Female 9-10-week-old C57BL/6 mice from the Animal Resource Centre (ARC) (Western Australia, Australia) were kept in individually ventilated cages under specific pathogen-free conditions with access to food and water ad libitum. All experiments were performed in accordance with National Health and Medical Research Council’s ethical guidelines with the animal ethics approval number 2019/1696 approved by the University of Sydney Animal Ethics Committee.

Mice were anesthetised with isoflurane prior to being infected intranasally with WNV (Sarafend, a lineage II strain of WNV) delivered in 10 μL of sterile PBS [as previously described (6, 45)]. Mice were infected with 1.2 x 10⁵ plaque forming units (PFU) or 7 x 10³ PFU of WNV, doses that are lethal in 100% and 50% of mice, respectively. Mice were sacrificed no later than days post infection (dpi) 7. Diseases scores were recorded based on the criteria shown in Supplementary Table 1 .

Plexxikon Inc. (USA) provided the PLX5622 which was formulated in AIN-76A standard chow by Research diets (USA) (1200 ppm). All mice were feed PLX5622 or control chow (AIN-76A) for 21 days prior to infection. Mice were fed either PLX5622 or AIN-76A for no longer than an additional 7 days post infection.

As per the manufacturer’s instructions, PKH26 cell-linker was mixed with diluent C (Sigma-Aldrich, USA) prior to use. PKH26 cell-linker was used at a 10-fold higher concentration than recommended and injected intravenously via the lateral tail vein 2 hrs prior to tissue collection.

Mice were injected intraperitoneally with 1 mg of bromodeoxyuridine (BrdU) (Sigma-Aldrich, USA) in 200 µL sterile PBS 3 hrs before sacrifice.

A virus plaque assay was performed as previously described (6) using virally susceptible BHK cells. Brain tissue was dissociated with a power homogenizer (TissueLyser, Qiagen) before being clarified by centrifugation. Tenfold dilutions of tissue homogenates were used to infect BHK cells. The inoculum was removed after a 1 hr incubation and replaced with an Agarose plug. Cells were incubated for a further 3 days before being fixed with 10% formalin (Sigma-Aldrich, USA). A 3% crystal violet (Sigma-Aldrich, USA) dye solution in 20% methanol (Fronine) was used to stain fixed cells. The PFU per gram was determined by factoring the number of plaques, the inoculum volume, and the dilution.

For RNA extraction, brain tissue was dissociated with TRI Reagent (Sigma Aldrich, USA) using a tissue homogenizer (TissueLyser, Qiagen, DE). The High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific, USA) was used to generate cDNA and the Power SYBR™ Green PCR Master Mix (ThermoFisher Scientific, USA) was used to conduct qPCR using primers all purchased from Sigma Aldrich, USA (See Supplementary Table 2 ), on the LightCycle^® 480 Instrument II (Roche, CH). Gene expression values were normalized to Rpl13a.

Prior to collection of spleen, brain and femurs, mice were anaesthetised and transcardially perfused with ice cold sterile PBS. Femurs were flushed with cold PBS using a 30-gauge needle, while spleens were gently mashed through a 70 μM nylon mesh sieve using a syringe plunger. RBC lysis buffer (Invitrogen, USA) was used to lyse erythrocytes in single cell suspensions of BM cells and splenocytes. Brains were processed into single cell suspensions in PBS and DNase I (DN25, 0.05 mg/mL) and collagenase (5 mg/mL) (Sigma-Aldrich, USA) using the gentleMACS dissociator (Miltenyi Biotec, DE). Subsequently, a 30%/80%Percoll gradient was used to isolate the cells from brain homogenates. After tissue processing, live cells were counted with trypan blue (0.4%) on a haemocytometer. Single cell suspensions were incubated with purified anti-CD16/32 (Biolegend, USA) and UV-excitable LIVE/DEAD Blue (UVLD) (Invitrogen, USA) or Zombie UV Fixable Viability kit (Biolegend, USA) and subsequently stained with a cocktail of fluorescently-labelled antibodies (See Supplementary Table 3 ). Cells were washed two times and fixed in fixation buffer (Biolegend). Intracellular antibodies were stained after surface staining, fixation and incubation with Cytofix/Cytoperm (BD Biosciences, USA). Anti-BrdU (3D4 or Bu20a, Biolegend, USA) was stained intranuclearly, as previously described (46). Briefly, after cell surface staining and fixation, cells were incubated in Cytofix/Cytoperm (BD Biosciences, USA), Cytoperm Permeabilization Buffer Plus (BD Biosciences, USA) and DNase (DN25, 30 U/sample) (Sigma-Aldrich, USA), prior to being stained with anti-BrdU.

