FACSCanto II instrument (BD Biosciences, San Jose, CA). Data were analysed using FlowJo Software (Data analysis Software, Ashland, OR).
Bright field and confocal microscopy. For conventional bright field microscopy, cells were cultured in flat bottom well plates. Cells were analysed and photographed using an EVOS XL Core digital inverted microscope with cell imaging system (Thermofischer Scientific). The image acquisitions were performed with the 10x and 40x objectives.
For confocal microscopy, cells were alternatively cultured in labteck or on autoclaved coverslips. After staining, slides were then mounted using Mowiol and left overnight at 4 °C for solidification. Cells were analysed and photographed using a confocal microscope A1 combined with an ECLIPSE Ti inverted microscope and a digital imaging NIS-Elements AR 3.30.02software (Nikon AG, Egg, Switzerland). The image acquisitions were performed with the 20x and 40x objectives. Images were treated with Imaris 9.1 software (Bitplane AG, Zürich, Switzerland), applying background subtraction, threshold applications, gamma correction and maxima.
Biosafety. All experiments using JEV were conducted in a biosafety level 3 bio-containment. After inactivation with 3% paraformaldehyde solution at room temperature for 15 minutes, samples were further analysed by flow cytometry and microscopy. Statistical analysis. Significant differences were determined with GraphPad Prism 6 software (GraphPad software Inc., La Jolla, CA) using the student t-Test (p < 0.05) or the Mann-Whitney Test (p < 0.05).

Data Availability
The datasets are available upon request