Fluorescently-tagged antibodies were measured on the 5-laser Aurora, Spectral cytometer (Cytek Biosciences, USA). Acquired data was analysed in FlowJo (v10.8, BD Biosciences, USA). Quality control gating including time, single cells, non-debris/cells and Live/Dead staining was applied to exclude debris, doublets and dead cells. Cell numbers were quantified using cell proportions exported from FlowJo and total live cell counts.

The FCS files were compensated and gated in FlowJo prior to exporting channel values (CSV). T-distributed stochastic neighbour embedding (tSNE) was applied to CSV files, in RStudio (1.1.453 or 1.4.1717) using CAPX (46) (package publicly available: https://github.com/sydneycytometry/CAPX) or Spectre (47) (package publicly available: https://github.com/ImmuneDynamics/Spectreusing default settings) using default settings i.e., perplexity = 30, theta = 0.5 and iterations = 1000.

The FCS files were compensated and gated in FlowJo prior to exporting median fluorescent intensity (MFI) signals from populations of interest. Heatmaps were applied to MFI’s in RStudio (1.2.1335) using the R package, pheatmaps (48).

Non-parametric statistical tests were applied to data in GraphPad Prism 8.4.3 (GraphPad Software, La Jolla, CA). Comparison of two groups was conducted using Mann–Whitney test, and three or more groups were compared using a Kruskal–Wallis test with a Dunn’s multiple comparison test. When two independent variables and three or more groups were being compared a Two-way ANOVA and a Šídák’s or Tukey’s multiple comparisons test was used. Error bars are shown as standard error of the mean (SEM).

The depletion of microglia using PLX5622 provides a non-invasive approach to study the in vivo functions of microglia during disease. Previously published studies show that microglia depletion exacerbates clinical scores and mortality in murine models of viral CNS infection (31, 36–40). To investigate this in more detail, we set up 3 principal groups, as follows. In the experimental group (PLX), we fed mice PLX5622-formulated chow for 21 days prior to infection and following this, for 5 or 7 dpi, to induce and maintain maximal microglia depletion. In the second group (PLX-Ctrl), after 21 days we replaced PLX5622-formulated chow with control chow (AIN-76A) from dpi 0 (the day of infection) onwards, to enable the replenishment of microglial numbers via proliferation of surviving microglia (34). In the master control group (Ctrl), we fed mice only control chow (AIN-76A) throughout the period until sacrifice at dpi 5 or 7 ( Figure 1A ).

In the mock-infected PLX group, 90% of microglia were depleted ( Figures 1B, C ). We also observed a significant decrease in the number of CD45⁺, Ly6G^-, CD3e^-, NK1.1^-, Ly6C^-, CD11b⁺, CD68⁺, MHC-II⁺ cells, a phenotype consistent with a subpopulation of border-associated macrophages (BAMs) ( Supplementary Figure 1 ). Further, in the brains of this group, we saw an unexpected increase in the number of neutrophils, T cells and NK cells, possibly recruited in response to dead microglia in the CNS ( Supplementary Figure 1 ). In PLX mice, a mean depletion of 85% microglia was evident in WNV-infected mice at dpi 5. Depletion remained at 85% on dpi 7, despite reduced chow consumption by infected animals on dpi 6 and 7 ( Figures 1B–D ). Strikingly, however, the WNV-infected PLX group showed a significantly lower clinical disease score at dpi 7 than the Ctrl group ( Figure 1F ). In contrast, in the PLX-Ctrl group, substitution of PLX5622 chow with control chow on dpi 0 of infection was associated with restoration of microglial numbers and a significant increase in disease score by dpi 7 ( Figures 1B, F ). Despite differences in disease score, all infected groups showed a similar degree of weight loss from dpi 5 onwards, losing >5% of their initial weight by dpi 7 ( Figure 1E ).

Since microglia are putatively protective during viral encephalitis (31, 36–40), we investigated whether the decrease in disease score in the PLX group resulted from microglial repopulation, with the decreased chow consumed by mice late in infection. The number and percent of proliferating microglia in the PLX group, measured using BrdU incorporation in vivo, was not significantly different from Ctrl or PLX-Ctrl groups at dpi 7 ( Figures 1G, H ). Thus, it is unlikely that the amelioration of disease scores conferred by PLX5622 was due to microglial proliferation, especially as disease scores in the Ctrl group were re-capitulated, if not exceeded, in the microglia-repopulated PLX-Ctrl group. Intriguingly, however, in both mock-infected and infected PLX5622-treated mice, microglia were still evidently proliferating ( Figures 1G, H ). At dpi 5, when peak microglia proliferation occurs in WNE (12), PLX5622 did not inhibit the proliferative capacity of remaining microglia, as the frequency of BrdU⁺ microglia (~9%) was no different from the Ctrl group ( Figure 1G ). Overall, these data suggest that PLX5622 is protective during lethal viral infection and that a CSF-1-independent mechanism of microglial proliferation during CSF1-R blockade becomes more conspicuous in WNE.

To investigate the protective effect of PLX5622 during WNV-infection, we assessed viral loads in the brain at dpi 5 and 7 using a viral plaque assay and q-PCR ( Figures 2A–E ). Compared to Ctrl mice, plaque assays on homogenized brains from PLX mice showed no change at dpi 5 and slight increase in PFU at dpi 7 ( Figures 2B, C ). Using q-PCR, we observed no significant differences at either timepoint ( Figures 2D, E ). However, there was an increase in viral RNA at dpi 7 (albeit non-significant) and a significant inverse correlation between viral RNA and numbers of microglia in the brain in PLX mice at dpi 7 ( Figures 2E, F ), supporting the association of microglia with viral clearance (31, 35, 36, 38). Interestingly, microglial restoration in PLX-Ctrl mice showed no difference in viral load via qPCR, compared to microglia-depleted PLX mice ( Figure 2E ). While a plaque assay was not performed on this group to confirm these findings, this could suggest that repopulated microglia are less capable of clearing virus. Supporting this, P2RY12, a nominal microglia-enriched (29) purinergic receptor was reduced by more than 50% on repopulating microglia in the PLX-Ctrl group and remaining microglia in the PLX group at dpi 7, relative to the Ctrl group ( Supplementary Figure 2 ). This molecule is associated with an enhanced microglial migratory capacity and ability to phagocytose infected neurons for viral clearance (35), thus a reduction in P2RY12 may have inhibited viral clearance by repopulated microglia. Nonetheless, the increased viral titres in the PLX group do not explain the protective effect of PLX5622 in WNV-infection.

Mice in the PLX group had reduced disease scores at dpi 7 despite higher brain viral titres ( Figures 1 and 2 ), while repopulation of microglia in the PLX-Ctrl group abrogated this protective effect ( Figure 1 ). To further investigate the ameliorative effect of PLX5622-treatment during infection, we examined single cell brain suspensions by spectral cytometry for changes in immune-cell infiltrates ( Figure 3 ). In the PLX group at dpi 5 ( Figures 3A–D ) and 7 ( Figures 3E–H ) there was a 68% reduction in cell numbers infiltrating into infected brains compared to the Ctrl group. This corresponded to statistically significant decreases of 68-73% in Ly6C^hi macrophages and 51-60% in NK cells at dpi 5 and 7 ( Figures 3C, D, G, H ). Reduced CNS infiltration in the PLX group also corresponded with reduced pro-inflammatory cytokine and chemokine expression in the CNS, compared to the Ctrl group ( Figures 4A, B ). In contrast, in the PLX-Ctrl group, there was a rebound increase in the number of immune cells infiltrating into the brain and a commensurate increase in the quantity of pro-inflammatory mediators, compared to the PLX group ( Figures 3F–H , 4B ). Since the number of microglia and proliferating microglia were similar to the Ctrl group ( Figures 1B, G, H ), the increased disease score in PLX-Ctrl mice ( Figure 1F ) is likely explained by the increase in CNS infiltration in this group.

Since the depletion of microglia reduced disease score, this data could suggest that microglia are pathogenic in our model. While more studies are required to confirm this, this seems unlikely considering several previously published articles investigating microglia during viral infection including WNE, have demonstrated the neuroprotective effect of microglia (31, 35–42). Moreover, we have previously shown that MCs are central to WNV immunopathology, with their reduction in the CNS improving disease phenotype and enhancing survival (6, 10). Thus, these data more likely suggest that the protective effect of PLX5622 in WNE is due to the substantial reduction in infiltrating MCs into the CNS. The reduction in NK cells is unlikely to have contributed to a decrease in disease score, as depleting these cells had no effect on this (data not shown). Reduced NK cell infiltration in PLX mice may be due to the reduced infiltration of MCs and reduced expression of CCL2, CCL3, CCL5 and CXCL16 in the CNS, which recruit NK cells (49). Irrespective, considering both resident and infiltrating myeloid cell populations were significantly reduced with PLX5622 treatment, it is unclear from these data which cell type principally recruits immune cells into the brain and produces these chemoattractants and pro-inflammatory mediators.

Notably, infiltrating MCs in the brains of the PLX group showed reduced CD11c expression and increased Ly6C expression, compared to the Ctrl group ( Supplementary Figure 3 ). While this implied monocyte immaturity, which could be due to the absence of microglia or an off-target effect of PLX5622, as previous studies have argued (31, 38, 40), it is also possible that monocytes in the PLX group merely represent recently-arrived cells that have spent less time undergoing macrophage differentiation in the brain. Examining marker expression in the BM, blood and brain in the Ctrl group shows that on entry into the brain these cells normally upregulate CD11c and MHC-II, as well as CD11b, CX3CR1, CD64, CD68 and CD86, and downregulate Ly6C ( Supplementary Figure 4 ). In mice treated with other monocyte-modulatory treatments that improve clinical outcomes and/or reduce MC infiltration into the WNV-infected CNS (6, 11), including clodronate liposomes injected intravenously at dpi 5 or anti-Ly6C antibody blockade injected intraperitoneally at dpi 5 and 6, Ly6C^hi macrophages also show increased Ly6C expression ( Supplementary Figure 5 ), suggesting this altered phenotype is not specific to PLX5622 treatment. Thus, this MC phenotype could also represent an immature monocyte phenotype arising from the BM as it attempts to increase the output of cells during infection.

To determine whether the protective effect of PLX5622 was due to depletion of microglia or an off-target effect, we reduced PLX5622 to a dose which was insufficient to reduce microglia numbers (50, 51) by combining approximately one-sixth of the depleting dose of PLX5622 chow with control chow. Using the same feeding regimen as shown in Figure 1A , low-dose PLX5622 (PLX^lo group) did not reduce microglia numbers, but reduced the number of infiltrating Ly6C^hi macrophages in the brain at dpi 7 by 20%, compared to the Ctrl group ( Figures 5A, B ). Similar to high-dose PLX5622 chow, removal of low-dose chow prior to infection at dpi 0 in the PLX^lo-Ctrl group resulted in a substantial rebound effect, with a statistically significant increase in Ly6C^hi macrophage numbers infiltrating into the CNS, compared to the Ctrl group ( Figure 5A ). The low-dose chow, however, did not alter disease scores. This is likely due to the substantial number of MCs still infiltrating the CNS. Thus, since 1) microglial cell numbers in the PLX^lo group remained at Ctrl group levels, 2) there was a reduction in CNS infiltration in the PLX^lo group and 3) there was a rebound CNS Ly6C^hi macrophage infiltration in the PLX^lo-Ctrl group ( Figure 5B ), the reduction in infiltrating cells in the brain associated with high-dose PLX5622 shown in Figure 1F and the corresponding decrease in disease score cannot simply be explained by the absence of microglia.

The number of Ly6C^hi macrophages in WNV-infected brains in the Ctrl group at dpi 7 correlates with both the number and proliferative status of mature Ly6C^hi monocytes (i.e. CD45⁺, Lin^-, CD11c^lo, CD117^-, Sca1^int/hi, CD48⁺, CD11b⁺, CD115⁺, Ly6C^hi cells) in the BM ( Figures 5C, D ). Therefore, we examined BM from mice fed the full dose of PLX5622 chow, showing the same level of microglial depletion depicted in Figure 1 , to determine whether this produced detectable off-target effects during microglial depletion ( Figures 5E–L ). The gating strategy and phenotypic markers used to identify the various developmental stages of the myeloid lineage in the BM are shown in Supplementary Figure 6 and Supplementary Table 4 . Monopoiesis in the BM is substantially increased with WNV-infection, presumably to enhance the supply of “emergency” monocytes to the CNS ( Figures 5E, J ). Strikingly, CSF-1R inhibition by PLX5622 reduced the number of mature Ly6C^hi monocytes in the BM of both mock-infected and infected mice by ~40% ( Figures 5F, G and Supplementary Figures 7, 8 ). This reduction in BM monocytes was also accompanied by a reduction in CSF-1R/CD115 expression ( Figure 5I , dpi 7 BM and Supplementary Figure 9 , mock-infected BM), as well as a reduction in the number ( Figures 5I, K, L ) and frequency (out of live cells) ( Supplementary Figures 10, 11 ) of proliferating mature monocytes in the BM of PLX5622-treated mice. Interestingly, expression of CSF-1R/CD115 on mature Ly6C^hi monocytes correlated with number and proliferative status of these cells in the BM of PLX mice ( Supplementary Figure 9 ), potentially providing a measure of PLX5622 efficacy in individual mice. Replacing PLX5622 with Ctrl-chow at dpi 0 in PLX-Ctrl mice reversed this effect, with increased CD115 expression on significantly increased numbers of BrdU⁺ mature BM monocytes at dpi 7 ( Figures 5G, I, L ). This rebound in BM monopoiesis on removal of PLX5622 likely explains the increased number of infiltrating monocytes in the brain of PLX-Ctrl mice. This also shows that exposure to PLX5622 does not permanently diminish the central recruitment of monocytes from the periphery, and that the migratory response of rebounding Ly6C^hi BM monocytes is not substantially altered.

Interestingly, despite the reduction in numbers of mature BM monocytes caused by PLX5622 in mock-infected and infected mice, the frequency of proliferating mature cells did not decrease, compared to the Ctrl group ( Figures 5K, L and Supplementary Figures 10, 11 ). This suggests that a proportion of mature Ly6C^hi monocytes in the PLX group are deleted and/or not replaced. It indicates that a quorum of cells in these mice, innately resistant to PLX5622, or that become resistant to inhibition, retains the capacity to proliferate and differentiate into mature Ly6C^hi monocytes.

The reduction in these mature monocytes in mock-infected mice may also be explained by the reduced proliferative status of precursor cells. Indeed, in PLX5622-treated mock-infected mice, we observed a reduction in the frequency of proliferating monocyte lineage cells that we identified as early immature monocyte 1 (CD45⁺, Lin^-, CD11c^lo, CD117⁺, Sca1^-, CD48⁺, CD11b^-, CD115^hi, Ly6C^hi cells) and late immature monocyte 2 (CD45⁺, Lin^-, CD11c^lo, CD117^-, Sca1^-, CD48⁺, CD11b^-, CD115⁺, Ly6C⁺ cells) ( Supplementary Figure 10 ). These are the presumptive precursors of mature Ly6C^hi monocytes in the BM. However, the proliferative status of these monocyte precursors in the infected BM at dpi 7 was not significantly reduced ( Supplementary Figure 11 ), suggesting that infection can to some extent overcome the PLX5622-induced reduction in proliferation status seen in mock-infected animals. Thus, the data suggest that the reduction in mature Ly6C^hi monocytes in infected, PLX5622-treated mice is due to their deletion, while in mock-infected mice, the reduction of mature monocytes may be due to a combination of deletion and/or direct inhibition of their renewal from precursor cells.

Importantly, the number and proliferative status of classical dendritic cells (cDCs) in the BM were also reduced in mock-infected, PLX5622-treated mice ( Supplementary Figures 7, 10 ). This is not surprising, since these cells also express CSF-1R (43). However, notably, these cells were unaffected by PLX5622 during infection. There was also some reduction in plasmacytoid DC numbers (pDCs) in the BM in infected PLX5622-treated mice, but this did not translate into reduced proliferation. Strikingly, however, in stark contrast to the rebound increase in monocyte proliferation and numbers seen in the BM of the PLX-Ctrl group, pDCs underwent a significant reduction in proliferation after removal of PLX5622 in this group, a response that was also observed in MHC-II^lo B cells. Other cell types in the spleen and BM were not reduced with PLX5622 treatment ( Supplementary Figures 7, 12 ). Taken together, PLX5622 has non-microglial effects on proliferation and renewal of various immune subsets. Importantly, the effect on the myeloid lineage evidently results in the reduction of monocyte immigration into the CNS, ameliorating the severe inflammation associated with infection.

Given PLX5622 targets CSF-1R, we used monoclonal anti-CSF-1R antibody blockade to determine if this approach could be used therapeutically. Since monoclonal antibodies (mAb) do not cross the intact BBB, this would protect microglia from adventitious effects of CSF-1R inhibition. Treatment of mice with anti-CSF-1R mAb at dpi 0, 4, 5 and 6 or at dpi 5 and 6 resulted in a striking ~50% reduction in the number of Ly6C^hi macrophages infiltrating the WNV-infected CNS without affecting numbers of microglia ( Figures 6A, B ). However, unlike PLX5622, the anti-CSF-1R mAb did not cause a reduction in disease score, nor increase survival in experiments using a 50% lethal dose of WNV (data not shown). Thus PLX5622 may be a more potent inhibitor of CSF-1R than the anti-CSF-1R mAb and, as a small molecule, better able to inhibit autocrine actions of endogenous CSF-1, which may not be amenable to mAb blockade (52).

Treating infected mice with anti-CSF-1R antibody not only significantly reduced CNS infiltration, but also impacted BM monopoiesis similar to PLX5622 ( Figures 6C–E ). Interestingly, treating mice only twice in the later phase of disease had a greater impact on BM monocytes than treating mice every second day from dpi 0 ( Figures 6C–E ). Perhaps, alternate mechanisms came into play to compensate for CSF-1R inhibition in mice treated for an extended period with anti-CSF-1R. Nevertheless, considering anti-CSF-1R had a similar effect on CNS infiltration and BM monocyte production, this suggests that the effect we saw with PLX5522 is likely due to inhibition of CSF-1R and not other tyrosine kinase receptors.

CSF-1R mAb blockade effectively reduced MC infiltration into the CNS, even when administered in the later phase of disease ( Figure 6B ). This suggests that CSF-1R blockade can rapidly inhibit BM monocyte production. However, considering previous studies have suggested a role for CSF-1 in monocyte migration (52–54), this data does not exclude the possibility that anti-CSF-1R modulates monocyte migration into the brain. To investigate this, infected mice, treated with anti-CSF-1R 14 hours previously, were injected intravenously with the membrane dye, PKH26, for 2 h prior to tissue collection ( Figure 7A ). In 16 hours, anti-CSF-1R had significantly reduced MC numbers in the brain ( Figure 7B ). However, the numbers of dye-positive Ly6C^hi macrophages recently infiltrating the brains of anti-CSF-1R-treated mice were similar to those seen in untreated mice ( Figure 7C ), suggesting that anti-CSF-1R blockade does not impact monocyte trafficking into the brain. Interestingly, the proportion of dye-positive Ly6C^hi macrophages was somewhat higher in anti-CSF-1R-treated mice ( Figure 7D ), presumably because anti-CSF-1R mAb-treated animals had a lower overall infiltration of MCs into the brain than the isotype control mAb-treated mice. Notably, we observed no change in the number or proportion of any BM monocyte subsets ( Figures 7E, F ). In this short period, however, the proportions of mature Ly6C^hi monocytes incorporating BrdU were reduced, albeit this was not statistically significant ( Figures 7G, H ). This indicates that CSF-1R blockade does not inhibit immigration of these cells into the brain and suggests that the reduced CNS recruitment of MCs in WNV-infected mice during CSF-1R inhibition is due to reduced monopoiesis and/or deletion.